Application of Biomarkers in Cancer Epidemiology

Edited by P. Toniolo, P. Boffetta, D.E.G. Shuker, N. Rothman, В. Hulka and N. Pearce

'ARC Scientific Publications No.142

International Agency for Research on Cancer, Lyon, 1997

Foreword

Although biological markers have been used in epidemiology for decades, it is oniy in recent years, that advances in cellular and molecular biology have greatly expanded their potential. Biomarkers have possibilities in measuring whole-body or organ-specific exposures, as indicators of biological change or of early disease, and in assessing individual susceptibility to exposure, These developments foster the incorporation of a more biological thinking into studies of cancer epidemiology aid open up new perspectives for elucidating the mechanisms of pathogenesis in vivo at the cellular and molecular level. Although the promise that biological markers hold in expanding the boundaries of epidemiological research is enormous, the criteria for their effective use in human observational research are as yet poorly understood, as are their limitations. Biomarkers have many useful future applications in cancer prevention and public health; particularly in the monitoring of exposure to hazardous substances, in identifying individuals at increased risk of disease, and in risk assessment at the population level. This volume aims to provide a set of state-of-the-art reviews of methodological issues iп the use of biological markers in cancer epidemiology. In addition, it provides a brief expert statement of our present understanding of these issues under three broad headings: study design and analysis; development of biomarkers; and application of biomarkers. The support of the United Kingdom Department of the Environment and the collaboration of the New York University are gratefully acknowledged. We thank Fred Garber for kindly designing the artwork for the cover of the book.

P. Kleihues Director, IAFIC International Agency For Research On Cancer The International Agency for Research on Cancer ('ARC) was established in 1965 by the World Health Assembly, as an independently financed organization within the framework of the World Health Organization. The headquarters of the Agency are in Lyon, France. The Agency conducts a programme of research concentrating particularly on the epidemiology of cancer and the study of potential carcinogens in the human environment. Its field studies are supplemented by biological and chemical research carried out in the Agency's laboratories in Lyon and, through collaborative research agreements, in national research institutions in many countries. The Agency also conducts a programme for the education and training of personnel for cancer research. The publications of the Agency contribute to the dissemination of authoritative information on different aspects of cancer research. А complete list is printed at the back of this book. Information about 'ARC publications, and how to order them, is also available via the Internet at: http:llwww.iarc.frl Workshop on the Application of Biomarkers to Cancer Epidemiology, Lyon, 20-23 February 1996 List of participants

R.J. Albertini S. Gaгte F, Perera E.Taioli Genetics Laboratory, Program in Environmental Division of Environmental Department of Environmental University of Vermont, 32 Carcinogenesis, NY University Sciences, Columbia University Medicine, NYU School of North Prospect St, Burlington, Medical Center, MSB Room Sсhool of Public Health Medicine, 341 East 25th VT 05401-0508, USA 209, 550 First Avenue Room 8109, 60 Haven Avenue Street, 2nd Floor, New York, NY 10010, USA New York, NY-10032, USA New York, NY 10010, USA B.K. Armstrong Cancer Control Information D. Gompertz J.D. Potter P Toniolo Centre, NSW Cancer Council MRC Institute for Environment Head, Cancer Prevention Head, Unit of Cancer P.O. Box 572, Kings Cross, and Health, University of Research Program Epidemiology, New York NSW 2011, Australia Leicester, 94 Regent Street Fred Hutchinson Cancer University School of Medicine, Leicester LEI 91N, UK Research Center 341 East 25th St, New York, Н. Bartsch 1124 Columbia Street NY 10010-2598, USA German Cancer Research K. Hemminki Seattle, WA 98104, USA Centre, Center for Nutrition and E. White 1m Neuenheimer Feld 280 Toxicology, Karolinka Institute, N. Rothman Cancer Prevention Research 69120 Heidelberg, Germany НdlSovagen 7, 14157 Environmental Epidemiology Program, Fred Hutchinson Huddinge, Sweden Branch, National Cancer Cancer Research Center С.W. Boone Institute, EPN 418, 6130 1124 Columbia Street, МP-702 Chemoprevention Branch B. Hulka Executive Boulevard, Bethesda, Seattle, WA 98104, USA Division of Cancer Prevention Department of Epidemiology MD 20892-7364, USA and Control, National Cancer School of Public Health T.F. Zhang Institute, Bldg 31, Am 10А52 University of North Carolina R. Saracci Department of Epidemiology 31 Center Drive, Bethesda, СВ NO. 7400 McGavran» lstituto Fisiologia Clinica CNR and Biostatistics, Memorial MD 20892-2580, USA Greenberg Hall, Chapel Hill, Epidemiology Section Sloan Kettering Cancer Center NC 27599.7400, USA Via Trieste, 41 1275 York Avenue, B ox 44 F.X. Bosch 56106 Pisa, Italy New York, NY 10021, USA Servei d'Epidemiologia D.J. Hunier Hospital Duran i Reynals Channing Laboratory, Harvard A. Schatzkin Autovia de CastelldeteIs, Km 2.7 School of Public Health Division of Cancer Prevention Observers 08907 Hospitalet de Llobregat 677 Huntington Avenue Studies Branch, National H. Auirup Barcelona, Spain Boston, MA 02115, USA Cancer Institute, 6130 Department of Environmental Executive Boulevard, and Occupational Medicine N. Caporaso M.T. Landi Bethesda, MD 20852, USA Aarhus University, Genetic Epidemiology Branch Genetic Epidemiology Branch Universitetsparken bygriing 180, National Cancer Institute, National Cancer Institute P.A. 5сhulte 8000 Aarhus C., Denmark EPN 439, 6130 Executive Executive Plaza North, Am Screening and Notification Boulevard, Bethesda, 439, 6130 Executive Section T. Ballard MD 20892-7377, USA Boulevard, Msc 7372, National Institute for Venetian Tumour Registry Bethesda, Occupational Safety and Via Giustiniaiii, 2, D. Coggon MD 20892-7372, USA Health 35100 Padua, Italy MRC Environmental 4678 Columbia Parkway, Epidemiology Unit A.J. McMichael Room 42, Cincinnati, University it Southampton Department of Epidemiology OH 45226-1998, USA Southampton General Hospital London School of Hygiene and 'ARC participants Southampton 5016 6YD, UK Tropical Medicine, Keppel D. 5huker R Boffetla Street, London WC1 E lIT, UK Toxicology Unit M.Freisen R Farmer University of Leicester P. Hainaut MRC Toxicology Unit. N. Pearce Hodgkin Building, Lancaster J. Hall R. Montesano University of Leicester, Wellington Asthma Research Road, Р.O. Box 138 Hodgkin Building, Lancaster Group, Department of Leicester LEI 91N, UK N.Munoz P. Pisani Road, Р.O. Box 138, Medicine, Wellington School of Leicester LE1 9HN, UK Medicine, P.O. Box 7343 E. Riboli Wellington South, New Zealand C. Wild Contents

Application of biomarkers to cancer epidemiology 1 Workshop report Transitional studies 19 P.A. Schulte and F.P. Perera Logistics and design issues ïn the use of biological samples in 31 observational epidemiology J.D. Potter

Methodological issues in the use of biological markers in cancer 39 epidemiology: cohort studies D.J. Hunter General issues of study design and analysis in the use of biomarkers in 47 cancer epidemiology N. Pearce and P Boffetta Saurces of variation in biomarkers 59 R Vineis Effects of biomarker measurement error on epidemiological studies 73 Е. White Markers of internal dose: chemical agents 95 D. Coggon and M.D. Friesen Biochemical markers of dietary intake 103 R. Kaaks, Е. Riboli and R. Sinha Biomarkers for biological agents 127 N. Mилог and EX. Bosch Carcinogen-DNA and carcinogen-protein adducts in molecular 143 epidemiology С.P. Wild and P Pisani Somatïc cell mutations in cancer epidemiology 159 R.J. Albertini and R.B. Hayes Cytogenetic end-points as biological dosimeters and predictors 185 of risk in epidemiological studies J.D. Tucker, D.A. Eastmorid and L.G. Littlefield Methodological issues in the use of tumour markers in cancer epidemiology 201 Z-F, Zhang, C. Cordon-Cardo, N. Rothman, A.N. Freedman and J.A. Taylor Quality control of biomarker measurement in epidemiology 215 D. Gompertz vi sample collection' processing and storage 223 M.T. Landi and N. Caporaso issues involving biomarkers in the study of the genetics of human cancer 237 N. Caporaso and A. Goldstein Gene-environment interactions iп the application of biomarkers of cancer 251 susceptibility in epidemiology S. Garte, C. Zocchetti and E. Taioli

Using and interpreting surrogate end-points иn cancer research 265 A. Schatzk;n, L.S. Freedman, J Dorgan, L. МсShапе, M.H. Sсhiffтаn and S.M. Dawsey Biomarker end-points in cancer chemoprevention trials 273 C. W. Boone and G.J. Kelloff The use of biological markers as predictive early-outcome measures 281 in epidemiological research A.J. Mclichael and A.J. Hall The use of biomarkers to study pathogenesis and mechanisms of cancer; 291 oesphagus and skin cancer as models R. Montesano, R Hainaut and J. Hall Comparing measurements of biomarkers with other measurements of 303 exposure R. Sаraссi Ethical and social issues in the use of biomarkers in epidemiological studies 313 P.A. Schulte, D. Hunter and N. Rothman

vii Application of Biomarkers in Cancer Epidemiology

(IARC Scientгfic Pubications No. 142) Errata

• List of participants attending the Workshop on the Application of Biomarkers to Cancer Epidemiology, Lyon, 20-23 February 1996, page v, should include the following participants:

R. Sinha Nutritional Epidemiology Branch, National Cancer Institute, EPN 443, Bethesda, MD 20892, USA

J.D. Tucker Biology and Biotechnology Research Program, Lawrence Livermore National Laboratory, P.O. Box 808, L-452, Livermore, CA 94551-9900, USA

. The Technical Editor for the Workshop, page y, was: S. Garber Epidemiology Program, New York Univ. School of Medicine, 341 East 25th St, New York, NY 10010-2598, USA

. The following persons contributed to the Workshop proceedings, page v, but were unable to attend:

P. Vineis Cancer Epidemiology Unit, University of Turin, Via Santena, 7, l-10126 Turin, Italy

R. Kaaks NTR/IARC

• In line 12 of the first paragraph of p. 309, replace 95.7% with 9.75%. Applicaliоп or Biomarkers iп Career Epidemiology Тоыоlо, R, Botfetta, P., shuker, D.E.G., Rothman, N., HuIka, B. arid Pearce, N., ode 11~RC Ccieriidie Poblicalione No. 142 Иlсгпаtlопаl Agency or Research on Carnier, Lyori, 1997 Application of biomarkers in cancer epidemiology

Workshop report*

In epidemiology, а biological marker (commonly abbreviated, for convenience, to biomarker) is any substance, structure or process that can be measured in the human body or its products and may influence or predict the incidence or outcome of disease. Biomarkers can be broadly classified into markers of exposure, effect and susceptibility. Biomarkers may include the following: xenobiotic agents and their metabolites in tissues or body products; normally occurring body constituents whether in physiological or pathological amounts; endogenous compounds that are not present under normal conditions; and inherited and acquired abnormalities of body chemistry, structure or function, including pathological manifestations of precursors to disease. Biomarkers should be distinguished from biomarker assays, specific laboratory tests aimed at measuring particular biomarkers, and biomarker measurements, the amounts of particular biomarkers present in specified units of tissues or body products as measured by biomarker assays.

Biomarkers have been used in epidemiological et aL, 1995; Institute for Environment and Health, studies of cancer for many years. Early examples 1996) on the use of biomarkers in epidemiology, include the classic studies of B. Maclahon and there remains a substantial need for relevant coleagues on the geographical correlation of uri- methodological research, wider discussion and nary estrogen concentrations with cancer of the understanding of the methodological issues, aid breast (MacMahon et al., 1974; Dickinson et ai, their incorporation into the formal and informal 1974), subsequent studies of the relationship between training of cancer epidemiologists and other scien- urinary and blood estrogens and breast cancer in tists pursuing careers in biological research on case—control studies (Cole & McMahon, 1969; cancer. Maclahon et aL, 1982, 1983) and analyses of the The general objective of the use of biomarkers relationship of cancer mortality to serum choles- in cancer epidemiology is the same as that of can- terol concentrations in the Framingham and other cer epidemiology itself: to gain knowledge about cohort studies (McMichaet et al., 1984). Two the distribution aid determinants of disease occur- recent, таjот contributions of biomarkers to epi- rence and outcome that may be applied to reduce demiological studies are the demonstration of the the frequency and impact of disease in human carcinogenicity of aflatoxine in combination with populations. There are, however, several specific hepatitis B in humans through a cohort study of objectives for the use of biornarkers in cancer liver cancer, including measurements of urinary epidemiology which should guide the evaluation metabolites and nucleic acid adducts of aflatoxin of proposals to develop or apply biomarkers. (Ross et al., 1992; Qian et al., 1994), and the iden- Principally, they are to increase the accuracy of tification of the major role of human papilo- measurements of genetic or other acquired suscep- maviruses (HPVs) in causing cancer of the cervix tibility to disease; of exposures that may cause or worldwide (Munoz etal., 1992; Bosch etal., 1995). prevent disease; of exposures that confound or In spite of the quite extensive current use of modify the associations between risk and other biomarkers in cancer epidemiology, methodologi- exposures; of disease itself; and of fаctoтs that may cal aspects of their use have not been extensively determine the outcome of disease, such as disease elucidated or described. While there have been *This paper is the consensus report of the workshop important contributions to this subject, including Application of Biomarkers in Cancer Epidemiology, which was textbooks (Hulka et al., 1990; Schulte & Pererа,1993) held at the International Agency for Research on Cancer, Lyon, and comprehensive meeting reports (Mendelsohn France, in February 1996. Application of Biomarkers iп Cancer Epidemiology

and disease. For example, the activity of an inducible, carcinogen-metabolizing enzyme is influenced by both exposure to inducing agents and the host genetic constitution; blood cholesterol concentration, an indirect indicator of fat nutrition, is almost certainly influenced by the presence of early cancer (International Collaborative Group, 1982; Sherwin et aL, 1987); presence of СС-ТТ mutations in the p53 gene in normal skin, a plausible marker of lifetime exposure to UV radiation (Nakazawa et a1., 1994), is also influenced by DNA repair enzymes (genetic susceptibility) arid may be influenced by the pathological consequences of p53 gene mutation (e.g. a proliferative advantage for the mutated cells). The practical consequences of this complexity have not yet been elucidated. In many cases, as measurements get closer Figure 1. Schematic representation showing how biomarkers to the biologically effective dose of an external may be used to measure phenomena that reflect the amount or agent at its target tissue, the relationship between effects of agents that influence or predict disease incidence or the biomarker and disease will be less likely to outcome. reflect the relationship between the amount of precursors and stages. Biomarkers may also be used external exposure to the agent and disease. to reduce, in the proposed study, the time interval However, exposures from multiple sources, such as between the relevant exposure and measurement nitrate, which may include exogenous pathways, of the putative effect and to increase the yield of may be better evaluated using biomarkers. information on disease pathogenesis. Ultimately, Because of the complexity of the interrelation- they should serve to increase the cost-effectiveness ship between biomarkers, their potential for of epidemiological studies, in the sense that, as a increasing knowledge about cancer epidemiology result of their use, more information is gained per and pathogenesis is now emerging. Research is unit cost. needed to explore the methodological characteris- Because of their origin in the structure and tics and biological functions of specific biomarkers function of the human body, biomarkers may pro- and how they relate to one another. Most particu- vide more complex measures of the underlying larly, their application in epidemiological studies exposure, process, etc. that they represent than should be undertaken after thorough preparation other measures that are common in epidemiology. and interdisciplinary collaboration. Even then, This complexity is illustrated in Fig. 1. The three interpretation of results must be undertaken with outer circles, genetic factors, environment (broadly appropriate caution. defined) and disease, represent the universe of phenomena that can be measured to reflect the Epidemiological study design and analysis amount or effects of exogenous or endogenous The use of biomarkers of exposure or dose, disease agents that may influence or predict the incidence or individual susceptibility aims to contribute to or outcome of disease. The inner cirde, biomarkers, the elucidation of the causal relationships in represents the subset of these phenomena that can human populations between diseases and factors be measured in the human body. That almost all of such as external exposures (via personal habits, the intersections between the three outer circles occupation and the ambient environment), genetic fall within the inner circle illustrates pictorially traits and interventions. that biomarkers, more than any other epidemio.. The types of epidemiological studies used in logical measure, may measure elements of any biomarker research closely parallel the types used two, or all three, of genetic factors, environment in other fields of epidemiology; biomarker studies,

2 Workshop report however, raise specific methodological issues Like other studies using biomarkers, transitional which are addressed in the following sections. studies raise ethical issues when the meaning of the biomaikei results is not known. The objectives Transitional studies of the research generally are not to identify health Tтaпsitional studies provide a bridge between the risks but to identify characteristics of the bio- use of biomarkers in laboratory experiments and marker or distribution by population subtypes. their use in cancer epidemiology studies (Schulte & There is a need to anticipate the impact of transi- Perera, this volume). They have as their outcome tional studies on study participants and plan to the characterization of biomarkers and of the address their concerns (Hunter, this volume). problems in their use; thus, they yield preliminary Cohort and case-control studies results rather than end. results about cancer etiol- Cohort and case-control studies represent the two ogy and prevention. Their intent should be to lay most widely used types of observational study the groundwork for the use of a marker in full-scale (Potter, this volume). They can be population- epidemiological studies by addressing the follow- based or based on patients at a hospital, clinic or ing aspects: intra- and intersubject variability; practice. For simplicity, studies based on patients the feasibility of marker use in field conditions; will be referred to throughout as clinic-based stud- potential confounding and effect-modifying fac- ies. tors for the marker; and mechanisms reflected by In clinic-based cohort studies, of treated patients the biomarker. Transitional studies can be divided or screened populations, the inclusion of biologi- into three functional categories: developmental, cal measures of exposure and susceptibility is both characterization and applied studies. methodologically sound and logistically feasible. Developmental studies involve determining In population-based studies, collection of biologi- the biological relevance, pharmacokinetics, repro- cal material for such markers is feasible but logisti- ducibility of measurement of the marker and the cally more complex. For early outcome markers, optimal conditions for collecting, processing and collection of material (e.g. precancerous lesions) is storing biological specimens in which the marker logistically feasible in a hospital setting, but is to be measured. becomes more difficult in the population setting. Characterization studies involve assessing inter- End-points assessed routinely at a variety of individual variation aid the genetic and acquired institutes may produce major problems of stan- factors that influence the variation of biomarkers dardization of methods (e.g. histological diagnosis). in populations. This includes assessing the fre- Except for some routinely registered biological quency or level of a marker in populations, identi- characteristics (e.g. receptor status in breast сап fying factors that are potential confounders or cer), even greater problems attend the issue of effect modifiers, and establishing the components identification of biologically or pathologically of variance in the biomarker measurement, la- defined disease subsets. boratory variability, intra-individual variation and Most of the considerations that make pros- interindividual variation. The ratio of intra- pective cohort studies an attractive study design individual variation to interindividual variation apply to all methods of exposure assessment, has important implications for study size and including biomarkers. The strengths of such stud- power. ies include the fact that exposure is measured Applied transitional studies assess the relation- before the outcome, that the source population ship between a marker and the event that it marks, that gave rise to the cases is explicitly defined, and namely exposure, pre-clinical effects, disease or that participation can be as high as 100% if speci- susceptibility. Applied studies are often of cross- mens are available for all subjects and follow-up is sectional or short-term longitudinal design, and complete. Weaknesses of prospective cohort stud- are not intended to establish or refute a causal rela- ies involving biornarkers include the usually small tionship between a given exposure and disease. number of cases of each of many types of cancer, Rather, they are intended to assess whether rele- the lack of specimen if the biomarker requires large vant correlations exist and if they are strong amounts of specimen or unusual specimens, degra- enough to be useful in full epidemiological studies. dation of the biomarkers during long-term storage,

s Applicatioл of Biomarkers in Cancer Epidemiology and the lack of detail on other potentially con- lected from all cohort members and stored at the founding or interacting exposures. Cohort studies beginning of the study; the controls are then arе, in principle, the preferred method for studying selected throughout the course of the study, ideally the temporal sequence of intermediate end-points; at the time that each case is diagnosed. An advan- however, the costs and complexity of repeated tage of the nested case—control approach is that screening may often make this difficult. biomarkers can be measured in specimens matched Important issues in prospective cohort studies on storage duration and casecontrol sets can be involving biomarkers, especially of exposure, analysed in the same laboratory batch, reducing involve the frequency and timing of specimen the potential for bias introduced by sample degra- collections. А major concern in cohort studies of dation and laboratory drift. On the other hand, a short duration (as in case—control studies) is the case—cohort design may be used when sample col- possibility that the disease process has influenced lection from the entire cohort is not feasible, or the biornarker level among cases diagnosed within when the cost of storage or analysis of each sample 1-2 years of the specimen being collected. Iп stud- is prohibitive. This design involves collecting ies of longer duration, there may be considerable biological samples from the cohort at the begin- misclassification of the aetrologicallÿ relevant ning of follow-up and then collecting samples exposures if specimens have been collected only at from cases as they occur. However, as samples are baseline. This misclassification occurs not only being taken at different times for cases and 'controls' because an individual's exposure level may change (i.e. the reference sample from the entire cohort), systematically over time, but also because there bias will be introduced if sample degradation or may be considerable intra-individual variation laboratory drift occurs over time. Furthermore, (from day to day or even hour to hour) in bio- the case—cohort approach may lead to laboratory marker levels. The effects of intra-individual varia- personnel being unblinded to case and control tion can be reduced by taking multiple samples, status. but this greatly increases the expense of sample In case—control studies, biomarkers of internal collection and storage and the burden on study dose or effective dose are appropriate when they subjects; similar considerations apply to taking are stable over a long period of time (e.g. carrier samples at several points in time in an attempt to status for infectious agents, such as HBsAg) or estimate time-integrated exposures or exposure when exposures have been constant over the rele- change. An alternative approach is to estimate the vant exposure period; however, it is essential that extent of intra-individual variation, and the mis- they are not affected by the disease process, diag- dassification involved in taking single specimens, nosis or treatment (except in those circumstances by taking multiple specimens in a sample of the when it is possible to collect specimens prior to the cohort. This information can be used to correct for commencement of treatment). For genetic suscepti- bias to the null introduced if the misclassification bthty markers, case—control studies are highly appro- is non-differential, and therefore de-attenuate priate. Clinic-based case—control studies are partic- observed relative risks. ularly suitable for studies of intermediate end- Use of biomarkers of disease susceptibility in points, as these end-points can be systematically cohort studies raises significant ethical issues, par- measured. Population surveys (or screening) are a ticularly if there is repeated contact with study sub- prerequisite for the identification of cases in popu- jects. In particular, informing the cohort members lation-based case—control studies of intermediate of their biomarker level is problematic if the bio- end-points. marker is not considered to be sufficiently predic- Clinic-based case—control studies are of particu- tive of disease, and if there are rio known preven- lar value for studying the etiology of precancerous tive steps cohort members can take to reduce their lesions (e.g. early cervical intraepithelial neopla- risk of the disease. sia). None the less, the relation between these A nested case—control study or a case—cohort intermediate disease markers and the cancers of design can be used to reduce the expenses of data which they are precursors is usually difficult to collection and sample analysis. In a nested study for both ethical and logistical reasons. In case—control study, biological samples may be col- case—control studies involving cancer as the end-

4 Wойuhop report point, biomarkers of exposure (though not of any studies, in particular in the control of confounding. given genotype) are of limited value, as the disease Such studies often do not seek to make causal infer- itself, its therapy, or behaviour changes iii patients, ences about exposure– disease relationships; may alter any exposure marker or biological rather, they are aimed to elucidate plausible mech- process thought to be a precursor of disease. anisms (Kuopio et al., 1995). Case–case comparisons can be useful in the study of biologically defined disease subsets when Cross-sectional studies differences in the etiology of such subsets are being Cross-sectional study designs, with single or examined. However, if a control group is not in- repeated sampling, have been frequently used in cluded, any heterogeneity in the exposure associa- the validation of markers of exposure and of dis- tions among disease subsets may be the result ease. They can provide a useful snapshot of the of increased risk in one subset as opposed to a relationship between exposure and susceptibility decreased risk in another. factors, on the one hand, and the selected biomarker, on the other, in populations whose expo- Intervention studies sure can be well characterized. Frequently, the Intervention studies (trials) assess prospectively studies have involved industrial populations, whether an intervention is efficacious in changing chemotherapy patients and smokers—all groups the frequency of a marker of exposure or dose (e.g. exposed tô 'model' carcinogens—and have com- measures to encourage smoking cessation and pared them to appropriate non-exposed groups. either serum cotinine or protein or DNA adducts) Such studies can also be used in healthy individu- or pre-clinical condition or disease (e.g. р53 mu- als to establish the possible range of measurements tation, dysplasia and cell proliferation in trials of protective factors, e.g. monitoring of vegetable using antioxidants to prevent second tumours). intake by measurement of caroterloids or phenolic Biomarkers can be particularly valuable in assess- compounds in plasma and urine. This design can ing compliance with the intervention (e.g. serum be used to establish an association between the R-carotene in a randomized trial of (3-carotene). biomarker and the exposure under study. Markers of susceptibility (e.g. GSTM1 polymorphism) could be used to determine whether groups Family-based studies in various intervention arms are comparable. Family studies are intended to address scientific Intervention studies may also be conducted as questions related to four general areas of research: transitional studies when there aie stiff questions about the response of biomarkers following an 1. Does familial aggregation exist for a specific intervention. Critical in all types of intervention disease or characteristic? studies is the need to address those same issues 2. Is the aggregation due to genetic factors or that are important in cohort studies, including lab- environmental factors, or both? oratory drift, loss to follow-up, and intervening 3. If a genetic component exists, how many exposures that may affect the biomarker. genes are involved and what is their mode of inheritance? Ecological studies 4. What is the physical location of these genes Ecological studies use groups rather than individuals and what is their function? as units of observation. They include studies of geographical differences and tinte trends in disease Using linkage methods, biomarkers that charac- incidence and prevalence. While ecological studies terize known DNA haplotypes have been instru- provide useful information on exposure, disease mental in locating a number of cancer-related genes. and modifying factors, they are, in most cases, These studies require relatively large amounts of inadequate to establish causal relationships. germline DNA (usually represented by DNA ex- Although, in general, ecological studies based on tracted from white blood cells). This is because biourarkers have the same advantages and limita- there are multiple marker loci that require evaluations as the other types of ecological studies, it is tion. Once a locus is identified, further DNA is possible to conduct carefully planned ecological necessary to identify the gene. Often, fibroblast or

5 Application of Biomarkers in Cancer Epidemiology

Epstein-Barr virus (EBV)-transformed cell lines are ships between exposure, marker and disease are, in required to provide adequate DNA for these studies. most cases, obscure, the capacity to establish the special concern must be paid to ethical and presence or absence of misclassification in the counselling issues in family studies in which bio- interpretation of the findings of biomarker-based marker results for an individual may also have pre- studies is limited. dictive value for other members of the family. The use of biomarkers of exposure does not reduce the need to control for confounding and, in General issues in study design and analysis some instances, the use of biomarkers may actually As with other epidemiological studies the design introduce confounding into a study (Pearce & and statistical analysis of a study involving bio- Boffetta, this volume). Confounding can be con- markers involves iп general: (1) relating a particular trolled in the analysis by stratifying the data into disease (or marker of early effect); (2) to а particular subgroups according to the levels of the con- exposure; while (3) minimizing bias; (4) controlling founder(s), or by the use of mathematical model- confounding; (5) assessing and minimizing ran- ling. However, problems of multicollinearity can dom error; and (6) assessing interactions (Pearce & occur when variables that are highly correlated Boffetta, this volume). (e.g. serum levels of various micronutrients) are Many studies using biomarkers of disease are of entered simultaneously into a model; this will cross-sectional design and measure the prevalence make the model unstable and can lead to invalid of the disease state, which is dependent on both its effect estimates. incidence and its duration. Thus, in a study exam- Some biomarkers may be surrogates for inter- ining markers of cell damage as an effect of expo- mediate stages in the disease process. For example, sure to known or suspected carcinogens, the results when the relationship between reproductive risk would depend on factors such as the turnover of factors and cervical cancer is adjusted for HPV sta- the cells in which the marker is measured and the tus, the relative risks for number of sexual partners capacity to repair the damage. are reduced. This suggests that HPV is a mediator When deciding whether to measure exposure of the relationship between the reproductive risk with biomarkers, it is important to consider how factor and cervical cancer, helping to substantiate useful and informative the biomаrkers are in rela- the causal role for HPV. In cases such as this, the tion to the study hypothesis. This is a particular biomarkers should not be treated as potential con- issue for exposures that change over tinge; for founders, i.e. they should not be adjusted for in the example, a biomarker may be more valid than analysis. None the less, analyses alternatively a questionnaire in assessing exposure to HPV, including and excluding the putative intermediate whereas a questionnaire will be more valid than a can be used to help determine causal pathways. biomarker in assessing cigarette smoking 10 years When choosing the method of exposure or dose previously. It should be stressed that the concept of assessment, it is important to consider their impli- useful and informative' depends on: (1) whether, cations on study size, particularly when an expen- in principle, it is most appropriate to study exter- sive or invasive method may be used. An additional nal exposure or internal dose (i.e. which hypothe- consideration in study size estimation is the ratio of sis is being tested); (2) what information is avail- the number of assays per individual and the number able for the etiologically relevant time period; (3) of individuals in the study. lany biomarkers show whether the exposure or dose measurement is sub- marked variation from day to day within the same ject to intra-individual variation; and (4) whether individual (in part from problems of repeatability it is desirable and possible to obtain (with a bio- of the laboratory tests, but also from genuine dif- marker) information on interindividual variation ferences), and in some cases the intra-individual in exposure or biological response. variation may be greater than the irnterindividual When exposure or disease is measured with bio- variation (e.g. 24-h urinary sodium measurements). markers, it is important to ensure that any mis- Thus, it is important to take into account the trade- classification is non-differential (i.e. it applies off between including more participants, on the equally to the groups being compared) and is as one hand, aid gathering multiple samples from small as possible. However, because the relation- each participant, on the other. It is essential that Workshop report the relative magnitude of intra-individual varia- processing, evaluating intra-individual and inter- tion of a biomarker is adеquately characterized. individual variation in the biomarker in target Biornarkers present better opportunities for populations, determining laboratory assay varia- assessing interactions between genetic and envi- tion, and studying the impact of exogenous and ronmental factors. In particular, to infer interac- endogenous influences on the biomarker. tion, biomarkers of genetic susceptibility should Regardless of the specific protocol used for the show a higher disease risk in exposed susceptible collection, processing and analysis of a specific bio- groups than in exposed non-susceptible and in marker, two important issues to consider are the non-exposed groups. None the less, such testing storage of biological samples and the evaluation of for interaction usually requires a substantial sources of variation in biomarker measurement. increase in study size. This section will therefore review storage and vari- As is the case for other aspects of epidemiological ation issues following the discussion of individual research, if the relative risk between a biomarker biomarkers. and exposure or outcome is not very strong (e.g. The following biomarkers have been selected to less than 10), then results from multiple studies demonstrate the wide range of assays available for will be required before agreement on the existence incorporation into epidemiological studies, and to of an association can be reached. It is critical, comment on their utility for various study designs therefore, that the results from all existing studies relevant to the study of cancer. This is not meant are available. Unfortunately, the tendency for to be a comprehensive list; other important cate- investigators and journals to publish only 'positive' gories of biomarkers in blood (e.g. immunological results may bias the literature in favor of positive markers) and in normal and preneoplastic tissue studies, a phenomenon often called 'report bias' are not addressed here. or 'puNicatiori bias'. Biomarker research may be even more prone to this bias than other fields of Biomarkers of biological agents epidemiological research, as multiple biomarkers Biological agents associated with chronic infection may be assessed, sometimes at relatively low cost. and subsequent development of cancer are mea- It is important that investigators publish, or sured using serological or nucleic acid markers make available in abstract form, all their results (Munoz & Bosch, this volume). An example of from studies in which the results are reliable, nucleic acid-based biomarker is HPV DNA detec- even if these results are 'null'. Publication bias tion where the presence of type-specific DNA at a can be particularly relevant when several analyses given time is measured by PCR-based assays. of the data are done on subgroups of the study Cohort studies having high-grade squamous population, defined according to сhaгaсtегistiсs intraepithelial lesions as the end-point have shown such as tumour markers or susceptibility gene that the infection preceded the disease. HPV DNA polymorphisms. Such multiple analyses increase infections are often transient, especially in young the probability to obtain significant results by women. Therefore, repeated sampling is required to chance only. Although these analyses may provide assess persistent HPV infections. Accurate serologi- useful information on effect modifiers, the selec- cal assays aimed at distinguishing transient from tive reporting of significant or 'positive' results persistent infections need to be developed. should be discouraged. Another example is the hepatitis B virus (HBV), for which there are serological markers that distin- Incorporation of biomarkers into epidemiological guish between past and persistent infections. HBV studies DNA detection in sera further refines the assessment Biological markers undergo a process of discovery, of exposure. These markers have been used in all characterization and refinement before use in types of epidemiological design (cross-sectional, epidemiological studies. Animal studies contribute case—control, cohort arid intervention studies). to the development and validation of biomarkers and provide insights into the mechanisms of the Biomarkers of internal dose for chemical exposures multistage process of carcinogenesis. TraпsitioпaI Biomarkers of internal dose of external chemical studies assist in this process by optimizing sample exposures are measurements of a parent compound

7 Application of Biomarkers in Cancer Epidemiology or its metabolite(s) in an accessible biological in using this approach. Biomarkers can be useful in matrix, such as serum or urine (Institute for nutritional epidemiological studies at two different Environment and Health, 1996). They have poten- levels (Kaaks et al., this volume). First, biomarkers tial applications in several types of epidemiological may be of interest as potentially more precise, studies (Coggon & Friesen, this volume). These more specific or more objective measurements (or include studies to establish the importance of dif- correlates) of the intake levels of specific foods or ferent sources of exposure as determinants of total food constituents, compared to measurements dose; studies to validate other methods of expo- obtained with questionnaire and interview methods. sure assessment such as the use of questionnaires; Secondly, most biomarkers can also be seen as indi- studies to establish that a chemical reaches a sus- cators of a nutritional/metabolic status, which may pected target tissue; cross-sectional studies relating be intermediate between `exogenous' nutritional exposure/dose to biomarkers further downstream in lifestyle factors (e.g. the composition of diet, phys- the exposure—disease continuum; and case—control ical activity) and disease risk. When used as an and cohort studies relating exposure/dose to disease. indicator of intake, they are useful in four different The utility of an internal dose marker in types of study: validation (comparison of a mea- case—control and prospective cohort studies surement against a gold standard), calibration depends, in part, upon the half-Iife of the external (comparison of two measurements), observational agent от its metabolites in the body; the pattern epidemiological and intervention studies. They are of the exposure it is measuring (e.g. regular, daily especially important in validation or calibration exposure versus infrequent, episodic exposure); studies where biomarkers represent an additional whether secular trends have occurred in that expo- category of measurement, the `random' errors (i.e. sure (e.g. smoking cessation); and direct or indirect variations that are independent of individuals' true influences of the disease process. The information habitual intake levels) of which can be assumed to that a biomarker of internal dose provides must be be statistically independent of those of the ques- compared to the availability and quality of other tionnaire and other methods. The factors that are sources of data (e.g. questionnaires, environmental important in deciding whether to use a given bio- measurements, medical records). Essentially, all marker in a particular study include whether it is exposure measures misclassify some subjects on a good indicator of intake; whether it is a long- or their usual pattern of exposure; it is the relative short-term marker; whether there is need for ability of different sources of data to place individ- multiple measurements; whether it is acceptable uals correctly into exposure categories that is to the researcher and the subject; and whether it is important. In case—control studies, questionnaires compatible with the design of the study itself remain the primary source of exposure data. There (case—control of early disease, case—control of late are, however, a few instances where a measure of disease, or a cohort study). Biomarkers can also be internal dose may be a more suitable measure of informative if they provide an integrated biologi- cumulative exposure, particularly in the study of cal measure of intake, lifestyle and metabolic cases with preneoplastic lesions or early disease. processes. For example, measurements of red cell The ideal biomarker should persist over time (e.g. folate coupled with serum homocysteine would fat-soluble substances such as DDT metabolites) offer a picture of medium-term intake of folic acid, and should not be affected by disease status. which is distinguishable from transient fluctua- Markers of internal dose may be useful in prospec- tions in dietary intake (Green & Jacobsen, 1995). tive cohort studies as long as components of variance of the biomarker are well characterized, since Biomarkers for endogenous hormones the problem of reverse causality (i.e. disease status Biomarkers are available to measure specific` affects the level of the biomarker) is minimized. endogenous hormones (Lemaster & Schulte, 1993; Hulka et al., 1994). Cross-sectional studies, where Biomarkers of dietary intake and nutritional status the biomarker is the outcome measure, are impor- Dietary intake is usually assessed by various types tant for the assessment of differences in marker of questionnaire and other methods (e.g. diaries); levels between subjects with different characteris- however, there are many sources of error involved tics (e.g. sex, race, anthropometry, geographical Workshop report location). A case—control study is not an appropri- Cross-sectional studies evaluate exposure— ate study design to asséss associations between adduct relationships in populations currently metabolic and hormone markers and disease out- exposed to agents of concern. Elevated adduct come, since marker sampling occurs after the levels in an 'exposed' versus 'unexposed' popula- clinical appearance of disease. This problem may tion may suggest that the exposure is associated be less significant от insignificant if the outcome is with a higher cancer risk, given the increasing evi- a pre-clinical condition or a very early lesion. dence that DNA damage represents a primary Cohort studies including case—control studies mechanism of carcinogenesls. In this instance, the nested within cohort studies, are ideally suited for adduct is not being used as a dosimeter, but rather evaluating thé association between hormone bio- as evidence of a potentially harmful response in markers and disease. sampling issues important in vivo. Such studies should be regarded as providing cohort studies are the frequency of sampling and supporting evidence only, until the association of the timing of collection in relation to events which adducts with subsequent development of cancer may influence measurement (e.g. stage of men- has been demonstrated. It is relevant to note, how- strual cycle, menopause, oophorectomy, ageing, ever, that in the presence of limited epidemiological medications). evidence of a chemical's carcinogenicity, the demonstration in humans that the chemical causes Macromolecular adducts as biomarkers of exposure a dose-dependent increase in macromolecular to reactive chemicals adducts or in other biomarkers that reflect geno- Chemicals can bind covalently to cellular macro- toxic damage provides supporting evidence that molecules such as nucleic acids and proteins (Wild the chemical is carcinogenic to humans. This type of & Pisani, this volume). The product of this addi- evidence has been used, for example, in the recent tion of a chemical moiety to a macromolecule is IARC evaluation of carcinogenicity of ethylene termed an `adduct'. The adduct may be highly spe- oxide within the Monographs programme (IARC, cific for the carcinogen of interest, but not neces- 1994). sarily specific for a given exposure because of mul- The utility of using adducts as markers of bio- tiple sources of the carcinogen within the envi- logically effective dose is limited in case—control ronment. Adduct formation normally occurs after studies due to the relatively short half-life of most the metabolic activation of the carcinogen; DNA adducts evaluated to date. They may have utility in prospective cohort studies, again with the caveats repair znлу follow adduct formation. As a result, measured adducts represent an integration over previously described for all exposure markers. time of carcinogen exposure and interindividual variations in carcinogen metabolism, DNA repair Biomarkers of somatic сеll mutations and other host factors. The persistence of adducts Somatic mutations provide evidence of irreversible is determined by the chemical stability of the genetic damage (Albertini & Hayes, this volume). adduct itself and the turnover of the macromole- Furthermore, specific mutations (mutation spec- cule to which the chemical is bound. In practice, trа) may, in principle, identify exposures to specific this gives a half-life of adducts on proteins (haemo- agents or mechanisms; however, measurements of globin and albumin) of a few weeks to months, mutations are usually less sensitive in this regard while DNA adducts may have half-lives of a few than are other biomarkers of exposure (e.g. hours to several years depending on the cell type metabolites, adducts). 5отаtiс mutations should concerned. Adducts of more remote exposure have their greatest utility in epidemiological stud- (such as modified amino acids in histone proteins ies when it is possible to establish a qualitative from non-dividing cells) would represent a major association between exposure and specific muta- advance in the utilization of these markers in epi- tions, thus identifying the agents of concern. They demiological studies. Adduct measurements can be also have promise as surrogate markers of out- made in blood and exfoliated cells, and metabo- come. It must be emphasized, however, that for Iites of adducts can be measured in urine. The this last purpose, mutations must initially be quantity of material required is dependent on the assumed to be associated with increased risk of can- assay sensitivity for a given adduct. cer. In-vivo mutational response may also be used

9 Application ы Biomarkers in Cancer Epidemiology to define interindividual differences iп sensitivity to painted arid unpainted chromosomes. Improve- mutagens or carcinogens. rпents in the techniques of molecular cytogenetics Somatic mutations have potential application continue to be made, and new DNA probes are in case control studies when the exposures of con- being developed iii a regular basis. Cytogenetic cern are chronic and stable, and the disease does assays have been performed on metaphase cells, not alter mutagenic responses in the test cells. but the more recent applications of F151 to inter- Their greatest utility, however, may be in studies phase cells means that cell culture is no longer that use mutations as an intermediate end-point required and that relevant target tissues can be to evaluate genetic responses under various expo- examined directly for chromosome reanange- sure conditions and in studies concerned with pre- ments. The analysis of interphase cells appears vention. There is a need in the future for rapid, iriex- to have significant promise for use in future epi- pensive and sensitive molecular assays that may demiological studies and may make the banking of increase the range of genes and tissues amenable for the relevant tissues from a large numbers of sub- study; this, in turn, should facilitate the evaluation jects feasible. and potential use of those biomarkers as surrogates of disease-relevant genotoxic responses. Biomarkers of genetic susceptibility Biomarkers have increasingly played a role in Biornarkers of cytogenetic damage studies that evaluate the role of genes in cancer Chromosomal aberrations are commonly found in (Caporaso & Goldstein, this volume). Family-based tumour cells and, in some cases, there is very good studies have led to the mapping of several cancer- evidence that they play a causative role in carcino- related genes characterized by high penetrance, genesis (Tucker et aL, this volume). Two prospec- high absolute and relative risk, but low attributable tive cohort studies have shown that chromosomal fraction for the most common cancers. The envi- aberrations measured in peripheral lymphocytes ronment plays a variable but arguably small role in were associated with an increased risk of cancer the manifestation of these genes. Other genes (e.g. (Hagmar et a1., 1994; Bonassi et aL, 1995). Non- metabolic polymorphism genes that lie in the target cells may be appropriate tissue for this metabolic pathways of carcinogens and therefore marker. Translocations, in particular, are useful for have a mechanistically plausible role) are charac- quantifying certain types of acute and chronic terized by modest relative risks, low absolute risks exposure, as well as exposure that occurred many and high attributable risks. The environment is years previously. Translocations are therefore a rea- crucial in determining the effects of these genes. A sonable and appropriate biomarker for use in cross- major contrast between these two is the low preva- sectional studies because they provide a direct and lence of the former and the high prevalence of the quantifiable indication of DNA damage. In case— latter. control studies, the analysis of stable transloca- For a given genotype, the phenotype is deter- tions in early, local disease may provide an insight mined by the penetrance. The use of phenotype or into accumulated cytogenetic damage and its link genotype in a specific study depends on numerous with cancer. It is possible, however, that the fre- factors. In general, the methodology of genotyping quency of translocations is affected by advanced has increased because of rapidly improving tech- disease, but this remains to be established. The use nology. Recent advances in the analysis of genetic of translocations in prospective cohort studies polymorphisms using DNA present in serum, in could have substantial utility but is limited by the formalin-fixed, paraffin-embedded pathology sam- logistic constraints of culturing or cryopreserving ples and in material collected non-invasively by lymphocytes from a large number of subjects. buccal swabs and oral rinses will provide new The analysis of translocations by fluorescent in- opportunities for incorporating biomarkers of situ hybridization (FISH) with probes that paint genetic susceptibility into epidemiological studies. whole chromosomes represents an important advance in cytogenetics; FISН is a rapid, sensitive Tumour biomarkers and highly reproducible technique which relies Tumour markers include characteristics of tumours on the detection of colour `junctions' between at the anatomical, histological, serum, chromosomal

10 Workshop report and molecular levels (Zhang et al., this volume). way that the error in one measure is not repeated Using tumour markers in the assessment of etio- in another, e.g. the two specimens are collected at logicaI heterogeneity represents а progression from different times over the relevant etiological time these previous studies and may have greater speci- period, and handled, stored and analysed with the ficity for particular patterns of exposure. Tumour variation in specimen collectors, laboratory tech- markers can readily be incorporated into case series nicians or batches that would occur in the parent and case—control studies, as well as into prospec- epidemiological study. tive studies that study the relationship between Similarly, reliability studies could be designed to biomarkers in an initial tитоит and the risk of partition the specific components of error: the developing second primary tumours or the rela- effects of handling and storage; laboratory varia- tionship between precursor lesions and subsequent tion; and short-, medium- and long-term biologi- malignancies. There is a rapidly increasing ability cal variation. In any case, researchers should not to evaluate а range of tumour markers in formalin- assume that small laboratory error implies that a fixed, paraffin-embedded tissue. Frozen tissues are measure is good, because these other sources of easier to assay, but more difficult and costly to error introduced by the design and needs of the obtain and store for epidemiological purposes. epidemiological study can be far greater than the laboratory measurement component of error. Evaluating sources of variation in biomarkers Before embarking on an epidemiological study storage of biological samples that uses a biomarker, it is important to under- The use of biological markers in epidemiological stand the potential measurement error in the bio- studies often requires storage of the relevant biomarker (Gompertz, this volume; Vineis, this vol- logical samples for a period of time, which may ume; White, this volume). The researcher's first vary from a few weeks up to years or even decades, concern should be to prevent or rule out, as far as depending on the study design (i.e. cross-sectional, is practicable, differential measurement error. prospective) and the timing of the laboratory Because differential measurement error can bias analysis (Landi & Caporaso, 1997, this volume). the odds ratio in either direction, its presence to Factors that can be considered in the choice of any appreciable degree in a biomarker will invali- storage method include the type of biological date its use in an epidemiological study. material; the type of laboratory analysis planned; Differential measurement error is a particular con- the duration of storage and spread over time of cern in case—control studies and among the early laboratory analysis; and the logistical and practical cases in cohort studies when the marker may be conditions. For long-term storage of a large collec- influenced by pre-clinical disease, by the effects of tion of blood samples, the safest and, in most cases, the disease after diagnosis or by treatment. the most efficient method of storage is at liquid When differential measurement error is not nitrogen temperature. likely to be present, the researchers should focus storage temperature should be as low as possible, on assessment of the non-differential error, or at depending. on the biological material, and should least some of the major components of error in the ensure stability of all potential analyses over long biomarker. Error components include laboratory periods of time. For very long-term storage, it is variation, variation from specimen collection and worthwhile to store pools of biological specimens storage, and biological variation. The latter may be containing known concentrations of the aпalytes short-term (e.g. day to day), medium-term (e.g. of interest in order to be able to monitor possible seasonal), or long-term (e.g. variation over an etio- degradation over time. A stable isotope standard logically relevant period of years). Ideally, one may be useful in such cases. would conduct a validity study in which the biomarker to be used was compared to a perfect (true) Application of biomarkers measure. However, much of the total error can be Biomarkers may be used in epidemiological studies measured in a well-designed reliability study to increase the information obtained from classical which requires the collection and analysis of two study designs and to expand the areas of scientific (or more) specimens from a group of subjects in a inquiry to which epidemiology can contribute. Use of Application of Biomarkers in Cancer Epidemiology biomarkers in epidemiological research has grown response rate with respect to collection of biologi- significantly in recent years and can be expected to cal samples than community-based studies, and increase rapidly in the years ahead. The areas of differential participation by cases and controls is application include research on etiology of disease; less of a problem. For casе-сontrol studies of pre- correlation between external exposure and internal cursor lesions, no population-based registries exist dose; susceptibility and gene- nvironment inter- from which to identify cases, whereas such cases action; clinical trials; disease mechanisms and can be identified in the health care setting and, pathogenesis; and cancer prevention. given appropiate survey conditions, in random samples of the population. In cohort studies, Etiological epidemiological research where the biomarker is the exposure of interest, In etiological research, biomarkers of exposure rep- uriexposed subjects, defined as people not carry- resent the independent variable and are used to ing the biomarker, may be identified through the predict disease occurrence. Research of this type health care setting. Tissue specimens may be avail- has been limited in the past because of the need for aЫe only through pathology departments associ- transitional studies to develop, characterize and ated with hospitals. Opportunities to obtain speci- evaluate biomarkers in their application. When mens may be enhanced if health care providers are biomarkers are tested as outcome variables they are co-investigators in the research. The method for useful in often being much more prevalent in selecting study subjects is an important potential the population than cancer itself and they occur source of bias in clinic-based studies. In studies much closer in time to the exposure than does can- based on biomarkers, however, it may be less cer; exposure will therefore be measurable more important than in other types of epidemiological accurately by whatever method is used. The major- studies, because the characteristics of the subject ity of epidemiological biomarker research reported measured by the biomarker (e.g. genetic polymor- to date has been from transitional studies. As this phism status) may not be linked to the probability body of literature has accumulated, it is now of inclusion in the study. similarly, it is unlikely becoming possible to use selected biomarkers in that markers of exposure and effect are affected by the context of classical etiological research. referral and detection, since the individual's bio- At current levels of development of biomarkers, marker status prior to the study is not known. In there will still be many circumstances in which most cases, however, it is unknown how subjects questionaire methods of assessing exposure are are selected with respect to the biomarker, and the complementary or preferable to technically sophis- passibility of selection bias should always be ticated biomarkers of the same exposure. Assays considered. of urinary levels of metabolites of tobacco components such as nicotine, for example, reflect only Exposure-internal dose and exposure-biological short-term consumption. 5uсh assays tell us about response correlations smoking habits in the relatively short time before Transitional studies of exposure frequently corre- measurement; questionnaires are still preferable to late external exposure measured in a traditional determine long-term exposure. fashion (e.g. interview or ambient monitoring) Etiological studies using biomarkers generally with internal dose measured in body fluids or employ case-control or cohort epidemiological tissues. These studies are generally cross-sectional designs. Modifications of these are the case-case or short-duration longitudinal in design. The design and the applied transitional study. The exogenous measure is considered the independent case-control study nested within a cohort offers variable, and the internal dose the dependent vari- distinctive opportunities by using blood or tissue able. Rather than an internal dose measure, the collected in the past. dependent variable can be a biological response The use of biomarkers in case-control studies may marker, e.g. a p53 signature mutation associated necessitate greater use clinic-based (or hospital- with aflatox1rn exposure (Links et al., 1995). Such based) studies (Potter, this volume), Clinic-based transitional studies frequently focus on heavily studies of biomarkers may be more feasible, practi- exposed (and unexposed) persons in order to min- cal and economical. They also have a higher imize sample size aid maximize the opportunity

12 Workshop report to identify arid characterize the biomarker. If the exposure', increasing risk of disease in the entire correlation between the two types of exposure study population. Usually, however, susceptibility measurements is high and can be quantified with is manifested as an effect modifier of exposure. If good precision, future studies may be able to the population is stratified into two groups, based employ the more easily obtained and inexpensive on presence or absence of the susceptibility gene, exposure measurement. the effect of the exposure variable may be more evident in the susceptible group. It may also be susceptibility and gene—environment interaction useful to stratify on levels of the exposure variable, Studies of susceptibility and gene environment e.g. smoking or age groups, to identify qualitative interaction are emerging as an important compo- and quantitative aspects of the association nent of epidemiological research (Garte et ai., this between the susceptibility marker and disease volume). This represents a significant development across levels of the exposure. in epidemiology; it emphasizes the importance of individual differences and the combined influence Cancer prevention of genes and environment in determining disease In cancer epidemiology, biomarkers can be used to risk. Although relevant genes lie in a continuum offer quantitative insights about biological events with iespect to prevalence and expressivity, a dis- occurring at different stages of the pathologic tinction should be made between single gene process (McMichael & Hal, this volume). The mutations of large effect and polymorphisms of improved knowledge of the natural history of can- small effect. BRCA1 exemplifies the former, and cer should have application to research aiming at metabolizing enzymes for substrates that form cancer prevention and control. At least three carcinogenic metabolites illustrate the latter. aspects of cancer prevention or cancer prevention Intermediate situations exist in individuals research may make use of biological maikers. They heterozygous for some genes, such as ataxia telang- are screening, community-based intervention tri- iectasia, which may affect a large proportion of the als and the monitoring of biomarkers as risk factors population, aid for which there is some evidence for disease. Screening may concentrate on the of increased risk of various cancers, including identification of precursor lesions or early stage breast cancer (Easton, 1994). Because estimates of disease (e.g. cervical cytology), high-risk individu- cancer risk in relation to single gene mutations als (e.g. prostate-specific antigen, PSA) or suscepti- have been derived from high-risk families, it is not bility markers (e.g. BRCA1). Cervical cytology is yet known what the cancer risk is for gene carriers well established as an effective tool to reduce within the general population. There is a need for mortality from invasive cervical cancer. PSA is con- epidemiological studies of cancer in the general troversial because it is not specific to invasive pro- population to study risk in single gene carriers who static cancer or its precursors. It cannot distinguish are not members of high-risk families. between precursor lesions that will invade and the Several methodological issues have arisen in majority that will not, resulting in a significant the context of these studies. Lack of concordance amount of over-treatment and morbidity. Further- between genotype and phenotype has been re- more, and crucially, there is no evidence that ported and it is not easy to discern which is more screening reduces mortality from prostatic cancer. informative with respect to disease effects. However, Screening for BRCA1 became possible only recently some advantages of studies of susceptibility and when the gene was cloned and sequenced (Futreal gene—environment interactions are the potential et aL, 1994; Miki еi аL, 1994). Currently, testing for for case—case studies (Begg & Zhang, 1994) and for BRCAI has been suggested only foi women in studying 'exposed' subjects only. high-risk families in whom the gene is known to Khoury etaI. (1988) has described six models for segregate, or for women with multiple first-degree gene—environment interaction. These concepts relatives who have had an early age at diagnosis of may be simplified by considering the possibility of breast cancer. Even in these situations, screening is a main effect and effect modification for both the not universally accepted because of the limited risk susceptibility marker and the exposure. For some and disease management options, which markediy genes, the susceptibility marker can behave as an reduce the utility of the genetic information. Even

13 Application of Biomarkers io Cancer Epidemiology

in the cases of high penetrance cancer genes, how- Certain laboratory and pathology studies may also ever, there is the possibility to contribute to cancer yield pertinent evidence, as in the case of the colon prevention by modifying the environmental fac- adenoma to carcinoma sequence. With markers tors that interact in the carcinogenic process. having attributable fractions substantially less Community-based intervention trials for risk than 1, there may exist alternative pathways to reduction may benefit from biological markers. For cancer that bypass the marker in question. An example, smoking cessation trials have found that exposure or intervention may operate through the urinary cotinine levels may be a better indicator alternative pathway(s) in a way that offsets the of smoking status, when measuring the trial cancer effect mediated by the original marker. For outcome, than reports on interview or question- such markers, inferences to cancer are problematic naire. (Schatzkin et al., 1990, and this volume). Monitoring biomarkers of exposure to carcino- Although there are well-recognized difficulties in genic agents in populations may reveal trends and extrapolation of carcinogenesis and chemopreven- assist in risk assessment in high-risk groups. hon data from animal experiments to humans, animal models provide opportunities for biomarker Clinical prevention trials research. Linkages between mechanistic pathways The use of biornarkers in chemopreventive trials in different species may ultimately be useful in raises important methodological issues. The test- resolving problems in inter-species extrapolation. ing of chemopreventive agents in clinical trials using the end-point of cancer incidence requires a Ethical aspects of biomarkers in cancer study period of many years, very large sample sizes epidemiology and great expense. Therefore, short-term, smaller The use of biological markers presents potential clinical trials that use surrogate end-point bio- ethical issues because Ыошarkеrs are obtained markers (5EB) have had to be developed. A SEB from an individual's unique tissues and can be may be defined as an early change during the used to provide important data about exposures, intraepithelial, pre-invasive phase of neoplastic biological effects and susceptibility to cancer progression, at the molecular, cellular or tissue (Schulte et al., this volume). The ethical issues arise level, whose response to a chemopreventive agent from the possibility of abuse or misuse of bio- predicts the effectiveness that the agent would marker data and failure to respect the rights of per- have in a large clinical trial using the end-point of sons participating in research. Biomarker data can cancer incidence reduction (Boone & Kelloff, this be misused by failing to keep data confidential and volume). Examples of SEBs include computer- by using it, or allowing it to be used, to stigmatize assisted quantitative image analysis of pathological research subjects. Research subjects have a right to specimens, which measures nuclear features privacy as well as a right to be told of risks of par- (altered size, shape, and chromatin texture), aid ticipating in biomarker studies and of any clini- cytological features by binding to chromages—anti- cally important findings. There remains a range of body conjugates. Markers and indices of prolifera- opinion whether banked specimens collected for tion (e.g. Ki-67, PCNA), oncogene mutation or one study or purpose may be used for another. amplification, and allelic loss and other alterations Generally, the degree of risk in participation and may also be useful SEBs. the extent to which the results are cknicaiy impor- Studies are needed, however, to evaluate the tant should be considered when using banked validity of SEBs. To be suitable for use as an SEB, specimens and determining whether subэects need there should be evidence that the marker is a nec- to be notified. essary step on the pathway to cancer, or at least very highly correlated with cancer occurrence, i.e. Recommendations for future studies of cancer the attributable fraction for the marker in relation epidemiology involving biomarkers to cancer must approach 1. This evidence may be The Workshop considered the main issues regard- obtained primarily from observational epidemio- ing the use of biomarkers in future studies in can- logical studies and clinical trials that incorporate cer epidemiology, and agreed a set of recommen- the marker and have explicit cancer end-points. dations. These are presented in Box 1.

14 Workshop report

2 в • в .

1. Biomarkers should be chosen because of their biologi- study population, a similar population should be used be- cal relevance to the question and on the basis of the sci- cause the ratio of within-person to between-person variance ence, not the technology. may vary substantially between populations due to differ- епсеs in the range of exposure, or fluctuations in exposure. 2. Biomarkers should be evaluated by transitional studies before use in full-scale studies. Regulatory and funding 7.. There should be strong and open communication agencies should become aware of the importance of transi- between laboratory scientists and epidemiologists with tional studies for the conduct of epidemiological studies regard to identification, investigation and control of specific based on biomarkers. sources of measurement error in the biomarker assay. Such issues. would include the need for replicate measurements 3. In addition to etiological research, appropriately validat- and standardized non-variable experimental protocols dur- ed biomarkers should be used in prevention: in screening, . ing the study; internal controls; operator blindness; monitor- as biomarkers of neoplasia or susceptibility, in clinical trials ing expérimental drift; sample handling; and development of (e.g.. Ыоmarkетs of dietary modification and compliance, study design by both epidemiologists and laboratory scien- and biomarkers of surrogate end points) and in monitoring tists to deal with these issues. exposure to carcinogenic agents (e.g. trends in biomarker status fn general populations and subgroups, .exposure sta- B: Epidemiologists .should work closely with laboratory tus and risk assessment iп high-risk populations, and expo investigators when developing protocols for the collection, sure/susceptibility interactions). processing and storage of biological samples to ensure that the samples can be analysed for the main biomarkers of 4. Epidemiological studies based oh biomarkers should interest and that the requisite quality control procedures are satisfy. thé requirements of good epidemiological research implemented. design, including identification and minimization of potential sources of bias and confounding. Research should be con 9. standardized procedures should be developed for the ducted on methodological aspects to clarify advantages and collection, handling, processing and storage of biological disadvantages of different options in the design of biomark- samples. A biospecirimen documentation sheet should be er based studies, in particular 10 determine wehfher, in the generated for each biological sample collected, document- clinic based (or hospital based) setting, studies of biomark ing critical variables that may affect biomarkers ers of exposure or susceptibility are less subject to selection biases than traditional studies of behavioural and enviroп- 10. storage of blood samples in multiple aliquots, and ide- mentâl risk factors of disease. ally in separate freezers, is advisable whenever long-term storage and multiple measurements spread over long time 5. Studies should be designed with appropriate statistical periods are planned The samples should be stored at the power to minimize the problem of random error and maxi- lowest feasible temperature. Assessment of sample stability mize the capacity to establish the presence and strength of would be facilitated by storage of aliquots of a pool prepared interactions, particularly between markers of susceptibility specially for this purpose and exposure. 11 samples should be collected not only for the main bio 6. It is important to define the components of variance in markers of interest in a particular study; rather, they should order to establish the capacity of the biomarker to measure, be processed and stored in a way that allows a wide range as precisely. as possible, the relevant exposure, process and of biomarkers to be tested in the future. In particular; con- outcome. For continuous measurements, it is highly desir sideration should be given to collecting a source of genom able that tha amount of random error due to within person is DNA in epidemiological studies to allow the evaluation of variability in the measurement be estimated in a subset of the genetic markers.: While this is optimally obtained from study population. If this cannot be estimated directly ih the peripheral blood samples, various non invasive methods for

15 Application of Biomarkers in Cancer Epidemiology

ь в ► 1 јйј

collecting genomic DNA are available and continue to be logical interpretations of results obtained from epidemiolog- developed, including the use of buccal swabs or oral rinses ical studies using blomarkers. An iterative process is envis- to collect exfoliated epithelial cells. aged in which epidemiological and laboratory components collaborate in efforts to define pathogenesis. 12. Where feasible, tumour pathology samples should be collected for analysis of tumour characteristics in addition to 15. Care should be taken to consider the possible ethical histological confirmation. This generally requires collecting issues when using biological markers. Biological marker more material than is usually obtained for histological eval- data should be kept confidential and not released uation alone. If tumour blocks are available, collection of in ways that will allow for identification of participants in multiple slides from formalin-fixed, paraffin-embedded research. blocks is desirable. Availability of frozen fresh tumour and normal tissue should be encouraged when relevant and fea- 16. When using banked specimens, the need for informed sible. consent and reporting back results to sub(ects should be based on the degree of clinical risk associated with the bio- 13. In studies of thé .natural history of malignant disease marker. and ils prevention, it is recommended that molecular and protein alterations be studied in normal tissues; benign 17. Investigators should be encouraged to publish, or make lesions, premalignant precursor lésions and the surrounding publicly available data involving biomarkers even if no asse unaffected tissue. This evaluation should also be extended ciations are observed. This applies in parhcular to .transi to non target tissues and to the other organ, in the case of tional studies. paired organs. Assays that can be applied to paraffin embedded formalin=fixed tissues are of particular impor 18. More opportunities for Joint training in both laboratory tance to epidemiological research. and epidemiological methods should be developed.. Encouraging effective collaboration . between. laboratory- 14. Special emphasis should be given to improving'correla based scientists and epidemiologists and the creation of lion ` between animal studies and human in vitro studies interdisciplinary programmes would foster research in this designed to further develop refine and allow for better bio area

16 Workshop report

References Institute for Environment and Health (1996) IEH Report Begg, C.B. & Zhang, Z.F. (1994) Statistical analysis of on the Use of Biomarkers in Environmental Exposure molecular epidemiology studies employing case-series. Assessment, (Report R5), Leicester, Institute for Cancer EpidemioL Biomarkers Prev., 3, 173-175 Environment and Health IARC (1994) 'ARC Monographs on the EvaIuatio o f Bonassi, S., Bolognesi, C., Аыюпdadolo, A., Barale, R., п Bigatti, P., Camurri, L., Dalpra, L., DePerrari, M., Fini, A., Carcinogenic Risks to Humans, Vol. 60, Some Industrial Lando, C., Padovani, P., Pasquini, R., Stella, М. & Puntorni, Chemicals, Lyon, pp. 73-159 R. (1995) Influence of sex on cytogeniic end points: evi- International Collaborative Group (1982) Circulating dence from a large human sample and review of the lit- cholesterol levels and risk of death from cancer in men erature. Cancer Epidemiol. Biomarkers Prey., 4, 671-679 aged 40 to 69 years. J. Am. Med. Assoc., 248, 2853-2859 Bosch, F.X., Manos, М.М., Munoz, N., Sherman, M., Khoury, M J., Adams, M.J., Jr & Flanders, W.D. (1988) An Jansen, A.М., Peto, J., Schiffman, M.H., Moreno, V., epidemiological approach to ecogenetics. Am. J. Hum. Kurman, R., Shah, K.V. & the IBSCC Study Group (1995) Genet., 42, 89-95 Prevalence of human papillomavirus in cervical cancer: a worldwide perspective. J. Nat! Cancer Inst., 87, 796-802 Kuopio, T., Ekfors, T.O., Nikkanen, V. & Nevlainen, T.J. (1995) Acinar cell carcinoma of the pancreas: report of Cole, P. & Maclahon, B. (1969) Oestrogen fractions dur- three cases. APIs, 103, 69-78 ing early reproductive life ш the aetiology of breast cancer. Lancet, i, 604-606 Lemaster, G.К. & Schultе, P.A. (1993) Biologic markers in the epidemiology of reproduction. In: Schulte, ~.A. & Dickinson, L.E., MacMahon, B., Cole, P. & Brown, Pегеra, F.P., eds, Мolecuiar Epidemiology: Princгples and J.B.(1974) Estrogen profiles of Oriental and Caucasian Practice, San Diego, CA, Academic Press. pp. 385-406 women in Hawaii. N. Eng!. J. Med., 291, 1211-1213 Links, J.M., Kensler, T.W. & Groopman, J.D. (1995) Easton, D.F. (1994) Cancer risks in A-T heterozygotes.let. Biomarkers and mechanistic approaches to environmen- J. Rad. Bio!., 66, 5177-182 tal epidemiology. Ann. Rev. Pub. Health, 16, 83-103 Futreal, P.A., Liu, Q., Shattuck-Eidens, D., Cochran, C., MacMahon, B., Cole, P., Brown, J.B., Aoki, K., Lin, T.М., Harshman, K., Tavtigian, S., Bennett, L.М., Haugen- Morgan, R.W. & Woo, NC. (1974) Urine oestrogen pro- Strano, A., Swensen, J., Miki, Y., Eddington, K., McClure, files of Asian and North American women. Iпt. J. Cancer, M., Frye, C., Weaver-Feldhaus, J., Ding, W., Gholarni, Z., 14, 161-167 Soderkvist, P., Teary, L., Jhanwar, s., Berchuck, A., Iglehart, J.A., Marks, J., Ballinger, D.G., Barrett, J. . MacMahon, B., Trichopoulos, D., Brown, J., Andersen, С A.P., Cole, P. deWaard, F., Kauraniemi, T., Skolnick, M.H., Катb, A. & Wiseman, R. (1994) BRCA1 mutations in primary breast and ovarian carcinomas. Polychronopoulou, A., Ravnihar, B., stormby, N. & Science, 266, 120-122 Westlund, K. (1982) Age at menarche, urine estrogens and breast cancer risk. let. J. Cancer, 30, 427-431 Green, R. & Jacobsen, D.W. (1995) Clinical implications of hyperhomocysteinemia. 1n: Bailey, L.B., ed., Folate in MacMahon, B., Cole, P., Brown, J.B., Paffenbarger, R., Health and Disease. New York, Marcel Dekker Inc. pp. Trichopoulos, D. & Yen, Ѕ. (1983) Urine estrogens, fre- 75-122 quency of ovulation and breast cancer risk: case-control study in premenopausal women. J. Nail Cancer lest, 70, Hagmar, L., Bragger, A., Hantseen, I.L., Heim, S., 247-250 Нёgstеdt, B., Knudsen, L., Lambert, B., Linnainmaa, K., Mitelman, F., Nirdenson, I., Reuterwall, C., Saliras, S., McMichael, А.J., Jensen, Q.М., Parkin, D.M. & Zaridze, D.G. (1984) Dietary aid endogenous cholesterol and Skerfving, S. & Sorsа, M. (1994) Cancer risk in humans predicted by increased levels of chromosomal aberra- human cancer. Epi. Reviews, 6, 192-216 tions in lymphocytes: Nordic Ѕtцду Group on the Health Mendelsohn, M.L., Peeters, J.P. & Normandy, M.J., eds. Risk of Chromosome Damage. Cancer Res., 54, (1995) Biomarkers and Occupational Health: Progress and 2919-2922 Perspectives, Washington, DC, Joseph Henry Press Hulka, B.S., Wilcosky, T.C. & Griffith, J.A., eds (1990) Miki, Y., Swenseri, J., Shattuck-Eidеns, D., Futreal, P.A., Biological Markers in Epidemiology, New York, Oxford Harshman, K., Tavtigian, S., Liu, Q., Cochran, C., University Press Bennett, L.М., Ding, W., Bell, R. Rosenthal, J., Hussey, C., Huika, B.S., Liu, E.T. & Lininger, R.A. (1994) Steroid hor- Tran, T., McClure, M., Frye, C., Hattier, T., Phelps, R., mones and risk of breast cancer. Cancer, 74 (Suppl. 3), Haugen-Strano, A., Katcher, H., Yakumo, K., Gholami, Z., 1111-1124 Shaffer, D., Stone, S., Bayer, S., Wray, C., Bogden, R., Dayananth, P, Ward, J., Tenin, P., Narod, S., Bristow, P.K.,

17 Application of Btomarkers in Cancer Epidemiology

Norris, F.H., Helvering, L., Morrison, P., Rosteck, Р. Lai, M., Barrett, JC ., Lewis, C.; Neuhausen, S., Cannon- Аl гighi, L., Golcigar, D., Wiseman, R., Kamb, A. & Skolnick, M.H. (1994) A strong candidate for the breast and ovarian cancer susceptibility gene BRCA1. Sсгепсе, 266, 66-71 MuEoz, N., Bosch, F.X., de 5anjosé, 5., Tafur, L., Izarzugaza, I., Gili, M., Viladru, P., Navarro, C., Martos, C., Ascunce, N., Gonzalez, L.С., Kaldor, J.M., Guerrero, E., Lбrincz, A., Santamaria, M., Alonso de Ruiz, P, Aristizabal, N. & Shah, K. (1992) The causal link between human papillomavirus and invasive cervical cancer: a population-based case-control study in Colombia and Spain. Jot. J. Cancer, 52, 743-749 Nakazawa, H., English, D., Randell, EL,, Nakazawa, K., Martel, N., Armstrong, R.K. & Yamasaki, H. (1994) UV and skin сапсer: specific p53 gene mutation in normal skin as a biologically relevant exposure measurement. Proc. NatI Aсад. Sel., 91, 390--364 Qian, G.S., Ross, R.K., Yu, M.C., Yuan, J.М., Gao, Y.T., Henderson, B.E., Wogan, G.N. & Groopman, J.D. (1994) A follow-up study of urinary markers of aflatoxin exposure and liver cancer risk in Shanghai, People's Republic of China. Cancer Epidemiol. Biomarkers Prey., 3, 3-10 Ross, R.K., Yuan, J.М., Yu, M.C., Wogan, G.N., Qian, G.5., Tu, J.T., Groopman, J.D., Gao, Y.T. & Henderson, WE, (1992) Urinary aflatoxiis biomarkers and risk of hepatocellular carcinoma. Lancet, 339, 943-946

Sсhаtzkiп, А., Freedman, L.S., Schiffman, M.H., & Dawsey, S.М. (1990) Validation of intermediate end points in cancer research. J. Nat' Cancer Inst., 82, 1746-1752 Schulte, P.A., & Perera, F.P., eds (1993) Molecular Epidemiology; Principles & Practices, San Diego, Academic Press Sherwin, R.W., Wentworth, D.N., Cutler, J.A., Hulley, S.B., Koller, L. & Stamler, J. (1987) Serum cholesterol levels and cancer mortality among 361,662 men screened for the multiple intervention trial. J. Am. Med. Assoc., 257, 943-947

18 Applicансгг of Biomarkers in Cancer Epidemiology roniolo, P., Bоffеtlа, P., Shukeг, D.E.G., Rathman, N., Hujka, B. arid Pearce, N., ads lARC Scientific PLiblicatiaus No. 142 ИlегпаtЁопаl Agency fer Research en Cancer, Lyon, 1997

Transitional studies

P.А. Schulte and,F.P' Perera

Transitional studies are studies using biological markers that bridge the gap between laboratory experiments and population-based epidemlology.The goal of these studies is to characterize and validate biomarkers and to assess the following: intra- and Inter-subject variability; the feasibility of marker use in field conditions; confounding and effect-modifying factors for the marker; and mechanisms reflected 'by the biomarker. Another goal is to optimize the conditions for the use of biomarkers' Transitional studies involving biomarkers of exposure or effect are distinguished from etiological studies because the biomarker is generally the outcome or dependent variable. Despite this difference, transitional studies can be epidemiological studies, but they may also include laboratory studies to assess reliability (and accuracy) and to identify parameters for collecting, рro- cеssing and storing biological specimens prior to assay. Generally, transitional studies involve healthy people, patients or workers with specific exposures. At some point in the validation of a biomarker the line between transitional and etiological studies becomes blurred. None the less, it is useful to identify transitional studies as a distinct set of efforts to validate and characterize biomarkers. Transitional studies can be divided into three functional categories: developmental, characterization and applied studies.

Transitional studies have been described as studies Transitional studies can be divided into three func- using biological markers that bridge the gap between tional categories (Table 1): developmental, charac- laboratory experiments and population-based epi- terization and applied studies (Schulte et aL, 1993; deniiology (Ilulka, 1991; Schulte, 1992; Schulte et Rothman et al., 1995). There is sri clear deniarca- Al., 1993; Rothman et al., 1995). The goal of these ion between the types of transitional study, but studies is to characterize and validate biomarkers the categories are useful in identifying the cluster- and to assess the following: intra- and inter-sub- ing of preparatory research that is needed before a ject variability; the feasibility of marker use in field biomarker is ready for population research. All conditions; confounding and effect-modifying three types of transitional study are efforts to deter- factors for the marker; and mechanisms reflected mine aspects of the validity of a biomarker. by the biomarker. Another goal is to optimize the Transitional studies use epidemiological methods conditions for use of biomarkers. Transitional stud- in the development, testing and validation of bio- ies involving biomarkers of exposure or effect are markers. These studies represent preparatory distinguished from etiological studies because the efforts to determine the parameters, limitations or biomarker is generally the outcome or dependent characteristics of a biomarker prior to its use in еti 'variable. Despite this difference, transitional studies ological, prevention or other kinds of intervention can be epidemiological studies, but they may also efforts. Transitional studies build on scientific include laboratory studies to assess reliability (and knowledge from various types of laboratory studies, accuracy, if possible) and to identify parameters for including tissue culture and animal studies. The collecting, processing and storing biological speci- conduct of transitional studies is not a one-time mens prior to assay. Generally, transitional studies step in the development of biomarkers but can involve healthy people, patients or workers with represent steps in an iterative process. Thus a bio- specific exposures. At some point in the validation marker may be developed in the laboratory, vali- of a biomarker the line between transitional and dated in a transitional study using epidemiological etiological studies becomes blurred. None the less, it methods, and then applied in an etiological study is useful to identify transitional studies as a distinct that raises new questions that stimulate new labo- set of efforts to validate and characterize biornarkers. ratory and transitional studies.

19 Application of Biomarkers in Cancer Epidemiology

determine whether a biomarker measures what it claims to measure, be it exposure, disease or susceptibility. A more epidemiological definition is Туре Description that a biomarker is valid to the extent that it measures tue Developmental Builds on laboratory and in vitro the marker populations, that is, with no measure- animal studies and involves the ment error (see White, this volume). test of assay in humans As shown in Table 2, validation can be considered to be a two-stage process, although many of Characterization Examines the variability of the steps require iteration as new information biomarkers in population subgroups and factors that could becomes available. The ultimate goal of the valida- confound associations between tion effort is to allow biomarkers to be used in markers and what they represent appropriate epidemiological and clinical applications, which will require a selection of the best bio- Applied . Assesses Ike relationship between a marker and an markers based on the criteria described below. underlying event such as an The first step is laboratory validation, in which exposure, a biological effect or the shape of the dose—response curve, low-dose susceptibility sensitivity (the ability to detect the exposure at levels low enough to be of biological interest), exposure specificity and reproducibility of the assay are tested. In the second step, known as epidemiological vali- Overview of the validation process dation, the population sensitivity and specificity, Validation of Ыomarkers involves the clarification intra-individual aid interindi'Tidual variation in of factors that influence the ability of the marker to response and persistence, predictive value, and bio- predict exposure or outcome in a test population. logical relevance and feasibility are evaluated. The issues covered here are discussed by Schulte & In this validation process, pilot studies in high- Pererа (1993) and Perera & Mooney (1993). During dose groups are useful. Molecular epidemiological the validation process, evidence is weighed to studies to monitor the carcinogenic potential of

Table 2. Criteria for molecular epidemiological validation of biomarkers

Dose—response curve Population sensitivity and specificity Detection limit or low-dose sensitivity Intra-individual variation over time: Exposure specificity . a. without altering exposure Reliability of the assay: b. when exposure is removed (persistence/half-life) or changed a. from run to run Inter-individual variation: b. from day to day a. response to a given exposure c. from one laboratory to another Ь. persistence of biomarker . Optimal conditions for sample collection, Half-life in surrogate tissues processing and storage Positive predictive value (yield of high-risk people) Feasibility: a. amount and availability of tissue b. cost c. time required for each assay Biological relevance to disease

Source: Регега R Mooney, 1993

20 Transitional studies industrial chemicals in workers have been a useful Another hindrance to good interdisciplinary approach to validation because occupational expo- collaboration is the tension between assay consis- sures tend to be higher and more controlled than tency and assay improvement (Rothman, 1993; ambient exposures (Perera & Weinstein, 1982; Wilcox, 1995). The relatively long term for epi- Pereyc & Sante1la, 1993). Initial research linked a demiological studies may span a period when number of occupational exposures to increases in improvements are made in the assay. The labora- adducts, mutation and other biomarkers in workers. tory collaborators will naturally want to take ad- State-of-the-art studies have broadened to include vantage of the improved assays and incorporate assessment of smoking and dietary factors (e.g. vit- them in the study without consulting the epi- amins, fat oxidative damage), inherent factors demiologist. This can be disastrous from the epi- such as genetic predisposition to biotransforma- demiological point of view. The use of two different tion and repair capacity, and endogenous levels of assay methods can ruin an epidemiological study. enzymes and promoters. As new and more reliable assays are developed, scientists will be persuaded to replace previous Interdisciplinary collaboration essays. When a marker assay is new, measurements Effective collaboration between laboratory scien- may differ from those in previous assays for the tists and epidemiologists is critical in conducting same marker. Although it may be reasonable to transitional studies. These studies involve the first wait until laboratory techniques and estimates of field testing of the markers and they require both variability display consistency, it is not reasonable laboratory and epidemiological expertise. The or feasible to wait until a technique is so standard interaction of different disciplines requires that that no refinement is likely to occur. However, attention be paid to the underlying assumptions, before modifying a technique during the course of paradigms and language of the various disciplines. a study, the old versus new results must be care- Epidemiologists generally speak in terms of groups fully evaluated for overall comparability. and risk to groups. Laboratory scientists tend to focus on individuals or components of an individ- Developmental studies ual. Epidemiology is an observational science Reliability whereas laboratory disciplines use controlled When a candidate biomarker is identified in the experimental designs. Epidemiologists and labora- laboratory, some very basic issues need to be solved tory researchers use the same word to mean different before it can be considered for use in population things. For example, when laboratory researchers studies. The first priority in evaluating a marker for speak of a valid marker, they are referring to the use in population studies is to determine its relia- characteristics of the assay for the marker, whereas bility or reproducibility. As long as the assay is reli- when epidemiologists speak of a valid marker they able, the ordering of subjects by the measure is are generally referring to one with а high predic- preserved (Rothman et al., 1995). Hence, there will tive value or correlation with exposure or disease. be consistency within a study. Since this 1s al that The scale of transitional studies may be a hin- is required for studying a marker—disease relation- drance to effective collaboration. Laboratory re- ship, reliability aid not accuracy is of primary search generally takes place on a small scale, importance (Rothman et aL, 1995). Reliability can whereas epidemiological studies can produce vast be assessed optimally through analyses of blind quantities of specimens. Even in a study that is replicate human samples that are representative of small by epidemiological standards, e.g. of 30 peo- a range of values likely to be found in human pop- ple, the volume of specimens is often more than ulations. the research laboratory is used to processing. The Reliability encompasses both unsystematic ran- laboratory workers can easily feel that they are not dom laboratory variation observed in repeated engaged in true scientific research but are merely measurements and bias caused by non-random providing a service. Meanwhile, the epidemiolo- variation (Vineis etaL, 1993). To assess random ezror, gist wonders why the laboratory is not able to han- multiple measurements are needed. Obviously, the dle the volume of specimens required for the study. random error in the arithmetic mean of several (Wilcox, 1995). measurements is smaller than the random error of

21 Application of Biomarkors in Cancer Epidemiology

an individual measurement, Quantitative indices bility; the high rates of misclassification in studies of the extent of random variation of a biological of the association between exposure and disease; marker can be used to determine whether the and the consequent underestimation of the associ- reliability of a given measure is sufficient for the ation between a health measure and the measured purpose being considered. The two most common extent of exposure to an environmental risk factor. indices are the standard error of the measurement All these factors pertain to studies using biological and the reliability coefficient (Massey, 1986). markers of exposure or effect. Fleiss (1986) recom- Analysis of random laboratory variation involves mends that unreliability be controlled by conducting several steps. First, multiple analytical measure- pilot studies and replicating measurement proce- ments of the same biological specimen (or several dures on each study subject. biological specimens from the same individual at Another aspect of developmental transitional the same point in time) must be made to estimate studies is to define the optimal conditions for col- the variability due to random analytical errors. lecting, processing and storing biological speci- This fraction of the total random variability of a mens. The issues mentioned here are discussed by biological marker is usually minor. Second, multiple Winn & Gunter (1993) and in other chapters in measurements of a marker must be made for one this volume. The major lessons are that, at all individual over time to estimate the intra-individual stages of specimen handling, variation and/or temporal variability. Third, multiple measure- error can be introduced, and careful attention is ments across different individuals must be made required to prevent these untoward aspects. In to estimate interindividual variability in the value addition to the reliability of biomarkers, it is of the marker. In the second and third cases, ran- important to understand the biokinetics and sta- dom error is a component of the variability but bility aspects before application in population systematic errors may also contribute (e.g. circa- studies. These issues have been addressed in previous dian cycles in the value of a marker or differences publications (Droz, 1993; Bernard, 1995). As dis- among individuals due to genotype). Most mol- cussed, valid biomarkers, particularly of exposure, ecular epidemiological research using biological will be those that have biological relevance, markers seldom requires large numbers of individ- defined pharmacokinetics and temporal relevance. ual measurements. Thus, a small number of indi- For markers of exposure, the parameters that best viduals can be used as a sample of the infinitely summarize the behaviour of a chemical in biolog- larger population to which the distribution refers. ical systems is the elimination half-life, which The standard error indicates how the mean of that reflects both the affinity of the chemical for the sample is distributed around the mean of the larger biological matrix and the efficiency of excretory or population. Hence, the standard error of the mean metabolic processes of elimination. Bernard (1995) reflects the reliability of the sample mean as an has suggested four categories for biomarkers of indicator of the population mean (Massey, 1986). exposure: half-life less than 12 hours; half-life This value may not be as informative as the rеliа- between 12 and 100 hours; half-life between 100 bшty coefficient for evaluating markers to be used hours and 6 months; and half-life greater than 6 in epidemiological studies. months. Generally, for etiological epidemiological The reliability coefficient is technically known studies, biomarkers in the latter two categories will as the intraclass coefficient of reliability (Shrout & be most useful. Fleiss, 1979; Fleiss, 1986) and ranges from Ito 1. If each measurement is identical, the intraclass co- Biological relevance efficient is 1.0. The greater the variation among The biological relevance of a biomarker is of prime measurements, the lower is the reliability. Fleiss importance in the selection and validation of a (1986) has evaluated the impact of unsystematic marker. The validation of a specific biomarker bias variation in measurement, described the prob- must include consideration of its biological rele- lematic consequences of unreliability, and recom- vance to the disease or exposure under study and mended how unreliability can be controlled. The the position in the continuum between exposure and consequences described by Fleiss (1986) include disease. Given the fact that many exposures cause the need to increase sample size to reduce unrelia- multiple diseases, the validation of a biornarker's

22 Transitional studies relationship to а disease first requires a clear defin- studies is a matter of judgment. If the objective is ition of the disease end-pdint. A well developed to determine simply whether an assay `works', it is hypothesis of the relationship between the bio- appropriate to test the assay on a small number of marker and other events in the exposure—disease available subi ects without regard for sample size or continuum is necessary. representativeness. However, in transitional stud- While knowledge of the stability and natural ies to determine the validity of a marker and its history are important before population applica- variability, it is important that subjects are selected tion of bioorarkers of effect, it is important to through defensible sampling designs. Critical in assess the attributable risk or proportion (Schatzkin this regard will be the need to minimize selection et al., 1990; Benichou, 1991; Trick, 1995). The bias and to generate adequate statistical power when attributable proportion associated with a particular null hypotheses of no difference among groups or biomarker is an estimate of the proportion of can- perhaps no association among biomarkers are to cer cases that must progress through the bio- be tested. marker. This is not simply the proportion of all For biomarkers of exposure, it has been shown cases that are positive for the biomarker, because to be useful to select subjects from groups known the biomarker will occur in some cases ('back- to have high exposures (such as chemotherapy ground cases') even when the exposure of interest patients or workers in specific occupational expo- or the biological events associated with the bio- sure groups) to maximize exposure differences marker are not etiological events for those cases between groups and to obtain an indication of (Trick, 1995). This assessment will help to identify whether an assay will identify a potentially large the possible mechanism by which the biomarker is `signal', and ultimately a dose—response relation- related to the causal pathway for the cancer and to ship. For biomarkers of effect, studies involving determine the extent to which the biomarker truly histologically defined subsets of patients are useful represents a biological event intermediate between in determining the association of biomarkers and exposure and cancer. a particular cancer. The calculation of sample size Shatzkiп et cd. (1990) have developed a frame- for transitional studies is the same as for etiologi- work for considering exposure—marker—disease cal studies. For a situation in which two groups relationships: (e.g. cases and controls) are compared on the basis of presence or absence of some biomarkers, the for- •A single marker is known to be linked causally mulation for the calculation of sample size for a and, hence, is necessary and sufficient for disease. given combination of significance level, power and 'Multiple pathways and multiple biomarkers size of expected difference as measured in terms of are identified and any marker is sufficient but relative risk is well known (Schesshnan, 1982; not necessary for disease. Hertzberg & Russek-Cohen, 1993). 'An exposure leading to disease operates When groups of subjects are being compared for through an unobservable event which in turn shifts in central tendency (such as mean values), lends to an intermediate marker that is not the sample size is calculated according to a well directly biologically related to the disease but known formula described in elementary textbooks which may correlate with its occurrence. that gives the sample size as a function of alpha, power, the size difference to be detected and the It is the role of developmental and applied tran- variance of marker levels within groups. When the sitional studies to refine the view of these various investigator is interested in inferring the relative mechanisms and to attempt to confirm which risk of a disease as a function of incremental mechanism describes the type of marker being change in the level of a given biomarker, the sample considered. It is important to understand underly- size is determined by a formulation involving type ing mechanisms in order to distinguish measure- 1 and II error rates, mean levels in cases and controls, ment error problems from multiple pathways. and variance in cases and controls, respectively (Hertzberg & Russeck-Coheп, 1993). Determining the number of subjects The determination of sample size reflects a ten- The number of subjects needed in transitional sion between feasibility and cost issues on the one

23 Application of Biomarkers in Cancer Epidemiology hand and confounding on the other. What are the if the responses of exposed and unexposed people best design and analysis strategies to address these overlap. Interindividual variation in response is problems? For biomarkers of exposure it is essential thought to result from differences in a variety of to have an unexposed comparison group от sub- factors, including personal exposure, ability to groups with a range of exposures. Matching metabolize carcinogens, and differences in DNA exposed and unexposed subjects on potential con- repair, immune surveillance and nutritional status. founding factors can increase the precision of the Therefore, people with the same apparent expo- measure of association between exposures and out- sure may vary widely in their response to carcino- come, but this may decrease the study's power gens. For example, interindividual variation in the while controlling for confounding on the match- level of xenobiotics adducted to DNA iesults from ing variables (lulka, 1990). Alternatively, a study processes that vary between people, affecting the design could involve restriction of suspected con- concentration at the target tissue after distribution founders, i.e. limiting the subjects eligible to those in vivo, metabolic activation and detoxification, with or without a particular potentially confound- capacity of repair, and persistence of the xenobi- ing factor. This may minimize confounding but otic and cellular target in vivo. will not allow for analysis of effect modification by A goal of molecular epidemiology is to elucidate the restriction variables. the mechanisms that explain why people vary in In pтаctice, the number of subjects will depend their risk of disease. To do this, we must under- on both the statistical power and the cost of stand which part of the biological or biomarker obtaining subjects, specimens and assays. Unlike measurement is due to laboratory variability or to more traditional epidemiological studies, in which intra-individual variation and which part reflects unexposed persons may be more accessible and true interindividual variation. The mechanisms cheaper to recruit than exposed persons, the орро- underlying interindividual differences in bio- site situation may occur in transitional studies marker response are important because they are where there is little incentive for non-exposed or the same mechanisms that the body uses to miti- non-diagnosed persons to provide biological sam- gate a disease outcome. This means that biomark- ples (Hulka, 1990). ers have potential as targets of interventions (Реrеrа & Mooney, 1993). Characterization studies Within-person biological variability may or Once a marker has been sufficiently developed in may not be time-dependent (Hulka & Margolin, terms of the reliability of the assay and optimal 1992). Biological specimens collected from the conditions for handling, it is necessary to assess its same person at different times might show characteristics in human populations (Schulte et changes because the person has aged, has been 01., 1993; Rothman et al., 1995). The objective of exposed to mutagenic substances or has been these studies is to identify factors that are con- exposed to a medical procedure. The actual bio- founders or effect modifiers which should be taken logical matrix, e.g. white blood cells, will have into account in etiological or public health studies. changed in the intervening time. Some cells will One type of characterization study iпvolves the have died, while others may have been generated. assessment of the frequency of a marker in various Time-independent biological variability will population subgroups, determined by characteris- also occur because the distribution of matter (e.g. tics such as age, race, sex, medical condition, xenbiotics) can vary across organs and cells. Each behaviour, etc. These types of study should be con- biological sample will not capture the same bio- ducted on a marker-by-marker basis depending on logical material (Hulka & Margolin, 1992). the underlying biology, mechanism and relation Intraperson variability has important iтnplica- to various host factors. tions for sample size and power in transitional The ultimate goal of characterization-type tran- studies (Hulka & Margolin, 1992). This has been sitional studies is to assess interindividual variation demonstrated in cytogenetic studies, such as that and the genetic and acquired factors that influence by Hirsch et aI. (1984) who found that cell-to-cell the variation. Large interindividual variation will variability in the frequency of sister chromatid make it more difficult to predict the risk of disease exchanges from an individual person was greater

24 Transitional sk dies than the variability in the mean frequency of sister tional studies are generally cross-sectional or short- chromatid exchanges from person to person. To term longitudinal designs and are not capable, in compensate for this intra-individual variability, and of themselves, of establishing or refuting a the number of cells scored per person was in- causal relationship between a given exposure and creased. interperson variability in biomarker mеа- disease. They do, however provide mechanistic sures is generally addressed in the same way as insight and may yield useful information on rela- other types of variable measures in epidemiological tionships between biomarkers arid the events they studies. The exception may be that the variability represent (Rothman et ai., 1995). Where case—con- of biomarkers between persons may reflect trol studies that link a biomarker with disease are acquired or inherited susceptibility which can conducted, inferences about causality can some- serve as an effect modifier in relation to exogenous times be made. Another type of applied transi- exposures of interest (Hulka & Margolin, 1992). tional study that has been reported involves the susceptibility to environmental agents is likely to use of biomarkers in intervention studies. These arise from complex interaction between activating include interventions with smokers, workers and and inactivating chemical metabolizing enzymes. the treatment of high-risk populations (e.g. in For example, Tang et al. (1995) have estimated that treatment with antioxidants, biomarkers are being the combination of an inherited deletion of a gene used to monitor the efficacy of the intervention) for the detoxifying enzyme glutathioпe-S-trans- (Регегa & Mooney, 1993). The studies are designed to test not only the presence of the biomarkers, but fегasе (GST11) and high PAH-DNA adduct levels confers a 12-fold risk of lung cancer. also the level of change with alteration of the risk The ability to measure biomarkers at the mole- factor. These studies are transitional in that they cular and genetic levels has resulted in the identi- establish the ability of the marker to serve as an fication of a degree of interperson variability not early or intermediate `end-point' with which to previously imagined. The major sources of this monitor efficacy or compliance. variability need to be accounted for prior to the use The objective of the studies is to determine one of biomarkers in etiological or public health appli- of the following relationships: cations. Characterization-type transitional studies may involve determining the prevalence of partic- • exposure/marker ular alleles in specific racial subgroups, evaluating •marker/disease the correlation between genotypic and phenotypic • exposure susceptibility marker (high disease assays, estimating the likelihood and impact of risk or low disease risk). allele misclassification and evaluating potential induction effects. Additionally, transitional studies Applied transitional studies can be used to can evaluate the biological plausibility of gene-- assess the attributable proportion of a particular environment interactions observed in etiological biomarker. Hence, they build on the mechanistic studies (Rothman et al., 1995). knowledge obtained in characterization-type transitional studies. Applied transitional studies Applied transitional studies are those that assess Exposure/marker the relationship between a marker and the event The assessment of the relationship between an expo- that it marks, namely exposure, disease or suscep- sure and a marker represents much of the previous tibility. These studies are often conducted on effort termed molecular epidemiology. Exposures to healthy subjects, and generally, in most cases in- carcinogens have been evaluated to determine volving biomarkers of exposure or effect, the bio- changes in DNA (DNA adducts) or proteins (e.g. marker is treated as the outcome variable. However, haemoglobin or albumin adducts), or in some with susceptibility biomarkers andin certain study cases cytogenetic changes (e.g. chromosornal aber- designs, a biornarker of exposure or effect may be rations, micronuclei). These have generally been used as an independent variable (e.g. serum cross-sectional in nature. It is critical in such stud- organochlorines and breast cancer; carcinogen ies to be rigorous not only in the assessment of the DNA adducts and lung cancer). Applied transi- marker but also in the assessment of exposure and

25 Application ы Biomarkers in Cancer Epidemiology

control for confounding. This is illustrated in the atively long time. The best and possibly only work of Mayer et aL (1991) and Schulte et aL (1992), example of such a study is the Nordic prospective who demonstrated the value of putting appropriate study on the relationship between peripheral lym- resources into the exposure characterizations when phocyte chromosome damage and cancer morbid- validating the relationship between exposure to ity in occupational groups (Bragger et aI., 1990). ethylene oxide and formation of haemoglobin Ten laboratories in four Nordic countries partici- adducts. Too often, emphasis is placed only Of the pated in a study of a combined cohort of persons assay and not on the measure of exposure. (mostly from occupational groups) who had been In characterizing the relationship between a cytogenetically tested. The cohort will be followed marker and exposure, it is important that all prospectively for cancer morbidity. The cohort sources and rates of exposure are considered, since comprises 3190 subjects, of whom 1986 (62%) a marker generally represents the integration of have been scored for chromosome aberrations these. Attention should also be given to the toxi and 2024 (63%) have been scored for sister chro- cokinetics and natural history of the marker in matid exchanges. Preliminaтy analysis indicates order to understand the appropriate sampling and that chromosomal aberrations are associated with specimen collection times. A marker with a short cancer. half-life will not be detectable in samples a long time after exposure has ended. Moreover there is Exposure/susceptiЬнl tу to disease a need to assess the relationship between the Validated biological markers of susceptibility can marker and different regimens of exposure (con- serve as effect modifiers in epidemiological studies. tinuous or intermittent). Effect modification is a term with statistical and biological aspects. Statistically, effect modification Marker/disease is analysed by examining the joint effects of two or The relationship between a marker and disease is more factors. The interpretation of effect modifi- often the most frequently considered issue in cation depends on the statistical method (e.g. mul- assessing whether a marker is ready for use in an tiplicative or additive) used to model interaction. epidemiological study. Of interest is how well the From the biological perspective, effect modifica- marker predicts or represents disease. These types tion can explain why two similarly exposed indi- of validation studies are difficult to accomplish viduals do not develop a disease. The answer, in because of the temporal factor. To identify an early dart, is individual variability in metabolic detoxi- change—i.e. a change in pathogenesis or a change fication and repair capabilities. predictive of disease—generally requires a prospec- To validate a susceptibility marker, it is impor- tive study, although cross-sectional clinical studies tant to minimize misclassification, which can occur of heavily exposed individuals and case-control as a result of laboratory or epidemiological factors studies can be used to great advantage. However, that affect phenotyping or genotyping (Rothman when not using a prospective design, care must be et al., 1993). Next, ills necessary to demonstrate taken to avoid biased associations. This is often dif- that the susceptibility marker either increases the ficult, and hence prospective studies are the best biologically effective dose or elevates the risk of approach for validation. However, prospective disease. studies are expensive and time-consuming, and Genetic susceptibility markers can be character- few are conducted. For example, despite the large ized in terms of the six possible patterns of number of studies on cytogenetic markers, there is gene-environment interaction identified by still little consensus on their predictive value, since Khoury et aL (1988). These are shown in Table 3. most of the studies have been cross-sectional and Khoury et al. (1988) recommend that an epidemi- suffer from temporal ambiguity. Specifically, in epi- ological approach be used to evaluate genetic demiological terry s, predictive value means the marker-disease associations and their interaction percentage of those who test positive for a marker with specific environmental risk factors. However, who actually develop the disease. To perform the prior to conducting such studies, a major challenge appropriate prospective studies of sister chromatid is to determine which environmental factors might exchanges would take a large population and a rel- be involved in the etiology of a specific disease.

26 Transitional studies

EIlect of genotype. in the absence Specifieity of environmental Pattern of the environment effect vis-à-vis genotype. Notations 1 ' Innocuous Specific R9 -1, le =1 Innocuous Non-specific R9.= 1 > 1 2 1е Specific R9 > 1, Re = 1. З Risk factor 4 . Risk factor Non-specific Rg > 1 „Re > 1 5 Protective 5pecifiç Rg < 1', Re =1: Rg <'1, Re 1 б Proteéfive Non-spécifie з

Source: Khoury e.t aI., 1988.

In cancer epidemiology, interest has been ration. This can be very useful but should be ori- focused on susceptibility genes that are common ented towards confirming a hypothesized link in a in the population and are generally considered to continuum. be polymorphisms (i.e. with a minor alele fre- It is of great importance in identifying a con- quency of more than 1%), that are probably asso- tinuum of events between exposures and disease ciated with relative risks under 10 (and as such do that the marker represents a 'critical effect'. not exhibit familial patterns of inheritance) and (Schulte, 1989; Borm, 1994). A critical effect is the that may interact with a particular exposure biological marker that is deemed most representa- (Rothman et a2., 1995). The objective of applied tive of a particular component in the continuum transitional studies with regard to susceptibility and is ultimately most pathognomonic. This markers is to determine whether the markers are requires a series of independent studies, primarily effect modifiers. As mentioned earlier, the distinc- toxicological, but also clinical and epidemiologi- tion between a transitional study aid an etiological. It is necessary to develop a hypothesis con- cal study becomes slightly bluned in this effort, so cerning the role of the marker in the development that such distinctions are arbitrary. The emphasis of the disease. As more causal components are in a transitional study is to determine whether the identified it becomes necessary to elucidate quan- marker can be used to investigate populations in titative relationships of the kinetics, natural his- terms of risk (i.e. serve as an outcome variable or as tory and rates of transition along the continuum. an independent variable). Determining level of effort Preparing for transitional studies After a candidate marker has been identified, it is $electing candidate markers useful to determine what field testing is required. Many more markers are identified in laboratories This is a useful exercise because it allows for than could reasonably be tested and used in the research planning and funding and assures that a field. For this reason, there is a need to select comprehensive approach is considered. The alter- among candidates which markers are beneficial native is that a marker becomes labelled as vali- for field testing and use. To gauge utility, it is help- dated and ready for field use when ills not ready ful to envision a framework for candidate markers or eligible, and this can lead to flawed and costly such as the continuum between exposure and studies. An example of a check list for what needs disease. A potentially useful candidate marker will to be considered is shown in Table 2. be one that can be related to some heuristic continuum and for which successful field testing will Selecting candidate populations add relevant information to various etiological or A key factor in transitional studies is selecting and public health questions. Iп some cases a transi- accessing populations with the kind of characteris- tional study may only provide mechanistic infor- tics thought to be important to the testing of a

27 Application of Biomarkers in Cancer Epidemiology

marker. These may be populations with contrasting their limits, they are intermediate rather than end levels of exposure, demographic or behavioural results about cancei causation or controls; yet factors, or ethnicity; or they may be populations without this effort, the end results may not be homogeneous for such factors. The similarity obtained. Currently, much of the work of transi- between these populations and those in whom the tional studies is subsumed in pilot or feasibility marker will ultimately be used for etiological or studies conducted prior to a larger study. This may applied purposes is important. not be the most effective way of conducting transitional research because researchers may be forced Reporting results of transitional studies to trade off those funds available for conducting an Transitional studies can be highly informative etiological study against those available for the regardless of their outcome. Those that show a pos- assessment of the utility and limits of a marker. itive relationship between a marker and exposure, More funding agencies should designate and sup- disease or susceptibility are obviously important. port transitional biomarker studies if a wide range However, those that show negative relationships of useful tools are to become available. between a biomarker and a particular event, or have small biomarker frequencies in a population References subgroup, provide useful information about the Benichou, J. (1991) Methods of adjustment for estimat- utility, generalizability or limitations of а marker. ing the attributable risk in case-control studies: a review. Such negative results in well conducted studies StaG Med., 10, 1753-1773 should be published in the peer-reviewed scientific Bernard, AM, (1995) Biokinetics and stability aspects of literature. Clearly, statistical power considerations biomarkers: recommendations For application in popu- should be discussed in such studies. lation studies. Toxicology, 101, 65-71 There appears to be a wide variation in the Borm Р,].A. (1994) Biological markers and occupational approaches used in reporting test and study results lung disease: mineral dust-induced respiratory disorders. to participants in the research. Some investigators Ехp. Lung Res., 20, 457-470 and their supporting organizations require all test Brogger, A. Hagmar, L., Hansteen, I.L., Heim, S., and subject results to be communicated to partici- Hogstedt, B., Knudsen, L., Lambert, Б., Linnainmaa, K., pants regardless of clinical relevance and with the Mitelman, F., Nordenson, I., Reuterwall, C., Salomao, S., most truthful interpretation possible, while others Skerfving, Ѕ. & Sorsa, M. (1990) An inter-Nordic prospec- limit reporting to those results that are clinically tive study on cytogenetic endpoints and cancer risk. relevant. This difference hinges on the tension Cancer Gene Cyfogenet., 45, 85-92 between beneficence and autonomy (see Schulte et Droz, P.O. (1993) Biological monitoring and pharmaco- al., this volume, for a discussion of the ethical kinetic modeling for the assessment of exposure. In: issues). Should participants be told only when Schulte, P.A. & Perera, RE, eds, Molecular Epidemiology; something can be done and not told when there Principies and Practices, San Diego, Academic Press, pp, might be anxiety without benefit, or should par- 13 7-15 7 ticipants have a right to information that is held Fleiss, J.L. (1986) Statistical factors in early detection of about them regardless of whether it is considered health effects. In: Underhill, D.M. & Radford, E.D., eds, by the holder to be of benefit? These questions New and Sensitive Indгcators of Health Impacts of need further consideration. Environmental Agents, Pittsburgh, PA, University of Pittsburgh Press, pp. 9-16

Support for transitional research Hertzberg, V.S. & Russek-Cohen, Е. (1993) 5tatisncа1 Transitional studies are important for the success- methods in molecular epidemiology. In: Schulte, P.A. & ful and effective use of biomarkers in cancer epi- Peiera, F.P. eds, Molecular Epidemiology: Principles and demiology. However, they are perceived as having Practices, San Diego, Academic Press, pp. 199-216 neither the excitement nor the appeal of basic lab- Hirsch, B., McCue, M. & Cervenka, J. (1984) oratory or etiological research or public health Characterization of the distribution of sister chromatid application. Thus, they are not widely and inten- exchange frequencies: implications for research design. sively supported by funding agencies. Since their hum. Genetics, 65, 280-286 outcome is the characterization of biomarkers and Hulka, B.S. (1990) Methodo1og1c issues in molecular epi

28 Transitional studies demiology. й: Ilulka, B.S. Wilcoskey, T.C. & Griffith, Oxford University Press J.D., eds, Biological Markers in Epidemiology, New York, Schulte, P.A. (1989) A conceptual framework for the vali- Oxford University Press, pp. 214-226 dation and use of biological markers. Env. Res., 48, 129-144 Huika, B.S. (1991) Epidemiologic studies using biological Schulte, P.A. (1992) The use of biological markers in markers: issues for epidemiologists. Cancer Epidemiol. occupational health research and practice. J. ТохгсоI. Env. Biomarkers Prey., 1, 13-19 Health, 40, 359-366 Hulka, B.S. S& Margolin, B.H. (1992) Methodological Schulte, P.A., Boeniger, 1., Walker, J.T., Schober, S.E., issues in epidemiologic studies using biologic markers. Pereira, M.A., Gulati, D.K., Wojсieсhowski, J.P., Garza, A., Am. J. Ep i L, 135, 200-209 гдет п Froelich, R., Stгauss, G. et al. (1992) Biologic markers ira hospital workers exposed to low levels of ethylene oxide. Khоurv, M.S., Adams, М.J., Jr & Flanders, W.D (1988) An epidemiologic approach to ecogenetics. Am. J. Hum. Mutat. Res. 278, 237-251 Genetics, 42, 89-95 Schulte, P.A. & Perera, F.P. (1993) Validation. In: 5chultе, Massey, B.S. (1986) Measures in Science and Engineering P.A. & Perera, F.P., eds, Molecular Epidemiology: Principles ,Chichester, Ellis Horwopd and Practices, San Diego, Academic Press, pp. 81-109 Mayer,L, Warbiuton, D., Jeffrey, A., Pero, R., Andrews, L, Schulte, P.А., Rothman, N. & Schottenfeld, D. (1993) Wales, B., Tom, M., Latriano, L., Tang, D., Tsai, W.Y., Design consideration иn molecular epidemiology. In: Kuroda, M. & Perera, F.P. (1991) Biologic markers in eth- Schulte, P.А. & Perera, F.P., eds, Molecular Epidemiology: ylene oxide-exposed workers and controls. Mutai. Res., Principles and Practices, San Diego, Academic Press, pp. 248, 163-176 159-198 . & Fleiss, J.L. (1979) l tradass correlations: uses Pеrerа, ЕР & Weinstein, I.S. (1982) Molecular epidemi- Sbiout, Р Е п ology arid carcinogen-DNA adduct detection: new in assessing rator reliability. Psycho. Ball., 86, 420-428 approaches to studies of human cancer causation. J. Tang, D.L., Chiamprasert, S., Sante1la, R.М. & Perera, F.P. Chronic Dis., 35, 581-600 (1995) Molecular epidemiology of lung cancer: carcinogen- Perera, F.P. & Mooney, L.A. (1993) The role of molecular DNA adducts, GST1Vf1 and risk. Proc. Am. Assoc. Cancer epidemiology in cancer prevention. In: DeVita, V.T., Res., 36, 284 Hellinan, S. & Rosenberg, S.A., eds, Cancer Prevention, Trick, B.J. (1995) Application of biological markers in Philadelphia, J.B. Lippincott, pp 1-15 cancer environmental epidemiology. Toxicology, 101, 93-98 Perera, F. & Santella, R. (1993) Carcinogenesis. In: Schulte, Vineis, P., Schulte, P.A. & Vogt, R.F., Ji (1993) Technical . . & Perera F., eds, Molecular Epidemiology: Principles and РА variability in laboratory data. In: Schulte, P.A. & Perега, Practices, San Diego, Academic Press, pp. 277-300 F.Р., eds, Molecular Epidemiology: Principles and Practices, Rothman, N. (1993) Epilogue. In: Schulte, P.А. & Pегera, пап Diego, Academic Press рр. 109-135 F.P., eels, Molecular Epidemiology: Principles and Practices, Wilcox, А.J. (1992) Molecular epidemiology: collision of San Diego, Academic Press, pp 199-2 16 two cultures. Epidemiology, 6, 561-562 Rothman, N., Stewart, W.F., Caporaso, N.E. & Hayes, R.B. Winn, D.M. & Gunter, E.W. (1993) Biologic specimen (1993) Misclassification of genetic susceptibility bio- banks: a resource for molecular epidemiologic studies. markers: implications for case—control studies and In: Schulte, P.A. & Регега, F.P., eds, Molecular Epidemiology: cross-population comparisons. Cancer Epidemiol. Principles and Practices, San Diego, Academic Press, pp. Biomarkers Prey., 2, 299-303 217-235 Rothman, N., Stewart, W.F. & Schulte, P.A. (1995) Incorporating biomarkers into cancer epidemiology: a matrix of biomarker and study design categories. Cancer EpidemioI. Biomarkers Prey., 4, 301-311 CorresporndnnQ author Schatzkin, A., Freedman, L.S., Schiffman, М.H. & P.A. Schulte . (1990) Commentary: validation of inter- Dawsey, S.М National Inslitute for Occupational Safвtу and Health mediate end points in cancer research. J. Nat' Cancer 4676 Columbia Parkway C 14; Cincinnati, 0145226; Iпst., 82, 1746-1752 USA` Schesselman, J.J. (1982) Case-control Studies, New York,

29 Appficativп ы Bioп1агkers in Carcer Epidemiology Toniolo, Р. Botfotta, P., Shuker, D.E.G., Rothman, N., Hulka, B. and Puarce, N., eds ARC 5eiепtиuc PuЫkaliом No. 14 lntеrиliоnаl Agency or пuseaгch or' Cancer, Lyon, 1997

Logistics and design issues in the use of biological samples in observational epidemiology

J.D. Potter

Standard epidemiological study designs are well suited to answering questions involving the collection of biological samples. However, different designs are better suited—both for design and logistic reasons—to different questions. The strengths and weaknesses of each design are discussed in relation to markers of exposure, susceptibility and early outcome, and to markers used to classify cancers into biologically defined subsets.

Epidemiology has, for over a century, encompassed by carcinogens (both chemical and viral) and pro- observational and experimental modes. The earli- moters that are intended to model but are essen- est modem epidemiological endeavour, which we tially unrelated to, the human disease. Examples all lеатп at our mentor's knee, is the story (usually, include the use of dimethylhydrazihe (DII) as a alas, oversimplified) of John Snow and the Broad carcinogen in rat colon cancer, dimethylbenz- Street pump. From very early on (but subsequent anthracine (DMBA) as a carcinogen in mouse to the Snow experiment), epidemiologists have breast cancer, and phorbol esters as promoters in studied both reported and recorded data, on the mouse skin models. As an aside, it is worth noting one hand, and biological samples on the other. that animal 'models' of cancer were originally Infectious disease epidemiology, particularly acute intended to mimic the end-point in order to study outbreaks, requires both `shoe-leather' epidemiol- it better. However, we appear to have lost this per- ogy and the collection of specimens. Chronic spective, and have made the error of assuming that disease epidemiology is similar, particnlarly where because these models mimic the end-point, they there are established biological markers that are also mimic the process. central to the disease process: cholesterol in coro- Since the 1960s, as a result of studying human nary heart disease; glucose and insulin levels in pathology of cancer, it has become dear that there diabetes mellitus; uric acid in gout. What is impor- are many early stages of cancer that are discernible tant about both infectious diseases and the above- and can be studied in their own right, e.g. cervical mentioned chronic diseases is that the clinical, intra-epithelial neoplasia, adenomatous polyps experimental and epidemiological models evolved of the colorecturn, proliferative dysplasia of the largely in concert with each other; biology, under- breast and Barrett's oesophagus. More recently, it stood in one area, rapidly became incorporated has also become clear that, even at a molecular into research endeavours in a different discipline. level, specific carcinogens do not always act in the Cancer is markedly different from other diseases manner that was previously hypothesized {Jin et in two ways. The first is that, unlike the infectious al., 1996) and do not always act in humans in the diseases arid the above-mentioned chronic dis- same way as in animals (Kakiuchi et al., 1995). eases, cancer is not regarded as a systemic disorder. What is now beginning to emerge in the study Consequently, treatment is largely local (surgery, of human carcinogenesis is the kind of conver- radiation), and there is little opportunity to under- gence of clinical, experimental and epidemiologi stand the disease process when the treatment is саl studies that has previously characterized the successfuI; the later stages of disease and their study of heart disease. Although, in the main, we systemic manifestations are regarded as signs of lack the systemic markers and manifestations to be the failure of treatment. Secondly, in experimental used as risk indicators of disease (such as lipopro- models of cancer in animals, the cancer is induced teins, cholesterol and its subtractions, and hyper-

31 Application of Biomarkers in Cancer Epidemiology

tension), there are, none the less, some рrоbаЫе A variety of strategies exist. Almost all require systemic markers for certain cancers, and we do collaboration across disciplines and a supportive, have an increasing understanding of the disease understanding study population. Matches between process at molecular level and of its manifestations design and the question asked are outlined in the at cell and tissue levels. Further, as a consequence following sections. An attempt is made to identify of several decades of largely 'paper-and-pencil' the best way of answering specific questions. A dif- occupational, environmental and dietary ерi ferent typology and approach have been proposed demiology, we have some fairly robust ideas about by Rothman et al. (1995). which exposures increase or decrease risk. Additionally, with developments in epidemiology Clinic-based cohort studies and biostatistics, we know to what extent misclas- Two different kinds of cohort studies can be estab- sification and measurement error attenuate our lished in the clinical setting. The first is a screen- findings. Finally, extensive development in mо1- ing-based cohort where individuals aie enrolled ecular and biochemical methods over the last two specifically following a negative screen (e.g. mam- decades have improved markedly our capacity to mography, colonoscopy or Pap smear), and where detect and quantify components of relevant bio- specific biological samples, such as blood and logical pathways in small samples collected on benign lesions, are stored to establish the relation- large numbers of people. ships between exposure, susceptibility, metabolic Against this background, it is appropriate to markers such as hormones or disease intermediate consider the methodological issues that arise markers (e.g. adenoma or cervix pathology) and regarding the incorporation of biological markers subsequent risk of disease. Examples of such stud- in epidemiological studies. This chapter covers the ies include the New York Breast Cancer Study following: consideration of some of the logistics of (Topiolo et al., 1991) and the US National Polyp sample collection in observational studies in the Study (Winawer et aI., 1993). clinical setting and in population-based studies; The second kind of cohort study that is well the issues surrounding the use of markers of various established in the clinic or hospital setting is the stages in the cancer process—susceptibility, exposure, study designed specifically to follow individuals early biological effect, surrogate or intermediate treated for disease with known pathology (includ- end-points, and markers defining subsets of disease ing first primary cancers) in order to establish the in each of these designs and settings; and the par- risk of further disease (including second primary ticular issue of studies to develop screening markers. neoplasms). Here, this is termed a 'pathology' In undertaking the collection of human sam- cohort. The biological markers that can be used ples for epidemiological studies, the best design in such studies include markers of susceptibility and the easiest logistics are sometimes in conflict. including specific genetic syndromes such as Li- For most etiological questions, population-based Fraumeni, ataxia telangiectasia, hereditary non- case-control studies are preferable to those in polyposis colon cancer (HNPCC), the characteris- which study subjects are recruited solely in the tics of the original tumour itself, and possibly bio- clinic/hospital population. For cohort studies, logical measures of therapeutic exposures. One recruits front the general population wЊ are in gen- example of this design is the Children's Cancer erally good health are preferable to patients (unless, Survivor Study (principal investigator Dr Leslie of course, the cohort is focused on complications of Robison) in the USA, which follows more than the initial disease and its therapy). 20 000 survivors of childhood cancer who have By contrast, the recruitment of a hospital/clinic- lived at least 5 years after diagnosis and were based population makes the logistics of sample col- disease-free at the time. Blood samples are being lection—susceptibility markers, exposure markers, used to establish genetic susceptibility, and the early end-points or pathology samples--consider- cohort is to be followed to a variety of end-points ably easier. As with many epiderniological design so that the effects of genetic susceptibility, envi- and execution issues, it is sometimes necessary to ronmental risk factors and therapeutic exposures establish some compromise between the method- (radiation, chemotherapy) on subsequent disease ologically desirable and the logistically feasible. can be ascertained.

32 Logistics and design issues

associаtàодs with biological markers will be clear, problems of generalizability may be more marked than is the case with a population-based cohort, • в • - в в в given the selective nature of such clinic-based study populations (e.g. who comes for screening, or who gets the first primary cancer). Design Logistics. Markers : From a design standpoint the timing of the col- Exposure (e.g. blood levels; 1-2a 1 lection of biological measures of exposure in rela- adducts) tion to disease outcome is always problematic. If Susceptibility (e.g. genes) 1-2 1 the period between collection and outcome is too Intermediate end-point .. . 2 2 . long, there may be increasing misclassification of exposure levels; too short a period may result i Stгatifiçаtiôп of diseases into 1 2 п biologically defined subtypes early disease influencing the relevant exposure marker. Family-based cohort studies have similar Biology af. first tumour 1 1 logistic issues to population-based studies (Sell xs (pathology cohort) or е pre neoplashc pathology et al., 1995). The pathology cohort, in which the (screening cohort) end-point may be a second primary cancer, in many cases will have biological material from the a1 = most optimal

33 Application of Biomarkers in Cancer Epidemiology

altering the biological markers processes (as well as, of course, specific behaviours) of interest in the Table 2' Optimality of design and relative etiology of the disease. Such a setting is therefore ease of sample collection logistics in the best approach to establishing the relation clinic-based case-control studies between exposures, whether measured by self-

1. . ,. report or by biological markers, and the early [11h stages of the neoplastrc process. Similar benefit Exposure (e.g. Ыюоd levels, 1--3a adducts) accrues to the study of specific genotypes (again both metabolic polymorphisms and specific susceptibility (e.g. genes) 1-2° 1 germline mutations) and the known от suspected Intermediate end-point 1. ; intermediate stages of the ceoplastic process. Stratification of diseases 1 It is worth noting that, with a few interesting into biologically defined . exceptions, the relationship between these inter- subtypes mediate disease markers and the cancers of which they are precursors is usually difficult to study, for a1 = most optimal (design) or easiest (logistics). both ethical and logistic reasons. For instance, in bAppropriate for early diseaserпtвrmediate end points but not the case of established colonic adenomatous С пicaty detected cancer 'The pattern of susceptibility génotypes may be distorted by polyps, CIN III and benign proliferative disease of the tendency of those with a family history to seek breast, the precursor lesions are removed and the asymptomalic screening rather.than presenting with early natural history of the disease is disrupted—a parti- diaclase as a resitll of symptoms.. cularly appropriate and satisfactory outcome for the patient but one which, none the less, removes the opportunity for studying a variety of patho- physiological and molecular processes. subtypes based on histopathology, gene expression, One interesting exception is Barrett's oesophagus, mutations, deletions, etc. No particular advantage where the progression of the dysplasia needs close is derived from the clinic-based cohort design for monitoring but the malignant potential is not this purpose except the rather general one that this sufficient to justify the severe intervention of an is likely to be a geographically restricted popula- oesophagectomy. tion; thus facilitating personal contact with Clinic-based studies, where the etiology of can- patients and their physicians, which in turn allows cer itself is the focus of study, allow the collection active follow-up. of end-point cancer tissue as well as blood. However, etiological questions using this material Clinic-based case-control studies are confined to questions of the role of genotypes Table 2 shows, similarly to Table 1, the ranking for and etiology. Exposure markers are not appropriate optimality of design and logistics when colecting in this setting, since the disease itself, the symp- various biological markers in the setting of clinic- toms 0f the disease and its therapy may all alter based case-control studies. These studies are of par- any exposure marker or biological process thought ticirlar value for investigating the etiology of pre- to be a precursor 0f disease. cancerous lesions. For these diseases [e.g. early cer- A particular additional use of this design is to vical intra-epithelial neoplasia (Brock et al., 1988), establish screening markers. What is important in colonic adenomatous or hyperplastic polyps, this variation 0f thé study design is that biological Barrett's oesophagus (Blount et al., 1991; Meltzer markers are sought which classify people with and et aL, 1994)1, very few if any, population-based without a specific cancer or its precursors. Here, registries exist and the detection of the disease, and unlike studies of etiology, the temporal sequence of therefore the identification of disease-free individ- the biological markers and the outcome of interest uals to be enrolled as controls, is invasive. Such is irrelevant if the association is consistent and studies are only possible in the clinical setting. allows a ready distinction between those with and Further, the early stage of the disease reduces, but those without the disease. The issue of study power does not eliminate, the problem of the disease needs to be borne in mind, but there are two facets

34 Logistics and design issues

of disease via exploration of specific intermediate markers and processes. However, unless the cohort Table З. Optimality of design and relative is routinely screened, there may be misclassifica- ease of sample collection logistics tion in the determination of such intermediate in population-based cohort studies end-points. One possibility of course, is to restrict comparisons to those who have been screened as a matter of routine care. The Harvard group have used such an approach to the study of adenoinatous .., polyps within their cohort studies (Giovanrnucci et -' , t 1 a1., 1993). Intermediate end-point :. 2 3 Subsequent exposure-marker collection is also Stratification of diseases into 1 2~e facilitated. This may be relevant when it is in- biologically defined subtypes . . tended to show that cumulative exposure, changes in exposure level or changes in metabolic state are a1 = most optimal (design) or easiest (logistics> important in the progression to cancer. This has bDisease stratification markers will vary in the degree of not been used much iп cohort studies of cancer, logistic complexity depending on whether the marker is although it has been used, for example, in cohorts of routinely measured and recorded in a registry (e.g. estrogen lead-exposed children. Some further biastatistical гесвplогs), wfеthér itis measured and recorded in â chait (e.g, cytogenetics FAB morphology; etc.), or whether it is methodological development to optimize the use measured for ihe study only of sequential exposure measures is warranted. As with the clinic-based cohort studies, the problem of the truncation of the disease process by the to be considered. Although underpowered studies removal of early stages of the neoplastic process are in some danger of rejecting, as statistically non- remains. significant, a promising marker that could be used A possible disadvantage of the cohort design, as in screening, it is nevertheless the case that for a compared with the case-control design, is the fact screening marker to be of major use, it should хеаd- that usually, cohorts are somewhat restricted in лу (i.e. with small study size) distinguish between their age range and thus may not provide the full those with and those without the disease. The picture if cancers at specific ages are determined added complication is the possibility that two or differently, i.e. if there is an interaction between more markers in combination may provide a use- exposure markers and age, or between metabolic ful screening battery. Input into the design of such markers and age. studies should be obtained from biostatisticians, The logistics of sample collection in a population- biology and clinical practice. Some further devel- based cohort study are the most problematic, as most opment of power calculations, methods and cohorts are not able to be recruited via a clinical facil- design for such studies seems warranted. ity. Whether recruited by mail or by a specially established clinic, costs of biological collection are high. Population-based cohort studies As with the clinic-based cohort study, blood samples The population-based cohort study (see Table 3) or buccal cells can be used to establish a wide variety has a number of advantages over other epidemio- of genotypes from very little material using PCR, but logical study designs. Its prospective nature studies of serum or red-cell markers of exposure (e.g. reduces bias, both in the collection of histories of serum nutrient levels, haen oglobin adducts) will exposure and in the use of biological measures of rapidly deplete the stored samples. Nested case- exposure. There are no specific advantages over control analysis rerпafпs the design of choice—as is case-control studies in the use of gentypiog data planned for the EPIC study (Riboli 1992). nor in the capacity to stratify the outcomes of interest on biological subtypes. Population-based case-control studies A specific advantage does accrue, however, if, This design is inferior to the cohort design in the using regular re-contact of cohort members, it is biological measurement of exposure and, of course, possible to establish aspects of the natural history in the far more limited focus of the outcome

35 Application of Biomarkers in Cancer Epidemiology

measured self-reported intermediate markers such as pigmented naevi), there are no opportunities to Table 4.Optimality of design and relative conduct population-based case-control studies of ease of sample collection logistics in intermediate markers. population-based case-control studies In summary, each of the standard studies used

1-i . in observational epidemiology acquires specific гп г advantages and disadvantages when biological Exposure (ë.g. blood levels, No Ь 3 adducts) material is used to measure exposure or susceptibility, or is obtained in order to define specific Susceptibility (e.g. genes) 1 3 intermediate end-points or to establish subsets of Intermediate end-point [1] Noc cancer outcomes. The screening clinic-based cohort study is most valuable in the follow-up to Stratificalion of diseases into 1 2_зд a specific cancer end-point in individuals known biologically defined subtypes to be free of the disease at baseline. The focus is on e1 = most optimal. (design) or easiest (logistics). both exposure and susceptibility markers relevant lt may be acceptable for studies of very early stage disease to that end-point. The clinic-based case-control (Т1; M; No), but the problem of disease affecting the biological study is of particular value in establishing marker of exposure precludes гnогé general use.. the etiology of pre-cancerous lesions and in 'Nith a few excephons e . Pap.smear registries or accurately developing promising screening markers. The self reported intermediate lesions such as pigmented nasal, population-based cohort study gives the best . there will be few opportunities to conduct population-based case-control studes on intermediate arid points opportunity for a generalizable study of etiology with exposure markers analysed in nested case- °See.fDotnate'Ь' 1oTaЫe 3 control fashion. The population-based case- control study provides the most cost-effective way of studying genetic susceptibility in the etiology of measures (single rather than multiple diseases). Its specific cancers. major advantages (Table 4) are its lower cost (although the cohort design almost certainly References involves a lower cost per cancer studied) and its Blount, P.L., Ramel, 5., Raskind, W.H., Haggitt, R.C., broader age range. The best use of the population- Sanchez, C.A., Dean, PJ., Rabinevitch, P.S. & Reid, BJ. based case-control study is in the exploration of (1991) 17р allelic deletions and p53 protein overexpres- the role of specific genotypes in etiology (Heckbert sien in Barrett's adenocarcinoma. Cancer Rca., 51, 5482-5486 et аL, 1992; Kadlubar et al., 1992) (again, both the tarer genetic mutations and the more common Brick, K.E., Berry, G., Mock, P.A., MacLennan) R., metabolic polymorphisms may be studied) and in Truswell, A.S. & Brmton, L.А. (1988) Nutrients in diet the end-point stratification by molecular, histo- and plasma and risk of in situ cervical cancer. J. Nat[ logical or other means. The number of samples Cancer Iпst., 80, 580-585 needed to answer either of these questions is Giovannucci, E., Stampfer, М.J., CoIditz, G.A., 0mm, markedly lower than in the cohort design, but the E.S., Trichopou1os, D., Rosner, B.A., Speizer, F.Е. & logistics of blood collection is likely to be just as Willett, W.C. (1993) Folate, methionine, and alcohol complicated as in the cohort, since neither cases intake and risk of colorectal adenoma, J. Nat1 Cancer inst. 85, 875-884 nor controls are being recruited via a clinical route. Exposure markers (for design reasons) and inter- Heckbert, S.R., Weiss, N.S., Hornung, S.K., Eaton, D.L. & mediate markers (for logistic reasons) frequently Motulsky, A.G. (1992) Glutathione S-transferase and cannot be studied in a useful fashion in this epoxide hydrolaae activity in human leukocytes in rela- design. The collection of exposure markers in a tion to risk of lung cancer and other smoking-related cancers. J. Nat! Cancer Inst., 84, 414-422 case-control fashion is not likely to be useful, as noted above in the discussion on clinic-based Jin, Z., Houle) B., Mikeev, A.M., Cha, R.S. & Zarbl. H. case-control studies. With a few exceptions (e.g. (1996) . Alterations in Hras1 promoter conformation during NMU-induced mammary carcinogenesis and population-based Pap smear registries or well-

36 Logisaics and design issues

pregnancy. Cancer Res., 56, 4927-4935 Sellers, T.A., Anderson, V.E., Potter, J.D., Bartow, S.A., Chen, R-L, Everson, L., King, R.A., Kuni, .C., Kushi, adlubar, F.F., Butler. M.A., Kaderlik, K., Chou, J. & Lang, N.P. С К L. ., McGovern, P.G., Rich, S.S., Whitbeck, J. & Wiesner, (1992) Poiymorphisms for aromatic amine metabolism Н G. (1995) Epidemiologgic and genetic follow-up study of Environ. in humans: relevance for human carcinogenesis. 544 breast cancer families: design and methods. Genetic Health Persil., 98, 69-74 Ергдеттiol., 12, 417-429 Kakiuchi, H., Watanabe, M., Ushijima, T., Toyota, I., Toniolo, P.G., Pasternak, B.S., Shore, .E., Sonnenschein, Sugimura, T. & Nagao, M. А Imai, K., Weisburger, J.H., E,, Koenig, K.L., Rosenberg, C., Strax, P. &Strax, S. (1991) (1995) Specific 5'-GGGA-3' - 5'-GGA-3' mutation of the Endogenous hormones and breast cancer: a prospective gene in rat colon tumors induced by 2-amino-l- Арс cohort study. Breast Cancer Res. Treat., 18, X23-S26 methуl-б-phenylirmdazo[4,5-h]pyridine. Proc. Nod Ac&l. Sci. USA, 92, 910-914 Winawer, S.J., Zauber, A.G., Ho, M.H., O'Brien, M.J., Gottlieb, L.S., Sternberg, S.S., Wayne, J.D., Sckapiro, M., Meltzer, S.J., Yin, J., Mania, B., Rhyu M.-G, Cottreil, J., Bond, J. ., Panish, J.F., Ackroyd, F., Shike, M., Kurtz, Hudson, E., Redd, J.L., Krasna, M.J., Abraham, J. . & Н М R.C., Hornsby-Lewis, L., Gerdes, H., Stewart, E.Т. & Reid, B.J. (1994) Microsatellite instability occurs fre- National Polyp Study Workgroup. (1993) Prevention of anexploid cell popula- quently and in both diploid and colorectal cancer by colonoscopic polypectomy. New tion of Barrett's-associated esophageal adenocarcinomas. Fig!. J. Med., 329, 1977-1981 Cancer Res., 54, 3379-3382 Riboli, E. (1992) Nutrition and cancer: background and rationale of the European Prospective Investigaцoд into Cancer and Nutrition (EPIC). Ann. licol., 3, 783-791 • J.D.Potter Rothman, N., Stewart, W,F. & Schulte, P. (1995) Cancer Prevention Research Program, Fred Hutchinson Incorporating biomarker's into cancer epidemiology: a Cancer Research Center, Seattle, WA,.USR matrix 0f biomarker and study design categories. Cancer Epidemiol. Biornarkers Prey., 4, 301-311

37 ApplicaUci of Biomerkera h Cancer Epidemiology Toпioio, Р., Billette, P, Shuker, [.E.G., Rolhman, N., lulka, B. and Penroe, N. eds IARC Scieпlilic Publicalions Ni. 142 International Аgепсц for Research on Cancer. Lyon, 1997

Methodological issues in the use of biological markers in cancer epidemiology: cohort studies

D.J. Hunter

In this chapter we summarize the major strengths and weaknesses of cohort studies; consider how these characteristics influence the use of biomarkers in cohort studies; briefly review considerations of statistical power, design and the influence of measurement error in cohort studies; and discuss some of the emerging ethical considerations that relate to the use of biomarkers in prospective studies.

Most of the considerations that make prospective Selection of controls studiés an attractive study design apply to expo- For results to be unbiased in a case-control study, sures in general, not only to biomaikers. the controls must be representative of the population that gave rise to the cases (Breslow & Day, strengths 1980); deviations from this principle may lead to Exposure measured before outcome selection bias. The choice of a control series that The major strength of cohort studies is that the meets this criterion is frequently difficult. Random temporal relationship between exposure and out- sampling from the underlying population may be come reflects the sequence of the causal pathway, impractical, even if this population can be defined; i.e. exposure is measured before disease rather than the source population for cases presenting at a afterwards, as in case-control(retrospective) studies. cancer referral hospital, for instance, may be very Thus, we can be assured that the Ыоmarker level in difficult to specify. In a cohort study, the source cases preceded the outcome. This timing has three population that gave rise to the cases is explicitly important implications: defined (cohort members at risk of disease), and 1. Provided that undiagnosed or pre-clinical thus selection bias should be minimal as long as disease has not altered the biomarker level (see follow-up rates are high. As it is rarely feasible to below) we can be sure that the biomarker has not measure biomarker levels of all cohort members at been influenced by the disease or its treatment. baseline, a sample of cohort members is usually 2. In a long-running cohort, a range of different tested, using either a nested case-control or case- times will exist between biomarker collection cohort design (see below). and diagnosis of the outcome. This can be used to investigate hypotheses relating to different Participation latent periods between exposure and outcome, Selection bias can also be introduced if less than although the timing of first exposure (if it 100% of cases and eligible controls choose to par- occurs prior to baseline) is usually uncertain. ticipate. Participation rates, already low in some 3. Other exposures that are assessed prior to dis- settings, such as large cities in the USA, may be ease, such as questionnaire measures of diet, further adversely affected in studies incorporating exercise or other lifestyle characteristics, should biomarkers, as collecting blood, urine or other also be free of the influence of diagnosis of dis- samples usually requires more effort from partici- ease, and biomarker-environment interaction pants than studies using questionnaire-based mea- analyses should not be biased by the differential sures alone. In a cohort study, participation theo- recall of cases and controls (recall bias) which retically should be 100% if reserves of specimen are may occur in case-control studies. available for all subjects and follow-up is complete.

39 Application ы Biomarkers in Cancer Epidemiology

Weaknesses In summary, for relatively conurion outcomes Statistiсаl power for which sufficient cases are available, then the The major weakness of cohort studies is that the prospective design of a cohort study utilizing bio- number of cases of even the more common dis- markers has substantial advantages over retrospec- eases is limited unless follow-up time is very long. tive designs, although detailed questionnaire- A large cohort may not produce enough cases of based or interview-based information on other uncommon diseases even if follow-up is main- exposures may be lacking. tained until the last cohort member dies. This may be particularly troublesome for studies using bio- Timing of specimen collection and disease markers, as only a subset of people willing to be in Iп any cohort study there will be a range of times a questionnaire-based cohort study may be willing between the date of specimen collection and the to undergo the additional inconvenience associ- date of disease, due to staggered dates of collection, ated with providing a specimen such as blood or as well as a variation in the onset of disease. urine for biomarker analysis. Solutions to this problem include pooling information from multi- The influence of pre -clinical disease ple cohorts [for example, Mueller et al. (1989) A major concern in cohort studies of short dura- obtained sera from cases and controls in several tion is the possibility that the disease process has longitudinal studies to study Epstein-Barr virus influenced the biomarker level, even if disease is antibody profiles and risk of developing Hodgkin's undiagnosed at time of specimen collection. This lymphoma], conducting a study in a high-risk problem is most likely to occur if the average follow- cohort (e.g. uranium mine workers for lung cancer) up time is months or 1-2 years, and can be mini- or conducting a case-control study, in which the mized if procedures for disease screening at base- population giving rise to the cases is much larger line are used—for instance, if the end-point is lung than is feasible for a typical cohort study. cancer, through the exclusion of persons with recent weight loss or lesions on chest X-ray. Specimen size and appropriateness Alternatively, excluding cases occurring in the first Some biomarkers require large amounts of speci- 1-2 years of follow-up may eliminate this effect of men (e.g. several hundred millilitres of blood for pre-clinical disease. However, distinguishing certain pesticide residues), от unusual specimens between this source of bias and a genuine associa- that it may be impossible to collect in a prospective tion of a biomarker with disease that is only man- manner from large numbers of subjects (e.g. adi- ifested early in follow-up (e.g. a late promotion pose biopsies). effect in cancer) can be difficult. The long-running controversy about whether low serum cholesterol Level of detail of other exposures levels are associated with increased risk of cancer is In general, the 1éve1 of detail for most question- a good example. In several prospective studies, naire-based exposures is lower in a prospective even those such as the Multiple Risk Factor study than in a retrospective study for which the Intervention Trial (MRFIT) in which a physical questions can be focused on the subset of expo- examination was conducted at baseline, persons sures of direct interest with respect to the outcome with lower serum cholesterol levels were at higher under investigation. Furthermore, unless exposure risk of cancer diagnosis; this effect was limited to history is updated frequently (many cohorts mea- the first 2 years of follow-up (Sherwin et al., 1987). sure exposures only once at baseline due to the Although the consensus (Lewis & Tikkanen, cost of obtaining updated measurements) then the 1994) is that this is an example of bias intro- exposure history between the baseline question- duced by pre-clinical disease, there are still those naire or interview and disease occurrence will be who argue that it is a valid and potentially causal missing in a cohort study. Obtaining follow-up association. measurements to supplément baseline measurements may be particularly problematic for bio- Inference from a single biomarker level markers, because the expense of collecting and A major weakness of cohort studies in general is storing specimens is substantial. that exposure data may be sparse and only avail-

40 Methodological "гssues: cohort studies

Observed relative risk given specified true relative risk Plasma hormone ICC I .5 2.0 2.5 l.5 1.9 23 s1ВG (nmoln) 0.92 Percentage of free estradiol 0.80 1.4 1.7 2.1 Estradiol (pg/m1) 0.68 1.3 1.6 1.9 Profactin (ng/ml) 0.53 1.2 1.4 1.6

aTrue relative risk. Юbsérцed relative risk. From нankinson et al. (1995).

able for a single point iп time (e.g. baseline in a Time-integration. In most studies of the epidemiol- prospective study which does not update expo- ogy of chronic diseases, the exposures of interest sure). This may have several consequences, as are long-term average exposures, because the described below. induction periods for diseases such as cancer or cardiovascular disease are usually thought to be of Misclassification due to random within-person variation. the order of years or decades. Thus, it is usually Within-person variation is a characteristic of most desirable that a biomarker reflect cumulative exposure over at least months or years. Time-integra- biоmarkess that are not fired characteiistics, such as genotype. Substantial variation may be minute- tion is often a function of the nature of the bio- to-minute (e.g. blood noradrenaline), hour-tological sample being assayed. For many nutrients, hour (e.g. blood glucose), day-to-day (e.g. 24-hour for instance, concentrations in erythrocytes are urine sodium excretion), month-to-month (e.g. less susceptible to short-term fluctuations than in plasma p-carotene), or longer, and may have a plasma or serum (Hunter, 1990). Concentrations in diurnal rhythm (e.g. plasma cortisol), may be adipose tissue frequently represent a longer expo- related to the menstrual cycle (e.g. plasma hor- sure history than those in erythrocytes. A 24-hour mones), or may have a seasonal basis (e.g. plasma urine sample is more likely to be representative vitamin D levels). If this variability is random and of long-term intake than is a random urine sample. non-differential between future cases and controls, Feasibility constraints are frequently paramount in then the consequence will be bias to the null in determining the choice of biological specimen to be the relative risk calculated from the biomarker. obtained in a cohort study; however, consideration This attenuation of relative risks is proportional should always be given to the implications for the to the amount of within-person variability, and time-integration of the biomarkers of interest. can be substantial. For example, Hankinson et l. (1995) calculated the intradass correlations Inference from multiple biomarker levels. An obvious between three measurements of levels of four hor- méthod for increasing the time-integration of mones among 79 postmenopausal women; these exposure is to obtain specimens at several points in ranged from 0.53 (prolactin) to 0.92 (sexhormone time. These can be averaged, or persons with sus- binding globulin, SHBG). For a true relative risk of tained high levels can be compared with those 2.5, the observed relative risk would be 2.3 for with sustained low levels. Methods have been SHBG, but only 1.6 for prolactin, if the relative risk given to calculate the number of replicate mea- was based on a single measurement rather than the surements required to estimate the 'true' underly- average of three (Table 1). ing mean value of a biomarker within a specified

41 Application of Biomarkers in Cancer Epidemiology range of error (Lui et al., 1979), if within-person introduced by the fact that a single measure non- variation is assumed to be random. If within-person differentially misclassifies subjects with respect to variation is not random, but is due to changes in their true average exposure. Even a single repeti- behaviour or secular trends in exposure, then mul- tion of a measurement can be used to estimate the tiple measurements may be usefuI to account for correction factor; in fact, it has been shown this source of exposure misclassification. (Willett, 1990) that the correction factor is more The obvious problem in a large cohort study is precisely estimated in data with two measurements that obtaining samples at more than one time on a large number of people than with multiple greatly increases the expense of sample collection measurements on a smaller number. To be gener- and storage, and the burden on study subjects. alizaЫe to the larger cohort, the sample asked to participate in a reproducibility study would ideally Inference from biomarker/disease associations. The be a random sample of the cohort. This is rarely strength of the underlying association between a feasible, particularly in geographically dispersed biomarker level and disease depends on whether cohorts, and is usually further limited by less than the biomarker is measured at the point in the perfect participation. Care should therefore be causal pathway when this assodation is maximal, taken in interpreting reproducibility study infor- or whether it is measured at some other point. mation from less representative populations. Conventional theories of cancer causation distinguish between early events in the causal pathway Specimen collection issues ('initiators') arid later events ('promoters'). It is Collection and storage of a large specimen bank is increasingly clear that the latency period between a complex operation with numerous pitfalls. exposure and cancer diagnosis can be very long, A consideration that is particularly important in e.g. three decades or more for the deleterious effect prospective studies is how to minimize the loss of of smoking on colorectal cancer (Giovanntrcci et al., information due to exhaustion of the sample. In 1994), compared with about one decade for the pro- case—control studies this may be less problematic, tective effect of aspirin on this cancer (Giovannucci as much more sample may be available than it is et al., 1995). measurement of a biomarker of a feasible to colect and store from the large num- putative initiating exposure may thus need to be bers of participants enrolled in a typical cohort several decades prior to disease; measurement of study. The problem of sample exhaustion is obvi- more proximal exposures, such as a late promoter, ously most acute with respect to incident cases of could be several years prior to exposure. If the mea- disease; if every biomarker of potential interest is surement is made at the wrong time, however, the entertained, then a few millilitres of plasma can association may be attenuated or missed altogether. vanish very quickly. Due to the advantages of The within-person variability of the biomarker cohort data, interest from the scientific commu- is also relevant to determining how close to the nity may be great, and requests for samples fre- time of maximum effect the biomarker needs to be quent. Refusal to part with precious samples may measured. At the extreme, if the within-person be met with hostility from those who are refused. variability is nil, e.g. a genotype marker, then the It may be wise for those in charge of specimen time of measurement prior to disease is not critical. banks to consider a number of common-sense pro- In contrast, if the correlation of levels of a contin- cedures including: uous biomarker over time is low, then time of sampling will need to be close to the time of maximum • colIaborating with laboratory colleagues to effect, if this effect is to be estimated accurately. minimize the amount of sample required to perform an assay with acceptable precision; for Measurement of within-person variability example, Laden et al. (1997) collaborated with If within-person variability in biomarker levels is Dr Mary Wolff of Mt Sinai Medical Center on random, then knowledge of the correlation in a the development of methods to reduce the population of a single measurement with the amount of plasma required to perform assays 0f average of multiple measurements can be used to DDE and PCBs from 1 ml to 0.5 ml with equiv- 'correct' for attenuation of relative risk estimates alent precision;

42 у1е1hodоЭоgiсal issues: соhоп studies

• seeking collaborators who can measure a ing or non-fasting, etc. Finally, batch-to-batch vari- range of biomarkers in simuItanеous multiplex ability in laboratory assays can be substantial; thus, fashion on one aliquot of sample; where possible, cases and controls should be • organizing an advisory board that can review run together in the same laboratory batches for requests for samples, establish priorities and maximum comparability, whatever the study share responsibility for decisions on the pur- design. poses for which samples are to be used. Selection of controls: nested case-control or Although these concerns are particulaтly acute case-cohort for case samples, it is also desirable not to exhaust Iп a large cohort study it is rarely desirable to mea- samples from participants selected as controls, as sure levels of a biornarker on all cohort members at controls may become future cases. For nested baseline, largely due to the expense this would case-control analyses of multiple biomarkers, it is entail. Some exceptions may exist—for example, usually most efficient to choose a single set of con- assays that require fresh specimen or can be auto- trols, in order to ensure comparabdity of relative mated, e.g. haemoglobin or white cell count. The risks associated with each biomarker, as well as to usual approach, however, is to select all the inci- minimize the demands of aliquoting and sample dent cases, and to sample the control information. handling. This may rapidly exhaust the sample from these controls however, and thus many Nested case-control investigators establish rules that mandate that a In a nested case-control study, the controls are new control is selected when the amount of sam- selected from the population which gave rise to ple for a control falls below some minimum value, the cases. Ideally, controls are selected for each case in order to preserve specimens for future analyses from the risk set of participants eligible to become if the control becomes a case. a case at the time the case was diagnosed. It may be computationally easier to select from the set of Matching controls to cases participants at baseline who did not become a case Matching of cases to controls on a variety of char- during follow-up. If the amount of specimen acteristics may be used by investigators to increase needed is large, this latter procedure may also be the efficiency of statistical analyses. In biomarker preferable as it means that all incident cases are studies, an additional level of matching is usually analysed; in the former risk set sampling, some case required to ensure the comparability of biomarker specimens may be unavailable if the case had been information between cases and controls. chosen as a control for a case occurring earlier in follow-up. It has been shown that if the cumula- sample comparability tive incidence of disease is low (e.g. less than 5%), Biornarker levels are not fixed attributes of specimens, sampling from the baseline cohort introduces neg- but may be altered by a wide range of procedures ligible bias (Langholz & Thomas, 1990). inherent in sample collection and storage. Thus, the overriding requirement for validity in bio- Case cohort marker studies is that case and control specimens One of the inconveniences of nested case-control are handled in the same way. In prospective stud- studies is that as the number of case-control sets ies, an additional factor to consider is length of increases, the laboratory work required to conduct storage; levels of many biomarkers degrade over analyses increases. A potential solution to this time, and even under ideal storage conditions this problem is the case-cohort design, in which the degradation may be substantial (Bole11i et al., population distribution of the biomarker of inter- 1995). Thus, it is usually necessary to match cases est is measured at baseline in a subcohort, and and controls on duration of storage, in addition to then information about cases is added subse- other factors such as season of specimen collection quently as they occur. The calculations necessary (for biomarkers with seasonal variation), number of to compute the variance of relative risks in this freeze-thaw cycles to which the specimen has been study design are complex (Prentice, 1986), but are exposed, whether the specimen was collected fast- now available in statistical packages (1-lirosoft

43 Application of Biomarkers in Cancer Epidemiology

International Corporation, 1993) which make their • no interventions are established that could calculation convenient. help a person at higher risk to avoid disease; The major problems with the case-cohort • if a screening method is available (e.g. mam- design in biomarker studies are as follows: mography, sigmoidoscopy), the recommendations on frequency of screening are similar, 1. If sample degradation over time occurs, dif- whether or not the information on biomarker ferential degradation will occur in case and sub- level is known; cohort specimens. 'it would be logistically difficult or impossible 2. If laboratory drift in the biochemical analysis to communicate the information to cohort is a concern, then having all the subcohort members; samples analysed at a different point in time • the biomarker is clinically available and rou- from the cases may introduce substantial bias. tinely performed (e.g. serum cholesterol); This problem tends to be less acute when the • the within-person variability of the biomarker biomarker is a qualitative trait such as genotype. is so high that a value from the past (e.g. cohort 3. Laboratory personnel are more easily unblinded baseline) would be poorly correlated with the to case and non-case specimens in this design. current value; • information (for cases) about pre-diagnostic These problems make the case—cohort design of levels has no prognostic significance, and thus limited utility in prospective biomarker studies. is not useful for someone who has already been diagnosed with disease. Choosing the cohort The range of exposures in a cohort of members of Another ethical norm in cohort studies is that the general population will reflect the exposures explicit consent to the exact biomarker being mea- typical in the general population, and rare exposures sured has not been mandatory, as this may vary or unusually low or high exposure levels may be rare depending what disease the case developed, and or even absent. If the research depends on having with the rapid developments in the field of poten- large numbers of persons at the extremes of exposure, tial biomarkers, it was impossible to predict at the or persons with rare exposures, e.g. certain occupa- time of enrolment of subjects which biomarkers tional exposures, then oversampling these persons might be of scientific importance in the future. or restricting to certain groups may be desirable. Thus, a global consent to research use of the bio- Iogical specimen has usually been assumed to be Ethical issues adequate for the conduct of biomarker studies. If a biomarker is known to be predictive of disease These norms are challenged by the rapid growth probability, or if our own research shows that this of potential markers of genetic susceptibility. High is the case, then the question of whether to inform penetrance susceptibility genes (relative risks may the participants in the study of their biomarker be 50 or higher) are usually identified through result should be asked. This issue is particularly ger- family studies, in which the occurrence of a parti- mane in prospective studies in which repeated cular disease leaves no one in the family in doubt of re-contact with cohort members may occur during what disease is being studied; in which a powerful follow-up. Iп general, communicating this m liг- incentive may exist for family members to partici- mation to cohort members has not been the norm pate in the research; in which issues of clinical in epidemiological studies for a number of reasons, treatment, screening aid prophylactic procedures including the followiпg: (e.g. prophylactic mastectomy in breast cancer kindreds) are substantial; and in which relatively few • relative risks of the order of 1.1-3.0 are con- participants are involved and thus relatively inten- sidered sufficiently low that the information is sive counselling is possible aid appropriate. In not sufficiently predictive to be clinically useful; addition, because these studies involve family • the biomarker level is not considered defi- members, and frequently children, major issues of nitely predictive of disease and more studies are confidentiality and/or coercion may arise, and needed; thus procedures for informed consent must be rig-

44 Methodological issues: cohort studies onus. For these same reasons, counselling to warn analysed in a nested case-control design. participants of the potential psychological impact Covering all possibilities would provide partici- of knowledge of inherited susceptibility, and post- pants with an overwhelming amount of infor- test counselling to mitigate potential harm, is fre- mation and potential choices, and would prob- quently intensive. ably corrupt rather than enhance the process of Biornarkers of lower degrees of risk may be estab- informed consent. lished through the `candidate gene' approach: the • New bioniarkers are becoming available that frequency of polymorphisms that are known to alter may not have been imagined at the start of the the function of genes which might plausibly be study, and thus prospective consent is impossible. involved in cancer etiology may be compared in • Reconsenting participants for every new marker cases and controls. This may result in the identifi- would involve multiple re-contacts, and partic- cation of biomarkers which convey much lower ipation rates would almost certainly fall, even in degrees of risk (e.g. relative risks of 1.5-3.0) or studies in which re-contact is possible. Many which only convey risk in the presence of certain prospective studies, however, are designed to take environmental exposures (if a gene- environment advantage of passive follow-up through cancer interaction exists). registries, aid re-contact may not be feasible. For all of the reasons specified above, prospective studies of both low and high penetrartce genes If epidemiology is to make its appropriate, and are preferable to retrospective studies, in order to socially desirable, contribution to these studies of provide the most unbiased information. This infor- biomarkers of genetic susceptibility, then guidelines ration may be important in advising persons with need to be established which are appropriately these genotypes on their age-specific probability of respectful of the rights and concerns of subjects, disease and on lifestyle modifications they could but which do not rule out prospective studies of make to reduce this probability and may have a this whole class of biomarker research. These bearing on whether general population screening guidelines should be drafted by epidemiologists, recommendations are adequate in their situation. ethicists, lawyers, clinicians, activists and repre- However, because genetic screening is involved, sentatives of research subjects, working together to some experts hold that ethical norms previously resolve these issues. thought to be adequate for the assumption of consent in prospective studies involving biomarkers References are inadequate. A recent statement by members of Воlеш, F., Nuti, P., Michell, A., Sciajno, R., Franceschetti, the Ethical, Legal Scientific, Implications (ELSI) F., Krogh, V., Pisani, E & Berrino, F. (1995) Validity for panel of the US National Institute of Human epidemiological studies of Iong-term cryoconservation of steroid and protein hormones in serum and plasma. Genome Research (Clayton et al., 1995) points out Cancer Epidemiol. Biomarkers o., 4, 509-513 the potential for hair to study participants if con- Рг fidentiality is breached in genetic susceptibility Breslow, N.Е. & Day, N.E. (1980) 5tatistiсаI Methods in research, and underlines the lack of consensus on Cancer Research, Vol. 1, The Analysis of Case-Coпtгоi appropriate standards of informed consent in this Studies (IARC Scientific Publications No. 32), Lyon, International Agency for Research on Cancer, area. The extreme position is that any biomarker of genetic susceptibility requires the same explicit Clayton, R.W, Steinberg, KK., Khoury, M.J., Thomson, and detailed informed consent that is the norm in E., Andrews, L., Kahn, M.E., Kopelman, L.М. & Weiss, J.O. (1995) Informed consent for genetic research on family studies no matter what the degree of risk stored tissue samples. J. Am. Med. Assoc., 274, 1786-1792 associated with this biomarker, and no matter what the state of the science is about this risk. Giovanmicci, E.L., Colditz, G.A., Stampfer, M.J., Hunter, Adoption of this extreme would render many epi- D.J., Rosner, B.A., Willett, W.C. & Speizer, F.E. (1994) A prospective study of cigarette smoking and risk of col- demiological studies unfeasible, in particular most orectal adenoma and cancer in women. J. Nat! Cancer cohort studies, for the following reasons: Inst., 86, 192-199 • It is difficult to predict who will be tested, and Giovannucd, E., Egan, K.M., Hunter, D.J., Stampfer, M.J., Colditz, G.A., Willett, W.C. & speizer, F.E. (1995) Aspirin for what, at the start of a study which will be

45 Application of Biomarkers in Cancer Epidemiology

and the risk of colorectal cancer in women. N. EngL J. R., Soltero, I., Stamler, R., Gosch, E, Stevens, E. & Stamier, Med., 333, 609-614 J. (1979) Assessment of the association between habitual salt intake and high blood pressure: methodological Hankiпson, S.E., Manson, J.~., Spiegelman, D., Willett, W.C., Longcope, C. & Speizer, F.E. (1995) Reproducibility problems. Am. J. EpidernioL, 110, 219-226 of plasma hormone levels in postmenopausal women Mueller, N., Evans, A., Harris, N.L., Comstock, G.W. over a 2-3 year period. Cancer Ерiдетiоl. Biornarkers Pieu, Je11um, E. Magnus, K., Orentreich, N., Polk, B.F. & 4, 649-654 Vogehnan, J. (1989) Hodgkin's disease and Epstein—Barr Hirosift International Corporation: EPICURE/PEANUTS virus. Altered antibody pattern before diagnosis. N. Engl. J. Med., 320, 689-695 Softwдre (1993) The PEANUTS Program. EPICURE User's Guide. Seattle, WA, Hirosoft Prentice, R.L. (1986) A case-cohort design for epidemio- Hunter D.J. (1990) Biochemical markers of dietary logic cohort studies and disease prevention trials. Biometrika, intake. In: Willett, W., cd., NulTitiorod Epidemiology. New 73, 1-12 York, Oxford University Press, pp. 143-216 Sherwin, R.W., Wentworth, D.N., Cutler, J.A., Huuiey, Laden, F., Wolff, .S., Niguid la, N.J., Spiegelman, D., S.B., Kuller, LI-f. & Stamler, J. (1987) Sегum cholesterol М Ievels and cancer mortality in, 361, 662 men screened Hankinson, S.Е. & Hunter, D J. (1997) Reduced aliquot size tir a plasma organochlorine assay for use in epi- for the Multiple Risk Factor Intervention Trail. J. Am. Med. Assoc., 257, 943-948 demiologic studies. Cancer EрidemioI. Biomarkers Prey. (in press) Willett, W.C, cd. (1990) Nutritional Epidemiology, New Langhoiz, B. & Thomas, D.C. (1990) Nested case—control York, Oxford University Press and case—cohort methods of sampling from a cohort: a critical comparison. Am. J. EpidemioL, 131, 169-176 D.J. Hunter Lewis, B. & Tikkanen, M.J. (1994) Low blood total cho- Harvard School of Public Health, Department of lesterol and mortality: causality, consequence and con- Epidemiology, 677 Huntington Avenue, Boston, founders. Am. J. СагдiоIоgy, 73, 80-85 MA 02115, USA Liu, K., Cooper, R., McKeever, J., McKeever, P., Byington,

46 Арр]icatioп of Biomaгkегs in lancer Epidemiology Тег ioыo, Р, Bofletta, P., 5hukег, D.E.G., Rothman, N., HWka, В. arid Pearce, N., rids IARC Scientilio PuЫicatioпs No. 142 1пlегпаliопаl Agеnoy for Research on Cancer, Lyon, 1997

General issues of study design and analysis in the use of biomarkers in cancer epidemiology

N. Pearce and P. Boffetta

Other contributions to this volume have discussed sources of variation (see Vineis) and measurement error (see White). In this article, we focus on statistical issues involved in the design and analysis of epidemiological studies that use biomarkers. We do not consider statistical issues of laboratory analyses.

In general, epidemiological research involves study- Measuring disease with biomarkers ing the external, modifiable causes of diseases in ipidemiological studies are usually based on a par- populations (McMichael, 1994, 1995) with the ш- ticular population followed over a particular period tention of developing preventive interventions. In of time. Miettinen (1985) has termed this study some instances, this activity can be enhanced by population the `base population' and its experience using internal biomarkeтs to obtain better mea- over time the `study base'. The different epidemio- surements of internal exposure (dose), disease от logical study designs differ only in the manner in individual susceptibility. Statistical issues in studies which the study base is defined and the manner in using biomarkers of exposure are not markedly dif- which information is drawn from the study base ferent from those involved in other epidemiological (Checkoway et "L, 1989). Thus, epidemiological studies based on measures of external exposure, studies may involve measuring either incidence or while studies using biomarkers of disease pose spe- prevalence of disease. This distinction is important cific prоЫеms due to the lack of persistence of when a biomarker is being used to measure either some of these markers, and the analysis of interac- the disease under study or early biological effects tion is of particular interest in studies using mark- that are considered to be valid predictors of disease ers of susceptibility. As with other epidemiological risk (e.g. Rothman eta?., 1995). In particular, many studies, the statistical analysis of a study involving studies measuring disease with biomarkers are of biomarkers involves in general: (1) relating a par- cross-sectional design and measure the prevalence ticular disease (or health outcome, such as a marker of the disease, which is dependent on its incidence of early effect) to (2) a particular exposure while (3) and its duration. Thus, in a study looking at markers controlling for systematic error, (4) assessing inter- of cell damage as an effect of exposure to known or actions with other exposures and (5) assessing the suspected carcinogens, the results would depend on possibility of random error. We will consider each factors such as the turnover of the cells in which of these five aspects of study design and analysis in the marker is measured or the capacity to repair turn. We will restrict our discussion to full-scale the damage. For example, there is evidence that epidemiological studies, i.e. we assume that the use chromosomal aberrations caused by some carcino- of biomarkers, be they of exposure, effect or suscep- gens, such as arsenic and benzene, last for longer tibшty, aims to contribute to the elucidation of the periods than aberrations caused by vinyl chloride, causal relationships in human populations between and the reason for this difference is not known diseases and factors such as external exposures, per- (Schwartz, 1990). sonal habits, genetic traits and interventions. Issues in the design and analysis of transitional incidence studies studies, in which the main aim is the validation of the Three measures of disease incidence are commonly markers themselves, are outside the scope of this used in incidence studies. The (person-time) chapter. incidence rate (or incidence density; Miettinen,

47 Application of Biomarkers in Cancer Epidemiology

1985) is a measure of the disease occurrence per Incidence case-control studies unit time. A second measure of disease occurrence Incidence case-control studies involve studying all is the cumulative incidence (incidence proportion; of the incident cases of disease generated by the Miettinen,1985) or risk, which is the propoTtion of study base and a control group sampled at random study subjects who experience the outcome of in- from the same study base. The relative risk measure terest at any time during the follow-up period. A is the incidence odds ratio; the effect measure that third possible measure of disease occurrence is the this estimates depends on the manner in which incidence odds (Greenland, 1987), which is the ratio controls are selected. Once again, there are three of the number of subjects who experience the out- main options (Pearce, 1993). come to the number who do not experience the One option is to select controls from those who outcome. As for the cumulative incidence, the in- do not experience the outcome during the follow- cidence odds is dimensionless, but it is necessary to up period, i.e. the survivors (those who did not specify the time period over which it is being mea- develop the outcome at any time during the follow- sured. In incidence studies involving biomarkers up period). In this instance, a sample of controls of disease, it is therefore important to consider chosen by cumulative incidence sampling will es- whether a particular biotuarker measures incidence timate the exposure odds of the survivors, and the or cumulative incidence. odds ratio obtained in the case-control study will Corresponding to these three measures of dis- therefore estimate the incidence odds ratio in the ease occurrence, there are three principal ratio base population. Controls can also be sampled measures of effect that can be used in cohort stud- from the entire base population (those at risk at ies. The measure of primary interest is often the the beginning of follow-up), rather than just from rate ratio (incidence density ratio), which is the the survivors (those at risk at the end of follow-up). ratio of the incidence rate in the exposed group to In such case-base sampling, the controls will esti- that in the non-exposed group. A second com- mate the exposure odds in the base population of monly used effect measure is the risk ratio (cumu- persons at risk at the start of follow-up, and the lative incidence ratio), which is the ratio of the cu- odds ratio obtained in the case-contrai study will mulative incidence in the exposed group to that in therefore estimate the risk ratio in the base popu- the non-exposed group. When the outcome is rare lation. The third approach is to select controls over the follow-up period, the risk ratio is approx- longitudinally throughout the course of the study imately equal to the rate ratio. A third possible (Miettinen, 1976); this is sometimes described as effect measure is the incidence odds ratio, which is 'risk-set sampling' (Robins et al., 1986), 'sampling the ratio of the incidence odds in the exposed from the study base' (the person-time experience; group to that in the non-exposed gюир. An anal- Miettinen,1985) or density sampling' (Kleinbaum ogous approach can be used to calculate measures et al., 1982). In this instance, the controls will esti- of effect based on the differences rather than the mate the exposure odds in the study base, and the ratios, in particular the rate difference and the risk odds ratio obtained in the case-control study will difference. therefore estimate the rate ratio in the study base. In incidence studies involving biomarkers of dis- In incidence case-control studies involving bio- ease, it is therefore important to consider whether markers of disease, it is therefore important to con- a particular biomarker measures incidence or cu- sider whether a particular biomarker measures mulative incidence. For example, if the 'disease' incidence or cumulative incidence. These issues under study is hepatitis B virus (HBV) infection, determine not only which measure of effect is then a survey of the prevalence of HBV markers, in being estimated, but also which method of control a cohort that has been followed over time, will in- selection is appropriate, and the resulting methods dicate the cumulative incidence of infection in the of data analysis. cohort (wit the exception of those who have died from any cause during follow-up or who no longer Prevalence studies show evidence of infection). It will not directly in- The term prevalence denotes the number of cases dicate the incidence rate of infections; this would of disease existing in the population at the time the require repeated prevalence surveys over time. study was conducted. If we denote the prevalence

48 Geпera{ issues of study design and analysis of a disease in the study population by P, and if we the different mechanisms involved are likely to re- assume that the incidence rate is constant over quire different biomarkers for measurement of the time, the population has reached a `steady state' etiologically relevant exposures. and there is no migration into or out of the prevalence pool, then it can be shown (Rothman, 1986) Measuring exposure wïth biomarkers that the prevalence odds is equal to the incidence Validly of biomarkers of exposure rate (I) multiplied by the average disease duration There are considerable shortcomings in many cur- (D): rently available biornarkers of exposure, including problems of measuring historical exposures; un- P1(1 — P) = I x D certainties as to what a biomarker is measuring; greater susceptibility to confounding in some in- Thus, the prevalence odds is directly propor- stances; problems of application to public health tional to the disease incidence, and the prevalence policy (Pearce et al., 1995); the disease process odds ratio is estimated to be: affecting the level of the biomarker; and problems of validity of laboratory measurements (Boffetta, 1995). These issues are covered elsewhere in this OR = I1D110Dп volume. In this section we concentrate on issues of An increased prevalence odds ratio may thus re- study design and analysis when measuring expo- flect the influence of factors that increase the dis- sure with biomarkers, particularly with regard to ease incidence and/or factors that increase disease time-related exposures with a relatively long in- duration. The different mechanisms involved in duction and latency period between exposure and Increasing disease incidence or disease duration are the subsequent occurrence of disease (as is the sit- likely to involve different time patterns of expo- uation in most cancer epidemiology studies). The sure and disease (see below), which in turn are issues we discuss are not unique to a particular likely to require different biomarkers for measure- study design (cohort studies, case—control studies, ment of the etiologically relevant exposures. cross-sectional studies) but rather apply to all studies in which the etiologically relevant time period Prevalence case—control studies involves a relatively long induction and latency Just as an incidence case—control study can be used period, thereby posing problems with the mea- to obtain the same findings as a full cohort study, surement of exposure during this period. a prevalence case—control study can be used to obtain the same findings as a full prevalence study in Time-related exposures a more efficient manner. In particular, if obtaining боте biomarkers measure factors that are fixed and exposure information is difficult or costly (e.g. if it do not change over time in an individual, e.g. genetic involves serum samples), then it may be more effi- susceptibility genes that may interact with xeno- cient to conduct a prevalence case—control study biotic factors in cancer causation (Rothman, 1995). by obtaining exposure information on all of the Other biomarkers measure factors that change over prevalent cases of the disease under study and a time, e.g. micronutrient levels in serum may change sample of controls selected at random from the from day-to-day (Willett, 1990). non-cases. In this instance, a sample of controls In studies (both prospective and retrospective) will estimate the exposure odds of the non-cases, of long-term health effects involving time-related and the odds ratio obtained in the prevalence exposures, it is important that the time patterns of case—control study will therefore estimate the pre- the study exposure and of the relevant con- valence odds ratio in the base population, which in founders should be taken into account in the turn estimates the incidence rate ratio, provided analysis (Pearce et cL, 1986). In particular, it is im- that the average duration of disease is the same in portant that the principal exposure under study the exposed and non-exposed groups. Once again, should be analysed in a time-related manner, an increased prevalence odds ratio may reflect the taking account of the likely induction and latency influence of factors that increase disease incidence periods, and the relative etiological importance and/or factors that increase disease duration, and of exposure intensity, exposure duration and

49 Application of Biornarkers in Cancer Epidemiology

cumulative exposure. The simplest approach is to flect recent rather than historical dietary intake analyse the cumulative exposure in a time-related (Willett, 1990); given the long induction time of manner, and this may suffice when the aim is most cancers it is usually exposures between 10 merely to consider whether or not there is an effect and 30 years previously that are etiologically rele- of exposure. However, once it has been provision- vant. While this may not be a limitation in cross- aIly assumed that an effect exists, attention then sectional studies, provided that the etiologically shifts to understanding the nature of the effect. In relevant time period is close to the time of data col- this context, the temporal pattern of exposure and lection, it is an important limitation iп cohort and outcome cari be considered by examining the case—control studies aiming to assess the effects of effects of exposures in specific time windows while historical exposures. Some biomarkers are better controlling for time-related confounders and for than others in this respect (particularly biomarkers the effects of exposures in other time windows. A for exposure to biological agents), but even the more sophisticated approach is direct fitting of a best markers of chemical exposures reflect only the theoretical model of carcinogenesis (Pearce, 1992), last few weeks or months of exposure. On the other which requires assumptions as to the relevance of hand, with some biomarkers (e.g. serum levels of the times in which the exposure occurred arid in ICDD; Johnson et al., 1992) it may be possible to which the marker was measured. This may be par- estimate historical levels if the exposure period is ticularly relevant in studies including the mea- known, if the half-life is relatively long (and is surement of the exposure, the marker and the dis- known) arid if it is assumed that f0 significant ease, which are mainly aimed at elucidating the exposure has occurred more recently, or if it is role of the marker in the exposure—disease rela- reasonable to assume that exposure levels have re- tionship (5chatzlon et al., 1993). An example of this mained stable over time. type of study is the investigation of the role of human However, historical information on exposure papillome virus (HPV) in the etiology of cervical surrogates will often be more valid than current cancer (the `exposure' in this case being factors direct measurements of exposure or dose. This such as number of sexual partners and age at first situation has long been recognized in occupational intercourse). In this case, the association with HPV epidemiology, where the use of work history infection, when properly measured with PCR-based records in combination with a job—exposure matrix assays, is of such a magnitude that there is little (based on historical exposure measurements of work concern about the relevance of the time periods areas rather than individuals) is usually more valid (Munoz et al., 1992). However, for cancers from than current exposure measurements (whether based other organs, the association with HSV infection is on environmental measurements ou biomarkets) less clear-cut, and the relevance of the timing of because of changes in exposure levels over time the infection may be one of the unknown factors (Checkoway et al., 1989). Similar problems may modifying the exposure—marker—disease relation- occur in the measurement of other carcinogenic ship. exposures. For example, even the best currently Thus, biological measures of time-related expo- available measures of exposure to tobacco smoke, sures must be able to measures changes in expo- such as plasma or urinary cotinine, appear to have sure levels over time. In particular, stored biologi- similar validity to questionnaires for the measure- саl samples may not provide valid measurements ment of current exposures; their very short half- of long-term patterns of exposure when there are life makes them inferior to questionnaires in the significant variations in exposure over time, unless estimation of historical exposures (Pearce et aI., samples have been taken repeatedly over the 1995). On the other hand, some biomarkers would course of the study (Armstrong et al., 1992). If it is appear to have value in the validation of question- not possible to take repeated biological samples over naires (Forastiere et al., 1993), which can then be time, then it is essential that the samples that are used to estimate historical exposures. taken relate to the etiologically relevant time period. Another example in which timing of sample Many currently available Ыоmarkers only indi- collection is of great importance is in the case of cate relatively recent exposures. For example, it is DNA adducts (Wilcosky & Griffith, 1990). Since well known that serum levels of micronutrients re- most adducts are readily repaired, any measure of

50 General issues of study design and analysis exposure based on DNA adducts will depend on is often not practicable, but selection bias can also the time between the end of exposure and sample be controlled in the analysis by identifying factors collection; this pattern can then be modified by that are related to subject selection and controlling factors such as the activity of repair enzymes, for them as confounders. The statistical issues in- which in turn may also have an independent iд volved in controlling for selection bias in the fluence on the outcome, i.e. may be associated analysis are essentially the same as those involved with case—control status. DNA adduct formation in controlling for sources of confounding (see and repair are particularly problematic, since: below).

• the extent to which the amount of adducts Information bias measured represents the amount of biologically Information bias involves misclassification of the active adducts formed (as discussed above) is study participants with respect to disease or expo- usually not known; and sure status. Thus, the concept of information bias • in the case of measurements taken during ex- refers to those people actually included in the posure (or immediately thereafter, such as at the study, whereas selection bias refers to the selection end of a working day) it is not known how of the study participants from the study base, and much of the adducts found would persist long confounding generally refers to non-comparabil- enough to be biologically important. ity of subgroups within the study base. The various methodological issues of validity, reproducibility Analysis based on pооlеd samples arid stability of markers are part of the more gen- These issues of the timing of sample collection are eral problem of information bias. of particular concern in analyses based on pooled Non-differential information bias occurs when samples. If the samples were not all taken at the the likelihood of misclassification of exposure is same etiologically relevant time period, then the the same for both cases and non-cases of disease pooled sample will represent the average of sam- (or when the likelihood of misclassification of dis- ples taken from a variety of time periods. Thus, ease is the same for exposed aid non-exposed per- there may be considerable misclassification with sons). Non-differential misclassification of exposure regard to the exposure levels at the etiologically generally biases the relative risk estimate towards relevant time period. the null value of 1.0 (Copeland etal.,1977). Hence, non-differential information bias tends to produce Systematic error 'false negative' findings and is of particular con- The major possible types of systematic error (bias) cern in studies that find no association between are the same in traditional epidemiology and in exposure and disease. studies involving biomarkers (Boffetta, 1995). The Differential information bias occurs when the various types of bias can be grouped into three likelihood of misclassification of exposure is differ- major classes: selection bias, information bias, and ent between cases and non-cases (or the likelihood confounding (Rothman, 1986). This section is not of misclassification of disease is different between intended to give a comprehensive review of these exposed and non-exposed persons). This can bias the types of bias; rather, we will concentrate on issues observed effect estimate in either direction, either involving data analysis. One solution is to pool towards or away from the null value. small sets of samples stratified on the basis of time Information bias can drastically affect the validity since collection; however, this procedure may sub- of a study. As a general principle, it is important to stantially reduce the advantages of pooling. ensure that the misclassification is non-differential, by ensuring that exposure information is collected Selection bras in an identical manner in cases and non-cases (or 5election bias involves biases arising from the pro- that disease information is collected in an identical cedures by which the study participants are chosen manner in the exposed and non-exposed groups). from the study base. Selection bias can be avoided In this situation, the bias is in a known direction by including all of the study base (i.e. a cohort (towards the null), and although there may be study) and obtaining a response rate of 100%. This concern that not finding a significant association

51 Application of Biomarkers in Cancer Epidemiology

(between exposure and disease) may be due to cern in studies using biomarkers, since the identi- non-differential information bias, at least one can fication 0f potential confounders depends on pre- be confident that any positive findings are not due vious knowledge of the relationship between these to information bias. Thus, the aim of data collection and the relevant variables of exposure and out- is not to collect perfect information, but to collect come, and such knowledge is, for most biomarkers, comparable information in a similar manner from very poor. the groups being compared, even if this means The most straightforward method of control- ignoring more detailed exposure information if ling confounding in the analysis involves stratify- this is not available for both groups. However, it is ing the data into subgroups according to the levels clearly importait to collect information that is as of the confounder(s) and calculating a summary detailed and accurate as possible, within the con- effect estimate that summarizes the information straits imposed by the need to ensure that infor- across strata. However, it is usually not possible to mation is collected in a similar manner in the control simultaneously for more than two or three groups being compared. confounders when using stratified analysis. This In general, cross-sectional and case—control problem can be mitigated to some extent by the studies based on biomarkers of exposure are more use of mathematical modelling, but this may in prone to differential misclassification than studies turn produce problems of multicollinearity when based on measurement of external exposure and variables which are highly correlated are entered disease, since the biomarkers may be influenced by simultaneously into the model. For example, the disease itself. This problem is less relevant in serum levels of various micronutrients may be prospective studies (or nested case—control studies) strongly correlated and multicollinearity may in which the marker is measured on biological occur when they are included in the same model. material collected before the onset of disease, pro- This will lead to unstable effect estimates with vided that cases diagnosed within a short interval large standard errors, and may in fact lead to the after sample collection are excluded. The fact that wrong' micronutrient showing the strongest asso- the relationships between exposure, marker and ciation (negative or positive) with disease. This disease are in most cases obscure limits the inter- may be one reason why numerous studies have pretation of the findings of biomarker-based studies shown that the consumption of green and yellow with respect to the presence or absence of infor- vegetables is protective against a range of cancers, mation bias. but the identification of the specific dietary micronutrients involved has remained elusive (Steinmetz Confounding & Potter, 1991). Confounding occurs when the exposed and non- Iп general, controI of confounding requires exposed groups (in the study base) are not compa- careful use of а priori knowledge, together with rable due to inherent differences in background assessment of the extent to which the effect esti- disease risk (Greenland & Robins, 1986) caused by mate changes when the factor is controlled in the exposure to other risk factors. The concept of con- analysis. Most epidemiologists prefer to make a founding thus generally refers to the study base, decision based on the latter criterion, although it although as noted above, confounding can also be can be misleading, particularly if misclassification introduced (or removed) by the manner in which is occurring (Greenland & Robins, 1985). The study participants are selected from the study base. decision to control for a presumed confounder can If no other biases are present, three conditions certainly be made with more confidence if there is are necessary for a factor to be a confounder supporting prior knowledge that the factor is pre- (Rothman, 1986). First, a confounder is a factor dictive of disease. that is predictive of disease (in the absence of the Misclassification of a confounder leads to a loss exposure under study); second, a confounder is of ability to control confounding, although con- associated with exposure in the study base; aid trol may still be useful provided that misclassifica- third, a variable that is intermediate in the causal tion of the confounder was unbiased (Greenland, pathway between exposure and disease is not a 1980). Misclassification of exposure is more prob- confounder. This latter issue is of particular con- lematic, since factors that influence misclassifica-

52 General issues of study design and analysis tian may appear to be confounders, but control of per day) than the medium or low РАН exposure these factors may increase the net bias (Greenland group (since some of the total РАН exposure comes & Robins, 1985). from cigarette smoke). The dose—response will When appropriate information is not available then be confounded by cigarette smoking, and the to control confounding directly, it is still desirable `high РАН exposure' group may show a higher to assess its potential strength and direction. For lung cancer risk which is not due to РАН exposure, example, it may be possible to obtain information but which is actually due to the other carcinogenic on a surrogate for the confounder of interest. Even constituents of cigarette smoke (Pearce et al., though confounder control will be imperfect in this 1995). One solution is to stratify the analysis on situation, it is stil possible to examine whether the cigarette smoking (as measured by questionnaire), main effect estimate changes when the surrogate is but any confounding control is likely to be imper- controlled in the analysis, and to assess the fect. This is even more of a problem if biomarkers strength and direction of the change. Alternatively, are used to measure tobacco smoking because of it may be possible to obtain accurate confounder the problems of measuring the etiologically rele- information for a subgroup of participants (cases vant constituents of tobacco smoke (as distinct and non-cases) in the study and to assess the effects from exposure to PAIs in tobacco smoke). Once of confounder control in this subgroup. again, confounding control is likely to be imper- A related approach involves obtaining con- fect and, therefore, to yield results that are still founder information for a sample of the study base confounded and less valid than those obtained by (Or a sample of the controls in a case—control only considering occupational exposures (using a study). For example, in a study based on question- job—exposure matrix). Furthermore, one can only naires, biomarkers may be used to validate control for known confounders (e.g. tobacco smoke) questionnaire information in a subgroup of study and cannot control for unknown confounders that participants. may also be subject to the same types of biases The potential for confounding is of major con- described above. Therefore, it is usually preferable cern in all epidemiological studies, including those to avoid confounding rather than to attempt to using biomarkers. The use of biomarkers of exposure control for it post hoc (which is why randomized does not reduce the need to control for confound- trials are the preferred method when they are fea- ing, and in some instances the use of biomarkers sible). Thus, it is preferable to consider only occu- may actually introduce confounding into a study. pational exposures to PAIs, using a job—exposure For example, in a study of lung cancer and РАН ex- matrix, and not to attempt to measure non-occu- posure in a group of factory workers, if the workers pational exposures to РАН using bionrarkers. are classified according to РАН exposure on the Finally, it should be stressed that the above issues basis of industrial hygiene monitoring, then the have been discussed in terms of studies in which percentage of smokers (and the mean number of exposure is measured prospectively; the problems cigarettes smoked per day) will usually be similar are even more acute when historical exposures are in the groups with low, medium and high levels of being assessed. occupational exposure to РАН (since these groups are defined purely on environmental levels of РАН Random error exposure in various job categories and depart- Random error will occur in any epidemiological ments, and these exposures will usually be unrelated study, just as it occurs in experimental studies. It is to cigarette smoking). In this situation, cigarette often referred to as 'chance', although it can per- smoking will fat be a major confounder. However, haps more reasonably be regarded as 'ignorance' if workers are classified according to PAН-DNA (Checkoway et "L, 1989). Even in an experimental levels, these will indicate total exposure to PAf Is study in which participants are randomized into from all sources, including cigarette smoking. 'exposed' and 'non-exposed' groups, there will be Thus, cigarette smokers will be more likely to be in 'random' differences in background risk between the 'high РАн exposure' group, and this group will the compared groups, but these will diminish in therefore contain a higher proportion of smokers importance (i.e. the random differences will 'even (and a higher mean number of cigarettes smoked out') as the study size grows. In epidemiological

53 Application of Biomarkers in Cancer Epidemiology studies, there is no guarantee that differences in very low (or very high) prevalence, thus requiring baseline (background) risk will even out between large samples to detect a difference between groups the exposure groups as the study size grows, but it (Hulka & Margolirn, 1992; Rothman et al., 1995). is necessary to make this assumption in order to Finally, it should be noted that some biomarkers proceed with the study (Greenland & Robins, are of interest in themselves, rather than function- 1986). In practice, the study size depends on the ing as surrogates for other exposures; in particular, number of available participants and the available no alternative methods exist for measuring mark- resources. Within these limitations, it is desirable ers of genetic susceptibility. Nevertheless, the ad- to make the study as large as possible, taking into ditional information provided by the use of such account the trade-off between including more par- markers should still be compared with that pro- ticipants and gathering more detailed information vided by alternative, larger studies in which the about a smaller number of participants (Greenland, marker is not used. 1988). An additional consideration in study size esti- A major problem with the use of biomarkers of mation in studies using biomarkers is the ratio of exposure and outcome in cancer epidemiology the number of assays per individual and the num- studies (and particularly in cohort studies) is that ber of individuals in the study (Boffetta, 1995), of small numbers. Even large multicemtre cohort such as the detection of chromosomal aberrations studies often struggle to obtain sufficient numbers or sister chromatid exchanges. In this case, studies to assess risks of rare cancers from occupational ex- must be based on adequate numbers of individuals posures (Fingerhut et at, 1991; Saracсi et at, 1991). and observations per individual (Hirsch et al., The use of biomarkers may be a major problem in 1984; Wharton, 1985). Many biomarkers show this regard, since the resulting expense and сот- marked variation from day to day within the same рlеxity may drastically reduce the study size, even in individual, and the intra-individual variation may community-based and nested case—control studies, be greater than the interindividual variation and therefore greatly reduce the statistical power (Armstrong et al., 1992). It may therefore be neces- for detecting an association between exposure and sary to take a large number of measurements to disease. As in traditional epidemiology studies, in accurately estimate the average exposure level for studies using biomarkers statistical power is a func- each individual; otherwise it may be impossible to tion of the prevalence of exposure and the magni- detect differences between individuals. For exam- tude of risk; a biomarker with low prevalence aid ple, for 24-hour urinary sodium, the within-person high relative risk can be evaluated in small popu- variation may be three times as high as the be- lations, whereas a biomarker with low prevalence tween-person variation; it has been estimated that and low relative risk requires a larger population. the misclassification resulting from taking only The optimal balance between precision and valid- one sample per person would result in a true rela- ity depends on a number of considerations, in- tive risk of 2.0 being reduced to an observed rela- cluding the relative costs of the various exposure tive risk of 1.2 (Armstrong et at, 1992). Thus, it may measurement techniques (Greenland, 1988). Thus, be necessary to take 10-15 24-hour urine samples for an expensive biomarker to be useful it must be in order to achieve reasonable accuracy in estimat- substantially better than less expensive and less ining average individual sodium intake levels. vasive approaches. However, it has been argued that the necessary study size in some molecular interaction epidemiology studies may be smaller than in tra- Interaction (effect modification) occurs when the ditional epidemiology (Hulka, 1990a,b; Hertzberg estimate of effect of exposure depends on the level & Russek-Cohen, 1993) because of Iarger differ- of another factor in the study base (Miettinen, ences in biomarker distribution, identification of 1974). The term statistical interaction denotes a subgroups at higher risk, and the use of continuous similar phenomenon in the observed data. Thé for- outcome variables (Boffetta, 1995). While this is mer term will generally be used here. Interaction is true in many cases (e.g. the detection of mutations distinct from confounding (or selection or infor- in critical genes as a marker of increased risk of mation bias) in that it does not represent a bias cancer), in other cases the biomarkers may show a which should be removed or controlled, but rather

54 General issues of study design and analysis a real difference in the effect of exposure in vaтious paring the sizes of the odds ratios (relating exposure subgroups that may be of considerable interest. For and disease) in different strata of the effect modi- example, in а cohort study of passive smoking and fier, rather than merely testing whether the overall asthma in children, the effect estimate for odds ratio is different €тот the null value of 1.0. passive smoking might be different in different age The power of the test for interaction therefore groups, or in males and females. The clearest depends on the numbers of cases and controls in example of interaction is when а factor is actually specific strata (of the effect modifier) rather than the hazardous in one group and protective in another overall numbers of cases and controls. For exam- group. More generally, the risk might be elevated ple, Smith and Day (1984) give an example of a in both groups, but the strength of the effect may case—control study that would have to be five times vary. A typical example of effect modification in larger to detect a difference between odds ratios of studies using biomarkers is the estimate of the risk 1.0 and 2.0 in the two different strata of an effect of disease due to an external agent in subgroups of modifier than it would have to be to detect an the population with a different genetic suscepti- overall odds ratio of 2.0 (ignoring the effect modi- bility marker, such as the polymorphism for an en- fier). In general, when considering possible inter- zyme implicated in the activation or detoxification actions, the size of the study needs to be at least of the agent (see Landi & Caporasi, this volume)' four times larger than when interaction is not con- In this situation, effect modification should be in- sidered (Smith & Day, 1984). Thus, in a study interpreted with considerable care, since the pres- volving markers of genetic susceptibility, the gain ence of statistical interaction may depend on the. in statistical power from considering such markers statistical methods used. In fact, all secondary risk (thereby yielding higher relative risks in some factors modify either the rate ratio or the rate dif- strata) may be offset by the decrease in statistical ference, as uniformity over one measure implies power from the need to consider interactions. non-uniformity over the other. If the assessment of However, if the exposure of interest is independent the joint effects of two factors is a fundamental from the genetic factor under study, case—case goal of the study, this can be done by саlculating comparisons can be used to study interactions stratum-specific effect estimates. It is less clear how with greater statistical power (Piegorsch et al., to proceed if statistical interaction is occurring, but 1984). assessment of joint effects is not an analytical goal. Some authors (e.g. Klеinbaum et al., 1982) argue Conclusions that it is not appropriate in this situation to calcu- In some instances, the increasing use of biomark- late an overall estimate of effect summarized across ers in epidemiological studies represents a major levels of the effect modifier. However, it is com- improvement in the discipline during the last years mon to ignore this stipulation if the difference in (Vineis, 1992). In many cases, biomarkers have effect estimates is not too great (Pearce, 1989). In helped to improve our knowledge of causes and fact, valid analytical methods (e,g, standardized mechanisms of both disease etiology and preven- rate ratios) have been specifically developed for tion. In other cases, however, it is unclear whether this situation (Rothman, 1986). they represent an improvement on traditional epi- Biomarkers may provide better opportunities demiological methods. Epidemiological studies for assessing interactions between two or more based on biomarkers are usually more complex genetic and/or environmental factors. In particu- than traditional epidemiological studies, because lar, genetic susceptibility genes should produce a information is available on a larger number of vari- higher disease risk for exposed susceptible groups ables whose biological meaning is often poorly than for non-susceptible and non-exposed groups known. The methodological considerations in- (Boffetta, 1995). volved in classical epidemiological studies on However, a major problem of testing for inter- issues such as measurement of disease, measure- action, e.g. in studies involving markers of genetic ment of exposure, selection bias, confounding, susceptibility, is that it usually requires a substan- precision and interaction also apply to biomarker- tial increase in study size. For example, in a case— based studies, and in most cases the methodologi- control study, testing for interaction involves corn- cal problems of this type of study do not require

55 Application of Biomarkers in Cancer Epidemiology

solutions different from those used in classical Hertzberg, V.S. & Russek-Cohen, E. (1993) Statistical studies. In some cases, however, the use of bio- methods in molecular epidemiology. In: Schulte P. & Perera, F.P, eds, Molecular markers may pose specific problems, which have Eрidemiology: Principles and Practices. New York, Academic Press, pp 199-216 to be addressed within the general framework of modern epidemiological methods. Hirsch, B., McCue, M. & Cervenka, J. (1984) Characterization of the distribution of sister chroinatid exchange frequencies: implications for research design. Acknowledgements Hum. Genet., 65, 280-286 N.P. is funded by a Professorial Research Fellowship and the Wellington Asthma Research Group is sup- Hulka, B.S. (1990а) Overview of biological markers. In: ported by a Programme Grant from the HeaLth luiRa, B.S., Wilcosky T.C. & Griffith, J.D., eds, Biological Research Coundl of New Zealand. Markers in Epidemiology, New York, Oxford University Press, pp. 315 References Hulka, B.S. (1990b) Methodologic issues irr molecular Armstrong, B.K., White, E. & Saracci, R. (1992) Principles epidemiology. In: luIRa, S.S., Wilcosky, T.C. & Griffith, J.D., cdi, Biological Markers in Epidemiology, New York, о f Exposure Measurement in Epidemiology. New York, Oxford University Press Oxford University Press, pp. 214-226 Boffetta, P. (1995) Sources of bias, effect of confounding Hulka, B.S. & Margolin, B.Н. (1992) Methodological is- in the application of biomarkers to epidemiological stud- sues in epidemiologic studies using biological markers. ies. Toxico!. Lett., 77, 235-238 Am. J. EpidemioL, 135, 200-209 Checkoway, H., Pearce, N. & Crawford-Brown, D. (1989) Johnson, E.s., Parson, W., Weinberg, C.R., Shore, D.L., Research Methods in Occupational Epidemiology. New York, Mathews, J., Patterson, D.G. & Needham, L.L. (1992) Oxford University Press Current serum levels of 2,3, 7,8-Теtгасhloгodibепzо-р- dioxin in phenoxy herbicide applicators and characteri- Copeland, К.T., Checkoway, H., McMichael, A.J. & zation of historical levels. J. Natl. Cancer inst., 84, Holbrook, R.H. (1977) Bias due to misclassification in the 1648-1653 estimation of relative risk. Am. J. EpidemioL, 105, 488-495 Kleinbaum, D.G., Kupper, L.L. & Morgenstern, H. (1982) Epidemiologic Research. Principles and Quantitative Methods, Fingerhut, M.A., Halperin, W.E., Marlow, D.A., Piacitelli, Belmont, CA, Lifetime Learning Publications L.A., Honchar, P.А., Sweeney, M.H., Greife, A.L., Dill, P.A., Steenland, K. & Suruda, .J. (1991) Cancer mortality tri McMichael, A.J. (1994) Invited commentary-'Molecular А epidemiology': new pathway or new travelling compan- workers exposed to, 2,3, 7,8-tеtгасhlогоdibеnгор-dioхјn. New Rugi. J. Mcd., 324, 212-218 ion? Am. J. EpidemioL, 140, 1-11 Fotastiere, F., Agabiti, N., Del1'orco, V., Pistelli, R., Corbo, McMichael, A.J. (1995) Re: invited commentary G.M., Brancato, G., Pacifici, R., Zuccaro, E & Perucci, Molecular epidemiology': new pathway or new travel- C.A. (1993) Questionnaire data as predictors of urinary ling companion? Am. J. EpidemioL, 142, 225 cotinine levels among eosmoking adolescents. Arch. Miettinen, O.S. (1974) Confounding and effect modifi- Environ. Health, 48, 230-23 4 cation. Am. J. Epidemiol., 100, 350-353 Greenland, S. (1980) The effect of misclassification in the Miettinen, 0.5. (1976) Estimability and estimation in presence of covariates. Am. J. Epidemiol., 112, 564-569 case-referent studies. Am. J. Epidemiol., 103, 226-235 Greenland, S. (1987) Interpretation and choice of effect Miettlnen, O.S. (1985) Theoretical Epidemiology, New measures in epidemiologlc analyses. Am. J. Epidemiol., York, Wiley and Sons 125, 761-768 Munoz, N., Bosch, F.X., Sari) ose, S., Izarzugaza, I., Gi1i, M., Greenland, S. (1988) Statistical uncertainty due to mis- Vi ladin, P., Navarro, C., Martos, C., Ascunce, N., Gonzalez, classification: implications for validation substudies. J. L.C., Kaldor J.M., Guerrero, E., Lotiriez, A. Santamaria, M., Clin. Epidemiol., 41, 1167-1174 Alsonso de Ruiz, P., Aristizabal, N. & Shah, K. (1992) The Greenland, 5. & Robins, J.M. (1985) Confounding and causal link between human papillomavirus and invasive misclassification. Am. J. Epidemiol., 122, 495-506 cervical cancer: a population-based case-control study Iii Colombia and Spain. Ivt. J. Cancer, 52, 743-749 Greenland, S. & Robins, J.M. (1986) Ideпtifiability, ек changeability and epidemiological confounding. Int. J. Pearce, N. (1992) Methodological problems of time-re- EpidemioL, 15, 4128 lated variables in occupational cohort studies. Rev. ЕрiдетгоI. 5апté ЛиЫ., 40, S4З-854

56 General issues of study design and analysis

Pearce, N. (1993) What does the odds ratio estimate in a Schatzlo, A., Freedman, L. & Schi€famn, М. (1993) An case-control study. lut. J. EpidernioL, 22, 1, 1189-1192 epidemiologic perspective on biomarkers. J. lut. Med., 233, 75-79 pearce, N.E. (1989) Analytic imphcations of epidemiological concepts of interaction. lot. J. EpidemioL, 18, Schwartz, G.G. (1990) Chromosome aberrations. In: 976-980 Hulka, B.S., Wilcosky, T.C. & Griffith, J.D., eds, Biological Markers in Epidemiology, New York, Oxford University Pearce, N.E. Checkoway, H. & Shy, C.M. (1986) Time-re- Press. pp. 146-172 lated factors as potential confounders arid effect modifiers in studies based on an occupational cohort. Sсaпд. Siemiatycki, J., Wacholder, Ѕ. Dewar, R., Wald, L., Bégin, J. Work Environ. Health, 12, 97--107 D., Richardson, L., Rosenman, K. & Gérin, M. (1988) Smoking and degree of occupational exposure: are inter- Pearce, N., san)ose, S., Boffetta, P., Kogevinas, M., Saracci, nal analyses in cohort studies likely to be confounded by R. & Savitz, D. (1995) Limitations of biomarkers of ex- smoking status? Am. J. Ind. Med., 13, 59-69 posure in cancer epidemiology. Epidemiology, 6, 190-194 Smith, P.G. & Day, N.E. (1984) The design of case control Piegorsch, W.W., Weinberg, C.R. &Taylor, J.A. (1994) Non- studies: the influence of confounding and interaction ef- hierarchical designs for assessing susceptibility in popula- fects. lit. J. Epidemiol., 13, 356-365 tion-based case-control studies, 5tat. Med., 13, 153-162 Steinmetz, K.A. & Potter, J.D. (1991) Vegetables, fruit and Robins, J. ., Gail, .H. & Lubin, J. . (1986) More on bi- М М Н cancer: I. epidemiology. Cancer Causes Control, 2, 325-357 ased selection of controls for case-control analyses of cohort studies. Biometrics, 42, 1293-1299 Vineis, P. (1992) Uses of biochemical and biological markers in occupational epidemiology. Rev. Epidem. Sante Rothman, K.J. (1986) Modern Epidemiology, Boston, Little, PuhL, 40, 563-569 Brown Whorton, E.B. (1985) Some experimental design and Rothman, N. (1995) Genetic susceptibility biomarkers in analysis considerations for cytogenetics studies. Environ studies of occupational and environmental cancer: Mutagen; 4 (Suppl), 9-15 methodologic issues. Toxicol. Left., 77, 221-225 Wilcosky, T.C. & Griffith, J.D. (1990) Applications of bi- Rothman, N., Stewart, W.F., Rabkin, C.S. & Schulte, P.A. ological markers. In: Hulka, B.S., Wilcosky, T.C. & (1995) The development, validation, and application of Griffith, J.D., eds, Biological Markers in Epidemiology, New biomarkers for early biologic effects. In: Mendelsohn, York, Oxford University Press, pp. 16-27 M.L., Peeters, J.P. & Normandy, M.J., eds, Biomarkers and Occupational Health: Progress and Perspectives, Willett, W. (1990) Nutritional Epidemiology. New York, Washington, D.C., Joseph Henry Press, pp. 109-115 Oxford University Press

5aгaсci, R. (1984) Assessing exposure of individuals in the identification of disease determinants. in, Berlin, A., Draper, M., Hemminki, K. & Vainio, Н., eds, Monitoring Human Exposure to Сarciпogeпic and Mutagenic Agents, (IARC Scientific Publications No. 59), Lyon, Inter- Correspon ling author national Agency for Research on Cancer N: Pearce Saracci, R., Kogevinas, M., Bertazzi, P., et al. (1991) Wellington Asthma Research Group, Cancer mortality in an international cohort of workers Wellington School of Medicine; PO Box 7343, . exposed to chiorophenoxy herbicides and chlorophe- Wellington, New Zealand no1s. Lancet, 338,. 1027-1032

57 Аpplicatioп of Biomarkars in Cancer Ёpkdеmиolоgy Toniolo, Р., Boffetta, h, 5hukeг, Q.E.G., RогКпеп, N. lLjlka B. апд Pearme, N,, ada 1АRC Scientific РuЫXahmns Ni. 142 Iniermtional Адепсц for Research cri Cancer, Lyon, 1997 Sources of variation in biomarkers

P. Vïneis

Epidemiology is interested in variation. The goal of epidemiological research is to infer cause-effect relationships by observing whether the occurrence of disease varies according to relevant exposures. Epidemiology is usually interested in intergroup' variation (e.g. between those who are exposed and those who are unexposed to the factor of interest), while `intragroup' variation is a source of noise.Therefore, the study design aims to increase intergroup variability, to limit intragroup variability and to reduce error by ensuring validity and reliability of measurements. The goal of this chapter is to summarize simple ways to recognize the main sources of measurement error and to estimate the extent and the impact of such error,

In fact, the impact of error in the measurement of type of assay, and the data shown in Table 1 are biomarkers within epidemiological research may measurements. be dramatic. For example, it has been claimed that in the measurement of oxidative damage to DNA, Validity and reliability routine phenol-based DNA purification procedures Validity is defined as the (relative)lack of systematic can increase 8-hydroxydeoxyguanosine levels 20- measurement error when comparing the actual fold in samples that are exposed to air following observation with a standard—a reference method removal of the phenol. Exposure to air alone that represents the 'truth'. Such 'truth' is in fact results in a fourfold increase compared with DNA the abstract concept we are interested in, e.g. 'can- samples that have been soiubilized in buffers cer' as defined through an histological characteri- purged with nitrogen (Wiseman et al., 1995). It is zation. While validity entails a 'standard', reliabil- obvious that such gross contamination would seri- ity concerns the extent to which an experiment or ously hamper an epidemiological study. The con- any measuring procedure yields the saine results sequences would be even greater if subsets (batches) on repeated trials (Carmins & Zeller, 1979). By coming from different subgroups in the study 'the same results' we do not mean an absolute cor- population (e.g. exposed versus unexposed) under- respondence, but a relative concept, i.e. a tendency go different technical procedures with different towards consistency found in repeated measure- levels of error. ments of the same phenomenon. Reliability is rel- An example of 9nterlaboratory variability, con- ative to the type and purpose of the measurement: cerning the measurement of DNA adducts in the for some purposes we may accept a level of unreli- same samples, is shown in Table 1. To achieve an accurate estimate of the association between any marker and disease, in epidemi- Table 1. 32Р-роstlabelling of aromatic ology we need reliable and valid measurements of DNA adducts (per 108 bases± 5i)) exposure, covariates (potential confounders and from foundry workers effect modifiers), and outcomes. Causal inference is impossible in the absence of such requirements. ri Т1.ТЁ' I will distinguish, in the following, between marker (any variable that can be measured and is infor- 2 3 . mative for the purposes of the study), assay (a spe- Control (n = 6.) 3.1 ± 1.7 — 1.7 ±0.7 . cific laboratory test which aims at measuring that Exposed workers 26 $ 43 12 t 10 9.2 + 23 marker) and measurement (the concrete act of (л = 35) measuring the value of a marker in an individual using a specific assay). For example, PAH-DNA Dasa from 5avela of al., 1989. adducts are a type of marker, ЭгР-ilostlabelling is a

59 Application ы Biomaгkвгs in Cancer Epidemiology

ability that is unacceptable for other purposes. In measurement error (normally distributed with addition to being reliable, the marker must be mean = 0 and variance = 2lk) (Tatou et aL, 1994). valid, i.e. it must provide au accurate representa- The normality of distribution, assumed in the till of some abstract concept. model, must be verified. In fact, many biomarkers Reliability focuses on a particular property of have distributions that are faz from being normal. empirical markers—the extent to which they pro- Normalization can be achieved through an appro- vide consistent results аcrоss repeated measure- priate transformation, e.g. log transformation. ments. Validity concerns the crucial relationship The model is based on a linear (additive) assump- between concept and marker. Both are a matter of tion, which implies that measurement errors are degree, riot an all or none' property. Validity and independent of average measurements. Such an reliability are independent: a measurement may be assumption must be verified case by case, e.g. by perfectly reliable (reproducible in different labora- checking whether errors are correlated with the tories and repeatable at different times) but con- mean. sistently wrong, i.e. far away from the true value. For example, a gun may be completely reliable if Impact of measurement error: random aid all the shots take place within a small area, but seri- systematic error ously biased if this area is far away from the target; Errors of marker measurement may have different conversely the gun is unbiased but unreliable if the impacts depending on the error distribution. If the shots have an average distribution around the epidemiological study has been conducted blindly, centre of the target but are dispersed in a large area. i.e. laboratory analyses have been done with no We are interested in both validity and reliability; knowledge of the exposed/unexposed or diseased! however, since validity is often not measurable, healthy status of the subjects, we expect that the reliability is sometimes used (incorrectly) as a sur- measurement error will be evenly distributed across rogate. strata of exposure or disease. Table 2 shows an An aspect that is clearly relevant to the discus- examрlе. Suppose we recruit a cohort of smoking sion of measurement error is timing. Any inference fathers, in order to investigate respiratory disease about the meaning of biomarker measures should among their children. We determine urinary nico- be strictly time-specific, since time influences the tine and cotinine only at recruitment (Т0 ), and not results in several different ways. at other time points during fo11ow-up. However, a The major components of biomarker variability number of fathers change their smoking habits in that affect the design of epidemiological studies are the course of time: at time Т1, 250 smokers have variability between subjects (intersubject), vari- ceased, while 150 non-smokers have started smok- ability within subjects (intrasubject) and variability ing, so that at time ТZ, there are 1100 smokers and due to measurement errors. The impact of these 850 non-smokers. Therefore, we have 'misclassifi- three categories of variability on the biomarker cation' of exposure, based on our spot measure- response can be represented by a linear model of ment of urinary nicotine. Misclassification the foIlowing form (Taioli et a'., 1994): depends, in this case, on the study design and the characteristics of the population, not on technical Y~1k =u + а1 + Ы + егlk limitations of the test (assay). The effect of misclassification is such that a 'true' ratio of 2.0 where Уй1k is the marker response for subject i at between the prevalence of respiratory symptoms time j and replicate measurement k; u is the true among children born from smokers and the preva- population mean response; а! is the offset in mean lence among children born from non-smokers response for subject i (assumed to be normally dis- becomes an 'observed' ratio of 1.48 (Table 2). This tributed with mean = 0 and variance = s?; the vari- underestimation of the ratio is due to a 'blurring' ance represents the extent of intersubject variabil- of the relationship between exposure and disease, ity), b1 is the offset in response at Lime j (assumed which occurs when misclassification of exposure, to be normally distributed with mean = 0 and as in this case, is evenly distributed according to variance = s2;; this variance represents the extent the diseased/healthy status (i.e. is not influenced of intrasubject variability), and е~1k is the assay by the outcome).

60 Sources of variation iп biomarkers

Both underestimation and overestimation of the association of interest may occur when nus- classification is not evenly distributed across the study variables. In the example, we may have a more general distortion of the etiological relationship, and not only a blurring of the association, if a 1 T2 classification of exposure depends on the outcome г т (diseased/healthy status). Blurring is 'bias toward Smokers 1200 250 quit 1100 the null' while distortion as a consequence of Non-smokers 750 150 start 850 uneven distribution of misclassification can be in The true prevalence of respiratory symptoms at rire Т2 either direction, both towards and away from the among the children is 10% among children of smokers . null hypothesis. (22012200 children) and 5% ameng children of non-smokers A model of study design used in large cohort (8511700 children) (prevalence, ratio = 2.0}. studies is the nested case—control study, which EaCrnated prevalence of respiratory symptoms among' those consists of identifying, in the course of the follow- classified as smokers at To: 10% (bue prevalence in smokers) up period, all the subjects who develop the disease x 950 + 5% (prevalence in non.smokors) x 250 = 107.511200 and drawing a sample of the healthy subjects (see = 8.9%. Pearce & Boffetta, this volume). The advantage of Estimated prevalence of respiratory symptoms amопg those such a design is that with less effort we may have classified as non-smokers at 7: 10 % (true prevalence in smokers) x 150 + 5% (prevalence in non-smokers) х 600 = estimates that are as stable as those that would be 451750-6%. obtained by analysing the whole cohort. This is Estimated prevalesce ratio = 8.9%16% = 1.48. ' particularly useful when measuring biochemical/molecular markers. In our example, suppose we collect urine samples from all 1950 fathers and knowledge of the disease status by the researcher is store them in a refrigerator. When a case of respi- related to degradation of analites when biological ratory disease is diagnosed in a child, we take the samples are stored for a long time. If the samples father's urine sample and, say, the urine samples of front the cases affected by the disease of interest two fathers whose children are healthy at that time and those from controls (within a cohort design) (cases and controls are usuallymatched for the are analysed at different times, bias can arise from time-lag between sample collection and sample differential degradation in the two series. For analysis, to take analite degradation into account; example, the researcher may decide (incorrectly) see below). In such a situation, bias due to uneven to analyse samples from the cases as soon as these distribution of error is a potential issue. For екат arise in the cohort, while controls are analysed at рlе, samples from fathers of diseased children the end of the study. Since the levels of, say, vita- might be analysed in a more accurate way than min C decrease rapidly with time, serious bias may samples from fathers of healthy children. Scientific arise from the differential timing of measurement curiosity might lead one to measure additional bio- in the two series. For this reason, biochemical markers among the fathers of diseased children analyses should be made after matching of cases (including, for example, DNA adducts in exfoliated and controls for time since sample collection. bladder cells or markers of oxidative damage to DNA), on the assumption that these subjects are hources of variability: intersubject, intrasubject, scientifically more interesting. This is a rather com- laboratory mon attitude, which may, however, lead to serious Variation in biomarkers includes interindividual bias. Bias can go both ways. In our example, if (intersubject) variation, intrasubject variation (i.e. oxidative damage to DNA is measured more accu- variation in the marker over a particular time rately among case subjects (i.e. if samples are solu- period), sampling variation (i.e. variation depending bilized in buffers purged with nitrogen) than on the framework of biological sample collection) among controls, we might find higher levels in the and laboratory variation. Sometimes intra-individ- latter. ual and/or sampling variation are so large that the A more realistic example of bias dependent on laboratory measurement variation makes a marginal

61 Application of Biornarkers in Cancer Epidemiology

variable in different points of the colon mucosa. Therefore, not only is intrasubject variation over time important, due to variable exposure to agents that induce cell proliferation, but measurements are also are strongly influenced by how and where the mucosa is sampled. For example, a study (Lyles Weight Age et al., 1994) estimated that 20% of the variability of the rectal mucosa Lutein -.0.06 (0.63) 008 (0.48) proliferation index (measured by Criptoxanthine —0.22 (0.05) —0.04 (0.69) nuclear antigen immunohistochemistry) is due to Lycopene —0.25 (0.03) 0.04 (0.73) subject, 30% is due to the biopsy within the sub- wCarotene —0.12 (0.29) --4.03 (0.78) ject, and 50% is due to crypts within a biopsy, In p-Carotene —0.27 (0.02) 0.11 (0.32) other words, as much as 80% of variation is related Tocopherol —0.12 (0.30) 0.08 (0.49) to sampling. Laboratory measurements can have many Measurements were made by Giuseppe Maiani, 1stituto sources, in particular two general classes of labora- Nazionale dalla Nuuzione, Roma. values are brackets (n = 72). tory errors: those that occur between analytical Р гп batches and those that occur within batches. An example of a study that was designed in order to assess the different sources of laboratory variation contribution to overall variation. A particular exam- is reported by Taioli et aL (1994), using the model ple of intrasubject variation is associated with error described above. Iп one experiment, they drew due to handling, processing and storing of specimens; blood from five subjects three times in three dif- such variability can be measured only if repeated ferent weeks (n = 5, k 3,] = 3) in order to measure samples from the same individual are collected. DNA—protein cross-links. Table 5 shows the results, Intersubject variability in marker response may which indicate that variation between batches is derive from factors such as ethnic group, gender, quite important and greater than variation diet or other characteristics. For example, Table 3 between subjects. An interaction between inter- shows interindividual variation concerning the subject variation and batch variation is suggested. levels of serum vitamins according to weight and Methodological issues should be discussed as age. We might have expected that vitamin levels much as possible within biomarker categories, due varied according to diet-related characteristics such to the specificities of each category. Table 6 shows as weight (thinner subjects are expected to have how methodological data can be organized accord- healthier diets and higher vitamin consumption; ing to biomarker type. Intra-individual and sam- P. Vineis, unpublished data). In this particular pling variations are considered, due to the extent example, age is not associated with vitamin levels, of their influence on actual measurements for although age always has to be considered as an most markers. important source of intersubject variation. Similarly, the marker response may vary within Measurement of variation the same subject over time due to changes in diet, Reliability health status, variation in exposure to the compound The extent of variability in measurements can itself of interest (for dietary items, season is an impor- be measured in several ways. Let us distinguish be- tant variable), and variation in exposure to other tween continuous measurements and categorical compounds that influence the marker response. measurements. A general measure of the extent of Table 4 shows intra-individual seasonal variation variation for continuous measurements is the соеf- in levels of micronutrients, indicating that temp- ficieпt of variation (CV = standard deviation! erature arid rainfall (as markers of season) had an mean, expressed as a percentage). A more useful important effect on most measurements. measure is the ratio between CV and CVw; CV,, Biological sampling variation is related to drcum- measures the extent of laboratory variation within stances of biological sample collection. For ехаm- the same sample in the same assay, CVb measures рlе, hyperproliferation of colonic cells is extremely

62 Sources of variation iп biomarkers the between-subject variation, and the CVЪ/CV, ratio indicates the extent of the between-subject variation relative to the laboratory error. Large degrees of laboratory enor can be tolerated if between-person differences in the parameter to be measured are large. Temperature Rainfall A frequently used measure of reliability for continuous measurements is the intraclass correla- Lutein —0.89 0.74' tion coefficient, i.e. the between-person variance di- ,x-Carotene 0.54 —0.17. a-Carotene 0.56. . —0.29 . vided by the total (between- plus within-subject) Retinol —0.69 variance. The intraclass coefficient is equal to 1.0 if a-Тocopherol —0.60 0.47 there 1s exact agreement between the two measures on each subject (thus differing from the Pearson From Colley et at., 1995: changes in seasonal levels were compared for each 2-month period tar 21 individuals over a correlation соеfficient which takes the value 1.0 1-year period: Pearson correlation cootticionis. when one measure is a linear combination of the aA 5 0.02. other, not only when the two agree exactly). A ЬP Ç 0.05. coefficient of 1.0 occurs when within-subject variation is null, i.e. laboratory measurements are totally reliable. The intraclass correlation coefficient can then be used to correct measures of association r = 0.12) have a dramatic effect on observed rela- (e.g. relative risks, RRs) in order to allow for labo- tive risks, which are all around 1.1. The observed ratory error. Table 7 shows an example where inti- RR was obtained using the formula RR = exp (In aclass correlation coefficients were computed for RRt к r), where RR, is the true relative risk. different laboratories which measured estrone lev- The intraclass correlation coefficient can be els, and where these coefficients were used to esti- used to estimate the extent of between-subject mate the effect of laboratory error on the observed variability in relation to total variability. The latter relative risks (for true relative risks of 1.5, 2.0 aid includes variation due to different sources (repro- 2.5). In fact, labora-tory error attenuates the causal ducibility, repeatability and sampling variation). relationship between the parameter of interest To measure reproducibility, i.e. the ability of two (estrone) and disease (breast cancer), as a conse- laboratories to agree when measuring the same quence of random misclassification. De-attenuation analite in the same sample, the mean difference of the relationship (i.e. a correction for measure- between observers (and the corresponding confi- ment error) can be achieved by applying the intra- dence interval) has been proposed (Brennan & class correlation coefficient. In Table 7, a high intra- 5ilman, 1992). Let us consider the example in class correlation coefficient (Lab. 4, r = 0.90) corre- Table 8: two pathologists interpeted the slides of 40 sponds to good agreement between true and subjects with bladder сапсеr, in order to quantify observed relative risks, while low coefficients (Lab. 1, р53 overexpression as measured by immunohlsto-

Table 5. Analysis of variance in a reliability study: DNA--protein cross-link

• г . Week 0.05 13.7 (27) <001 Between sub]act 0.02 3.4 (4, 7) . -0 05 Week - subject 0.01 2.3 (7, 40) 6.045 Error 0.03

From: Taioli etal., 1984. dl degrees of troodirn;

63 Application of Biomarkers in Cancer Epidemiology

' ь , 1F. rn 1 *1i *t W 1 ј г гТ гј Е ф гвт 1 Х 'рј

Biomârker category jntra-individual variation Biological sampling variation Internal dose (blood) Hormones .. Yes (diurnal variation) No Water-soluble nutrients Yes (short half-life) No 0rganochlorine No (long half-life) No Biologically effective dose Peripheral white blood cells Yes (half-life: weeks to months) No Exfoliated urathelial cells DNA adducts Yes (half-life: months) Yes Early biological effects Lymphocyte metaphase chromosome aberrations More or less stable ? Somatic cell mutations glycophorin A Probably low No (?) Intermediate markers Cervical dysplasia Yes Yes Colonic hyperproliferation Yes Yes Genetic susceptIbiIity Genotype assay ` No No Non inducible phenotype No No Inducible phenotype Yes No Tumour markers Yes Yes

chemistry. Table 8 gives the percentage of positive- pressor gene) is considered to be a feature of malig- staining cells according to each pathologist and nancy. Data are expressed as proportions of cells the difference between the two observers for each which stain positive, i.e. a continuous variable. patient. The proposed measure of interobserver Therefore, a simple measure of agreement between agreement is the mean difference (in this case 0.9) readers is the correlation coefficient (Table 8). All and the corresponding 95% confidence interval, coefficients are strongly statistically significant which is between —20.9 and +22.7 (I have used 2 (P. 0.0001) and range from 0.6 to 0.9. However, for as the probability point of the t-distribution, clinical purposes p53 expression is generally used although the number of observations is only 37; in in a categorical (binary) way, as in Table 9. Some theory this choice would be appropriate only with researchers use 20% as a cut-off point to suggest a n > 50). What we can conclude is that there is good more aggressive clinical strategy. If we compare the average agreement between the two pathologists first and the second pathologist, we see that one (mean difference = 0.9), but a large confidence counts 11137 positive cases, and the other 16137 interval indicating large overall variability. Iii the (three have been considered as not interpretable case of immunohistochemistrу a very important for technical problems). The discrepancy between source of variation is sampling, since pathologists the two proportions is more informative than the usually do not read the saine fields, i.e. the same simple correlation coefficient, since it suggests that cells, at microscope. at least five patients would receive different clinical Let us now consider categorical (binary) data. suggestions on the basis of the two readings. Four different pathologists have read specimens of However, if we look at the detail of the table, we bladder cancer, specially stained to reveal the con- see that the two pathologists agree about the posi- centration of the p53 protein (an immunohisto- tivity (~20%) in only 10 cases; in the other 20 they chemical technique). Accumulation of the p53 agree on the lack of positivity (<20% of stained cells), protein (i.e. overеxprеssioп of the p53 tumour sup- while for seven patients there is disagreement.

64 Sources of variation b biomarkers

Simply to state that agreement concerns 30/37 tioiisly since there are some methodological pit- readings is insufficient, because agreement may falls; for example the value of kappa strictly arise as a consequence of chance. Let us suppose that depends on the prevalence of the condition which the two pathologists completely ignare the process is being studied; with a high underlying prevalence underlying their observations, or even throw dice we expect a high level of agreement (Brennan & to guess the p53 positivity. Nevertheless, they will Silman, 1992), agree in a number of cases. Let us imagine that the In addition to reproducibility, i.e. agreement two pathologists have a propensity to diagnose between readers on the same set of observations, 11/37 arid 16/37 positive cases, but the allocation with similar techniques we can measure repeata- of subjects is left to chance alone, and the two bility, i.e. agreement within the same observer at readings are completely independent. Then we will different times (repeat observations). have the results shown in Table 10, in which a is 26 x 21/37 = 15, and so on (multiplication of mar- Validity ginais is an expression of statistical independence Until now we have considered reliability as a prop- of the two readers). The level of agreement due to erty of the assay in the hands of different readers chance alone is 20/37. Therefore, agreement (reproducibility) or at repeat measurements beyond chance is (30 — 20)137 = 10/37. However, (repeatability). Let us now consider validity of the total potential agreement between the two assessment, i.e. correspondence with a standard. It readers cannot be 100%, but only (1— 20/37)%; i.e. is essential to bear in mind that two readers may to be fair we must subtract chance agreement from show very high levels of agreement, as measured 100% to have an estimate of total attainable agree- for example by the Pearson correlation coefficient ment. The measure (10/37)1(1 — 20/37) = 0.59 is (i.e. r = 0.9), even if the first consistently records caled the kappa index and measures agreement twice the value recorded by the second observer. beyond chance. In this case, the two pathologists Alternatively (e.g. when using the intraclass cone- agreed in 59% more cases than they would have lаtiоп coefficient, ICC), two readers could show agreed on the basis of chance alone, i.e. by throw- high levels of agreement (e.g. ICC = 0.9) but poor ing dice. There are other measures of agreement validity if the same errors repeat themselves for beyond chance, and kappa has to be used cau- both readers.

True relative risks . RR= 15 ARt = 2.ü RRt = 25 Laboratory .. r Observed relative risks Laboratory 1 0 12 1.1 1.1 1.1 Laboratory 2 Analysis 1 0.82 1.4 1.8 21 Analysis 2 0.53 1.2 1.4 1.6 Laboratory 3 Analysis 1 0.57 1 3 1.5 ... 1.7 Analysis 2 0.65 ` 1.3 1.6 1:,8

Laboratory 4 6.90 1.4 . f .9 . 2.3

Observed RR= exp (In ,RRr x r}. From Hankinson etal. ,1994.

65 Application of Biomarkers in Cancer Epidemiology

In the case of immunohistochemistry we might decide that one pathologist is the standard, but this would not be an ideal Specimen Pathologist choice unless we are organizing a training course in which younger ber First Seco d Difference пит п pathologists аre instructed to rеад 1 16.00 21.00 -4.00 specimens. According to the del- 2 83.00 60.00 23.00 nition of validity, we are inter- a -9.so ested in the correspondence of 4 19.40 19.4~ 0.00 5 9640 72.80 23.60 the measurement with a concep- 6 4.90 4.50 0.40 tuai entity, i.e. accumulation of 7 . - 3.20 - the p53 protein as a consequence 8 21.00 16.20 4.80 of gene mutation (in fact, with- 9 11.40 5.40 6.00 out a mutation the protein has a ii 13.00 26.О0 -13.00 very short half-life and rapidly i1 9.40 6.Dо 3.40 12 4.60 1.70 2.90 disappears from the cells). Tables 13 9.70. 1.80 7.90 11 and 12 show data on the cor- 14 20.80 32.10 -11.30 respondence between linmuno- 15 '43.50 49.00 -5.50 histochemistry and p53 muta- 16 5.60 12.50 -6.90 tuons. Sensitivity of immunohis- 40.t i7 б X8.00 12-10 tochemistry is estimated as 85%, 1В 11.50 7.00 4.5U 19 14.60 13.50 1.10 i.e. false negatives are 15% of all 20. - 27.00 - samples containing mutations; 21 7.50 2.80 4.70 specificity is estimated as 71% , 22 . 7.40 21.00 -13.60 i.e. 29% of samples not contain- 16. 440 11.90 . . зо ing mutations are falsely positive 24 21.00 46:50 -25.50 at immunohustосhетisuу. A corn- 28 29.0 34.00 -4.20 bined estimate of sensitivity and 27 ' 82.00' 69.50 12.50 specificity is the area under the 2e 75.00. 75.20 -д.20 receiver-operating curve (ROC), 29 700 2400 -17.00 i.e. a curve which represents зо 7~2о 8.40 -1.20 graphically the relationship Эi 5.70 о.5о 5.2о between sensitivity and specificity. s2 15.~ 6,1 ~.sa а а In the example shown in Fig. 1, зз о.5о 100 -о.5о 34 1800: 7.50 10.50 the area under the ROC curve is 35 9.80 1100 -1.20 90.3% (Cordon-Carlo etal., 1994). 36 .4.20 10.50 -6.30 It is usually believed (Fletcher et 37 15,20 . 16.00 -2.80 al., 1988) that sensitivity and 38 . 9.30 22.00 -12,70 _ _ _ specificity indicate properties of Э9 40 - 52.60 29.30 23.30 a test irrespectively of the fre- . quency of the condition to be Difference parameters detected (however, this is an ri Mean sD Minimum Maximum assumption that requires to be 37b 0.9 ' 10.907 -25.5 23 6 verified). In the example 01 95% range for agreement: mean ±2 х 5D(diff)=09±2 x 109= -2091022.7 Tables 11 and 12, the proportion of samples showing a mutation аAercentagе of pleuve cells as measured by two different palhologeffs in 1hе saте samples. ' 1s high (32/73 = 44%); it would bThree missing чaiuвs: Data ktndly provided by Renato Coda n= 40). . be much lower, for example, in patients with benign bladder

66 Sources of variaton in biomarkers

Table 10. Expected distribution of observations if Pathologist 1 and Pathologist 2 of Table 9 agree by chance alone Pathologist 1

0 <20% г2О% Total Pathologist 2 ь• 0 O o 0 o <20% O 20 1 21 (3 NI) 20% 0 6 10 16 Total 0 26 11 37 Pathologist 3 o 0 a 0 rectly predicts mutations in 69% of the positive <20% 1 15 3 (5N1) cases. Let use suppose, however, that the pTeva- г2O% 2 6 8 lence of mutations is not 44% , but 4.4% (32/730). With the same sensitivity and specificity values Pathologist 4 0 0 0 0 (85% and 71%, respectively), we would have a pos- <20% 0 18 0 (7N1) itive predictive value of 11.8%, i.e. much lower. 20%% 1 7 7 The predictive value is a very useful measure, because it indicates how many true positive cases Pathologist 2 we will obtain within a population of subi ects who Ρ20% 0 <20% г test positive with the assay we are applying. Pathologist 3 However, we must bear in mind that the predictive 0 0 2 1 value is strongly influenced by the prevalence of <20% 0 20 .1 (5 NI) the condition: a very low predictive value may г20 % 0 3 8 simply indicate that we are studying a population in which very few subjects actually have the con- Pathologist 4 . o O 1 0 dition we want to identify. H2O% 0 2 2 (7N1) Tables 13 and 14 show another set of estimates J20% 0 0 7 of validity based on the clinical outcome. In this case, the aim is to understand how useful p53 Pathologist 3. immunohistochemistry may be in predicting the 0 <20% 20% risk of metastases. While in the previous example Pathologist 4 the conceptual entity to be predicted by irnmuno- 0 0 1 0 histochemistry was р53 gene mutation, here the <20% 1 21 3 (7 NI). conceptual entity is the aggressiveness of malig- г2О% O 3 6 nancy, as expressed by the occurrence of lymph NI, not interpretable node invasion in the follow-up of patients who were tested for p53 at diagnosis. In this case we are obviously interested in the positive predictive value which, for example, is 60% at stage 1. However, we also want to establish whether im conditions or in healthy subjects. A measure that munohistochemistry is worth measuring, i.e. is useful in predicting how many subjects (among whether it adds to what we already know from those testing positive) are really affected by the other clinical examinations. The best predictor condition we aim to detect 1s the positive predic- of outcome is usually the clinical stage, based on tive value. In the example among 39 patients test- the tumour size at diagnosis. We see that among ing positive at immunohistochemistry, 27 actually stage 1 tumours, seven out of 25 (28%) manifest have mutations, Le. irnmiinohistochemistry cor- lymph node metastases in the follow-up. The ques-

67 Application of 8iomarkers in Cancer Epidemiology

Tables 13 and 14 show that in stage 1, the measurement of p53 allows the probability of metastases to increase from 28% (a priori) to 63% (а posteriori), while in stages 2 and 3 the contribution of p53 measurement is totally insignificant.

p53 nuclear reactivity Laboratory drift, study design, quality controls (immunohistochemistry) When we organize and analyse an epidemiological study employing markers, we want to minimize p53 mutations Ыо total intragroup variability, in order to identify by SSGP — + ++ Total intergroup differences (e.g. between exposed and No mutation 29 7 5 41 unexposed or between diseased and healthy sub- All mutations 5 8 19 32 jects) if they exist. Total intragroup variation is Total 34 15 24 73 the weighted sum of imtersubject, intrasubject, sampling and laboratory variations, with weights Data from Esrig et al., 1993. that are inversely correlated to the numbers of sub- sensitivity of immunohislochemistry (+ and ++) = 27132 = 85% Specificity = 29141 = 71%. jects, measurements per subject and analytical Positive predictive value = (8+19)1(15+24) = 27!39 = 69%. replicates used in the study design, respectively. Obviously, if we do not have detailed information we cannot adjust for intragroup variation. This is the reason why in epidemiological studies till is: does p53 immunohistochemistry add employing biomarkers it is important to collect, any information to simple knowledge of the stage whenever possible, (1) repeat samples (day-to-day, at diagnosis? A way to answer this question is to month-to-month or year-to-year variation may estimate the a posteriori probability of metastases, be relevant depending on the marker); (2) poten- after measurement of p53, and to compare it to tially relevant information on subject characteris- the a priori probability (i.e. in the absence of p53 tics that may influence intersubject variation; measurement), which in this case is 28%. The (3) conditions under which samples have been a posteriori probability is computed using the collected and under which laboratory analyses relationship: have been conducted (batch, assay, specific procedures). a posteriori odds = a priori odds x likelihood ratio Concerning item (3), measurement variation may occur as a consequence of many different aspects which are related not only to the choice of the assay but also to:

• collection of the sample (how and when a blood sample was drawn; the type of test tube utilized; the amount of biological material col- Mutations lmmuvohistochemistry lected; for some measurements, whether the subject was fasting; avoidance of exposure to — +1++ Total light if we are interested in vitamin C); Yes 5 27 32 • processing of the sample (e.g. speed of cen- No 496 202 698 trifuging to separate different blood compo- Total 501 229 730 nents; use of a gradient to separate lymphocytes); • storage (in a simple refrigerator at —20°C; at Sensitivity = 27132 = 85%. Specificity = 4961698 71%. —70°C; in liquid nitrogen at —196°C; for how long); Positive predictve velue • laboratory analyses (interlaboratory variation; assay; technician performing the assay; batch; accidental contamination of the sample).

68 Sources of variation in biomarkers

1.0

0.9

0.8

0.7 p53 immunohistachemistry stage j, lymph node invasion Negative Positive Total 0.6 No 14 4 18 a, Yes 1 6 7 7 0.5 ÿ Total 15 10 25 C Ф ° 0A Data kindly provided by Professor A. Fontana. Predictive va!ue = 6110 = 60%. . 0.3 Likelihood raiid = (6/7)1(4/18) = 3.9 A prion probability = 7/25 28%, a priori odds = probablity/ 0.2 (1 — probаЬildу) = 39%. A gosteriâri odds = a priori odds X likelihood 0.1 ratio = 39%п х 3.8 = 1.5.

0.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 compounds that undergo substantial changes 1-specificity during the day, such as hormones, requires very accurate timing. Area under the curve =50.3% • subject—physiological compounds such as proteins, iron and cholesterol can be increased by 5-15% in the standing position in comparison Figure 1. Receiver operating curve (ROC) statistical analysis of the sensitivity and specificity of immunohistochemistry as with supine position. it relates to PCR-sSCP and sequencing results, represent- • Haemolysis may occur as a consequence of ing identification of p53 mutaiions. The area under the curve tube transport and manipulation. as a measure of diagnostic accuracy was estimated to • Storage conditions. be 90.3%. Redrawn from Cordon-Carlo et а1., 1994. Laboratory drift is a special problem, but one Therefore, in order to minimize intragroup varia- that is not peculiar to laboratory analyses (e.g. a tion, technical details should be considered. As an drift in the quality of interviews typically occurs example, for blood collection the following vari- during longitudinal epidemiological studies). ables need controlling (Young & Bermes, 1986; Laboratory drift is a consequence of changes in Pickard, 1989): procedures and accuracy over the course of time, so that the first samples that are analysed tend to • Collection tubes contamination—e.g. in the differ from subsequent samples. The avoidance of case of trace metals, all materials (needles, laboratory drift requires a monitoring programme tubes, pipettes, etc.) should be of a type that consisting of repeated quality controls. For exam- does not release metals. ple, measurements may be compared with a stan- • Types of additive—e.g. collection of plasma dard at different points in time. entails the use of heparin. Another source of drift, which cannot be tech- • Order of collection tubes—to avoid carry-over nically avoided, is the degradation of analites of trace additives, tubes without additives when they are stored for a long time. A typical should be proceeded first. example is given In Table 15, which shows the • Time of venipuncture—e.g. measurement of results of a case- control study nested within a

69 Application of Biomarkers in Cancer Epidemiology cohort. After blood collection, as soon as а case of cancer occurred, two suitable controls, matched to the case by time since blood collection, were sampled and their blood was analysed for retinal levels. ТаЫе-15 shows that among control subjects the A posteriori concentration of retinol decreases with the time that A priori probability (when has elapsed since blood collection. The impression probability p53 is positive) one gets when looking at retinol levels after less Stage 1 28% 63% than 1 year since blood collection is that cancer arose Stage 2 73% 80% . in those who had low vitamin levels compared 5tage 3 100% - with controls (the difference between cases and a Calculated from the same data as. in Table 13 (as described controls is statistically significant). However, after in the footnotes of that table) for stages 1, 2 and 3 3 years the difference has disappeared, indicating that the first measurement among the cases was likely to have been influenced by the metabolic since validity is often not measurable, reliability is impairment that occurs in cancer patients. In other sometimes used (incorrectly) as a surrogate. One words, two events occurred: degradation of retinal aspect that is highly relevant to the discussion of in stored samples, which is evident among control measurement error is timing: any inference about subjects; and an inversion of the cause— effect rela- the meaning of biomarker measures should be tionship as a consequence of the metabolic impair- strictly time-specific, since time influences the ment in cancer cases. Dealing with this source results in several different ways. of drift requires a strict matching of cases and Generally speaking, a blurring of the relationship controls for time since blood collection when between marker and disease occurs when errors of measurements are planned. In addition, measure- measurement are evenly distributed according to ments in cases should not be made shortly after the diseased healthy status (i.e. are not influenced blood collection, to avoid the effects of metabolic by the outcome). Both underestimation and over- impairment. estimation of the association of interest may occur when measurement errors are not evenly distrib- Conclusions uted across the study groups. Blurring is 'bias In the context of epidemiological studies using toward the null', while distortion as a consequence biomarkers, as well as in 'traditional' question- of uneven distribution of measurement errors can naire-based studies, we are interested in both validity be in either direction, both towards and away from and reliability. Validity refers to the truthfulness of the null hypothesis. However, there are some estimates, and is therefore the key issue. However, exceptions to this simple rule.

Time between blood collection and cancer onset Less than 1 year , 1-2 years >3 years n Mean n Mean n Mean g ( 911) (p9Il) (Ч ј1) Cases 66 641 45 650 116 694 Controls 132 722a 90 701k 232 633

From Wald et al., 1986. aР < 0.001. ЬР < 0.01.

70 Sources of variation in biomarkers

When we organize and analyse an epidemio- Fletcher, R.H., Fletcher, S.W. & Wagner, E.H. (1988) logical study employing biomarkers, we aim to Clinical ЕрјdепйоТo'lу—.Тhе Essentials, 2nd edn, Baltimore, minimize total intragroup variability and to iden- Williams and Wilkins tify intergroup differences (e.g. between exposed Lyles, C.M., sandier, R.S., Keku, T.O., Kupper, L.L., aid unexposed or between diseased and healthy Millikan, R.C., Murray, S.C., Bangdiwala, 5.I. & Ulshen, subjects. Total intragroup variation is the sum of M.H. (1994) Reproducibility and variability of the rectal intersubject, intrasubject, sampling and laboratory muicosal proliferaton index using proliferating cell nuctear antigen immunohistochemistry. Cancer variations. Obviously, if we do not have detailed EpidernioL Biomakers Pieu, 3, 597-605 information we cannot adjust for intragroup variation. This is the reason why in epidemiolo- Hankinson, S.E., Manson, J.E., London, 5.J., Willett, W.C. gical studies employing biomarkers it is im- & Speizer, F.E. (1994) Laboratory reproducibility of endogenous hormone levels in postmenopausal women. portant, whenever possible, to collect repeat Cancer Epidemiol. Biomarkers Prev., 3, 51-56 samples, potentially relevant information on subject characteristics that influence inter- Pickard N.A. (1989) Collection and handling of patients subject variation, and conditions under which specimens. Iп: Kaplan, L.A. & Pesce, A.J, eds, Clinical and Correlation, 2nd , CV samples have been collected and laboratory analy- Chemistry: Theory, Analysis едп Mosby, St Louis ses conducted. 5аvеlа, K., Hemminki, K., Phillips, D.Н., Hewer, A., Putman, K.L. & Randerath, K. (1989) Interlaborato y Acknowledgements г comparison of the 32P-postlabelliпg assay for aromatic I am grateful to Professors Renato Coda, Benedetto DNA adducts in white blood cells of iron foundry work- Terracini, Enzo Bailatori and Roberto Pagni, and to ers. mut. Res., 224, 485-492 Drs 5tefano Bonassi, Emily White and Nat Rothman for comments and suggestions. This Taioli, F., Kinney, P., Zhitkovich, A., Fulton, H., Voitkun, V., Cura, G., Frenkel, K., Toniolo, P., Garte, Ѕ. & Costa, work has been partially funded by the Associazione M. (1994) Application of reliability models to studies of Italiana per le Ricerche m1 Cancro and the Italian biomarker validation. Environ. Health Perspect., 102, National Research Council (Progetto Finalizzato 306-309 ACRO, Grant No. 93.04716. 04). СТ Wald, N., Boreham, J. & Bailey, A. (1986) Serum retinol and subsequent risk of cancer. Br. J. Cancer, 54, 957-961 References Brennan, P. & Silman, A. (1992) Statistical methods for Wiseman, H., Kaur, H. & Hailiwell, B. (1995) DNA dam- assessing observer variability in clinical measures. Br. age and cancer: measurement and mechanism. Cancer Med. J., 304, 1491-1494 Lett., 93, 113-120 Carmines, E.G. & Zeller, R.A. (1979) Reliability and Young, D.S. & Bermes, E.W. (1986) specimen collection and processing: sources of biological variation. Iii: Tiertz, цaIidity Assessment. London, Sage Publications NW, ed., Textbook of Ctirdcal Chemistry, Philadelphia, WB Cooney, R.V., Frankle, A.A., Hankin, J.Н., Custer, L.J., Saunders Wilkens, L.R., Haywood, Р.J. & Le Marchand, L. (1995) Seasonal variations in plasma micronutrients and antioxidants. Cancer EpidemioL BiomarkersPm '., 4, 207-215 Cordon-Carlo, C., Dalbagni, G., Saez, G.T., Oliva, M.R., Zhang, Z.F., Rosai, J., Reuter, V.E. & Pellicer, A. (1994) p53 mutations in human bladder cancer: genotypic vs. P. Vineis phenotypic patterns. lot. J. Cancer, 56, 347-353 Unit of Cancer Epidemiology, Dipartimento di Scienze Esrig, D., Spruck C.H., III, & Nichols, P.W. (1993) p53 Biomediche e Onçologia Umana, Ospedale di nuclear protein accumulation correlates with mutations Ѕ. Giovanni Battistâ. e Universitb di Torino, via Saoteua 7, Tonne, Italy in the p53 gene, tumor grade and stage in bladder cancer. Am. J. PathoL, 143, 1389-1397

71 Rpр1iсat[on of Biomaгkeгs in Сапсег Epidemiology Toniolo, Р, Botfetta, P., 5huker, D.Е . Rothman, N., H,jlka, B. arid Pearce N., eds 'ARC Scientific Pubheatiorir' No. 142 l пternatioпal Agency cr Research vn Cancer Lyon, 1997

Effects of biomarker measurement error on epidemiological studies

Е white

This chapter presents an overview of how measurement error in a biomarker affects epidemiological studies which use the biomarker. To estimate the effects of biomarker error, one must first measure the error using an appropriate validity or reliability study design and using appropriate parameters, i.e. parameters that are informative about the effects of measurement error on the parent' epidemiological studулhese measures of the biomarker measurement error from the validity or reliabiIity study can then be applied to what is known about the association under study in the parent study, in order to estimate the effects of the biomarker error on the results of the epidemiological study.

An epidemiological study that uses a biomarker underlying biomarker without laboratory or other can be of several types. For simplicity, we assume sources of error, and if the measure can fluctuate that the biomarker h an intermediate marker over time, the true biomarker would be integrated between an externat exposure and a disease, i.e.: over the time period of etiological interest (e.g. the mean value over a 10-year period). exposure - biomarker disease There are numerous sources of measurement error in biomarkers. Examples are given in Box 1. and that the parent epidemiological study assesses Some have been discussed in the previous chapter the association between the biomаrker and the (Vineis, this volume). Measurement error can be external exposure or between the biomarker and introduced by errors in the laboratory method the disease (or between the biomarker and another selected to measure the exposure of interest; for intermediate marker). Two common study designs example, if the true biomarker of etiological in epidemiology that include biomarkers are: (1) a importance is (i-carotene, the choice of total study in which a biomarker is compared between carotenoids in the blood as the biomarker test will those with a high level of exposure and those with include other exposures not relevant to the epi- low or no exposure; and (2) a case-control (or demiological true exposure. Further sources of nested case-control study) in which the biomarker error include failure to specify fully the protocol in is compared between those with the disease of terms of timing and method of specimen collec- interest and controls. Because both types of study tion, specimen handling and storage, and labora- are essentially two-group comparisons (either tory technique. An additional source of error is the exposure groups or disease groups) on the distrib- 'random' variation that occurs between batches ution of the biomarker, these will be discussed and between laboratory technicians even when the together as 'two-gтопр studies'. Measurement error protocol is well specified. in the biomarker leads to bias in the measure of 0f particular importance is the fact that mea- association (e.g. bias in the odds ratio for the asso- surement error can be due to short-term (e.g. week ciation of biomarker to disease) in the parent to week) biological variation and long-term change study; this bias is called information bias or mis- in the bioniarker within subjects. This type of error classification bias. is often ignored when assessing laboratory mea- The measurement error for an individual can be surement error, but it can have a large impact defined as the difference between his/her 'mea- on an epidemiological study. This is because, sured' (the biomarker 'test') and true' biomarkers. unless the biomarker is a fixed characteristic within individuals, the underlying true biomarker The true Ь опiагker can be conceptualized as the 73 Application of Biomarkers in Cancer Epidemiology

that influences the disease of interest is rarely an study. A primary concern in case-control studies individual's measured biomarker on a single day, of biomarker-disease associations is that the bio- but rather the average over time. Thus, even a logical effects of the disease or treatment may affect perfect measure of the biomarker at a single point the biomarker so that it is no longer an equally in time could be a poor measure of the 'true' good measure of the true (pre-disease) biomarker for biomarker. both cases and controls. The term differential mea- Differential measurement error occurs when the surement error can be more generally defined as biomaтkеr error differs between those with and measurement error that differs between the com- those without the disease to be studied in the parent parison groups (e.g. disease groups or exposure groups) to be used in the parent study. Non-differential measurement enor occurs when the measurement error does not differ between the comparison groups in the parent epidemiological study. The effects of both differential and non- в в в в r в differential error will be discussed in this chapter. `Validity' is the relationship between the bio- Errors in the laboratory method as a measure of thé exposure of interest marker test (the mismeasured bior arker) and Method may not measure all sources of the biological the true biomarker in a population of interest. true (etiological) exposure of interest Measures of validity are parameters that describe Method may measure other exposures that are not the the measurement error in the population. A valid- true exposure of interest ity study is defined here as one in which a sample Method may be influenced by 'subject characteristics of individuals is measured twice—ornce using the (other than the true exposure) that thé researcher biornarker test of interest and once using a perfect cannot manipulate, e.g. by the disease under study or measure of the true biornarker. To design a validity by other diseases study that reflects the amount of measurement Errors or omissions in the protocol error that will occur in the parent epidemiological Failure to specify the protocol in sufficient detail regard- study, several design issues must be considered. ing timing and method оf specimen collection, speci- First, the sample of individuals for the validity men handling, storage and laboratory analytical pro- study should be a random sample from the parent cedures . study. If this is not possible, the subjects in the Failure to include standardization of the instrument periodically throughout the data collection validity study should be comparable to the subjects in the parent study in terms of age, sex and other Errors due to biological variability within subjects parameters that could influence the distribution Short-term variability (hour-to-hour, day-to-day) in bio- (variance) of the biomarker and/or the error of the logical characteristics due to, for example, diurnal biomarker test. Second, the biomarker test should variation, time since last meal, posture (sitting versus lying down). be collected, processed, stored and analysed in the Medium-term variability (month-ta-month) due to, for validity study using the same procedures that will example, seasonal changes in diet be used in the parent study. Third, a perfect measure Long-term change (year-to-year) due to, for example, of the true biomarker is needed for all individuals deliberate dietary changes over time in the validity study. This true measure must reflect Errors due to variations in execution of the protocol the underlying true biomarker without error, Variations in method of specimen collection including error due to variation in laboratory pro- Variations in specimen. handling or preparation cedures and variations over time. The last issue is Variations in length of specimen storage particularly problematic, because the true bio- Variations in specimen analysis between batches marker of interest is often the average value over (different batches of chemicals, different calibration of many years. In addition, one should consider con- instrument) . ducting separate validity studies on those with the Variation in technique between laboratory technicians disease and those without the disease, to assess Random error within batch whether the measurement error differs between cases and controls. Finally, a validity study should

74 Effects of biomarker measurement error be analysed using parameters that provide infor- of reliability studies (Dunn, 1989; Armstrong et cL, ration about the effect of biomarker measurement 1992). Techniques to reduce biornarker measure- error on the parent epidemiological study. ment error, and therefore to reduce the bias in the Often, a perfect measure of the biomarker does results of the parent study caused by measurement not exist iris not feasible to use in a validity study. error, are covered throughout this book. Reliability studies, in which repeated measurements For the purposes of this chapter, sources of bias of the biomarker are taken on a group of subjects, other than misclassification bias in epidemiological usually only measure part of the measurement studies (i.e. bias due to the effects of confounding error. However, reliability studies can sometimes factors, unrepresentative selection of subjects, and be used to measure the validity of a biomarker, sampling a finite number of subjects) are assumed without having a perfect measure of that bio- to be absent. Furthermore, to make possible an marker. explicit analysis of the effects of measurement error This chapter is divided into six sections. The for continuous Ыоmarkers, certain simplifying first looks at quantifying the measurement error in assumptions are made about the form of the error. binary biomarkers using a validity study. The sec- The purpose is to give the reader some insight into ond section is a discussion of the effects of the evaluating the effects of measurement errors. error in a binary biomarker on the odds ratio or other measures of association in the parent epi- Mearürиng the error in a ыnary biomarker demiological study. The third considers quantify- Binary biomarkers are those that classify an analyte ing measurement error in continuous biomarkers or characteristic as present (positive) or absent using a validity study, and the fourth section covers (negative) for each study subject. Measurement the effects of the error in a continuous biomarker error in a binary (dichotomous) biomarker is on the parent epidemiological study which uses usually referred to as misclassification. Binary bio- the biomarker. Although one generally needs to markers are subject to all of the sources of mea- conduct a validity study for binary biomarkers, in surement error as described in the introduction the case of continuous biomarkers, reliability stud- and Box 1. ies can be designed that yield information about the validity of the biomarker. Thus, information Measures of misclassification—sensiivity and from a reliability study can answer questions about specificity the effect of biomarker error on the parent study. The degree of misclassification iii a binary biomarker This is covered in the fifth section. is measured by its sensitivity and specificity. The last section presents techniques to adjust Sensitivity and specificity cari be measured in a the results of the parent study for the effects of validity study in which the biomarker under evalu- measurement error. The researcher can then estimate ation (the misaneasured biomarker) is compared the value of the true odds ratio (or other measure with a perfect measure of the underlying bio- of association) that might have been observed in marker (true biomarker) in the population of inter- the absence of biomarker error. est. Individuals are then cross-classified by their Many related topics are beyond the scope of this results on each test: chapter. A discussion of the effect of measurement error in a susceptibility biomarker (a factor that True biomarker modifies the relationship between exposure and + - disease) on the results of the parent study using the Classified by + susceptibility marker will not be discussed. The biomarker test - reader is referred to other sources for information on the effects of measurement error in a categorical The sensitivity (sens) of the biomarker under measure (Walker and Blettner, 1985; de K1eтk еt а2., evaluation is the proportion of those who are true 1989; Armstrong et al., 1992), the effects of measure- positives (positive on the criterion test) who are rient error on sample size and power (Armstrong correctly classified as positive by the biomarker test: et al., 1992; McKeown-Eyssen, 1994; White et al., 1994), and the design and anaylsis of other types sens = a/(a + c)

75 Application of 8iomarkers in Cancer Epidemiology

(Note that the definition given here of sensitivity true biomarker of interest would be presence of the is different from the meaning in some laboratory protein expression in the normal tissue 5-10 years contexts, i.e. the lowest level detеctаЫe by a test.) before diagnosis, while the measured biomarker is The specificity (spec) is the proportion of those expression in normal tissue at the time of diagno- who are true negatives who are classified as nega- sis, which could be influenced by the presence of tive by the biornarker test: disease. Assuming the protein expression could be measured in a stored tissue sample, the validity spec= dI(b + d) study could be conducted on two groups: the subset of cases and the subset of controls who had Even though both sensitivity and specificity can prior breast biopsies, say, 5-10 years before the range from 0 to 1, it is assumed that sensitivity plus index biopsy, with the true measure being expres- specificity is greater than or equal to 1. In other sion from the earlier biopsy. This would yield sep- words, for the biomarker test to be considered a arate measures of sensitivity and specificity for measure of the true biomarker, the probability that cases and controls. the biomarker test classifies a truly positive person as positive (sensitivity) should be greater than the prob- Effects of the error in a binary biomarker on ability that it classifies a truly negative individual epidemiological studies as positive (1 — specificity), i.e. sens> (1— spec), or Тhе effect of differential misclassification of a binary (sens + spec) > 1. Thus the parameter (sens + spec— 1), biornarker in a two-group study called the Youden misclassification index (Kotz & The effects of misclassification are straightforward Johnson, 1988), is a good measure of the total for studies in which a binary variable is compared degree of misclassification. If the Youden index is between two groups (Brass, 1954; Newell, 1962; 1, the biomarker testis perfect; if it is zero, the test Gullen et al., 1968; Goldberg, 1975; Barron, 1977; has no association with the true biomarker; and if Copeland et al.,1977; Fleiss, 1981; Kleinbaum etaL, it is less than zero, the test is inversely related to 1982). the biomarker. In an unmatched two-group study of a binary For a validity study to measure the sensitivity biomarker, under the assumption that the group and specificity of a biomarker, the study sample status is correctly classified, the effect of misclassi- can be made up of subjects not restricted by their fication of the biomarker is to rearrange individu- biomarker status, or one can sample subjects who als in the true 2 x 2 table into an observable 2 x 2 are true positives and those who are true negatives table. individuals in the first group (diseased or by the criterion test. However, one cannot sample exposed group) remain in the first group but subjects based on the results of the mismeasured may be misclassified as to biomaoker status, and biomarker test and correctly compute sensitivity the second group (non-diseased or unexposed) is and specificity. also rearranged, as follows: As noted in the introduction, separate validity studies should be conducted on the comparison True classification groups to be used in the parent study, particularly if one suspects that the sensitivity and/or speci- Disease or exposure group ficity may differ between groups. Specifically, if the parent epidemiological study is a case—control Group 1 Group 2 study in which the disease could influence the bio- + - marker test, then separate validity studies of the True biomarker + Рt РZ biomarker should be conducted on a group with the — 1 — Р1 1 — Р2 disease and on a control group. For example, suppose the parent study were a study to measure the pres- Measures of association Difference = Рt — РZ ence of a specific protein expression (present versus 1 (1— 2 absent) in normal breast cells among women with RT Р Л ) concurrent breast cancer compared with those О = pz (1— 1) without cancer at the time of a breast biopsy. The Р

76 Effects of biomarker measurement error

Observable classification Observable difference -р 1 —р2 (2) (misclassification) or Disease or exposure group р1(1— рг) OR0 = Z Group 1 Group 2 р (1-p1) (3) where 1 and z are from equations (1). Biomarker test + р р p1 рг Differential misclassification can have any — 1 — рr 1— рг effect on the difference: the observable difference, could be closer to the null hypothesis or Measures of association Difference = p1 — р1 — рz р2 cross over the null hypothesis (i.e. have a different sign) compared with the true difference, Рr — РZ. ORo = pr Similarly, compared with the true odds ratio, the r) observable odds ratio can be closer to the null рг (1— р hypothesis of OR = 1, be further from the null or Note that Р1 and Рz are the true proportions of cross over the null. biomarker-positives in groups 1 and 2 respectively, Returning to an earlier example, let us suppose it is hypothesized that a certain protein expression and similarly p1 and р2 refer to the proportions that would be 'observable' as positive by the biomarker in normal breast tissue is associated with increased test in the two groups. The term observable is used risk of breast cancer. This could be tested in a to mean that which would be expected on average case—control study, using normal tissue removed when there is measurement error (the actual at the time of diagnosis from breast cancer cases and observed parameter in the parent study would be tissue from controls undergoing a breast biopsy. an estimate of the 'observable' parameter). suppose that the test was insensitive among both There is differential misclassification when the cases and controls, such that only half of those sensitivity of the biomarker test for group 1 (sensr) with true protein expression were classified as differs from that for group 2 (sense) arid/or the positive (sens1 = sense = 0.5). Let us assume further specificity of the biomarker test for group 1 (spec1) that the specificity was perfect for controls differs from that for group 2 (specZ). The misclassifi- (specZ = 1.0), but that, among cases, 10% of those cation leads to the observable p1 and рz being without expression in normal tissue were classified different from the true P1 and P2 (Goldberg, 1975): as positive due to the influence of the breast tumour (specr = .90). If there were no true associa- р1 = sens1 к Р1 ±(1 — syеc1) x (1— Р1) tion between breast cancer and the protein expres- р2 = sense к Рz ±(1 — specz) x (1— Рz) (1) sion in normal tissue (e.g. suppose 10% of each group were protein-positive), then applying equa- The first equation states that a proportion tions (1) and (3) would yield the following: (sens1) of the true biomarker-positives (Р1) in the True classification: first group plus a proportion (1— specr) of the true biomarker-negatives (1 — Р1) in the first group will Breast cancer be classified by the biomarker test as positive (p1) + - in the first group. The second equation expresses True + 0.10 0.10 the same concept for group 2. biomarker — 0.90 0.90 The association between group and biomarker ORT = 1.0 in the parent study would typically be measured by the difference in the proportion biomarker posi- Observable classification: tive if the two groups were exposure groups and by the odds ratio if the two groups were disease р1 = 0.5 x О.1 + 0.1 к 0.9 = 0.14 groups. When there is measurement error, these рZ = о.s хО.1+0=0.05 measures of association are biased because they are OR = 0.14 (0.95) based on р1 and р2: б 0.05 (0.86)

Application of Biomarkers in Cancer Epidemiology

Thus, a true odds ratio of 1.0, i.e. no association collect more adequate DNA specimens, and the between the disease and Ыomarker, could appear specimens were tested by an L1 consensus primer as a strong association (OR0 = 3.1). polyrnerase chain reaction (PCR) technique. The second study is considered here as an epidemlo- The effect of non-differential misclassification of a logical study without biomarker measurement binary biomarker in a two-group study error, and the fiist study as one with biomarker Non-differential misclassification in a two-group error. Had a perfect measure been available at the study would occur when the sensitivity of the time of the earlier study, a validity study could biomarker test is the same for both groups have been conducted so that the sensitivity and (sens1 = sens2) and the specificity is the same for specificity of the first test could be computed. both groups (spec1 = spec2). Then the effect of mea- Adapting data from this report (modified to serve surement error in the biomarker on the difference as an example), such a validity study might have between groups can be computed as in equations yielded the following: (1) and (2), which simplifies to: True (2nd) test r - 2) (sens + spec -1) (4) HPV+ р1 - pZ = (Р Р HPV- Misclassified (1st) HPV + 10 6 This states that the observable difference bet- test HPV- 5 31 ween the proportions that are positive in group 1 Total 15 37 and group 2, p1 -р2, based on the biomarker test is equal to the true difference between groups, P1 and sensitivity and specificity could be computed -Р 2, multiplied by a factor equal to sensitivity plus as follows: the specificity minus 1. If the biomarker test at a mimimum classifies a true positive person as positive sens = 10115 = 0.67 on the test with greater or equal probability than spec 31/37 = 0.84 it classifies a true negative person as positive (i.e. the Youden index (sens + spec - 1) is in the range Here we assume that the mea ѕuтепјепt error is 0-1), then the observable difference is always less non-differential, i.e. that the sensitivity and speci- than the true difference and does not change sign. flcity are the same for cases and controls. The effect For example, if the biornarker test had sens = 0.7 of this magnitude of biomarker error on the true and spec = 0.8, which could be considered reason- relationship between HPV and CIN can be estimated able test accuracy, then (by equation 4) the observ- by applying equations (1) and (3) to the true rela- able difference between groups would only be half tionship between HPV and CIN as seen in the sec- of the true difference (i.e. sens + spec -1 = 0.5). ond epidemiological study: The effect of non-differential misclassification of a biomarker on the odds ratio in a case-control CIN study can be computed by calculating p1 and рZ as Cases п (%) Controls n (%) in equation (1) (except there would be a common True (2nd) HPV + 381 (0.81) 80 (0.18) sensitivity and a common specificity for the dis- test HPV- 89 (0.19) 375 (0.82) eаsед and non-diseased groups) and then саlculаt- 470 (1.0) 455 (1.0) ing the odds ratio (equation 3). As an example, data are adapted from a report by 5chiffшan & OR = 0.81 (0.82) = 20 Schatzkin (1994). They report on two case-control T 0.18 (0.19) studies of human papillomavirus (HPV) infection and cervical intra-epithelial neoplasia (CIN). A р1 = о.67x0.в1+ о.1jx о.19= о.57 major difference between the two studies was the рZ =0.67x0.18+0.16x0.82=0.25 accuracy of the test for HPV In the first study, a 3 ml cervicovaginal lavage was tested by Southern OR = 0.57 (0.75) =4.0 blot DNA hybridization techniques. In the second ° 0.25 (0.43) study, a 10 m1 cervicovaginal lavage was used to

78 Effects of biomarker measurement error

This shows that a study using the misclassi6ed The theory of measurement error in continuous HPV test would find 57% of cases positive, 25% of variables and its effects on studies of a continuous controls positive (rather than 81% and 18% based outcome were developed in the fields of psycho- on the accurate test) and an odds ratio of 4.0 rather metrics, survey research and statistics (Hansen than the true odds ratio of 20. In fact, the first et al., 1961; Cochran, 1968; Lord & Novick, 1968; study using the inaccurate HPV test observed an Nunnaцy, 1978; Allen & Yen, 1979; Bohmsted, 1983; odds ratio of 3.7. Fuller, 1987). The effects of measurement error Non-differential misclassification leads to an have also been derived in the context of epidemio- attenuation of the odds ratio towards the null logical studies of a continuous exposure variable value of 1 (Gullen etaL, 1968). The degree of atten- and a dichotomous disease outcome (Prentice, 1982; uation iп the observable odds ratio depends not Whittemore and Grosser, 1986; Armstrong et al., only on the true odds ratio and the sensitivity and 1989). specificity of the biomarker test, but also on the proportion of the non-diseased group who are true A model of measurement error biomarker-positives (]?2). The observable odds ratio A simple model of measurement error in a contin- does not 'cross over' the null of 1. uous measure X is: Table 1 gives further examples of the effect of non-differential misclassification on the odds ratio Х1 = Tг +b+Ei for reasonable values of sensitivity (0.5-0.9), specificity (0.8-0.99) and Р2 (0.1, 0.5) and for true odds where µ~ = 0 and рTЕ = 0. In this model, for a ratios of 2 and 4. As can be seen from the table, the given individual i, the measured biomarker Х; attenuation in the odds ratio can be considerable. differs from its true value Т by two types of When the proportion who are truly positive is low measurement error. The first is the systematic (e.g. РZ = 0.1 in upper half of the table), the atten- error or bias, b, that would occur (on average) uation of the odds ratio is severe, except when the for all measured subjects. The second, E', is the specificity is very high (e.g. spec = 0.99). When the additional error in Х~ for subject L E will be proportion who are truly positive is high (e.g. referred to as the subject error to indicate that it РZ = 0.5 in lower half of the table), the observed OR varies from subject to subject. It does not refer is strongly attenuated from the true OR except just to error due to subject characteristics; rather when the sensitivity is very high (e.g. sens = 0.9). Even it includes all of the sources of error outlined in strong associations between the true biomarker Box 1. and disease would be obscured by moderate values For the population of potential study subjects, of sensitivity and specificity. For example, for X, T and E are variables with distributions, e.g. sens = 0.7, spec = 0.8 and OR = 4.0, the observable the distribution of E is the distribution of sub- odds ratio would be 1.64 for РZ = 0.01 and 1.83 for ject measurement errors in the population of РZ = 0.5. These obsетvаые odds ratios would not be interest. X, T and E would have expectations detectable as different from the null vaiue of 1 unless (population means over an infinite population) the epidemiological study sample size were large. denoted by i, µT and µE respectively, and variances denoted by , 6.1, and вE. Because the Measuring the error in a continuous biomarker average measurement error in X in the pop- using a validity study ulation is expressed as a constant, b, it follows Often a biomarker assay yields quantitative infor- that µr, the population mean of the subject niation about the amount of an analyte in a bio- error, is zero. The assumption of the model that logical specimen; these measures can usually be the correlation coefficient of T with E, PTT' is considered to be continuous variables. This section 0 states that the true value of the biomarker is covers the parameters of measurement error that not correlated with the measurement error. In can be derived from validity studies in which each other words, individuals with high true values subject in the validity study is measured twice— are assumed not to have systematically higher once using the mismeasured biomarker and once (or lower) errors than individuals with lower true using a perfect (true) measure. values.

79 Application of Biomarkers in Cancer Eрidеmiolo9Y

.. - - в в l в в iI(•] в •

True OR=2.0 True OR=4.0

Biorriarker test sensitivity Biomarker test specificity ORô ORob Р2 = 0.5 0.80 1.14 1.38 0.7 0.80 1.23 1.64 0.9 0.80 1.32 1.92 0.5 0.90 1.28 1.76 0.7 0.90 1.39 2.09 0.9 0.90 1.48 2.41 0.5 0.99. 1.75 3.06 0.7 0.99 1.83 3.33 0.9 0.99 1.89 3.61

Рz = 0.5a 0.5 0.80 1.24 1.46 0.7 0.80 1.40 1.83 4.9 0.80 1.64 2.59 0.5 0.90 1.35 1.69 0.7 0.90 1.50 2.07 0.9 0.90 1.73 2.85 0.5 0.99 1.48 1.96 0.7 0.99 1.61 2.33 0.9 0.99 1.82 3.11

aР2 is the prevalence of true biomа kег-positives in the non-diseased group. Р1, the prevalence of true biomarker-positives in the diseased group, is, by detinitton, Р1 =. Р2 к ORT!(1 + Рz (ORS 1 j]. boRb from equatons (1) and (3).

Measures of measurement error-bias and validity of E, which is often called the within-subject vari- coefficient ance. (Note that the model is formulated so that Two measures of measurement error are used to the two measures of measurement error have sep- describe the relationship between X and T in the arate parameters. The average error is given by b, population of interest, based on the above model and b does not contribute to the variance of the and assumptions. One is the bias, i.e. the average error. The error that varies from subject to subject measurement error in the population: is parametrized by E and measured by oE, and E does not contribute to the average error.) A more important measure of precision is the correlation b = ~х - of T with X, рTX, termed here as the validity coef- The bias in X can be estimated from a validity ficient of X. The measure ртX is important because study as: it relates the within-subject variance б to the total variance бК, and it is this ratio, along with the bias, b=X-T that measures the impact of biomarker error on the parent epidemiological study. Using the above The other is a measure of the prеcisian' of X, i.e. the model, it can be shown that ртх is the square root variation of the measurement error in the popula- of 1 minus the variance of E relative to the variance tion. One measure of precision is o, the variance of X (Allen & Yen, 1979), i.e.:

80 Eiiects of biomarker measurement error

= d1— (вÉ1 ) (5) groups in the parent study, e.g. there could be dif- Ртх ferential error between cases and controls in a Using the above equation, it can be seen that case—control study. the smaller the error variance, the greater is ртх. The terminology surrounding measurement would range between 0 and 1, with a value of error varies between fields. In this chapter, the 1Ртк jndicating that X is a perfectly precise measure of terms validity, accuracy and measurement error are T. PTX is assumed to be zero от greater, i.e. for X to used as general terms reflecting the relationship be considered to be a measure of T, X must be at a between X and T, including both the concepts of minimum positively correlated to T. bias and precision. Iп laboratory quality control, can be estimated in a validity study by the the terms validity and accuracy are sometimes used PearsonРтх correlation coefficient of X with T. Thus, to refer to unbiasedness only. T can be interpreted as the proportion of the vari- рanceк of X explained by T. Using the above model: Effects of error in a continuous biomarker on epidemiologicai studies ' Q" When the bias and validity coefficient of the bio- Ртк бт marker (X) are known, one can estimate the For example, if ртх were 0.8, this would reflect that impact of the degree of measurement error in X on only 64% of the variance in X is explained by T, the parent epidemiological study. with the remainder of the variance being due to Several types of epidemiological study designs are the error. discussed, and both differential and non-differential To further understand the concepts of bias and measurement errors are considered. First, however, precision, consider a situation in which X only has the effect of measurement error on a single study a systematic bias, with E/ = 0 for all subjects. For population is discussed. example, suppose that the only source of error in a measure of serum cholesterol was that it quanti- The effect of measurement error on the observable tated each individual exactly 100 mgldl too high. mean and variance Then, in a population, the variable X, even though In a single study population, both the mean and it has systematic measurement error, could be used variance of the measured biomarker X would differ to order each person in the population perfectly from the true mean and variance due to measure- by their value of T. X would be biased but have ment error. Using the above model, the population perfect precision. However, if Е~ varied from person mean of X would differ from the trie mean (the to person (around the mean µE = 0), the ordering is population mean of [) by b: lost. The greater the variance of Е, relative to the variance of X, the less precise is X as a measure of T. µ =µT+b The degree of measurement error is not an inherent property of a biomarker test, but rather is The population variance of X, based on the a property of the test applied using a particular model and assumptions, would be (Allen & Yen, protocol to a specific population. Therefore, the 1979): error not ог lу will vary between two assays which measure the same biomarker, but also can vary Ρ T+ É ΡT1 в X = в б в Ртх (6) for a single test when applied using a different protocol or when applied to different population Thus, the variance of X in the population is groups. Moreover, the validity coefficient is de- greater than the variance of T, due to the addition pendent on the variance of the true biomarker in of the variance of the measurement error. For the population (q), so that even if the error example, if the validity coefficient (р TX) were 0.8, Ρ variance в É wer.e the same for two populations, then the variance of measured X would be 56% ΡT/0.82 = 1.56 ртх would differ if вT differed. Therefore, a greater than the variance of T ( = в validity study done on one population may not di- вT by equation 6). rectly apply to another study population. Finally, Figure 1 demonstrates the effect of measure- measurement error could differ between the study ment error on the distribution of X in a popula- Application of Biomarkers jn Cancer Fрidemiology

For a two-group epidemiological study, the biomarker measure Х11 for the ith person in the first T group (exposed or diseased group) differs frот that person's true exposure Т11 by the systematic bias 'й -- X (b1) in X within that group and by error in subject о i's measure:

Хrј = Т;+bi+ Еr, wr µх Biomarker (X or T) and, similarly, for the second group (unexposed or non-diseased group): Figure 1. Distribution of true (T) and measured (X) biomarkers. хZ,= тz,+ ьZ + ЕZ1 tion, using a normally distributed biomarker aid normally distributed error as an example. The bias Differential measurement error occurs when b1, in the measure causes a shift in the distribution of the bias in the first group, differs from b2, the bias X compared with T. The increased variance of X in the second group, and/or thé variance of Е1 dif- compared with T (measured by Pix) causes a flat- fers from the variance of Еz. tening of the distribution of X. Even if a measure Errors in the measurement of the biomarker X were correct on average (b = 0), there could still be would affect the measure of the association in the substantial measurement error due to lack of pre- epidemiological study. The effect of differential cision, which could lead to a greater dispersion in measurement error on the observable difference the measured exposures. between groups in the biomarker means is:

The effectif differential measurement error in a 2) ~ i - цXZ = (µri- ~т2) + (bt - b (7) continuous biomarker in a two-group study х While measurement errors have an effect on the This states that the observable difference observable mean and variance of an exposure between groups is equal to the true difference plus variable within a single population, a greater the difference in the biases of the biomarker test concern would be the effect of measurement between groups. This equation holds even when errors when comparing a biomarker between two the 'perfect' measure T is not perfectly precise, as groups. This section and the next section cover long as T is unbiased (or at least not differentially epidemiological study designs in which the biased between groups). biomarker is compared between two groups. The effect of differential measurement error When the two groups being compared are expo- in X on the odds ratio can only be easily quan- sure groups, e.g. a highly exposed group and an tified when certain simplifying assumptions are unexposed group, the common measure of asso- made. Results can be derived for two-group ciation is the difference in the group means studies under the following assumptions: (a) X1 of the biomarker. When the groups being com- and ХZ are modeled as above with рTЕ = 0 for pared are a case group and a control group, the each group, (b) Т1 and ТZ are normally distributed common measure of association between a bio- with means µr1 and µTZ respectively and the same marker and disease is the odds ratio, which is variance, сT, and (c) Е1 and Е2 are normally dis- often expressed as the odds ratio of disease for one tributed with mean zero and common variance, level of the biomarker versus another (usually бT. The later assumption means that X1 and X2 lower) level, or as the odds ratio of disease for a are equally precise, so only differential bias is u unit increase in the level of the biomarker. (These considered. results do not apply to odds ratios expressed The above assumptions imply a logistic regres- as odds of disease for the upper quantile of the sion model for the probability of disease (pr(d)) as biomarker versus the lowest quantile—see a function of true biomarker T, with a true logistic Armstrong et al., 1992.) regression coefficient RT (Wu et aI., 1986):

82 Effects of biomarker measurement error

log 1—pr(d) where Rr = FLт1 µт2 T2 The true odds ratio for any u unit increase in T µ l L 2 wr, would be ORf = еРтu х, г тгг Х Biomarker (X or 7) With measurement error in the biomarker test X, the assumptions also lead to a logistic model: Figure 2. Effect of differential measurement error (b ~ b2) on distribution of true (T) versus measured (X) biomarker in at two- rd group study. log p = аo + (30Х, 1--pr(д) where the null value of 1(it would slope downwards from 1 as X gets larger, rather than upwards). po = ( z)+(bib2) µri т Differential measurement error should be a con- бт, ТХ cern in a case—control study when the biomarker is measured within the pre-clinical disease phase The observable logistic regression coefficient, R , before diagnosis, or any time after diagnosis, and differs from p~ due to the measurement error in X. the marker is not fixed. This concern is illustrated by po can be expressed in terms of RT (if R~~ 0) as fol- several case—control studies that found low serum lows (Armstrong et al., 1989): cholesterol at the time of diagnosis to be a risk factor for colon cancer (Law & Thompson, 1991), which could be an artefact if increased catabolism or other po= i 1+ b1 b ( г Rт 8) effects of colon cancer reduce serum cholesterol. As 1 N i µ z) Ртх т т an example of the interpretation of equation (8), if Since (b1 — bZ) can be any magnitude and can be true mean serum cholesterol for cases 10 years either positive or negative, the observable logistic before diagnosis (before the effects of disease) was regression coefficient, Ro, could be greater than, 6.0 mnol/l µT1, while the true mean serum choles- less than, or even have a different sign than the terol among controls at a comparable time period true coefficient fly. The observable odds ratio for was 5.8 µ.f , then the true odds ratio for colon can- any u unit increase in X, ORo =elou, could be cer wоulд2bе greater than 1, say ОАT = 1.05 for a towards the null value of 1, away from the null or 1 nmol/l increase (Q= li (1.05) = 0.05). However cross over the null value compared with the true at the time of diagnosis (and a comparable time odds ratio. period for controls), there was substantial error in Figure 2 gives a graphical presentation of differ- the measure of serum cholesterol (X) as a measure ential measurement error, in particular differential of T. suppose the effect of the disease were to cause bias between cases and controls. In the figure, the serum cholesterol at the tinre of diagnosis for cases true mean biomarker level in the diseased group, (X1) to be 0.5 nmol/l lower, on average, than 10 years µgr, is greater than the true mean biornarker Level earlier (b1 = — 0.5), while among controls - = in the non-diseased group, µ~Z. This would lead the serum chorlesterol increased over the 10 years to a positive slope in the true odds ratio curve. In by 0.5 nmol/l (b2 =µX —µTZ = 0.5}. furthermore, this example, the bias for the non-diseased group suppose that ртx = 0.8. Then, by equation (8); is positive, so the distribution of X2 is shifted to r -- ~.5 — 0.5 the right relative to Т2, and the bias among those (30 = 1+ 0.82 x 0.05 with disease is negative so that the distribution of L 6.0 — 5.8 X1 is shifted to the left relative to T1. This would lead the observable odds ratio curve to cross over =-0.13

83 Application of Biomarkers in Cancer Epidemiology

and ORo = p = 0.88 Ρ е о (1.Lx1 - uх2) = (Irr1 - µт 2) (9) Under these assumptions, a true small increased However, the lack of precision flattens each dis- risk (ORT =1.05) associated with serum cholesterol tribution and leads to more overlap and less would appear as a decreased risk (OR0 0.88). distinction between the distributions of X1 and X2 Differential bias is a greater concern than dif- compared with the true distributions. The variance ferential precision, because, as described above, of the observed difference between groups in the differential bias can lead to a shift in the distribu- mean biomarker would be expected to increase by tion of the biomarker in one group relative to a factor 11 ртк the other. Differential measurement error will Under non-differential misclassification, the also occur if differs between groups. If there odds ratio curve is flattened towards the horizon- were no differential bias but the 6E values differed tal line of odds ratio equal to 1 for all X. Equation (and бT values were equal for the two groups), (8) can be simplified to (Whittemore & Grosser, the shape of the odds ratio function could 1986; Wu et al., 1986): change. For example, the observable odds ratio curve could be U-shaped when the true ехро- R = (10) sure-disease relationship is increasing (Gregorio о р хРт 1985). et aL, If ORт = е1Tu is the true odds ratio for a u unit increase in T, aid ORo = еR0 is the observable odds The effect of non-differential mеа5игетвпt error in a ratio for a u unit increase in X, then: continuous biomarker in a two-group study When the assumptions given in the above section ORo = рRpтx (11) hold, there is non-differential misclassification when there is equal bias (Ы = b2) and equal error This states that the observable odds ratio for any variance (or equivalently equal ртx) in the bio- fixed difference in units of the biomarker is equal marker test when applied to the two groups in the to the true odds ratio for the same fixed difference parent epidemiological study. Figure 3 illustrates the to the power р7X Since 0 < Р X < 1, the observable effects of non-differential misclassification. Under odds ratio will be closer to the null value of 1 (no non-differential misclassification, the two distrib- association) than the true odds ratio. The observ- utions may shift, but they are not shifted with able odds ratio does not cross over the null value if respect to each other, became there is equal bias for X and T are, at a minimum, positively correlated. the two groups. Thus, based on the model pre- While the difference in the bias of X between sented, the observable difference in the mean values groups (b1- b7) can play a major role in the bias in of X between groups is equal to the true difference: the odds ratio under differential measurement error (equation 8), equation (12) shows that the attenuation in the odds ratio under non-differen- T2 т, tial misclassification is a function of the precision of X (measured by 2), but not of the bias in X. Prentice (1982) has shown that, under similar

Ш assumptions, equation (10) also applies approximately to estimates of R obtained from the proportional hazards model for data from cohort and matched case-control studies. N'Хг iLТ2l.LХ W7i Examples of the effects of non-differential mea- Biomarker (X or T) surement error in a normally distributed biomarker on the odds ratio, based on the attenuation equa- Figure 3. Effect of non-differential measurement error (equal tion (11), are given in Table 2. The table shows that bias and precision) on distribution of true (T) versus measured biomarkers with a validity coefficient ртх of 0.5 (X) biomarker in a two-group study. would obscure all but the strongest associations.

84 Effects of biomarker measurement error

For example, if the true odds ratio for a u unit change in the biomarker were 4.0, this would be attenuated to an observed odds ratio of 1.41. Furthermore, measures as precise as р = 0.9 still lead to a modest attenuation-for example, a true odds ratio of 4.0 would be attenuated to 3.07. True True The effect of independent measurement error in a OR-2.ü OR=4.0 a continuous bio- study of the correlation between Px Oho OR0 marker and a continuous variable Ртх 0.50 0.25 t19 1.41 Often, the parent study which uses a biomarker 0.60 0.36 1.28 1.65 will be one in which a continuous biomarker X will 0.70 0.49 1.40 t97 be compared with another continuous variable Y, 0.75 0.56 1.48 2.18 e.g. a questionnaire measure of an exogenous ex- 0.80 0.64 1.56 2.42 posure or another biomarker. Suppose both X and 0.85 0.72 1.65 2.72 Y are measured with error, with the measurement 0.90 0.81 1.75 3.07 error given by the model presented in `the model 0.95 0.90 1.87 3.49 of measurement error' section, with the true value of X denoted by TX the true value of Y denoted by is the validity coefficient of X. р» by X and the x is the reliability coefficient of X under the parallel test ТY, the subject error in X denoted Е р . model (see text): = subject error in Y denoted by Е There is indepen- Px Ртн dent nieasurenient error if Е is uncorrelated with The true OR is lite odds ratio tir a u unit difference in т. ORo is the observable odds ratio for a u unit difference in . 1,, Y is uncorrelated with ТX, and X is uncorre- Х Т Е Е Computed from equation (11). See text for model and lated with Е1,. Under these assumptions, the observable correlation coefficient of X with Y is given assumptions. by (Allen & Yen 1979):

Y° X PTX 1Y (12) (X) with 1 year average serum il-carotene (ТX) is 0.8, Рх РтХтУ Рт and the correlation of the food frequency estimate This equation states that the observable correla- of dietary 3-carotene (Y) with true dietary intake of tion р Y in the parent study is equal to the true cor- (3-carotene (T1,) is 0.6. Also assume that the errors X and Y and of relation рTXT times the validity coefficient of X in X and Y are independent of true ) times Jte validity coefficient of ( T~Y). Note each other. Then the observable correlation coeffi- (Рт~х У р that the bias in X and the bias in Y do not enter the cient between X and Y (рXy) in the study would be equation. This is because adding a constant to X (by equation 12) 0.7 x 0.8 x 0.6 = 0.34, a consider- and Y does not change the correlation between X able attenuation from the true association of 0.7. and Y. As an example, suppose one were conducting a Measuring the error in a continuous biomarker study of serum R-carotene (X) in relation to dietary using a reliability study p-carotene as measured by a food frequency ques- The term reliability is generally used to refer to the tionnaire (Y). Assume that the true time period of reproducibility of a measure, i.e. how consistently interest is the prior year, but serum il-carotene has a measurement can be repeated on the same sub- error primarily because it is only measured once jects. Reliability can be assessed in a number of during the year. The food frequency measure has ways, but only one type will be covered in this error primarily because individuals cannot accu- chapter. Intramethod reliability studies measure rately recall diet over the past year. suppose that the reproducibility of an instrument on the same the true correlation of serum R-carotene over the subjects repeated two or more times using the prior year (ТX) and dietary intake of il-carotene identical method or with some variation. For over the past year (T1,) is 0.7. Assume that the cor- example, a comparison could be made of a bio- relation of a single measure of serum 3-carotene marker from a single specimen analysed in two

85 Application of Biomarkera in Cancer Epidemiology batches or by two laboratory technicians оr from measured once, by either X1 or X2 (e.g. either by two specimens on each subject collected at two laboratory technician 1 or 2). Therefore, any sys- points in time. Reliability studies in which two dif- tematic difference between X1 and X2 would not be ferent analytic methods are compared, with one a systematic bias affecting everyone in the parent better than the other but neither perfect (inter- study, but would vary between subjects because method reliability studies) are not covered here (see some are measured bу Х1 and some by X2. Because Armstrong et al., 1992). Measures of reliability are X1 and X2 will be used as interchangeable measures primarily important for what they reveal about the of X in the parent study, and because more than validity of a biomarker test, because the bias in the two replicates per subject can be used to compute odds ratio in the parent epidemiological study is a the intraclass correlation coefficient, the reliability function of the validity of the Ъiomатker measure. coefficient of X can also be written as Px- This section covers the interpretation of measures Two measures of the validity of a continuous of reliability in terms of measures of validity for con- exposure measure were shown to be important iп tinuous biomarkers. The purpose is to provide some assessing the impact of measurement error: the general concepts for the interpretation of reliability bias and the validity coefficient. Unfortunately, studies. Only continuous measures are covered be- reliability studies generally cannot provide infor- cause the relationship between the reliability and mation on the bias in Х. The inability of many reli- validity of categorical measures is more complex. ability study designs to yield information on bias, particularly on differential bias between study A model of reliability and measures of reliability groups, is a major limitation. It should be recalled, suppose each person in a population of interest is however, that in the case of non-differential mea- measured two or more times using the same bio- surement error (and certain other assumptions), marker test to be used in the parent study. For a the attenuation equations depend only on the given subject i, two (or more) biomarker measure- validity coefficient and not on the bias. The relia- ments, Х11 and Х~2, are obtained. (Note: in this sec- bility coefficient does provide information about tion, X1 and X2 refer to the two measures per sub- the validity coefficient, and thus can be used to ject, not to two groups of subjects as in the earlier estimate the effects of measurement error on the sections.) The simple measurement error model parent study under the assumption of non-differ- described above applies to each measure: ential measurement error.

i1= ,+b1 + Relationship between reliability and validity under Х Т Егr X1= + b2 -ј- Е2 the parallel test model When certain assumptions are met, reliability stud- Both Хгr and Х12 are measures of the subject's true ies can yield information about the validity coeffi- biomarker Т1, but with different errors. In a reliability cient. One such set of assumptions is the model of study, information is available on X1 and X2 for each parallel tests (Lord & Novick, 1968; Nunnally, subject, but not on Т. A reliability study can yield 1978; Allen & Yen, 1979; Bohrnstedt, 1983). The estimates of the mean of X1 and Хz(.tх1 and µX2) first assumption of the parallel test model is that and of the correlation between the two measures, the error variables, Е1 and ЕZ, are not correlated TX V termed the reliability coefficient. р Т with the true value Т. It is further assumed that Еr The intradass correlation coefficient is generally and ЕZ have equal variance, бT. This also implies used as the reliability coefficient [see Neiss (1986) that X1 arid X2 have equal variance and that X1 and or Armstrong et al. (1992) for formulas for its cor- X2 are equally precise (р 1= р Z). This is usually a putationj. The intraclass correlation differs from reasonable assumption in intramethod reliability the Pearson correlation coefficient in that it studies, since X1 aid X2 are measurements from includes any systematic difference between X 1 and the same instrument. Finally, it is assumed that Е1 X2 (Le. any difference between b1 and b2) as part of is not correlated with ЕZ. This important (aid the subject error Е (the enor that varies from sub- restrictive) assumption implies, for example, that ject to subject). The assumption is that in the par - an individual who has a positive error, Е1, on the ent epidemiological study, each subject will be first measurement is equaly likely to have a positive

86 Effects of biomarker measurement error or а negative error, Е2, on the second measurement. one measure is not more likely to be higher than These assumptions are often summarized by saying her true value on another measure). This study also that two measures aie parallel measures of Tif their measures most sources of error—blood processing, errors are equal aid uncorrelated. storage, random laboratory error (within-batch Under the assumption of parallel tests, it can be error) and variation due to changes in day-to-day shown that (Allen & Yen, 1979): and long-term variations of plasma hormones within women. Thus, the estimated validity coef- (13) ficient (ртX) for a single measure of total estradiol бT= 1_бr - Рx - zX 6Xz Ртх (X) as a measure of average estradiol over 1-2 years б (T), based on equation (14), is 0.71, and it is 0.88 or equivalently for percentage unbound estradiol. Based ors equation (13), the results in the last (14) section on the effects of measurement error could Р тк = ~Рx have been (and often are) expressed in terms of рX These equations state that the reliability coeffi- rather than рTк. For example, equation (11) can be cient, Px' is equal to the square of the validity coef- written: ficient of X, ртх' This result is important, because it shows that if the assumptions are correct, the relia- ОRo = ORr (15) bility coefficient, which is a measure of the correlation between two imperfect measures, can be used Examples of the bias in the odds ratio from var- to estimate the correlation between T and X, with- ious degrees of unreliability are given in Table 2. out having a perfect measure of T. The correlation of For example, a biomarker with a reliability coeffi- the replicates of X is less than the correlation of X cient of 0.64 (under the parallel test model) would with Tbecause each replicate has measurement error. attenuate a true odds ratio of 4.0 to an observed А reliability study of a Ыоmaтker test can often odds ratio of 2.4. When Px is substituted for рTX in be assumed to have equal and uncorrelated errors the attenuation equation, the equation applies if (1) the replicates are sampled over the entire time only when the meaning of the reliability coeffi- period to which the true biomarker is intended to cient is restricted to the correlation between paral- relate; (2) the specimen handling, storage and ana- lei measures of T. However, the term 'reliability lytical techniques vary in the reliability study as coefficient' is often used to refer to the correlation they will in the parent study; and (3) the true expo- between repeated measures, рX, even when the sure is defined as the mean measure over the rele- assumptions of parallel tests do not hold. vant time period of repeated measures of the assay. А study by Toniolo et aL (1994), which exam- Relationship between reliability and validity when ined the reliability of serum hormone levels in 77 the errors are correlated postmenopausal women, provides a good example. 1n real reliability studies, the assumptions of parallel Selected women in a prospective study of serum tests are often incorrect. One assumption of the estrogens and breast cancer zisk who had blood model of parallel tests that is often violated is the drawn 2-3 times over a 1-2 year period were assumption of uncorrelated errors. Often, рЕIЕ2 > 0 included in the reliability study. The blood was In other words, the error in one measure is posi- stored and analysed as in the parent study. The reli- tively correlated with the error in the other. ability coefficient (intraclass correlation coeffi- Correlated errors occur when the sources of error cient) was 0.51 for total estradiol and 0.77 for per- in the first measurement on a subject tend to centage unbound estradiol. The repeated measures repeat themselves in the second. For example, if, in in this study are close to parallel test model: the a reliability study, blood was drawn once on each errors on each of the repeated measures can be subject and analysed twice in different batches, assumed to be equal because the same test proce- and the true marker of interest were mean dure was repeated, and the errors are likely to be R-carotene for the 2 years surrounding the time of independent (i.e. a woman whose hormone mea- measurement, there would be correlated error. This sure was higher than her 'true' average value on is because a person whose p-carotene level on the

87 Application of Biomarkers in Cancer Epidemiology first measure was higher than their true mean level dent (Walker & Blettner, 1985): (perhaps due to a seasonal variation in intake of (3- carotene) would also be likely to have a J3-carotene рn 0), then the validity coefficient study, each matched set of cases and controls were is less than the square root of the reliability coeffii- to be analysed within a batch so that batch differ-

Equation From equation Differential or non-differential Binary biomarker, two-group comparison True difference = P, -. Р2 2) ORт= Рi (f - Р (1 - 1) Рг Р (1-3) Differential non differential where P - 1 + s ес 1 séns +:s eс 1) апд = (p -1 + s ec Р2 2 р г)1(sen52 + spec2 -1) )1(sens'+ spec -1) P1 - Рг = (Р1 - F2 (4) Non-differential Continuous biomarkerT two group comparison !rr - ( Differential = (, - ~кZ} - (b1 - b2) 7) b - ОRr= е тwhere 3 = I(.1 b2 lI Ro l- Differential 1 !Р (S) (9) . Non differential цт, - Њ, = , - µ ," ОRT= (11) Non differential

ORT= ОR!°x (15) Non differential Continuous biomarker; continuous outcome Non differential ВтXтr РхуKPтКх X.Рr,(У) (12)

аSee text for the notation the assumptions used iп derivation of equations and the. interpretation of equations.: .

88 Effects of biomarker measurement error ences would not contribute to measurement error. The coefficient of variation, ' provides only С 4. Example of adjustment for effects limited information about measurement erroi ТаЫе of differential measurement error because the CV is the ratio of вF to X, but it is the ratio of E to ~ (as in equation 13 above) that is needed т в Serum cholesterol ("mollI) to understand the impact of measurement error. X T' Adjusting the results of epidemiologic studies Case—control study for biomarker error (parent study) The emphasis of this chapter has been an expla- Cases (n = 43) 5.56 nation of how the observable odds ratio or other Controls (n = 43) 6.47 measure of association in the parent epidemiological study is a function of the measurement error in Validity study the biomarker and the true measure of association. Cases (n= 16) 5.75а .6.21. However, the same models can be applied in the Controls (n = 16) 6.30a 5.72 opposite direction. By use of estimates of the biomarker measurement error from a validity or relia aData created to serve as an example. bility study, it is possible to adjust the observed measure of association from the parent epidemiological study to yiеlд an estimate of the true associa- unbiased, even if T is not perfectly precise.) A corn- tion. However, because the assumptions used in the panson of X with T for these 32 subjects could serve models cannot be known to be correct (see below), as a validity study to estimate bias. The bias in X the results of the adjustment equations should not among the cases could be estimated from the validity be considered to be the `true association', but rather study (with X, T serving as estimates of µx, µT) as: an indicator of the degree of bias in the observed odds ratio or other measure of association. b=X1 — T7 =5.75-6.21=-0.46, These `adjustment' equations and the equations from which they were derived are given in Table 3. i.e. serum cholesterol at the time of diagnosis Because under non-differential measurement error underestimates true serum cholesterol by 0.46 the observable association is attenuated from the among cases. The bias in X among the controls true association, these adjustment equations are could be estimated from the validity study as: called `deattentuation' equations when non-differential measurement error is assumed. b2 = — = 6.30-5.72 = 0.58, Data from a study by Winawer et al. (1990) are adapted as an example of adjusting for the effects i.e. serum cholesterol (at time of diagnosis of cases) of differential measurement error in a two-group overestimates true serum cholesterol by .0.58. study of a continuous biomarker. In this study, Clearly, there is differential bias. serum cholesterol at the time of diagnosis (X) was To correct the observed difference in serum cho- compared between 43 cases with colon cancer and lesterol in the parent study for the effects of this 43 controls (see Table 4). differential bias, the correction equation from The observed case—control difference in serum Table 3 based on equation (7) is used, with X and cholesterol was: T serving as estimates of µx and г : = (-0.91) — (-0.46-0.58) = 0.13 i — 5.56 —6.47 =-0.91 (!.LT1 — г 2) Thirty-two subjects (16 cases and 16 controls) had This suggests that there is little difference between serum cholesterol measures available from 10 years colon cancer cases in 'true' (before disease) serum prior to diagnosis, which can serve as T because it cholesterol. is unlikely to be biased. (As noted earlier, the bias As an example of de attenuation under non- in X can be estimated by comparison with Tif T is differential measurement error, suppose a cohort

89 Application of Biomarkers in Cancer Epidemiology

study of occupational exposure to chlorophenates, errors in the repeated measures. Then, if equation assessed by urinary chlorophenate concentration, (15) is applied, this leads to a conservative (closer in relation to soft tissue sarcoma yielded an odds to the null) estimate of the true odds ratio under ratio of 1.3 for each 100 µg/L increase in concen- non-differential measurement error. Iп other tration. Suppose also that a validity study among words, the estimate would only be 'deattevuated' a subset of the subjects yielded an estimate of 0.6 for the random part of measurement error, and not for the validity coefficient, ртx, between urinary for the part that was repeated across measures in cholorophenate concentration and industrial the reliability study. records of exposure (assumed here to be a near-per- While adjustment procedures may aid in under- fect measure). Then information from these two standing the results of a study, caution should be studies could be used to adjust the observed odds exercised in interpreting these results. First, the ratio to yield an estimate of the true odds ratio assumptions used in the derivation of the equations (based on the equation in Table 3 derived from in Table 3 may not be appropriate. In particular, an equation (11), provided the assumptions hold): assumption of non-differential measurement error could be incorrect, so it is preferable if the biomarker . ORS = (1.З)zlо ь' = 2.1 measurement error can be assessed separately for the two study groups to account for differential This suggests that biomarker measurement error misclassification. For continuous exposures, the may have led to the weak observed association assumptions of the simple measurement error between the biomarker and the disease, because model, normality of the biomarker and its error, and the observed odds ratio is consistent with a true the logistic model of disease—biomаrkеr relationship odds ratio of 2.1 for each 100 цg/1 increase in often also fail to hold. Second, both the observed chiorophenate concentration. measure of association between the biomarker and Information from reliability studies can also be outcome and the estimated measurement error have used in adjustment procedures, to the extent that sampling errors; this needs to be considered in esti- the reliability study provides information about mating the true association. Third, the estimates the validity of the exposure variable. If the relia- of the measurement error should be estimates from bility study were of two parallel measures, рX could the same population(s) as the parent study to be be substituted for р fx as in the equation in Table 3 corrected, yet such estimates may not be available. derived from equation (15). As an example, the Finally, the presence of covariates modifies the prospective study by Toniolo et al. (1995) of serum effect of biomarker measurement error. Information hormone levels and breast cancer risk among 7063 on the multivariate measurement error structure of postmenopausal women (among whom 130 devel- the biomarker and covariates is required in order to oped breast cancer), observed an odds ratio of 1.8 correct fully for measurement error. Therefore, for total estradiol >44 pg/ml (versus <20) and an unless these issues have been accounted for, the odds ratio of 2.0 for percentage unbound estradiol emphasis of the adjustment procedure should be >1.53% (versus <1.20%). These odds ratios can be on interpretation of the observed estimate of corrected for the effect of measurement error, effect, and not of the corrected estimates. based on the reliability coefficients of 0.51 and A great deal of work has been done on adjust- 0.77 for total estradiol and percentage unbound ruent procedures, or, more generally, procedures estradiol, respectively, from the reliability study that incorporate information from a validity or described in an earlier example. Using equation reliability study into the statistical analysis of the (15), the estimated 'true' odds ratio is 3.2 for total disease—exposure relationship (Tenebein, 1970; estradiol >44 pg/m1, and 2.5 foi percentage Barron, 1977; Copeland etaL, 1977; Prentice, 1982; unbound >1.53%. This suggests that total estradiol Greenland & Kleinbaum, 1983; Clayton, 1985; might be strongly related to breast cancer risk, but Stefandn & Carroll, 1985; Whittemoie & Grosser, its effect is obscured by the error in its measure- 1986; Espeland & lui, 1987; Fuller, 1987; ment. Armstrong etal., 1989; Rosier etal., 1989;Qizilbash Reliability studies yield only an upper limit for et al., 1991; Pepe & Fleming, 1991; Carroll et ai., the validity coefficient when there are correlated 1995). Many of these procedures take into consid-

90 Effects of biomarker measurement error eration some of the issues discussed above; in par- for binary markers and equation (8) for continuous ticular, some make less restrictive assumptions biomarkers). Generally, very little differential error about the error model, yield confidence intervals is acceptable. that incorporate sampling error from the reliability When differential error is unlikely to be a prob- study, and/or allow a multivariate measurement lem, the researchers should focus on assessment of error structure. These methods may prove to be use- the non-differential error, or at least some of the ful in accounting for measurement error. major components of error in the biomarker, To assess the total error (the laboratory variation, Summary variation from specimen collection and storage, Before embarking on an epidemiological study and variation from short-, medium- and long-term which uses a biomarker, it is extremely important biological variability), one would ideally conduct to understand the measurement error in the bio- a validity study in which the biomarker to be used marker. For a binary biomarker, the validity (the is compared with a perfect (true) measure. Then, measurement error in а population) of the bio- the effect of non-differential measurement error marker is quantified by its sensitivity and speci- on the odds ratio can be estimated by equations (1) ficity. For a continuous biomarker, X, the validity aid (3) for binary markers and equation (11) for can be estimated by the bias X— T and by the valid- continuous markers. For continuous markers ity coefficient, рхт (correlation coefficient if X (assuming a simple measurement error model), the with T), a measure of precision. These estimates of effect of non-differential measurement error measurement error ideally utilize a true measure of depends only on the validity coefficient and not the biomarker (T) which has no sources of error on the bias (see equation 11). aid which integrates the biomarker over the time Because validity studies are rarely possible, it is period of etiological interest (geпeraIly years). To important to understand that one can measure assess whether the error is differential between much of the total error in a well-designed геliabil- cases and controls, separate studies on a group of ity study. Much of the error can be measured if one cases and a group of controls are needed. collects and analyses two (or more) specimens on The researcher's first concern should be to rule a group of subjects in such a way that the error in out, as far as is practicable, the possibility of differ- one estimate is not repeated on another, e.g. the ential measurement error. Because differential two specimens are collected at different times over measurement error can bias the odds ratio in any the relevant etiological time period and are han- direction, the presence of differential error in a bio- dled, stored and analysed with the same degree of marker can invalidate the epidemiological study. variation (different specimen collectors/laboratory Differential measurement error is a particular con- technicians/batches) as would occur in the parent cern in case—control studies (and among the early epidemiological study. Reliability studies are gen- cases in cohort studies) when the biomarker is not erally analysed by the kappa coefficient for binary a fixed marker (e.g. genotype) and therefore could be variables and by the intraclass correlation coeffi- influenced by pre-clinical disease, by the physical cient for continuous variables (see Fleiss, 1981, 1986; or emotional effects of the disease after diagnosis, Armstrong et al., 1992). The intraclass coefficient and/or by treatment. Assessment of differential provides information about the validity coefficient error requires pre-diagnostic specimens on a sample (see equations 14 and 16) and therefore provides of cases and comparable early specimens on con- information about the degree of attenuation of the trols. These serve as the 'true' markers (without odds ratio due to non-differential measurement differential bias) for computation of sensitivity and error. The degree of acceptable measurement error specificity separately for cases and controls. For would depend on the magnitude of the true odds continuous variables, differential bias has the most ratio (see equation 15), but generally an intraclass untoward effects; this can be estimated by com- correlation (including all sources of errors) of less paring X — T for cases with X— T among controls. than 0.5 would not be acceptable. Whether the degree of differential error is acceptable Whether or not the total error can be studied, can be determined by estimating the effects of the one should attempt to estimate some or all of the error on the odds ratio (see equations (1) and (2) components of error. The epidemiologist and

91 Application of Biomarkers in Cancer Epidemiology laboratory scientists should develop studies of the Banon, S.A. (1977) The effects of misclassification on the laboratory error and other components of error. To estimation of relative risk. Biometrics, 33, 414-418 assess the laboratory error, a binded test—retest Bohrnstedt, G.W. (1983) Measurement. In: Rossi, P.H., reliability study on spIlt samples from a single Wright, J.D. & Anderson, A.B., eds, Handbook of Survey specimen froze each of a group of subjects, Research, Orlando, FL, Academic Press, pp. 70-121. analysed in separate batches and by different labo- Bross, 1. (1954) Misclassification ir► 2 x 2 tables. ratory technicians (in a way that reflects the design Biometrics, 10, 478-486 of the parent epidemiological study), would yield Carroll, R.J., Ruppert, D. 5tefanski, L.A. (1995) an intraclass correlation coefficient that measures Measurement Error in NoпIinear Models, London, the laboratory component of error. Similarly, other Chapman & Hall reliability studies could. be designed to test the Clayton, D. (1985) Using test-retest reliability data to effect of handling, storage and short-, medium- improve estimates of relative risk: an application of and long-term biological variation. Epidemiologists latent class analysis. Stat. Med., 4, 445-455 should not assume that a small laboratory error Cochran, W.G. (1968) Errors of measurement in statis- implies that a measure is good, because these later tics. Tecpnometrics, 10, 637-666 sources of error introduced by the design and needs of the epidemiological study could be far Copeland, K.T., Checkoway, J., McMichael, A.J. & greater than the laboratory component of the error. Holbrook, R.H. (1977) Bias due to misclassification in the estimation of relative risk. Am. J. EpidemioL, 105, Often, multiple sources of error can be partitioned 488-495 in a nested study design (see Dunn, 1989). When only some components of error are measured, the de K1erk, N.H., English, D.R. & Armstrong, B.K. (1989) A resulting intraclass correlation only provides an review of the effects of random measurement error on relative risk estimates in epidemiologic studies. fut. J. upper estimate of thé validity coefficient (see equa- EpidemioL, 18, 705-712 tion 16). However, by estimating the components of error, the researcher can seek to improve those Dunn, G. (1989) Design and Analysis of Reliabiliiy Studies, New York, Oxford University Press aspects having the most adverse effects. For example, enhanced laboratory quality control procedures Espeland, M.A. & lui, S.L. (1987) A general approach to would reduce laboratory error, or use of multiple analyzing epidemiologic data that contain misdassifica- specimens (over time) per subject would reduce the tion errors. Biometrics, 43, 1001-1012 error due to shift- or medium-term biological vari- Fleiss, J.L. (1981). Statistical Methods for Rates end ation (see Armstrong etaL, 1992). Proportions, 2nd edn, New York, Wiley, pp. 188-236 Finally, after the parent epidemiological study Fleiss, J.L. (1986) The Design and Analysis of CliпicaI has been completed, estimates of biomarker mea- Ехpеrimеpts, New York, Wiley suierrient error can be used to adjust the results of Fuller, W.A. (1987) Measurement error Models, New York, the parent study for the effects of measurement Wiley error (or for those components of error that have Goldberg, J.D. (1975) The effects of misclassification on been measured) (see Table 3). Cautious use of these the bias in the difference between two proportions and methods can add insight into the bias caused by the relative odds in the fourfold table. J. Am. Stat. Assoc., biornarker measurement error on the results of an 70, 561-567 epidemiological study. Greenland, S. & Kleinbaum, D.G. (1983) Correcting for misclassification in two way tables and matched-pair studies. lot. J. EpidemioL, 12, 93-97 References Gregorio, D.I., Marshall, J.R. & Zielenzyny, M. (1985) Armstrong В.G., Whittemore, A.S. & Howe, G.R. (1989) Analysis of case-control data with covariate measure- Fluctuations in odds ratios due to variance differences ii-i ment error: application to diet and colon cancer. Stat case-control studies. Am. J. Epidemiol., 121, 767-774 Med., 8, 1151-1163 . Gu11en, W.H., Bearman, J.E. & Johnson, E.A. (1968) Effects of misclassification in epidemiologic studies. Armstrong, B.К., White, E. & Saracci, R. (1992) Principles of Exposure Measurement in Epidemiology, Oxford, Oxford Public Health Reports, 83, 914-918 University Press, pp. 49-136 Hansen, 1.1., Gurwitz, W.N. & Bershad, M. (1961)

92 Hlects of biomarker measurement error

Measurement errors in censuses and surveys. Bull, fit. Tenenbein, A. (1970) A doubling sampling scheme for Stat.Inst., 38, 359-374 estimating from binomial data with misclassifiction. J. Am. Stat. Assoc., 65, 1350-1361 Юeiнbаum, D.G., Kupper, L.L. & Morgeristern, H. (1982) ЕрiдетгоIоgгс Research, Belmont, CA, Lifetime Learning Toniolo, P., Koenig, K.L., Pasternack, B.S., Banerjee, S., Publications, pp. 183-193, 220-241 Rosenberg, C., Shorе, C.E., Strax, P. & Levitz, M. (1994) Reliability of measurements of total, protein bound and Kotz, . & Johnson, N.L. (1988) Encyclopedia of Statistical Ѕ unbound estradiol tri serum. Cancer Epidemiol. Sciences, vol. 9, New York, Wiley, pp. 662-663 Biomarkers Prev., 3, 47-50 Law, M.R. & Thompson, S.G. (1991) Low serum choles- Toniolo, P.G., Levitz, M., Zelenluch-Jacquotte, A., terol and the risk of cancer: an analysis of the published B nerj , S., Koenig, K.L., Shore, RE, Strax, P. & prospective studies. Cancer Cases Control, 2, 253-261 а ее Pasternack, B.S. (1995) A prospective study of endo- Lord, P.M. & Novick, M.R. (1968) Statistical Theories of genous estrogens and breast cancer in postmenopausal Mental Test Scores, Reading, MA, Addison-Wesley women. J. Nail Cancer Inst., 87, 190-197 McKeown-Eyssen, G.E. & Tibshirani, R. (1994) Walter, A.M. & Blettner, M. (1985) Comparing imperfect Implications of measurement error in exposure for the measures of exposure. Am. J. Epidemiol., 121, 7 83-90 sample sizes of case-control studies. Am. J. Epidemiol., White, E., Kushi, L. &Pepe, M. (1994) The effect of expo- 139, 415-421 sure variance and exposure measurement error on study Newell, D.J. (1962) Errors in the interpretation of errors sample size: implications for the design of epidemiologic in epidemiology. Am. J. P . Health, 52, 1925-1928 иЫ studies. J. СIгniсаl Epidemiol., 47, 873-880 Nunnaly, J.С. (1978) Psychometric theory (2nd edn). New Whittemore, A.S. & Grosser, S. (1986) Regression meth- York, McGraw-Hill, pp. 190-225 ods for data with incomplete covariates. In: Moolgavkar, Prentice, R.L. ,(1982) Covariate measurement errors and 5.1, & Prentice, R.L, eds, Modern statistical methods in parameter estimation in a failure time regression model. chronic disease. New York, Wiley, pp. 19-34 Biometrika, 69, 331-342 Winawer, S.J., Fiehinger, B.J., Buchalter, J,, Herbert, E. & Shike, M. (1990) Declining serum cholesterol levels prior Qizilbash, N., Duffy, S.W. & Rohan, Т.E. (1991) Repeat measurement of case-control data: correcting risk esti- to diagnosis of colon cancer. J. Am. Med. Assoc., 263, mates for misclassification due to regression dilution of 2083-2085 lipids in transient ischemic attacks and minor ischemic Wu, M.L., Whittemore, A.S. & Juпg, D.L. (1986) Errors in strokes. Am. J. Epidemiol., 133, 832-838 reported dietary intakes: I. Short-term recall. Am. J. Rosner, B., Willett, W.C. & Spiegelman, D. (1989) Epidentiol., 124, 826-835 Correction of logistic regression relative risk estimates and confidence intervals for systematic within-person measurement error. Stat. Med., 8, 1051--1069 Schiffman, M.H. & Schatzkin, A. (1994) Test reliability is critically important to molecular epidemiology: Ass example from studies of human papillonra infection and E. White cervical neoplasa. Cancer Res., 54, 1944s-1947s Cancer Prevention Research Program, Fred Hutchirnson Cancer Research Center, Stefanski, L.A. & Carroll, R.J. (1985) Covariate measure- 5eattle, WA 98109, USA ment error in logistic regression. Ann. Stat., 13, 1335-1351

93 Application of Biomarkers in Cancer Epidemiology zonioln, P.' Boffetta, P., shuknr, DEC., Rothman, N., Нцlkа, Band Pearce, N„ eds lARC SnienUfic Pablicat]ona No. 142 International Agency for Research on Cancer, Lyon, i 997

Markers of internal dose: chemical agents

D. Coggon and M.D. Friesen

Biomarkers of internal dose measure the level of a carcinogen or one of its metabolites in a tissue or a body fluid such as urine or blood. The choice of a biomarker of internal dose for a particular epidemiological study or type of study requires careful consideration of the period of exposure to which the biomarker relates, host factors related to carcinogen metabolism, in- vasiveпess of sampling, reliability and cost of the biomarker. Before a new biomarker is adopted, it is important to assess these characteristics in transitional studies to ensure that the biomarker will be applied appropriately, Biomarkers of internal dose have been applied most successfully in ecological studies and nested case—control studies, and are especially useful when they provide information about long-term carcinogen exposure.

Human exposure to carcinogens can be estimated been applied in epideiniological studies. The appli- by measuring available levels in food, water or air. cability of a binmarkgiv of internal dose to cancer Air levels, for example, have been widely used in epidemiology depends on certain features which industrial hygiene to asses industrial exposure to need to be characterized in transitional studies carcinogens. Measuring the external exposure, (Halka, 1991) before application in the field. however, does not necessarily provide accurate information about the amount of carcinogen асtu- To what period of exposure does the biomarker a11y absorbed by an individual, a parameter which is relate? of great importance to epidemiologists. This para- Although biomarkers of internal dose generally meter, which has been called the `internal dose', give information about recent exposure, the half- reflects total exposure by all routes (inhalation, life of a carcinogen or its metabolite in human ingestion and dermal absorption) and is influenced body fluids can vary from less than an hour to more not only by the level of carcinogen in external media than 10 years (Fig. 1). For example, the level of uri- (air, food, water, etc.) but also by factors such as the nary methylhippuric acid, which has a half-life of amount of air breathed or food consumed by the only a few hours, is useful in occupational settings individual during the period of exposure and the as a biomarker of recent exposure to xylene (Inoue efficiency of absorption from the lung or gut. et al., 1993; Huang etal., 1994), while urinary cad- This chapter will examine the use of bioniarkers mium, which is related to body burden with an of internal dose to chemical agents. Other types of excretion half-life of over 10 years, is a useful long- biomarkers, which measure the biologically effective term biomarker of exposure to cadmium (Ghezzi dose or early biological effects of the carcinogen, are et aL, 1985). Biomarkers of internal dose are being also currently available or under development. increasingly used in the field of industrial hygiene, Several excellent books have reviewed the use of all where validated biomarkers for measuring occupa- these kinds of biomarkers of human carcinogen tional exposures to a wide range of carcinogens in exposure (Hulka et aL, 1990; Groopman & Skipper., blood, urine or breath have been proposed by 1991; Armstrong etal., 1992; Schulte &Perera, 1993). national industrial hygiene authorities (Table 2). In general, it is past or long-term exposure that Biomarkers of internal dose is of interest in cancer epidemiology; however, a Biomarkers of internal dose measure the level of the Ыоmarker that reflects recent exposure may still be carcinogen or one of its metabolites in a tissue or of importance and provide information superior to body fluid such as urine or blood. Table 7 gives that obtainable with traditional methods of expo- examples of how biomarkers of internal dose have sure assessment. For habitual or repeated exposures

95 Application of Biomarkers in Cancer Epidemiology

Table 1. Examples of bïomarkers of internai dose applied in еpidеmиobgЁcü1 studies

S • -

Aflatoxin B, and metabolites Urine Aflatoxin in the diet Liver Ross et at, t992; Qian et a1., 1994 Dioxins and furans Blood fat Herbicide production All organs Flesch Janys et at, 1995 •Lead Blood Lead battery production Gastrointestinal tract Gerhardsson et at., 1995 N-nitrosamino acids Urine Endogenous formation Oesophagus Wu et at., 1993 Arsenic Urine Copper smelting Lung Entetline & Marsh, 1982 Cotinine Urine Tobacco smoking Lung Dе Wаагд et аt, 1995. DDE Serum DDT Breast Wolff et at, 1993; Krieger et at, 1994

such as smoking or mycotoxins in the diet, for (РАН), and the heterocyclic aromatic amines 2- example, а small number of measurements with a amino- l-methyl-6-phenуlimidazo [4, 5-b]pyridine biomarker of recent exposure may give а reason- (PhIP) and 2-amino-3,4-dimethylimidazo[4,5-л- able measure of longer-term exposure (Ross et al., quinoxaline (MeIQx) (Sinha et al., 1994) after con- 1992; Qian etaL, 1994; De Waard et aL, 1995). The sumption of charbroiled or pan-fried beef by use of such biornarkers of recent exposure to asses human volunteers. longer-term exposure requires transitional studies The situation is more complex in the case of to determine the appropriate number of samples agents to which exposure occurs by multiple and frequency of sampling. routes and at levels that are difficult to measure: One example of such a complex situation is expo- Now are bromarker levels affected by metabolic sure to РАН, for which uiinary excretion of 1- host factors? hydroxypyrene is a marker (Sherson et al., 1992; The level of biomarker measured in a tissue or Santella et aI., 1993). body fluid depends on metabolic factors specific to the individual under examination. These include Is sampling invasive? how the carcinogen and its metabolites are parti- Samples of urine and breath are fairly easy to tioned in elimination or storage compartments obtain, and in adults, blood sampling does not such as urine or adipose tissue, individual levels of usually pose ethical difficulties or cause major phase I and phase II enzymes, and the influence on problems with compliance. On the other hand, these enzyme levels of other substances to which some tissues may be more difficult to sample. For the subject is exposed. example, although the level of DDT accumulated For carcinogens occurring in food, where the in adipose tissue is an excellent measure of long- intake in a normal diet can be accurately determined, term exposure to the pesticide, it would be hard it is feasible to validate exposure biomarkers in to obtain samples from large numbers of people studies with human volunteers. This approach is who were not already undergoing surgery. being used in transitional studies to characterize However, blood levels of DDE, a metabolite of int.erindividual differences in the urinary excretion DDT, have been used to measure long-term expo- of 1-hydroxypyrene-glucuronide (Kang et аL, 1995), sure to DDT (Wolff et al., 1993; Krieger et al., a metabolite of polycyclic aromatic hydrocarbons 1994).

96 Markers ы internal dose: chemical agents

♦Dichloromethane ♦ Blood • Styrene • Urine • Trichloroethylene ♦Carbon disulphide • Trichlorotriiuoroethane •Benzene • Methylhippuric acid (xyiene) • Mandelic acid (styrene) • Toluene ♦Carbon monoxide • Thiothiazоlidine-4-carboxylic acid (Cs2) ♦1,1,1-Trichloroethane • Trichloroethanol (Trichloroethylene) ♦Perchloroethylene • Chromium • Ethoxyacetic acid (ethoxyethanol) • Nickel ♦Mercury • Trichloroacetic acid (trichloroethylene) • Arsenic ♦Lirtdane • Cobalt • Pentachlorophеnol • Mercury • Dieldrin Polychlorinated • Biphenyls • Cadium

O 1 hour 10 hours 1 week 1 year 10 years

Figure 1. Excretion half-life of various biomarkers ы internal dose in blood and urine (from data published by the UK Health and Sately Executive).

How much does the assay cost? Cohort studies relating exposure to cancer. Many assays for measurement of internal dose, Because cancers occur relatively infrequently, cohort such as radiоunmunаssay for the measurement of studies require large numbers of subjects. For use in urinary eotivin as an index of exposure to envi- a standard cohort study, biomarkers of internal dose ronmental tobacco smoke (De Waard et al., 1995), can be a good choice, if they are relatively cheap can be carried out on a large scale at a relatively and non-invasive and if they reflect exposure over low cost. Other markers such as measurement of a relevant time period—usually months or years. An dioxins in blood fat (Flesch Janys et aI., 1995) are example is the use of bloo.d lead measurements in a still very expensive and thus applicable only to cohort study of occupational exposure to lead among small numbers of samples, perhaps in a subset of a lead smelter workers (Gerhardsson et al., 1995). cohort under study. For biomarkers that are more expensive to measure, an alternative is to use a nested case-control Use of biomarkers of internal dose in design. Here, biological samples are collected from epidemiological studies ai cohort members at baseline and stored. An Biomarkers of internal dose should be evaluated important issue in investigations of this kind, within the framework of epidemiological study therefore, is the stability of the markers during design (Rothman et al., 1995). storage (Riboli et al., 1995). Collected samples are

97 Application of Biomarkers in Cancer Epidemiology

analysed only for those subjects who go on to as a marker for DDT exposure in a case—control develop a cancer of interest and for suitably chosen study of breast cancer in the United States of controls. This approach was used successfully in America (USA) (Wolff etal., 1993). This approach is Shanghai iii China (Ross etaL, 1992; Qian etaL, 1994) feasible because organoeh1orine compounds such where a nested case—control analysis showed highly as DDT or PCBs accumulate in adipose tissue with signifïcant associations between the presence of an elimination half-life of several decades. urinary aflafoxins, serum hepatitis B surface antigen However, a biomarker with a shorter half-life positivity and risk for hepatocellular carcinoma may still be applicable in case—control studies, pro- (НСC). A cohort analysis including the same cases of vided patterns of exposure remain fairly constant ICC revealed no strong or statistically significant over time and are not modified by the develop- association between 1-ICC and dietary aflatoxin ment of disease. The potential for such use would consumption, as determined using food frequency be greatest in studies involving early, asympto- questionnaires and aflatoxin levels in local foods. matic cases of cancer (Rothman et a1., 1995).

Case—control studies relating exposure to cancer Ecological studies relating exposure to cancer In this case the biomarker should reflect exposure In ecological studies, exposures and disease оиt over many years and must not be altered as a con- comes are compared in populations or groups of sequence of the disease process or its therapy, Опе people. Por example, levels of N-nitrosaniino acids such example is the measurement in serum of DDE in the overnight urine of men living in 69 counties

'ARC carcinogenicity ACGII,1992 DFG,1992 Fill 1993 UK HSE, 1991 . evaluation Group 1 Benzene. U-phenol U pheпof,B boizene B-benzene B-benzene Ethylene oxide B-ethylene oxide Vinyl chloride U-thiodiglycolic acid Arsenic U-volatile As compounds U-Asэ+ ASS+ U-As3+, As5+, MMA+ DMA Cadmium U-Cd B-Cd U-Cd, B-Cd U-Cd, B-Cd U-Cd, B-Cd Chromium U-Cr U-Cr U-Cr B-Cr, U _ .. -Cr . Nickel .0-Ni U-Ni U-Ni

Group 2a MICA , U-MICA Tetratrachforoethylene B tetrachloroethylene B-tetrachloroethylene B-tetrachloroothylene . Tnbhloroethylene U TCA U-TCA, U-TCA, U-TCA B-tпchloгoethaпol B-tгichloгoethaпol U-trichloroethanot Group 2b _ Cobalt U Co U Co U Co Lead B-Pb, U Pb, B ZPP B-Pb,'U-ALA B Pb, B ZPP B-Pb, U ALA, B ZPP

Abbreviations: GI , American Conference it Governmental Industrial Hygienists; DFG, Deutsche Fo АС Н гsçhungsgémeinschaft; пОН Fпnish institute of Occupational Health; UK HSE, United Kingdom Health and Safety Executive; U, urine; B, blood lIA monomethylarsinFc acid; DMA, diniethylarsinic acid; MICA rtiethyEene bis(2-chloroaniline); TCA, trichioroacetic acid; ALA;$ aminole vulinic acid; ZPP, erythrocyte zinc protoporphyrin

98 Markers of internal dose: chemical agents

in China have been geographically associated with For example, in a retrospective study of people oesophageal cancer mortality (Wu et al., 1993). For manufacturing phenoxy herbicides and chloro this type of study, some imprecision in the mea- phenols (Manz et al., 1991; Flesch Janys et al., surement of biomarkers can be tolerated, provided 1995), exposure to dioxins was assessed from job that a relatively large number of samples are mea- histories and information about processes and sued in each geographical unit, since analysis is operating methods at the plant where they based on estimates of population means. worked. Blood and adipose tissue levels of dioxins, measured in a subset of the cohort, confirmed the Cross-sectional studies linking exposure with early departments with the highest exposure and markers of carcino,genesis allowed quantitative estimates of personal expo- Cross-sectional studies can be used to examine sure for all members of the cohort. whether suspected carcinogenic exposures are Studies have also been carried out to compare associated with early markers of carcinogenesis, exposure levels measured by different analytical such as chromosomal aberrations. Biomarkers of approaches. For example, urinary Ievels of 1- internal dose sometimes provide a convenient hydroxypyrene, a biomarker for exposure to РАН, measure of exposure for such investigations. For were shown to correlate well with levels of example, a number of studies looking at the asso- PAH-DNA adducts in the white blood cells of alu- ciation of styrene with cytogenetic damage in lym- minium workers (van Schooten et aI., 1995), but phocytes have used urinary mandelic acid as a not in foundry workers exposed to lower levels of marker for exposure (IARC, 1994). For a study РАН (Santella et al. 1993) . designed in this way, the bioznarker of exposure should also cover a period relevant to the early Conclusion marker of cardnogenesis (months or years). In summary, biomarkers of internal dose have found successful application in cancer epidemiol- Studies to assess the contribution of different ogy. The choice of a biomarker of internal dose for sources of exposure to total dose a particular epidemiological study or type of study Epidemiology may be used to investigate not only requires careful consideration of the period of expo- the risk of cancer from different levels of exposure sure to which the biomarker relates, host factors to a carcinogen, but also the importance of differ- related to carcinogen metabolism, invasiveness of ent sources of exposure in determining total dose. sampling, reliability and the cost of the assay. In this type of study, biomarkers of internal dose Before a new biomarker is adopted, these charac- which relate only to recent exposure are often the teristics must be assessed in transitional studies to most useful. For example, concerns have been ensure that it will be applied appropriately. expressed that clusters of leukaemia in the vicinity of petrochemical works might be attributable to References releases of benzene into ambient air. However, Armstrong, В.K., White, E. & Saracci, R. (1992) Principles studies by Wallace et ai. (1987) in the USA indicate of Exposure Measurement in Epidemiology, Oxford, Oxford that residence near petrochemical works has negli- University Press gible impact on personal doses of benzene De Waard, F., Kemmeren, J.M., van Ginkel, L.A. & (assessed by measurement in exhaled breath). Far Stolkег, A.A. (1995) Urinary cotinine and lung cancer risk more important is whether the individual smokes in a female cohort. Br. J. Cancer, 72, 784-787 or is exposed to environmental tobacco smoke and Enterline, P.E. & Marsh, G.М. (1982) Cancei among how he/she makes use of motor vehicles. workers exposed to arsenic and other substances in a copper smelter. Am. J. EpidemioL, 116, 895-911 Studies to evaluate other methods of exposure Flesch Janys, D., Berger, J., Gum, P., Manz, A., Nagel, 5., measurement Waltsgott, H. & Dwyer, J,H. (1995) Exposure to роlу- Even where a biomarker cannot be used on a large сЫоrйated dioxins and forais (PCDD/F) and mortality scale (e.g. because of expense or ethical con- in a cohort of workers from a herbicide-producing plant straints), it may still be useful in the evaluation or in Hamburg, Federal Republic of Germany. Am. J. refinement of other methods of exposure assessment. Epidemial., 142, 1165-1175

99 Application of Biorriarkers in Cancer Epidemiology

Gerhardsson, L., Hagmar, L., Rylander, L. & Skerfving, S. Henderson, BE., Wogan, G,N. & Groopman, J.D. (1994) A (1995) Mortality and cancer incidence among secondary follow-up study of urinary markers of alatoxin exposure lead smelter workers. Occup. Environ. Med., 52, 667-672 and liver cancer risk in Shanghai, People's Republic of China. Ghezzi, L, Toffoletto, F., 5esana, G., FagioIi, M.G., Cancer Epidemiol. Biomarkers Prev., 3, 3-10 Micheli, A., Di Silvestro, P., Zocchetti, C. & Alessio, L. Riboli, E., Haley, N.J., De Waard, F. & Saracci, R. (1995) (1985) Behaviour of biological indicators of cadmium in Validity of urinary biomarkers of exposure to tobacco relation to occupational exposure. Int. Arch. Occup. smoke following prolonged storage. mt. J. Epidemiol., 24, Environ. Health, 55, 133-140 354-358

Groopman, J.D. & Skipper, P.L. (1991) Molecular Ross, R.K., Yuan, J.М., Yu, M.C., Wogan, G.N., Qian, G.S., Dosimetry and Human Cancer: Analytical, Epidemiologi са', Tu, J,T., Groopmari, J.D., Gao, Y.T. & Henderson, В.E. and Sociаl Considerations, Boca Raton, CRC Press (1992) Urinary afiatoxin biomarkers and risk of hepatocellular carcinoma. Lancet, 339, 943-946 Huang, М.Y., Jin, C., Liu, Y.T., Li, B.H., Qu, Q.S., Uchida, Y., Inoue, O., Nakatsuka, H., Watanabe, T. & Ikeda, M. Rothman, N., Stewart, W.F. & Schulte, P.A. (1995) (1994) Exposure of workers to a mixture of toluene and Incorporating biomarkers into cancer epidemiology: a xylenes. I. Metabolism. Occup. Environ. Mcd., 51, 42-46 matrix of biomarker and study design categories. Cancer Hulka, B.S. (1991) Epidemiological studies using biolog- Epidemiol. Biomarkers Prey., 4, 301-311 ical markers: issues for epidemiologists. Cancer Epidemiol. Santella, R.M., Hemminki, K., Tang, D.L., Paik, M., Biomarkers. Prev., 1, 13-19 Ottman, R., Young, T.L., Savela, K., Vodickova, L., Dickey, Hu1ka, B.S., WiIcosky, T.C. &c Griffith, J.D. (1990) C., Whyatt, R., et al. (1993) Polycyclic aromatic hydro- Biological Markers in Epidemiology, Oxford, Oxford carbon-DNA adducts in white blood cells and urinary I - University Press hydroxypyrene in foundry workers. Cancer Epiderniol. Biomarkers Лгс"., 2, 59-62 ТАКС (1994)'ARC Monographs on the Evaluation of Carcinogenic Risk to Humans, Some Industrial Chemicals, Schulte, P.A. & Perera, F.P. (1993) Molecular Epidemiology: Vol. 60, pp. 233-320 Principles and Practices, San Diego, Academic Press lieue, O., Seiji, K., Kawai, T., Watanabe, T., Jin, C., Cat, Sherson, D., Sigsgaard, T., Overgaard, E., Loft, S., Poulsen, S.X,, Chen, Z., Qu, Q.S., Zhang, T. & Ikeda, M. (1993) H.E. & Jongeneelen, F.J. (1992) Interaction of smoking, Excretion of methylhippuric acids in urine of workers uptake of polycyclic aromatic hydrocarbons, and exposed to a xylene mixture: comparison among three cytochrome P450IA2 activity among foundry workers. xylene isomers and toluene. Int. Arch. Occup. Environ. Br. J, m d. Med., 49, 197-202 Health, 64, 533-539 Sinha, R., Rothman, N., Brown, E.D., Mark, S.D., Hoover, IPCS, (1993) Biomarkers and Risk Assessment: Concepts and R.N., Caporaso, N.E., Levander, O.A., Knize, M.G., Lang, Principles. (Environmental Health Criteria 155), Geneva, N.P. & Kadlubar, F.F. (1994) Pan-fried meat containing World Health Organization high levels of heterocyclic aromatic amines but low levels of polycyclic aromatic hydrocarbons induces Kang, DI., Rothman, N., Cho, S.H., Poirier, M.C., cytochrome P4501A2 activity in humans. Cancer Res., 54, Greenberg, A., Hsu, CI., Schwartz, B.S., Baser, M.E., 6154-6159 Weston, A„ Groopman, J.D. & Strickland, P.T. (1995) Interindividual differences in the concentration of 1- van Schooten, F.J., Jongeneelen, F.J., Hi11ebrand, M.J., hydroxypyrene-glucuronide in urine and polycyclic r - van Leeuwen, F.E., de Looff, AJ., Dijkmans, А.P., van а о Rooij, J.G., Den Engelse, L. & Kriek, E. (1995) olycyclic таtiс hydrocarbon-DNA adducts in peripheral white Р blood cells after charbroiled beef consumption. aromatic hydrocarbon-DNA adducts in white blood Camnogencsis, 16, 1079-1085 cell DNA aid 1-hydroxypyrene in the urine from aluminum workers: relation with job category and syn- Krieger, N., Wolff, M.S., Hiatt, R.A., Rivera, M,, ergistic effect of smoking. Cancer ЕргdеmгоI. Biomarkers Vogelman, J. & Orentreich, N. (1994) Breast cancer aid Prey., 4, 69-77 serum organochlorines: a prospective study among Wallace, L.A., Pellizzari, E.D., Hartwell, T.D., Sparacino, white, black, and Asian women. J. Nat' Cancer Inst., 86, 589-599 C., Whitmore, R., Sheldon, L., Zelon, Н. & Perritt, R. (1987) The TEAM (Total Exposure Assessment Manz, A., Berger, J., Dwyer, J.H., Flesch Janys, D., Nagel, Methodology) Study: personal exposures to toxic sub- 5. & Waltsgott, H. (199 I) Cancer mortality among work- stances in air, drinking water, and breath of 400 residents ers in chemical plant contaminated with dioxin. Lacet, of New Jersey, North Carolina, and North Dakota. 338, 959-964 Environ. Res., 43, 290-307 Qian, G.S., Ross, А.K., Yu, M.C., Yuan, J.M., Gao, У.T.,

100 Markers of internal dose: chemical agents

Wolff, M.S., Toniolo, P.G., Lee, E.W., Rivera, M. & Dubiri, N. (1993) Blood levels of оrgaпoсh1orine residues and risk of breast cancer. J. Nat[ Cancer Inst., 85, 648-652 Corresponding author Wu, Y , Chef, J., Ohshima, H., Pigatelli, B. Boreham, J., Q. Coggon Li, J., Campbell, T.C., Peto, R. & Bartsch, H. (1993) MRC Environmental Epidemiology Unit, Geographic association between urinary excretion of University. of Sоuthaтqtоn, N-nitroso compounds and oesophageal cancer mortality Southampton, UK in China. mL J. Caecer, 54, 713-719

101 Applicalioп of Biomorkirn in Canner Epidemiology Tnniolo, P., Bottetta, P., Shaker, D.E.G., Rghman, N., Hu1ka, Band Pearce, N. eds lARC Scientific Pabkcatiorna No, 142 lnlerпational Agency for Research оп Cancer, Lyon, 1997 Biochemical markers of dietary intake R. Kaaks, E. Riboli and R. Sinha

The primary objective of nutritional epidemiology is to identify, in combination with other forms of research, which aspects of diet and nutritional factors are causally related to cancer development. However, traditional epidemiology can evaluate with onIy a limited degree of specificity to which individual dietary factors an increased occurrence of cancer can be au ibuted.he two main reasons for this are: (1) dietary intake levels of specific foods or food constituentsг can be strongly intercorrelated; (2) dietary intake levels of specific food constituents are generally measured with rather large errors. Biochemical markers are increasingly seen as measurements that may help to overcome some of the above-mentioned methodological problems in nutritional epidemiology.

The most promising application of biomarkers of attractive class of measurement in epidemiological diet is in prospective cohort studies, where the studies. marker is unlikely to have been influenced by the disease, as biological specimens can be collected In prospective cohort studies, biomarkers of diet and stored well before cancer develops and can be used as an independent measurement of becomes clinically manifest. dietary intakes iп order to evaluate the association Two main types of markers of diet should be dis- between these intakes and disease risk, or as an tinguished: additional measurement of diet to be combined 1. Markers of absolute quantitative intake levels statistically (e.g. in latent variable models) with such as urinary nitrogen for protein. These mark- measurements obtained by questionnaire. This latter ers are based on a time-related balance between application is especially useful in validity studies to intake and output. estimate the magnitude of systematic and random 2. Markers based on the concentration of a specific errors of dietary exposure measurements obtained substance in biological fluids, or tissues (e.g. the from questionnaires. concentration of vitamins in blood plasma). Such The primary objective of nutritional epidemiol- markers have no time dimension, and provide only ogy is to identify, in interaction with other forms a correlate of dietary intake level. This class repre- of research, to which specific aspects of diet or sents the vast majority of Biomarkers of diet. other nutritional lifestyle factors (e.g. physical activity) observed variations in incidence of dis- The main advantages of biomarkers of diet are ease can be attributed. In theory by identifying as follows: causal factors with a highest possible specificity, • Biomarkers may provide more accurate тeаsите- accurate recommendations can be given about ments—or at least а correlate of intake levels of nutrition-related lifestyle factors, with the aim of specific chemical constituents—than the tradi- reducing the overall incidence of disease. Besides tional dietary assessment methods. providing insight into how to optimize nutritional • Biomarkers are objective measurements, the lifestyle patterns, it is hoped that epidemiological validity and precision of which are independent of research may also help to identify individual the subjects memory or capacity to describe foods chemical agents that, even when administered as consumed. In statistical terms, this means that the isolated compounds, may reduce cancer risk. measurement errors of biomarkers can be assumed Examples of agents for which such potentially to be uncorrelated with those of the questionnaire `chemopreventive' action has been postulated measurements. It is this statistical independence of include vitamins C, Е and A, carotenoids, these errors that makes biomarkers of diet such an flavonoids, indoles or phytoestrogens (Steinmetz

103 Application of Biomarkers in Cancer f pidemiology

& Potter, 1991; Wattenberg, 1992). 1n practice, The first class are markers based ôп knowledge however, epidemiological studies can evaluate about the metabolic balance between the intake with only a limited degree of specificity to which and excretion of specific chemical components; individual dietary risk factors (foods, nutrients or that is, the per cent recovery of the compound or other chemical food constituents) an increased its metabolites in excretion products (mainly urine occurrence of disease can be attributed. The two and breath) is known. All markers belonging to major reasons for this are: (1) intake levels of spe- this first category are time-related; that is, the cific foods, food groups or chemical food con- markers are based on a balance between intake and stituents (nutrients or non-nutrient substances) output over a representative time period, usually of can be strongly intercorrelated; and (2) the intake 24 h, and thus can be translated into estimates of levels of specific foods or food constituents are absolute intake level over 24 h as well. Probably generally measured with rather large errors. the best known example is the 24-h urinary excre- In nutritional epidemiology, biochemical mark- tion of nitrogen as a marker of average 24-h ргo- ers are increasingly seen as measurements that may tein intake (Bingham & Cummings, 1985). This help to overcome some of the above-mentioned marker is based on the knowledge that, in individ- methodological problems in nutritional,epidemi- uals who are in nitrogen balance (i.e., in practical ology. In this chapter, we discuss the potential uses terms, individuals who have a stable body mass and limitations of biochemical markers of dietary aid composition), the 24-h urinary nitrogen excre- intake levels in studies of the relationships tion represents a practically constant proportion of between diet and the risk of chronic diseases such nitrogen ingested (mostly in the form of protein). as cancer. Although much of this discussion will Likewise, the urinary excretion of potassium can apply to the use of bîomarkers in practically any be used as an indicator of potassium intake, even type of epidemiological design, including ecologi- though the percentage recovery in urine of cal correlation studies and case-control studies, we ingested potassium is more variable between indi- shall focus mainly on their use in prospective viduals than the recovery of nitrogen (Bingham et cohort studies. Prospective cohort studies have the al., 1992, 1995). A high intake of potassium can be advantage that the level of the marker is unlikely taken as an indicator of a diet rich in vegetables to have been influenced by the presence or and fruits (Williams & Bingham, 1986). A third, absence of disease, as biological specimens can be well-known example of a biomarker based on a collected well before disease develops or becomes metabolic balance between ingestion and excre- clinically manifest. А new generation of prospec- tion is the 'doubly labelled water' method to esti- tive cohort studies on diet and cancer, including mate average daily energy expenditure (Schoel1er, large banks of biological specimens (blood, urine, 1988). If, on average, a group of individuals is in toenails), is currently being conducted in different energy balance (i.e. there are no gains or losses in parts of the world (Toasiolo et cd., 1995; Berrino et body weight or composition), on average the daily cd., 1996; Riboli & Kaaks, 1997). energy expenditure is equal to daily energy intake; thus, the doubly labelled water method can be Types of biomarker of diet used to estimate dietary energy intake. Most markers currently in use are related to the The second class of markers of dietary intake are intake of a specific chemical compound ingested measured as concentrations of specific substances together with a food (Riboli et aI., 1987). These in: biological fluids (plasma, urine, saliva), specific compounds may be natural food constituents (e.g. tissues or cells (e.g. adipose tissue, white blood the traditional nutrients); they may be formed dur- cells), lipoproteins, cellular membranes, DNA and ing processing or treatment (e.g. heterocyclic specific proteins. Here, the word `concentration' amines formed in foods: Phi?, McIQ trarssfatty must be taken in a broad sense, in that the denoni- acids); or they may be contaminants from either a inator determining the concentration does not natural origin (e.g, aflatoxin) or a xenobiotic origin need to be measured in units of volume. Examples (e.g. DDT, PCBs). Prom a methodological point of of different types of 'concentration' are the frac- view, it is useful to distinguish two major classes of tion (ppm) of DNA bases with specific adducts (e.g. bhmarkers of diet. of PhIP), the fraction of amino acids in a given

104 8iochemicn1 markers of dietary intake type of protein (albumin, haemoglobin) that car- (i.e. using a questionnaire); nevertheless, replicate ries a specific type of adduct (PhIP, afutoxin), the measurements of the second type (written records concentration of vitamin E in lipoproteins or in or short-term recalls) can be used in smaller sub- the lipid phase 0f cell membranes, and the fatty studies for the `validation' and/or 'calibration' of acid composition of an adipose tissue biopsy. This the questionnaire measurements. Either type of second class of marker represents the vast majority measurement—whether based on a questionnaire of markers of dietary intake and, contrary to the about habitual food consumption or on the markers based on a balance over time between recording or recall of the actual consumption on intake and excretion, are without a time dimen- given days--can be biased as a result of unknown, sion. Also in contrast with the first class of mark- or unquantifiable, subject-associated factors such ers, the quantitative relationship between this sec- as a lack of memory, an incapacity to identify or ond class of markers and dietary intake level can- describe accurately types of food consumed, or a not be as clearly defined from knowledge about its tendency to deviate from the habitual diet on days physiology, and this quantitative relationship may that written records of food consumption are being vary between populations or even between popu- kept. When measurements are repeated by the lation subgroups, depending on the presence and same method, individuals may tend to make errors relative impact of determinants other than diet. of a similar size and magnitude on two or more The markers belonging to this second category can occasions. Therefore, errors may be correlated pos- therefore only provide a correlate of dietary intake itively, i.e. the measurements may have less than level; that is, they can provide a more or less accu- perfect validity (see below). Between different rate relative ranking of individuals by intake level types of dietary assessment method—e.g. between and/or disease risk, but cannot be translated questionnaires of habitual dietary intake and directly into absolute (recommended) intake lev- records or short-term recalls of actual dietary els. A list of examples of nutritional markers intake levels on specific days—errors are often belonging to this second category that are often assumed to be practically independent. This used in epidemiology is given in Table 1. assumption will be questionable, however, whenever some of the potential sources of error are the Advantages of biomarkers as measurements of same. For example, subjects may vary in their dietary intake motivation to provide complete answers to each To appreciate the potential advantages of bio- type of method, aid may thus underestimate chemical markers as measurements of diet, it is dietary intake levels to a similar degree by both useful to review briefly the characteristics and im- approaches (i.e. again resulting in a positive corre- itations of the more traditional measurements of lation of errors between the two types of measure- dietary intake level. The lattez can be divided into ment). assessments of an individual's habituai, long-term An initial reason for using biochemical markers intake of foods, by means of a structured ques- of dietary intake level is that they may provide a tionnaire or interview (Cameron & van Staveren, more accurate measurement (or at least a correlate) 1988; Willett, 1990a), and measurements of the of intake levels of specific chemical constituents actual consumption of foods on one or more spe- than can be obtained by the traditional methods. cific days by means of written food consumption The inaccuracy of the traditional methods may be records (Cameron & van Staveren, 1988 ) or by due, for instance, to the lack of reliable food com- means of short-term (usually 24-h) recalls (Witschi, position data. Food composition tables generally 1990). contain figures only of those chemical food con- Food composition tables are used to translate stituents ('essential nutrients') that provide energy either type of measurement of food consumption or that must be consumed regularly to prevent spe- into estimates of nutrient intake. In casе- сontrol cific, known deficiency diseases that would other- or cohort studies on the relationship between diet wise develop relatively quickly. Although in many and the risk of developing chronic disease, mea- countries databases on food composition are cur- surements of habitual exposure levels can usually rently being extended so as to include a wider be obtained only by the first type of measurement range of chemical compounds (i.e. extending

105 Application of Biomarkers in Cancer Epidemюlogy

Nutrients or food No. of Category of Biochemical Correlation Author groups subjects subjects index coefficient Vitamin A and carotenoids Roidt et al., 1988 Total preformed 302 men and All serum retinol 0.08 retinol women Serum a-carotene 0.00 Serum p-сагоtепе 0.03 Total vitamin A. 302 men and All Serum retinol 0.08 women Total dietary 302 men and 5егит retinol 0.02 carotenoids women Serum a-carotene 0.26 Serum p-carotene 0.21 Romieu etaL, Carotenoids 370 men and Non-smokers plasma R-carotene 0.37 1990 women

van't Veer et al, Fruits and vegetаыеs 140 women All (dietary history) Plasma carotenoids 0.22 1990, 1993

Coates et a!„ 1991 Total carotenoids 50 women Non-smokers Serum provitamin A 0.45 carotenoids Total carotenoids Ρ Serum Д -сатоtепе 0.43 Provitamin A Serum provitamin A 0.38 carotenoids carotenoids a-Carotene serum a-carotene 0.38 p-Carotene Serum u-carotene 0.32 Cryptoxanthine Serum cryptoxafrthine 0.36 Lycopene Serum lycopene —0.06 Lutein Serum 0.00 luteimzeaxanthin Total carotenoids 41 women Smokers Serum Arvitormn A 0.17 carotenoids Total carotenoids Serum 1-carotene 0.23 Provitamin A Serum provitamin A carotenoids carotenoids 0.17 a-Carotene Serum a-carotene 0.21 Ρ Д -Carotene Serum h-carotene 0.23 Cryptoxanthine Serum cryptoxanthine 0.18 ' Lycopene Serum !ycopene 0.00 Lutein Serum 0.08 lutein/zeaxanthiп

106 Biochemical markers of dietary intake

Nutrients or food No. of Category of Biochemical Correlation Author groups subjects subjects index coefficient Ascherio et a1., Carotenojfls 110 men Non-smokers Plasma a-carotene 0.52 1992 Plasma гсагоtепс 0.34 Plasma a plus 0.34 p-carotene Plasma lycopene 0.13 Plasma tutein 0.36 Plasma zeaxanthin 0,11 Plasma retinol 0.11 162 women Non-smokers Plasma a-carotene 0.37 . Plasma R-carotene 0.30 . Plasma a plus 0.30 p carotene Plasma lycopené 0.01 Plasma lutein 0.19 Plasma zeaxanthin 0.02 Plasma retinol 0.13

Jacques et a1., Vitamin A 42 men and 71 Non supplement Plasma retinol 0:02 . 1993 women users 55 men and 82 Supplement and Plasma retlnol 0.07 women non supplement users

Jacques et at, Carotanoids 53 men and 81 Non-supplement Plasma carotenoids 0.29 1993 women. users supplement and . Plasma carotenoids 0.37 non-supplement . users

Campbell et a1., . , Total fruits and 50 men and 49 All (FF0). Plasma lutein 0.39 1994 vegetables women Plasma p-carotene 0.44 Plasma lycopene —0.04 . Plasma cc-carotene 0.54 Plasma (3 carotene 0.43 Plasma sum of 0.54 carotenoids .

Hebert etal., 1994 Retinol 167 men and All Plasma retinol 0.14 women Plasma R-carotene —0.02 Total (i-carotвne Plasma u-carotene —0.17 Plasma retinol 0.29

107

Aрp1iсauoп of Biomarkers in Cancer Epidemiology

Nutrients or food No. of Category of Biochemical Correlation Author groups subjects subjects index coefficient Yong etal., 1994 tx-Carotene 98 women All (7-day records) Plasma a-carotene 0.59 p-Carotene Plasma R-carotene 115 j3-Cryptoxanthin Plasma 0.49 p-cryрtoхапthin Lutein Plasma lutein 029 Lycopene Plasma lucopene 0.41 Total Plasma total 0.51 a-Carotene All (FF0) Plasma a-carotene 0.52 . R-Carotene Plasma ji-саrotепе 0.44 R-Cryptoхanthm Plasma 0.30 (3-cryptoxanthin Lutein Plasma lutoin 0.29 Lycopene Plasma lutein 0.28 T0ki1 Plasma total 0.43

Porrini et a!, 1995 p-Carotene 11 men and All (FF0) Plasma (3-carotene –0.07 33 women All {7-day records} Plasma p-carotene 0.44

Enger eta!,, 1995 Carotenoids 215 men and All Plasma a-carotene 027 women Plasma ji-carotene 0.22 Plasma 0.35 p-cryрtoxanthin Plasma lutein + 0.33 zeaxarthin Lycopene 0.36

Kardinaal etal., Retinol 82 men and All (FF0) Plasma retinal 0.06 1995 R-Carotene women Plasma (3-carotene 0.15

Vitamin E Roidt вt а1., 1988 Vitamin E 50 women Non-smokers— Serum a-tacopherol 0.36 food source only Non-smoker—total Serum a-tocopherol 6.38 41 women Smokвrs—food Serum a-tocopherol 0.02 source only Smokers—total 5егum a-tocophera1 0.32 Romieu etal., Vitamin E 339 men and Non-supplement Plasma a-tocopherol 0.16 1990 women users

108 Biochemical markers of dietary intake

Nutrients or food No. of Category of Biochemical Correlation Author groups subjects subjects index coefficient Sinha of al., 1993 Vitamin E 65 men All Plasma a-tocapherol 0.32 46 men Nan-supplement Plasma a-tocopherol 0.17 user 19 men supplement user Plasma a-tocopherol 0.37 65 men All Plasma y-tocopheroI —0.33 46 men Non-supplement Plasma y.tocopherol —0.08 user 19 men Supplement user Plasma y-tocopherol —0.25

Jacques of a1., Vitamin E 41 men and 70 Non-supplement Plasma a-1oсaphetu1 0.35 1993 women users 55 men and 70 supplement and Plasma r-tocophérol 0.53 women non-supplement . users

Porrini etal., 1995 Total a-tocophérol 11 men and 33 All (measured by Plasma a-tocophérol —0.22 women FFQ) All (measured by Plasma a-tocopheroI 0.10 7-day record)

Kardinaal etal., roTocopherol 82 men and All (FFQ) Plasma a-totopherol 0.11 1995 a-Tvtaphero1: women Plasma a-tocophero1: 0.22 cholesterol ratio cholesterol ratio

Other vitamins Jacques of ai., Vitamin D 42 men and 66 Non-supplement Plasma 25-0Н 0.25 1993 women users vitamin D 55 men and 80 Supplement and Plasma 25-OH 0.35 women non-supplement vitamin D users Thiamin 42 men and 67 Non-supp[ement Red blood cell 0.01 women users thiamin 56 men and 82 Supplement and Red blood cell 0.02 . women non-supplement thiamin users Vitamin B2 42 men and 66 Non-supplement Red blood cell —0.21 women users riboflavin Supplement and Red blood cell —0.13 non-supplement riboflavin users 0.05 . Vitamin ВБ 42 men and 66 Non-supplement Red blood celI. ВБ (pyridoxine) women users

109 Application of Biomarkers in Cancer Epidemiology

Nutrients or food No. of Category of Biochemical Correlation Author groups subjects subjects index coefficient

Jacques et al. Vitamin 8 56 men and 82 Supplement and Red blood cell 8в —0.15 1993 (Contd) (pyridoxine) women non-supplement users Vitamin 612 40 men and 65 Non-supplement Piasma В12 0.19 (cyanocobalamine) women users 54 men and 7B Supplement and Plasma Вi2 0.35 women non-supplement users Plasma folate 40 men and 65 Non-supplement Plasma folate 0.61 women users Supplement and Plasma folate 0.63 non-supplement users

Wild et L, 1994 Folate 16 women Control subjects Serum folate 0.81 16 women Women with infants Serum folate —0.29 with neural tubé defect . 16 women Control subjects Red blood cell folate 0.72 . Women with infants Red blood cell folate —0.08 . with neural tube defect

Sinha et a1., 1994 Vitamin C 493 women From food Serum ascorbic acid 0.19 From food + Serum ascorbic acid 0.32 suppiement From food (one 5erum ascorbic acid 0.36 24-h recall) From food + Уегит ascorbic acid 0.56 supplément

Sinha et al., 1992 Vitamin C 68 men All Plasma ascoгЬc acid 0.43 Jacques of al., Vitamin C 39 men and 60 Non-supplement Plasma total 0.38 1993 women users : vitamin C 55 men and 80 Supplément and Plasma total 0.43 women non-supplement vitamin C . users

Porrini et a1., 1995 Ascorbic acid 11 men and 33 From food (FF0) Whole blood —4.22 . women ascorbic acid From food Whole blood 0.44 (7-day records) ascorbic acid ь

110 Biochemical markers of dietary intake

Author Nutrients or food No. of Category of Biochemical Correlation groups subjects subjects index coefficient Protein O'Donnell et a1, Protein 24 mon All (16-day Serum urea 0.21 1991 weighed records) Serum creatinine 0.12 i Unne nitrogen (24-h) 018 All (FF0) Serum urea -0.004 Serum creatinine —0.07 Urine nitrogen (24-h) 0.36 28 women Ail (16-day Serum urea 0.07 weighed records) 5еruт creatinine 0.39 Urine nitrogen (24-h) 0.19 All (FF0) Serum urea 0.09 5егит creatinine 0.12 Urine nitrogen (24-h) 0.28

Selenium Snook eta! 1983 Selenium 155 men and All (three 24-hr Plasma selenium : 0л2 . women recalls)

Yang etal., 1989 Selenium Аpprox. 150 3-day duplicate Whole blood 0.88 . families plate measured selenium for selenium Breast milk 0.90 24-hr urine 0.86

5waпson etaL, Selenium 24 men and Minimum of six Serum selenium 0.51 1990 20 women duplicate plates Whole blood selenium 0.55 for selenium Toenail selenium 0.53 measurement Urine Selenium 0.87 van 't Veer etal.,' Selenium 243 women All (dietary history) Plasma selenium 0.15 1990, 1993 239 women Erythrocyte selenium-0O5 360 women Toenail selenium —0.01

Other minerals O'Donnell eta! Sodium 24 men All (16 day Urine sodium 0.008 1991 Potassium weighed records) Urine potassium 0.19 . Zinc Serum ziпс . 0.08 Sodium All (FF0) Urine sodium 0.25 Potassium: Urine. potassium 0.08. Zinc Serum zinc 0 16 Applicatюn of 8iomarkers in Cancer Epidemiology

Author Nutrients or food No. of Category of Biochemical Correlation groups subjects subjects index coefficient O'Donnell etal. Sodium 28 women All (16-day п Urine sodium 0.09 1991 (Contd) Potassium weighed records) Urine potassium 0.44 Zinc Serum zinc —0.02 Sodium All (FFQ) Urine sodium 0.24 Potassium Urine potassium 0.04 Zinc Serum zinc 4132

Jacques et al., Magnesium 39 men and 70 Non-supplement Sérum magnesium 0.15 1993 women users 56 men and 82 Supplement arid Serum magnesium 0.27 women non-supplement users Zinc 44 men and 76 Non supplement Serum zinc 0.10 women users 56 men and 82 Supplement and Serum zinc 0.11 " women non-supplement users

Matkiovic etal., Calcium 381 pre teen All (3 day diary) Urine calcium 021 1995 Sodium girls Urine sodium 0.21

Donavan & Gibson, Tгоп 124 women All (3 dayweighed Sérum iron 0.26 1995 Zinc records) Serum zinc 0.16

Shaw: eta! 1995 : Iron 43 men and 71 All (6 day weighed Plasma haemoglobin 0.26 women food records)

Fatty acids van 5tavereп etal., Р:5 rat'ioa . 162 men and All subjects (2 day : Adipose PS ratio 0.38 1986 M:P ratio. women records) Adipose M ratio 0.36 P Adipose P 040 P:$ ratio 59 women Average 19 Adipose Р:$ .ratio 057 M.P ratio 24 h recalls Adipose M Р ratio 0.63 P , Adipose P 0.68 Linoleic:S ratio Adipose linoleic 0.62 S ratio M: linoleic ratio Adipose м linoleic 0.63 ratio lanoleic acid Adipose linoleic 0.70

112 Biochemical markers of dietary intake

Author Nutrients or food No. of Category of Biochemical Correlation groups subjects subjects index coefficient London et aL, 1991 5 115 FFQ Adipose 5 0.02 M postmenopausal Adipose М 0.08 Oleic acid women Adipose oleic acid 0.06 Trans fatty acids Adipose trans 0.41 fatty acids Р Adipose Р 0.15 Linoleic acid Adipose linoleic acid 0.13 Linolenic acid Adipose linoleiiic acid 0.07 n-3 fatty acids of Adipose n-3 fatty 0.43 manne origin acids of marine origin P1s ratio Adipose Р15 ratio 0.28

Hunter et ai 1992 5 118 men FFO Adipose S 0.14 M Adipose М 0.21 p Adipose Р 0.10 Adipose P:S ratio 0.43.. Р:S ratio Paimitic Adipose palimitiç 021 Oleic. Adipose oleic 0.28 Linoleic Adipose lirioleic 0.10 Eicosapentaenoic Adipose 0.43 ekosâpentaënoic Trans-isomers Adipose trans-isomers 0.21 S 118 men 2 weeks of diet Adipose 5 : 0.25 records Adipose 4.16. М. > М p Adipose P 0.09 Adipose :5 ratio 0:40 Р:S ratio, Р Feunekss et ai P:S rafioa 55 men and FFO. Adipose Р:S ratio 0.32 : ratio women Adipose : ratio 0.24 1993 М Р.. М Р р Adipose Р 0.24 Luoleic:S ratio Adipose lirioleic 0:33 5 ratio, Adipose :linol 0.25 М:linoleic ratio м ею ratio Linoleic acid Adipose linoleiç 0.28 :S ratioa Dietary history Adipose : ratio. . 0.28 Р Р Ѕ Adipose M: ratio 0.34 M:Р ratio Р p Adipose Р 0.29 Adipose linoleic: 0.30. Liпolsiс/S ratio S ratio

113 Application of Biomarkers in Cancer Epidemiology

Author Nutrients or food No. of Category of Biochemical Correlation groups subjects subjects index coefficient Feu nekes et al. M:linoleic ratio Adipose M:linoleic 0.44 1993 (Contd) ratio Linoleic acid Adipose linoleic 0.34 Р:5 rаtioa 99 men and FF0 Erythrocyte 0.22 women membrane P:5 ratio M:P ratio . Erythrocyte 0.37 membrane M:P ratio P Eгy4hrocytе 0.33 membrane Р Linoleic:S ratio Eтythгocyte 0.40 membrane linoleic: . 5 ratio М linoleic ratio Erythrocyte 0.41 membrane M:linoleic ratio Linolsic acid Erythrocyte 0.44 . membrane linoleic

T]onneland et al., S 67 men and 121 FFQ Adipose S 0.24 1993 М women Adipose М 0.05 P Adipose P 0.44 18:2 h-6 Adipose 18:2 n-6 0.44 19:3 n-3 Adipose 18:2 л-Э. 0.12 i 20:5 n-3 Adipose 20:5 n-3 0.47 22:6 n-3 Adipose 22:6 л-Э 0.41 S . 67 men and 121 Two 7-day diaries Adipose S 0.46 М . women . Adipose ЛA 0.19 P Adipose P 0.57 192 n-6 Adipose 182 n- 0.51 19:Э п-3 Adipôse.18:2 л-3 0.36 205 n-3 Adipose 20:5 n-3 0.44 226 n-3 . Adipose 22 6 п-3 0.55

Hebert et a?. 1994 Fat 167 men and Plasma cholesterol 0.02 women Plasma -0.12 KIL cholesterol Cholesterol Plasma cholestérol 0.02 Plasma HDL cholesterol X1.05 S Plasma cholesterol . ¢0.01 Plasma HDL-cholésterol -0.12

114 Biochemical markers of dietary intake

Author Nutrients or food No. of Category of Biochemical Correlation groups subjects subjects index coefficient Hebert et a1. M Plasma cholesterol 0.02 1994 (ctontd) Plasma HDL-cholesterol -0.12 P Plasma cholesterol 0.01 Plasma HDL-cholesterol -'109

Ma eta]., 1995 3570 men and FF0 Plasma cholesterol women esters: S S 0.23 M M 0.01 P P 0.31 P:S ratio P:S ratio 0.30 16:0 16:0 0.19 • 18:1 n-9 16:1 n-9 0.03 182 n-6 182 n3 0.28 18;3 n3 16:31-3 (.21 20:5 n-3 20:5 e3 0.23 22:6 n-3 22:6 n3 0.42 ь Plasma phospholipids: g S 0.15 M M 0.05 P P 0.25 P:5 ratio P:S ratio 0.24 16:0 16:0 0.16 18:1 n-9 18:1 n-9 0.08 18:2 n-6 18:2 n-6 0.22 18:3 n-3 18:3 n-3 0.15 20:5 n-3 20:5 n-3 0.20 22:6 n-3 22:6 n-3 0.42 aР, polyunsaturated fatty acids; i', monounsaturated fatty acids; S, saturated fatty acids.

115 Application of Biomarkers in Cancer Epidemiology

beyond the traditional nutrients), an important term recalls or food consumption records. It is this problem is that chemical constituents may have statistical independence of the errors which in the- rather variable concentrations within given types ory makes biomarkers of diet a very attractive addi- of food, due to natural variations between cultivars tional class of dietary intake measurement for use (e.g. indues in cabbage; Preobrazhenskaya et al., in epidemiological studies. 1993; Guengerich, 1995), due to variations in the soil content of specific minerals and trace elements The use of biomarkers of diet in prospective cohort (e.g. selenium; U11rey, 1981), or due to effects of studies storage, processing and cooking methods (e.g. for- A first type of application of biomarkers in mation of aromatic amines such as PhIP, McIQ nutritional epidemiology is their use as an inde- Siпha et al., 1995; Skog et al., 1995). Information pendent measurement of dietary intake levels, to on these sources of variation (type of cultivar, soil, evaluate the association between these intake lev- cooking methods) often cannot be assessed reliably els and disease risk. There are numerous examples using questionnaire methods, and it will therefore of this type of application, for instance in studying be impossible to apply correct food composition the relationships between cancer risk and plasma values by which the intake of these specific food levels of carotenoids, vitamin C, selenium or the constituents can be computed from the estimated fatty acid composition of blood lipids (e.g. 5thelin food consumption Ievels. et al., 1991; London et ai', 1992; Garland et aL, A second possible advantage of biomarkers 1993; van den Brandt et al., 1993; Zheng et al., often mentioned is their objectiveness; that is, the 1993; Berg et al., 1994; Gann et al., 1994; Kabuto et validity and precision of the markers can reason- al., 1994). As mentioned above, an initial argument ably be assumed to be independent of the subjects' for this type of application is that the marker may memory or capacity to describe foods consumed. classify individuals more accurately than traditional Only markers based on 24-h urinary excretion of a dietary assessment methods by habitual intake level given compound may be influenced directly by of some food constituents. For instance, dietary factors related to subjects' motivation, as urine col- intake levels of selenium may not be calculated very lections may be incomplete. However, the соm- reliably from individuals' estimated levels of food pleteness of urine collections can be checked by consumption, because selenium concentrations can measuring the recovery of a small amount of РABА vary widely irn individual foods. Likewise, intake lev- (para-arnino benzoic acid), ingested in the form of els of specific types of carotenoids (e.g. lycopene, small tablets at regular intervals during the 24-h zeoxanthine) or specific types of fatty acid may not collection period (Bingham & Cummings, 1986). It be estimable because of a lack of food composition must be noted that the term 'objectiveness' does data. A further argument for using a biomarker as not necessarily mean that the validity and preci- a measurement of dietary intake level is that a rela- sion are the same for cases with a given disease and tionship between a marker and disease risk maybe for disease-free controls; this may be true only if interpreted as independent confirmation of a rela- biological specimens are collected well before any tionship found between disease risk and dietary metabolic changes occur that may be a conse- intake measurements obtained by more traditional quence rather than a cause of disease and that may techniques (e.g. dietary questionnaires) that have have influenced the level of the marker (i.e. using different sources of error (although some caution a prospective study design). The more relevant should be made against oversimplistic interpreta- interpretation of the word objectiveness is that tions of biomarkers as a measurement uniquely of random errors in the marker—i.e. any variations dietary intake level; see the section on 'Validity that are not correlated with the individuals' trie, and confounding' below). habitual dietary intake levels—can be assumed to A second type of application of biomarkers of be statisticaIly independent of subjects' capacity or diet is to combine it, as an additional measurement motivation to give an accurate response to tradi- of diet, with measurements obtained by question- tional methods of measurement. Thus, the errors naires, recalls or food consumption records. None of biomarkers can be assumed to be uncorrelated of the existing methods of measuring individuals' with those of questionnaire measurements, short- true, habitual (long-term) intake levels of specific Biochemical markers of dietary intake foods or nutrients provides error-free measure- rents of different types, assuming independence ments. Food frequency questionnaires are the of errors between some, minimum, number of method used most often to measure individuals' variables may sometimes allow the estimation of dietary intake levels at baseline in prospective correlations of some other errors). To increase the cohort studies. systematic and random errors in likelihood that the assumption of statistical inde- the questionnaire measurements lead to bias in pendence of errors is actually valid, measurements estimates of relative risk or other measures of should preferably be taken by different methods, diet—disease association. By estimating the average with very different potential sources of error. magnitudes of these errors corrections may be Traditionally, the validity of dietary question- made for the bias in relative risk estimates. For naire measurements is evaluated by comparison instance, assume that questionnaire measurements with repeated measurements of actual food con- are related linearly to a true intake level T, i.e. sumption on a given day, obtained by food consumption records or 24-h diet recalls (both of these Q = с + jIQT + will also be referred to as 'reference' measurements (R), as, at a group level, they are often assumed to Unknown parameters in this model are the con- provide approximately unbiased measurements of stant and proportional scaling factors, аQ and p , intake, conditional on true intake level, i.e. R = T + н~ the mean (µ.1) and variance ( of the population Using this traditional design, the correlation coef- distribution of true intake values T, and the popu- ficient рQT is estimated by: lation variance бQ of random measurement errors EQ. From estimates of the parameters pQ, бQand б4, 1. calculating the crude coefficient of conela- the coefficient of correlation рQT between question between questionnaire measurements and tionnaire measureriients and true intake levels can the individuals' averages of their replicate refer- be computed. When relative risks are calculated for ence measurements (food consumption records quantile levels (e.g. quintiles) of the population or 24-h diet recalls); and distribution of intake levels, this correlation coef- 2. correcting this crude coefficient for attenua- ficient determines the degree of underestimation tion due to residual, within-subject variation in ('attenuation bias') due to random errors in the the individuals' average reference measure- questionnaire measurements, aid the loss in sta- ments, using a urцvariate analysis of variance tistical power that such underestimation entails to estimate within- and between-subject com- (de K1erk et al., 1989). ponents of variance in the average reference By making comparisons between questionnaire measurements (Rosier & Willett, 1988). measurements and other types of dietary intake As mentioned above, however, for the estimated assessment in valгдity subsfudies, it may be possible to estimate key parameters аQ, ¢a, µ~ бi, and 5 in correlation coefficient PQT to be valid, crucial the measurement model above. Estimation of assumptions to be made are as follows: these parameters then relies on statistical models in which true intake levels, T, are considered to be • Random measurement errors are statistically values of a latent variable. A crucial assumption for independent between the questionnaire aid the parameters to be estimated is that any associa- the replicate reference measurements. tion between complementary measurements of • Random measurement errors are statistically diet must be due only to the relationship between independent between the replicate reference each of these variables and the same latent vari- measurements themselves. able (true dietary intake level); that is, variations that are uncorrelated with true intake level ('ran- Violation of the first assumption in the form of dom errors') should also be uncorrelated between a positive correlation between errors of question- the different types of measurement. Generally, such naire and reference measurements will lead to an independence of errors cannot be proven in prac- overestimation of the validity coefficient рQT; on tice, but only assumed (although, in more complex the other hand, a positive correlation between the analyses of reliability studies with replicate measure- errors of replicate reference measurements will lead

117 Application of Biomarkers in Cancer Epidemiology

to an insufficient correction for attenuation effects, between random errors of questionnaire measure- and thus to an underestimation of the correlation ments (Q) and food consumption records or recalls coefficient РQT. If neither type of violation can be (R) cannot be excluded (but that the random errors ruled out, it is difficult to predict whether the of markers would be independent of those of mea- correlation coefficient рQ~. will be overestimated or surements Q and R), then the correlation coeffi- underestimated. cient р, T estimated from the triangular compari- As mentioned above, biomarkers have the son between measurements Q R and М can at least advantage that random errors can be assumed to be interpreted as an estimated upper limit of the be independent of those of more traditional types true value of PQT (Kaaks, 1997). of dietary intake assessment. Thus, any correlation $o far, substudies to evaluate the validity of between marker and questionnaire or reference dietary questionnaire measurements have usually measurements is assumed to be due entirely to the been conducted either on an external study group fact that these measurements are associated with of subjects not involved in the main epidemiolog- the same latent variable (true intake level of a ical (cohort) study, or on a relatively small sub- given dietary constituent). Nevertheless, it is sample of main study participants (nested within a important to note that generally biomarkers will cohort). Thus, since only a negligible number of not have a perfect correlation with true intake cases with a specific disease would occur in the level, even when the mean of many replicate mea- subsample; additional reference measurements surements of the marker over time is taken (i.e. (food consumption records, 24-h recalls) or addi- most markers have less than perfect validity as a tional measurements based on a biomarker could measure for classifying individuals by habitual, not be used as additional predictors of disease rel- true intake levels; see below). Therefore, the corre- ative risk. However, when biological specimens are lation between questionnaire measurements and collected from all members of a cohort study, and markers, even though adjusted for attenuation are available for later biochemical analyses, mark- effects due to random variation in the marker mea- ers can be measured in the specimens of subjects surements themselves, can only be taken as an esti- who subsequently develop disease and in individ- mated lower limit for the correlation coefficient р . uals who have remained disease-free (e.g. using a A somewhat different use of biomarkers in dietary nested case—control or a case—cohort design). In validity studies proposed more recently involves such situations, a triangular comparison between inferences based on simultaneous comparisons questionnaire measurements (Q), marker (М) and between questionnaire measurements (Q) 24-h disease status (D) may in principle allow the esti- food consumption records or 24-h recalls (R) and mation of relative risks unbiased by the random marker (M) (Plummer & Clayton, 1993a,b; Kaaks et errors in either type of exposure measurement. al., 1994a). For instance, from the a simple trian- However, the statistical methods for the analysis of gular comparison between single measures of Q R studies on the association between diet and disease and М, estimates of the correlations between each risk, using biomarkers as an additional measure- of the three types of measurements and individu- ment of dietary exposure, still require further als' true habitual intake levels (рq , PRT' M ) can be investigation. obtained using a structural equationsт modelР т (Kaaks of al., 1 994a) or, more directly, an elementary fac- Sources of variation, reproducibility and validity tor analysis model (Kaaks, 1997). The advantage of Intervening factors such a triangular comparison is that independence The interpretation of a biomarker as a measure of between random errors may need to be assumed an individual's (relative) intake level of a given only between the three diffèrent types of measure- food component obviously relies on the assump- ment (which indeed may have relatively indepen- tion that within- or between-subject variations in dent sources of error iii practice), and that no such true intake levels of a given compound of interest assumption is needed between replicate measure- are a primary determinant of within- and between- ments obtained by the same method (replicate subject variations in the marker. Variations in the weighed food records or replicate 24-h diet recalls). measured level of a marker of diet may, however, If it is felt that some level of positive correlation have many other exogenous or endogenous deter-

118 Biochemical markers of dietary intake minants which may affect the absorption, distrib- in the marker that are unrelated to true intake ution over body compaitments or tissues, metabolism level. Over wider ranges of intake, however, home- or excretion of a given compound. For example, ostatic mechanisms may cause significant devia- active smoking, as well as exposure to environ- tions from linearity. For example, for plasma mental smoke, have been reported to be associated retinol (vitamin A) concentrations, dose-response with decreases in plasma levels of vitamin C curves may be quasi-linear at habitual intake levels (Schectпnад et aL, 1989; Tribble etaL, 1993; Tappia that are so low as to induce a retinol-deficient (or et al., 1995) or p-carotene (Stryker et al., 1988; borderline deficient) state. At higher habitual Saintоt et al., 1994) that could not be explained intake levels, however, similar to those in most (only) by differences in diet, and which may be well-nourished populations, the dose-response due to increased oxidation of these antioxidant vit- curve flattens off to a practically constant level amins. Another example, concerning the validity that is under homeostatic control by the liver. of biomarkers of fat-soluble compounds such as Clearly, in the flatter parts of this type of carotenotds, retinol от vitamin E, is the simultane- dose-response curve, the association between ous intake of dietary fats, which enhance their marker and intake level will be weaker, or the asso- intestinal absorption. Endoger►ous metabolic fac- ciation may even be practically absent (see exam- tors include mechanisms of homeostatic regula- ple of retinol in Table 1). lion of specific nutrient levels in blood plasma (e.g. homeostatic mechanisms controlling the intesti- Repгaducibиltц validity and reliability na1 absorption and circulating plasma levels of Two useful notions in describing the quality and iron; homeostatic mecbaniszns controlling plasma usefulness of a bionrarker of diet are its repro- levels of retinol, by secretion of retinol—bound to ducihility (also called precision) and its validity. High retinol-binding protein--from the liver), or factors reproducibility implies that a measure of a bio- that affect the distribution of a compound over marker leads to very similar results for biological body compartments and different tissues (e.g. vit- specimens collected on two independent occasions amin C in plasma andin white blood cells). Other from the same individuals. A practical measure of factors may be related to the metabolism and/or the reproducibility is the correlation between the elimination of compounds. For instance, the levels replicate measurements. of adducts to lymphocyte DNA of many poten- Validity means that a marker actually measures tially toxic substances, such as N-nitroso-com- what it is intended to. As mentioned earlier, most pounds formed in the stomach from dietary pre- markers cannot be translated into a measure of cursors, or heterocyclic amines formed in meat daily intake on an absolute scale, but at the very during cooking, can be co-determined by genetic best are only correlated with intake level; that is, or phenotypic polymorphisms in the activity of the marker can oniy be used to provide a relative enzymes that play a role in the detoxification and classification of individuals by high or low dietary elimination of such compounds, or in the forma- intake levels. Therefore, we shall use the term tion of reactive intermediates that can form cova- validity here in a somewhat restricted sense, to in- lent bonds with DNA. These intervening factors dicate that the average of an infinite (or very large) may cause considerable variation in the marker number of replicate measurements of the marker that is unrelated to a dietary factor of interest. This (on independent occasions) has a correlation of additional random variation ('random' in the (close to) 1.0 with the individuals' true intake lev- sense that it is not predictable from the variations els. Furthermore, we shall call a measurement of a in diet that the marker is supposed to reflect) will marker of diet 'reliable' when it combines a high attenuate the association between markers and validity with a high reproducibility. true dietary intake levels. Perfect validity of a marker of diet implies that Within a restricted range of intake levels, the variations that are uncorrelated with true intake dose-response relationship between intake levels level ('random errors') are also statistically inde- of a given dietary constituent and a biochemical pendent between replicate measurements taken on marker is often approximately linear, and inter- the same individual (after a sufficiently long time vening factors may cause mainly 'random' variations interval). When the validity is less than perfect,

119 Application of Biomarkers in Cancer Epidemiology part of the between-subject variation in the marker of measurements of plasma vitamin C levels, and must reflect determinants (known от unknown) the correlation of plasma Ievels with dietary intake other than the dietary factor of interest. Foi example, measurements, can also be quite high (Bingham et plasma levels of a fat-soluble compound such as a2., 1995); these high correlations suggest that indi- vitamin E may be influenced, not only by vitamin viduals' dietary intake Ievels of vitamin C may be E intake, but also by plasma concentrations of low relatively stable and, hence, that a fast metabolic density lipoproteins, which, in turn, may depend turnover may have a relatively minor impact on on the type and amounts of fat consumed, on the the reliability of this marker). Similarly, the day- degree of obesity, or on genetic factors causing to-day variability of fatty acid composition is between-subject variations in lipid metabolism. higher for plasma triglycerides (especially in non- Reproducibility is generally seen as an initial fasting plasma) than for plasma phopholipids, practical criterion for the selection of a marker of cholesteryl esters and erythrocyte membranes, and diet as a potential measure of exposure—by defin- is almost null in subcutaneous adipose tissue biop- ition a poorly reproducible marker also has low sies, where the half-life of fatty acid turnover is reliability, irrespective of what the marker was sup- estimated at 18-24 months. posed to measure. Nevertheless, if random varia- A question of practical interest is whether a tions in the marker are only weakly correlated more reproducible marker—e.g. the fatty acid com- between replicate measurements, the overall relia- position of an adipose tissue biopsy, rather than bility may be increased substantially by taking the that of plasma triglycerides may generally also be average of replicate measurements of the marker expected to be moie valid. This would be true if on the same individual. Inversely, when the ran- the metabolic factors that cause the increased sta- dom errors of replicate measurements are highly bility over time primarily smoothen the effects of correlated (i.e. the validity is low), the reliability of day-to-day variations in true intake level, but do a marker may not be improved much by taking the not induce new between-individual variations that average of multiple measurements per individual. are independent of the individuals' habitual intake In practice, the reproducibility of a biomarker levels. In reality, this may not always be the case. will be a function of the stability of individuals' For example, from a physiological point of view, dietary intake level of a given compound (i.e. the the fatty acid composition of (non-fasting) plasma ratio of within-subject to between-subject varia- triglycerides, which consist mainly of fats absorbed tions in intake level), as well as of its metabolic immediately after a meal, should reflect very well turnover rate. The turnover rate may depend the composition of the fats just ingested. As a mea- largely on whether the compound is mainly fat- sure of the average composition of fats consumed soluble or water-soluble, as well as on the type of in the long term, however, this marker may be rel- medium (e.g. plasma, urine, saliva, blood cells or atively unreliable, as there can be substantial day- other tissue) chosen for the measurement. For to-day variation in the fatty acid composition of example, after intake of a high dose of vitamin C, foods consumed. Inversely, turnover rates of fats a water-soluble compound, a relatively large pro- are low in adipose tissue, and the fatty acid com- portion will be excreted in the urine within a very position of such adipose tissue biopsies is a highly short time. Fat-soluble compounds, on the other reproducible measurement, at least over intervals hand, can accumulate more in the body. The day- of less than 1 year. Nevertheless, whereas some to-day variation of vitamin C concentrations iп types of fatty acid from diet will be absorbed and lymphocytes is lower than that of vitamin C levels finally stored in adipose tissue without modifica- in blood plasma. Thus, if the vitamin C concen- tion and in proportion to their presence in plasma trations in lymphocytes can be taken as a weighted triglycerides, much of other fatty acids may be average of varying plasma levels in the days or modified before or upon storage, and part of the weeks before a blood sample is taken, the vitamin fatty acids found in adipose tissue may actually C concentrations in lymphocytes may be a more have been synthesized endogenously from carbo- reliable marker of average intake levels over a hydrates. Therefore, although more stable over longer time period. (Note, however, that results from time, the fatty acid composition of adipose tissue several studies have shown that the reproducibility may only be a reliable marker of some (not all)

120 Biochemical markers of dietary intake types of fatty acid in the diet (e.g. the linoleic acid VLDLs) (Hunter, 1990, 1993). High plasma levels content of adipose tissue shows moderately high of LDLs and VLDLs are associated with a nutri- correlations with lirnoleic acid intake from diet, tional/metabolic profile (e.g. marked by insulin whereas for palmitic acid or stearic acid this corre- resistance) that is often found in obese individuals lation tends to be lower: van Staveren et al., 1986; and which may be a risk factor of disease indepen- London etaL, 1991). dently of, and possibly in an opposite direction to, Studies to evaluate the reliability (ranking vitamin E. It has been stated that, iп this type of capacity) of markers as a measurement of diet have situation, the intervening factors may be seen as generally been based on a simple comparison (cor- (positive or negative) confounders of the observed relation) with traditional dietary intake assess- marker-disease relationship. In many of these ments (see Table 1 for examples). For some markers, situations, however, this view may be overly sim- e.g. caxotenoids and vitamin li, correlations with plistic or even incorrect. questionnaire measurements have been reported to If the intervening factors mainly cause random be rather low (rable 1; see also reviews by Willett, variation in the marker, unrelated to the level of 1990b and van't Veer et al., 1993), whereas the cor- intake of a given dietary component of interest, relation can be as high as 0.6 or 0.8 for vitamin C while at the same time affecting disease risk, then or finoleic acid (van Staveren et al., 1986; Bingham the marker may not only reflect differences in et aI., 1995; Riboli et al., 1997). Adjustments for intake levéls of the dietary component, but partly attenuation bias due to time-related, intra-individ- will also be a reflection of the effects of the internal variations in markers—especially those based vening factors; in this situation, the interpretation on a concentration of a given compound in tissues of the marker-disease relationship as reflecting or body fluids, rather than on a balance between uniquely a relationship between intake level and intake and excretion—often do not lead to great disease would be simply invalid. In a slightly more improvements in the correlation with the dietary complicated case, intervening factors causing vari- intake assessments, which suggests that the validation in both the markers and disease risk may also ity of these markers may be limited. Obviously, have some correlation with the dietary variable however, a less than perfect correlation between that a marker was actually supposed to measure. markers and other measurements of dietary intake For instance, smokers may consume less fruit and (e.g. from a questionnaire) may also be due to vegetables and have a lower vitamin C intake errors in the latter type of measurement. The use of (Schectmaп et al., 1989; 5ubar et aI., 1990; Ross et latent variable methods for the analysis of validity al., 1995), while, independently, inhalation of studies based on comparisons between replicate tobacco smoke may also lower plasma vitamin C measurements of a biomarker, questionnaire mea- levels by increasing its oxidation (Schectman, surements of habitual diet and/or replicate mea- 1989; Tribble et al., 1993; Tappia et a2., 1995). surements of actual food consumption on given Furthermore, smoking augments the risk of devel- days may help to obtain more accurate estimates of oping many chronic diseases, including cancer. In the validity and precision of markers. this case, the relationship between disease nsk and intake of vitamin C is indeed confounded by smok- Validity and confounding ing, in the sense that intake levels of vitamin C Besides being additional sources of variation in the could be found to be inversely associated with dis- levels of markers, exogenous and endogenous ease risk even when vitamin C intake was un- intervening factors may also be independent pre- related to the etiology of disease. To the extent that dictors of the risk of developing a given disease the marker is interpreted purely as a measurement under study. For example, high plasma levels of of vitamin C intake level, the estimated relation- fat-soluble compounds—such as (3-carotene, vita- ship between marker and disease risk must then min E or DDT metabolites—may reflect high also be considered to be confounded by smoking. dietary intake levels of these compounds, but they Again, however, as part of the variation in the may also reflect high concentrations of low- marker may actually reflect the vitamin C-lower- density aid very low-density lipoproteins (LDLs, ing effects of smoking rather than vitamin C

121 Appücation of Biomarkera in Cancer Epidemiology

intake, one can argue that the interpretation of the that are currently being developed (e.g. biomarkers marker purely as a measurement of vitamin C for intake of bioflavonoids, induIes or phytoestro- intake level is simply invalid. To avoid such ambi- gens from certain plant foods, or for intake of aro- guities in interpretation, it seems conceptually matic amines formed during cooking of meat or more appropriate to consider the marker as an fish). In validity sub-studies, as well as in full-scale intermediate variable that is determined by dietary epidemiological studies of diet—disease relation- variables of interest together with various inter- ships, this limitation can be overcome if another vening factors, and at the same time as a de- type of measurement is available (e.g. weighed terminant of disease risk. Under this alternative food consumption records) which does provide view the relationship between vitamin C intake estimates of intake level on an approximately valid and the plasma concentration of vitamin C as an scale. In theory, a few markers that can be translated intermediate (outcome) variable is confounded into correctly scaled, absolute daily intake levels by smoking, but not necessarily the relation- (e.g. 24-h urinary nitrogen excretion to estimate ship between disease risk and plasma vitamin daily protein intake) could also be used as a refer- C as an indicator of vitamin C status, which could ence measurement for caIibrаtion substudies. The actually be a valid reflection of the relationship latter can be used to adjust for bias in relative risks between the biological availability of vitamin C estimated for absolute, quantitative differences in and disease risk. dietary intake level (rather than between different quantile levels of the intake distribution) (Kaaks summary and concluding remarks et al., 1995), and can improve the comparability of Biomarkers as a measurement of dietary intake dietary intake measurements between different level can be useful when estimated intake levels of study cohorts (Kaaks et al., 1994b; Plummer et al., foods obtained from questionnaires or from 1994). records or short-term recalls of food consumption As described in this chapter, there can be many cannot be translated reliably into intake levels of applications of biomarkers as indicators of diet in specific compounds, for instance because of a lack epidemiological studies. Nevertheless, some words of accurate food composition tables. In addition, of caution may be needed against overenthusiasm even when dietary intake levels can be assessed in favour of markers as the solution to the prob- with reasonable accuracy by means of a question- lems of mismeasurement of dietary intake levels naire, or by keeping records of actual food con- by moie traditional methods (questionnaires, food sumption, a marker can be a useful complemen- consumption records, short-term recalls). Results tary measurement given that its random errors are from many studies suggest that correlations be- likely to be truly independent of those of the first tween biomarkers and true intake levels of specific two categories of measurement. Thus, markers can dietary constituents are generally not substantially be useful, especially in dietary validity studies, to higher (in fact, they are often lower) than those estimate the average magnitude of systematic and for the more traditional types of measurement, random errors of dietary exposure measurements, as and even the reproducibility of markers can be obtained in an epidemiological study. Knowledge relatively low. The rather low reliability of many of the average magnitude of such errors allows cor- markers may be due partly to a comparatively large rection for bias in relative risk estimates. intra-individual variation in true intake level ovei А limitation of most markers as indicators of time and relatively fast metabolic turnover rates diet is that, at the very best, they can only provide of given dietary compound, or may partly be an estimate of relative ranking of individuals by explained by the effects of intervening variables intake level, but cannot be translated into absolute that affect the validity of the marker as a measure intakes on a valid scale. This limitation is generally of diet. Another aspect that may explain the low present when the marker is based on measurement reproducibility of a marker are errors in measure- of a concentration of a compound, rather than on ments due to inaccuracy of laboratory techniques. the recovery over time of a compound in excretion Because of random measurement errors, and irre- products, and will be shared by many biomarkers spective of the type of application of markers in

122 Biochemical markers of dietary intake nutritional epidemiology—as an independent pre- References dictor of disease risk, or as an additional measure- Alpha-Tocopherol, Beta-Carotene Cancer Prevention ment of intake level in combination with more tra- Study Group (ATBC) (1994) The effect of vitamin E and ditional measurements of diet—the statistical beta-carotene on the incidence of lung cancer and other cancers in male smokers. N. Engl. J. Mcd., 330, power of studies using markers will be low unless 1029-1035 they include very large numbers of subjects. Strong caution must also be exercised against Ascherio, A., Stamрfеr, M.J., Colditz, G.A., Rimm, E.B., Litin, L, & Willel, W.C. (1992) Correlations of vitamin overinterpтetatioп (unjustified extrapolation) of the results of epidemiological studies based on A and E intakes with the plasma concentrations of carotenoids and tocopherols among American men and markers, in that it may not be entirely clear which women. J. Nutr., 122, 1792-1801 type of intake variable is specifically represented by a marker. For example, in several observational Berg, J.P., Glattre, E., Haldorsen, T., Hostmark, A.T, Bay, & Je11um, E. (1994) Longchain serum (epidemiological) studies, lung cancer risk was LG., Johansen, A.F. fatty acids and risk of thyroid cancer: a population-based found to be associated inversely with plasma levels case-control study in Norway. Cancer Causes Control, 5, of ft-carotene (ATBG, 1994; Omenn et aI., 1994) 433-439 and this seemed to confirm results showing simi- Beixjno, F., loti, P., Micheli, A., Bolelli, G., Krogh, V., 3- lar inverse associations with intake levels of Sciajrio, R., Pisani, P, Panico, S. & Secreto, G. (1996) carotene as measured by dietary questionnaires. Serum sep hormone levels after menopause and subse- Nevertheless, in the Finnish ATBC trial and in the quent breast cancer. J. Nat! Cancer Inst , 88, 291--296 CARET trial, supplementation with ft-carotene (at supraphysiological doses) led to a significant Bingham, S.A. & Cummings, J.H. (1985) Urine nitrogen as an independent validatory measure of dietary intake: increase, instead of the expected decrease, in lung a study of nitrogen balance in individuals consuming cancer risk. These conflicting results suggest that their normal diet. Am. J. Clin. Nufr., 42, 1276-1289 the decrease in lung cancer risk observed in epidemiological studies may have been attributed Bingham, S.A. & Cummings, J.H. (1986) Creatinine and РАBА as markers for completeness of collection of 24- wrongly to 3-carotene per se, and that high esti- hour urine samples. Horn. Noir. Clin. Nutr., 40, 473-476 mated intake levels of p-carotene or high measured plasma concentrations were actually indicators of Biugham, S.A., Cassidy, A., Cole, T.J., Welch, A., other protective factors, e.g. associated with a high Runswick, S.A., Black, А.E., Thurnham, D., Bates, C., Khaw, K.T., Key, Т.J. & Day, N.E. (1995) Validation of intake of vegetables and fruits (although an alter- weighed records and other methods of dietary assess- native interpretation of these studies may be that ment using the 24-hour urine nitrogen technique and ft-carotene at physiological doses does actually other biological markers. Br. J. Nutr., 73, 531-550 confer some protection against lung cancer but Bingham, S.A., Murphy, J., Waller, E., Runswick, S.A., that supraphysiological doses are harmful). Neale, G., Evans, D. & Cummings, J.Н. (1992) Para- This example shows that, in spite of a greater amino benzoic acid in the assessment of completeness of apparent specificity as a measurement of the rela- 24-hour urine. Collections from hospital outpatients and tive intake level of a given dietary constituent, bio- the effect of impaired renal function. Eur. J. Clin. Nutr. markers in many (if not most) cases may only be 1992, 46, 131-135 interpreted as indicators of more complex dietary Cameron, M.E. & van Staveren, W.A. (1988) Manual on risk factors. Indeed, there is increasing awareness Methodology for Food Consumption Studies, Oxford, Oxford that it may be more appropriate to search for University Press patterns of dietary risk factors associated with Campbell, .R., Gross, M.D., Martini, M.F.C. Grandits, increased risk. In future studies, different markers Л G.А., Slavin, J,L & Potter, J.D. (1994) Plasma carotenoids combined may be used as indicators (`markers' in as biomarkers of vegetable and fruit intake. Cancer the true sense) of such more complex patterns. Epidemiol. Biomarkers Prey., 3, 493-500 Coates, R.J., Eley, J.W., Block, G., Gunter, E.W., Sowell, Acknowledgements A.L., Grossman, C. & Greenberg, Es. (1991) An evaluation We thank Dr David Hunter, Harvard Sсhool of of a food frequency questionnaire for assessing dietary Public Health, for reviewing this manuscript and intake of specific carotenoids and vitamin E among for his helpful comments. low-income black women. Am. J. Epiderniol. 134, 658-671

123 Application of Biomarkors in Cancer Epidemiology de Kierk, N.H., English, D.R. & Armstrong, B.K. (1989) A measurements. Am. J. Clin. Nutr. (in press) review of the effects of random measurement error on Kaaks, R. Plummer, M., Riboli, E., Estève, J, & van relative risk estimates in epidemiological studies. Jot. J. 5taveren, W. (1994b) Adjustment for bias due to errors in EpidemioL, 18, 705-712 exposure assessments in multicenter cohort studies on Donovan, U.M. &c Gibson ~.S. (1995) Iron and zinc sta- diet and cancer: a calibration approach. Am. J. Clin. Nutr., tus of young women aged 14 to 19 years consuming veg- 59,2455-2505 etarian and omnivorous diets. J. Am. Coil. Nutr., 14, 463-472 Kaaks, R., Riboli, E., Estève, J., van Kappel, A.L. & van Staveren W.A. (1994а) Estimating the accuracy of dietary Enger, S.M., Longnecker, M.R, Shikany, J.M., Swenseid, questionnaire assessments: validation in terms of struc- M.E., Chef, M., Harper, J.M. & Haile, R.W. (1995) tural equation models. Stat. Med., 13, 127-142 Questionnaire assessement of intake of specific t - сато е Kaaks, R., Riboli, E. & van St veren, W. (1995) noids. Cancer EpidemioL Biomarkers Prev., 4, 201-205 а Calibration of dietary intake measurements in prospec- Feunekes, G.I.J., Van Staveren, W.A., De Vnies J.H.M. tive cohort studies. Am. J. EpidemioL, 142, 548-556 Burema, J. & Hautvast J.G.A.J.(1993) Relative and bio- Kabuto, M., Imai, H., Yonezawa, C., Neriishi, K., Akiba, marker-based validity of a food-frequency questionnaire S., Kato, H., Suzuki, T., Land, .E. & Blot, W.J. (1994) estimating intake of fats and cholesterol 1-3. Am. J. Clin. С Prediagnostic serum selenium and zinc levels and subse- Nutr., 58, 489-496 quent risk of lung and stomach cancer in Japan. Cancer Gann, P.H., lennekens, C.Н., Sacks, F.M., Grodstein, F., Epidemiol. Biomarkers Prey., 3, 465-469 Giovannucci, E.L. & Stampfer, M.J. (1994) Prospective Kardinaal, A.F.M., van't Veer, P., Brants 13.4.1., van den study of plasma fatty acids and risk of prostate cancer. J. Berg,H., van Schoorihoven, J. & Hermus R.J.J. (1995) Natl Cancerinst., 86, 281-286 Relations between antioxidant vitamins in adipose tis- Garland, M., Willett, W.C., Manson, J. Е. & Hunter D.J. sue, plasma, and diet. Am. J. ЕрiдетиоL,141, 440-450 (1993) Antioxidant micronutrients and breast cancer. J. London, 5.J., Sacks, EM., Caesar, J., Stampfer, M.J., Siguel, Am. II. Nutr., 12, 400-411 Са E. & Willett, W.C. (1991) Fatty acid composition of sub- Guengeiich, EP, (1995) Influence of nutrients and other cutaneous adipose tissue and diet in postmenopausal Us dietary materials on cytochrome P-450 enzymes. Am. J. women. Am. J. СIiп. Nutr., 54, 340-345 Clin. Nutr 61, 651S-65В$ London, S.J., Stein, Е.А., Henderson, I.C., Stampfer, М.J., Hebert, J.R., Hигlеy, T.G., Hsieh, J., Rogers, E., Stoddard, Wood, W.C., Remisse, S., Dmochowski, J.R., Robert, N.J. A.М., Sorensen, G. & Nicolosr, R.J. (1994) Determinants & Willett, W.C. (1992) Carotenoids, retinol, and vitamin of plasma vitamins and lipids: the working well study. E and risk of proliferative benign breast disease and Am. J. Eрideтiol. 140, 132-147 breast cancer. Cancer Causes Control, 3, 503-512 Hunter, D. (1990) Biochemical indicators of dietary Ma, J., Folsom, A.R., Shahar, E. & Eckfeldt, J.H., for the intake. In: Willett, W., cd., Nиtnhоnц Epidemiology, New Atherosclerosis Risk in Communities (ARIL) Study York, Oxford University Press, pp. 375-418 Investigators (1995) Plasma fatty acid composition as an indicator of habitual dietary fat intake in middle-aged Hunter, D.J. & Kelsey, К.T (1993) Pesticide residues and breast cancer: the harvest of a silent spring? J. Nat! Cancer adults. Am. J. Clin. Nutr., 62, 564-571 Inst., 85, 598-599 Matkovic, V., I1ich, J.Z., Andin, M.B., Hsieh, L.C., Hunter, D.]., Rimm, E. ., Sacks, F.M., Stampfer, M.J., Tzagournis, M.А., Lagger, B.J. & Goe1, P.K. (1995) Urinary В caldum, sodium, and bone mass of young females. Am. Colditz, G.A., Litin, L.В. & Willett, W.C. (1992) Comparison of measures of fatty acid intake by subcuta- J. Clin. Nutr., 62, 417-425 neous fat aspirate, food frequency questionnaire, and O'Donnell, M.G., Nelson, M., Wise, P.H. & Walker, D.M. diet records in a free-living population of Us men. Am, J. (1991) A computerized diet questionnaire for use in diet EpidemioL, 135, 418-427 health education 1. Development aid validation. Br. J. Nutr., 66, 3-15 Jacques, EF,, Sulsky, S.I., Sadowski, J.А., Phillips, C.C., Rush, D. & Willett, W.C. (1993) Comparison of micronu- Omenn, G.5., Goodman, G.E., Thornquist, M.D., Balmes trient intake measured by a dietary questionnaire and J., Cullen, M.R., Glass, A., Keogh, J.P., Keyskens, F.L., biochemical indicators of micronutrient status. Am. J. Valanis, B., Williams, J.Н., Barnhart, 5. & lammar, S. CIin. Nutr., 57, 182-189 (1996) Effects of a combination of beta carotene and vit- Kaaks, R. (1997) Bionnarkers as an additional measure- amin A on lung cancer and cardiovascular disease. New ment in studies on the accuracy of dietary questionnaire Engl. J. Med., 334, 1150-1155

124 Biochemical markers of dietary intake

. (1988) Measurement of energy expendi- Рlиттеr, M. & Clayton, D. (1993а) Meаsшement error Schoeller, D.А in dietary assessment: an investigation using covariance tore in free-living humans by using doubly labeled watec structure models. Part I. Stаt. Med., 12, 925-935 J. Nutr., 118, 1278-1289 Plummer, M. & Clayton, D. (1993Ь) Measurement error Shaw, N.S., Chin, C". Fr Pan, W.H. (1995) A vegetarian in dietary assessment: an investigation using covariance diet rich in soybean products compromises iron status in structure models. Part I. Stat. Med., 12, 937-948 young students. J. Nutr., 125, 212-219 Sinha, R., Block, G. & Taylor, P.R. (1992) Determinants of Рlиттег, M., Clayton, D. & Kaaks, R. (1994) Calibration in multi-centre cohort studies. Int. J. EpidemioL, 23, 419-426 plasma ascorbic acid in a healthy male population. Cancer Epidemiol. Biomarkers Prey., 1, 297-302 Porrini, M., Gentile, M.G. & Fidanza, F. (1995) Biochemical validation of a self-administered semi-quantitative Sinha, R., Frey, C.M., Kammerer, W.G., IcAdams, M.J., food-frequency questionnaire. Br. J. Nutr. 74, 323-333 Norkus, E.P. & Zeigler, R.G. (1994) lmportancе of sup- .1. plemental vitamin C in determining serum ascorbic acid Preobrazhenskaya, M.N., Bukhman, V.M., Korolev, А in controls from a cervical cancer case-control study: & Efimov, S.A. (1993) Ascorbigen and other indole- Implications for epidemiological studies. Nutr. Cancer, derived compounds from Brassica vegetables and their 22, 207-217 analogs as anticarcfnogenic and immunomodulating agents. PhannacoI. Ther., 60, 301-313 Sinha, R., Patterson, B., Mange's, A.R., Levander, O.A., Gibson, T., Taylor, P.R. & Block, G. (1993) Determinants Riboli, E. & Kaaks, R. (1996) The EPIC project: rationale of plasma vitamin E in healthy males. Cancer Epidemiol. and study design. Iot. J. Epidemiol. (in press) Biomarkers Prev., 2, 473-479 holm, H., Saracci, R. (1987) Biological Riboli, E., Rdnп Sinha, R., Rothman, N., Brown, E.D., Salmon, C.P., Knize, markers of diet. Cancer Surv., 6, 685-7 18 M.G., Swanson., С.А., Rossi, S.С., Mark, S.D., Levander, Riboli, E., Еlmstâhl, S., Saracci, R., Gullberg, B. & O.A. & Felton, J.S. (1995) High concentrations of the car- Lindgârde, R (1997) The Malmd food study: validity of cinogen 2-amino-1-methyl-6-phenylimidazo-[4,5-b]pyri- two dietary assessment methods for measuring nutrient dine (PhIP) occur in chicken but are dependent on the intake. Int. J. Eрidеmiol. (in press) cooking method. Cancer Res., 55, 4516-4519 Roidt, L., White, E., Goodman, GE, Wahl, P.W., Omenn, G.S., Skog, K., Steineck, G., Augustsson, K. & Jagerstad, M. Rollins, B. & Karkeck, J.M. (1988) Association of food fre- (1995) Effect of cooking temperature on the formation of quency questionnaire estimates of vitamin A intake with heterocyclic amines in fried meat products and pan serum vitamin A levels. Am. J. E idemioL, 128, 645-654 residues. Carsrinogenesls, 16, 861-867 Romieu, I., Stamfer, M.J., Stryker, W.S., Hernandez, M., Snook, J.T., Palmquist, D.L., Moxon, A.L., Cantor, A.H. & Kaplan, L., Sober, A., Rosner, B. & Willett, W.C. (1990) Vivian, V.M. (1983) Selenium status of a rural (predomi- Food predictors of beta-carotene and alpha-tocopherol: nantly Amish) community living in a low-selenium area. validation of a food frequency questionnaire. Am. J. Am. J. Clin. Nutr., 38, 620-630 ioL, 131, 864-876 E iдет Stielin, H.S., Gey, K.F., Eichholzer, M., Ludin, E., Rosner, B. & Willett, W.C. (1988) Interval estimates for Bernasconi, F., Thurneysen, J. & Brubacher, G. (1991) correlation coefficients corrected for within-person vari- Plasma antioxidant vitamins and subsequent cancer ation: implications for study design and hypothesis test- mortality in the 12-year follow-up of the prospective ing. Am. J. Eрidemiol., 127, 377-386 Basel Study. Ans. J. Epidemiol., 133, 766-775 Ross, M.A., Crosley, L.К., Brown, K.M., Duthie, 5.3., 5teirnmetz, K.A. & Potter, J.D. (1991) Vegetables, fruit, Collins, A.C., Arethur, J.R. & Duthie, G.G. (1995) Plasma and cancer. II. Mechanisms. Cancer Causes Coпtroi, 2, concentrations of carotenoids and antioxidant vitamins 427-442 in Scottish males: influences of smoking. Era. J. Clin. Stryker, W.S., Kaplan, L.A., Stein, .~., Stampfer, M.J., Nutr, 49, 861-865 Е Sober, A. & Willett, W.C. (1988) The relation of diet, cig- Saiдtot, M., Astre, C., 5cali, J. & Gerber, M. (1995) arette smoking, and alcohol consumption to plasma Within-subjects seasonal variation and determinants of beta-carotene and alpha-tocopherol levels. Am. J. inter-individual variations of plasma beta-carotene. Int. J. Epidemiol., 127, 283-296 Vitam. Nutr. Res., 65, 169-174 Subar, А.F., Harlan, L.C. & Mattson, M.E. (1990) Food Schectman, G., Byrd, J.D. & Gris chow, 11.W. (1989) The and nutrient intake differences between smokers and influence of smoking on vitamin C status in adults. Ans. non-smokers in the US. Am. J. Public Health, 80, J. Public Health, 79, 158-162 1323-1329

125 Application of Biomarkers in Cancer Epidemiology

Swanson, C.A. Longnecker, M.P., Veillon, C., Howe, S.M., Wild, J., Seller, M.J., Schoгah, С.J. & Smithells, R.W. Levander, O.А., Taylor, P.R., IcAdam, P.A., Brown, C.C., (1994) Investigation of folate intake and metabolism in Stampfer & 1J., Willett, W.C. (1990) Selenium intake, women who have had two pregnancies complicated by age, gender, and smoking in relation to indices of sele- neural tube defects. Br. J. Obstet. Gynaccol., 101, 197-202 nium status of adults residing in a selenifezous area. Am. Willett, W.C. (1990a) Food frequency methods. I J. Clin. Noir., 52, 858-862 п: Willett, W., cd., Nutritional Epidemiology, New York, Tappia, P.S., Troughton, K.L., Langley-Evans, S.C. & Oxford University Press, pp. 69-91 Grjmble, R.F. (1995) Cigarette smoking influences Willett, W.C. (1990b) Reproducibility and validity of cytokine production and antioxidant defences. Clin. Sci. food frequency questionnaires. In: Willett, W., Coach., 88, 485-489 cd., Nutritional Epidemiology, New York, Oxford University Tjenneland, A., Overad, K., Thorling, E. & Ewertz, M. Press, pp. 92-126 (1993) Adipose tissue fatty acids as biomarkers of dietary Williams, D.R. & Bingham, S.A. (1986) Sodium and potas- exposure in Danish men and women. J. CIiп. Nutr., 57, 629-633 sium intakes in a representative population sample: estimation from 24 h urine coilections known to be com- Toniolo, P.G., Levitz, M., Selеniuch Jaсquotte, A., Bаnerjее, plete in a Cambridgeshire village. Br. J. Nutr., 55, 13-22 S., Koenig, K.L., Shore, R.E., Strax, P. & Pasternack, B.S. Witschi, J.C. (1990) Short-term dietary recall and record- (1995) A prospective study of endogenous estrogens and ing methods. In: Willett, W. ed., Nutritional Epidemiology, breast cancer in postmenopausal women. J. Natl Cancer New York, Oxford University Press, pp, 52-68 Inst., 87, 190--197 Yang, G., Zhou, R., Yin, S., Gu, U1tr y, D.E. (1981) Selenium in the soil-plant-food chain. L., Yan, B., Liu, Y., Liu, Y. е & Li, X. (1989) St dies of safe maximal daily dietary sele- In: Spa11holtz, J.E. et al., eds., Selenium in Biology and ц nium intake in a seleniferous area in China. L Selenium Medicine, Westport, CT, AVI Publishing, pp. 176-191 intake and tissue selenium levels of the inhabitants. J. Tribble, D.L., Guiliana, L.J. & Fortmann, S.P. (1993) Trace EIem. Electrolytes Health Dis., 3, 77-87 Reduced plasma ascorbic acid concentrations in non- Yong, L.C., Forman, M.R., Beecher, G.A., Graubard, B.I., smokers regularly exposed to environmental tobacco Campbell, W.S., Reichman, M.E., Taylor, P.R., Lanza, E., smoke. Am. J. Clin. Nulr., 58, 886-890 Holden, J.М. & Judd, J.T. (1994) Relationship between van den Brandt, P.А., GoIdbohm, R.A., van't Veer, P., dietary intake and plasma concentrations of carotenoids Bode, P., Dorant, E., Hermus, R.J. & Sturman, F. (1993) in premenopausal women: application of the USDA-NC' A prospective cohort study on selenium status and the carotenoid food-composition database. Am. J. Clin. NuIr., risk of lung cancer. Cancer Res., 53: 4860-4665 60, 223-230 van Staveren, W.A., Deurenberg, P., Katan, М.B., Burema, Zheng,W., Blot, W.J., Diamond, E.L., Norkus, E.P,, Spate, J., de Groot, L.C. & Hoffmans, M.D. (1986) Validity of V., Minis, J.5. & Comstock, G.W. (1993) Serum micronu- the fatty acid composition of subcutaneous fat tissue trients and the subsequent risk of oral arid pharyngeal microbiopsies as an estimate of the long-term average cancer. Cancer Res. 53, 795-798 fatty add composition of the diet of separate individuals. Am. J. Epiдemio1,,123, 455-463 van't Veer, P., Widen, V.D., Kok, F.J., Hermus, R.J,J. & Sturmans, F. (1990) Selenium in diet, blood, and toenails in relation to breast cancer: А case-control study. Am. J. Epidemiol., 131, 987-994 Corresponding author van't Veer, P., Eardtnaal, A.F., Bausch-Goldbohm, R.A. & R. Kaaks Kok, F.J. (1993) Biomarkers for validation. Eur. J. Clin. Unit of Nutrition and Cancer, International Agency for Nutr., 47 (Suppl. 2), 558-563 Research on Cancer, 150 cours Albert Thomas, 69372 Lyon Cedex 8, France Wattenberg, L, W. (1992) IпЫbitioп of carcinogenesis by minor dietary constituents. Cancer Res., 52, 20855-20915

126 Application o1 Biomorkera iп Cancer Epidemiology Toninl , P., Boffotla, P., 5huker, D.E.G., Ro mаn, N., Ни1Ка, 6, arid Pearce, N., eds 1ARC Scientific Pubications No. 142 Iriternalionol Agency for Research on Cancer, Lyon, 1997

Biomarkers for biological agents

N. Mur"toz and F.X. Bosch

Biomarkers of exposure to biological agents have proved to be extremely useful in establishing causal associations between infections and human cancer, as well as in tracing the natural history of the relevant agents.This chapter will focus on biomarkers for biological agents currently recognized as causally associated with various human cancers after evaluation at the IARC.These are human papillomavirus (HPV), hepatitis B (IBV) and C virus (1CV) and Helicobacterpylori. A prototype of nucleic-acid-based biomarkers is detection of HPV DNA. It measures the presence of type-specific DNA at a given point in time. PCR-based assays are considered the method of choice for epidemiological investigations. The test requires collection of exfoliated cells or biopsies. 5pеcimens can be kept stored at -20°C for long periods of time. DNA degradation is low. Some of the limitations of the marker for cancer epidemiology lie in the fact that IPV DNA infections are often transient, especially in young women. In repeated measurements, HPV DNA may fluctuate, but the reasons for this are unknown. Antigens and antibodies from the IBV and HCV can be viewed as prototypes for serological biomarkers. Fог IBV, there are markers able to distinguish between past and persistent infections. IBV DNA detection in sera further refines the assessment of exposure. 5tапdaгdizеd serological assays are available and widely used in developing countries, For HCV antibodies, serological assays are standardized and widely available. RNA detection in sera by PCR is under development. A limitation of the currently available assays is the type and subtype variation of HCV by geography, which requires further research and standardization. In low-risk populations, numerous false-positive results arise and confirmatory tests are required.

Biomarkers of exposure to biological agents have fication of viral proteins in the cells. Serological proven to be extremely useful in establishing assays, currently under development, could be of causal associations between infectious agents and great help in epidemiological studies as they might human cancer, as well as in tracing the natuial his- be able to assess cumulative lifetime exposure or at tory of the relevant agents. This chapter will focus least recent, as opposed to just current, infecLion. The in biomarkers for biological agents currently rec- colposcopic, cytological and histological diagnoses ognized as causally associated with various human of HPV are not very reflable as they lack sensitivity cancers after evaluation at the International and specificity (Barrasse, 1992; Sherman et aL, 1994) Agency for Research on Cancer (IАRС). These are and as such have very limited use in epidemiolog human papillomavirus (HPV) (IARC, 1995), ical studies. Likewise, the identification of viral hepatitis B (HBV) and C virus (HCV) ('ARC, I 994а) proteins such as HPV capsid antigens in infected and Helicobacter pylori (‚ARC, 1994b). cells is also of limited epidemiological value as they can only be detected in the case of productive HPV HPV and cervical cancer infection. These methods will not be discussed Over 75 types of I-{PV have so far been identified. here. We will concentrate on methods for HPV One of the most common methods of virus detec- DNA detection and on serological markers of HPV. tion, namely the isolation of virus in tissue culture, is not available for HPVs because HPVs cannot be Methods for HPV DNA detection propagated in culture. The presence of the virus HPV DNA detection methods have been developed can be inferred from clinical, cytological, histolog- most extensively for genital tract HPVs, because ical or serological findings, but the most accurate these infections have been linked to cervical can- methods are based on the demonstration of HPV cer. Of the 30-40 HPVs that infect the genital tract, DNA in cells and, to a lesser extent, on the identi- some types (e.g. HPV types 16, 18, 45, 31 and 33)

127 Application of Biomaгkетs in Cancer Epidemiology are closely associated with cervical cancers, Sources of variability whereas others (e.g. HPV types 6, 11, 42, 43, 44) are High interlaboratory variability has been reported rarely found in cancers. Therefore, in clinical and for FISH, Southern blot and the Hybrid CaptureT1 epidemiological investigations, HPV diagnosis assay. There is a significant interlaboratory varia- aims at identification of specific types of HPVs and tion in detection and typing of HPVs by Southern at separating cancer-associated HPVs from low-risk hybridization. In a study of 40 clinical samples, HPVs. pairwise agreement between four laboratories Two types of 11W DNA hybridization methods ranged from 66 to 97% for HPV detection, and are currently used with aid without amplifica- from 77 to 96% for typing of positive specimens tion. In the former, targeted viral sequences are (Brandsma et aI., 1989). first amplified by polymerase chain reaction (PCR) Iп a study of interlaboratory variation in Hybrid and then identified by hybridization assays. Most Capture results in three laboratories, specimens authors would consider PCR methods as the cur- were tested with probe pools A and B. Kappa values rent standard for epidemiological studies. These in interlaboratory pairwise comparisons for posi- methods have been reviewed recently (Gravitt & tivity to either group A or group B ranged from Manos, 1992; Walboomers et al., 1994) and sum- 0.61 to 0.83. 0f specimens that were positive for marized in IARC (1995). group B by the reference standard, 74% were posi- Serological assays are still under development. tive for group B in all three laboratories. The inter- Currently, ELISA (enzyme-linked immunoab- laboratory correlations of HPV quantitative data sorbent assay) antibody assays using virus-like par- for probe B types ranged from 0.60 to 0.90. ticles as the antigen appear to be the most promising Probable false-positive results were occasionally and have raised high expectations. However, the encountered, in less than 3% of the tests antibody response is low-titred and not detectable (Schiffman et al., 1995). in all patients with documented infections. Interlaboratory variation for two PCR-based Likewise, at present, antibodies to E6 and Е7 proteins methods using L1 consensus primers has been are the most promising markers of HPV-associated shown to be small ('sing et al., 1996). invasive cancer. HPV DNA detection in a single specimen measures the presence of the virus at a given point in Logistics of sample collection, processing and time, which is clearly different from assessing life- storage time exposure to HPV At present, there are rio reli- Exfoliated cells can be collected from easily acces- аЫе methods (hybridization or serological) to mea- sible sites such as uterine cervix, penis, bladder or sure 1-IPV lifetime exposure or to distinguish tran- oral саvity but collection from other organs such as sient from persistent HPV infections other than oesophagus, stomach or lung requires invasive pro- repeated HPV DNA detection. cedures, i.e. endoscopy. Various methods are used to HPV types are defined by genomic analysis and collect exfoliated cells: swabs, scrapes, brushings, therefore represent genotypes. At present, an 11W lavages and urinary sediments. Tissue specimens genome is described as a new HPV type if the are obtained from biopsies or surgical specimens nucleotide sequences of its Еб, Е7 and L1 genes and therefore only from subjects with disease. (about one-third of the genome) differ by more Ethical restriction on taking biopsies from healthy than 10% from those of any previously described individuals precludes the use of tissue specimens in HPV type. Although this definition was established control subjects. arbitrarily, it appears to be sufficient to define the Whatever the specimen, the major drawback of five natural taxonomic units. HPV types differing PCR-based methods is the risk of contamination as by 2-10% are called subtypes and those differing by a result of the high sensitivity of these methods. less than 2% are called variants. Short-terпt longitu- The entire process (from sample collection to the dinal studies, such as those aimed at determining PCR assay) is susceptible to contamination. Thus, the natural history of HPV infections, ideally should it is recommended that strict negative controls be rely on the identification of variants, but genome strategically interspersed during every step of the sequencing is too time-consuming and costly to be assay (Gravitt & Manos, 1992). applied to large-scale epidemiological studies.

128 8iomarkers for biolog"scal agents

Applications of the biomarker in eрidеэluо1ogiсal 15 to 73 foi high-grade CIN, and 15 to 46 for inva- studies sive cervical cancer (Munoz & Bosch, 1996). Biornarker as the end-point Serological assays using various early or late Natural history of HPV infections. Prevalence surveys HPV antigens have shown that, in general, anti- of HPV DNA in various populations have shown body responses are higher in women with cervical that most HPV infections are latent or subclinical, as neoplasia than among control women, but the most individuals in whom HPV DNA is detected do magnitudes of the ORs are lower than those based not develop signs or symptoms. Short-term longitu- on HPV DNA detection (Dinner, 1994). dinal studies have shown that most HPV infections are transient, especially at younger ages, and that per- Cohort studies. Several cohort studies have been sistent infections, although less common, pose a reported, but only three have used high-grade CIN higher risk of pre-neoplastic lesions. as the end-point and accurate methods of HPV DNA detection. All of them suggest that persistent Transmission studies. Results from various longitu- infection with high-risk HPV types precedes the dinal studies indicate that genital HPVs are trans- development of high-grade CIN (Munoz & Bosch, mitted primarily through contact with infected tis- 1996). Recent observations suggest that persistent sues during sexual intercourse. Infections of the HPV infection, particularly with a high viral load, oropharynx and larynx may be related to orogenital is a predictor of progression to advanced CIN, contact, and mother to infant transmission has although the truly prospective estimates were low been proposed to explain recunent respiratory (Ho eta!,, 1995). papillomatosis in infants and children. An association between HPV and other anogen- ital cancers (cancers of the anus, penis and vulva) Use of biomarker in case—control/prospective cohort has been reported in a few case—control studies and studies of cervical cancer has been suggested for cancers of the upper aerodi- The epidentiological evidence indicating that some gestive tract (IARC, 1995). HPV types are human carcinogens at least for the uterine cervix is compelling and it is based on an Limitations of the evidence. Case—control studies suf- impressive and largely consistent set 0f case series fer from inherent temporal ambiguity concerning and case—control studies and some cohort studies exposure and disease outcome. Thus, the higher ('ARC, 1995). prevalence of HPV DNA among cases than among Case series from many populations around the controls could be interpreted in two ways. If we world have shown the presence of HPV DNA in the assume that a single measurement of HPV DNA is majority of high-grade CIN lesions (CIN II—III) and a good marker of cbronic persistent infection, HPV cervical carcinomas. In the largest study, in which DNA detected at recruitment of cases and controls about 1000 invasive cancer specimens were could be regarded as a marker of an HPV infection analysed in a single laboratory using a reliable that preceded the cancer development. PCR-based assay, a prevalence of 93% was found, Alternatively, HPV DNA could be more readily with no significant geographical variation in over- detected in tumoral cells than in normal cells or could be a marker of an opportunistic infection with HPV. аll positivity. HPV 16 was present in 50% of the specimens, HPV 18 in 14%, HPV 45 in 8% and Direct evidence in support of the first possibil- HPV 31 in 5% (Bosch et aI., 1995). ity can only be derived from long-term follow-up studies, and few such studies are now available. In Case—control studies. Although a large number of addition, indirect evidence may be obtained from studies have been published, only 10 fulfil the epi- the trend of HPV DNA prevalence by time since demiological requirements of a case—control study last sexual intercourse, because the major route of and have used accurate HPV DNA assays. Six were transmission is sexual. Data from studies in Spain on high-grade CIN lesions and four were on inva- and Colombia show a stable high rate of 12V DNA sive cervical cancer, In all of them, a strong associ- positivity both in women with cervical cancer who ation with certain HPV types (16, 18, 31, 33, 35) reported being sexually active at the time of the was detected, with odds ratios (ORs) ranging from interview and in women who had their last sexual

129 Application of Biomarkers in Cancer Epidemiology

intercourse many years before entry into the study genital HPVs (Lungu et cl., 1995). Standardized (Munoz et al., 1992; Bosch et al., 1993). reagents for these three methods are needed. The possibility of enhanced detectability in Although the two methods that amplify a region of tumoral cells is unlikely, because the HPV DNA the L2 gene are able to detect HPV DNA in over prevalence in precursor lesions (CIN II—III) is as 90% of invasive cervical cancers, recent studies high as in invasive cervical cancer. Running con- indicate the possibility that neither of these systems trary to the argument that 12V is an opportunistic will readily detect some of the HPV intratypical infection is a great deal of laboratory data indicat- variants of 12 genital HPVs (Stewart et al. 1996). ing that DNA and transcripts of specific HPV types PCR-based assays in an ELISA format to distin- are usually detected in tissue specimens from cer- guish high-risk HPVs (16, 18, 31, 33, 35, 39, 45 51, vical cancer and its precursor lesions, and that 52, 54, 56 and 58) from low-risk HPVs (6, 11, 34, 40, high-risk HPVs are able to immortalize human 42, 43 and 44) have been developed (Tacobs et al., cells arid their oncoproteins are able to interfere 1995) and are being commercialized. These methods with the functions of negative cellular regulators. will be of great help in screening programmes and Results from case—control studies and our inter- clinical management of HPV-associated lesions and national prevalence survey of HPV DNA in inva- as a first step in epidemiological investigations. sive cervical cancer indicate that over 95% of these The ELISA assay to detect antibodies to virus-like tumours can be attributed to certain HPV types. particles was, until recently, considered the most The fact that only a small minority of the persis- promising serological assay for the detection of HPV- tent HPV infections progress eventually to cancer specific antibodies. However, this assay appears to indicates that there must be othei factors or cofactors have low sensitivity and specificity, as antibodies are that increase the progression rate to malignancy. not detected in 25-50% of women with HPV I6- Two types of cofactors may be of importance: positive cervical cancers, and cross-reactivity with other HPV types has not yet been excluded. In addi- • Host factors that modulate the effect of HPV, tion, the antibody response is low-titred and it such as genetic factors [HLA or major histo- declines with time. Accurate serological assays which compatibility complex (МНС) haplotypes], aie able to distinguish past transient infections from genetic or induced immunosuppression, persistent infections need to be developed. endogenous hormonal factors, reflected in the associations with high parity detected in our Biomarkers for Helfcohacter pylori studies, as well as early age at first sexual inter- I-Ielicobacter pylori is a spiral or slightly curved course, which could be regarded as a surrogate Gram-negative bacterium that is uniquely adapted of early age at first I-IPV infection. to survive the acidic environment of the stomach of • Exogenous factors. In the studies in Spain, humans and non-human primates. H. pylon strains Colombia and Brazil, only long-term use of oral are genetically heterogeneous, can be cultured in a contraceptives emerged as a cofactor among microaerophilic environment, are sensitive to HPV-positive women. However, our observa- most antibiotics in vitro and are characterized by tions need to be confirmed in other populations very strong crease activity. H. pylon is recognized as and in larger studies. the main cause of gastritis and peptic ulcer, and has also been associated with stomach cancer and Future directions gastric lymphoma. Colonization of the gastric Epidemiological studies aimed at investigating the mucosa and subsequent development of gastritis natural history of HPV infections (prevalence, inci- are dependent on bacterial factors such as motility, dence, persistence) and their role in various dis- strong unease activity aid specific adherence to eases rely on PCR-based assays. The two most gastric epithelium. widely used methods, which amplify a region of the L1 gene of genital HPVs (МУб9—МУ11 and Methods for detection of H. pylori and some GP5+/GP6+), need to be formally compared with sources of variability each other and also with the recently described Methods for detection of H. pylon infection have PCR-based assay that amplified Eб sequences of been summarized in 'ARC (1994b). In brief, these

130 Biomarkers for Ьiological agents use as biological specimens gastric biopsies, smears Applications of the biomarker in epidemiological on gastric material, gastric juice, breath tests and studies serology. Bromarker as the end-point Gastric biopsies are collected before treatment Seroprevalence surveys have shown that H. pylon from both the antrum and the corpus with stan- prevalence is highest in developing countries, dard forceps and can be cultured after (1) placing where 80-90% of the population may be infected them in saline (analysis within 4 h); (2) placing by early adulthood, whereas in developed сип them iп transport medium (analysis after upto 24h); tries, in the age group 25-34 years, the proportion or (3) freezing them at —70°C or in liquid nitrogen of antibody-positives ranges from 10-60%. The (delayed analysis). prevalence increases gradually with age; it is higher Methods based on biopsies include: (1) a rapid in lower socio-economic groups and when crowd- urease test; (2) histological examination stained ing in childhood is reported. Transmission occurs with the standard haematoxylin-eosin or the mod- from person to person, but the exact mode has not ified Giemsa stain; (3) bacteriological tests includ- been elucidated. ing Grain staining or culture; (4) PCR with primers that encode urease, 1БS ribosomal RNA or specific Use of biomakers in case—coлtrol/prospective cohort genes of pathogenic relevance, such as the cagA studies of gastric cancer gene. The epidemiological evidence linking H. pylon to The techniques used for gastric biopsy speci- stomach cancer aid gastric lymphoma was sum- mens can also be used with gastric juice samples. marized in IARC (1994b) as follows: Techniques based on faecal spedmens and cultures from dental plaque and saliva are still in an еаrlу • Ecological studies. Six out of eight studies j1 stage of development. which the prevalence of H. pylori was correlated Urea breath tests were developed taking advan- with concurrent or earlier incidence or mortality tage of the ability to hydrolyse urea by the strong from cancer of the stomach show no correlation urease of H. pylorL In the urea breath test, urea between the two variables, and in the other two labeled with z3С02 is fed and subsequently elimi- studies a weak correlation was observed. nated in the breath. Breath samples are collected • Case series. In 10 case series of stomach cancer before and 30 min after absorption of labelled urea in which H. pylon was detected by histology, the and analysed for 13С02 by mass spectrometry. High proportion of positives ranged from 43 to 83%, levels indicate infection. Similar tests involve the and it was over 90% in two case series of gastric use of 14С-иxеа, as 14С02 can be measured easily Iymphomas. with a scintillation counter, but some concern has • Case—control studies. At least 10 case—control been expressed over the use of a radioactive iso- studies have addressed the association between tope. Low-dose tests are being developed to over- prevalence of H. pylon antibodies arid gastric come this problem. cancer with inconsistent results. In low-risk A variety of serological assays have been used, countries for gastric cancer three out of four including complement fixation, haemagglutina- studies reported significantly increased risk with tion, bacterial agglutination, immunofluorescence ORs around 2-3. In high-risk countries, four out and ELISA, which is currently the technique of of six studies have been negative (Mufioz & choice. Pisani, 1994). Various ELISÂ commercial kits have been com- •Cohort studies. Five case—control studies pared and large variations have been observed, nested within cohorts have been reported, and especially when the assay was repeated on different in three of them an increased risk of gastric can- days (Feldman & Evans 1995). cer was observed, with prospective RRs ranging Currently, the most accurate 'gold standard' for from 2.8 to 6.0, while the other two failed to a positive is probably culture supplemented with a show an association (Munoz & Pisani, 1994). histology or biopsy urease test. Iii epidemiological studies involving children, the breath test is usu- Although the working group at the LARC judged ally used as gold standard. the available evidence as sufficient to classify

131 Application of Biomarkers in Cancer Epidemiology

H. руIoтi as carcinogenic to humans, it seems to us antigen (HBsAg), the hepatitis B surface antibody that the quality of most of the epidemiological (HBsAb), the hepatitis B core antibody (IBcAb), studies evaluated is such that biases and confound- the hepatitis B e antigen (HBeAg) and the hepatitis ing cannot be ruled out (Munoz & Pisani, 1994). B e antibody (iBeAb). A recognized limitation of current serological PCR technology is rapidly evolving and several methods of 1. pylori detection for epidemiological methods and variants aie being proposed to studies is that the amount of H. pylon antibodies increase the accuracy of detecting HBV markers may be affected by the presence of stomach cancer. and to introduce quantitative measurements of It has been observed that H. pyioni does not colo- viral DNA (see recent examples in Gunther et al., nize gastric mucosa that has undergone intestinal 1995; Maia et al., 1995; Ranki et al., 1995). Other metaplasia. Thus, patients with gastric cancer methods are being developed to amplify sequences showing extensive areas of intestinal metaplasia in of HBV and HCV simultaneously(Nedjar et al., the non-cancerous mucosa might have a low bac- 1994; Hu et al., 1995). If validated, these methods will terial load and, consequently, low titres of anti- make a remarkable contribution to field epidemiol- bodies which may be difficult to detect. ogy. PCR also facilitates analysis of the sequence of the amplified genomes. PCR followed by sequenc- Future directions ing has also been used for subtyping HBV and for It is possible that H. pylon strains prevalent in high- characterizing and identifying HSV mutants. risk populations for stomach cancer differ anti- HBV is almost unique for cancer epidemiology genically from those prevalent in low-risk areas. in that typical patterns of serological markers in Thus, development of serological assays using anti- HBV infection have been described and are useful gens derived from local H. pylori strains and vali- in interpreting HBV exposure (see Table 1). dated in the populations under study should be encouraged. Logistics of sample collection In addition, two types of H. pylon strains are now Blood samples are among the easiest specimens to recognized. Type 1 strains possess cagA (cytotoxin- collect in field studies. Preprocessing and long-term associated gene A), a gene encoding a high-molecular- storage for HBц markers detection requires little weight immunodominant antigen and vac A more than standard laboratory equipment and (vacuolating cytotoxia gene). Type II strains do not –20°C freezers. Several of the assays described above have cog A or vac A genes. Type I strains of H. pylon are standardized, commercially available and rou- have been associated with an increased risk of gas- tinely used in blood banks and hospital laborato- tric cancer, but also of duodenal ulcer and atrophic ries. Technology transfer to developing countries ii gastritis (Figura et at, 19$9; Fox et al., 1992; Blaser Africa and Asia has been successfully achieved. et al., 1995; Kuipers et al., 1995). ELISA serological assays to detect antibodies to cag A and vac A рrо- Sources of variation across studies tein have been developed (Fox et at, 1992; Blazer Sources of variation between studies include the et at, 1995), but have not yet been validated. These following: assays, once validated, may be of value in epidemiological studies addressing the association 1. The laboratory assays used. between H. pylori and stomach cancer. 2. The lack of detail on the histological classification of cases. Hepatitis B viral markers 3. The use of hospital controls with different Methods for HBV detection and гnterprelation of stringency criteria for inclusion and exclusion. This results is particularly relevant with respect to the inclusion HBV infections--acute, persistent or resolved—are of patients with other liver diseases as controls. detected on the basis of serological assays for viral 4. In several instances, the HBsAg prevalence antigens and antibodies, as well as detection of among hospital controls is used to estimate the nucleic acids. In serum specimens, the most com- population attributable fraction (AF). It has monly used markers in epidemiological studies been reported that hospital patients have related to liver cancer are the hepatitis B surface higher HBsAg prevalence rates than the general

132 Biomarkers for biological agents

Anti-HBc

Infection status HBsAg 1gM Total HBeAg Anti-HВе Anti-HBs Acute infeetiona + + + + - — . Chronic infection with high levels of viral replication + — + + — - + - Chronic infection with low levels of viral гeplXаlionЬ + — + — Recovery from acute infection before development of anti-HBs — + + — + — Low titre; possible false-positive — — + — — — t- High titre; possiЫe'low-level carrier . — — — + - Recovery from acute infection, indicating immunity —. — + — + + Vaccine response° — — — — + SusceptiЫe to HBV infection — — — — —

aReactivaled chronic disease may have this pattern with sensitive anti-HBc 1gM assays. b5оте patients may be seronegative for HBeAg . °in unvaccinated individuals, a high titre may represent immunity or be ncn-spocilic; low titres are often non-specific. end aпli-НВС (From 1ARC, 1994а.)

population (Maynard et al., 1989; Val Mayans the undiluted specimens, and 40% had errors with et al., 1990), thus underestimating the OR and the diluted specimens (Quint et al., 1995). perhaps overestimating or underestimating the АР. Although the biological matrix of interpretation of the HBV profile is fairly consistent, in some sit- Evidence is now accumulating that HBV carriers uations the wtu1agica1 results are affected by the might not have HBsAg ni their sera. These are usu- coexistence of additional factors. ally IBcAb-positives without HBsAg who are Among subjects with inflammatory hepatopathies, shown to harbour HBV DNA in the sera and in the patients on haemodialysis, intravenous drug addicts, liver by the recently applied DNA amplification homosexuals and also in a fraction of spontaneous techniques (PCR), monoclonal antibodies and blood donors, anti-HBc may appear as the sole Southern blot assays (Nakajima et al., 1989). The marker of HBV exposure (reviewed by Levine etal., use of these methods in epidemiological studies 1994). Studies have shown that, in these patients, might increase the accuracy of our estimates of the HBsAg is often immunocomplexed (80% in in- relative risk for hepatocellulaI carcinoma (ICC) flammatory hepatopathies, 40% in patients on among HBV carriers arid the fraction of the disease haemodialysis, 63% in intravenous drug addicts), that is attributable to HBV. escaping detection by standard tests. Moreover, Contamination remains a ma]от difficulty of among patients with complexed HBsAg, 48% were studies based on PCR methods. Negative and posi- HBV-DNA-positive by nested PCR, compared with tive control samples, including reaction mixtures 19% among patients without complexed HBsAg. without DNA, should be analysed in each test These results suggest that some of the patients with (Dusheiko et aI., 1992; Seelig et al., 1992). One large anti-Ilk may only develop an active HBV infec- validation study using a panel of serum specimens tion Uоller- еmelka et al., 1994). was organized in 39 laboratories to evaluate variation in the detection of HBV DNA using non- Epidemiological studies using HBV biomarkers standardized PCBs- Of 43 sets of results, 10 (23.3%) Biomarkers as the end-point correctly identified the entire panel of specimens The majority of the epidemiological studies on (seven HBV-DNA-positive specimens, five HPV-DNA- HBV and ICC have been based on the one-time negatives and two dilution series). Nineteen (44.2%) detection of HBsAg, HBsAb and IBcAb. Some stud- had some false-positive or false-negative results with ies have also evaluated iBeAg and HBeAb. The Application of Biornarkers in Cancer Epidemiology

persistence of HBsAg in the serum following inf ес- 1. The geographical variation in the prevalence tim can be viewed as а surrogate marker of the risk of HBsAg carriers and its correlation with the of developing ICC and an intermediate end-point incidence of ICC (Szmunеss, 1978). for natural history studies. 2. The intrafamilial and intravillage clustering Likewise, the persistence of detectable levels of of IBV infections (Barrett et al., 1977; Whittle HBsAb can be interpreted as а measure of protection etal., 1991). conferred by natural infection or following vacci- 3. The strong association of HBV and ICC in nation. Some examples and principal results of more than 40 case—control studies. HBsAg car- studies conducted using these HB markers include riage is the strongest determinant of ICC. In HBV transmission studies and studies to evaluate some studies, the presence of IBcAb is also HBV vaccine efficacy. associated with the risk. Iп low-risk countries in The five basic IBV markers have been extensively Europe and the USA, the OR estimates range used to investigate HB transmission from HBsAg-car from 5 to 40 and the population AF ranges from rier mothers to newborns, between sexual couples or 1 to 50%, with a median value around 20%. In between siblings. The detection of HBeAg/Ab has high-risk areas in China and South-East Asia, been useful in ascertaining the increased risk of ORs (10-30) and AFs (range 40-90%) tend to be transmission from mothers to newborns if the higher. mother is a carrier of the IBeAg. Furthermore, 4. The independence of HBV from other risk recent work from the Gambia has shown that it is factors and the specificity of the association unlikely that bed-bug bites are responsible for НB with liver cancer (Iargely ICC) as opposed to transmission (Vail Mayans et al., 1990). other cancers metastatic to the liver. Vaccine efficacy has been assessed using short- 5. The risk of ICC among IBsAg carriers in sev- term follow-up studies in which the development eral prospective cohort studies (Beasley et al., of HBsAb in vaccinated children is used as one of 1981; Iclahon et aI., 1990; Hall et al., 1985). the end-points. Prevention of the development of In some cohort studies, the risk of HCC was the IBsAg-carrier state (i.e. by comparing with a highest for HBsAg carriers who were also HBeAg control group) provides surrogate estimates of the carriers (Sakuma et al., 1988). As in case—control prevention of ICC to be expected irr the future. studies, risk estimates for IBsAg and attribut- These follow-up studies are also useful in estimat- able fractions are high in high-risk countries ing the duration of the protection afforded by 113 such as Taiwan (Beasley et al., 1981) and in vaccination Asian migrant populations in low-risk countries Newly developed biomarkers may modify our such as the USA (Sherman et al., 1995). The risk current estimates of the risk related to lEV. One of progression is low or very low in the native relates to the existence of 113V mutants that may populations of low/intermediate-risk countries escape detection using conventional methods; the such as Canada (Villenueve et al., 1994) and other source of HBV exposure misclassification Italy (De Francis et al., 1993) relates to the HBV infections with low replication 6. The efficacy of lEV vaccines in preventing level in which PCR techniques are required to the IBsAg-carrier state in all settings (McMahon detect minute quantities of HBsAg or HBV DNA. et aL, 1987; Hsu et al., 1988; Tsen et al., 1991; Whittle et al., 1991). Use of biomarkers in case—control/prospective 7. Interactions of HBV and aflatoxin (AF) in cohort studies of NCC the causation of liver cancer. Epidemiological Hepatitis B is remarkable in that several antibody studies are increasingly using biomarkers of profiles can be described and attributed with some exposure to AF. Two of the most promising are exceptions to specific stages of the host/viral inter- urinary metabolites and adducts of AF and the actions (Table 1). These markers were sufficient to mutation patterns of the p53 tumour suppressor establish a strong link between the persistence of gene. IBsAg in serum and the risk of HCC. In brief, In addition, HBV markers have been used to epidemiological studies have been able to demon- identify high-risk groups and in the evaluation of strate: treatment modalities:

134 Biomarkers for ыоlogiса agents

1. Regular screening of HBsAg carriers for early [any of countries where AF exposure is rare liver cancers has been attempted in high-risk (Ozturk, 1991). Other reports with limited num- populations. The results in terms of improving bers of cases indicated high rates (i.e. >50% of the survival of the cases identified are not consis- p53 codon 249 mutations) in ICC specimens from tent (McMahon et al., 1991; De Francis et al., Quidong (Hsu et al., 1991), Senegal (Coursaget et 1993; Villenueve et aI., 1994; Colombo 1995). aL, 1993) and South Africa (Bressac et aI., 1991). 2. HBV vaccines are also under trial as a treatment Lower frequencies were reported from ICC in for IBsAg carriers. In these studies, measure- Thailand, where AF contamination of foods has ments of HBV load are used as biomarkers in the been documented in the past (Hollstein et ai,, evaluation of therapy (Poll et aL, 1993). 1993), and among Eskimos (De Benedetti et ы., 1995). The genetic changes associated with ICC The AFB,-N-Guа adducts and other AF metabo- are just beginning to be understood (Tabor, 1994). lites have been used in an ongoing cohort study in These new biological markers may represent a Shanghai, Chia (Ross et aI., 1992; Qian et aL, real breakthrough in the field of ICC epideniiol- 1994). This study showed that seropositivity to ogy. In particular, the new genetic markers can be IBsAg and the presence of one to four AF exposure determinant in quantifying the responsibility of markers in urine (AFB , АFР„ AFМ1 and AFВ1N-Gua) AF as an independent cause of ICC and in evalu- were strongly related to ICC (OR = 59.4; ating the likely interactions with the hepatitis 16.6-212.0). There was also a moderate increase in viruses in humans. risk for subjects exposed to AF who were HBsAg- However, a word of cation should be raised negative (OR = 3.4; 1.1-10.0) (Qian et al., 1994). concerning these pioneer studies in relation to: Subsequently, it was indicated that the anti-HCV (1) their small sample size and limited methodol- prevalence in the ICC cases found in the cohort ogy with respect to criteria for specimen inclusion; was about 1%, aid in the control group was about (2) inadequate adjustment of the correlations for 0.2% (OR = 5.0; 0.3-79.9) (Yuan et al., 1995). This exposures to other (viral and non-viral) risk factors study is so far the best evidence of an interaction at the individual level; (3) limited information on between HBV and AF in the causation of human the sensitivity and specificity of the assays to assess ICC. However, the number of ICC cases in which proposed genetic markers; and (4) insufficient the interaction was explored is small (13 cases AF- knowledge on the additional genetic changes asso- positive and HBsAg-negative) and there is room for ciated with ICC development. misclassification of cases in relation to their viral exposure (thus risk estimates become unstable). Future directions Better measurements of HBV exposure (i.e. detec- In spite of the large number of studies addressing hon of HBV DNA in serum by PCR) have shown natural history issues, we still lack an explanation that in areas where HBV is prevalent in South-East for the occurrence of a substantial fraction of HBV Asia, HSV DNA can be identified in up to 20% of infections in most settings. Natural history studies the HBsAg-negative ICC cases (reviewed by should be pursued based upon the newer methods Paterlini & Brechot, 1994) and in 30-32% of of НВV detection currently available. patients with chronic liver disease (Zhang et al., PCR methods and new assays based on different 1993). The follow-up of this cohort is awaited with regions of the viral genome (i.e. HBxAg and anti- great interest. HBx; Horiike et al., 1991), pre-5 antigens and anti- The presence of AF-specific mutations in the bodies (Itoh et al., 1986; Coursaget et ai., 1990) genome of HCG cells was suggested by an ecologi- may be useful in the future to increase precision in cal study iп which ICC biopsies from patients in the quantification of the HBV involvement in the 14 countries were investigated. Among 72 ICC incidence of 1CC. Studies on virai integration may specimens from South Africa and the south-east contribute to the description of the mechanisms coast of Asia (assumed to have been exposed to of carcinogenesis. AF), 12 (17%) showed a specific G to T mutation at The coexistence of HBV and HCV, with or with- codon 249 of the р53 gene. The mutation was not out 'DV, may be greater than 10%. The mecan- found in any of 95 1CC specimens from a miscel- isms by which HCV seems to suppress HBV and

135 Application of Biomarkers in Cancer Epidemiology

HDV replication remain unknown. The impact of hepatitis, with the results of second-generation multiple hepatitis infections on 1CC incidence anti-HCV assays and with Liver histology, and are requires further research. Follow-up studies of useful in monitoring the response of patients to haemophiliacs and drug addicts may provide interferon therapy. Nevertheless, reproducibility unique opportunities for research. between laboratories, each using their own proto- Molecular techniques are identifying genetic cols, for detection of anti-HCV has been poor changes related to exposure to liver carcinogens, (Zaaijier et al., 1993). such as aflatoxins (Ozturk, 1991), or to cell foci Well-controlled procedures for handling sam- believed to be ICC precursors (for a review see ples, extraction and purification of nucleic acids, Tabor, 1994; Hsia et al., 1994). avoidance of laboratory contamination and use of Polymorphism of the МНС has been suggested appropriate negative and positive controls are as playing an important part in the establishment essential prerequisites for the PCR assay. Selection of the HBV persistent infection (Thursz et al., of primers from the highly conserved 5' non- 1995). Other work suggests that non-response to coding region is also important for sensitivity and HBV vaccines may be related to 'LA groups (A. has allowed identification of a broad range of Zuckerman, personal communication, 1996). More genotypes (Okamoto et яL, 1990); studies using these susceptibility markers are 5equence variation analysis allowed the descrip- needed. tion of distinct types of HCV with some geograph- The investigation of HBV mutants is in constant ical variation in the type-specific prevalence progress. The clinical significance of some of (Dusheiko et al., 1994). Six major types were rec- these mutant viruses in the induction of ICC is ognized in 1994 (Simmonds et al., 1994). However, unknown. these classifications are contingent on completing Increasing the precision of the estimates con- sequencing and it is expected that a recognized sys- cerning HBV (and 1CV) will allow better estimates tem of HCV classification will be available in 1996. of the role of other putative factors such as Some commercial kits are available for HCV detec- tobacco, alcohol and oral contraceptives. It should tion and typing (Bréchot & Thiers, 1996). also help in understanding the role of liver cirrho- HCV RNA testing in mononuclear peripheral sis and other chronic liver conditions as interme- blood cells offers the possibility to detect HCV diate pre-neoplastic stages. replication in patients who are HCV-RNA-negative iп serum (Muratori et al., 1994). The clinical value Hepatitis C viral infections of these tests is uncertain, but it has been suggested Methods for the detection of HCV infection and that HCV replication iп mononuclear cells can be sources of variability responsible for the reactivation and acute episodes The detection of infection by HCV is based upon of HCV hepatitis during chronic infection and assays for viral antibodies and viral nucleic after transplantation (Qian et al., 1992). These acids. Tests to detect exposure to HCV have been in markers have not been used in epidemiological constant development since first reported in 1989 studies. (Kuo et al., 1989). Second- and third-generation FUSA assays, confirmatory assays with recombi- Logistics of sample collection nant imniunoblot assays (RIBА 1-4) and quantita- Epidemiological studies have been based on sero- tive tests based on PCR are now available. logical assays. The logistics required for blood col- TICV infection can also be assessed by detecting lection in the field are similar to the description HCV RNA by reverse transcription (RT) and PCR, pertaining to HBV. which is highly sensitive and has been used for early diagnosis. Quantitative PCR can be used to Applications of biomarker in epidemiological detect 5-30 molecules of synthetic HCV RNA studies (Hagiwara et al., 1993). Biomarkers as the end-point Testing by PCR has become the standard for The presence of HCV markers in serum can be some workers. The results of these tests correlate interpreted as a surrogate measure of the risk of well with the risk for transmitting post-transfusion developing liver cirhosis and ICC. Most of the

136 Biomarkers for biological agents available evidence refers to natural history studies Use of biomarker in case—control/prospective cohort and therapeutic intervention studies. studies of NCC More than half of the cases of HCV exposure Epidemiological studies are largely consistent in remain of obscure origin, in spite of intensive ques- showing a strong association between anti-HCV tioning. Person to person exposure is the most and ICC. The specific potential of each of the likely source, but the mechanism is poorly defined. ACV types and their variants to induce ICC, as Among haemodialysis patients, HCV contamination well as the potential for other factors from the host has been documented in wards with strict control and the environment to inteiact with HCV in the of blood products with second-generation assays. origin of ICC, still requires further research. DNA sequencing of the hypervariable region of the More than 30 case series on anti-HCV preva- E2 gene in PCR products identified clusters of lence in ICC, liver cirrhosis and chronic hepatitis closely related viruses, suggesting common sources have been reported. The range of HCV seropositivity other than blood (Allender et cL, 1995). Other in ICC cases from developed countries is 20-76%, studies based on PCR have shown the presence of with most studies in the 50-70% range. Apart from HCV RNA in tears and other body fluids, suggest- exceptions, the prevalence rates from studies that ing a possible alternative source of exposure used early assays are not significantly distinct from (Feutch et al., 1994; Su et al., 1994). 1CV RNA titre those that used second-generation assays. in the serum correlates with the presence of the The coexistence of IBV and HCV markers is rel- viral RNA in seminal fluid or saliva of infected atively rare among ICC patients. Studies based on males (Liou et L, 1992) and with the rate of trans- second-generation anti-1CV assays are consistent mission from mothers to infants (Ohto et a?., in showing prevalence rates of dual infection 1994). These results suggest that serum HCV RNA among cases below 10%. Relatively high rates titre is a marker of HCV infectivity. (7.7%) have been found in Italy (Stroffolini et cL, The mode of acquisition of HCV is of relevance 1992) and 5enegal(4.1%) (Couгsagеt et aI., 1992). to progression to chronic hepatitis and liver drrho- These results are consistent with the hypothesis sis. Several studies have shown that transfusion- that a substantial fraction of the anti-HCV-positive acquired HCV conveys a poorer prognosis than subjects observed in the developed countries result both sporadic hepatitis C Uovë etaL, 1988; Mattson from transfusions with IBsAg-screened blood et aL, 1989) and hepatitis C following Lv. drug use products. (IDU) (Gordon et aI., 1993). Studies of viral load Cohort studies have shown an increased risk for using serum titre of HCV RNA have shown that ICC among anti-HCV carriers. It is estimated that transfused patients had a higher viral load than about 50% of the HCV infections will lead to both patients infected through 'DU and health chronic liver disease, of which 20% will develop workers (Lau et cL, 1993). Other studies suggest that liver cirrhosis and perhaps 10% of these with cir- long-term outcomes of HCV infection may be pre- rhosis will progress to ICC ('ARC, 1994a). Some dicted by the severity of the initial hepatic lesion studies estimate the average time interval from postinfection (Schmid eta?., 1982; Titi et al., 1990). post-transfusion HCV hepatitis to chronic liver dis- Although there are few long-term follow-up ease, liver drrhosis and 1CC at around 10, 20 and studies of HCV-infected individuals, current evi- 30 years, respectively (Кiуоsаwа et ai', 1984). dence suggests that few people resolve 1CV infec- Case—control studies on ICC using first- and tons spontaneously, and HCV RNA can be found second-generation assays are numerous and larger by PCR in most instances (Alter et al., 1992). consistent in showing elevated risk estimates There is also evidence of familial clustering of among anti-HCV carriers. In Japan, a study of 91 HCV. Mother to child transmission has been doc- 1CC cases and 410 controls from the general population reported an OR of 52 (24-111) with an umented. Sеxuаl contacts of anti-HCV carriers are at a moderately high risk of HCV. Patients attend- 1CV AF of 60.5% (Tanaka et al., 1991). In Italy, a ing STD (sexually transmitted diseases) clinics study of 65 cases and 99 controls reported an OR show higher HCV prevalence than the general of 27 with an AF of 77% (Stroffolini et al., 1992). population, although sexual transmission seems to Increased risk estimates have occasionally been occur at a lower rate than for НIV or IBV. obtained even in areas where 11W is the

137 Application of 8iomarkers in Cancer Epidemiology

predominant etiological factor of HCG, e.g. in Publications No. 119). Lyon, International Agency for Guanxi, southern china (Okun et aI., 1994) and Research on Cancer, pp. 67-74 in Korea (Lee et al., 1993). In these two areas, the Barrett, D.H., Burks, J1,, McMahon, B., Elliott, S., prevalence rates of anti-HCV in all ICC cases (pos- Berquist, K.R., Bender, Т.R. & Maynard, J.E. (1977) itive and negative for HBsAg) were 5 and 17%, Epidemiology of hepatitis B in two Alaskan communi- respectively, and in HBsAg-negative ICC cases ties. Am. J. Ерidеmiol., 105, 118-122 they were 1.8 and 43%. In a series of 23 studies Beasley, Rb, Hwang, L.Y., Lin, C.C. & Chien, S.C. (1981) from different countries published in 1991-1993, Hepatocellular carcinoma and hepatitis B virus: a pгоsреc- the range of the reported ORs is 1.1-134. These tive study of 22707 men in Taiwan. Lancet; ü, 1129-1133 estimates were statistically significant in 18 of the Blaser, M'I., Pérez-Pérez, G.I., Kieanthous, H., Cover, T,L., 23 studies (for a review see also IARC, 1994a). Peek, R.M., Chyou, P.H., Stemmermann, G.N. & Other risk factors such as smoking and alcohol do Nomura, A. (1995) Infection with Helicobacter pylori not modify the strength of the association between strains possessing cagA is associated with an increased 1CV and ICC. risk of developing adenocarcinoma of the stomach. Cancer Rn., From case—control studies, risk estimates are 55, 2111-2115 consistently higher for individuals who are anti- Bosch, F.X., Munoz, N., de Sanjosé, S., Navarro, C., HCV-positive and HBsAg-positive (e.g. see Chuang ivloreo, P., Ascunce, N., Gonzalez, L.C., Tafur, L., Gill, M., Larrafiaga, I., Viladiu, P., Daniel, R.W., Alonso de Ruiz, P. еt яl., 1993). It has recently been estimated that the attributable fraction for both HBV and HCV is Anistizabal, N., Santamaria, M,, Guerrero, E. & Shah, K.V. (1993) Human papillomavirus about 50% in developed countries. and 90% in the and cervical intraepithe- lia1 neoplasia grade III!carcinoma in situ: a case-control developing world (Pisani et al., 1996). study in Spain and Colombia. Cancer EpidemioL Biomarkers Prev., 2, 415--422 Future directions Bosch, F.X., Manos, M.M., Mufioz N., Sherman, M., Genome sequencing, completion of the phyloge- Jansen, A.M., Peto, J„ Schiffmarl, M.H., Moreno, V., netic tree and agreement on a standard typification Kunman, R., Shah, K.V. & The IBSCC Study Group (1995) of HCV would help in evaluating international Prevalence of human papillomavirus in cervical cancer: variation in HCV types. Follow-up studies of 1CV- a worldwide perspective. J. Natl Cancer Inst., 87, 796-802 infected individuals should provide information on Brandsma, J.L., Burk, R.D., Lancaster, W.D., Pfister, H. & the risk factors for progression to advanced liver schiffman, M.H. (1989) Inter-laboratory variation as an disease. Natural history studies and population explanation for varying prevalence estimates of human prevalence surveys will be useful in identifying papillomavirus infection. ut. J. Cancer, 43, 260-262 background prevalence, routes of transmission and Brëchot, C. & Thiers, V. (1996) Evaluation du kit de sérotу- preventive strategies. More sensitive methods of page du VIC (Murex technical report), lurex HCV detection aid, in particular, quantitative PCR systems should be used in natural history studies. Bressac, B., Kew, M., Wands, J. & Ozturk, M. (1991) Selective G to T mutations of p53 gene in hepatocellular carcinoma from southern Africa. Nature, 350, 429-430 References Allender, T., Gruber, A. Mojgan, N. Beyene, A., Chuang, W.L., Chang, WY., Lu, S.N., Lin, Z.Y., Chen, S.C., Hsieh, .Y., Wang, L, SSderstгбш, T., BjOrkholm, I., Grillner, L. & Persson, М У., You, S.L. & Chen, C.J. (1993) The role of hepatitis C virus in chronic hepatitis М.А.А. (1995) Frequent patient-to-patient transmission of hepatitis C virus in a haematology ward. Lancet, 345, B virus infection. Gastroenterol. Jpn, 28 (suppi. 5), 23-27 603-607 СоlотЬо, M. (1995) Should patients with chronic viral Alter, М.J., Margolis, H.S., Krawczynski, K., Judson, F.N., hepatitis be screened for hepatoceIular carcinoma? Viral Hepatitis Rev., 1, 67-75 Mares, A., Alexander, W.J., lu, P.Y., Mi11еr, J.K., Gerber, M.A., Sampliner, R.E., leeks, E.L. & Beach M.J. (1992) Coursaget, P., Buisson, Y., Bourdil, C., Yvonnet, B., The natural history of community-acquired hepatitis C Infinie, C., Diop, M.T., Chiron, J.P., Bao, 0. & Diop-Mar, in the United States. New Eng1. J. Mcd. 327, 1899-1905 I. (1990) Antibody response to preS1 in hepatitis-B-virus- Barrasso, R. (1992) Colposcopie diagnosis of HPV cervi- induced liver disease and after immunization. Rn. ViroL, 141, 563-570 cal lesions. Iп: Mufioz, N., Bosch, F.X., Shah, K.V. & Meheus, A., eds, The Epidemiology of Human Coursaget, P., Leboulleux, D., Le Carin P., Bai, O. & Coil- Pap illomavirus and Cervical Cancer (IARC Scientific Seck, А.~. (1992) Hepatitis C virus infection in cirrhosis 138 Biomarkers for biological agonts

and primary hepatoce11ular carcinoma in Senegal. Trans. drug use. Hepatology, 18, 1338-1343 R. Soc. Trop. Med. Hyg., 86, 552-553 Gravitt, P.E. & Manos, M.M. (1992) Polymerase chain Coursaget, P., Depril, N., Chabaud, M., Nandi, R., reaction-based methods for the detection of human Mayelo, V., Le Cann, P. & Yvonnet, B. (1993) High preva- papillomavirus DNA. In: Mufioz, N., Bosch, F.S., Shah, lence of mutations at codon 249 of the p53 gene in hepa- K.V. & Meheus, A., eds, The Epidemiology of lamai tocellular carcinomas from Senegal. Br. J. Cancer, 67, Papillomavirus and Cervical Cancer (‚ARC Scientific 1395-1397 Publications No. 119). Lyon, International Agency for Researcho on Cancer, pp. 121-133 De Benedetti, V., Welsh, J., Trivers, G., Harpster, A., Parkinson, A., Lanier, A., McMahon, B. & Bennett, W. Gunther, S., Li, B.С., Miska, 5., Kruger, D.H., Meisel, H. & (1995) P53 is not mutated in hepatoce11ular carcinomas Will, H. (1995) A novel method for efficient amplifica- from Alaska Natives. Cancer EpidemioL Biomarkers Prey., 4, tion of whole hepatitis B virus genome permits rapid 79-82 functional analysis and reveals deletion mutants in immunosuppressed patients. J. Viril., 69, 5437-5444 De Francis, R., Meucci, G., Vecchi, M., Tatare'la, M., Colombo, M., del Ninno, E., Rami, М.~., Donato, M.F. & Hagiwara, H., Hayashi, N., Mita, E., Naito, M., Kasahara, Ronchi, G. (1993) The natural history of asymptomatic A., Fusamoto, H. & Kamada, T. (1993) Quantitation of hepatitis B surface antigen carriers. Ann. Intern. Mcd., hepatitis C virus RNA in serum of asymptomatic blood 118, 191-194 donors and patients with type C chronic liver disease. lepatology, 17, 545-550 Dinner, J. (1994) Antibody responses to defined HPV epi- topes in cervical neoplasia. Papillomavirus Report, 5, Наl1, A.J., Winter, Р.D. & Wright, R. (1985) Mortality of 35-41 hepatitis B positive blood donors in England aid Wales. Lancet, 1, 91-93 Dusheiko, G., Xu, J. & Zuckermann, A.j. (1992) Clinical diagnosis of hepatitis B infection: applications of the Ho, G.Y., Burk, R.D., Klein, S., Kadish, А.S., Chang, C.J., polymerase chain reaction. In: Becker, Y. & Darai, G., eds, Palan, P., Basu, J., Tachezy, R., Lewis, R. & Romney, S. Frontiers of virology. Berlin, Springer Verlag. pp. 67-85 (1995) Persistent genital human papillomavirus infection as a risk factor for persistent cervical dysplasia. Dusheiko, G., Schmilovitz-Weiss, H., Brown, D., J. Nail Cancer lnst., 87, 1365-1371 Mcimish, F., Yap, P.L., Sherlock, S., McIntyre, N. & ., Wild, C.P., Bleicher, F., Chutimataewin, Siшшопds, P. (1994) Hepatitis C virus genotypes: an Holistein, M.С investigation of type-specific differences in geographic S., Harris, C.C., Srivаtaпakul, P. & Montesano, R. (1993) origin and disease. HepatoIogy, 19, 13-18 P53 mutations and aflatoxin B, exposure in hepatocellular carcinoma patients from Thailand. ut. J. Cancer, 53, Feldman, R.A. & Evans, S.J.W. (1995) Accuracy of diag- 51-55 nostic methods used for epidemiological studies of Helicobacter pylori. Alime . Pharmacol. Therap., 9 (suppl. Horiike, N., Blumberg, B.S. & Feitelson, M.A. (1991) 2), 21-31 Characteristics of hepatitis BX antigen, antibodies to Xantigen, and antibodies to the viral polymerase during Feutch, H.H., Polywka, S., Zollner, B. & Laufs, R. (1994) hepatitis B virus infection. J. In fect. Dis., 164, 1104-1112 Greater amount of HCV-RNA in tears compared to blood. ., Kleiner, D.E. & Tabor, E. MiaoЫоl. Immunol., 38, 157-158 Hsia, C.C., Di Bisceglie, A.М (1994) Expression of hepatitis B surface and core anti- N., Guglielmetti, P., Rossolino, A., Barberi, A., Figura, gens aid transforming growth factor-a in "Oval Cells" of ., Russi, & Quaranta, S. Cusi, G., Musmanno, R.А М. the liver in patients with hepatocellular carcinoma. J. (1989) Cytotoxin production by Campylobacter pylori Viril., 43, 216-221 strains isolated from patients with peptic ulcers and trous patients with chronic gastitis only. J. Clin. Microbiol., 27, 'su, A.W., Burk, R.D., Liaw, K.-L., Chen, C.-J., Zhang, 225-226 T., Schiffman, M., Greer, C.E., You, S.-L., Hsieh, C.-Y., Huang, T.W., Wu, T.-C., O'Leary, T.J., Seidman, J.D. & Fox, J.G., Correa, P., Taylor N.S. Thompson, N., Fontham, Manos, M.M. (1996) I terlaborato y agreement in a E., Janney, F., Sobhan, M., Ruiz, B. & Hunter, F. (1992) п т polymerase chain reaction-based human papillomavirus High prevalence and persistence of cyt toxin-positive о DNA assay. CancerEpidemioi. BiomarkersPrev., 5, 483-484 Helicobacter pylori strains irs a population with high prevalence of atrophic gastritis. Am. J. GastroenteroL, 87, Hsu, H.М., Chen, A.S., Chuang, С.H., Lu, J.C.F., Two, 1554-1560 ~.М., Lee, C.C., Lu, H.C., Cheng, S.H., Wang, Y.F., Wang Chen, C., et aI. (1988) Efficacy of a mass hepatitis B vac- Gordon, S.C., E1loway, R.S., Long, J.C. & Dmuchowski, cination program in Taiwan. JAMA, 260, 2231-2235 C.F. (1993) The pathology of hepatitis C as a function of mode of transmission: blood transfusion vs. intravenous Hsu, I.C., Metcalf, R.А., Sun, T., Welsh, J.A., Wang, N.J. & Application of Biomarkers in Cancer Epidemiology

Harris, C.C. (1991) Mutational hotspot in the p53 gene circulating antibodies to a major etiologic virus of in human hepatoce11ular carcinomas. Nature, 350, human non-A, non-B hepatitis. Sсiепсе, 244, 362-364 429-430 Lau, J.Y.N., Davis, G.L., Kniffen, J., Qian, K.P, Urdea, Hu, K.-Q., Vu, C-H., Lee, S., Villаmil, F.G. & Vierling, J. M.S., Chan, C.S., Nfizokami, M., Neuwald, PD. & Wilber, (1995) Simultaneous detection of both hepatitis B virus J.С. (1993) Significancer of serum hepatitis C virus RNA DNA and hepatitis C virus RNA using a combined one- levels in chronic hepatitis C. Lancet, 341, 1501-1504 step l m s chain reaction technique. Hep atology, роу етае Lee, H.S., Han, C.J. & Kim, C.Y. (1993) Predominant etiologic 21, 901-907 association of hepatitis C Virus with hepatocellulаr car- 'ARC (1994a) L4RC Monographs on the Evaluation of the cinoma compared with hepatitis B virus in elderly Carcinogenic Risks to Humans. Vii. 59, Hepatitis viruses, patients in a hepatitis B-endemic area. Cancer, 72,2564-2567 Lyon, International Agency for Research on Cancer Levine, 0.5,, Vlahov, D. & Nelson, K.E. (1994) IARC (1994Ъ)'ARC Monographs on the Evaluation of the Epidemiology of hepatitis B virus infections among Carcinogenic Risks to Humans. ‚ToI. 61, Schistosomes, liver injecting drug users: seroprevalence, risk factors, and flukes and Helicobacter pylori, Lyon International viral interactions. EpidemioL Rev., 16, 418-436 Agency for Research on Cancer Hou, I-C., Chang, T,-T., Young, K.-C., Lin, X.-Z., Liii, C.-Y, IARC (1995)'ARC Monographs on the Evaluation of the & Wu,H.-L. (1992) Detection of HCV P'NA in saliva, urine, Carcinogenic Risks to Humans. Vol. 64, Human papillo- seminal fluid, and ascites. J. Med. Virol., 37, 197-202 mavirus, Lyon, International Agency for Research on Lungu, 0., Sun, X.W., Wright Cancer Т.C., J.R., Ferenczy, A., Richart, R.M. & Silverstein, S. (1995) A polymerase chain Itoh, Y., Takal, E., Ohnuma, H., Kitajima, K., Tsuda, F., reaction-enzyme-linked immunosorbent assay method Machida, A,, Mishiro, S., Nakamura, T., Miyakawa, Y. & for detecting human papillomavirus in cervical carcino- Mayumi, M. (1986) A synthetic peptide vaccine involv- mas and high-grade cervical cancer precursors. Obstet. ing the productif the pee-S (2) region of hepatitis В virus Gynecol., 8, 337-341 DNA: protective efficacy in chimpanzees. Proc. NatlAcad. Maia, M., Takahashi, H., Adler, K., Garlick, R.K. & Wands, Sci. USA, 83, 9174-9178 J.R. (1995) Development of a two-site immuno-PCR Jacobs, M.V., de Roda Husman, A.M., van den Bruie, assay for hepatitis B surface antigen. J. Viril. methods, 52, A.J.C., Snijders, Р.J.F., Meijer C.J.L.M & Walboomers, 273-286 J.M.M. (1995) Group specific differentiation between McMahon, B.J., Rhoades, E.R., Heyward, W.L., Tower, E., high- and low-risk human papillomavims genotypes by Ritter, D., Lanier, A.P., Wainwright, R.B. & Helminiak, C. general primers-mediated polymerase chain reaction and (1987) A comprehensive programme to reduce the inci- two cocktails of oigonucleotide-probes. J. Clin. dence of hepatitis B virus infections and its sequelae in Microbiol., 33, 901-905 Alaskan natives. Lancet, 2, 1134-1136 Jo11 r-Jemelka, H.1., Wicki, A.N., Grob, P.J. (1994) е Iclahon, B.J., Alberts, S.R., Wainwright, R.B., ulkow, Detection of lBs antigen iii 'anti-Hic alone" positive В L. & Lanier, A.P. (1990) Prospective study in 1400 hepati- sera. J Hepatology, 21, 269-272 tis B surface antigen-positive Alaska Native carriers. Arch. Jové, J., Shnchez-Tapias, M., Bruguera, M., Mas, A., Costa, Intern. Med., 150, 1051-1054 J., Barrera, J.M. & Rodes, J. (1988) Posttransfusional vs. McMahon, B.J., Wainwright, R.W. & Lanier, A.P. (1991) sporadic non-A, non- chronic hepatitis: a clinicopatho- В The Alaska native ICC screening program: a population- logical and evolutive study. Liver, 8, 42-47 based screening program for hepatocellular carcinoma. Kiyosawa, K., Akahane, Y., Nagata, A. & Furuta, S. (1984) In: Tabor, E., Di Bisceglie, A.M. & Purcell, R.H., eds, IepatoceIular carcinoma after non A non B posttrans.. Etiology, Pathology, and Treatment of Hepatocellular fusion hepatitis. Am. J. Gastroenteral., 79, 777-781 Carcinoma in North America (Advances in Applied Kui ers, E.J., Përez-Pérez, G.I., Meuwissen, S.G. & Blaser, Biotechnology Series, Vol. 13). Houston, PPC-Gulf р Publishing Company, pp. 231-241 М.J. (1995) Heiicobacter pylori and atrophic gastritis: importance of the cagA status. J. Nat! Cancer Inst., 87, Mattson, L., Weiland, O. & Glaumann, H. (1989) Chronic 1777-1780 non-A, non-B hepatitis developed after transfusion, illicit Kuo, G., Ohio, Q.-L., Alter, 11.3., Gitnick, G.L,, Redeker, self-injections or sporadically: outcome during long-term A.G., Purcell, R.H., Miyamura, T., Dienstag, J.L., Alter, follow-up—a comparison. Liver, 9, 110-127 M.J., Stevens, C.E., Tegtmeier, G.E., Bonino, F., Colombo, Maynard, J.Е., Kane, M.A. & Radler, S.C. (1989) Global M., Lee, W.S., Kuo, L., Berger, K., Shuster, B.R., Overby, control of hepatitis B through vaccination: role of L.R., Bradley, D.W. & Houghton, M. (1989) An assay for hepatitis B vaccine in the expanded programme on

140 Blomarkers for biologhal agents inimunizatiori. Rev. Infect. Dis., 11 (suppl. 3), s574-s578 Аеtiоlоgiсаi and Progression Factors. CRC Press, pp 167-190 Pisani, P., Parkin, D.M., Munoz, N. & Ferlay, J. (1997) Mufioz, N. Bosch, F.X., de Saп)osé, S., Tafur, L., lzarzugaza L Gill, M., Viladiu, P., Navarro, C., Martos, C,, Ascunce, N., Cancer and infection: estimates of the attributable frac- Gonzalez, L.C., Kaldor, J.M., Guerrero, E., Ldrincz, A., tion in 1990. Cancer Epidemiol. Biornarkers Prey. (in press) Santamаria, M., Alonso de Ruiz, P, Aristizabal, N. & 5hah, Poll, 5., Driss, F., Carnot, F., Michel, M.-L., Berthelot, P. & K. (1992) The causal link between human papillomavirus Brechot, C. (1993) Vaccination against hepatitis B virus; and invasive cervical cancer; a population-based case- an efficient immunotherapy against hepatitis B multi- control study in Colombia and Spain. Int. J. Cancer, 52, plication. C.R. Acad Sri. Paris, Life/sciences, 316, 688-691 743-749 Qian, C., Camps, J., Maluenda, M.D., Civeira, M.P. & Mufioz, N. & Pisani, P. (1994) He1icobactex pylori and Prieto, ], (1992) Replication of hepatitis C virus in gastric cancer. Eur. J. Gаstroenterol. Hep atot , 6, 1097-1103 peripheral blood mononuclear cells. Effect of alpha- Mufioz, N. & Bosch, F.X. (1996) Current views on the interferon therapy. J. Hepatology, 16, 380-383 epidemiology of HPV and cervical cancer. Papillomavirus Qian, G.-S., Ross, R.K., Yu, M.C., Yuan, J.-M., Gao, Y.-T., report. In; Lacey, C., ed., Papкllomavirus Reviews: Current Henderson, B.E., Wogan, G.N. & Groopman, J.D. (1994) Research on Papillotnaviruses. Leeds, Leeds University A fallow-up study of urinary markers of aflatoxin expo- Press. pp. 227-237 sure and liver cancer risk in Shanghai, People's Republic Muratori, L., Giostra, F., Cataleta, Ni., trancesconi, R., 0f China. Cancer Epidemiol. Biomarkers Prey., 3, 3-10 Ballardini, G., Cassani, F., Lenzi, M. & Bianchi, P.B. Quint, W.G. Ieijtink, R.A., Schirm, J., Gerlich, W.H. & (1994) Testing for hepatitis C virus sequences in periph- Niesters, Н.G. (1995) Reliability of methods for hepatitis eral blood mononuclear cells of patients with chronic B virus DNA detection. J. Clin. icrobi ., 33, 225-228 hepatitis C in the absence of serum hepatitis C virus М ы RNA. Liver, 14, 124-128 Ranki, M., Schatzl, H.M., Zachoval, R., Uusi-Oukari, M. & Lehtovaara, P. (1995) Quantification of hepatitis B virus Nakajlma, E., Thiers, V., Larzul, D., Thiollais, P. & DNA over a wide range from serum for studying viral Brechot, C.H. (1989) Hepatitis B virus (18V) DNA as a replicative activity in response to treatment and in rесиr markers of HBV infection. In: Bannasch, P., Keppler, D, & rent infection. Hepatology, 21, 1492-1499 Weber, G., eds, Liver CeP Carcinoma. Falk Sympоsium No. Si. Dordrecht, Kluwer Academic Publishers. pp. 107-111 Ross, R.K., Yuan, J.-M, Yu, M.C., Wogan, G.N., Qian, G.- S., Tu, J.-Т., Groopman, J.D., Gao, Y.-T. & Henderson, B.E. Nedjar, S., Mitchell, F. & Biswas, R. (1994) Simultaneous (1992) Urinary aflatoxin biomarkers and risk of hepato- amplification and detection of specific hepatitis B virus cellular carcinoma. Lancet, 339, 943-946 and hepatitis C virus genomic sequences in serum samples. J. Med. Viral., 42, 212-216 Sakuma, K., Saitoh, N., Kasai, M., Jitsukawa, H., Yoshino, I., Yamaguchi, M., Nobutomo, K., Yamumi, M., Tsuda, F., Ohto, H., Terazawa, 5., Sasaki, N., lino, K., Ishiwata, C., Komazawa, T.,Nakamura, T., Yoshida, Y. & 0kud, K. Kaki, M., цjiie, N., Endo, C., Matsui, A., Okamoto, H., & (1988) Relative risks of death due to liver disease among Mishiro, S. (1994) Transmission of hepatitis C virus from Japanese male adults having various statuses for hepati- the mothers to infants. New Engl. J. Med., 330, 744-756 tis Bs and e antigen/antibody in serum: a prospective Okamoto, H., Okada, 5., Sugiyama, Y., Tanaka, T., Sugai, study. Hepatolо у, 8, 1642-1646 Y., Akahane, Y., Machida, A., Mishiro, 5., Yoshizawa, H., Schiffman, M.H., Kiviat, N.B., Burk, R.D., Shah, K.V., Miyakawa . & Mayumi, M. (1990) Detection of hepati- У Daniel, R.W.; Lewis, R., Kupers, J., Manos, М.M., Scott, tis C virus RNA by a two-stage polymerase chain reaction D,R., Sherman, М.E., Kurman, R.J., Stoler, 1.1., Glass, with two pairs of primers deduced from the 5'-nonсоd- A.G., Rush, B.B., Mielzynska, I. & Lbrincz, A.T. (1995) ing region. Jpn J. Exp. Med., 60, 215-222 Accuracy aid interlaboratory reliability of human papil- Okuno, H., Xie, Z., Lu, B., Qin, X., Takasu, M., Kano, H., lomavirus DNA testing using Hybrid Capture. J. CIiп. Microbiol., 33, 545-550 5hiozaki, У. & moue, К. (1994) A low prevalence of anti- hepatitis C viгus antibody in patients with hepatocellu- Schmid, M., Pirovirig, M., Altorfer, J., et al (1982) Acute 1ar carcinoma in Guangxi Province, southern China. hepatitis non-A, non-B: are there any specific light Cancer, 73, 5 8-62 microscopic features? Liver, 2, 61-67 Ozturk, M. (1991) 53 mutation in hepatocellular carci- р Seelig, R., Renz, M. & Seelig, H.Р. (1992) PCR (polymerase noma after aflatoxin exposure. Lancet, 338, 1356-1359 chain reaction) in the diagnosis of viral hepatitis. Ann. Paterlini, P. & Brechot, C. (1994) Hepatitis B virus and primary Med., 24, 225-230 liver cancer in hepatitis B surface antigen positive and Sherman, М.E., Schiffman, M.H., Lorincz, A.T., Manos, negative patients. In: Brechot, C., cd., Primary Liver Cancer M.M., Scott, D.R., Kurman, RJ., Kiviat, N.B., Stolen, M.,

141 Appicalion of Biomarkers in Cancer Epidemiology

Glass, A.G. & Rush, В.B. (1994) Toward objective quality Tsen, Y.J., Chang, M.H., Hsu, H.Y., Lee, C.Y., Sung, J.L. & assurance in cervical cytopathology. Corrrelation of cyto- Chen, D.S. (1991) Seroprevalence of hepatitis B virus pathologic diagnoses with detection of high-risk human infection in children in Taipei, 1989: five years after a mass papillomavirus types. Am. J. Clin. PatiwL, 102, 182-187 hepatitis B. vaccination program. J. Med. Viril., 34, 96-99 Sherman, M., Peitekian, K.M. & Lee, C. (1995) Screening Vail Mayans, M., Calvet, X., Bruix, J., Bruguera, M., for hepatocellular carcinoma in chronic carriers of Costa, J., Estève, J., Bosch, F.X., Bm, C. & Rodés, J. (1990) hepatitis В virus: incidence and prevalence of hepatocel- Risk factors for hepatocellular carcinoma in Catalonia, luiar carcinoma in a North American urban population. Spain. Iпt. J. Cancer, 46, 378-381 Hep atology, 22, 432-438 Villenueve, J.P., Desrochers, M., Infante-Rivard, C., Simmonds, P., Alberti, A., Alter, H.J., Bonino, R, Bradley, Willems, B., Raymond, G., Boursier, M., Сôté, J. & Richer, O.W. & Brechot, C. (1994) A proposed system for the G. (1994) A long-term follow-up study of asymptomatic nomenclature of hepatitis C virus genotypes. Hep atology, hepatitis B surface antigen-positive carriers in Montreal. 19, 1321-1324 GastraenteroIogy, 106, 1000-1005 Stewart, A'-C., Eriksson, A.M., Manos, M.M., Munoz, N. Walboomers, J.M.M., de Roda Husman, A,-M., van den & Wheeler, C.М. (1996) Intratype variation in 12 human Bruie, A.J.C., Snijders, P.J.F. & Meije, C.J.L.M. (1994) papillomavirus types: a worldwide perspective. J. Viril., Detection of genital human papillomavirus infections: 70, 3127-3136 critial review of methods and prevalence studies in rela- tio to cervical cancer. In: Stern, P.L. & Stanley, M.A., eds, Stroffolini, T., Chiaramonte, M., Tiribelli, C., Villa, Е., п Human P llamav rus arid Cervical Cancer: Biology and Simonetti, R.G. Rapicetta, M., Stazi, М.A., Bertin, T., ари г Croce, S.L., Trande, P., Magliocco, A. & Chimie, P. immunology. Oxford, Oxford University Press, pp. 41-71 (1992) Hepatitis C virus infection, HBsAg carrier state Whittle, H.G., Inskip, H., Hall, A.J., Mendy, M., Downes, and hepatocellular carcinoma: relative risk and popula- R. & bare, Ѕ . (1991) Vaccination against hepatitis B and tion attributable risk from a case-control study in Italy. J. protection against chronic viral carriage in The Gambia. Hepatology, 16, 360-363 Lancet, 337, 747-750 Su, Q., Liu, Y., Zhang, . L1, D. & ang, J, (1994) Expression Ѕ У Yuan, J.M., Ross, R.K, Staпczуk, F.Z., Govindarajan, S. & of insulin-like growth factor II in hepatitis B, cirrhosis and Gao, Y.T. (1995) A cohort study of serum testosterone hepatocellular carcinoma: its relationship with hepatitis B and hepatocellular carcinoma in Shanghai, Chaina. lot. virus antigen expression. Hepatology, 19, 788-799 J. Cancer, 63, 419-493 Szmuness, W. (1978) I-iepatoce11ular carcinoma arid the Zaaijier, H.L., Cuypers, H.T.M., Reesink, H.W., Winkel, hepatitis B virus: evidence for a causal association. Frog. IN., Gerken, G. & Lette, P.N. (1993) Reliability of poly- Med. Viril., 24, 40-69 merase chain reaction for detection of hepatitis C virus. Lancet, 341, 722-724 Tabor, Е. (1994) Tumor suppressor geler, growth factor genes, and ondogenes in hepatitis B virus-associated Zhang, У.Y., Hansson, B.C., Kuo, L.S., Wide11, A. & hepatocellular carcinoma. J. Viril., 42, 357-365 Nordenfelt, E. (1993) Hepatitis B vims DNA in serum and Tanaka, K., Hirohata, T., Kopa, s., Sugimachi, K., liver is common!y found in Chinese patients with Kanematsu, T., Ohryohji, F., Nawata, H., Ishibashi, H., chronic liver disease despite the presence of antibodies to Maeda, Y., Kiyokawa, H., Tokunaga, K., luta, Y., HBsAg. Hepafology, 17, 538-544 Takeshita, S., Arase, Y. & Nishino, N. (1991) Hepatitis C and hepatitis B in the etiology of hepatocellular carcinoma in the Japanese population. Cancer Res., 51, 2842-2847

Thursz, M.R., Kwiatkowski, D., Aisopp, С.Е.М., Greenwood, B.M., Thomas, H.C. & Ill, A.V.S. (1995) Association between an МНС class II allele and clearance of hepatitis B virus in the Gambia. New Rugi. J. Med., 332, 1065-1069 Corresponding author N. Munoz . Tito, L., Sinchez, J.M., Costa, J., Jove, J., Vilella, A., Unit of Field and Intervention Studies, International Bruguera, M. & Rodes, J. (1990) Long-term follow-up of Agency for Research on Cancer,` 150 cours Albert chronic NANB (C) hepatitis C [abstractJ. . Hepatoligy Thomas, F-б9Э72, Lyon Cedex 08, France (suppl. 2), S61 Ј

142 AppliсаУlon at Biomarkers iii Cancer Epidemiology Toпiolo, R, Bouette, P., Shuker, B.E.G., flouinrrian, N., HuIka, B. arid Pearce, N., ads 'ARC ttdentjtic Fublicaticrie No. 142 International Аgепсу for Research on Cancer, Lyon, 1997 Carcinogen-DNA and carcinogen-protein adducts in molecular epidemiology

C.P. Wild and Р. Pisani Carcinogen-DNA and carcinogen-protein adducts provide an integrated measure of carcinogen exposure, uptake and âbsorption, metabolism, DNA repair and cell turnover. As such they promise to provide a more objective and relevant measure of exposure than that which can be derived from questionnaires and measures of ambient levels of carcinogen. Nevertheless, the interpretation of adduct measurements made in human tissues and body fluids requires an understanding of a number of factors. These include the sensitivity and specificity of the measurement, the temporal relationship between exposure and adduct level and the mechanistic role of the adduct in the process of carcinogenesis.The application of such biomarkers in epidemiological studies therefore necessitates careful consideration of optimal study design. The above issues are illustrated in this chapter with examples from studies in both animal models and human populations.

Chemicals can covalently bind to cellular macro- validity of this type of measurement. These issues molecules including DNA, RNA and proteins. The are of both a conceptual and a practical nature. product of this addition.' of a chemical moiety This paper therefore aims to evaluate current to a macromolecule is termed an adduct. These knowledge regarding the validity of adducts as adducts may be measured in tissues, exfoliated measures of carcinogen exposure and cancer risk in cells, peripheral blood or urine. Other types of DNA different epidemiological study designs. . damage result from exposure to chemicals or through biological processes such as inflammation, Rationale for use of adducts in exposure e.g. 8-oxo-deoxyguanosine (8-oxo-dG), although assessment these are not strictly adducts by the above definition. The rationale for using measurements of carcinogen- However, for simplicity these types of DNA dam- DNA and carcinogen-protein adducts in human age are also referred to as 'adducts' in this paper. exposure assessment is based on the assumption The measurement of carcinogen-nucleic acid or that DNA adducts formed in vivo are responsible carcinogen-protein adducts is a promising ap- for genetic alterations in genes critical for carcino- proach to determining exposure to DNA-damaging genesis and that the protein adducts formed carcinogens in humans. The degree to which such through the same processes reflect the formation measurements can be quantitatively associated with of DNA adducts. From this basic assumption, a cancer risk at the individual level is not understood. number of extrapolations are made to allow the The measurement of adducts is of potential measurements to be carried out with the available value in exposure assessment for a number of reasons. methodologies in available human samples, e.g. in First, it may provide a more objective and relevant surrogate tissues as opposed to the target tissue. As measure of individual exposure than can be ob- presented in Fig. 1, these extrapolations include tained by a questionnaire approach. Second, it is that from the specific DNA sequence level within potentially highly specific for the exposure of specific genes to the level of genomic DNA, that interest. The specificity of adducts makes them from the single cell to the target tissue or organ, particularly suitable when defined compounds' and that from the target organ to peripheral blood. carcinogens are investigated. Third, the measure- In addition, as mentioned above, extrapolations ments are often of a high sensitivity. There are, are also made from DNA to protein adducts. This however, numerous issues that can impinge on the illustrates that, in the majority of instances, what

143 Application of 8iomarkers in Cancer Epidemiology

Determinants of adduct levels in vivo Once a chemical carcinogen enters the body, the level of adduct formation will depend upon a number of factors, including the absorption and Decreasing proximity to critical lesions distribution of the chemical around the body and the organ-specific and cell-specific metabolism (activation and detoxification), The measured adduct level at any one point in time will further depend on the chemical stability of the adduct, any repair processes (for DNA adducts) and cell turnover. Thus, the carcinogen-DNA adduct repre- DNA Cell Organ Peripheral sequence blood/urine sents an integration of these different parameters at the individual level, and measures the dose of carcinogen reaching the target molecule for r- Gene Tissue Exfoliated са cells cinogenesis; this has been termed the `biologically effective dose'. It is therefore an integration of both exposure and interindividual variations in Figure 1. Measurement of carcinogen—DNA and DNA—protein metabolism and DNA repair, each of which may adducts: proximity to critical lesion. be determined genetically and/or be influenced by other environmental exposures. Some DNA adducts is actually measured is far removed from what one are removed by excision repair processes, e.g. N- would ideally measure. 5uсh 'surrogate' measures methylpurine glycosylases (Bessho et al., 1993), or require validation for their relationship to the are lost spontaneously, as for example in depuriria- 'ideal measure. tim, and these addicts can be detected in the urine An example of the above issue is the case of afla- as free bases or deoxynucleosides (Shukеr & Farmer, toxiпs where a specific mutation at the third 1992). The carcinogen-protein adducts, while not nucleotide of codon 249 of the p53 tumour sup- representing events thought to be critical in car- pressor gene in human hepatoce11ular carcinomas cinogeriesis, are of interest because they may reflect (HCCs) has been suggested to result from aflatoxin the level of DNA adducts in internal organs (Skipper exposure (Ozturk, 1991). Assuming, for argument's & Tannenbaum, 1990; Skipper et al., 1994). sake, that this is the case, the 'ideal' measure would be aflatoxin-DNA adducts in codon 249 of the р53 Dose-response studies: adducts, mutations and gene in human hepatocytes. In reality a few mea- cancer surements have been performed on total genomic The above rationale for adduct measurement de- DNA from liver (Zhang et al., 1991) but most epi- pends on the proposed causal link between DNA demiological studies have been performed using adducts, mutations and tumour induction. DNA aflatoxin-nucleic acid adducts excreted in the adducts induced by specific carcinogens have been urine (Autrup et al., 1987; Groopman et al., 1993) linked to mutations in a number of different types or aflatoxin-albumin adducts in serum (Wild et of study, including shuttle vectors containing aL, 1993). Nevertheless, the presence of urinary adducts of specific carcinogens (e.g. Maher et al., AFB1-N7-guanine adduct was associated with an 1989; Kat & Thilly, 1994). Mutation spectra in animal increased risk of ICC in a prospective cohort study tumours in genes involved in the carcinogenic pro- in the People's Republic of China (Qian et al., cess are generally consistent with those expected 1994). It is noteworthy that, prior to its successful from a knowledge of the types of DNA adduct application in this study, this biomarker had induced by specific chemicals (GreenЫatt, 1994). undergone extensive validation (Gioopman, 1994) Poirier and Beland (1992) reviewed data in to establish the dose-response relationships experimental animals where steady-state DNA between exposure and adduct levels in liver, urine adduct levels after 1-2 months' chronic exposure and albumin, both in animal models and in have been compared with tumour induction fol- human populations exposed to dietary aflatoxins. lowing lifetime exposure to the carcinogen, e.g.

144 Carcinogen adducls in cancer epidemiology aflatoxin B1 (AFB1) for rat and trout liver; diethyl- adducts and tumour incidence at the level of the nitrosamirie for rat liver; 2-acetylaminofluorene individual animal. One attempt to perform such (2-AAP) for mouse liver and bladder; ethylene oxide an experiment was reported by Fischer and Lutz for rat spleen mononuclear ccli leukemia; 4-methyl- (1995) who examined the induction of papillomas nitrosamino)-1-(3-рyridyl)-1-butanone (NNK) for on skin after exposure of outbred mice to 7,12- These authors rat lung tumours; and 4-aminobiphеnyl(4-ABP) for dimethyibenz[aj-anthracene (DMBA). mouse liver and bladder tumours. For some of these used the latency period to appearance of the first chemicals, a linear dose-response for both DNA papilloma as an indicator of individual risk. adducts in the target organ and tumour induction DMBA-DNA adducts, 8-OHdG adducts and cell were observed (e.g. rat and trout liver with AFB1). division were determined in the dermis of the However, in other situations, including bladder of treated skin area 2 weeks after the appearance of female mice with 2-AАF and male mice with 4-ABP, the first papilloma. The data indicated a positive adduct levels increased at all doses but tumours correlation between shorter latency period of were only induced at the highest doses (Poirier & papilloma induction and both 8-OHdG levels and Beland, 1992; Poirier et al., 1995). Iп the case of the cell proliferation. However, DMBА DN А adducts 2-AAF experiments, the tumour induction was seen were significantly higher in the animals with longer at dose levels associated with increased cell prolif- latency period. This was explained by the lower eration in the bladder (Cohen & E1lwein, 1990). rate of cell division in these animals compared There are examples of differences in response to with those with short latency, resulting iп a slower carcinogens between males and females apparently `dilution' of the DNA adduct concentration after as a result of differences in cell proliferative the end of treatment. This again indicates that response rather than DNA adduct formation. One measurement of carcinogen-DNA adducts does example is that of rats treated with 2-amino-l- not always reflect individual risk of tumour induc- HdG of particular interest methyl-6-phеnylimidazo[4,5-b]pyridinе (PhIP) tion. The data on 8-О ате (Ochiai et al., 1996). In this experiment rats were given the growing body of evidence showing that given PhIP in the diet for up to 12 weeks, and DNA secondary reactions resulting from carcinogen adducts, cell proliferation and aberrant crypt foci exposure could be of importance in addition to the were measured in the colon. PhIP-DNA adduct primary formation of DNA adducts (see below). levels were the same in male and female rats, but These above data suggest that DNA adducts are the incidence of aberrant crypt foci was threefold necessary but not sufficient to result in turnours higher in male than in female rats. This increase with these chemical carcinogens. Consequently, was associated with a 42% increase in labelling while measurements of DNA adducts in a given index in the male colon compared with the female tissue are certainly evidence of exposure of the colon after 8 weeks treatment. target organ to the carcinogen, they are unlikely to In the studies of rat lung with NNK, the deter- be sufficiently informative alone to predict tumour mination of DNA adducts in specific cell popula- incidence in that organ at the individual level. The tions within the lung was important if the dosе- important role of cell proliferation both in car- response curve for tumour induction in that organ cinogenesis and in determining adduct levels is was to be understood (Belinsky et aL, 1987, 1990). supported by the above observations. The degree This study illustrated that adduct measurements iп to which the above mechanisms interfere with the DNA extracted from total tissue are not necessarily association between exposure and DNA adduct lev- related to tumour induction, but rather that the els in human populations needs to be taken into target cell population needs to be identified. account when incorporating these markers into The studies cited above compared the DNA epidemiological studies. adduct levels in one set of animals with the tumour incidence in a parallel set of animals. The Methods for measuring carcinogen adducts former were treated with carcinogens under the Many sophisticated techniques have been developed same regimen but for a shorter duration, sufficient to measure adducts in human tissues and cells. These to allow steady-state adduct levels to be reached. generally include a clean-up step(s) followed by an There are few studies which have related DNA analytical step. Clean-up steps comprise solvent

145 Application of Biomarkers in Cancer Epidemiology

standardization is clearly an important process if Dietary affatoxins Diet Tobacco large-scale, multicentre studies are to be performed with sample analysis being conducted in different laboratories. BaP-DNA adducts Multiple adducts from individual carcinogens As mentioned above, the specificity of adduct mea- Aflatoxin-albumin Occupation Air pollution adducts surements for exposure to a particular carcinogen is one of the potential benefits of this approach. Measurement of specific adducts is based on the elu- Specific Non-specific cidation of the chemical structure of the adduct. With this information, adduct standards can be pre- Figure 2. Is the adduct specific for the exposure under study? pared, antibodies raised to the adduct, the fluores- BaP, benzo[a]pyrene. cent, UV and electrocheinical properties etc. investi- extractions, solid-phase extraction, immunoaffinity, gated, aid appropriate analytical techniques devel- liquid and gas chromatography, etc. Analytical tech- oped. It is this basis that permits the measures to be niques comprise mass spectrometry, immunoassays, specific for the agent of interest. Other methodolo- fluorescence, UV or electrochemical detection, etc. gies to measure DNA adducts, notably the zР-post- These approaches all have inherent strengths arid labe11ing technique, as originally developed, or weaknesses which have been reviewed in depth in a immunoassays analysing intact DNA, forfeit a degree number of comprehensive reviews and are not of specificity but are appropriate as an approach to repeated here (Wogan, 1988, 1992; Shuker & Farmer, measuring exposure to a defined class of carcino- 1992). In general, the above methods are of a high gens. This is valuable in situations of exposure to sensitivity, i.e. they can measure the adduct at low complex mixtures of chemicals (Perera et "L, 1990). levels in the biological sample. For example, the 32Р- To date, several hundred carcinogen—DNA postlabelhng technique has permitted detection of adducts have been identified (Hemminki et "L, adducts at the level of one adduct per mammalian 1994). In the majority of cases even a specific car- genome (Chacko & Gupta, 1988). The value of such cinogen forms adducts at different sites on the same sensitivity is illustrated by the ability to identify 4- molecule. For example, a simple environmental arninobiphenyl—Hb adducts in the blood of non- methylating agent such as dimethyl-nitrosamine smokers exposed to environmental tobacco smoke results in at least 12 different DNA adducts (Hammond et a1., 1993). The presence of carcinogen (Margison & O'Conner, 1979). hn this case the most adducts in human cells and body fIuids following abundant adduct, 7-mеthylguапiпe, does not appear environmental, occupational and therapeutic expo- to be the most relevant for induction of mutations sures has been documented and reviewed (Wogan, but rather the minor O-methylated bases, 0- 1992). methylguanine (O6-mеG) and 04-methуlthуmidiпe, Interlaboratory comparisons of methods have are the promutagenic adducts (Pegg, 1984; been relatively few and appear to have been Richardson et"L, 1987). organized mainly on an ad hoc basis between A compound may also form chemically distinct research groups. An exception to this is an interna- types of adduct, which differ in their degree of speci- tional coltaborative study organized by the US ficity for a given exposure. For example, the tobacco- Environmental Protection Agency and supported specific nitrosamine (TSNA) NNK can induce both by the European Union, in collaboration with the methylation adducts aid pyridyl-oxobutylation NCI and NCTR, USA, coordinated by the 'ARC, adducts. While the O-methylation adducts (e.g. Об France (Dr M. Castegnaro), and the CRC, UK (Dr D. meG) may be the source of mutations induced by Phillips). This collaboration aims to develop stan- this compound, the adduct is not specific to TSNfl dard protocols for the analysis of specific adducts but is also formed as a result of exposure to other by the ЗгР-postlabel1ing approach. In addition, environmental niethylating agents as well as to DNA adduct standards for different chemical car- endogenously formed methylating agents (Bartsch dnogens are being prepared. This type of method & Montesano, 1984). In this case, the pyridy-

146 Carcinogen adducts in cancer epidemiology

Absorption bioactivation Chemical exposure DNA adduct detoxification

e.g. AFBi e.g. AFBi—Ni—G \ - -

secondary DNA damage - - - - Y

e.g. apurinic site AFB1—Fгpy -~

Biological process Oxidative DNA damage

e.g. inflammation e.g. 8—oxo—G lipid peroxidation

Figure 3. Formation of DNA damage via different pathways. loxobutyl adducts would be а more specific marker carcinogen is activated to an AFBI 8,9-epoxide for exposure to TSNA. which binds predominantly, although perhaps not A highly specific measure of a given adduct is exclusively, to the N7 position of guanine (Martin therefore not equivalent to having а highly specific & Garner, 1977; Essigmann et al., 1977). However, marker for a given exposure. For example, while afla- this adduct is highly unstable and through either toxin—albumin adducts are specific for dietary expo- active DNA repair or spontaneous depurinatioin sure to aflatoxins, benzo[а]pyrene—DNA adducts can results in formation of an apurinic site in DNA. result from tobacco smoke, air pollution, diet and/or The mutation spectra observed in bacterial sys- occupation (Fig. 2). In this case, use of this latter tems, in mammalian cells hi vitro and in rodent marker as a measure of exposure to air pollutants and human tumours are consistent with a miscod- would need to be complemented by measures of the ing event resulting from this apurinic site (Trottier possible confounding exposures mentioned. et aL, 1992). The free AFB1-N7-guanine base is It is necessary to ensure that adducts do not result excreted in the urine and can be detected by from sources other than from the covalent binding immunoaffinity/HPLC as a marker of human of a chemical to a macromolecule in the body. exposure (Groopman et aI. 1993); this is a good specific examples are the difficulties of discriminat- marker of initial hepatic DNA adduct levels ing the naturally occurring ribonucleoside 7-methyl- (Groopman et iL, 1992). However, the AFB1-N7- guanosine from the DNA adduct 7 methylguanine in DNA can also undergo a chemically deoxyguanosine (Bianchini et al., 1993) and the induced imidazole-ring cleavage such that the presence of 3-mеthylаdeniпe as a natural compi- major DNA adduct in rat liver 48 h post-treatment lent of the diet (Shuker & Farmer, 1992). is the imidazole ring-open AFB1-N7-guanine, A further complication regarding specificity and known as AFB1-Papy (Hertzog et al., 1982). There- relevance of adduct measurements is that the ini- fore, the adducts measured in human liver tissue tial adduct formed in DNA may not be the critical are most probably the AFB1-Papy rather than the lesion involved in mutagenesis. This is illustrated presumed mutagenic lesion. The situation is fur- by the example of AFB1 presented in Fig. 3. This ther complicated by the fact that, at least in rat

147 Application of Biomarkers in Cancer Epidemiology

liver, treatment with AFB1 results not oniy in alternative to DNA as target molecules (Skipper & direct binding of AFB1 to DNA but also in an Tannenbaum, 1990; skipper et al., 1994). This is oxidative stress with a consequent increase in because proteins are often available in larger oxidative DNA damage (5hen etaL, 1994, 1995). In amounts than DNA and also because the lack of this type of exposure situation, the choice of the repair of protein adducts results in a greater persis- most informative biomarker would be facilitated tence (see below), and therefore the measured level by an understanding of which adduct is involved is more closely correlated with past exposure. In in the generation of mutations in critical genes for practice, in epidemiological studies the measure- carcinogenesis. The AFВ1-N7-guanine adduct and ment of protein adducts has been restricted to the AFB1 albumin adduct are the most specific albumin (Aib) and haemoglobin (lb) because of markers of exposure to AFBI, but the complemen- the ready availability of these proteins in periph- tary information from measurements of oxidative eral blood. The identification of adducts on his- damage in the same subjects would be valuable. It tones has been reported for some carcinogens is interesting in the above example that both AFB 1 (Ozba1 et al., 1994). The longer half-life of histones and oxidative stress appear to preferentially target compared with Alb and lb is one advantage of mutations to the third nucleotide of codon 249 of their use as dosimeters, as is the proximity of his- the p53 gene (Aguilar et al., 1993; Hussain et aL, tones to DNA. In addition, histones may also play 1994), a mutation in human hepatoceIlular carci- a role in gene regulation, and, consequently, the nomas that has been suggested to be induced by modification of histones by carcinogens may have aflatoxin exposure (GrеeпЫatt et al., 1994). a functional role iп carcinogenesis. The use of The need to understand the adducts involved in other long-lived proteins, including collagen, has mutageaesis is further illustrated by a study of also been discussed aid could be applicable to mutational spectra in the c-Ha-ras gene in mouse molecular epidemiological studies in the future skin papillomas induced by several aromatic (skipper et aL, 1994). hydrocarbons (Chakravarai et al., 1995). In this study the authors argued that the mutations (G to Target organs and surrogate tissues T or A to T transversions) were consistent with As mentioned above, one of the potential advan- apurinic sites resulting from depurination at gua- tages of examining DNA adducts is that they can be nines and adenines rather than being the result of measured in the target cells for carcinogenesis. stable DNA adducts, the latter being what would However, in many cases these cells are not avail- normally be measured in molecular epidemiology able and alternative sources of DNA or proteins are studies. For example, this observation was sug- used. Many studies have utilized DNA from periph- gested to be the case with benzo[a]pyrene which eral blood cells (PBC), bronchiolar lavages or oral forms a significant proportion of stable C-8 gua- cavity as well as adducts excreted in the urine. In nine adducts, which have been the basis of many only a few cases has the association between adduct measurements of carcinogen exposure in human levels in the target organ and those in the surrogate tissues. The above data at least suggest that assays tissue or body fluid been examined in humans. of apurinic sites would be a valuable addition to In animal studies, the ratio between DNA methods for monitoring DNA damage in humans. adducts in liver and PBC has been examined for a Overall, these examples illustrate the need for number of carcinogens aid these data were detailed information on adduct formation and the recently reviewed (Bianchini & Wild, 1994а, mutageuic potential of those adducts in choosing 1994b). The ratios varied from 0.1 to 100 with even an appropriate end-point for exposure assessment. the small number of carcinogens studied. In a In addition, the need to understand the different series of four methylating agents having different exposures that can yield a specific chemical саr- target organs for carcinogenesis, the ratio between cinogen adduct is indicated. liver and PBC for 7-mеG levels was relatively constant, but the ratio between the target organ (lung, Protein adducts as alternatives to DNA adducts colon, oesophagus and liver) and PBC ranged from Although proteins are not the target for muta- 3.6 to 200. These data suggest that the methylation tional events, they are nevertheless an attractive of PBC DNA is occurring in the liver by transfer of

148 Carcinogen adducts iп cancer epidemiology active methylating species from the hepatocyte to repair processes are present, and the turnover of the the PBC. macromolecule to which the chemical is bound. This In humans, aromatic DNA adducts in buccal parameter of persistence is particularly critical in mucosal cells were highly correlated with levefs of determining the appropriate epidemiological study the same adducts in oral biopsies from the same design in which to apply adduct measurements. individuals (Stone et aI., 1995). This type of corre- In the case of DNA adducts, the chemical lation might be expected in cells coming from the stability is highly variable. For example, the AFB1- same tissue, especially where cell turnover is rapid N7-G adduct is lost by depurination with a half-life and therefore adduct levels reflect mainly recent of around 8 h in rat liver (Groopman et al., 1992). exposure. There was also a good correlation be- A proportion of the initial adducts are, however, susceptible to imidazole ring-opening to form a tween 7-теG levels in bronchial tissue and PBC in one study based on only five subjects. A low but more stable AFB1-Fapy adduct which has been significant correlation of 0.34 between po1ycyclic reported to persist in rat liver DNA as long as 19 aromatic hydтocarbon (PAH)—DNA adducts in PBC weeks after the last treatment. and lung cancer tissue was reported in the study by In addition to the inherent chemical stability of Tang et a2. (1995). In the same study, the DNA the DNA adduct, there are also active repair adduct levels in PBC did not correlate with levels enzymes. For example, in human cells, the methy- in non-tumour lung tissue. Another study reported lated DNA bases N7-methу1dеoxyguaпasiпе and no correlation between the total PAl—DNA adduct N3-methуldeoхyadeпasiпе, as well as 8-oxo-guanine, levels in lung and PBC, but a significant correla- are repaired by the N-methylpurine glycosylase tion in the case of the major DNA adduct (van enzyme, although at markedly different rates Schooten et aL, 1992). Aromatic DNA adducts mea- (Male etaL, 1987; Bessho et аL, 1993). The ОЬ-теG adduct is repaired efficiently in most human tis- sured by 32Р-labelling assay were highly correlated in non-tumour lung tissue of lung cancer patients sues by the ОЬ-alkylguanine DNA alkyltransferase. and peripheral blood mononuclear cells (r = 0.74) The situation is rendered more complexby the fact (Wiencke et al., 1995); moreover, adduct levels in that DNA repair activity is not uniform throughout both tissues correlated negatively with years since the genome, with moie rapid repair of some types quitting smoking. The study suggested that aro- of DNA adduct occurring in transcribed as opposed matic DNA adducts in peripheral blood mononu- to non-transcribed genes and with а strand bias for clear cells could be an indicator of the DNA dam- the transcribed strand (Bohr, 1991; May et al., age caused by smoking in lung tissue. Phillips et al. 1993). This appears to be of importance in the (1990) carried out a similar study, looking at aro- process of carcinogenesis as there is a predomi- matic DNA adducts in non-affected lung tissue and nance of mutations in the non-transcribed strand SBC. They found that adducts in the lung corre- of the p53 tumour suppressor gene (Greenblatt et lated with smoking habits, but that the marker аI., 1994). assessed in PBC did not. In around 10 larynx can- Persistence of adducts is also dependent on the cer patients the levels of aromatic DNA adducts as stability of the molecule to which the chemical is the case of DNA this is essentially trans- determined by 32Р-postlabelling in larynx tumour bound. Iп tissue and surrounding non-tumorous tissue correlated as the rate of cell turnover. Thus adducts in lated well with adducts in PBC (Szyfter et al., 1994). epithelial cells lining the large bowel or in Thus, to date, there are some published data to neutrophils in peripheral blood (half-life = 8 h) support a correlation between DNA adducts in PBC will be far less persistent than adducts in non-pro- and those in internal organs, although the limited liferating tissues such as brain or liver or in long- number of studies should be noted and further lived T-lymphocytes. In fact, one would predict similar studies conducted where possible. that DNA adduct levels in the same individual should be higher in lymphocytes than totalleuko- Persistence of adducts cytes. This has been rarely studied, but in those Once an adduct has been formed, its persistence cases where it has been examined the levels were will depend on a number of factors, namely its higher in the longer-lived cells (Mustorien & inherent chemical stability, whether any active Hemminki, 1992). It is interesting that the differ-

149 Application of Biomarkers in Cancer Epidemiology

еncеs in adduct levels between smokers and non- Adduct stability during storage and analysis smokers in this study were observed in totalleuko- The attraction of applying measures of adducts in cytes, granulocytes and lymphocytes, suggesting prospective studies is easy to appreciate. However, that recent exposure over the past 1 or 2 days was a particular issue of concern in the context of long- correlated with past exposure of a few months. The term storage of biological samples is the stability of differences may not be so consistent in exposures adducts. This is of concern both in the initial of more intermittent pattern. phases of sample collection and processing and in One consequence of the above complexity of the storage of biological samples in freezers over DNA adduct formation and persistence is that many years. Unfortunately there are few data on half-lives of different DNA adducts need to be this topic. It is generally accepted that rapid pro- empirically determined.. A second consequence cessing of biological samples to storage at low tem- is that most measurements of DNA adducts peratures is desirable to avoid continued enzyme in human organs will be an average measure activity, which would degrade DNA and proteins of events at the sequence, gene, cell and tissue lev- or repair adducts after isolation. Postmortem sta- els. bility of DNA has been examined (Bar et al., 1988). The situation for carcinogen-protein adducts However, there is little information on the impact is somewhat simpler than for the DNA adducts. on adduct levels in, for example, peripheral blood In general it has been assumed that in the absence cells that have been isolated and frozen im- of active repair the adducted protein would have mediately after isolation or several hours later. the same half-life as the unadducted тоleсиle- Similarly, there are few data on the levels of i.e. in the case of haemoglobin, a half-life of adducts in the same samples after storage under around 120 days, and for albumin adducts different temperatures for different lengths of around 20 days. Surprisingly, this has not been time. The stability of benzo[ajpyrene diolepoxide confirmed for many adducts by longitudinal adducts in rat liver, lung and heart has been tested studies either in animals or humans. Carmella under various conditions (Izzotti et aI., 1993). and Hecht (1987) examined the half-life of a These authors reported that adducts were stable in tobacco-specific nitrosamine-Hb adduct in rats tissues up to 72 h at 4°C, but significant decreases and reported that, with a half-life of 9.1 days, this were observed at 20°C or 37°C even after periods as was significantly shorter than expected. Given the short as 24 h. Under conditions that mimicked first-order kinetics of adduct loss, this was most autopsy conditions, i.e. 16 h at 20°C or 16 h at probably the result of instability of the adduct 20°C followed by 24 h at 4°C, no significant varia- rather than a preferential clearance of the chemi- tion in adducts was observed. Similar studies for cally modified haeunog1obiri. other adducts, in different tissues and under dif- Iп humans, only a few studies have been ferent conditions of storage, would be of value. reported, for the most part relating to subjects Another issue is the artefactual formation of who stopped smoking for an agreed period of time. adducts during sample processing, where the reac- For example, in a study of 4-аminobipheпуl-Hb tion that leads to adduct formation in vivo can (4-ABP-Hb) adducts, smokers stopped smoking occur during sample storage or analysis. This is for a period of up to 80 days and the observed rate possibly best illustrated by the situation with oxi- of decline of the adducts was significantly in dized bases where oxidation reactions must be excess of that expected (Maclure et al., 1990). In avoided during sample preparation and analysis. the study by Mooney et aI. (1995) both 4-ABP-Hb This can occur during DNA preparation (Claycamp, and PAH-DNA adduct levels were monitored in 1992) or during analytical procedures, e.g. the for- volunteers for up to 14 months after they stopped mation of 8-oxo-dG and other bases during the smoking. Both markers were significantly de- silylation process prior to GC-EI1S analysis (Doukl creased 10 weeks after smoking cessation and the et aL, 1996). The appearance of hydroxyethylvaine half-lives were estimated to be 23 weeks (95% CI, in haemoglobin during storage of glоЫп under dif- 10.5-36.3) and 12 weeks (95% CI, 9.9-14.0) for ferent conditions has also been noted (Torngvist, PAH-DNA adducts and 4-ABP-Hb adducts respec- 1990). While the artefactual formation of adducts tively. is not likely to be a problem with chemical car-

150 Carcinogen adducis in cancer epidemiology cinogens, it is worth noting that it is not only lim- chemicals in the workplace can be performed by ited to oxidized bases. Scates et al. (1995) recently linking job descriptions obtained normally by illustrated the artefactual formation of DNA interview with documented usage and presence of adducts by bile acids in vitro, and the danger of the chemicals in the occupational activity от induction of W photoadducts during DNA purifi- industrial process described. This linkage can be cation has been noted (Widlak et aL, 1995). systematized in the so-called job-exposure matri- The possible artefactual formation of DNA ces OEMs) (Coggon etal., 1984). The JEM approach adducts therefore merits attention during method represented a substantial improvement on the clas- development. sical analysis of cancer risks by job title, because JEMs provide a reference category of non-exposed Epidemiological study design and adduct individuals and increase the study power by pool- measurements ing all the individuals exposed to the same agent A number of the above aspects of carcinogen—DNA from different jobs into one `exposed' group. and carcinogen—protein adducts have implications However, there are limitations in that categories of for the appropriate use of adducts in epidemiolog- exposure are only probabilistic; intensity of expo- ical studies. These are now discussed in relation to suie cannot be assessed on a common scale for the specific study designs. A comprehensive review on majority of job titles in the population; high mis- this topic has been published by Rothman et aL classification rates across exposure categories have (1995). been documented; and, finally, transferability of Application of adducts in case—control studies JEMs developed for one specific study to a new is limited when current and past exposures do not investigation is limited (Goldberg et al., 1993). correlate within individuals; this may happen as In spite of these limitations of JEMs, cross- a direct consequence of the disease, e.g. dietary sectional individual assessment of internal dose or habits, or because the exposure is not related to biologically effective dose (adducts) cannot replace personal habits and, therefore, is more likely to questionnaire information, because current expo- change over time, e.g. occupational activities. 1n sure is a bad indicator of past exposure in this the case of diet and occupation, objective markers instance. Most occupation-related cancer cases of exposure to specific chemicals would greatly occur in retirement age, and even in the active age increase the informativeness of the studies. How- groups the proportion of those who move from ever, to date, adduct measurements have not gen- one economic activity to another, changing envi- erally offered arr improvement over questionnaire ronment, work tasks and consequently exposures, data on exposure in case—control studies. In a tends to be rather high. A better application of case—control study of aflatoxin exposure and liver adduct measurements in this context is the valida- cancer the aflatoxin—albumin adduct level was cor- tion of associations between jobs and cancer risk related with measurements of aflatoxins in house- detected by case—control JEM-based studies, by hold foods in the controls but not in the cases determining average biological effective dose in (Hall & Wild, 1994). This suggests that either the workers classified as 'exposed' and `non-exposed' dietary intake of aflatoxins was affected by the dis- by the matrix. This type of transitional study can ease or the disease itself modified the metabolism substantially improve the evidence provided by of the carcinogen (Hall & Wild, 1992). In a study of a classical case—control study, and can also add lung cancer, PAH—DNA adducts in PBC of cases specific knowledge to help in understanding the were significantly higher than in controls, with an carcinogenic process. odds ratio of 7.7 (95% CI 1.7-34, P<0.01) (Tang et Adducts as markers of exposure and effective аI., 1995). In this study, adduct levels were signifi- dose have a good application in ecological studies, cantly associated with questionnaire-derived which relate exposure to a chemical experienced indices of smoking among the cases but not among by different populations to their risk of the disease the controls, suggesting a difference in biological at the group level. In this case the disease does not response to smoking between these two groups. affect the exposure level assessed in spite of the In case—control studies of occupational cancer, cross-sectional nature of the observation. Adduct retrospective assessment of exposure to specific markers have an ideal application when investi-

151 Application of Biomarkers in Cancer Epidemiology gating the effects of food-borne chemicals and air increased PAH—DNA adduct level in peripheral pollutants that cannot be measured at the individ- blood cells of subjects from the former population uaI level through interview. One example is the (Perera etal., 1992). The authors showed that after study of the relation between aflatoxin intake and adjusting for current smoking, another potential the risk of hepatoce11ular carcinoma. Many eco- source of PAH—DNA adducts, adduct levels were logical studies have been conducted on this subject, still positively associatéd with air pollution. with aflatoxin intake indirectly estimated by the Among ail observational studies, the prospec- level of contamination of food samples. The results tive study is the design in which the potential for of these studies are heterogeneous and sometimes comparison biases is the least serious: the adduct is inconsistent. In one study of this type, conducted detected in material collected when the disease was in Thailand, aflatoxin—albumin adducts were not clinically manifest and, although the presence determined in population samples (Srivatanakul of latent disease cannot be excluded, analysis by et Ri. , 1991). Another much large study in the time since exposure assessment can help in inter- Реорlе.'s Republic of China used urinary aflatoxirn preting the results. metabolites as a measure of exposure and found no The most cost-effective way to analyse cohort correlation with liver cancer rates (Campbell et al., studies, particularly when assessment of biological 1990). In this case some of the urinary alatoxin material is involved, is the nested case—control metabolites, in contrast to the AFB1-N7-G adduct, design, which consists of the analysis of the cases were later shown to be unrelated to dose and thus detected during the follow-up of the cohort, and of may have led to misclassification of exposure only an appropriate sample of those who do not (Groopman, 1994). Ecological studies of the effect develop the disease; this sample is in general a very of air pollution in the general population could small fraction of the entire cohort. The first exam- benefit from the assessment of adduct levels to a ple of a measure of a DNA adduct being associated chemical commonly contaminating the atmos- with an excess cancer risk was provided using this phere of urban areas. Many specific and non- type of study design (Ross et al., 1992; Qian et al., specific chemicals aid agents contaminate the air 1994). In this instance, AFB1-N7-G adducts in a of urban areas, and this multiplicity of exposures single urine sample collected from a cohort in may be seen as a serious limitation to the use of Shanghai, People's Republic of China, was associ- markers of specific exposures. However, it has been ated with an increased risk of hepatoce11ular carci- documented that urban air pollutants follow the noma. This type of study is needed to strengthen same circadian, weekly and seasonal variations, the rationale for using adducts as measures of out- which are mainly determined by atmospheric come in epidemiological studies. conditions and cyclic intensity of vehicle traffic. Besides potential confounders, which depend on Therefore, a few sentinel markers may be good the disease investigated, some additional match- indicators of a general non-specific exposure to ing of variables in nested case—control studies polluted urban air in addition to being markers of needs to be considered; these are the determinants internal biologically effective dose to defined car- of the accuracy of the adduct measurements previ- cinogens. It should be mentioned, however, that ously described. Samples of cases and matched ecological studies of air pollution should not rely controls should be analysed in the same batches in only on markers such as adducts, which are the order to distribute equally the effects of sample result of exposure to specific chemicals irrespective processing; in addition, controls should be selected of the external source of that exposure. Con- so that their samples have been stored as long sequently, while adducts can be incorporated into as those of their matched cases. This design descriptive studies of this type to improve under- avoids the introduction of comparison biases if standing and interpretation, it is necessary to preservation of the biological material affects the carefully consider possible independent sources stability of adducts with time (see above). Long- of exposure to the same chemicals as potential term effects of preservation can be monitored confounders. For example, a study comparing within the study bank by storing control material, PAH DNA adducts in a highly industrialized area which is assayed at the time of storage to deter- of Poland with those in rural areas showed an mine the baseline level and then re-assayed,

152 Carcinogen adducts in cancer epidemiology together with the material of cases and controls, within the time unit it is necessary to have some each time that analysis is performed. However, it is quantitative knowledge of both intra-individual difficult to ensure that assays performed years and interindividual variance (Armstrong et al., apart will be characterized by the same measure- 1992). Transitional studies to describe these aspects ment error, and therefore long-term monitoring of the marker in the population studied are there- should aim at detecting only substantial effects of fore necessary when planning a prospective inves- preservation. tigation. On the other hand, systematic trends One major advantage of prospective observa- over time are Iikely to be relevant to the carcino- tions in the assessment of biological markers is genic process, and therefore variations over long that the collection of the biological material that time periods should be studied as independent fac- will eventually be used to measure exposure can tors and not as sources of random misclassification be scheduled in order to reduce the effect of error. intra-individual variation, particularly if it is of In studies where external exposures have been a cyclical nature. Examples of this are seasonal determined at the individual level and correlated variation of grain contamination with aflatoxins with specific adducts in the same individuals, e.g. and circadian variation of exposure to air pollu- number of cigarettes smoked versus aromatic DNA tants in urban areas. These sources of random adducts in lung tissue (Phillips et al., 1988), there error can be controlled tri a prospective design In is still a high degree of interindividual variation in two ways: the amount of adduct formed for a given exposure. This interindividual variation can result, for • by imposing restrictions on the scheduled example, from differences iп carcinogen metabo- time of collection at the enrolment stage of the lism, and some studies have identified an impact cohort (more appropriate for the air pollution of these effect modifiers on the dose-response example); relationship between exposure and adducts. For • by matching nested controls to the cases for example, Vineis et aI. (1994) reported that, among similar conditions at the time of enrolment, cigarette smokers, those with a slow acetylator provided that the relevant information was phenotype have higher average 4-ABР-НЬ adduct recorded and is available. levels than rapid acetylators aid that this effect was particularly evident at low exposure levels. Intra-individual variability that cannot be Adduct assessment can therefore find a major attributed to known and controllable external fac- application in applied (transitional) studies trying tors can instead be reduced with repeated samples to elucidate proposed mechanisms of action for a from each individual. The timing and frequency of defined carcinogen. Lang et al. (1994), for example, repeated measures are often governed by financial studied the interaction between Р4501А2 and constraints rather than scientific considerations. NAT-2 polymorphism and a preference for well- However, the optimal use of available resources cooked meat, as assessed by questionnaire in a should be driven by some knowledge of the extent case-control study of colon cancer. A pathway to of the variation and its possible cyclic nature. link the intake of heterocyclic amines from cooked Intra-individual variation must be separated from meat to the endogenous formation of carcinogenic systematic time trends, e.g. hormonal levels in N-acetoxyarylamine was proposed, and indeed the women vary in a regular fashion over days and results conoborate the proposed metabolic path- months; at the same time, important trends are way. High and low intakes of heterocyclic amines documented with age. As already mentioned, were indirectly estimated by a preference for well- cyclic variations within relatively short time peri- cooked or lightly cooked meat. Increased excretion ods can be controlled by fixing constraints on the of heterocyclic amine has been shown in the urine schedule of collection if external causes of vari- of consumers of pan-fried meat, but DNA adduct ability are known. Alternatively, repeated samples levels, which should be the final products of the from the same subject can be obtained in a short proposed mechanism, in consumers and non- time period and pooled in an 'average' sample. To consumers of well-done meat have not been assess the optimal number of repeats per subject, described.

153 Application of Biomarkers in Cancer Epidemiology

The above examples illustrate the use of adducts described if adducts are to be used as outcome vari- to test the impact of polymorphisms in the metab- ables in intervention studies. olism of carcinogens on one step of the carcinogenic pathway. As a consequence of such work, the AcknowEedgements rationale for the integration of genetic polymor- C.P.W. was partially supported by a grant from the phic markers for such genes into case-control stud- NIEHS, USA, no. 2-P01 ES06052. ies can be strengthened. The fact that adducts represent an integration References of external exposure and interindividual vari- Aguilar, F., Hussain, 5.P. & Cerutti, P. (1993) Aflatoxin B., ability in carcinogen metabolism, DNA repair, induces the transversion of G- Т in codon 249 of the etc. makes them attractive because they may p53 tumor suppressor gene in human hepatocytes. Proc. provide a more relevant measure of exposure, Nat? Acad. sri. USA, 90, 8586-8590 i.e, the biologically effective dose. However, it also Armstrong, B.K., White, E. & 5aracсi, R., eds (1992) means that the relationship between exposure and Principles of Exposure Measurement in Epidemiology, adduct level is not simple. As public health deci- Monographs in Epidemiology and Biostatistics, Vol. 21, sions are normally based on ambient exposure lev- Oxford, Oxford Medical Publications els, more understanding of the relationship Autrup, H., 5eremet, T, Wakhisi, J. & Wasunna, A. (1987) between these levels and adducts needs to be Aflatoxin exposure measured by urinary excretion of developed if the latter are to be used in making aflatoxin B 1-guanine adduct and hepatitis B virus infec- such decisions. tion in areas with different liver cancer incidence in To complete this review, the possible applica- Kenya. Cancer Bes., 47, 3430-3433 tions of adducts in intervention studies to modify В r, W., Kratzer, A., Machler, M. & Schmid, W. (1988) the risk of cancer or premalignant conditions are Postmortem stability of DNA. Forensic Sci. fit., 39, 59-70 worth mentioning. In the context of the appLica- Bartsch, H. & Montesano, R. (1984) Relevance of пitro- tion of adducts, these can be divided into inter- samines to human cancer. Carcinogenesis, 5, 1381-1393 ventions that prevent exposure to genotoxic Belinsky, S.A., White, CM,, Devereux, T.R., Swenberg, agents and those which tend to modify the metab- J,А. & Anderson, M.W. (1987) Сен selective alkylation of olism of pro-carcinogens once exposure has DNA in rat lung following low dose exposure to the occurred. In these cases, adducts would provide an tobacco specific carcinogen 4-(N-methyl-N-nitтosаm пo) early marker of the intervention efficacy and could -1-(3-pyridуl)-1-butanone. Caflcer Res., 47, 1143-1148 be used in ad interim analyses for the purpose of 3e1lnsky, S.A., Foley, J.F., White, C.M., Anderson, M.W. & monitoring of compliance. An example of this Maronpot, R.R. (1990) Dose-response relationship type of study is an ongoing short-term interven- between 06-mеthylguaпine formation in Clara cells and tion to reduce aflatoxin-alburnin adducts using induction of pulmonary eeoplasia in the rat by 4- the drug of tipraz (Bolton et al., 1993). A different (methyluitгоwminо)-1-(3-pуridyl)-1-ъutaпone. Cancer type of study is illustrated by Verhageri et ai. (1995) Res., 50, 3772-3780 who attempted to modulate the urinary levels of 8- Bessho, T., Roy, R., Yamamoto, K., Kasai, H., Nishimura, oxodeoxyguanosine by increasing the dietary 5., Tano, K. & Mitra, S. (1993) Repair of 8-hydroxygua- intake of fresh vegetables. Тhi approach is an nine in DNA by mammalian N-methylpurine-DNA gly- interesting one for testing hypotheses of mecha- cosylase. Proc. Nat! Acad. ici. USA, 90, 8901-8904 nisms of carcinogenesis in humans. If adducts are Bianchhsi, E & Wild, С.P. (1994а) 7-Methyldеохуguanosine to be used as surrogate end-points for disease out- as a marker of exposure to environmental methylating come in intervention studies then considerably agents. Toxico?. Left., 72, 175-184 more understanding of the relationship between Bianchini, F. & Wild, C.P. (1994b) Comparison of 7- these events is required. medG formation in white blood cells, liver and target Tt should be also be noted that because of the organs in rats treated with methylating carcinogens. ethical implications of any intervention, it is Carrinogenesis, 15, 1137-1141 mandatory that the metabolic pathway leading to Bianchini, E, Montesano, R., Shuker, DE., Cuzick, J. & Wild, the production of adducts, together with its possi- C.I? (1993) Quantification of 7-methyldeoxyguanosroe ble confounders and effect modifiers, is carefully using unmunoaffinity purification and HPLC with elec-

154 Carcinogen adducts in cancer epidemiology trochemica1 detection. Carcinogenesis, 14, 1677-1682 dimethylbenzo[а]antliracene. Proc. Nat! Acad. Sci. USA, Bibi, V.A. (1991) Gene specific DNA repair. Carcinogenesis, 92, 5900-5904 12, 1983-1992 Goldberg, M., Kromhout, H., Guenel, P., Fletcher, А.С., Germ, M., Glass, D. ., Heederik, D., Kauppioen, . & Butor', M.G., Munoz, A., l bs , L.P., Groopman, J.D., С Т асо оп Ponti, A. (1993) Job exposure matrices in industry. lit. J. Mâxuitenko, .Y., Roebuck, B.D. & Kensler, T.W. (1993) У Epidemiol., 22, 510-15 Transient intervention with oltipraz protects against aflatoxin-induced hepatic turnorigenesis. Cancer Res., 53, Greenblatt, M.S., Bennett, W.P., Hollstein, M. & Harris, 3499-3504 C.C. (1994) Mutations in the p53 tumor suppressor gene: clues to cancer etiology and molecular pathogenesis. Campbell, T.G., Chen, I., Liu, C., Li, J. & Parpia, B. (1990) Cancer Res., 54, 4855-4878 Non-association of aflatoxin with primary liver cancer in a cross-sectional ecologic survey in the People's Groopman, J.D. (1994) Molecular dosimetry methods for Republic of China, Cancer Res., 50, 6882-6893 assessing human aflatoxin exposures. In: Eaton, D.A. & Groopman, J.D, eds, The Toxicology o f A flatoxi s, Human Carmella, S.G. & Hecht, . . (1987) Formation of hemo- п Ѕ Ѕ Health, Veterinary and Agricultiva! significance. San Diego, globin adducts upon treatment of F344 rats with the Academic Press, pp. 259-279 tobacco-specific nitrosamines 4- (methylnitrosamino)-1- (3-рyridyl)-1-butanоnе arid N'-nitrosonornicotine. Groopman, J.D., Hasler, J.A., Trudel, L.J., Piku1, A., Cancer Res., 47, 2626-2630 Donahue, P.R. & Wogan, G.N. (1992) Molecular dosimetry in rat urine of flatoxfn-N7-guanine and other aflatoxin Chacko, M. & Gupta, R.C. (1988) Evaluation of DNA а metabolites by multiple monoclonal antibody affinity damage in the oral mucosa of tobacco users and non- chromatography aid immunoa(finity/high performance users by 32 - ct assay. Carcinogenesis, 9, 2309-2313 Радби liquid chromatography. Cancer Res., 52, 267-274 Chakravarti, D., g, J.C., Cavaiieri, E.L. & Rogan, Решп Groopman, J.D., Wild, C.P., Hasler, J., Junshi, C., Wogan, E.G. (1995) Relating aromatic hydrocarbon-induced G.N. & Kensler, T.W. (1993) Molecular epidemiology of DNA adducts and c-H-ras mutations in mouse skin papil- aflatoxin exposures: validation of aflatoxin-N7-guanine bras; The role of apurinlc sites. Proc. Nat! Acad. Sci. USA, levels in urine as a biomarker in experimental rat mod- 92, 10422-10426 els and humans. Environ, Health Perspect., 99, 107-113 Claycamp, H.G. (1992) Phenol sensitization of DNA to Hall, .J. & Wild, C.P. (1992) Aflatoxin biomarkers [let- subsequent oxidative damage in 8-hydroxyguanine А ter; comment]. Lancet, 339, 1413-1414 assays. Carcinogenesis, 13, 1289-1292 Hall, A.J. & Wild, C_P. (1994) Epidemiology of aflatoxin- Coggon, D., Pannett, B., Acheson, E.D. (1984) Use of related disease. In: Eaton, D.A. & Grooprnari, J.D., eds, job-exposure matrix in an occupational analysis of blad- The Toxicology o fAflatoxins, Human Health, Veterinary and der cancers on the basis of deaths certificates. J. Nat' AgticultnrrI significance. San Diego, Academic Press, pp. Cancer Inst., 72, 61-65 233-258 Cohen, S.M. & Eliwein L.B. (1990) Proliferative and Hammond, S.K., Coghlin, J., Gann, P.H., Paul, M.,. genotoxic cellular effects in 2-acetylaminotluorene blad- Taghizadeh, K., Skipper, P.L. & Tannenbaum, S.R. (1993) der and liver carcinogenesis: biological modeling of the Relationship between environmental tobacco smoke App!, Phar !., 104, 79-93 EDO1 study. Toxico!. гегсо exposure and carcinogen-hemoglobin adduct levels in Douki, T., Delatour, T., Bianchini, F. & Cadet, J. (1996) nonsmokers. J. Natl Cancerinst., 85, 474-478 Observation and prevention of an artefactual formation Hemminki, K., Dipple, A., shuker, D.E.G., Kadulbar, F.F., of oxidized DNA bases and nucleosides tri the GC-EIMS SegerbSck, D. & Bartsch, H., eds (1994) DNA Adducts, method. Carcinogenesis, 17, 347-353 identification and Biological significance (IARC Scientific Essigmann, J.M., Croy, R.G., Nadzan, A.M., Busby, W.F., Publications No. 125). Lyon, International Agency for Jr, Reinhold, V.N., Buchi, G. & Wogan, G.N. (1977) Research on Cancer Structural identification of the major DNA adduct Hertzog, P.J., Smith, J.R. & Garner, R.C. (1982) Proc. Nat[ Acad. S L USA, formed by aflatoxin B1 in vitro. с Characterisation of the imidazole ring-opened forms of 74, 1870-1874 trans-8, 9 -dihydro-8, 9 -dihydтo-8-(7-guanyl) 9 -hydroxy Fischer, W.H. & Lutz, W.K. (1995) Correlation of indi- aflatoxin B1. Carcinogenesis, 3, 723-725 vidual papilloma latency time with DNA adducts, 8- Hussain, S.P., Aguilar, F., Amstad, P. & Cerutti, P. (1994) hydroxy-2'-deoxyguanosine, and the rate of DNA syn- Oxy-radical induced mutagenesis of hotspot codons 248 thesis iп the epidermis of mice treated with, 7, 12-

155 Application of Biomarkers in Cancer Epidemiology and 249 of the hитал р53 gene. Oncogene, 9, 2277-2281 Epidentiol. Biomarkers Prey., 4, 627-634 Izzotti, A., Bagnasco, M., Scafolini, L., Rovida, A. & De Mustonen, R. & Hemminki, K. (1992) 7-Methylguaпinе Flora, S. (1993) Post-mortem stability of benzo[alpyrene levels in DNA of smokers' and non-smokers' total diolepoxide-DNA addicts in rat organs. Carcinogenesis, white blood cells, granulocytes and lymphocytes. 14, 2185-2187 Carcinogenesis, 13, 1951-1955 Kat, A.G. & Thi11y, W.G. (1994) Mutational spectra of Ochiai, M., Watanabe, M., Kushida, H., Wakabayashi, K., endogenous genes in mammalian cells. In: Hemminki, Sugiinura, T. & Nagao, M. (1996) DNA adduct formation, K., Shuker, D.E.G., Kadlubar, F.F., Segerb~ck, D. & cell proliferation and aberrant crypt focus formation Bartsch, H., eds, DNA Adducts, Identi fгcatioп andBiological induced by PhIP in male and female rat colon with rele- signifIcance (‚ARC Scientific Publications No. 125). Lyon, vance to carcinogeriesis. Carcinogeuesis, 17, 95-98 Inte national Agency for Research on Cancer. pp. г OzbaI, C.C., Velic, L, Soohoo, 371-383 С.K., Skipper, P.L. & Tannenbaum, S.R. (1994) Conservation of histone Lang, N.P., Butler, M.A., Massengill, J., Lawson, M., carcinogen adducts during replication: implications Stotts, R.C., Hauer Jensen, M. & Kadlubar, F.F. (1994) for long-term molecular dosimetry. Cancer Res., 54, Rapid metabolic phenotypes for acetyltransferase and 5599-5601 cytochrome 4501 2 and putative exposure to food- Р А Ozturk, M. (1991) 53 mutation iii hepatocellular carci- borne heterocyclic amines increase the risk for colorectal р noma after aflatoxiri exposure. Lancet, 338, 1356-1359 cancer or polyps. Cancer Epidemiol. Biomarkers. Prey., 3, 675-682 Pegg, А.E. (1984) Methylation of the 06 position of guanine in DNA is the most likely initiating event in Maclur , M., Bryant, M.S. & Skipper, P.L. . (1990) Decline е А carcinogenesis by methylating agents. Cancer Invest., 2, of the hemoglobin adduct of 4-amйobipheпyl during 223-231 withdrawal from smoking. Cancer Res., 50, 181-184 e r , E, Jeffrey, A., Santella, R.M., Brenner, D., Mayer,]., Maher, V.M., Yang, J-L., Mah, C-M. & McCormick, J.J. Ргеа Latniano, L., Smith, S., Young, T.L., Tsai, WY., Hemniluki, (1989) Comparing the frequency and spectra of mu- K. & Brandt-Rauf, P. (1990) Macromolecular adducts and tations induced when and 5V-40 based shuttle vector related biomarkers in biomonitoring and epidemiology containing covalently bound residues of structurally- of complex exposures. I : Vainio, H., Sorsa, M. & related carcinogens replicates in human cells. Mutat. Res., п Iclichael, A.J., eds, Complex Mixtures and Cancer Risk 220, 83-92 (1ARC Scientific Publication No. 104). Lyon, International Male, R., Haukanes, B.I., Helland, D.E. & Кlерре, K. Agency for Research on Cancer, pp. 164-180 (1987) Substrate specificity of 3-methyladenine-DNA g1y- Perera, EP., Hemminki, K., Gryzbowska, E., Motykiewicz, cosylase from calf thymus. Eur. J. Biochem., 165, 13-19 G., Michalska, J., Santella, R.M., Young, T.L., Dickey, C., Margison, G.P. & O'Connor, Р.J. (1979) Nucleic acid Brandt-Rauf, P. & DeVivo, I. (1992) Molecular and modifications by N-nitriso compounds. In: Grover, P.L., genetic damage in humans from environmental poIlu- ed., Chemical Carcinogens and DNA, Vol. 1, Boca Raton, till in Poland. Nature, 360, 256-258 FL, CRC Press, pp. 111-159 Phillips, D.H., Hewer, A., Martin, С.N., Garner, R.C. & Martin, C.N. & Garner, R.C. (1977) Aflatoxin B-oxide King, M.M. (1988) Correlation of DNA adduct levels in generated by chemical or enzymic oxidation of aflatoxin human lung with cigarette smoking. Nature, 336, 790-792 I causes guanine substitution in nucleic acids. Nature, В Phillips, D.H., Schoket, B., Hewer, A., Bailey, E., Kostic, S. 267, 863-865 & Vincze, I. (1990) Influence of cigarette sXoking on the May, A., Nairn, R.S., Okumoto, D.S., Wassermann, K., levels of DNA adducts in human bronchial epithelium Stevnsner, T., Jones, J.C. & Bohr, V.A. (1993) Repaii of and white blood cells. mt. J. Cancer, 46, 569-575 individual DNA strands in the hamster dihydrofolate Poirier, .C. & Beland, F. reductase gene after treatment with ultraviolet light, alky- М А. (1992) DNA adduct measurements and mmli incidence during chronic carcino- lating agents, arid cisplatin. J. Biol. Chem., 268, 1650-1657 gen exposure in animal models: implications for DNA Mooney, L.A., Santella, R.M., Covey, L., Jeffrey, A.M., adduct-based human cancer risk assessment. Chem, Res. Bigbee, W., Randall, M.C., Cooper, T.B., Olirnan, R., Tsai, ТохгсоL, 5, 749-755 W-Y., Wazneh, L., Glassman, ., Young, T-L. & erera, АН Р Poirier, M.C., Fullerton, N.F., Smith, . F.P. (1995) Decline of DNA damage and other biomarkers ВА & Beland, RA. (1 995) DNA adduct formation aid tumonigenesis in mice in peripheral blood following smoking cessation. Cancer dining the chronic administration 0f 4-aminobiphenyl

156 Carcinogen adducts in cancer epidemiology at multiple dose levels. Carciaogenesis, 16, 2917-2921 Banaszewski, J. & Yang, K. (1994) Aromatic DNA adducts in larynx biopsies and leukocytes. Carcinogenesis, 15, G.S., Ross, R.K., Yu, .C., Yuan, J.M., Gao, Y.T., Qian, М 2195-2199 Henderson, B.E., Wogan, G.N. & Groopman, J.D. (1994) A follow-up study of urinary markers of aflatoxin expo- Tang, D., Santella, R.M., Blackwood, A.M., Young, T.-L» sure and liver cancer risk iii Shanghai, People's Republic Mayer, J., Jaretzki, A., Grantham, 5., Tsai, W.-Y. & Perеra, of China [see comments]. Cancer Еpkdermo2 Biornarkers. F.P. (1995) A case-control molecular epidemiological Prev., 3, 3-10 study of lung cancer. Cancer Ерыепsиd. Biomarkers Prey., 4, 341-346 Richardson, F.C., Beauchamp, R.O., Jr & Swеnberg, J.A. (1987) Properties and biological consequences of Tomgvist, M. (1990) Formation of reactive species that olkylpynniidirae deoxyribonhcleosides. Pharmacol. Ther., lead to hemoglobin adducts during storage of blood sam- 34, 181-213 ples. Carcinogenesis, 11, 51-54 Ross, R.K., Yuan,).M., Yu, M.C., Wogan, G.N., Qian, G.5., Trotuer, Y., Waithe, W.I. & Anderson, A. (1992) Kinds of Tu, J.Т., Groopman, J.D., Gao, Y.T. & Henderson, B.E. mutations induced by aflatoxin B1 in a shuttle vector (1992) Urinary aflatoxin biomarkers and risk of bepato- replicating in human cells transiently expressing cellular carcinoma, Lancet, 339, 943-946 cytochrome Р4501А2 cDNA. MoL Carcinog., 6, 140-147 Rotthman, N., Stewart, W.F. & Schulte, P.A. (1995) van Schооtеn, F.J., Hillebrand, M J., van Leeuwen, F.E., Incorporating biomarkers into cancer epidemiology: a van Zandwijk, N., Jansen, H.M., Den Engelse, L. & Kriek, matrix of biomarker and study design categories. Cancer E. (1992) Polycyclic aromatic hydrocarbon-DNA adducts Epidemiol. Biomarkers. Prey., 4, 301-311 in white blood cells from lung cancer patients: no correlation with adduct levels in lung. Carcinogenesis, 13, Scates, D. ., Spigelman, .D. & Verritt, . (1995) Appearance К А Ѕ 987--993 of artefacts when using 32P рostlаёеllјг1g to investigate DNA ackluct formation by bile acids in vitro: lack of evi- Verhagen, H., Poulsen, H.E., Loft, S., van-Poppel, G., dence for covalent binding. Carcinogenesis, 16, 1489-1491 Wi[1ems, M.I. & van Bladeren, P.J. (1995) Reduction of oxidative DNA-damage in humans by brussels sprouts. Shen, H.M., Shi, C.Y., Lee, H.P. & Ong, C.N. (1994) Carcinogenesis, 16, 969-970 Aflatoxin B1-induced lipid peroxidation in rat liver. Toxicol. AppL Pharmacol., 127, 145-150 Vineis, P, Bartsch, H., Caporaso, N., Harrington, A.M., Kadlubar, F.F., Landi, M.T., Malavei11e, C., Shields, P.G., Shen, 1,1., Ong, C.N., Lee, B.L. & Shi, C.Y. (1995) Skipper, P., Talaska, G. & Tannenbaum, S.R. (1994) Aflatoxin B1-induced 8-hydro yd oxyguan sine forma- х е о Genetically based N-acetyltransferase metabolic poly- tion in rat hepatic DNA. Carcinogenesis, 16, 419-422 morphism and low-level environmental exposure to car- Shuker, D.E.G. & Farmer, Р.B. (1992) Relevance of urinary cinogens. Nature, 369, 154-156 DNA adducts as markers of carcinogen exposure. Chem. Widlak, P., Bykov, V.1. & Hemmi ki, K. (1995) Formation Toxico!., 5, 450-460 п Res. of UV-photoadducts durings DNA purification. Mutat. Skipper, P.L. & Tannenbaum, S.R. (1990) Protein adducts Res., 347, 117-119 in the molecular dosimetry of chemical carcinogens. Wiencke, J. ., K lsey, KT., Varkonyl, A., Semey, K., Wain, Carcinogenesis, 11, 507-518 К е J.С., Mark, E. & Claristiani, D.С. (1995) Correlation of Skipper, P.L., Peng, X., Soohoo, C.K. & Tannenbaum, S.R. DNA adducts in Ыоод mononuclear cells with tobacco (1994) Protein adducts as biomarkers of human carcino- carcinogen-induced damage in human lung. Cancer Res., gen exposure. Drug Metab. Rev., 26, 111-124 55, 4910-4914 Srlvatanakul, P., Parkin, D.M., Jiang, Y.Z., Khlat, M., Kао- Wild, C.P., Fortuin, M., Dornato, F., Whittle, H.С., Hall, Ian, U.T., Sontipong, Ѕ . & Wild, C.P. (1991) The role of A.J., Wolf, С.R. & Montesano, R. (1993) Aflatoxin, liver infection of Opisthorchis viverrini, hepatitis B virus and enzymes, and hepatitis B virus infection in Gambian aflatoxin exposure ш the aetiology of liver cancer in children. Cancer Epidemiol. Biomarkers. Prev., 2, 555-561 Cancer, 68, 2411-2417 Thailand. Wogari, G.N. (1988) Detection of DNA damage in stud- Stone, J.G., Jones, N.J., McGregor, A.D. & Waters, R. ies on cancer etiology and prevention. In: Bartsch, H., (1995) Development of a human biomonitoring assay Hernminki, K. & O'Neill, LK., eds, Methods for Detecting using buccal mucosa: comparison of smoking-related DNA DNA Damaging Agents in Humans: Applications in Cancer addicts in mucosa versus biopsies. CnncerRes., 55,1267-1270 Epidemiology and Prevention (IARC Scientific Publications Szyft r, K., Hemminki, K., Szyfter, W., Szmeja, Z., Ni. 89). Lyon, International Agency for Research on е Cancer, pp. 32-51

157 Application 01 Biomarkers in Cancer Epidemiology

Wogan, G.N. (1992) Molecular epidemiology in cancer risk assessment and prevention: recent progress and avenues for future research. Erlüiroi. Health Perspect., 98, 1б7-178 Corresponding author Zhang, У.J., Chen, C.J., Lee, C.S., Haghighl, B., Yang, С.P. Wild Г.Y., Wang, L.W., Feitelson, M. & Santella, R. (1991) Molecular Epidemiology Unit, Research 5droo1 of A#latoxin B1-DNA adducts and hepatitis B virus antigens Medicine, Algernon Firth Building, University of Leeds, in hepatoce]lu1ar carcinoma and non-tumorous liver tis- Leeds LS2 9JТ, UK sue. Carcinogen cris, 12, 2247-2252

158 'ARC Scientific PuЫcatioпs No 14 lnternalioпal Agency for Research on Cancer, Lyon, 1997

Somatic cell mutations in cancer epidemiology

R.J. Albertïni and R.B. Hayes

Somatic cell mutations arising in vivo in reporter genes and in cancer-associated genes may now be measured in humans. Background mutation levels and mutational responses following various mutagen exposures are reviewed ïn this chapter. The detection methods are compared for similarities and differences based on the underlying biology of the systems. Currently available data on molecular mutational spectra are reviewed and the utility of such information is discussed in terms of mutagen exposure characterization and for defining the mutagenic basis of carcinogenesis. In addition to the reporter gene assays, recently developed assays for mutation in cancer-associated genes are considered. The strengths and limitations of using somatic cell mutations for cancer epidemiology and areas for future research are discussed

Somatic cell mutations occur regularly and univer- Reporter gene somatic cell mutation assays: sally in all humans. Sоmе are 'naturally occurring', genetic basis and methods arising continuously as spontaneous replication The haernoglobin (Nb) genes errors or in response to endogenous mutagens or Genetic basis (Table 1). lb is a tetrameric protein DNA metabolism, while others are induced by that constitutes >99% of the protein in non-nucle- external mutagens which are ubiquitous in the ated mature red blood cells (RBCs) (reviewed in environment. Mutations are central to human car- Stamatoyannapoulos et al., 1984). Several роlypер- cinogenesis, and, consequently, exposures that tiде chains, encoded by genes at several linked loci, cause mutations are also suspect of causing cancer. constitute the intact lb molecule, depending on the stage of development of the individual. Iп prindple, subjects who have experienced greater exposure to genotoxic agents will accumulate Following birth, adult IbA (аZ 32) constitutes >95% greater numbers of mutations, so that quantitation of the lb in mature RBCs, with a small amount of of the frequency of mutational events serves to HbA2 (a2Б2) also present. The c y (two loci) S and (3 identify heavily exposed individuals. In addition, lb genes are on chromosome lip; the t and a specific genotoxic agents may produce specific (two loci) lb genes are on chromosome 1бр types of mutations, implying that characterization (Deisseroth et al., 1977, 1978). of mutational errors (mutational spectrum) will be Three classes of mutations affect the lb genes. useful for identifying exposure to specific agents. Two, the thalassaemias and hereditary persistence Modern in-vivo somatic mutation studies may of fetal Hb (HPFH), produce abnormal levels of lb employ any of seven assays for measuring in-vivo polypeptides. These are not useful for mutational mutations in five different reporter genes, i.e. not studies however, as non-genetic influences can functionally related to carclnogenesis. There are reduce lb levels in RBCs (Stamatoyannapoulos also methods for detecting mutations in cancer et ai., 1984). The third class of lb mutations pro- genes (oncogenes and tumour suppressor genes). duces structural alterations in lb. Over 500 such Extensive databases are available for two of the changes have been described, most being single reporter gene assays (GPA and hprt), while limited amino acid changes in the а or the p chain. Of information is available for all others. Here we these, the best known is an A to T transversion in describe the genetic bases and methods for the the gene that gives 'sickle cell' haemoglobin, or major assays, summarize the results of applications lbs. Although the gene is autosorna1 and contains in humans, and discuss the uses and limitations of three exons spanning 2 kb, the effective the methods for future epidemiological studies. mutational target size for single base changes is a

159 Application of Biomarkers in Cancer Epidemiology

. в в - в t' - в ~• - в '

Cells Gene(s) Chromosome Gene size Target size

RBC lb gene Autosome, lip Э exons, 2 kb bp (small) lb gene Autosome, 16р Э exons, 2 kb bp (small) RBC GPA Autosome, 4q 7 exoпs, 44 kb >44 kb (large) T-lymphocytes HPRT X-linked, Xq 9 exils, 44 kb >44 kb (large)

T-lymphocytes HLA-A Autosome, бp 7 exils, 5 kb >5 kb (large) Т lymphocytes TCR Autosome 1 4q Large 'multi -gene Probably large TIR Autosome, 7q Large multi gene' Probably large single base pair. The assay is thus hnhted to the at approximately 5 x 105 molecules/cell (Furthmayer, detection of point mutations, which limits its prac- 1977; Gahmberg et cL, 1979). The GPA gene on tical sensitivity and restricts to а narrow range the chromosome 4q spans 44 kb and contains seven spectrum of mutational mechanisms detected. exons (Kudo & Fukuda, 1989). In principle, it is a large target for mutation. It has two codiminantly lb variant frequency (VF) assays (Table 2). G.P. expressed (M and N) alleles (Furthmayer, 1978) Stamatoyaranopoulos and co-workers originally (except for rare nulls) which have approximately developed an assay based on highly specific antibod- equal frequencies in all populations (Cartron et al., ies that discriminate the different mutaпf haemoglo- 1990). Thus, approximately 50% of humans are bins from each other and from normal haemoglobin, I/N heterozygotes with both forms of the glyco- requiring samples of 1 ml blood (reviewed in protein on h IC surfaces. The Maid the N molecules Stamatoyannopoulos et al., 1984). Scoring tech- differ by two non-adjacent amino acids and can- niques for rare variant cells with a mutated lb are not be interconverted by simple point mutations. based on fluorescence labelling of fixed RBCs on slides. As haemoglobin is an intracellular protein, fix- GPA VF assays (Table 2). The currently used GPA ation and permeabilization are required for antibody assay measures the frequency of variant cells that entrance. Recent refinements have included newer have lost expression of the M form in blood samples methods for permeabilizing and Iabelling RBCs in from heterozygous individuals (M/N) (Langlois et suspension for automated cell sorting and counting аl., 1990; Jensen & Bigbee, 1996). A small (<1 ml) (Bigbee et al., 1981). More recently, automated sample of blood is treated soon after blood draw to microscopy has been used to screen large numbers of fix spherical erythrocytes which are then kept at cells on slides (Tates et al., 1989). One or more struc- 4°С until analysis. Variant cells are detected by tural lb mutations may be scored in an assay. flow cytonietry with distinguishable fluorescent- Although there have been no systematic des- labelled monoclonal antibodies specific for the N criptions of aI1 technical sources of error in this assay, and M forms and counting variant cells that bind quality of RBCs and extent of permeabни atiоn could the anti-N but not the anti-M antibody. Simple loss be major sources, as could the quality of the iinmuno- mutations, many representing 'point mutations', logical reagents and ion-specific staining. Poisson are expressed as rare 0/N cells (Grant & Bigbee, counting of rare events produces relatively large 1993). These are referred to as'hemizygous variants'. errors. More complex mutations are expressed as rare N/N cells on the I/N background, i.e. a loss of func- The GPA gene tion of one allele with double expression of the Genetic basis (Table 1) Glycophortn-A (GPA) is a other. These are thought to arise from mutational polymorphic glycoprotein on RBC surfaces present events that lead to cellular homozygosity (or loss or

160 Somatic cell mutations in cancer epidemiology

Gene and cell Bample size Method(s) Time for assay Potential for error lb in RBC 1 ml Immunological staining Month RBC quality —Manual slides Days Permeabilization —Automated slides Fours–days Immunological reagents --Cytometry Non-specific staining . Poisson counting

GPA in RBC 1 ml Immunological staining Hours–days RBC quality —Cytometry Immunological reagents Non-specific staining Poisson counting NPAT in T-cells 10-50 ml Short-term Cycling cells give phenocopies Autoradiography Observer error Manual slides Days–week Cell culture factors Automated slides (potential) BrdU incorporation with differential fluorescence Non-specific staining changes Manual slides Days–week Poisson counting Automated slides Days Instrumentation for Cytometry (potential) (Short) ` automaled assays Poisson counting Cloning Weeks Inversé corrélations Selection in tissue culture between CE and MF Observer error Reagents and culture conditions Poisson counting

1-iLA in T-cells ° 10-50 ml Cloning Weeks Non-specific lack of I mmunoselection (cytotoxicity) imrnunocytotoxicity in vitro with cell culture Observer error outgrowth Reagents and culture conditions Immunological reagents and complement Poisson counting

5-10 ml Immunological staining Non-specific failure to label Cytometry immunological reagents Instrumentation Poisson counting

7CR in T-oeils Few ml Immunological staining Non-specific failure to label Cytometry . immunological reagents instrumentation

161 Application of Biomarkers in Cancer Epidemiology heterozygosity), such as chromosome missegrega- dependent cytotoxicity of the purine analogue TG tion, somatic recombination or gene conversion (Albertini et aI., 1990). Both assays require similar (Grant & Bigbee, 1993). These variants are termed collection and fractionation protocols. Although 'homozygous variants'. The rare 0/N and N/N cells the assays can be performed on 5-10 ml of blood, that have lost expression of the M allele fall within it is preferable to obtain 30 ml or more. prescribed areas of the cytogram. The variant frequency (VF) is defined as: The VF assay (Table 2). The first assay to be devel- number of 0/N or N/N variant erythrocytes oped was a short-term phenotypic assay which has the advantage of speed and potential for automa- VF total number of red blood cells analysed tion but the disadvantage of consuming the TGI The quality and age of the ABCs in the blood T-cells which are then not available for study sample can affect results, necessitating the rapid (Strauss & Albertini, 1977, 1979; Staтk et aL, 1984). fixation of collected samples. Antibody lot, quality As there is no way to unequivocally demonstrate and preparation are potential sources of technical the mutational basis of the TG` cells, they are variability, as are changes in cytometer perfor- termed variants and their frequencies in blood are mance. Poisson counting of small numbers pro- termed VFs. For assay, cryopreserved MNCs are duces relatively large errors. A gold standard has thawed and stimulated with PHA in replicate recently been developed for the GPA assay {Jensen short-term cultures, with or without TG, and incu- & Bigbee, 1996). Cryopreservation of formalin- bated until culture termination at 24-30 h (cryo- fixed spherical RBCs in medium and DMSO, and preservation is required to avoid labelling of nor- storage. at —80°C or —150°C maintain samples with maI T-cells that are in 'in cycle' in vivo; AIbertiпi et stable VFs for up to б months. Reference standards аI., 1981). At termination, fixed cells are added in are available for interlaboratory quality control measured volumes to microscope slides, and then and standardization. stained, autoradiographed and scored. Recent protocol modifications use BrdU staining and scoring The NPRTgene by differential fluorescence (Ostrosky-Wegman Genetic basis (Table 1). The HPRT gene encodes the et al., 1988), approaches suitable for automation HPRT enzyme which is constitutively expressed using either cell cytometry or image analysers. but dispensable in virtually all mammalian cells. There are several technical sources of variability HPRT phosphoribosylates its normal substrates in the short-term autoradiographie assay. Only rare hypoxanthine and guanine for conversion to labelled cells are counted, so Poisson errors inherent inosinic acid (Stout & Caskey, 1985) and is re- in counting small numbers can be large. The slide- quired to phosphoribosylate purine analogues based method is laborious and susceptible to ob- such as б-thioguanine (TG) to their cytotoxic server error. Automated assays will have concerns forms (Albertini, 1985а ; Stout & Caskey, 1985). with instrumentation. T-cells that are cycling in Cells with normal IPRT activity are susceptible to vivo may become labelled and be scored as mutants, the cytotoxicity of TG and related agents; mutants unless measures such as cryopreservation are em- are resistant, thus providing a basis for selection. ployed to remove this effect (Albertini et al., 1981). The BPRT gene is X-linked and expressed as a single copy in all cells. The gene spans 44 kb and The MF assay (Table 2). The second HPRT assay де- contains nine exons (Patel et al., 1984). A total of pends on direct in vitro cloning in TG and, 55 kb of DNA including and surrounding this gene although laborious, allows for mutant isolation, has been sequenced, making the region one of the in-vitro propagation and molecular analyses best characterized for mutational studies. In prin- (Albertini et aI., 1982; Morley et al., 1985; ciple, НРјТ is a large target for mutation. Henderson et aI., 1986; O'Neill, et al., 1987). Fresh or thawed, cryopreserved MNCs are washed and 'PAT VF and mutant frequency (MF) assays (Table plated in culture medium for direct cloning in the 2). There are two assays for assessing in vivo ITIPRT presence or absence of TG. (Cryopreservation is mutations in human T-lymphocytes. Both are not required here to remove 'phenocopies' as it is based on the resistance of mutant cells to HPRT- in the autoradiographie assay, but it is convenient.)

162 Somatic cell mutatгons in cancer epidemiology

A mutant frequency (MF) (and its confidence inter- is separated. As for HPRT, this assay involves inoc- val) is calculated from the ratio of the cloning effi- ulation of MNCs into the wells of microtitre plates ciency of T-cells in the presence of TG to the in limiting dilutions. Selection is due to cytotoxic- cloning efficiency in its absence. There are large ity by a relevant antibody in the presence of com- methodological differences between the two kinds plement. Following selection, cells aie directly of HPRT assays, which may give somewhat differ- inoculated into microtitre plates essentially as in ent results. the HPRT cloning assay. Colonies are scored at There are several sources of technical variability 16-20 days by inverted phase microscopy. in the cloning assay (Robinson et aL, 1993). All One major source of technical variability irn the investigators have reported a strong inverse corre- HLA assay is lack of killing by specific antibody. lation between control or non-selected cloning However, these 'phenocopies' can be recognized by efficiencies and calculated mutant frequencies; testing growing colonies for resistance to the specific observer error with failure to recognize slow-grow- selecting antibody. The other sources of technical ing mutant colonies may result in variability. variability are inherent in cell culturing. The sources Attention to reagents and conditions is critical. of immunological reagents and complement and the Fortunately, the ability to cryopreserve MNCs for conditions of immunoselection can also produce subsequent testing provides a protocol standard to technical variation. It should be possible to produce control for intralaboratory drift. a cryopreserved cell standard as described for In PRТ . The 'LA gene VF assay (Table 2). A short-term flow cytometry Genetic basis (Table 1). The several linked HLA loci assay has also been described as measuring HLA include two classes of genes that encode cell sur- loss mutations in T-cells (Kushiro et al., 1992). The face recognition or restriction molecules of impor- method uses specific biotin-labelled anti-HLA anti- tance for antigen presentation in immune responses bodies to label HLA gene products on T-cells, FITC- (Bodmer, 1984; Janeway & Travers, 1994). These conjugated monoclonal anti-CD3 antibody to label genes constitute the major histocompatibility all T cells, and two-colour cytometric analysis to score complex (МНС) in humans. Although the loci are CD3+ T-cells lacking the target HLA antigens. extremely polymorphic, some alleles are present in The rare cells that have lost the target HLA anti- a high proportion of individuals, i.e. approxi- gen, i.e. less than 1/25 x that 0f normal HLA+ cells, mately 50% of the population is heterozygous for fall within a prescribed area of the cytogram. The either the НLА A2 or the НLA-A3 allele. variant frequency is defined as: The HLA complex is on chromosome bp. In practice, mutation studies have been confined to the number of CD3f target HLA antigen-lacking HLA-A gene for primary detection, although mol- VF = - lymphocytes ecular studies have defined loss of other linked genes. total of CD3+ lymphocytes The HLA-A gene is autosomal, spans 5 kb and contains seven exons; it is a large target for mutation. Sources of error in the assay are non-specific failure to label cells, immunological reagents, FILA mutant frequency and variant frequency assays instrumentation and Poisson sampling. When (Table 2). There are also two assays for assessingНLA used only in the cytometry mode, the assay may be mutations in human T-cells. The constitutional HLA subject to serious pheraocopy errors. However, cell genetic background of an individual must be known sorting can and has been employed to clone and to measure somatic mutations at this locus, i.e. an propagate single variant cells in vitro for molecular individual must normally express one of the test HLA analyses. In principle, the genetic basis of all vari- antigens in heterozygous form. The assays measure ants detected in this assay could be verified by mol- cellular loss of one codominantly expressed antigen. ecular analyses.

Cloning (MF) assay (ТаЬе 2). In the cloning assay, The TCR gene Uanatipour et a2., 1988; McCarron et a1., 1989), Genetic basis (Table 1). There are at least four T-cell peripheral blood is obtained and the INC fraction receptor (TCR) genes, i.e. the a, the R, the yard the

163 Application of Biomarkers in Cancer Epidemiology

S genes, located on chromosomes 14q (u and, within normal T-cells. Variant cells are CD4+ cells that it, 8), 7q ((i) arid 7p (y). Germ-line TCR genetic seg- have lost expression of CD3. VFs are calculated as ments consist of vахiаЫе (V) joining (J), diversity the number of CD3- cells, i.e. ceI1s with CD3 (D) (j and y TCR genes only) and constant (C) expression level less than 1/25 x that of normal regions Qaneway & Travers, 1994). Rearrange- CD4 cells, divided by the total number of CD4+ ments occur during the differentiation of T-cells T-cells. and are mediated by a recomЫnasе system, termed Sources of technical variability for the TCR gene V(D)J recombinase, through which the different V, mutation assay are variabilities in immunological D and J regions of any TCR germ-line gene are reagents aid in flow cytometer performance. joined in all possible combinations. This confers the Furthermore, as detection of the cell surface mol- TCR gene uniqueness found in mature T-cells. The ecules are the scored phenotype, the condition of rearranged genes characterize and identify a specif- the cells, surface perturbations, etc. could be potent- ically reactive T-cell and its clonai descendants. ial sources of technical variability. Large numbers The complete TCR on the T-cell surface consists of variant cells are scored in this assay, and so of a constant molecule, termed CD3, and the TCR errors inherent in counting small numbers should heterodimeric molecule consisting of either an u not be a problem. As it is not possible to clone and fi or a y and l polypeptide chain, encoded by most of the TCR gene variants in vitro for further the respective TCR genes (Clevers et ai., 1988; analyses, the mutational basis of the measured Kyoizumi et aL, 1990). A given T-cell expresses phenotype is usually not demonstrable. either the u/a heterodimeric TCR or the у/Б heterodimeric TCR—never both. More than 90% of Reporter gene somatic mutations the peripheral blood T-cells express the u/ TCR. General population studies The TCR gene mutational assay currently in use There is an extensive literature on in-vivo somatic focuses only on. CD4+/TCRu/jЭT-ce11s. mutations in humans, with extensive databases for If one of the molecules of the TCR heterodimer GPA and HPRT mutations (reviewed in Albertini et is defective for any reason, the СD3TCRu/RТСR a1., 1990; Cole & Skopek, 1994). This section pre- complex fails to form on the T-cell surface, and the sents representative results for in-vivo variant and CD3 molecule accumulates in the cytoplasm. mutant frequencies and molecular mutational Allelic exclusion operates for the TCR genes, ren- spectra. dering any cell functionally hemizygous for gene expression. Therefore, mutational 1oss of gene lb variant frequency. The lb mutation assay has function is not masked by a second allele. been used in demonstration studies of very few TCR variant T-cells are recognized by the subjects, showing in the original report a mean absence of the CD3 molecule from the surface of a background VF of 11.0 x 10-g (range 4-30 x 10- ) for CD4+ T-cell. The presumed mutation arises in one the single base changes producing the gene HbS of the TCR genes (u or 1). (As alleLic exclusion is and HbC mutations and, in a more recent report not operative for the СD3 gene, mutational loss of on five subjects, a lower mean background VF function would require somatic mutation of both value of 3.7 x 10-g (stamatoyannopoulos & Nute, alleles at this locus.) 1981; Bernini et aL, 1990). Intra-individual VF val- The TCR genes are large genetic segments when ues have remained stable over relatively long time in the germline configuration. However, somatic intervals, i.e. at least months, and smoking report- mutations may arise iп rearranged genes, which edly elevates the haemoglobin V1 values approxi- are much smaller. It is not possible at this time to mately twofold. define a target size for the TCR gene mutations. GPA variant frequency. Hundreds to thousands of The TCR VF assay (Table 2). The TCR gene muta- individuals have been studied for in-vivo GPA mu- tion assay is also based on analysis by flow cytome- tations using different versions of the assay ter (Kyoizumi et al., 1992). Commercially available (reviewed in Cole & Skоpek, 1994). 1n general, anti-СD3 is phycoerythrin-labelled and anti-CD4 mean hemizygous arid homozygous VF values for is FITC-labelled, allowing for double labelling of adults have each averaged 10 x 10-, with the for-

164 Somatic cell mutations in cancer epidemiology mer somewhat lower and the latter somewhat HLA variant frequency. HLA VFs determined by cyto- higher. There is wide interindividual variability iп metric assay are at least five times higher than the VF. Values f от newborns and children have been clonally determined HLA MFs reported earlier, i.e. 1ow.er than in adults, indicating a clear age effect, 1.5 x 10- and 0.7 x 10 for HLА A2 and HLA-А24 with about a twofold increase by age 70. smoking loss variants, respectively (Kushiro et al., 1992). (It is shows a weak association with increased VF, but noteworthy that the anti-HLA-A2 antibody was the systematic differences are not found by gender or same for the dining and cytometric assays.) The ethnicity. reasons for this and for the difference between VFs determined for the two HLA-A alleles are unknown. HPRT variant frequency. For the short-term tests, most f-WRT T-cell VFs have been determined using TCR variant frequency. There are relatively few pub- the autoradiograpliic version of the assay. Mean lished reports of background TCR gene VFs. As this values in normal populations range from 10-6 to is a phenotypic assay, the CD3 loss T-cells are 10-5 (reviewed in Albertini et al., 1990; Cole & referred to as variants. The background VFs are skopek, 1994). VFs aie increased in adults com- remarkably high compared with other reporter pared with newborns, and in adults they generally genes (except HLA loss determined by cytometry), show an increase in frequency with age. i.e. 2.5 x 10-4(Kyoizumi etaL, 1990, 1992). The rea- Interindividual variability ranges from 10-fold to sons for this are unknown, although speculation 30-fold and smoking has generally been associated has centred on the role of this genetic region in with an increase in variants. Repeat sampling from immune diversification, which may require hyper- individual donors has shown differences of mutability (HLA also is involved in immune diver- twofold to fourfold in most instances, although sification, but only shows high frequencies of differences as great as ninefold have been reported. variants in the cytometric assay). Mean TCR VFs increase with age and are reported to be one- to HPRT mutant frequency. The Iargest in-vivo somatic fourfold higher in males. No published reports of mutation database is for НРRТ in T-cells, as deter- studies on newborns or young children, or of mined by cloning assay (reviewed in Albertini et smoking effects are available. al., 1990; Cole & Skipek, 1994). As these mutations are being confirmed by molecular studies, Studies in the inherited genetic instability frequencies of TG` cells are termed IF. A recent sta- Marked increases of several VFs and 1v1Fs have been tistical analysis of four large data sets (Sussex, UK; observed in patients who are homozygous for Vermont, USA; Paris France; and Leiden, the certain rare genetic instability syndromes due to Netherlands), including 72 newborns, 70 children defects in DNA repair (reviewed in Albertini et al., and 418 adults, showed ranges of IF values of 1990; Cole & skopek, 1994). For example, ataxia 0.1-14.7 x 10 for newborns, 0.5-39.5 x 10 for telangiectasia (AT) homozygotes have shown children, and 0.8-81.7 x 10- for adults, with a con- increases in GPA hemizygous (0/N) and homozy- sistent trend from birth to old age (Robinson et al., gous (N/N) VFs, HPRT MFs and TCR VFs, although 1994). smoking was clearly associated with no effect was seen for lb VFs. Marked excesses in increased MFs in the Sussex and Leiden data sets somatic cell mutations have generally also been only, although a recent analysis in Vermont also found in Bloom's syndrome and Fanconi's anaemia showed a smoking effect. Neither sex nor ethnicity (FA) patients. However, a recent study did not detect has been found to influence HPRT MFs. increases in HPRT MFs but did find GPA VF increases in FA (sala-Trepat etaL, 1993). (An earlier 'LA mutant frequency. As ‚LA mutations determined study did show HPRT IF increases in FA; by cloning can be confirmed by molecular studies, Vijayalaxmi et al., 1985). For xeroderma pigmen- frequencies of antigen loss cells are also termed tosum (XP), increases were found for HPRТ T-lym- mutant frequencies. Mean HLA MF values are phocyte MFs, but not for GPA or lb VFs. As the 2-3 к 10 -5. 'LA MFs increase with age, but the effects cause of genetic instability in Х1' is hypersensitiv- of tobacco use have not been investigated (reviewed ity to UV, this finding supports the occurrence of in Albertini et al., 1990; Kyoizumi et al., 1992). somatic mutations in T-cells as they circulate

165 Application of 8iomarkers in Cancer Epidemiology

throughout the body—in this case in skin, where response are both consistent with the multipotent they are subject to UV irradiation. By contrast, the bone marrow stem cells being the mutational tar- RBC precursor cells, in which GPA and lb muta- gets at the time of the bomb blasts. As this stem tioпs must occur, do not receive similar UV expo- cell pool contains a limited number of cells, high- sures. Therefore, somatic mutation appears linked dose irradiation reaching it leaves few survivors to a single body compartment for the RBC precur- and wide fluctuations of induced mutants among sors, but not for the T-cells. individuals. Although the numbers of variants among individuals will have an enormous range, Studies of environmental exposures to genotoxic the mean VFs in the population at the different agents (reviewed in Cole & 5kд pеk, 1994) radiation exposure levels will reflect the dose. lb variant frequency. Increased HbS and HbC VFs Other studies, in Goiana, Brazil, and among vic- were found among subjects exposed to X-rays, and tims of the Chernobyl accident, have reported sim- H ,de1 VFs were increased after accidental expo- ilar results (Straume et al., 1991; Jensen et ai., ЬLе sure to 1i7Cs (stamatoyannopoulos & Nute, 1981; 199$). Berniui et al., 1990). Another study found HbS VFs By contrast to the atomic bomb results, patients of 8, 18, 27 and 43 x 10-d in four individuals fol- receiving local irradiation to solid tumours lowing their exposure to ethylene oxide, compared (Hodgkin's disease, prostate) have shown no ele- with a background range in unexposed individuals vations of GPA VFs (Mendelsohn, 1990; Grant & of 0-8 x 10 (Tates etal., 1989). However, the three Bigbee, 1994). The geometry of radiation expo- exposed individuals with the most elevated values sures, as well as intensity and duration, appear to were also smokers. There have been too few stud- be important in determining the GPA mutational ies of lb mutations following mutagen exposures response. This result is consistent with the lack of to evaluate its performance. VF elevations in XP patients noted above, in that both point to the bone marrow as the sole site of GPA variant frequency (cytometric assay). By con- in vivo mutations. However, the results of a study trast to the lb system, there is a wealth of data on in Japanese patients who had received the emitter GPA mutations following human mutagen expo- Thorotrast (12Тh) in the 193оs and 1940s for radi- sures. ographic visualization are somewhat at variance with this (Umeld et ai., 1991; Kyoizumi et a]., Radiation. Among the earliest population studies 1992). ARhough this agent accumulates in body with the GPA system were those of atomic bomb tissues, there were no significant elevations of GPA survivors. Significant GPA mutation inductions VFs in 10 patients who, as a group, did show ele- with radiation exposures were reported in two vations of TCR gene T cell VFs (see below). It is pos- studies, with the latter showing 63.0 x 10-/Gy, sible that the bone marrow received significantly 32.0 x 10-6/Gу and 0.14 x Ы-61Gу for 0/N, 0/M lower doses than did other tissues. and М/М variants, respectively (Lang1ois et al., 1987; Kyoizumi et aI., 1989). (These were per- Chemotherapy-related exposures. Earlier studies of a formed with an earlier version of the GPA assay heterogeneous group of 30 cancer patients treated that could also measure loss of the N form of GPA.) with a variety of cytotoxic agents and a more Similar but somewhat lower values for hemizygous recent study of breast cancer patients receiving the variant increases have been reported more recently combination CAF (cyclophosphamide, adriamycin, (Langlois et al., 1993). All results are population 5-fluorouracil) showed consistent elevations of 0/N means; in the absence of pre-exposure reference VFs, while patients receiving CMF (methotrexate values, individual VFs could not be used to deter- substituted for adrlamycin) showed lower and mine individual radiation exposures, as greatly ele- more variable elevations (Bigbee et al., 1990). The vated VFs were widely variable in high-dose sub- VF elevations in chemotherapy patients treated jects. These values were obtained more than 40 with these S-phase specific agents were transient years after the mutagen exposure, indicating that with values returning to normal in several months, the GPA marker can be Iong lived. This long mem- consistent with the half-life of RBCs (Bigbee et al., ory and the wide interindividual variabilities in 1990; Grant & Bigbee, 1994).

166 Somatic cell mutations jn cancer epidemiology

In children, chemotherapy post-treatment ele- bone marrow stem cells, which may be the types of vations were significant. By contrast with the adult relevant pathogenic events in benzene-induced studies, however, when plotted against time post- leukemias (Rothman et аI., 1995). therapy, VFs for both the 0/N and N/N variants A recent study of reinforced plastics workers in have remained significantly elevated for more than Finland exposed to styrene revealed significantly 10 years (Hewitt & Mott, 1992; Mott et al., 1994). elevated GPA N/N VFs among the most heavily Radiation alone did not elevate GPA VF in the exposed workers, particularly women (Bigbee et al., childhood studies. 1996). Lastly, a study of platinum therapy in adult patients with germ cell tumours reported signifi- HPRT variant frequency (short term assays): radiation. cant elevations of both 0/N and N/N VFs at differ- Elevated HPRT VFs have been reported for individ- ent time points throughout therapy, with some sig- uals who received heavy accidental exposures to nificant increases persisting up to 6 months post- garmria irradiation from a Б6Со source in Mexico, therapy (Perera et aI., 1992a; Grant & Bigbee, 1994) for individuals in Kiev, Ukraine, at the time of the (HPRT T-cell mutations by cloning assay were Chernobyl accident, and for subjects exposed to reported as only marginally elevated in this gamma irradiation from the Сs source in Goiana, study—see below). Brazil (Tates et aL, 1989; Ostrosky-Wegman et al., Unlike localized radiotherapy, therefore, chemo- 1990). therapy clearly induces GPA mutations, which An early study reported radiation-induced ele- most probably occur in the differentiated rather vations of HPRT VFs among 12 cancer patients than the multipotent RBC precursors depending who received 1.8-2.0 Gy/day localized radiother- on the chemotherapeutic agent in question. The apy for totals of 20-60 Gy, with the range of VF life spans of the resultant mutants are therefore values in the treated individuals being higher than some multiple of an RB C's life span, producing even the highest background values in control memories of months to years, depending on when smokers. VFs returned towards normal, from $ to in differentiation most mutations occui. In chil- 32 weeks post-treatment (Ammenheuser et al., dren there may be a tendency for mutations to 1991). occur in early progenitor cells, with some arising in the multipotent stem cells. The GPA memory for Chemotherapy-related exposures. The earliest auto- chemically induced mutations appears to be longer radiographic studies of in vivo HPRT mutations in in children than in adults. humans were in cancer patients receiving cytotoxic chemotherapies (Albertini, 1985a, 1985b). Other exposures. Measurements of GPA variants, However, the most detailed time-series study of IPRT mutant T cells and DNA adducts were made induced I-IPRT VF elevations involved multiple in iron foundry workers exposed to po1ycycflt aro- sclerosis patients receiving iv. holus infusions matic hydrocarbons (PAHs), and all results were of cyclophosphamide (750 mg/mг per month) related to ambient PAl exposure levels determined (Ammenheuser et al., 1988). Non-smoking and by personal and area monitoring (ferera et al., smoking patients were reported to have mean VFs 1993, 1994). 0/N hemizygous VFs increased before treatments of 1.52 x 10-6 and 6.56 x 10-6, slightly, but not significantly, with ambient expo- respectively. Two weeks after the first treatment, sures, while N/N VFs were unrelated to the expo- mean VFs were 29.07 x 10 6 (п = 4). All treated indi- sines. Neither set of VFs correlated with PAl-DNA viduals showed elevated VFs, the lowest being adducts (the HPRT mutations in this study are dis- 11.61 x Lo-6. At 4 weeks after the first treatment, cussed below). the mean value fell to 5.58 x 10-6 (n = 4), aid A study of 24 workers in China heavily exposed declined further to 3.50 x 10-6 (n = 5) at 7-13 weeks to beizene and 23 matched controls showed a sig- after the last treatment. Although clearly indicat- nificant increase in mean N/N VF (13.9 x 10 6 ver- ing a mutagenic effect, these rapid rises and falls in sus background of 7.4 x 10-6) but no change in 0/N VFs suggest that cytotoxicity, cell division and its VFs in benzene-exposed workers, suggesting that abatement may have had some effects at these benzene induces gene-duplication mutations in high acute doses.

167 Application of 82omarkers гп Cancer Epidemiology

Other exposures. An early study of nurses who dis- response curve (Hakoda et aL, 1988; Akiyama et aI., pense chemotherapy showed no VF elevations 1990; Hirai et al., 1995). This was initially inter- over background (Albertini et al., 1988). However, preted to indicate that HART is relatively insensi- all safety practices had been observed. By contrast, tive to ionizing radiation. It is now realized, how- another early study (using a version of the autora- ever, that the mutational signal has decayed in the diograpahic assay that did not take steps to elimi- 40+ years between the exposure and testing, as a nate phenocopies) did show a statistically signifi- consequence of mutations arising in peripheral cant (five- to sixfold) increase in VFs in cyclophos- T-lymphocytes without renewal from the bone phamide workers (Hitter et aL, 1990). marrow stem cell compartment in adults. In survivors Several recent occupational studies using the who were young at the time of the exposures, the autoradiographic HPRT assay have shown HPRT mutations also arose in precursor bone тах increases in VFs at exposure levels considered to be row stem cells—even in rnultipotent stem cells. in the acceptable range. Three have involved expo- This is probably the reason for any elevation at all sures to butadiene. Thirteen non-smoking workers in mean HPRT MF in the exposed populations. in a monomer production plant showed mean Indeed, a molecular study in one survivor showed HPRT VFs of 1,0 z 10-6 for the six non-exposed con- that at least one mutational event arose in a stem trols, 1.2 x 10-6 for the five low-level exposed work- cell before differentiation into the B-, NK and T cell ers (0.03 ppm butadiene) and 4.0 x 10-6 for the Lineages had occurred (Hakoda et al., 1989). eight high-level exposed workers (3.5 ppm butadi- Several groups have found increased HPRT ene). The mean VF of the highly exposed group mutational responses to radiotherapy in cancer was significantly elevated compared with the other patients. From an early study of 12 breast cancer groups. Furthermore, urinary levels of 1,2-dihy- patients receiving 2 Gy/day local irradiation sam- droку-4- (N-acetylcysteinyl)butane, a marker of in- pled shortly after treatment, it was estimated that vivo butadiene dose, were significantly elevated in T-cells circulating through the radiation field the highly exposed group and correlated signil- would have received 4 Gy irradiation, giving a radi- cantlywithlPRTVFs (Ward etal., 1994). А second ation-induced increase of 6.9 x 10-6 niutants/Gy study of the same worker population and an ongo- (Messing & Bradley, 1985). Others have found ing study of workers in styrene-butadiene produc- excesses for patients with breast cancer and tion tend to confirm these results (Ward et al., Hodgkin's disease, although some investigators 1996). Thus, these studies suggest that butadiene report only weak associations (Nicklas et al., 1990, exposures at this level have genotoxic effects. 1991; Sala-Trepat et al., 1990; Branda et al., 1991). (These results using the autoradiographic assay are There have also been measurements of HPRT remarkable because two studies with the cloning T-cell mutations in patients receiving extremely assay (in China and the Czech RериЫic; see below) low irradiations (10-15 mGy) during nuclear med- have failed to show increases in HPRT MFs in pop- icine scans. Initial studies reported significantly ulations with comparable butadiene exposures.) increased mean MFs following these low exposure The BrdU staining method of short-term assay procedures (Seifert et al., 1987), while later investi- has been used for only a single study of chemical gations did not find such an effect (Bachand et al., exposure, i.e. to arsenic in Mexico (Ostrosky- 1991; Kelsey etal., 1991; Van Dam etal., 1991). It Wegman et al., 1991). Although the HPRT VFs in a seems fair to conclude that the extremely low irra- high-level exposure group were reportedly twice diations encountered in nuclear medicine scans the levels in a low-level exposure group (mean are not measurable as mutagenic to HPRT and that 5.0 vs mean 2.4 x 10-6), this difference was not sta- the differences among studies were due to techni- tistically significant. cal variables. In contrast to patients, two early studies of radi- HPRT mutant frequency (cloning assay): radiation. ation therapy and nuclear medicine technicians Cloning assays have been used to study HPRT mu- showed them to have elevated MFs relative to hos- tations in Japanese atomic bomb survivors. All pital controls based on doses received (2 mSv) in studies have shown low-level MF elevations the previous 6 months (Messing etal., 1986, 1989). in exposed individuals, with a 'shallow' dose- This same investigative group has recently studied

168 Somatic cell mutations in cancer epidemiology workers exposed to low-level chronic irradiation than in the treated AIL patients (1.7 x 10-6) and in -6) ( irota et iL, 1993). (mean 1.34 гi v) tri a Quebec factory (Seifert et iL, healthy controls (1.1 x 10 Н 1993). Although no group differences were found Fifteen of the ALL patients had IFs >10 к 10-6 and between exposed and non-exposed workers, radia- elevations persisted for years. A potentially signifi- tion doses received 48--68 weeks before sampling cant difference between the HPRT mutagenic were significantly positively correlated with IFs. responses induced by chemotherapy in adults and The authors calculated a rate for induction of in children is the usually transient nature of the mutants of 0.7-3.4 x 10~/sv, a rate which is simi- former and the persistence of the latter. This may lar to, but somewhat less than, that determined indicate that, in children, a relatively larger pro- earlier for the radiation therapy and nuclear med- portion of the mutations are induced in bone mar- icine technicians. row stem cells. Of note, however, is а study of 36 workers at the Sellafield nuclear reprocessing installation, 18 of Other exposures. Oncology workers, nurses and whom had cumulative recorded radiation doses of pharmacists have all shown elevated HPRT IFs in <50 mSv and 18 had cumulative doses of 500 mSv different studies, although cloning efficiencies and accumulated over many years, but which was lуmphoсÿte subpopulations have also occasionally clearly negative for HPRT mutations (Cole et iL, been affected by the exposures, which may have 1995). axtefactually elevated values (Chrysostomou et iL, 1984; Dubeau et al., 1994). Factory workers exposed Chemotherapy-related exposures. Many groups to nor-nitrogen mustard clearly showed increased have used the cloning assay to measure HPRT IFs HPRTIFs compared with controls (Cole & Skopek, in adult patients receiving chemotherapy 1994). (Dempsey et al., 1985; Palmer et iL, 1988; Sala- Two, groups of workers chronically exposed to Trepat et al., 1990; Branda et ai., 1991; Caggana et ethylene oxide, i.e. nine hospital workers and 15 al., 1991). Increased IFs were found among a factory workers (40 h time-weighted average expo- group of patients with solid tumours and lym- sures of 0.25 ppm for the hospital workers and phomas, among breast cancer patients, other 5 ppm for the factory workers) showed mean IFs malignancies, and among patients treated with for the hospital workers of 12.4 x 10 6 and for the low doses of cyclophosphamide for connective tis- factory workers of 13.8 x 10-6 (Tates et iL, 1991). sue disease. The latter elevation was significant compared with A recent study examined HPRT IFs in 15 can- controls. The sensitivities of the various biomarkers cer patients (10 with testicular cancer) and found employed in this study for detecting ethylene that cyclophosphamide and ifosfamide were the oxide exposure were haemoglobin adducts > sister most mutagenic agents, while adriamycin, 4-epi- chiomatid exchanges > chromosome aberrations> adriamycin and bleomycin produced equivocal micronuclei > HPRT mutations. By contrast, acute responses (Tates et ai., 1994a). Cisplatin and the high-level exposures to ethylene oxide in seven etoposide VP16 (a topoisomerase inhibitor) did not workers had no effect on HPRT, 5СЕ or any other increase HPRT mutations. These results are in biomarker evaluated, leading the authors to con- accord with another study of multiple biomarker clude that transient exposures to this agent, even responses in germ cell tumour patients treated at high doses, produce no genotoxic consequences with platinum-based chemotherapeutic regimens, (Tates et al., 1995). In styrene workers, two HPRT which showed marginal elevation in HPRT IFs studies have been negative and one suggestive of (but increased GPA VF5) (Perera et iL, 1992а). an effect in five subjects (Cole et aL, 1989; Tates et The mutagenicity of chemotherapy in children al., 1994b; Vodicka et al., 1995). In iron foundry has been evaluated by the HPRT doling assay. A factory workers exposed to PAHs at low levels, i.e. study of 45 children with acute lymphoblastic from <50 to 200 ng/m, HPRT mutations did cor- leukaemia (ALL), 13 children with acute myeloge- relate with borderline significance with estimated nous leukaemia (AIL) aid 28 age-matched РАН exposure, but much more significantly with healthy controls showed a significantly higher РАН adduct levels (Perera et al., 1993, 1994). (By mean IF (7.8 x 10-6) in the treated ALL patients contrast, GPA mutations in this study were not sig-

-169 Application of Biomarkers in Cancer Epidemiology

ntiicantly elevated and did not correlate with DNA administered dose were calculated as 0.30 x 10-d adducts.) In a study of bus maintenance workers, TCR variantslGBq. Similarly, TCR VFs in six of 10 both the HPRT MFs and the РАН adduct levels were patients who received z3zTh for radiographic visu- highest in the most heavily exposed workers, with alizations were elevated compared with concurrent a highly significant increase in individual MFs controls (Umeki eta?., 1991; Kyoizumi et cL, 1992). Z Z observed with increasing adduct levels (Hou et al., 3 Тh remains in the body and produces 1995b). Two recent studies that used the cloning constant irradiation. By contrast, GPA VFs also assay to measure HPRT MFs in butadiene-exposed measured in these patients showed no significant workers have been negative (Hayes et al., 1996; elevations. Tates et al., 1996). These butadiene studies contrast with those mentioned above which used the Mutation spectra in reporter genes autoradiographrc assay. The T-сеll cloning assays allow for molecular analy- In summary, ionizing radiation and chemicals ses of mutations. By far the largest database is for at exposure levels found in accidental, household HPRT, where mutation events ranging from single and occupational settings do induce HPRT muta- base changes to deletions, translocations and tional responses. The autoradiographic form of the recombinations have been identified, and muta- assay may detect exposures with greater sensitivity tional spectra under a variety of circumstances have than the cloning assay, but this must be demon been described. 'LA molecular analyses have been strated for the same populations receiving the limited to Southern blots which have revealed a same exposures for any valid conclusion. high frequency of events such as somatic recombination that involve the homologous chromosome. HLA mutant frequency. The cloning assay for HLA In contrast to autosomal genes, the X-chromosomal mutations has not been used, thus far, to study HPRT gene cannot undergo homologous recornbi- humans exposed to mutagens. nation. However, the mutation spectra defined to date have been for HPRT mutations. HLA variant frequency (cytometric assay). The cytometric assay for 'LA mutations was used to study HPRT mutational spectra 69 atomic bomb survivors, chosen as low-dose Thousands of `spontaneous' IPRT mutations aris- exposed (DS 86 doses = 0 Gy; 1986 estimates) or ing lu vivo in human T-cells have now been high-dose exposed (DS 86 doses >1 Gy) (Kushiro et analysed at the molecular level. There have been cL, 1992). There were no significant increases in several recent reviews of molecular studies, and a VFs with radiation. It was concluded from this computerized database of published results is avail- study that 'LA mutations cannot be detected in able (Albertini et cl., 1990; Cariello et aL, 1992; T-lymphocytes after the 40-50 years' lag between Cariello & Skopek, 1993; Cole & Skоpek, 1994; exposure and assay. Cariello, 1994). The adult background HPRT mutational spec- TCR variant frequency (cytometric assay). This assay trum differs from that in the fetus and in young has been used only to study radiation exposures. children. In adults, <15% of mutations arising in No significant dose effects were found in a large vivo show gross structural alterations such as dele- study of atomic bomb survivors, indicating again tions, insertions or other rearrangements on that mutational signals in T-cells have decayed in Southern blots (Nicklas et al., 1989) (these the 40+ years between exposure and test (Kyoizumi 5outhem blot alterations involve >300 base pairs). et cL, 1992). A single individual exposed to 3-1 Gy The remaining 85% of adult HPRT mutations have in Chernobyl was found to have an elevated VF of been classified as `point mutations' and include 21.1 x 10-d 3.5 years later. base substitutions, frame-shifts, smaller deletions Eighteen thyroid cancer patients treated with and insertions, соmрlек alterations and uncharac- 1311 from 2 months to 5 years earlier showed a sig- terized splice site changes. By contrast, the back- nificant linear relationship between the amount of ground HPRT mutations in placental cord blood radioactivity administered and the TCR gene VFs show 75-85% to have gross structural alterations (Kyoizumi eta?., 1992). The induced mutations per and the remainder to be `point mutations' (Finette

170 Somatic cell mutations in cancer epidemiology et aL, 1996), although maternal lifestyle factors ated with particular mutagens, e.g. butadiene and exposures may alter this pattern (Manchester et exposures in mice (Cochrane & 5koрek, 1994). at, 1995). Approximately 35% of fetal mutations Therefore, some degree of mutational specificity show a single kind of DNA gross structural alter- from different mutagen exposures has been dis- ation, as discussed below. This reversal of the adult covered at HART. patters persists in children until approximately the age of five. HLA mutational spectra. (Table 7) Thus far, the HLA The impetus for characterizing background mutations in human T cells have been analysed mutation spectra in reporter genes is for compari- only by Southern blots (Turner et al., 1988; Morley son with mutational spectra following exposures et al., 1990; Grist et at, 1992). However, the inves- to specific environmental mutagens. The expecta- tigations have defined loss of the specific target tion is that specific znutagепs or classes of muta- HLA-A gene in the mutants and the presence or gens will induce characteristic mutational changes. absence of the HLA-B and other linked genes. Once identified, these changes can then be used in Approximately 65% of background HLA mutants subsequent population studies to define the nature show no change on 5outhern blots, 2-8% show of exposures. In this sense, it is hoped that reporter simple deletions, and 34% show changes compat- gene mutations may serve as restricted biomarkers ible with mitotic recombination. Few show gene of exposure by providing specificity for the offend- conversion. The 'no change' and mitotic recombi- ing mutagen. nation classes increase significantly with age. The discovery of induced in-vivo mutational In contrast to background, 75% of the HLA spectra is just beginning. Ionizing radiation pro- mutants induced by ionizing radiation in vitro duces an HPRT mutational spectrum that becomes show changes compatible with deletion (Kushiro et increasingly dominated by large structural alter- al., 1992). Therefore, ionizing radiation is also ations (such as deletions) as radiation doses characterized by deletion mutations at HLA. increase, i.e. ionizing radiation, at least low energy transfer (LET) ionizing radiation, produces dele- Cancer-relevant mutational changes `captured' in tions at HPRT (Nicklas etal., 1990, 1991). Studies of reporter genes (Table 3) chemical mutagen exposures in humans have thus The rationale for defining in vivo reporter gene far given mixed results, probably because of msuf- mutation spectra was originally to provide `speci- ficient numbers of mutants analysed. An early ficity' for identifying exposures. However, when report indicated that in-vivo exposures to ethylene used as biomarkers of effect, reporter gene muta- oxide induced a G to A transition at G197 (Cariello tions must also reflect events with pathogenic sig- et al., 1992). In-vitro controlled mutagenicity nificance occurring elsewhere in the genome. To experiments are beginning to show characteristic be useful as surrogates for cancer genes, reporter mutational spectra, e.g. the pesticide malathion genes must undergo mutagenic processes that have may be associated with characteristic deletions in carcinogenic potential. exon 3 (P1uth et aI., 1996). Animal studies also Both the GPA and the HLA systems reflect show characteristic HPRT mutation spectra associ- mitotic recombinations among the mutants. This

GPA HLA HPRT Mitotic recombination: Mitotic recombination Large deletions with topoisomerase Il breakpoints Gene conversion Gene conversion Fusion gens Chromosome reduplication Chromosome reduplication V(D) J recombinase-mediated recombinations.

171 Application of Biomarkers in Cancer Epidemiology mutational mechanism is known to underlie loss 5omatic mutations in cancer genes of heterozygosity (LOI) in several tumour sup- Several assays have been developed to detect gene pressor genes, and, therefore, is an important mutations in cancer genes or their products. mutational step in carcinogenesis. Agents that cause such changes in reporter genes, e.g. benzene- Cancer-associated 'mutation'assays (Box 1) induced `homozygous' GPA mutations as noted Phenotypic `mutation' assays. Malignant transfor- above, may be human carcinogens (Rothman et al., mations are associated with mutations of onco- 1995). Monitoring for such specific mutational genes and tumour suppressor genes. In the case of changes may be important for predicting individ- the former, mutations result iп overexpression of a ual cancer outcomes. normal protein (or expression of an aberrant pro- Although the HPRT gene cannot undergo tein) essential for cell proliferation. Many inca- homologous somatic recombination, it does cap- proteins and tumoui suppressor gene proteins are ture a variety of other carcinogenic mutagenic detectable by immunological techniques, e.g. mechanisms. Large deletions and translocations, nnmunoblot, ELISA, in body fluids such as serum common in human tumours, are frequent changes from cancer patients (Brandt-Rauf, 1991, 1992). in HPRT mutations (Nicklas et al., 1989), particu- Studies of banked sera have revealed that, in many larly those following ionizing radiation (Nicklas et patients, the increased oncoprotein levels were а1.,1990, 1991). The breakpoints of deletions often present months to years before diagnosis. For envi- occur in DNA sequences with high homology to ronmentally related malignancies, it has been pos- topoisomerase II consensus cleavage sequences tulated that the causative carcinogen mutated the where similar breakpoint sites are seen in the relevant gene(s) before the onset of clinical cancer. leukemias (Rainville et al., 1995). HPRT mutations Measurements of these proteins might therefore be also may produce fusion genes, another change used as biomarkers of effect in epidemiological frequently observed in cancer. studies to detect cancer-relevant somatic muta- One specific mutational change in HPRT is par- tions. Although many oncoproteins and tumour ticularly striking because it mimics so weI1 an event suppressor proteins have been studied, attention seen tri virtually all lymphoid malignancies (Finger has focused on the ras p21 protein, the extracelIu- et aI., 1986; Boehm & Rabbitts, 1989; Tycko & Sklar, lar domain (ECD) of the c-еrbВ-2 (HER-2, feu) p185 1990; Breit et al., 1993). This is the specific intra- protein, the R-transforming growth factor ((3-TGF) genic deletion that occurs in vivo during fetal life protein and the p53 protein. and early childhood, which is the most frequent sin- Genotypic `mutation' assays. Two polymerase chain gle class of background HPRT mutations during this reaction (PCR)-based molecular assays have been period of life (Fuscoe et a1., 1991; Manchester et al., 1995; Finette et aL, 1996). These mutations show all of the characteristics of the ц(D)1-mediated recombination that characterizes the TCR gene rearrangements, and they are virtually identical, at the в sequence level, to the known cancer-related V(D)J Phenotypic assays recombinase-mediated mutations in lymphoid malignancies. This mutagenic mechanism with car- • Oncogene protein in serum cinogenic potential is precisely captured in HPRT. ras р21 As specific mutational mechanisms of carcino- ECD of aerb 8-2 pi В gerdc significance are recognized, it becomes feasi- (i-TG F ble to develop PCR and other molecular tech- • Tumour suppressor proteins in serum niques for their rapid identification. Monitoring Р53 for these precise events, rather than reporter gene lenotypic assays mutations in general, may be the relevant measures in reporter genes when they are used as bio- Hybrid TIR genes markers of effect in cancer epidemiology. ВСL-2 rearrangements

172 Bomalic call mutations in cancer ерVdemiology introduced to detect 'mutations' associated with several human studies. Depending on the methods used and the products studied, these molecules сапсeт. Strictly speaking, both measure forms of chromosome aberrations mediated by aberrant may be found at low levels in healthy control indi- V(D)J recombinase activity, as described above. viduals. Therefore, different studies have used These two assays are described here because they different definitions of `positive'. Furthermore, measure directly either a cancer-relevant mutation although some studies measured truly mutant pro- mechanism or a known cancer-related chromo- teins, most have simply measured increased levels some translocation. of the marker protein, using methods that сoulд The first of these assays detects hybrid TCR not distinguish between mutant and normal mol- genes, i.e. VR -Jy or Vy JR, formed by inversions of ecules. The assumption is that elevated levels of chromosome 7 (iпv7[р13.; g321) (Lipkowitz et al., even normal proteins are probably due to somatic 1992). This inversion joins the ТСRу gene on chro- mutations and correlate with cancer. p21 protein in healthy mosome 7р13-15 to the ТСАR gene an 7g32-35. An early study of the ras (This molecular assay may measure events similar Finnish foundry workers exposed to PAHs in the to the cytometric TCR gene somatic mutation workplace showed that one of the eight exposed assay described above, but direct comparisons have and none of 10 unexposed individuals had de- not been made.) By chosing primers for PCR from tectable serum levels (Brandt-Rauf, 1992). 1n another appropriate 3 and yTCR gene regions, products are study, three of 16 hazardous waste workers ех- formed only when the inversion has occurred. The posed to a wide variety of mutagens showed assay is performed on peripheral blood lympho- detectable serum ras p21 proteins, as did two of 17 cytes, presumably the T-cells. unexposed workers (Brandt-Rauf, 1992). Both of The second molecular method (BCL-2 translo- the latter, however, were heavy smokers and were cation assay) measures an aberrant V(D)J recombi- only 'trace' positive. One of the exposed workers nase-mediated event of direct relevance to cancer developed a premalignant colon lésion 1.5 years (Liu et al., 1994). The chromosome translocation after testing. Removal of this lesion normalized the t(14:18) (g32:q21), which occurs at high frequency serum ras p21 protein level. in non-Hodgkin's lymphomas, brings together the Serum ras p21 proteins were determined in a Polish population study where a doubling in the B-cell leukеmiа!lymphопtа-2 (BCL-2) locus on oncogene expression (>2SD chromosome 18 aлд the 1g heavy chain joining U) frequency of high ras region on chromosome 14, dysregulating the for- control serum levels) was found in individuals exposed to environmental pollutants (Perera et al., mеr and resulting in delayed programmed cell death. Again, by choosing appropriate primers for 1992b). A more recent study measured the ras р21 PCR, products are formed only when the translo- serum proteins in butadiexie-exposed workers and cation has occurred. The assay is performed оп found ria elevations (Anderson et aL, 1996). normal peripheral blood lymphocytes, presumably A study in vinyl chloride (VC) workers moni- the B-cells. tored the specific mutant Asp13c-Ki-rяs p21 protein (DeVivo et aL, 1994). Four of five exposed workers Cancer-associated gene mutations in humans with liver angiosarcoma and eight of nine with The phenotypic assays. Immunological detection of liver auguras had detectable mutant proteins (by oncogene and tumour suppressor gene products in immunoblotting) in their sera. Importantly, 22 of sera was initially carried outil cancer patients to 45(49%) of the VC-exposed workers with no evi- define the percentage of positives for the diffeтent dence of liver neoplasia also showed detectable lev- cancers for the different products. As testing els. A significant linear trend was found for mutant p21 protein in serum and increasing duration of mоvед to serum banks, it became possible to assess sera retrospectively to determine if significant ele- VC exposure. None of 28 non-exposed individuals vations of the serum markers could be found had detectable serum mutant p21 protein. before the onset of cancer. Thus, the utility of these Other oncoproteins have also been studied in assays was originaly for early diagnosis. mutagenlcarcinogen-exposed individuals. Forty- Oncogene and tumour suppressor gene proteins six pneumoconiosis patients (32 asbestos, 10 sili- have now Ьееn used as biomarkers of effect in cosis) were studied for a variety of oncoproteins

173 Application of Biomarkers in Cancer Epidemiology

(Brandt-Rauf et al., 1992). Five of 18 with cancer mal values. Importantly, individuals heavily ех- had elevated serum ras p21 proteins (total) posed to pesticides have had inversion frequencies defined as a fivefold elevation over normal by dilu- intё rmediatе between normal control and AT tion, compared with only two of the 28 without patients. cancer. There were many pre-diagnosis positive Standardization of the PCR method for detecting serum values in this study. Pneurnoconiosis BCL-2 translocation frequencies is just beginning patients also had significant elevations of platelet- (Liu et al., 1994). Quantification is achieved by a derived growth factor (PDGF) serum proteins. multiple tube method based on Poisson distribu- Another study of multiple serum oncoproteins tions. Individuals with no detectable transloca- showed that the (3-TGF proteins were elevated in 14 tions (dl-6) are considered as `negative.' Twenty- of 33 fire-fighters exposed to a variety of pulmonary four of 53 blood samples from normal individuals mutagens/carcinogens (Ford et aL, 1992). No eleva- were negative for this translocation. 0f the positions were found in unexposed controls. tives, translocation frequencies varied from 0.8 to studies are being reported of serum levels of the 32,0 x 10-6, a 40-fold difference between the lowest extrace11ular domain (ECD) of the c-еrbB-2 protein and the highest (values d0-6 can be detected when p185 and the epidermal growth factor receptor frequencies are based on several analyses). When (EGF) protein (Brandt-Rauf et al., 1994; Partanen et stratified into age groupings of 0--20 years, 21-40 яl., 1994). These are elevated in certain premalig- years, 41-60 years and >60 years, these frequencies nant conditions, in early cancers, and in some were 0.29, 0.77, 1.43 and 3.39 x 10-', respectively. exposure situations with a high risk of cancer. Variability also increased with age. (It should be Studies of p53 serum protein in early cancers are noted that, although termed `translocation fre- ongoing. AI are being pursued with the goal of quencies', many of the translocations from specific developing biomarkers of effect for human muta- individuals represent large clones, some of which genicity monitoring. have persisted for long intervals. Therefore, in The history of the of cancer gene biomarkers is these cases the measured frequencies greatly over- that the original study populations were individu- estimate the translocation events.) als with cancers. Studies then progressed to individuals with early cancers then to individuals with Issues in the application of somatic mutations in premalignant conditions, and finally to individu- epidemiological studies ais exposed to mutagenslcarcinogerrs with a high strengths and limitations of the methods risk of cancer. This contrasts with the history of the At present, seven assays are available for assessing reporter gene mutation assays. In this case, devel- mutations of five reporter genes using two cell opment progressed from studies of healthy indi- types. Mutations scored in RBCs occur in nucleated viduals exposed to mutagenic agents and were precursor cells, thus limiting the in-vivo site of designed to detect exposures. Only recently has mutation to the bone marrow. Moreover, the RBC attention focused on cancer-relevant mutagenic assays, i.e. lb and GPA, do not allow for molecular mechanisms. Thus, development of these two analyses of the mutations or the development of kinds of assays have progressed from opposite direct molecular detection methods. The advan- directions towards each other, i.e. reporter genes tages of the RBC assays is that they require very moving from exposure to disease, with the cancer small blood samples and are rapid, inexpensive gene mutations moving from disease to exposures. and simple. Furthermore, the GPA assay, although phenotypic, allows for assessment of mitotic The genotypic assays. The PCR-based method for recombination, an important mechanism in LOH. detecting the inv7 (р13 X32) chromosome aberra- Finally, when mutations arise in the multipotent tion has been used to study normal individuals, bone marrow stem cells, they are potentially long- patients with AT, and individuals exposed to pesti- lived and may be used for remote dose reconstruc- cides (Lipkowitz et a1., 1992). Normal background tions or nested case—control studies. (However, frequencies are 10, as determined by limiting di u- chemically induced mutations in RBC precursors tions. AT homozygotes have a 100-fold increase in appear to arise, in part, in more differentiated cells frequency, while heterozygous carriers have nor- with much shorter persistence, i.e. onlymonths.)

174 Somatic cell mutations in cancer epidemiology

The greatest strength of the lymphocyte assays and the interindividual differences in somatic is that they provide nucleated cells for molecular mutations. Since these genotypes are associated analyses. Mutation spectra can be determined and with different responses for the biomarkers of direct molecular assay methods developed. Further- exposure, i.e. DNA adducts, there is every reason more, as mutations occur in peripheral T-cells, to expect that somatic mutations will also be mutations can arise in all body sites. А limitation increased in susceptible individuals. A suggested of T-cells, however, is that tissue culture methods association between increased HPRT MF in indi- must be used for molecular analyses to be possible. viduals of GST11 null genotype has recently been These are usually slow, costly and labour-intensive. reported (Hou eta?., 1995a). Eventually, epideniio- In addition, relatively large blood samples must be logical studies will have to take all of these into obtained. The short-term lymphocyte assays cir- account in assessing somatic mutations. cumvent this, using inexpensive and rapid meth- Physiological, nutritional and pathological ods, but do not allow for molecular analyses. The changes can also influence interindividual and even mutational memory in the peripheral T-cells is intra-individual replicate sample variations in probably a matter of months, at least in adults. somatic mutations. For example, recent studies Therefore, these mutations will not be of value for have shown that HPRT mutations are inversely asso- remote past exposures. ciated with serum folate levels (Branda et al., 1991). The time of appearance of mutants in peripheral Viral infections may increase the mutability of some cells has not been precisely defined for all of the somatic cells (Havre et al., 1995). Autoimmune dis- mutation assays. Mutants appear as early as eases raise in-vivo T-cell IPKT MFs, probably as a 2 weeks after mutagen exposures, as shown for the consequence of increased cell proliferation НPВТ autoradiographic assay (Ammenheuser et a?., (Theocharis eta?., 1995). Haematopoietic stress may 1988). Usually, however, the time for optimal also affect RBC assays. Further studies will also have occurrence will be a period of months, Additional to relate VFs and MFs to these factors. longitudinal studies are needed to establish the This variability obviously affects the ability to time of optimum mutant appearance for various detect differences in VFs or MFs between groups in exposures. In any case, mutant cells will usually epidemiological studies. A systematic analysis of appear long after the metabolites, SCEs and chu- required sample sizes, given various coefficients mosomal aberrations when these biomarkers are of variation, has been accomplished for the HPRT used with somatic mutations in epidemiological cloning assay (Robinson et al., 1994). Similar studies. The use of multiple biomarkers, however, is analyses must be undertaken for all of the systems. usually advantageous in epidemiological studies as Superimposed on the biological variability is the some, i.e. DNA adducts, are the best estimators of technical variability. Cryo- or other standards must in-vivo mutagen doses. Several studies have shown be developed and validated over time for all of the correlations of HPRT mutations to these biomarkers assays used in epidemiological studies. of exposure but not to ambient exposures. Somatrc mыations as biomarkers of exposure Assay variability Exposure/dose assessment. Somatic mutations in All of the currently used reporter gene assays show reporter genes are used to assess mutagen/carcinogen large interindividual variation. Much of this is bio- exposures. A major determinant of their utility in logical, and epidemiological studies will help to this regard is their sensitivity relative to other bio- establish the precise causes. It has been shown that logical end-points. Although mutations do detect the rare genetic instability syndromes are associ- exposures, it has been shown that, compared to ated with large increases in VFs and MFs. It is not other measures, e.g. metabolites or protein or DNA known however, if lesser deficiencies of DNA adducts, they are among the least sensitive for repair will also be reflected in detectable increases chemical mutagens/carcinogens (Tates et al., 1991). in mutations. This will require methods to assess This might have been expected as the metabolites more accurately the repair capabilities themselves. and adducts are more proximal to the exposures. An important emerging area of research is the cor- However, even for acute ionizing irradiation, the relation between the various metabolic genotypes most sensitive of the somatic mutation assays is no

175 Application of Biomarkors in Cancer Epidemiology

better and probably less sensitive than are chro- of reporter gene mutations for human bioinoni- mosome aberrations. The reason is that the latter, toting may eventually be as surrogates for cancer although clearly reflecting a genotoxic effect, is a mutations. For this to become a reality, it must very large target, i.e. the entire genome. be demonstrated that mutations in reporter Although lacking in sensitivity, there aie occa- genes, measured in tissues of convenience, are sions when reporter gene mutations may be used valid surrogates for mutations in cancer genes as exposure dosimeters. First, those assays that are occurring in target tissues. simple and require only small blood samples may Current evidence that reporter gene mutations be easier and less expensive to use than other are valid surrogates in this regard is indirect but biomarkers. Moreover, mutations may be used for positive. First, there is the analogy to another bio- exposure assessment for unknown mutagens/ marker of effect, namely non-specific chromosome carcinogens when no other biological end-points aberrations. Two recent retrospective follow-up are available. Another use for certain somatic studies have shown that individuals with high mutations for exposure/dose assessment may be frequencies of non-specific chromosome changes for dose reconstructions of remote exposures, as in have relative risks of developing cancer in the next nested case-control studies. However, only muta- decade of greater than 2.0 (Hagmar et al., 1994; tion assays that detect events in stem cells, i.e. the Bonassi et aI., 1995). Regression analyses indicated RBC assays, will be useful in this regard. Finally, that the aberrations conferred risk beyond that somatic mutations in the fetus, i.e. as HPRT in T- associated with exposure per se (Bonassi et al., ce11s, may be useful for population exposure assess- 1995). It is noteworthy that SCE frequencies, ments because the newly induced mutations arise which often detect exposures more sensitively, on a low and characteristic background. were riot associated with increases in cancer risk. Animal studies have shown that agents that Exposure characterization. Somatic mutations as produce cancers in various tissues also produce biomarkers of exposure may have their greatest HPRT T-cell mutations in vivo (idoo et al., 1991). utility in terms of exposure characterizations. Thus, HPRT is a functional surrogate for cancer in Mutation spectra are being defined with the expec- these species. Finally, the administration of radio- tation that the naturally occurring background protective agents in mice receiving ionizing radia- spectrum will differ from those рrодисед by dif- tion reduced both the induced malignancies and ferent mutagens or classes of mutagens. If so, char- HPRT T-cell mutations, again relating the reporter acterizing the mutations will allow a diagnosis of a events to disease-causing events (Grdina et al., specific exposure, i.e. will provide specificity. This 1991, 1992). will be possible only for those mutations that are The best evidence that reporter mutations recovered for molecular analyses—currently re- reflect the occurrence of cancer gene mutations in quiring the T-cell assays and laborious tissue cul- humans is the discovery that mutagenic mecha- ture methods. nisms with carcinogenic potential are captured It has been shown that ionizing radiation pro- in these reporter genes. The occurrences of the duces a characteristic spectrum of deletions for both somatic recombinations, the deletions with char- HPRT and BLA mutations. A good deal of effort is acteristic breakpoints, the fusion genes and the now being expended in defining chemical molecu- V(D)J rеcoшЫnase-mediated mutations in the var- lar mutational spectra using both in-vivo and in- ious reporter genes have been described. vitro systems. The identification of such spectra for The use of the various assays detecting events in exposure characterizations may become an impor- cancer genes for epidemiological studies remains tant reason for mutagehicity monitoring in humans. to be defined.

Somatic mutations for detecting genotoxic effects Future directions Somatic mutations unequivocally reveal genotoxic Next-generation assay development effects. Although such effects in reporter genes do The next generation of assay development will not necessarily indicate genotoxic effects in cancer probably involve both reporter and cancer genes. genes, the most important potential application Reporter genes are useful for defining the molecular

176 Somatic cell mutations in cancer epidemiology bases of in-vivo genetic damage. However, the are poorly understood. As our understanding of assays that allow this are costly and difficult to per- cell replication and death increases, it may be pos- form. New developments will therefore be in sible to distinguish more clearly long-term muta- methods to identify and quantify mutations that tional events. This would clearly be advantageous do not require tissue culture. for the purposes of dosimetry. This insight, how- The five current reporter gene mutation assays ever, may also be relevant for understanding the use only blood cells. The development of molecular- time-dependent relationship between exposure based assays will also allow the sampling of other and the development of cancer. For example, radi- non-selectable genes and will include examination ation-induced leukemia is detectable after just a of other tissues. Eventually, a large armamentar- few years, reaches a peak at about 5 years after ium can be developed, using the tissue culture exposure, and decreases thereafter (Committee on method to design and characterize systems and the the Biological Effects of Ionizing Radiation, 1990). molecular technologies for application to large- A similar pattern, but with a substantially longer scale human studies. , wave period, has been described for radon-induced The development of mutation assays for changes lung cancer (Lubin et al., 1994). Studies to evaluate in cancer genes is just now beginning. Future assay the time-dependent occurrence of somatic muta- development will likely focus on exposure-specific tions in cell populations may provide insight into mutations or mutated protein products. Specific this process. and sensitive immunological and molecular assays With the further development of mole culai require further methodological development. approaches to somatic mutation assessment, it will increasingly be possible to investigate multiple Increased understanding of the biological process of markers in human studies. This approach can pro- somatic mutation vide internal consistency checks and may broaden The interpretation of somatic mutations in human mechanistic understanding. populations will be enhanced by increased uлдех- standing of the biological basis of their occurrence Validation of somatic mutations as predictors of and of how this relates to human carcinogenesis. disease outcome Advances in our understanding of the biological A limitation for the interpretation of the reporter basis for seveial of the major gene instability syn- gene somatic cell mutation assays is uncertainty dromes have led to an appreciation of the undez- about the significance of these events as surrogates lуiпg mechanisms for the frequent occurrence of for mutations leading to cancer in humans. For somatic mutations in these syndromes. Other mutations in cancer genes, problems in interpreta- mechanistic links may be expected, as indicated tion also arise because the mutations observed may above for V(D)) recombinase. It has recently been not lead to functional alterations or may be 'after demonstrated that transfection of human cells in the fact', i.e. may indicate existing but subclinical vitro with HPV proteins may inactivate p53 and cancer, and therefore not be useful for the purposes increase mutagenesis of the HPRT reporter gene of prevention. (Havre et cL, 1995). Thus, somatic mutations in With improvements in assay methodology, human populations may be determined by infec- it will increasingly be possible in large-scale tions as well as by chemical' physical mutageп epidemiological studies to determine if specific exposures and underlying genetic susceptibilities. somatic cell mutations are statistically associated Studies in human populations exploring the mech- with increased risks for cancer. Case—control stud- anisms for somatic cell mutation will contribute to ies can be used for this purpose, but will probably an understanding of these interrelated factors, and be of limited value for the 'short memory' markers consequently of a possible relation to cancer. and because of the possibility that disease status may Although the 'memory' of somatic mutations influence marker outcome. 'Nested' case—control appears to diminish with time, the temporal se- studies of a subset of cases and comparison subjects quence of development, appearance in the periph- for whom samples were collected prior to disease eral blood (or other potentially assessed tissues), development as part of a large cohort investigation and eventual disappearance of somatic mutations provide a methodologically sound alternative to

177 Application of Biomarkers in Cancer Epidemiology

the case—control study. Several such cohorts have Albertini, R.J., 5u11ivan, L.S., Berman, J.K., Greene, С.J., now been established, but procedures for the col- Stewart, J.A., Silveira, J.M. & O'Neill, J.P. (1988) lection and storage of biological materials are suit- Mutagenicity monitoring iii humans by autoradiographic assay for mutant T-lymphocytes. Mut. Res., 204, 481-492 able only for molecular studies and not for the standard cell-based assays, unless special proce- Albertini, R.J., Nicklas, J.А., O.'Neill, J.Р. & Robison, 5.H. dures such as cryopreservatlon are undertaken. (1990) 7п vivo somatic mutations in humans: measure- Although a case can be made for the special storage ment and analysis. Ann. Rev. Genet., 24, 305-326 procedures, they are laborious and expensive. Ammenheuser, M.M., Ward J.B. Jr, Whorton, E.В., Jr, Because of this, direct validation studies will usu- Killian, J.M. & Legator, M.5. (1988) Elevated frequencies ally require molecular approaches. of 6-thioguаnine resistant lymphocytes in multiple sclerosis patients treated with While such validation studies are clearly needed cyclophosphamide: a prospective study. Mut. Res., 204, 509-520 to describe the statistical relationship between somatic mutations and cancer risk, a positive asso- Ammenheuser, M.M., Au, W.W., Whorton LB. Jr, Belli, ciation does not establish a causal link. Ultimately, J.A. & Ward J.B., Jr (1991) Comparison otlPRTvariarnt frequencies and chromosome aberration frequencies in lym- the usefulness of the somatic ceI1 mutation assays phocytes from radiotherapy and chemotherapy patients: a will be based upon our understanding of their prospective study. Environ.Mol. Mutagen. 18, 126-135 biological basis and of how this relates to human carcinogenesis. Anderson, D., Hughes, J.А. Brinkworth, lvii., Peltonen K. & Sousa, M. (1996) Levels of ras oncoproteins in human plasma from 1,3-butadiene-exposed workers and References controls. Mut. Res., 349, 115-120 Aidoo, A., Lyn-Cook, L.E., Mittelsyaedt, RA., Heflich, R.H. & Casciano, D,A. (1991) Induction of 6-thiogua- Bachand, M., Seifert, A.M. & Messing, K. (1991) Nuclear nine-resistant lymphocytes in Fischer 344 rats following medicine patients do not have higher mutant frequen- in vivo exposure to N-ethyl-N-nitrosourea and cyclophos- cies after exposure thallium-201. Mut. Res., 262, 1-6 phamide. Environ. Mod. Mutagen., 17, 141-151 Bernini, LE, Nataraj an, A.T., Schreuder-Rotteveel А.Н.I., Akiyama, M., Kyoizumi, 5., Hirai, Y., Hakoda, M., Giordano, P.C., Ploem, J.S. & Tates, A. (1990) Assay for Nakamura, N. & Awa, A.A. (1990) Studies on chromosome somatic mutation of human hemoglobins. In: aberrations and IPRT mutations in lymphocytes and Mendelson, M.L. & Albertini, R.J., ids, Mutation and the GPA mutation in erythrocytes of atomic bomb survivors. Environment Part c: Somatic and Heritable Mutation, In: Mendelson, ML. & Albertini, R.J., eels, Mutation and Adduction, and Epidemiology, Vol Progress in Clinical and the Environment Part c: Somatic and Heritable Mutation, Biological Research, 340e, New York, Wiley-Liss, pp, 57-68 Adduction, and Ерiдеmiоlоgy, Progress in Clinical and Bigbee, W.L., Branscombe, E.W., Weintraub, H.B., Biological Research, 340e, New York, Wiley-Liss, pp. 69-80 Papayannapoulou, T & 5tamatoyannopoulos, G. (1981) Albertini, R.J. (1985а) Somatic gene mutations in vivo as Cell sorter irrununofiuoiescence detection of human ery- indicated by the 6-thioguanine-resistant T-lymphocytes throcytes labeled in suspension with antibodies specific in human biood. Mut. Res., 150, 411-422 for hemoglobin Sand C. J. Immunol. Methods, 45, 117-127 Albertini, R.J. (1985b) The use of human T-lymphocytes to Bigbee, W.L. Wryobek, R.G., Langlois, R.G., Jensen, R.H. monitor mutations in human populations. In: Li, A.P., ed. & Everson, А.B. (1990) The effect of chemotherapy on New Approaches in Toxicity Testing and Their Application in the in vivo frequency of glycophorin A "null" variant Huma" Risk Assessments. New York, Raven Press. pp. 51-66 erythrocytes. Mut. Res., 240, 165-175 Albertini, R.J., Allen, E.F. Quinn, A.s, & Albertini, M.R. Bigbee, W.L,, Grant, 5.G., Langlois, R.G., Jensen, R,H,, (1961) Human somatic cell mutation: in vivo variant lym- Anttila, A., Pfaffli, P,, Pekari, K. & Norppa, H. (1996) Glycophorin A (GPA) somatic cell mutation frequencies phocyte frequencies determined by 6-thioguaпiпe resistant lymphocytes. In: Hook, PB. & Potter, J.H., eds, in Finnish reinforced plastics workers exposed to styrene. Population and Biological Aspects of J-lamai Mutation, New CancerEpideminl. Biomarkers Prev., 5, 801-810 York, Academic Press, pp. 235-263 Bodmer, W.F. (1984) The HLA system, 1984. In: Albeit, Albertini, R.J., Castle, K.L. & Borcherding, W.R. (1982) T- E.D., Baur, M.P. & Mayr, W.R., eds, Histocompatibility Testing 1984, Berhn, 5pring r Verlag, pp. 11-22 cell cloning to detect the mutant б-thioguadne-resistant е lymphocytes present in human peripheral blood. Proc. Boehm, 'Г. & Rabbitts, T.H. (1989) The human T cell NatlAcad. Sсi. USA, 79, 6617-6621 receptor genes are targets for chromosomal abnormali-

178 Somatic cell mutations in cancer epidemiology

ties in T cell tumors. PAsEB J, 3, 2344-2359 Chrysostomou, A., Seshаdri, R.S. & Morley, A.A. (1984) Mutation frequency in nurses and pharmacists exposed Bonassi, 5., AbboпdandoIo, A., Camuuri, L., Dal Pal, L., De Ferrari, M., Degrassi, F., Forni, A., Lamberti, L., Lando, to cytotoxic drugs. Aus. NZ J. Mcd., 14, 831-834 C., Padovani, R, Sbrana, I., Vecchio, D. & Puntoni, R. Clevers, I.В., Alarcon, T., Wileman, C. & Terhost, C. (1995) Are chromosome aberrations in circulating lym- (1988) The T cell rесeptor/CD3 complex: a dynamic pro- phocytes predictive of future cancer onset in humans? tein ensemble. Ann. Rev. Immucol., 6, 629-662 Preliminary results of an Italian cohort study. Cancer Cochrane J.E. & Skopek, T.R. (1994) Mutagenicity of Genet. CytogeiwL, 79, 133-135 butadiene and its epoxide metabolites: II. Mutational Branda, R.F., 0.Net]1, J.P., Sullivan, ~.М. & Albercini, R.J. spectra of butadiene, 1,2-epoxybutenе and diepoxybu- (1991) Factors influencing mutation at the HART locus in tane at the HPRT locus in spleaic T cells from exposed T-lymphocytes: Women treated for breast cancer. Cancer В6СЗF1 mice. Carcinogcnesis, 15, 719-723 Res., 51, 6603-6607 Cole,J. & Skopek, T.R. (1994) International Commission Brandt-Raul, P.W. (1991) Oncogene proteins as biomaik- for Protection against Enviormental Mutagens and ers in tire molecular epidemiology of occupational car- Carcinogens, Working paper 3, Somatic mutant fre- dnogenesis, The example of the ras 0ncogene-encoded quency, mutation rates and mutational spectra in the p21 protein. Intl Arch. Occup. Environ. Health, 63, 1-8 human population in vivo. Mut. Res., 304, 33-106 Brandt-Raul, P.W. (1992) Oncogenes and oncoproteins Cole, J., Arlett, С.F., Green, M.H.L., James, S.Е., in occupational carcinogenesis. Scaпd. J. Work Environ. Henderson, L., Cole, H., Sala-Trepat, M., Benri, R., Price, Health, 18, 27-30 М.L. & Bridges, B.A. (1989) Measurement of mutant frequency to 6-thioguanine resistance i circulating T-lym- Brandt-Rauf, P.W., Smith, S., Hemminki, , Koskinen, п Ю phocytes for human population monitoring. In: Jolies, I-L, Vainio, H., Nimarn, H. & Ford, J. (1992) Serum onco- G. & Cordier, A., eds, New Trends in Genetic Risk proteins and growth factors in asbestosis aid silicosis Assessment. London, Academic Press, pp. 175-203 patients. Ill. J. Cancer, 50, 881-885 Cole, J., Arlett, C.F., Green, .H., Ioldsworth, D., Brandt-Raul, P.W., Pincus, M.R. & Carney, W.P. (1994) М Yawn, E.J. & Bridges, B.A. (1995) Mutant frequencies in The c- rb -2 protein in oncogenesis: molecular structure е в workers at the Sellafield installation. Health Phys., 68, to molecular epidemiology. Cdt. Rev. Oncogen, 5, 313-329 388-393 Br ik, T.M., Mol E.J., Wolvers-Tettero, I.L.M., Ludwig W.- е Committee on the Biological Effects of Ionizing D, van Wering, E.R. & Van Dongen, J.J.M. (1993) Site- Radiation (1990) Health Effects o(Exposure to Low Levels of specified deletions involving the tal-land si genes are Ionizing Radiation, B.E.LR V. Washington, DC, National restricted to cells of the T-cell receptor lineage: T-cell Academy Press, pp. 1-421 receptor gene deletion mechanism affect multiple genes. J. Exp. Med., 177, 965-977 Deisseroth, A., Nienhuis, A,, Turner, P., Velez, R., Anderson, W.F., Ryddle, E., Lawrence, J., Creagan, R. & Caggana, M., Liber, H.L., Mauch, P.M., Coleman, C.N. & Kucherlapati, R. (1977) Localization of the human alpha- Kelsey, K.T. (1991) I vivo somatic mutation in the lym- п globin structurai gene to chromosome 16 in somatic cell phocytes of Hodgkin's disease patients. Environ. MoL hybrids by molecular hybridization assay. Ccl1, 12, Mutages., 18, 6-13 205-218 Cariello, N.F. (1994) Software for the analysis of muta- Deisseroth, A., Nienhuis, A., Lawrence, J., Giles, , tions at the human HPRT gene. Mut. Res., 312, 173-185 Ю Turner, P & Ruddle, FI. (1978) Chromosomal localiza- Cariello, N.F., Craft, Т.R., Vrieling, H., van Zeeland, А.A., tion of human beta-globin gene on human chromosome Adams, T. & Skopek, TR. (1992) Human HPRT mutant 11 in somatic cell hybrids. Proc. Natl Acad. 5cî. USA, 75, database: software for data entry and retrieval. Environ. 1456-1460 MoL Mutagen., 20, 81-83 Dempsey, J.L., Seshadri, R.S. & Могlеy, A.А. (1985) Cariello, N.E & Skopek, T.R, (1993) Analysis of mutation Increased mutation frequency following treatment with occurring at the human HPRT locus. J. Mol. Biol., 231, cancer chemotherapy. Cancer Res., 45, 2873-2877 41-57 DeVivo, L, Marion 1.-J, Smith, S.J., Carney, W.P. & Caтtron J.-P, Collin, Y., Kudo, S. & Fukuda, M. (1990) Brandt-Rauf, P.W. (1994) Mutant c-Ki-ras p21 protein in Molecular genetics of human erythrocyte sialoglycopro- chemical carcinogenesis in humans exposed to vinyl teins: Glycophorins A, B, C and D. In; Harris, J.R., ed,, BIood chloride. Cancer Causes Control, 5, 273-278 Cell Biochemistry (Vol. 1). New York, Plenum, pp. 299-335 Dubeau, H., Zazi, W., Baron, C. & Messing, K. (1994)

179 Application of Biomагkеrs in Cancer Epidemiology

Effects of lymphocyte subpopulations on the clonai Skeтfving, S. & Soma, M. (1994) Cancer risks in hшnans assay of HPRT mutants: occupational exposure to cyto- predicted by increased levels of chromosomal aberra- static drugs. Mat. Res., 321, 147-157 tions in lymphocytes: Nordic study group on the health Finette, B.A., Poseno, T. & Albertini, R.J. (1996) V (D)J risk of chromosome damage. Cancer Res., 54, 2919-2922 recombinase mediated HPRT mutations in peripheral Hakoda, M., Akiyama, M., Kyoizumi, S., Awa, A.A., blood lymphocytes of normal children. Cancer Res., 56, Yamakido, M. & Otake, M. (1988) increased somatic cell 1405-1412 mutant frequency in atomic bomb survivors. Mut. Res., Finger, L.R., Harvey, R.C., Moore, R.C.A., Showe, L.C. & 201, 39-48 Croce, G.M. (1986) A common mechanism of chromo- Hakoda, M., lirai, Y., Shimba, H. & Kusunoki, Y. (1989) somal translocation in T- and B-cell neoplasia. Science, Cloning of phenotypically different human lymphocytes 234, 982-985 originating from a single stem cell. J. Exp. Med., 169, Ford, J., Smith, S., Lui J.-C, Friedman-Jimanez, G., 1265-1276 Brandt-Rauf, h., Markowitz, S., Garibaldi, K. & Niman, H. Havre, P.A., Yuan, J., Hedrick, L., C1m, K.R. & Glazer, PM. (1992) Serum growth factors arid oncoproteins in fire- (1995) P53 inactivation by НРVI6 E6 resuslts in fighters. Occup. Med., 42, 39--42 increased mutagenesis in human cells. Cancer Research, Furthmayer, Fl. (1977) Structural analysis of a membrane 55, 4420-4424 glycoprotein: glycophorin A. J. Supramol. Struct., 7, Hayes, R.B., Xi, L., Bechtold, W.E., Rothman, M., Yao, M., 121-134 Henderson, R., Zhang, L., Smith, M.T., Zhang, D., Furthmayer, H. (1978) StyucturaI comparison of gly- Wiemels, J., Dosemeci, M., Yin, Ѕ . & O'Neill, J.P. (1996) cophorins and immunchemieal analysis of genetic vari- HPRT mutation frequency among workers exposed to ants. Nature, 271, 519--524 1,3-butadiene in China. Toxicology, 113, 100-105 Fuscoe, J.C., Zimmerman, L.J., Lippert, J J., Nicklas, J.A., Henderson, L., Cole, H., Cole, J., James, S.E. & Green, M. O'Neill, J.P. & Albertini, R.J. (1991) V (D)J gecombinase- (1986) Detection of somatic mutations in man: evalua- like activity mediates 'PET gene deletion inhuman fetal tion of the microtitтe cloning assay for T-lymphocytes. T-lymphocytes. Cancer Res., 51, 6001-6005 Mutagenesis, 1, 195-200 Hewitt, M. & Mott, .G. (1992) The assessment of in vivo Gahmberg, С.G., Jakinen, M. & Aridersson, L.C. (1979) М Expression of the major red cell sialoglycoprotein, gly- somatic mutations in survivors of childhood malig- nanci. Br. J. Cancer, 66, 143-147 cophorin A, in the human leukemic cell l е К 562. J. ВиоI. Chem., 254, 7442-7448 Hirai, Y., Kusunoki, Y., Kyoizumi, S., Awa, A.A., Pawel, D.J., Grant, S. & Bigbee, W.L. (1993) In vivo somatic mutation Nakamura, N. & Akiyama, M. (1995) Mutant frequency and segregation at the human glycophoin A (GPA) at the HPRT locus in peripheral blood T-lymphocytes of locus. Mut. Res., 288, 163-172 atomic bomb survivors. Mut. Res., 329, 183-196 Grant S.G. & Bighee, W.L. (1994) Bone marrow somatic Hirota, H., Kubota, M., Hashimoto, H., Adachi, 5., mutation following genotoxic cancer therapy. Lancet, Matsubara, K., Kuwakado, K., Akiyama, Y., Tsutsui, T. & 343,1507-1508 Mikawa, H. (1993) Analysis of HPRT gene mutation following anti-cancer treatment in pediatric patients with Grdina, D.J., Carnes, B.А., Grahn, D. & Sigdestad, C.P. acute leukemia. Mut. Res., 319, 113-120 (1991) Protection against late effects of radiation by Hou, S.-M., Fait, S. & Steen, A.-M. (1995a) Hprt mutant 5-2-(3-amјгiopтоpyhmiпo)-еthуlphosphoгothioic acid. Cancer Res., 51, 4125-4130 frequency and GST11 genotype in non-smoking healthy individuals. Environ. MoI. Mutagen, 25, 97-105 Grdiisa, D.J., Kataoka, Y., Basic, I. & Perrin, J. (1992) The radioprotector WR-2721 reduces neutron-induced muta- Hou, S.-M., Lambert, B. & Hemminki, K. (1995b) rions at the hypoxanthine-guane phosphoriboyyl trans- Relationship between HPRT mutant frequency, aromatic ferase locus in mouse splenocytes when administered prior DNA adducts and genotypes for GSTMI and NAT2 in bus to or following irradiation. Carcinogenesis, 13, 811-814 maintenance workers. Carcinogenesis 16: 1913-1917 Grist, S.A., McCarron, M., Kutlaca, A., Turner, D.R. & H0ttner, E., Mergney, U., Braun, R. & Schoneich, J. (1990) Morley, A.A. (1992) In vivo human somatic mutation: fre- Increased frequency of 6-tMoguanипе-resistant lympho- quency and spectrum with age. Mut. Res., 266, 189-196 cytes in peripheral blood of workers employed in cyclophosphamide production. Mut. Res., 243, 101-107 Hagmar, L., Brogger, A., Hansteen, 1.-L, Heim, S., Hdgstedt, B., Knudsen, L., Lambert, B., Linnainmaa, K., Janatipou , M., Trainor, K.J., Kutlaca, R., Bennett, G., Hay, J., Turner, D.R. & M l , A.A. (1988) Mutations in Mltelman, E, Nordenson, l., Reuterwall, C., SaIошaa, S., огеу

180 Somatic cell mutations in cancer epidemiology human lymphocytes studied by an HLA selection sys- for somatic mutations at the glycophorin A locus in tern. Mut. Res., 198, 221-226 humans. Cytometry, 11, 513-521 Janeway, C.A., Jr & Travers, P. (1996) Chapter 4: antigen Langlois, R.G., Aldyama, M., Kusunoki, Y., Dupont, BR., recognition by T-lymphocytes. In: Immunobiology: The Moore D.H., I.I., Bigbee, W.L., Grant, S.G. & Jensen, R.H. Immune System in Health and Disease, 2nd edn, London, (1993) Analysis of somatic cell mutations at the gly- Current Biology Ltd, pp. 38-48 cophorin A locus in atomic bomb survivors: a comparative study of assay methods. Rad. Jensen, R.H. & Bigbee, W. (1996) Direct immunofluores- Res., 136, 11-117 сепсе labeling provides an improved method for the gly- Lipkowitz, S., Garry, V.F. & Kirsch, I.R. (1992) Igiteгlocus cophorin A somatic cell mutation assay. Cytometsy, 23, V-J recombination measures genomic instability in agri- 337-343 culture workers at risk for lymphoid malignancies. Proc. Nat? Acad. Sri. USA, 89, 5301-5305 Jensen, R,H., Langlois, E.G., Bigbee, W.L., Grant, S.G., Moore D., Ш, Piliлskaya, M., Vorobtsova, I. & Pleshanov, P. Liu, Y., Hernandez, A.M., Shibata, D. & Cortopassi, G.A. (1995) Elevated frequency of glycophorin A mutations in (1994) BСL2 translocation frequency rises with age in erythrocytes from Chernobyl accident victims. Rad. Res., humans. Proc. Nat? Anal. Sc?. USA, 91, 8910-8914 141, 129-135 Lubin, J.Н., Bocce, J.D., Jr, Edling, C., Hornung, R.W., Kelsey, K.T., Donohoe, J.K., Memisoglu, A., Baxter, B., Howe, G., Kunz, E., Kusiak, R.A., Morrison, H.I., Radford, Caggana, M. & Liber, H.I. (1991) In vivo exposure of Е.P., Samet, J.М., Tirmarche, I., Woodward, A., Xiang, human lymphocytes to technetium-99т in nuclear Y.S. & Pierce, D.A. (1994) Radon and Lung Cancer Risk: А medicine patient does not induce detectable genetic Joint Analysis of 11 Underground Miners Studies (HIH effects. lut. Res., 264, 213-218 Publication No. 94-3644), Bethesda, MD, US Kudo, S. & Fukuda, M. (1989) Structural organization of Department of Health and Human Services, Public Health Service, National Institutes of Health, pp 1-136 glycophorin A and B genes: glycophorin B gene evolved by homologous recombination at Alu repeat sequences. IcCarron, M.A., Kutlaca, A. & Morley, A.А. (1989) The Proc. Nail Асад. Sci. USA, 86, 4619-4623 HLA-A mutation assay: improved technique and normal lut. Kushiro,J.-I., Huai, Y., Kusunoki, Y., Kyoizumi, S., Kodama, results. Res., 225, 189-193 Y., Wakisaka, A., Jeffreys, A., Cologne, J.E., Dohi, K., Manchester, D.K., Nicklas, J.A., 0.'Nei11, J.P., Lippert, Nakamura, N. & Akiyama, M. (1992) Development of a M.J., Grant, S.G., Langlois, R.G., Moore, LI., D.H., Jensen, flow-cytometric HLA-A locus mutation assay for human R.H., A1bertini, R.J. & Bigbee, W.L. (1995) Sensitivity of peripheral blood lymphocytes. Mut. Res., 272, 17-29 somatic mutations fn human umbilical cord blood to maternal environments. Environ. Kyoizumi, S., Nakamura, N., Hakoda, M„ Awa, A.A., Mol. Mutages., 26, 203-212 Bean, М.A., Jensen, А.H. & Akiyama, M. (1989) Detection of somatic mutations at the glycophorin A locus in ery- Mendelsohn, M.L. (1990) New approaches for biological throcytes of atomic bomb survivors using a single beam monitoring of radiation markers. Health Phys., 59, 23-28 flow sorter. Cancer Res., 49, 581-588 Messing, K. & Bradley, W.E.C. (1985) In vivo mutant fre- Kyoizumi, S., Akiyama, M., Hirai, Y., Kusunoki, Y., quency rises among breast cancer patients after exposure Tanabe, K. & Umeki, S. (1990) Spontaneous loss and to high doses of gamma radiation. lut. Res., 152, alteration of antigen receptor expression in mature CD4+ 107-112 T cells. J. Exp. Med. 171: 1981-1999 Messing, K., Seifert, A.M. & Bradley, W.E.C. (1986) Iп vivo Kyoizumi, S., Umeki, S., Akiyama, M., Hirai, Y., mutant frequency of technicians professionally exposed Kusunoki, Y., Nakamura, N., Endoh, K., Konishi, J., to ionizing radiation. In: Sorsa, I. & Norppa, H., eds, Sasaki, M.S., Mori, T., Fujita, S. & Cologne, J.B. (1992) Monitoring o f Occupational Genotoxicants, New York, Alan Frequency of mutant T lymphocytes defective iп the R. Liss, pp. 87-97 expression of the T cell antigen receptor gene among Messing, K., Ferraris, J., Bradley, W.E.C., Swartz, J. & radiation-exposed people. Mut. Res., 265, 173-180 Seifert, A.M. (1989) Mutant ftegьencу of radiotherapy Langlois, R.G., Bigbee, W.L., Kyoizumi, S., Nakamura, N., technicians appears to be associated with recent dose of Bean, M.A., Akiyama, M. & Jensen, R.H. (1987) Evidence ionizing radiation. Health Phys., 57, 537-544 for increased somatic cell mutations at the glycophorin Morley, A.A., Trainor, К.J., Dempsey, J,L. & Seshardi, R.S. A locus in atomic bomb survivors. Science, 236, 445-448 (1965) Methods for study of mutations and mutagenesis Langlois, R.G., Nisbet, B., Bigbee, W.L., Ridingei, D.N. & in human lymphocytes. Mut. Res., 147, 363--367 Jensen, R.H. (1990) An improved flow cytometnic assay Могlеу, A.A., Grist, 5.A., Turner, DR., Kutlaca, A. &

181 Application of Biomarkers in Cancer Epidemiology

Bennett, G. (1990) Molecular nature of in vivo mutations genesis in asbestosis patients. JIM, 36, 1324-1328 in human cells at the autosomal HLA-A locus. Cancer Patel, P.I., Nussbaum, R.I., Framson, P.E., Ledbetter, D,H., Res., 50, 4584-4587 Caskey, C.T. & Chinault, A.C. (1984) Organization of the Mitt, M.G., Boyse, J., Hewitt, M. & Radford, M. (1994) IPRT gene and related sequences in the human genome. Do mutations at the glycophorin A locus in patients Som. Cell Моi. len., 10, 483-493 treated for childhood Hodgkin's disease predict sec- Perera, F.P., Motzer, R.J., Tang, D., Reed, E., Parker, R., ondary leukaemia? Lancet, 343, 828-830 Warburton, D., 0.'Neill, J.Р., Albertini, R., Bigbee, W.L., Nicklas, J.A., Hunter, T.C., 0.'Nei11, J.P. & Albertini, R.J. Jensen, R.H., Santella, R., Tsai, W.-Y., Simon-Cеreijldo, (1989) Molecular analyses of in vivo HPКТ mutations in G., Randall, C. & Bosl, G. (1992а) Multiple biologic human T-lymphocytes: III. Longitudinal study of HPRT markers in germ cell tumor patients treated with plat- gene structural alterations and T-сеll clonai origins. Mut. inum-based chemotherapy. СапсетRеs., 52, 3558-3565 Res., 215, 147-160 Perera, F.P., Hemminki, K., Gryzbowska, E., Motykiewicz, Nicklas, J.A., Falta, M.T., Hunter, T.C., O.'Neiil, J.P., G., Michalska, H., Satella, R.M., Young, T.L., Dickey, C., Jacobson-Kram, D,, Williams, J. & Albertini, R.J. (1990) Brandt-Rauf, P.W., DeVivo, I., Blaner, W., Tsai, W.-Y. & Molecular analysis of in vivo InTPRT mutations in human Chorazy, M. (1992b) Molecular and genetic damage in lymphocytes V. Effects of total body irradiation sec- humans from environmental pollution in Poland. ondary to radioimmunoglobulin therapy (RIT). Nature, 35, 256-258 Mutagenesis, 5, 461-468 Perera, EP., Tang, П .L., 0.'Neill,J.P., Bigbee, W.L., Albertini, Nicklas, J.A., O'Neill, J.Y., Hunter, T.C., Falta, M.T., А.J., 5аntella, R., Ottman, R., Tsai, W.-Y., Dickey, C., Lippert, M.J., Jacobson-Kam, D., Williams, J.R. & Mooney, L.A., Savela, K. & Hemminki, K. (1993) HPRT Albertini R.J. (1991) In vivo ionizing irradiations produce and glycophorin A mutations in foundry workers: rela- deletions in the HPRT gene of human T-lymphocytes. tionship to РАН exposure and to PAH-DNA adducts. Mut. Res., 250, 383-396 Carcinogenesis, 14, 969-973 O'Neill, J.P., McGinniss, M.J., Berman, J.K., Sullivan, Perera, F.P., Dickey, C., Sarste1la, R., O.'Nejll, J.P., L.M., Nicklas, J.А. & Albertini, R.J. (1987) Refinement of Albertini, R.J., Ottman, R., Tsai, W.Y., Mooney, L.A., a T-lymphocyte cloning assay to quantify the in vivo Savela, K. & Hemminki, K. (1994) Carcinogen-DNA thioguanine-resistant mutant frequency in humans. adducts and gene mutation in foundry workers with low- Mutagenesis, 2, 87-94 level exposure to polycyclic aromatic hydrocarbons. Ostrosky-Wegman, P., Montero, R.M., Cortinas de Nova, Carcinogenesis, 15, 2905-2910 C., Tice, R.R. & Albertini, R.J. (1988) The use of bromod- P1uth, J.M., Nicklas, J.А., O'Neill, J.P. & Albertini, R.J. eoxyuridйe labeling in the human lymphocyte HGPRT (1996) Increased frequency of specific genomic deletions somatic mutation assay. Mut. Res., 191, 211-214 resulting from in vitro malathion exposure. Cancer Res., Ostrosky-Wegman, P., Montero, R., Palao, A., Cortinas de 56, 2393-2399 Nava, C., Hurtado, C & Albertini, R.J. (1990) 6-thiogua- RairwiIle, I.E., Albertini, R.J., O'Neill, J,P & Nicklas, J.A. nйe-resistant T-lymphocyte autoradiographic assay. (1995) Breakpoints and junctional regions of intragenic Determination of variation frequencies in individuals deletions in the HPRT gene in human T-cells. Som. Cell suspected of radiation exposure. Mut. Res., 232, 49-56 Mol. Gm., 21, 309-326 Ostrosky-Wegman, P., Gonsebatt, M.E., Montero, R., Robinson, D., Goodall, K., Albertini, R.J., O'Neill, J.P., Vega, L., Barba, H., Espinosa, J., Palao, A., Cortinas, C., Finette, B., Sala-Tгepаt, M., Moustacchi, E,, Tates, А.D., Garcia-Vargas, G., Del Razo, L.M. & Cebrian, M. (1991) Beare, D.M., Green, M.H.L. & Cole, J. (1994) An analysis Lymphocyte proliferation kinetics and genotoxic find- of in vivo HPRT mutant frequency in circulating T-1ym- ings in a pilot study on individuals chronically exposed phocytes in the normal human population: a compari- to arsenic in Mexico. Mut. Res., 250, 477-482 son of four databases. Mut. Res., 313, 227-247 Palmer, E.G., Smith-Burchnell, C.А., Pelton, B.K., Hylton, Rothman, N., Haas, R., Hayes, R.B., Li, G.-L., Wiemels, J., W. & Denman, A.М. (1988) Use of T cell cloning to Camp1eman, 5., Quintana, P,J.E., Xi, L. J., Dosemeci, M., detect in vivo mutations induced by tyclophoshamide. Titenko-Holland, N., Meyer, E.g., Lu, W., Zharng, L.P., Arth. Rheum., 31, 757-761 Bechtold, W., Wang, Y.-Z., Kolachana, R, Yin, 5.-N., Blot, W. Partanen, R., Hemminki, K., Koskinen, H., Luo, J.-C., & Smith, М.T. (1995) Benzene induces gene-duplicating Carney, W.P. & Brandt-Rauf, P.W. (1994) The detection of but not gene-inactivating mutations at the glycophorin A locus in exposed humans. Proc. Natl Acad. 3 i. USA, 92, increased amounts of the еxtгace11ulaт domain of the cpi- с dermal growth factor receptor in serum during carcino- 4069-4073

182 Somatic cell muiations in cancer epidemiology

Sala-Trepat, M., Cole,J., Green, M.H.L., Rigaud, 0., Vilcoq, Verwoed, N.P., Cole, J., Green, M.H.L., Arlett, C.F. & JR. & Moustacchi, E. (1990) Genotoxic effects of radio- Norris, P.N. (1989) Detection of somatic mutations in therapy and chemotherapy on the circulating lymphocytes man: HPRT mutations in lymphocytes and hemoglobin of breast cancer. III: Measurement of mutant frequency mutations in erythrocytes. Mut. Res., 213, 73-82 to 6-thioguanine resistance. IvIutagenesis, 5, 593-598 Tates, A.D., Grurmnt, T., Tornqvist, M., Farmer, P.B., van Sala-Trepat, M., Boyse, J., Richard, P., Papadopoulo, D. & Dam, FJ., van Mossel, H., Schoemaker, H.М., Osterman- Moustacchi, E. (1993) Frequencies of HPRT- lymphocytes Golkar, S., Uebel, C., Tang, Y.S., Zwinderman, A.H., and glycophorin A variants erythrocytes in Fanconi ane- Natarajan, A.T, & Ehrenberg, L. (1991) Biological and mia patients, their parents and control donors. Mut. Res., chemical monitoring of occupational exposure to ethyl- 289, 115426 ene oxide. Mut. Res., 250, 483-497 Seifert, A.M., Bradley, W.C. & Messing, K. (1987) Tates, AD., van Dam, F.J., Natasajan, A.T., Zwinderman, A.H. Exposure of nuclear medicine patients to ionizing radia- & Osanto, s. (1994а) Frequencies of HPRT mutants and tionis associated with rises in HPRT- mutant frequency micronuclei in lymphocytes of cancer patients under in peripheral T-lymphocytes. Mut. Res., 191, 57-63 chemotherapy: a prospective study. Mut. Res., 307, 293-306 Seifert, А.M., Demers, C., Debeau, H. & Messing, K. Tates, А,D., Grummit, T., van Dam F.J., de Zwart, F., (1993) HPRT-mutant frequency and lymphocyte charac- Kasper, F.J:, Rothe, R., Stirn, H., Zwinderman, All. & teristics of workers exposed to ionizing radiation on a Natarajan, A.T. (1994Ь) Measurement of frequencies of sporadic basis: a comparison of two exposure indicators, HPRT mutants, chromosomal aberrations, micronuclei, job title and dose. Mut. Res., 319, 61-70 sister-chromatid exchanges and cells with high frequencies of SCEs in styrene/dichlorometharie-exposed work- stamatoya raopoulos, G.P. & Nute, Р Е , (1981) screening of human erythrocytes for products of somatic mutation, ers. Mut. Res., 313, 249-262 an approach and critique. In: Hook, E.B. & Porter, I.H., Tates, A.D., Boogaard, P.J., Darroudi, F., Natarajan, A.T., eds, Population and Biological Aspects of Human Mutation, Caubo, М.E. & van Sittert, N.J. (1995) Biological effect New York, Academic Press, pp. 265-273 monitoring in industrial workers following incidental Mut. Stamatoyanopoulos, G., Nate, P., Lindsley, D., Farguhar, exposure to high concentrations of ethylene oxide. Res., 329, 63-77 М.B., Nakamato, B. & Papayannepoulou, T.N. (1984) Somatic-cell mutation monitoring system based on Tates, A.D., van Dam, FJ., de Zwart, F.A., Darroudi, F., human hemoglobin mutants. In: Ansari, A.А. & de Serres, Natarajan, A.T., Rossner, P., Peterkova, K., Pe1tonen, K., F.J., eds, single Cell Monitoring System, Topics in Chemical Demopoulos, N.A., V1achodimitropoulos, D. & Sram, R.J. Mutagenesis (Vol 3), New York, Plenum Press, pp. 1-35 (1996) Biological effect monitoring in industrial workers Stark, M. ., Tucker, J.I1., Thomson, E.J. & Perry, FE, from the Czech Republic that were exposed to low levels Н of butadiene. Toxicology, 113, 91-99 (1984) An automated image analysis system for the detection of rare autoradiographica11y labeled cells ira the Theocharis, S., Sfikakis, P.P., Lipnick, R.N., K1ipple, G.L., human lymphocyte HGPRT assay. Сytomelrу, 5, 250-257 Steinberg, A.D. & Tsokos, G.C. (1995) Characterization of mutated T cell clones from patients with systemic Stout, J.T. & Caskey, C.T. (1985) Hprt: gene structure, in vivo lupus erythematosus. Clin. Immunol. Immunopath., 74, expression, and mutation. Arm. Rev. Genet., 19, 127-148 135-142 Straume, T., Langlois, R.G., Lucas, J., Jemsem, R.H., Bigbee, W.L., Ramalho, A.T. & Brandao-Mello, C.E. Turner, D.R. Grist, S.A., Janatipour, M. & Morley, A.A. (1988) Mutations in human lymphocytes commonly (1991) Novel biodosimetry methods applied to victims involve gene duplication and resemble those seen in can- of the Goiania accident. Health Phys., 60, 71-76 cer cells. Proc. Nat! Acad. Sc!. USA, 85, 3189-3193 Strauss, G.H. & Albertini, R.J. (1977) 6-thioguaпine resistant lymphocytes iii human perpheral blood. In: Scott, Tycko, B. & Sklar, J. (1990) Chromosomal translocations in lymphoid neoplasia: a reappraisal of the recombinase D., Bridges, B.A. & Sobels, F. ., eds, Progress in Genetic Н model. Cancer Cells, 2, 1-8 Toxicology: Developments in Toxicology and Environmental Sciences (vol. 2), Amsterdam, Elsevier/North Holland Umeki, S., Kyoizumi, S., Kusunoki, Y., Nakamura, N., Biomedical Press, p. 327 Sasaki, M., Mori, T., Ishikawa, Y., Cologne, J.В. & Akiyama, M. (1991) Flow cytometric measurements of somatic cell Strauss, G.H. & A1 ertinl, R.J. (1979) Enumeration of 6- Ъ mutations in Thorotrast patients. Jm J. Cancer Res., 82, thioguanine resistant peripheral blood lymphocytes in man as a potential test for somatic cell mutations arising 1349--1353 in 4vo. Mut. Res., 61, 353-379 van Dam, F.J., Camps, J.A., Woldring, V.1., Natarajan, .T., van der Wall, E.E., Zwinderman, A.H., Lohman, Tates, A.D., Bernini, L.F., Natajaran, A. ., Ploem, J,S., А Т FF1.1., [bowels, E.K.J. & Tates, A.D. (1991) Radionuclide

183 Application of Biomarkers in Cancer Epidemiology angiography with Technetium-99m in viva labeled cry- Ward, J.B., Jr, Ammenheuser, M.M., Whorton, E.B:, Тr, tiirocytes does not lead to induction of mutations in the Bechtold, W.E., Kelsey, K.Т. & Legator, M.S. (1996) HPRT gene of human T-lymphocytes. J. Nod. Med., 32, Biological monitoring for mutagenic effects of occupa- 814-818 tional exposure to butadiene. ТохгсоIоgy, 113, 84-90 Vijayalaxmi, Wunder, E. & Schroeder, T.M. (1985) Spontaneous 6-thioguaniпe-resistant lymphocytes in Fanconi anemia patients antI their heterozygous parents. Hum. Genet., 70, 264-270 Vodicka, P., Bastlova, T., Vodickova, L., Peterkova, K., Lambert, B. & Hemminki, K. (1995) Biomarkers of styrene exposure in lamination workers: levels of 06- guaisine DNA adducts, DNA strand breaks and mutant frequencies in the hypoxanthine guanine phosphoribo- syltransferase gene in T-lymphocytes. Ccrrcinogenesfs, 16, 1473-1481 Corresponding author. R.J. Albertini Ward, J.B., Ammenheuser, М.M., Bechtold, W.E., Whorton, E.B., Jr & Legator, M.S. (1994) HPRT mutant Genetics Laboratory, University of Vermont, 32 N. Prospect St, BurLington, VT 05401-0505, U5A lymphocyte frequencies in workers at a 1,3-butadiene production plant. Environ. Health Perspect. 102, 79-85

184 Аpplicalion of Biomarker5 iп Corcer Epidemiology Toпiolo, P., Boffetta, P., Shuker, E.Е.G., RoЁ hmaп, N., Hulka, B. апд Pearce, N., eds 1ARC Scientific PuЫicalioпs No. 142 lпleгпatioпal Ater'cy for Research on Cancer, Lyon, 1997 Cytogenetis end-points as biological dosimeters and predictors of risk in epidemiological studies

J.D.Tucker, D.A. Eastmond and L.G. Littfefïе d

Cytogenetic end-points have been successfully used in epidemiological studies for many years. Conventional end-points are now being replaced by procedures that utilize molecular methods, with greatly increased sensitivity, specificity and precision. In this paper we briefly review the most common cytogenetic assays that are useful in epidemiological settings, including structural chromosome aberrations, micronuclei, sister chromatid exchanges and analysis of interphase cells for aneuploidy. We describe new developments of each assay, where applicable, and discuss the strengths and weaknesses of the assays for detecting exposures and estimating risks. Finally, pertinent information concerning each of the assays that is useful in designing epidemiological studies is summarized in a table. It is hoped that the information presented here will be useful to individuals who are interested in applying biomarkera to studies of human environmental exposure and disease.

Several cytogenetic end-points have potential use 1960s. Information gleaned from a multitude of as biomarkers of exposure to clastogens. These radiation cytogenetic studies has also provided an include structural chromosome aberrations, insight into the mechanisms that govern the micronuclei and sister chromatid exchanges (SCEs). induction, persistence, accumulation and elimina- Structural aberrations have been studied for more tion of chromosome damage and into the rele- than half a century and thousands of papers have vance of such damage to risk of late effects in been published on the effects of DNA-damaging exposed populations. Attempts to derive recom- agents in inducing chromosome damage. At the mendations regarding the applicability of cytoge- molecular level, aberrations are believed to result netic techniques in population epidemiology must from double-strand breaks ira DNA that remain take into account a wealth of information that has unrepaired or that undergo aberrant rejoining, been accumulated over the past 30 years. One of giving rise to chromosomal rearrangements. the compelling arguments for studying cytogeпet- While dozens of chemical, physical and biolog- ics is the relationship between chromosome aber- ical agents induce DNA damage, ionizing radiation rations and tumour formation. It is of particular is among the most efficient in inducing double- interest to discover whether aberration frequencies strand breaks that lead to the prompt formation of can be used to predict cancer. Several recent reports chromosome aberrations. It has been known for (Hagmar et al., 1994; Bonassi et яI., 1995) have more than three decades that radiation induces shown that aberration frequencies are increased dose-dependent increases ira asymmetrical (i.e. prior to the clinical manifestation of disease. unstable) and symmetrical (i.e. stable) chromosome If this relationship holds true upon further in- aberrations in cells throughout the body, including vestigation, then it will be important to determine mature cells and progenitor stem cells in the the statistical strength of this predictability for vari- haematopoietic lineage. Induced aberrations in ous types of cancer, and to decipher the biological human lymphocytes have been used as biological mechanisms responsible for these intriguing obser- dosimeters to gauge exposure levels since the early vations.

185 Applгсation of Biomarkers in Cancer Ерidеmiology

Micronuclei contain chromosomes or portions aberrations such as dicentrics and rings are readily of chromosomes that exist separately from the detected using these techniques. Symmetrical main nucleus of a cell. Their existence has been chromosome rearrangements are not easily known for decades, and they are frequently used to observed, and early studies used clinical laboratory quantify exposure to chemicals and radiation. methods ofrdiogramming or karyotyping of pho- Micronuclei can be observed in almost any cell tographic prints to quantify translocation fre- type, and for this reason many variations of the quencies (e.g. Buсkton et al., 1962; Littlefield & assay exist. Compared to other cytogenetic assays, Joiner, 1978). Each metaphase cell had to be there are several advantages of quantifying counted and mentally karyotyped by locating and micronuclei, including speed and ease of the comparing the sizes and centromere location of analysis, and especially the absence of the require- chromosome pairs 1, 2, 3 and 16, and by counting ment for examining cells in metaphase. Thus, and comparing the centromere locations of the B, many cell types are amenable to analysis of D, F and G group chromosomes. Typically, any micronuclei, and most can be employed in epi- metaphase cell that contained an abnormal mono- demiological studies. Several mechanisms of centric chromosome was photographed and kary- action contribute to the formation of micronuclei otyped. Interstitial translocations and those (I-Ieddle et al., 1983), including chromosome involving exchanges of small pieces of chromatin breakage (clastogenesis) and spindle disruption were not detected. Comparisons of translocations (aneuploidogenesis). detected by these methods and even more labori- SCEs are the cytological manifestation of inter- ous banding techniques have demonstrated that, changes between DNA replication products at when performed by highly skilled staff, group apparently homologous loci. SCEs have been com- analysis techniques detect upwards of 75% of the monly employed to evaluate cytogenetic responses cells with symmetrical rearrangements iп the to chemical exposure, and hundreds of chemicals entire genome (Buckton, 1976; Ohtaki et al., 1982). have been evaluated in a wide variety of in-vitro Group analysis and banding methods are slow and and in-vivo short-term experiments (Tucker et al., tedious and require highly-trained staff who have 1993a). Numerous studies have also been con- complete familiarity with the human chromosome ducted to evaluate possible environmental expo- complement. Thus, the methodology is not applic- sure in humans. Unfortunately, the mechanism of able for the large-scale screening of populations SCE formation remains unknown, thus limiting that is frequently of interest in epidemiological the use of this biomarker for evaluations concern- surveillance. 11g the effects of exposures. The application of cytogenetic techniques in The objective of this article is to provide perti- population epidemiology is becoming more feasi- lent information concerning these cytogenetic ble with the recent development of fluorescence assays that is applicable when designing epidemi- in-situ hybridization (FISH) using chromosome- ological studies of human environmental exposure specific DNA probes that paint specific chromo- aid disease. Table 1 summarizes the salient fea- some pairs along their entire lengths (e.g. Lucas et tures of these end-points with respect to their util- al., 1989; Natarajan etaL, 1991; Tucker et al., 1993b). ity as biomarkers in such studies. These methods allow the selective identification of each pair of chromosomes in the human genome Structural chromosome aberrations and permit rapid recognition of chromosome Since the early 1960s, three methods have been breakage and exchange events between painted used to observe and quantify structural chromo- and non-painted chromosomes. Symmetrical ек- some aberrations in human cells. These are analy- changes can be readily observed, and these new sis of unbanded chromosomes, banded chromo- methods have refocused the attention of the somes and painting. Classical studies typically scientific community on the translocation as a used DNA dyes such as Gien~sa to stain all chrо- potential retrospective biomarker of exposure. mosomes in the complement, and the earliest Because chromosome painting techniques typi- methods did not allow unique identification of all cally quantify only those exchanges that occur the chromosomes in the complement. Unstable between painted and non-painted chromosomes, Cytogenetic end-pulls as biological dosimeters and predictors of risk it must be assumed that the observed exchange currently available for detecting recent exposures events in the target chromosomes are accurate to clastogenic agents (e.g. Bender et al., 1988). surrogate measures of those occurring in the entire Unstable aberrations such as dicentrics are easily genome. Estimates of total genomic frequencies identified in metaphase cells of cultured peripheral can be calculated (Lucas et cL, 1992) by taking into blood lymphocytes, and because these occur with consideration the percentage of the total chromo- Iow background frequency (-1-2 per thousand) in somal DNA that is painted, and by assuming that lymphocytes of persons with low levels of envi- random breakage and exchange occurred between ronmental or occupational exposures to clasto- all pairs of chromosomes. In-situ hybridization gens, relatively minor increases above background procedures require some knowledge of molecular can be readily detected. For example, when suffi- biology techniques, expensive reagents and a cient numbers of metaphase cells are evaluated good-quality fluorescence microscope. Procedures from lymphocyte cultures initiated within a few for hybridization typically result in some loss of weeks of exposure, scoring of dicentrics can read- chromosomal morphology, as a result of subjecting ily detect recent whole-body exposures of -10-20 cells to extremes in temperatures, high concentra- cGy in individuals and -5 cGy in populations dons of salts, as well as enzymes and formamide, exposed to ionizing radiation. Dicentrics and other which can cause problems in ascertaining accu- types of unstable aberrations are lethal when cells rately some of the aberration types. None the less, undergo division, and are therefore quickly elimi- painting methods offer a number of advantages in nated from proliferating cell compartments such terms of population monitoring, including the rel- as bone marrow. Dicentrics will only be observed ative ease of scoring large numbers of metaphase in populations of mature lymphocytes that have cells to obtain estimates of translocation frequen- not undergone in-vivo cell divisions in the interim cies, objectivity of scoring (which is expected to between exposure and collection of blood for cyto- reduce ioterscorer variability), and the potential genetic evaluation. The average life expectancy of for full computerization of data collection, which T lymphocytes appears to be bimodal, with means would make large-scale monitoring a feasibility. of approximately 1.1 arid 6.3 years (Bogen, 1993), and Prior to the publication of painting methods, lymphocytes carrying dicentrics will be useful for dozens of cytogeaetic studies were conducted biological dosimetry purposes. Thus, in terris of popu- using classical metaphase analysis to evaluate the lation monitoring, unstable aberrations in lym- induction and persistence of structural chromo- phocytes are useful as transient biomarkers of recent some aberrations in various types of somatic cells exposure and reflect only a snapshot of a person's in individuals and populations with medical, occu- exposure history. In general, samples obtained pational or accidental exposures. A number of con- within 6 weeks of a single acute exposure should clusions derived from these studies have now been be suitable for biodosimetric analysis by scoring verified and expanded upon with chromosome dicentric chromosomes. After this tire, dicentric painting techniques. In the following paragraphs frequencies begin to decline and precise dosimetry we summarize the current status of knowledge becomes more difficult. Back-extrapolation of regarding the applications and limitations of struc- dicentric frequencies obtained after 6 weeks may tural chromosomal aberrations as prompt or retro- be possible, but additional sources of error are spective biomarkers of exposure and predictors of introduced. Measurements of translocation fre- risk in epidemiology studies. quencies become the preferred method for quantifying exposure due to the stability of the lesions. Sепsitivlty of structural aberrations for detecting recent exposures Translocations as retrospective biodosirneters Studies of radiation-induced chromosome aberra- As first observed and discussed in the early 1960x, tions in hundreds of persons with accidental when persons are exposed to physical or chemical radiation exposures have firmly established that agents capable of inducing chromosome damage, unstable (or asymmetrical) types of chromosome stable chromosome rearrangements, such as recip- aberrations in cultured peripheral blood lympho- rocal translocations, are also induced in mature cytes are the most sensitive biological end-points cells and in progenitor cells in various stem cell

187 Application of Biomarkers in Cancer Epidemiology

ф ja C ~ N

Z p m, Eл .5 o ) и Û ~ и

° ы ô ô _ .`~ O O O L Ф Û f1 0. - - ф S] ф й 1 ф O` гn 6 cб a = 7 и Sг m - 21 Й Û Й Ô И >~ O ❑ Z ш a д 0 (Л гc d .E Q â ._moi Q a E Y Ф' O U Q O ф Z C ф O И c O7 Ci 9 и 1- U ' N '- a -o С0 Ф ф 1 п `7+ E 1 Ф d p E ~ ' V -0 O L - 1] '5~ Ф -=2 ц Sо i Й 1 [] Ш 2 о 9 O .и O ❑ Z ш 2 o2 cЛ m 06.. 5 . EL. Q -

ф0 И . и b - ô ô a °- É ° гô Eф '5 и o C Ô ф O 5 ф ф S]

o, z .zo oO ~ Q

Ç [U 02 и 'Q i o N и - p ÿ ô o . _ Ф Ф m ~? ф ô n ô O t O N й O ф y tC a (л Z о D .E Q a Q T _

'Q я ..- =Й 05 a 5n ô fEд ô m ф ° й -â r ш 5 9 m Ç 0 m 9 a ~ И ô ди Ф `a Q â и a01 a° i1 ¢ ~° .

C и a '5и Q ô É E ~o Ф 2ш ф ' 2 '5E m J ш ф Й ~_ ш ~ m 0 .j И й ~_ й - 'В m

фC иC иC C -C O C C -' C _C

V O ф 5 O П 3 O [ц 5 o 1 -s O 9 Ф pб 'y p '_ ' Q ô â ÿ 7 ôD C о O о O с Ф N CDI (rO C6 И U [C CDI d 0) C 07 ф Ш Ш Ф Ф Ф d1 й E ES C й E E S - 5 E É Ç F d Q U) ( lî И й Ф Q Л (J1 й ¢ CЛ гЛ й

188 Cytogenetic end-points as biological dosimeters and predictors of risk

L Ô 0) 2 = Ô Ш Ф C O ГJ o Ô ç C " L7 O O CO И Q O cб Ф ro - O E ro й t m д cl ô °_ ô ° E m Q1 m 0Ё л E ш m io m :1 и о 7 00. 01 [6 O a [[ v ¢ a â â oCс w C п з й I O p~ И И _ a) И ~ ~ C1" O d 1 t_ • ~ -Ш rn O C O Q U и ы7 C m о ~ Ç Ф cб Ô

m Ci _ 0) 'п °cб toд Ф a L2 O- O_ z z z OC z C_ V a a -U O N ~ ° Ici R д c6 [C 0. Q O- 'rй ф иo ô ô â z z a ш (j) (c m 9 -O л1 L a U Uб O Ô E Û Û O) ° Ç ~ d 2 a И d d И N С') N fЛ ° C m z0 z0 0 йэ (#) .0 z

a —C 9 мo ce ci E м m m ci a) 0. â Q ф ÔO O ы й z z a . ш z 2 m C E Et 'D 0) 2 ci и C CИ ~ E о -o .Q ф -O .0 п -ô й 7 Ô .._ O c E U ~ б ~, 'â o_ . фc a E d гц а) й й a .CS .~ o is a Cil) = Ci") 0 0) 0) } â o O v o c =p C m Cii m c0 N‚Di O 07 }+ 0)0)0)и o) ` 0) 0) Ф Ф Ф л Q C N ~в , ç. . C E } Q .Е m Е Е _ ~C_Е ФL n .- cice (d d ci 9c И ci ci '0 с Q. И ci '5 С ¢ u] [Л й 7 ¢ и u) й И

169 Application of Biomarkers in Cancer Epidemiology

compartments. Because these types of rearrange- observations suggest that progenitor subsets of ments typically do not impede cellular progression T-lymphocytes may undergo selective clonal through mitosis, translocations are passed on to expansion under certain conditions. Studies in daughter cells during cci proliferation. If cells populations with exposures over a wide range of bearing translocations do not have a selective doses (i.e. A-bomb survivors in Hiroshima and advantage or disadvantage, compared to those Nagasaki, and women irradiated locally for cervical with normal chromosome complements, then one cancer) have demonstrated that the mean fre- might exped them to persist throughout a lifetime quencies of Iymphocytes bearing translocations and to reflect broadly the person's cumulative show excellent regression with radiation dose for exposure to clastogens (Kleinerman et a1., 1989). upwards of four decades after exposure (Awa, 1974; Much information on the behaviour of these Kleinerman et al., 1989). Thus, stable aberrations persistent staЫе types of chromosome aberrations appear to be sensitive retrospective biomarkers has been gleaned from long-term follow-up studies of exposure levels for populations as a whole. of hundreds of persons accidentally exposed to However, because considerable variability has been ionizing radiation, or exposed during various med- observed between persons within dose groups, ical procedures. Evaluations using classical methods there are questions regarding the precision of sta- and painting techniques have demonstrated that ble aberrations iп lymphocytes as retrospective the majority of stable aberrations observed in Iym- dosimeters for individuals studied many years after phocytes are reciprocal or non-reciprocal translo- exposure (Kleinerman et al., 1989). As discussed, cations (e.g. Ishihara & Kumatoтi, 1969; Awa, clonal expansion of subsets of memory T-cells in 1974; Littlefield & )oiner, 1978; Tucker et al., 1994; response to random antigenic challenge would Ramsey et al., 1995). In studies of relatively large tend to increase interperson variability in the cohorts of non-exposed persons, translocation fre- observed frequencies of cells with radiation- quencies observed in cultured lymphocytes are induced translocations within dose groups with several times higher arid considerably more vari- increasing time after a specific exposure. able than dicentric frequencies. Data from classical analyses have shown that mean frequencies of Translocations as predictors of risk of late effects stable aberrations in lymphocytes are higher in As early as 1962, consideration was given to the groups of older women (Littlefield et aL, 1991), and possible relationship between persistent chromo- recent data from painting studies have shown a some rearrangements in lymphocytes and the risk strong age effect in controls (Tucker et al., 1994; of late effects (Buckton et al., 1962). Recently, it has Ramsey et al., 1995). The fact that variable back- been postulated that the burden of stable chromo- ground frequencies are observed in normal controls somal rearrangements in lymphocytes may be tue- is a confounder that must be taken into consider- fui as a surrogate biomarker that gauges the relative ation when attempting to gauge the sensitivity of Ievels of cumulative spontaneous or agent-induced translocation analyses as an index of exposure. genetic damage sustained by the haematopoietic For irradiated populations, increased frequen- stem cell compartment (Littlefield et al., 1991). cies of lymphocytes with stable aberrations are Because chromosomal rearrangements are thought observed for upwards of 40 or more years after to be relevant to, and play a role in, carcinogenesis exposure, and for any individual, the numbers of (e.g. Heir & Mitelman, 1987; Sandberg, 1990; lymphocytes bearing radiation-induced stable Cleary, 1991; stanbridge, 1992), it can be reasoned aberrations remain relatively constant over time that the frequencies of stable aberrations in (e.g. Buckton etaL, 1962; Awa, 1974). Infrequently, somatic cells would be positively correlated with three or more metaphase cells with apparently the risk for cancer in exposed populations. identical translocations or insertions have been A major prospective study by the Nordic 5tuду observed in individuals in several study cohorts, Group is underway, the purpose of which is to including A-bomb survivors, ankylosing spondy- gather information on the correlation between lytis patients, Y-12 accideд t survivors, persons various cytogenetic end-points and future cancer exposed at Chernobyl (Salassidis et al., 1995), as risk (Brogger et al., 1990; Hagmar et al., 1994). well as in controls (Tucker et al., 1994). These Cytogenetic studies in this cohort were conducted

190 Cytogenehc end-points as biological dosimeters and predictors of risk between 1970 and 1988 and aberrations include quence of at least one, if not two, genes. Assuming chromatid as well as chromosome breaks and the average gene is -20--25 kb long, including exchanges. Eighty-five cancers were diagnosed in litrons, and assuming 100 000 genes and 3.3 x 109 the Nordic Study Group cohort during the period base pairs/haploid genome, it is obvious that no 1970-1991. In the second follow-up report (Hagmar more than -25-40% of the genome occupies the et ai., 1994), no significant trend in the standard- spaces between the genes. Although a small frac- ized incidence ratio was observed for either sister tion of all genes appears to be important for chromatid exchanges or micronuclei, but there tumorigenesis, DNA-damaging agents that increase was a statistically significant linear trend for chro- the total burden of translocations in somatic cells mosomal aberrations with regard to cancer risk. would have a finite probability of inducing genetic Similar results have been seen by Bonassi et al. alterations at sites that are relevant in carcinogen- (1995). esis. Chromosome painting techniques will now In another study, attempts weré made to derive allow for large-scale cytogenetic epidemiology correlations between stable radiation-induced studies of populations known to be at risk for chromosome aberrations in lymphocytes and can- cancer. Such studies are needed and are likely to cer risk in women who received high-dose, local- produce important information regarding the ized radiotherapy for cervical cancer or low-dose association between induced chromosome aberra- exposures to treat benign gynaecological disease tions in somatic cells and subsequent cancer risk. (Kleinerman et al., 1994). Leukaemia risk is nearly the same in both groups, despite a 10-fold differ- Micronuclej ence in average dose to their bone marrow. The The existence of micronuclei has been known for cervical cancer patients received fractionated, many years, and they are frequently used to high-dose-rate radiotherapy, with average accumu- quantify exposure to chemicals and radiation. lated doses of up to 40 Gy to the whole pelvis, and Micronuclei can be observed in almost any cell it has been postulated that their lower excess type, and for this reason many variations of the leukaemia risk per unit dose may be attributed to assay exist. Micronuclei can be evaluated in cell cell killing in the high-dose fields, with wasted types that are not amenable to the analysis of radiation effect. Classical studies using group structural aberrations, although the most common analyses conducted nearly two decades after expo- cells examined do belong to the haematopoietic sure demonstrated that the rate of stable translo- system, including lymphocytes and erythrocytes. cations in lymphocytes was only slightly higher One caveat, however, is noteworthy. In humans, among the cervical cancer patients. Assuming that erythrocyte micronuclei persist only iп splenec- the fraction of cytogenetica11y aberrant stem cells tomized individuals because the spleen efficiently that survive radiotherapy contribute to the leukae- filters out erythrocytes containing micronuclei. mogenic process, then the cytogenetic findings are Nevertheless, sufficient numbers of splenecto- consistent with epidemiological findings of com- mized people exist, at least in countries where sur- parable overall leukaemia risks seen in these two gical treatment of traumatic injuries is common, so irradiated populations. These data further suggest that some limited studies involving these people that in instances of high-dose localized exposures, have been performed (e.g. Schreunemachеrs & persistent translocations in lymphocytes appear to Everson, 1991). Other cell types that can be exam- serve as biomarkers of effective risk as well as bio- ined in humans are urothehal cells and exfoliated markers of dose. cells from the buccal and nasal mucosa. In humans, Cytogenetic evaluations in each of these studies involving bone marrow and other interпaI cohorts were undertaken prior to the introduction tissues are technically possible, but because their of painting techniques, and thus a large data bаsе acquisition requires invasive procedures they are has not yet been accumulated that would allow not commonly employed. Peripheral lymphocytes prospective studies to determine the correlation are a commonly examined tissue, and may be between the burden of translocations in somatic evaluated for micronuclei with or without cultur- cells and the risk of late effects. It is likely that most ing. If the cells are cultured, cytochalasin B maybe translocations will damage or interrupt the se- used to inhibit cytokinests, and mlcronuclel are

191 Application of Bbmarkers in Cancer Epidemiology scored in binucleated cells. 1f care is taken to evalu- more often than expected by chance (Hando et al., ate only those cells with intact membranes, the 1994; Nath etaL, 1995). investigator can be certain that all the products of a mitosis are present, which has important impli- Micronuclei and cancer risk assessment. cations for studies where the mechanism of The relationship between micronuclei and adverse micronucleus formation is of interest. health risks is not as well substantiated as it is for 5eves1 mechanisms of action contribute to the structural aberrations. The primary reason is that formation of mkroasudgii (1-leddle et al., 1983), micronuclei are formed by moie than one mecha- including chromosome breakage (clastogenesis) nism and unless the contents of each micronu- and spindle disruption (aneuploidogenesis). Until cleus are known (especially with respect to the a few years ago, cytological evaluation of micronu- presence or absence of centrorneres), the data may ciel provided little or no information concerning be subject to errors of interpretation. Furthermore, the type of damage that led to their formation. even when using micronuclei as a biomarker to There was considerable debate concerning whether evaluate exposure to an agent with a known mech- the size of a micronucleus could be used to elicit anism, a method should be employed that enables mechanisms. The basic premise was that larger discrimination between classes of micronuclei. micronuclei were caused by spindle disruption and There are several reasons for this. First, few agents would contain whole chromosomes, while smaller have a single mechanism of action. Even known micronuclei would consist of one or more chro- spindle disrupters appear to have some clastogenic mosome fragments. While the relationship be- activity, and radiation induces a significant num- tween micronucleus size and its mechanism of ber of kinetochore-positive micronuclei (Eastmond origin may be generally true, we now know that & Tucker, 1989). Thus, understanding the types there is a significant amount of overlap between and distribution of micronuclei will provide a these classes of micronuclei, with the result that it moie full description of the mechanism(s) of is impossible to know for certain how a single action of the compound in question. For agents micronucleus arose. with unknown or mixed mechanisms of action, The solution to this problem has been to use evaluating the type(s) of micronuclei induced can molecular methods to identify the contents of lead to important insights into risks resulting from micronuclei. These methods consist of two general exposure. The second reason to use centromere approaches: anti-kinetochore antibodies and DNA detection methods is that spontaneous micrопu- probes. Anti-kinetochore antibodies bind to the clei arise by a variety of mechanisms and, as site of spindle fibre attachment to the chromo- already mentioned, the contents of micronuclei somes and thus serve as a marker for the сеn- are not a random representation of the genome. tromere. Micronuclei with a kinetochore signal Understanding the types of micronuclei in the can reasonably be assumed to contain one or more unexposed cells will enable a more refined statisti- whole chromosomes (the number of which can cal determination of exposure. For example, for often be determined) and are indicative of processes analyses limited to clastogenic processes, only that disrupt the mitotic spindle. Kinetochore- those micronuclei iп the unexposed (control) negative micronuclei are indicative of clastogenic group that lack a centromere need to be consid- processes (e.g. Eastmond & Tucker, 1989). A con- ered when determining whether the compound sensus sequence repetitive pan-centromenic DNA induced micronuclei containing chromosome probe has been used to make similar determina- fragments. This will have the effect of increasing tions (e.g. Titenko-Iiolland et al., 1994). Unique the statistical power of the analysis. sequence DNA probes aid repetitive probes for a To date, only a few studies have used molecular single chromosome type can also be used to inves- approaches to label centromeres of micronuclei in tigate the contents of micronuclei. For example, human populations (Hando et al., 1994; Titenko- several studies have shown that the frequency of Holland et aL, 1994; Catalan et aI., 1995; Nath et chromosomes in nnicronuclei is non-random. In aI., 1995) and none of the subjects had received particular, the X and Y chromosomes appear in an occupational or environmental exposure. The micronuclei of females and males, respectively, far reasons may simply be that these methods require

192 Cylogenetic end-points as biaiogical dosimeters and predictors of risk additional effort and that not every cytogenetics pounds (Tucker et aL, 1993a). SCEs are generally laboratory is equipped for the molecular aspects more sensitive indicators of genotoxic effects than of the work. In spite of these limitations, the are structural aberrations, and require less effort to advantages of including molecular methods in analyse. Like structural aberrations, the existence of micronucleus analyses are significant and would threshold effects for SCE induction is possible but substantially improve the ability of micronucleus difficult to prove or disprove, especially without an assays to serve as a meaningful biomarker. As more understanding of the mechanism of formation. micronucleus data including cerstromere informa- A variation of the basic method involves quan- tion are gathered, so our understanding of the use- tifying cells with a high frequency of SCEs (Moore fulness of the data for cancer epidemiology will & Carrano, 1984). In this approach, the distribu- improve. Iп the meantime, the analysis of micro- tion of SCEs per cell is determined for each mdi- nucleus data can be performed in a manner which vidual, and cells exceeding an arbitrarily chosen assumes that all micronuclei are formed either by number of exchanges (typically those in the upper clastogenic or by spindle disruptive processes. 5a/o of the range) are identified as 'high frequency Although this will lead to an overestimation of the cells' (HFCs). The number of HFCs per subject is risk estimate, at least an upper confidence bound then compared between groups of individuals could be obtained. In spite of some limitations, using non-parametric statistics. This method may micronuclei can be very useful as a biomarker. be useful for detecting differences between individuals or groups of people when differences in the Sister chromatid exchanges (SCEs) mean frequencies are small. SCEs can be measured in any eukaryotic cell that can be grown in bromodeoxyuridine (or a suitable Interphase cytogenetic analyses alternative) for two cell cycles and examined in Interphase cells are routinely obtained during the metaphase. SCEs have been evaluated in numer- collection of clinical specimens such as Pap smears, blood and urine samples, and skin biopsies. These оus studies involving human exposure. The assay yields quantifiable data from every metaphase cell, clinical samples are evaluated for normal and and the genotoxic potential of a chemical can be pathological features and provide valuable infor- determined more rapidly than with some other mation for clinical diagnosis. However, historically cytogenetic methods (SCEs are not a good indica- these samples have had limited usefulness in iden- tir of exposure to ionizing radiation). Besides the tifying genetic changes occurring in the tissues, disadvantage of an unknown mechanism, the use organs or body fluids from which the sample was of SCEs has been criticized because significant obtained. Developments in molecular cytogenetics increases have been obtained from compounds over the past 10 years are allowing new and impor- such as NaCI and KCl which clearly are not tant types of cytogenetic information to be ob- expected to be genotoxic (Galloway et al., 1987). tained for cells and cell types that has not been Thus, a positive SCE response does not necessarily possible previously. One type of molecular tech- mean that a compound is genotoxic. However, SCE nique which is proving to be particularly valuable assays remain widely used for evaluating the abil- for these studies is FISH, which allows information ity of chemicals to induce genetic damage. Unlike on chromosome number and, to a limited extent, some types of structural aberrations (especially re- chromosome structure to be simply and rapidly ciprocal translocations), SСEs evaluated following obtained from interphase cells (Pinkel, etal., 1986; in-vivo exposure show limited persistence and Gray, et al., 1994; Eastmond & Rupa, 1995). accumulation, which appears to reflect DNA repair Interphase studies using FISH focus primarily on processes as well as cell turnover. aneuploidy, a condition in which the chromosome Nevertheless, SCE data do have some appealing number of the cell or individual differs from the characteristics. The distribution of SCEs per cell is normal number in that cell type—typically 46 for close enough to normal to allow parametric statis- somatic cells aid 23 for germ cells in humans. tical analyses in most situations. Large SCE data- Aneup1oidy in germ cells has been associated with bases have been generated, making it possible to infertility pregnancy loss, congenital anomalies and compare relative potencies for hundreds of cor- mental retardation, whereas in somatic cells this

193 Application of Biomarkers in Cancer Epidemiology

condition has been linked with cell death and car- some loss is influenced by the superimposition of cinogenesis (Hook, 1985; Oshimura & Barrett, 1986; signals, associations of сеntromeric regions in Hecht Sr Hecht, 1987). FISH studies for arieuploidy interphase cells and inefficient probe penetration. typically utilize DNA probes specific for the cen- It is also Iikely that other cellular and hybridization trorneric region of one (or occasionally several) phenomena can significantly affect the number of chromosome(s) of interest (Eastniond & Rupa, cells scored as hyperdiploid. For example, breakage 1995). Following hybridization with the probe, occurring within the chromosomal region targeted two brightly stained fluorescent hybridization by the DNA probe can be incorrectly identified as regions are observed at the position of the chro- an additional chromosome (Eastmond, et al., 1994; mosomal region in both metaphase and interphase Rupa, et aL, 1995). Although this illustrates a poten- somatic cells. Aneuploid cells are identified by tial imitation of this FISH assay, D.A. Eastmond counting the number of fluorescent spots in each and co-workers have shown that, by modifying the nucleus. Cells are considered to be aneuploid when standard single probe hybridization assay through the number of hybridization signals differs from the use of multicolour FISH with two adjacent the expected two per chromosome in the nuclei of probes, this approach can also be used to detect somatic cells and one in the nuclei of germ cells. chromosomal breakage in interphase somatic and For technical reasons, this assay is more sensitive germ cells (Eastmond et aL, 1994; Rupa, et al. 1995; for detecting gains in chromosome number D.S. Rupa, unpublished data). Recent studies have (hyperploidy) than for detecting chromosome shown that this tandem FISH approach is effective losses (hypoploidy) (Eastmond & Pinkel, 1990). for detecting breakage in cells exposed in vitro to In-vitro studies have shown that this FISH assay genotoxic agents, as well as in workers occupa- 1s effective in detecting hyperdiploidy induced by tionally exposed to pesticides (Rupa, et al., 1995). aneuploidy-inducing agents (Eastmond & P1nke1, One of the advantages of these interphase FISH 1990). To date, only a limited number of studies assays is that they can be performed on most cell using FISH to detect aneuploidy in human pори- types, allowing information on chromosomal lations have been conducted. Some initial studies, alterations to be obtained from the target organs of primarily in sperm and exfoliated cells, have used carcinogens or germ-cell-damaging agents. For probes in an attempt to establish the levels of example, by assaying cells exfoliated in the urine, hyperploidy in the cells of individuals without it may be possible to measure damage occurring in known exposure to genotoxic agents (e.g. Moore et the bladder of workers or patients exposed to al., 1993; Robbins etaL, 1993; Bischoff et al., 1994). agents that induce bladder cancer. Studies have These have been followed by a number of studies been initiated that have shown that interphase (e.g. Robbins et aL, 1994; Rupa et al., 1995; Smith FISH can be performed on buccal mucosal cells, et al., 1995) in which FISH, using centromeric DNA urothelial ceis and sperm, as well as blood cells probes, has been employed to detect increases in such as granulocytes and lymphocytes. In addition, aneuploidy in individuals exposed to benzene, pes- it is Iikely that other cells isolated through skin ticides and chemotherapeutic drugs. In most of biopsy, and cells isolated during bronchoalveolar these recently published or soon-to-be published lung lavages or Pap smears could also be assayed. studies, a significant increase in hyperploidy was For most of the cell types tested to date, relatively seen in the exposed individuals when compared simple preparation procedures can be used. For with control individuals. Although additional example, standard cytogenetic fixation procedures studies will be required to validate this assay thor- can be used to prepare lymphocyte and granulo- oughly, the results of these initial studies indicate cyte slides. Interphase analyses for aneuploidy and the promise and feasibility of this technique in breakage can also be performed on blood smears human biomonitoring. following simple fixation procedures. However, Since aneuploid cells are identified by counting this approach is not recommended as a routine the number of hybridization regions in the nucleus, method, as relatively few nucleated cells are re- hybridization artefacts and other technical prob- covered on a standard blood smear and scoring is lems may significantly affect this assay (Eastmond tedious. It is likely that with relatively minor mo- etaL, 1995). For example, the detection of chromo- difications, such as the use of hypotonic Rd to lyse

194 Cytogerietic end-points as biological dosimeters and predictors of risk the red blood cells in the blood sample, nucleated detection of bound probe, (2) visualization of probes cells can be concentrated to facilitate scoring as hybridized to smaller targets, (3) reduction of signal well as to improve probe penetration. This should fading, and (4) reduced need for computerized allow biomonitoring to be performed on minimal image processing, which is expensive and beyond amounts of blood, such as that obtained by prick- what most laboratories can afford. ing a finger, and would eliminate the requirement The field of interphase cytogenetics has seen for cell culturing. However, in some cases cell cul- significant advances in recent years, also due to ture may facilitate the detection of chromosome applications of FISH. The detection of structural aberrations such as aneuploidy, as cells must pass aberrations in interphase with closely spaced through a mitosis in order for numerical alter- probes is significant. Extension of this basic ations to occur. Recent studies by Rupa etaL (1996) method to other chromosomes and in other have shown that culturing is not required for the species could lead to significant insights into the detection of chromosomal breakage in interphase frequency and distribution of chromosome dam- cells. Iп these studies, similar frequencies of breakage in cell types that are not currently accessible. age were seen in cultured and non-cultured lym- Such probes would not necessarily have to be tai- phocytes, and in granulocytes irradiated to vtttu. geted to repetitive sequences, as it is now possible to produce unique-sequence probes to specific Future cytogenetic techniques chromosomal regions by microdissection (e.g., The use of biomarkers for assessing cytogenetic Yokoyama & Sakuragawa, 1995). damage has seen tremendous advances in recent It is also possible to perform FISH on tissue slices years, the most notable of which is FISH. Early of archived pathological specimens (Sauter et al. efforts were limited to a single colour and repeti- 1995). Application of these methods to epidemio- tive sequences, while the current state of the art logical applications would create new avenues of involves multiple colours used in combination to research in which samples gathered many years identify a wide range of targets comprised of previously would become amenable to sensitive and unique and repetitive DNA sequences. The sensi- ghly specific molecular analyses. Retrospective tivity and specificity of the various F151 assays case-control studies could be performed with rela- have increased accordingly, but there is still a need tive ease on existing samples. Prospective cohort for additional fluorochromes in the visible range. and nested case-control studies could be planned Ideally, at least 24 unique colours are needed, one because the sample storage requirements are for each of the human chromosome types. The straightforward. The relatively small amounts of actual number of fluотoсhromes needed to accom- material that are typically required for molecular phsh this is considerably less than 24. For example, analyses could mean that numerous assays might two colours are now commonly mixed to yield a be performed on only a few tissue slices, which third colour. Three fluorochromes can be mixed to would be a distinct advantage for very small or rare yield seven colours (three individually, plus three samples. unique pairs, plus all three together). In general, n Automation of cyto genetic assays has been fluorochromes can theoretically be mixed to yield a goal for many years. Some progress has been Zn - 1 colours, assuming that the fluorochromes made, for example in the area of karyotyping are mixed in equal ratios. Alterations of the ratios and aberration detection (e.g. Luudsteen & Piper, will lead to additional colours. As many as 12 1989; Mayali et al., 1990; Huber et aI. 1995). colours have now been detected simultaneously However, fully automatic analysis of metaphase (Dauwerse et al., 1992), but the use of this number chromosomes is yet to be realized. Automatic of colours simultaneously is far from routine. metaphase finding shows significant promise for Regular application of large numbers of probes in enhancing throughput, and consequently for different colours would significantly increase the increasing the sensitivity of aberration detection frequency of the detectable events, including struc- (e.g. Piper et al., 1994; Vrolijk et al., 1994). Some tural aberrations, chromosomal contents of progress has also been made towards automatic micrornudei and aneuploidy. Similarly, the applica- detection of micronuclei in erythrocytes (e.g.. tion of brighter colours would enable (1) enhanced Zetterberg & Grawe, 1993) and nucleated cells

195 Application of Biomarkers in Cancer Epidemiology

(e.g., Vrai et al., 1994), but except for isolated Special considerations on the use of cytogenetic instances, such methodology is not used on a rou- end-points in epidemiology studies tine basis. Automated systems for spot counting in Epidemiology studies involving cytogenetic end- interphase cells would be useful for detection of points have generally required advanced planning aneuploidy. In general, automation is expected to because of the need to culture cells to bring them permit the analysis of more cells than would oth- to metaphase. This is true for studies seeking to erwise be possible, with the result that the sensi- measure structural chromosome aberrations, tivity for performing biological dosizn.etry will be whether by FISH or by classical methods, as well as increased. for studies involving SCEs. This limitation severely Improved techniques for hybridization are restricts the use of archived material for these end- needed, particularly for slides that have not points. A further limitation for studies using FISH been stored in NZ gas at-20°C. While some success is the requirement for microscope slides to be at hybridizing archived material has been frozen in nitrogen gas to preserve their ability to achieved, routine F1SH on old material is often hybridize. The reasons for this are not well under- difficult. A method that consistently results in stood and might be overcome by additional high-quality FISH signals would eliminate the research. An alternative approach to freezing slides need for side storage at —20°C, and would open is to freeze cells in a manner that maintains their up new avenues of research on existing material. viability. With this approach, selected samples For prospective epidemiology studies, cryopreser- can be thawed and analysed at a later date. vation of white blood cells offers an attractive Disadvantages of this method include the expense alternative to real-time biomonitoring of at-risk involved in cryopreservation and diminished cell populations, since ceI1s can be stored long-term viability upon thawing. For all of these reasons, and evaluated as needed in the future. When metaphase-based analyses are most commonly frozen for viabi1ity, lymphocytes may be readily used for case—control or cross-sectional studies, and thawed and cultured (McFee et al., 1995) for stud- not for prospective cohort or retrospective studies. ies of structural chromosome aberrations, The advent of interphase-based analyses as micronuclei or SCEs. After culture, slides may be described in this paper appears to have significant prepared for FISH immediately or cells may be potential for many types of epidemiology studies. stored in fixative at —20°C for periods of at least a Prospective cohort and nested case—control studies year. Similarly, interphase analyses using сеп- could take advantage of these methods for several trоmere, chromosome or DNA probes can be per- reasons. The absence of a requirement for tissue formed on lymphocytes or granulocytes from culture greatly simplifies the work needed to store selected persons within cohorts. Such approaches samples, although freezing of the microscope may be useful in case—control or casecohort stud- slides is still required for FISH. Interphase analyses ies of populations with specific exposures at some might be possible on lymphocytes obtained from time in the past. gravity sedimentation of whole blood cells ('buffy One goal of all these methods is to improve the coats'). These cells could be smeared on micro- sensitivity and specificity of cytogenetic analyses, scope slides and stored for later analysis. The sim- so that their utility as biomarkers of exposure and plicity of this method, the inherent low cost aid effect can be improved. Developments such as new the small amount of material required would make DNA probes and probe combinations, enhance- this approach applicable to a broad range of studies, ments in the area of automation and the hybridiza- including those that require very large sample sizes. tion of archived slides, as well as improvements iп Finally, a comment on inter-reader variation of the number and intensity of fluorochromes, are all cytogenetic analyses is in order. Epidemiologists expected to increase the rate of analysis and to often express concern about the inherent subjec- improve the sensitivity and specificity of detec- tivity of cytogenetic analyses. Any method that tion. The result will be that lower levels of expo- requires human judgment as part of the analytical sure can be detected with increased confidence and process has some potential for confounding due to reliability in populations that could not previously interindividual differences in decision-making. be evaluated. One of the significant advantages of molecular

196 Cytogeneic end-points as biological dosimeters and predictors of risk cytogenetic methods, compared to the older, clas- Genet. Суtоgeпеt., 79, 133-135 sical methods, is that the end-points of interest are Brandriff, B.F., Gordon, L,A., Moore, D., II & Carrano, more easily detected and therefore less subjective A.V. (1988) An analysis of structural aberrations in in their interpretation. For example, a colour'junc- human sperm chromosomes. Cyto,Fenet. Cell Genet., 47, tion' which is evidence of a chromosome translo- 29-36 cation in painted cells is much more obvious than Brogger, A., Hagmar, L., Hansteen, I.L., Heim, S., the equivalent event seen in banded preparations. Нбgstedt, B., Knudsen, L., Lambert, B., Kinnainmaa, K., This increase in objectivity should diminish con- Mitelman, F. Nordenson, I., Renterwall, C., salomaa, S., cerns inherent in any large study where there is 5kerfving, S. & Sorsa, M. (1990) An inter-Nordic prospective study on cytogenetic endpoints and cancer risk. more than one person who performs the routine Cancer Genet. CytogeneL, 45, 85-92 analyses and where unavoidable turnover in the technical staff occurs. Buckton, ICE. (1976) Identification with G and R bапд- In summary, compared with classical cytoge- iпg of the position of breakage points induced in human chromosomes by in vitro X-irradiation. Int. J. Radiat. netic approaches, molecular cytogenetic biomark- BioL, 29, 475-488 ers offer a number of significant advantages that make them appealing for use in a wide variety of Buckton, ICE., Jacobs, P.A., Court-Brown, W.M. & Doll, R. (1962) A study of the chromosome damage persisting epidemiological studies. after X-ray therapy for ankylosing sgondylitis. Lancet II, 676-682 Acknowledgements This work was performed in part under the aus- Catalan, J., Autio, K., Wessman, M., Lindholm, C., pices of the US DOE by the LLNL under contract Kiiuutila, S., Sorsa, M. & Norppa, H. (1995) Age-associated micronuclei containing centromeies and the X No. W-7405-ENG-48 and supported by NIH grant chromosome in lymphocytes of women. Cytogenet. Cell CA59431 ()DT). LGL was supported by NCI intera- Genet., 68, 11-6 gency agreement 01- 3-0573 and DOE/ORAU У СР Cleary, M.L. (1991) Oncogenic conversion of transcrip- contract DE- CO5-760800033. DAE was sup- А tion factors by chromosomal translocations. Cell, 66, ported by a US EPA grant (R 820994-01-1). 619-622 .K., Breuning, l'AI. & References Dauwerse, J.G., Wiegant, J., Raap, А Awa, A.A. (1974) Cytogenetic and oncogenic effects of Van Omen, G.J.B. (1992) Multiple colorsby fluorescence the ionizing radiations of the atomic bombs. In: in situ hybridization using ratio-labelled DNA probes create a molecular karyotype. Human Mol. Genet., 1, German, J., ed., Chromosomes and Cancer, New York, John 593-598 Wiley, pp. 637-674 Eastmond, D.A. & Tucker, J.D. (1989) Identification of Bender, M.A., Awa, A.A., Brooks, A.L., Evans, Н.J., Groer, P.G., Littlefield, L.G., Pereira, C., Preston, R.J. & aneuplliidy-inducing agents using tytolwesis-blocked human lymphocytes and an aritikinetochore antibody. Wachholz, B.W. (1988) Current status of cytogenetic procedures to detect and quantify previous exposures to Environ. Mol. Mutagen., 13, 34-43 radiation. Mutation Res., 196, 103-159 Eastmond, D.A. & Pirtkel, D. (1990) Detection of anen- ploidy and aneuploidy-inducing agents in human lym- Bischoff, F.Z., Nguyen, D.D., Burt, К,J. & Shaffer, L.G. (1994) Estimates of aneuploidy using multicolor fluores- phocytes using fluorescence in situ hydridization with chromosome-specific DNA probes. Mutation Res., 234, cence in situ hybridization on human sperm. Cytogenett. 303-318 Cell Genet., 66, 237-243 Bogen, K.T. (1993) Reassessment of human peripheral T- Eastmond, D.A., Rupa, D.S. & Hasegawa, L.S. (1994) lymphocyte lifespan deduced from cytogenetic and cyto- Detection of hyperdrploidy and chromosome breakage human lymphocytes following exposure to toxic effects of radiation. Int. J. Radiat. Biol., 64, 195-204 in interphase the benzene metabolite hydroquinone using multicolor Bonassi, S., Abbondandolo, A., Camurri, L., Dal Pra, L, fluorescence in situ hybridization with DNA probes. De Ferrari, M., Degrassi, F., Forni, A., Lamberti, L., Lando, Mutation Res., 322, 9-20 C., Padovani, P., Sbrano, I., Vecchio, D. & Pvntoni, R. (1995) Are chromosome aberrations in circulating iym- Eastmond, D.A. & Rupa, D.S. (1995) Fluorescence in situ - phocytes predictive of future cancer onset in humans? hybridization: application to environmental niutagene Environmental Preliminary results of an Italian cohort study. Cancer sis. In: Phillips, D.H. & Venitt, S., eds,

197 Application of Biomarkers in Cancer Epidemiology

Mutagenesis, Bios Scientific Publications, Oxford, pp. Machado, S.G., Bfettner, M., Peters, L.J. & Boice, J.D., Jr 261-290 (1989) Chromosome aberrations in peripheral lymphocytes and radiation dose to active bone marrow in Rastmond, D.А., Schulег, M. & Rupa, D.S. (1995) Advantages and limitations of using fluorescence in situ patients treated for cancer of the cervix. Radiation Res., hybridization for the detection of aneuploidy in inter- 119, 176-190 phase human cells. Mutation Res., 348, 153-162 Kieinerman, R.A., Littlefield, L.G. Tanne, R.E., Sayer, Galloway, S., Deasy, D. Bean, C., Kraynak, R., Armstrong, A.M., Cooktair, D.L., Wactawski-Wende, J., Iraskip, P.D., M. & Bradley, O. (1987) Effects of high osmotic strength Block, A,M.W., Ramesh, K.H. & Boice, J.D., Jr (1994) on duomosome abenations, sister chromatid exchanges Chromosome aberrations in lymphocytes from women and DNA strand breaks, and the relation to toxicity. irradiated for benign and malignant gynecological dis- Mutaf ion Res., 189, 15-25 ease. Radiation Res., 139, 40-46 Gray, J.W., Pinke1, D. & Brown, J.M. (1994) Fluorescence Littlefield, L.G. & Joiner, E.E. (1978) Cytogenetic follow-up in situ hybridization iii cancer and radiation biology. studies in six radiation accident victims (16 & 17 years Radial Res., 137, 275-89 post-exposure). In: Late Effects of Ionizing Radiation, Vienna, International Atomic Energy Agency, pp. 297-308 Hagmar, L., Btogger, L., Hansteen, I.L., Heim, S., Hogstedt, B,, Knudsen, L., Lambert, B., Linnainmaa, K., Littlefield, L.G., Kleinerman, R.А., Sayer, A.M., Tamone, R. & Boice, J.D., Jr (1991) Chromosome aberrations in lym- MiteImaп, E, Nordenson, i., Reuterwall, C., Salomaa, S., Skerfving, S. & Sorsa, M. (1994) Cancer risk in humans phocytes—Biomonitors of radiation exposure. In: predicted by increased levels of chromosomal aberrations Gledhill, B.L. & Mauro, F., eds, New Horizons in BiаIоgгсаl in lymphocytes: Nordic Study Group on the Health Risk Dosimetry, New York, Wiley-Liss, pp. 387-397 of Chromosome Damage. Cancer Res., 54, 2919-2922 Lucas, J.N., Tenzin, T., Straume, T,, Pinke1, D., Moore, D., II, Litt, M. & Gray, J.W. (1989) Rapid human chromo- Harido, J.C., Nath, J. & Tucker, J.D. (1994) Sex сhrото- somes, micronudei and aging in women. Chromosoma, some aberration analysis using fluorescence in situ 103, 186-192 hybridization. Jot. J. Radiat. Biol., 56, 35-44 (Erratum: Int. J. Radiat. Biol., 56, 201) Hecht, F. & Hecht, B.K. (1987) Aneuploidy in humans: dimensions, demography, and dangers of abnormal Lucas, J.N., Awa, A.A., Straume, T., Poggensee, M., numbers of chromosomes. In: Vig, . . & Sandberg, A.A., Kodama, Y,, Nakano, M., Ohtaki, K., Weier, 1.-U., Pinkel, ВК D., Gray, J. & Littlefield, L.G. (1992) Rapid translocalon eds, Aneuploidy Part А: Incidence and EtгoIо у , New York, Alan R. Liss, Inc pp. 9-49 frequency analysis in humans decades after exposure to ionizing radiation. lit. J. Radiat. BгоI., 62, 53-63 Нe l , J. ., Hite, M., Kirkham, B., Movournin, K., дд е А Lundsteen, C. & Piper, J., eds. (1989) Automation of MacGregor, J.T., Newell, G.W. & Salamone, М.F. (1983) The eduction of micronuclei as a measure of genotoxi- Cytogenetics, Berlin, Springer-Verlag city. A report of the U.S. Environmental Protection McFee, A.F., Sayer, A.M. & Littlefield, L.G. (1995) Agency Gene-Tox Program. Mutation Res., 123, 61-118 Cryopreservation of lymphocytes for cytogenetic analysis. I. Prompt separation is needed. Environ. Mol. Mutagen, Неim; S. & Mitelman, R (1987) Cancer Cytogeneiics, New York, Alan R. Liss 25 (suppl. 25), 34 Hook, E.B. (1985) The impact of aneuploidy upon pub- Mayali, B.H., Tucker, J.D., Christensen, M.L., van V1iet, lic health: mortality and morbidity associated with L.J. & Young, I.T. (1990) Experience with Athena, a semi- human chromosome abnormalities. In: Dellarco, V., automated karyotyping system. Cytometry, 11, 59-72 Voуtek, Р.E. & Hotlaender, A., eds, Aneuploidy, New York, Moore, D.H., II & Carrano, A.V. (1984) Statistical analy- Plenum Press, pp. 7-33 sis of high SCE frequency cells in human lymphocytes. Huber, R., Kulka, U., Lorch, T.H., Braselmann, H. & In: Tice, R.R. & Hollaender, A., eds, Basic Life Scтences Vol. Bauchger, M. (1995) Automated metaphase finding: an 29, New York, Plenum, pp. 469-479 assessment of the efficiency of the METAFER2 system ira Moore, L.E., Titenko-Holland, N. & Smith, М.Т. (1993) a routine mutagenicity assay. Mutation Res., 334, 97-102 Use of fluorescence in situ hybridization to detect chro- Ishihara, T. & Kumatont, T. (1969) Cytogenetic studies on mosome-specific changes in exfoliated human bladder and fishermen exposed to fallout radiation in 1954. Jpn. J. oral mucosa cells. Environ. Mol. Mutagen., 22, 130-137 Genet., 44 (Sпpp1. 1), 242-251 Natarajan, A.T., Vyas, R.C., Wiegant, J.C. & Curado, M.P. Kleinerrnan, R.A., Littlefield, L.G., Tanne, R.E., (1991) A cytogenetic follow-up study of the victims of a

198 Cytogenetic end-points as biological dosimeters and predictors of risk radiation accident in Goiania (Brazil). Mutation Res., 247, symmetrical translocations and the influeiice of clonai 103-111 aberrations for retrospective dose estimation. Int. J. Radiai. BioL, 68, 257-262 Nath, J., Tucker, J.D. & Hando, J.С. (1995) Y chromosome aneuploidy, micronuclei, kinetochores and aging Sandberg, A.A. (1990) The Chromosomes in Human Cancer in men. Chromosoma, 103, 725-731 and Leukemia, 2nd edn, New York, Elsevier

Ohtaki, K., Shimba, H., Awa, A.A. & Sofuni, T. (1982) Sauter, G., Mock, H., Gasser, T.C., Mihatsch, М.~. & Comparison of type and frequency of chromosome aber- Waldman, F.М. (1995) Heterogeneity of chromosome 17 rations by conventional and G-staining methods in Iii-o- and erbB-2 gene copy number fn primary and metastatic shima atomic bomb survivors. J. Radiat. Res., 23, 441-449 bladder cancer. Cytometry, 21, 40-46

Оsbimша, M. & Barrett, J.C. (1986) Chemically induced Schreunemachers, D.M. & Everson, R.B. (1991) Effect of aneuploidy in mammalian cells: mechanisms and biolog- residual splenic function and folate levels on the fre- ical significance fn cancer. Environ. Mutagen., 81, 129-159 quency of micronucleated red blood cells in splenec- tomized humans. Mutat. Res., 263, 63-67 Pinkel, D., Straume, T. & Gray, J.W. (1986) Cytogenetic analysis using quantitative, high-sensitivity, fluorescence Smith, М.T., Zhang, L., Rothman, N., Wang, Y., Hayes, hybridization. Froc. Natl Acad. sri., 83, 2934-2938 R.B., Li, G.L. & Yin, SN. (1995) Interphase cytogenetics of workers exposed to benzene. Toxicologist, Piper, J., Poggensee, M., Hill, W., Jensen, R., Ji, L., Poole, 15, 87 1., Stark, M. & Sudar, D. (1994) Automatic fluorescence Stanbridge, E.J. (1992) Functional evidence for human metaphase finder speeds translocation scoring in FISH tumour suppressor genes; chromosome and molecular painted chromosomes. Cytometry, 16, 17-16 genetic studies. Cancer Sm-v., 2, 5-24

Ramsey, M.J., Moore, D.H., II, Bnier, J.F., Lee, D.A., Titenko-Holland, N., Moore, L.E. & Smith, М.T. (1994) Olsen, L.A., Seлft, J.R. & Tucker, J.D. (1995) The effects of Measurement and characterization of micronuclei in age aid lifestyle factors on the accumuiation of cytoge- exfoliated human cells by fluorescence in situ hybrktiza- netic damage as measured by chromosome painting. tion with a centromeric probe. Mutat. Res., 312, 39-50 Mutat. Res., 338, 95-106 Tucker, J.D., Auletta, A., Cimino, M.C., Dearfield, K.L., Robbins, W.A., Segraves, R., Piiikel, D. & Wyrobek, A.J. Jacobson-Kram, D., Tice, R.R. & Carrano, A.V. (1993a) (1993) Detection of aneuploid human sperm by fluores- Sister-chromatid exchange: second report of the Gene- cence in situ hybridization: evidence for a donor differ- lox pTogram. Mutat. Res., 297, 101-180 ence in frequency of sperm disomic for chromosomes 1 Tucker, J.D., Ramsey, .~., Lee, D.A. & Minkler, J.L. and Y. Am. J. Hum. Genet., 52, 799-807 М (1993Ь) Validation of chromosome painting as a bio- Robbins, W.A., Cassel, M.J., Blakey, D.H., Meistrich, М.L. dosimeter in human peripheral lymphocytes following & Wyrobek, A.J. (1994) Induction of aneuploidy in the acute exposure to ionizing radiation in vitro. Inn. J. sperm of Hodgkin's disease patients treated with NOVD Radiat. Bio'., 64, 27-37 chemotherapy (detection by multi-chromosome fluores- Tucker, J.D., Lee, D.A., Ramsey, M.J., Briner, J., Olsen, L. cence in situ hybridization). Envi ron. MoL Mutagen., 23 & Moore, D.H., II. (1994) On the frequency of chromo- (suppi. 23), 58 somal exchanges iп a control population measured by Rupa, D.S., Hasegawa, L. & Eastmond, D.A. (1995) chromosome painting. Mutat. Res., 3131, 93-202 -1g12 Detection of chromosomal breakage in the lсеп Vrai, A., Verhaegen, F., Thierens, H. & de Ridder, L. (1994) interphase human lymphocytes using multi- region of The in vitro cytokinesis-block micronucleus assay: a color fluorescence in situ hybridization with tandem detailed description of an improved slide preparation DNA probes' Cancer Res., 55, 640-5 technique for the automated detection of micronuclei in Rupa, D.S., Schuler, M. & Eastmond, D.A. (1996) human lymphocytes. Mutagenesis, 9, 439-443 Detection of aneuploidy and breakage in the 1cen-1g12 Vrolijk, J., Sims, W.C., DanouB, E, Natarajan, А.T. & Tanke, region of human lymphocytes induced by genotoxic H.J. (1994) A system for fluorescence metaphase finding agents. Fundamental AppL ToxicoL, 30, 46 and scoring of chromosomal translocations visualized by Salassidis, K., Georgiadou-Schumacher, V., Braselmann, in situ hybridization. ut. J. Radia t. BioL, 66, 287-295 H., Muller, R, Peter, RU. & Bauchinger, M. (1995) Yokoyama, Y. & Sakuragawa, N. (1995) Improved simple Chromosome painting in healthy irradiated Chernobyl generation of GTG-band specific painting probes. victims: a follow-up study to evaluate the stability of Cytogenet. Cell Genet., 71, 32-36

199 Application of Biomarkers in Cancer Epidemiology

Zetterberg, G. & Grawe, J. (1993) Flow cytometric analysis of micronucleus induction in mouse erythrocytes by gamma-inadiation at very low dose-rates, lot. J. Radiat, iii!., 64, 555-564

Corresponding author J.U. Tucker Biology and Biotechnology Research Program, PO Box 808, L-452, Lawrence Livermore National Laboratory, Livermore, CA 94551, USA

200 Application ы Siomarkecs in Cancer Epidemiology Torüob, R, Выfеttа, Р. Shukec, D.E.G., Rothman, N., Halka, B. arid Pearce, N., ads IARC Scientific Publications No. 142 IпlerпatгaпaI Ageпcy foг Research cri Gercer, Lyle, 1997

Methodological issues in the use of tumour markers in cancer epidemiology i-F, Zhang, C. Cordon-Carda, N. Rothman, A.N. Freedman and J.A. Taylor

In this chapter, we review major methodological and practical issues associated with the use of tumour markers. At this stage of development, studies with a combination of tumour, susceptibility and exposure markers are needed to illustrate the link between exposure and biological response and to assess the interactive effects of tumour susceptibility markers in this process. Several practical issues related to the application of tumour markers are discussed, including banking of tumour tissue, setting a laboratory strategy and performing etiological heterogeneity analysis.

Recent advances in technology aimed at identify- carcinogenesis, such as tumour initiation and pro- ing genetic mutations and abnormally expressed motion, arid may be measured quantitatively or products have led to an interest in incorporating qualitatively by biochemical, immunochemical, these alterations as tumour markers in epidemio- cytogenetic and molecular techniques in human logical studies in an attempt to better understand biological materials, including tumour tissues, the natural history and the etiology of cancer. blood and urine samples, etc. These markers may There is a concomitant need for research into be employed to predict pnmary or secondary methodological issues including study design, sta- tumour risk, to establish tumour burden, to sub- tistical analysis, and interpretation as well as prac- classify tumours beyond histological classification tical issues such as tumour tissue storage. (in addition to pathological classification), to pre- Tumour markers have been defined by Schwartz dict tumour prognosis, to determine treatment (1989) as: strategy and to evaluate chemoprevection or inter- `... substances which can be measured quanti- vention efficacy. They can be employed in primary tatively by biochemical or immunochemfcal prevention or etiological research by detecting the means in tissue or body fluids to identify the relationship between environmental exposure and presence of a cancer, possibly the organ where it specific mutations; in secondary prevention studies resides, to establish the extent of tumour bur- (i.e. early detection and diagnosis) by identifying a den before treatment as well as to monitor the precursor or tumours at the early stage; and iп ter- response to therapy.' tiary prevention (i.e. enhanced prognosis) if they predict tumour progression and patient survival. The term has been defined more narrowly as In this chapter, we will briefly review some (Юavins, 1989): tumour markers of relevance to etiological studies of d. .. substances that are produced by cancer cells cancer and discuss methodological issues associated and present in the circulation of cancer patients with the use of those markers in cancer epidemiol- at a higher concentration than in individuals оgy, including issues of study design and statistical without cancer. It can be used to monitor the analysis. We will also discuss several tumour- course of the disease.' marker-related practical issues, including the collection and storage of tumour-marker-related biolog- More broadly, tumour markers can be defined as ical materials, the laboratory strategy for measuring those abnormal biological products or molecular tumour markers, and issues associated with the use alterations related to any sequence of multistage of tumour markers in prognostic studies.

201 Application of Biomarkers in Cancer Epidemiology

Etiological heterogeneity 1991). The study of the etiological heterogeneity of The theoretical framework for the use of tumour tumours at the molecular level may provide great markers in cancer epidemiology is based on the insight into the mechanisms and causal pathways hypothesis of etiological heterogeneity. The con- to carcinogenesis, which may lead to appropriate cept of etiological heterogeneity was first suggested preventive strategies to reduce the incidence of by Kreyberg (1937), and was based on the fact that cancer. a relationship might exist between histological types and biological or etiological factors. In epi- Tumour markers demiological studies of smoking and lung cancer, Tumour markers include all of the biological prod- Kreyberg (1952, 1954) suggested that lung cancer ucts related to the development and progression of was histologically and biologically heterogeneous neoplastic disease. Several markers are quantita- and related to particular exposures. The relation- tively measured and may indicate the burden or ships between stage, grade, histology, anatomic extent of the cancer, such as serum carcinoembry- location of the tumour and etiological factors have onic antigen (CEA), alphafetoprotein (AFP) and been explored for a number of different tumour prostate-specific antigen (PsA). Since those mark- types (Pall! et al. 1992; Sturgeon et al., 1994; ers have been discussed frequently in the literature Vaughan et al., 1995). The study of etiological het- (Chu, 1987; Sell, 1992; Kramer & 5гivastava, 1994), erogeneity using tumour markers represents a pro- we will not discuss them in this chapter. In the gression from studies of the relationship of envi- following section, we will briefly review three types ronmental exposures at the anatomical arid mor- of tumour markers with etiological implications: phological levels to studies at molecular and cytogenetic markers such as chromosome aberra- genetic levels. tions, oscogenes such as the ras family, and In perhaps one of the earliest studies of this tumour suppressor genes such as the ТР53 gene. kind to examine etiological heterogeneity at the molecular level, Taylor et al. (1992) conducted a Cytohenetic markers study of the relationship between occupational Since the use of cytogenetic markers has been exposures and the activation of the ras oncogene extensively reviewed by Tucker et al. (this volume), in the etiology of acute myeloid leukaemia (AML), we will limit our discussion to the cytogenetic employing a conventional control group identified analysis of lymphohaematopoietjc malignancies. by random digit dialling. In this study, in addition Cytogenetic markers in this context include chro- to calculating unadjusted and adjusted odds ratios mosome aberrations, sister chromatid exchange characterizing the association between occupa- and micronuclei, which can be measured in both tional factors and AIL, the investigators also tumour tissues and peripheral lymphocytes. Specific assessed the association between the occupational cytogenetic changes were found to be non-random risk factors and mutations of the ras gene in cases events involved in multistep carcinogenesis. Similar with AIL. Several recent studies have now cytogenetic changes were identified in second pri- expanded this paradigm by evaluating the etiolog- mary acute non-lymphocytic leukaemia (ANLL) ical heterogeneity of tumours, exploring the rela- after chemotherapy (Mitelman et al., 1981; Rowley, tionship between environmental exposures and 198 3) and in patients with occupational exposures ТР53 mutations(р53 ozerexpression in tumour (litelman et al., 1981; GoIomb et al., 1982; samples from a variety of tumour types, including Rowley, 1983; lite1inan et al., 1984). A high cor- lung cancer (Kondo et al., 1992; Suzuki et al., 1992; relation was observed between sites of cytogenetic Taylor et al., 1994), head and neck cancer (Field aberrations arid positions of oncogene or tumour et al., 1991), oesophageal cancer (Hollstein et al., suppressor gene Ioci (Rowley, 1984). 199 lа), bladder cancer (Spruck et al., 1993; Zhang etaL, 1994b), colorectal cancer (Zhang et al., 1995b; Environmental exposures, cytogeneic markers and Freedman et al., 1996a, 1996b), prostate cancer lymphohaematopoietic malignancies. Sandler and (Zhang et al., 1994a), stomach cancer (Zhang et al., Coilman (1987) have extensively reviewed issues 1995a), liver cancer (Bressac et al., 1991; Hsu et al., concerning cytogenetic and enviroлmentaI factors 1991; Ozturk 1991) and skin cancer (Brash et al., in the etiology of acute leukaemia. Sandler et al.

202 Methodological issues in the use of tumour markers

(1993) investigated the leukaemia risk associated myeloid cells. Loss of part or entire chromosomes with cigarette smoking and cytogenetic changes in 5 and/or 7 was noted in 23 of 25 patients with ane a multicentre case—control study of acute leukaemia ирlоiду. Loss of chromosome 5 was noted only in in adults. Smoking was found to be more common patients who previously had malignant lymphoma among patients with specific chroniosonje abnor- whereas loss of chromosome 7 was seen in these malities in acute myeloid leukaeniia (AIL) [-7 or patients as well as in those who had other malig- 7q—, —Y, +13] and in acute 1ymphocytic leukaemia nancies. Abnormalities of both chromosomes 5 (ALL) [t(9;22)(g34;q11)]. in a case—case study of 59 and 7 occurred in 53% of the patients treated with patients with newly diagnosed AIL (Fagioli et aI., combined therapy and in only 27% of patients 1992), 18 patients had prolonged contact with pes- treated with either modality alone. Cytogenetit ticides and seven patients had exposure to organic changes were analysed in 76 cases of secondary solvents. Cytogenetic studies confirmed the fre- myelodysplasia (sIDS) and acute non-lymphocytic quent occurrence of 5q and/or 7q aberrations in Ieukaemia (sANLL) (Johanson etal., 1991). Among patients with occupational exposure (10 out of 25 the 36 sМDS patients, 23 (64%) displayed clonal cases). These findings revealed that ACL in patients chromosomal abnormalities. The most common occupationally exposed to toxic substances might aberrations were —7, 5q—, —5 and +8. 0f the 40 sANLL represent a distinct cytogenetic entity. patients, 30 (75%) cases displayed clonai chromo- In a case—case study of 162 patients with ANLL somal abnormalities. The most frequent aberrations (Mitelman et al., 1981), 52 patients were occupa- were —7, —5, +8, 7q—, —17 and +21. When the inci- tionally exposed to chemical solvents, insecticides dences of characteristic cytogenetic abnormalities or petrol products, and 110 patients had no his- were correlated with the type of previous therapy, tory of occupational exposure to potential muta- —7 was found to be more frequent in sDS and genic оr carcinogenic agents. Clonai chromosomal BANLL patients who had been exposed to chemo- aberrations were present in 75% of exposed pa- therapy, whereas 5q—was associated with previous tierits, compared with only 32% in the unexposed exposure to ionizing radiation in sID5 patients. group. The incidence of certain characteristic kary- Those results suggest that cvtogenetic measures otypic abnormalities (5q—, 7q—, +8, +21, t(8;21), maybe employed to predict the risk of second pri- and t(9;22)) were decidedly more common in mary tumour after treatment such as radiotherapy exposed than in unexposed patients. In another and chemotherapy for the first primary tumour. case—case study of 74 patients with ANLL (Golomb In summary, many chromosomal abnormalities et a1., 1982), 25 of the 58 (43%) unexposed patients identified will have diagnostic, prognostic arid ther- had a clorial chromosome abnormality, compared apeutic implications. The identification of chro- with 12 of the 16(75%) exposed patients (P=0.02). mosomal abnormalities directs us to investigate Either a —5/5q— or a —7/7q— was present in 67% of abnormal loci of the genome that harbour the mol- the exposed patients with a chromosome abnoт- ecular basis responsible for malignant transforma- таlity compared with 28% of the aneuploid unex- tion and progression (IcClay, 1989). The imple- posed patients. These studies support the observa- mentation of interphase cyto genetics by tech- tion that a subset of patients with de nova ANLL have niques such as fluorescence in-situ hybridization a history of occupational exposure and a unique (FISH) will lead to the more frequent use of cytoge- pattern of clonal chromosomal aberrations. netic markers in cancer epidemiological studies.

Gytogenetic changes in patients with second acute non- Oncogenes and tи п our suppressor genes Eymphocytic leukaemia after treatment of a primary Inherited (germline) or acquired (somatic) gene malignancy. Cytogenetic studies were conducted mutations and altered gene products controlling on 26 patients who developed ANLL or a dys- cell growth, cell death and differentiation are con- myelopoietic syndrome after treatment of a ptimary sidered to be crucial steps iп human carcinogenesis. malignancy (Rowley et aL, 1981). Fifteen patients Molecular studies have defined that aberrations had radiotherapy and chemotherapy, seven had only affecting two major types of genes, proto-oncogenes chemotherapy, and four had only radiotherapy. and tumour suppressor genes have a direct role in Twenty-five patients had an abnormal karyotype in turnorigenesis and cancer progression.

203 Application of Biomarkers in Cancer Epidemiology

Oneogenes. Proto-oncogenes are normal cellular 1995), although it is important to point out that genes involved in a wide variety of functions, such mutational patterns differ for other smoking- as cell growth arid signal transduction. The act!- related cancers such as bladder cancer. vation of a proto-oncogene into an oncogene yields The wide range of involvement of ТР53 in a gain of function or dominant event. Oncogene human tumours and the broad spectrum of muta- activation usually occurs by a somatic mutation, tions make the gene a good candidate for molecu- mainly gene amplification or point mutation. These lar epidemiological studies (Holistein et al., 199 lb; alterations can convert a proto-oncogene from a Jones et a?., 1991; Harris, 1993; Harris & Hollstein, normal cellular gene to an oncogene and can lead 1993). ТР53 mutations have been suggested as to uncontrolled or neoplastic growth (Taylor, DNA fingerprints of exposure in a variety of tumours. 1989). . Dietary АFВ, exposures are associated with AGG-AGT mutations at codon 249 in liver cancer Tumour suppressor genes. These genes are also nor- (Ozturk et al., 1991; Bressac et al., 1991; Hsu et al., mal cellular genes. Tumour suppressor genes con- 1991). UV exposure may induce СС-ТT muta- tribute to oncogenicity through their loss of functions in skin tumour (Brash et al., 1991). Radon tions, and are considered recessive events. The end exposures are related to AGG-ATG mutations in result is that the products of these genes are absent codon 249 in lung cancer (Taylor et al., 1994). or inactivated in the malignant cells. Tumour sup- Newly developed molecular biological methods, pressor gene mutations are the most frequently e.g. polymerase chain reaction (PCR), automated observed genetic events in cancer. In general, sup- sequencing techniques and comparative genomic pressor gene inactivation occurs by a point mu- hybridization (CGH), will accelerate the process of tation of one allele and a deletion of the remain- characterizing DNA alterations. The CGH method ing contralateral allele. The loss of both alleles represents a recently developed molecular cytoge- (homozygous deletion) is an alternative but netic screening technique that can be employed to uncommon mechanism. These alterations result in survey entire genomes for variations in DNA an inability to suppress cell proliferation, and it is sequence copy number, as well as to map chromo- followed by tumour development. Sinсe many some regions with amplifications or deletions in mutagens are capable of altering tumour suppres- tumour DNA prepared from fresh or archived sor genes, it has been hypothesized that tumour materials (Kallioniemi et aL, 1993; Thompson suppressor genes contribute to the development of & Gray, 1993; Houldsworth & Chaganti, 1994; cancer and may be a critical area in which to study Kallioniemi et al., 1994). CGH is a powerful cancer etiology (Hollstein etal.., 1991b; Jones etal.., adjunct to traditional cytogenetic techniques and 1991; Harris, 1993). a useful tool with which to screen for molecular and genetic defects, which will eventually lead to The use of oncogenes and tumour suppressor genes the identification of tumour suppressor genes in cancer epidemio]ogy. The use of molecular and and oncogenes in solid as well as haematological genetic alterations of tumour suppressor genes and tumours. By combining advanced methods for proto-oncogenes in cancer epidemiology has characterizing exposure to carcinogens and mea- advanced our understanding of cancer biology and suring tumour markers, there is a great potential carcinogenesis. Point mutations in tumour sup- for further elucidating the etiology of cancers pressor genes (e.g. ТР53) and oncogenes (e.g. ras) and for the development of strategies for cancer may be specific for both tumour type and the crit- prevention. ical environmental exposure (etiological heterogeneity). Lung tumours from smokers show a high Study designs for the use of tumour markers frequency of G to T traaversiori in both K-ras and Tumour marker studies provide some interesting ТР53, and may reflect the molecular fingerprint of new challenges in study design and statistical carcinogenic constitution of tobacco smoke (]ones analysis. A detailed discussion of study designs for et a?., 1991). Such molecular epidemiological evi- the use of biomarkers in epidemiology is provided dence supports the well established association by Rothman et aL (1995), Hulka & Margolin between smoking and cancer (Vineis & Caporaso, (1992) and Hulka & Garrett (1996).

204 Methodological Essues in the use of tumour markers

Case—case study design Mantel—Haenszel procedure or logistic regression. It has been suggested that case-case (or case—series A test of the null hypothesis that yr(W) =1 is a test or case-only) studies can be employed to evaluate of the hypothesis of etiological heterogeneity, i.e. gene—environment interactions. The critical as that the strength of Y as a risk factor is different for sumption is that the exposure and genetic factors the two case groups (e.g. р53+ and X53—). occur independently and the disease is rare. Under these assumptions, the case—case approach is valid Case—control study design and offers better precision for estimating gene— In case—control studies, etiological heterogeneity environment interactions than does the case— has traditionally been evaluated by two separate control approach (Piegorsch et al., 1994). The case— analyses: marker-positive cases versus controls and case study design can be employed to assess the marker-negative cases versus controls. The analytic association between exposure and tumour markers strategy is to use polychotomous logistic regression aid to evaluate etiological heterogeneity between (Dubin & Pasternak, 1986). In this model, the rela- marker-positive and marker-negative tumours tionships between marker-positive cases and (Taylor, 1989; Begg & Zhang, 1994). This study controls, and between marker-negative cases and design may be used to evaluate the hypothesis that controls are both modelled concurrently using two the two categories of cases, distinguished by the separate (logistic) regression functions. Let X31 be presence or absence of the tumour marker, are the coefficient of the primary risk factor in the characterized by etiological heterogeneity, as evi- logistic regression relating marker-positive cases denced by differences in the strengths of effects of and controls, arid let R2 be the corresponding para- the risk factor in the two case groups. The differ- meter relating marker-negative cases and controls. ences could be due to the fact that the causal path- If there are no interactions between Y and W, then way differs, or they could merely reflect a F is the conditional log odds ratio of the risk different magnitude of effect via the same mecha- factor on marker-positive disease, and RZ is the con- nism. Empirical evidence of such etiological hetero- ditional log odds ratio of the risk factor on marker- geneity with respect to one or more risk factors negative disease. To test the null hypothesis that would provide strong justification for more detailed the two diseases possess etiological heterogeneity investigations of the specific mechanisms of action. with respect to the risk factor, one can test the null This study design consists of a series of incident hypothesis that X31= 1 2' i.e. that the two odds ratios cases. Ideally, this would be a consecutive series of are equal. Such a comparison can be accomplished population-based incident cases. If the ascertain- by using, for example, a likelihood ratio test. ment is not complete, or if, for example, the study Quantitative evidence of the degree of departure is hospital-based, we must assume that case selec- from the hypothesis can be characterized by the tion for the two disease categories is not influenced difference in these coefficients, l — This is the differentially by the risk factors associated with logarithm of the ratio of the two adjusted relative case ascertainment. risks of the risk factor, i.e. the relative risk with respect to marker-positive cases and with respect to 5uрpose that Y is the risk factor of primary interest, assumed for simplicity to be binary, and marker-negative cases, respectively. that W denotes the set of remaining risk factors, where Yf indicates the presence of the risk factor Case—case versus case—control study design and Y indicates its absence. Let X+ (X—) denote It has been shown by our group (Begg & Zhang, the presence (absence) of the tumour marker. 1994) that the odds ratio from the case—case study Furthermore let yr(W) be the odds ratio relating Y is theoretically equivalent to the parameter p, — (3Z and X, conditional on W. In the context of our in the polychotomous logistic regression model, study of p53 overexpression and bladder cancer and thus evaluation of etiological heterogeneity (Zhang et al., 1994b), Y represents smoking status, does not require a conventional control group. X represents the presence or absence of TР53 mu- We illustrate the method using data from our tations in the tumour samples, and W represents own case—case study of the relationship between the remaining risk factors. We can evaluate 'y(W) smoking and TР53 mutations in patients with using standard statistical methods such as the bladder cancer (Zhang et al., 1994b). The raw

205 Application of Biomarkers in Cancer Epidemiology

estimated directly from an appropriately designed case—case study without the need for a control group (Zhang et ai., 1994a, 1994b, 1995а, 1995b; Freedman et al., 1996а, 1996b). The odds ratio Cases obtained from case—case study needs to be interpreted with great caution, since this measure does smoking p53+ p53— Controls not indicate the directions of individual relative status risk of marker-positive or marker-negative disease. Smoker 34 43 81 Therefore, the use of a control group is necessary if Non-smoker 10 21 64 we wish to estimate the actual relative risks for marker-defined tumour subtypes.

frequencies are contained in Table 1. For illustra- Practical issues in tumour tissue banking tive purposes we have employed а control group The research strategy to incorporate exposure, sus- consisting of patients with other cancers believed ceptibility and tumour markers for a case—control to be unrelated to smoking, although this would study includes collection and storage of all related not be an ideal control group for a case—control biological specimens, such as blood samples (for study in general. The odds ratios and confidence cases and controls), tumour and normal tissues (for intervals are presented in Table 2. The unadjusted cases only), ïп addition to the collection of epi- odds ratios are calculated directly from the cross- demiological exposure data. By collecting and products, as usual, i.e. yr = (34 x 21)1(10 x 43), Q= storing blood specimens for all cases and controls, (34 x 64)/(10 x 81) = 2.69, and 02 = certain tumour susceptibility markers, such as (43 x 64)/(21 x 81). Note that yr = 0 1/02. polymorphism in GST M1, NАТ2 and CYP1A1, can We have shown that the odds ratio relating an be assessed. Other markers such as mutagen sensi- environmental risk factor to the presence of a bio- tivity, DNA repair capacity and haemoglobin/DNA logical marker is an appropriate measure for char- adducts can also be measured. Proper collection acterizing the degree of etiological heterogeneity and storage of tumour and normal tissues for cases between the disease groupings defined by the allow the characterization of tumour suppressor marker. These observations legitimize the common genes and oncogener, as well as the DNA replica- recent practice of exploring gene—environment tion repair defect phenotype, and microsatellite associations in case—case studies (Piegorsch et a1., instability. By collecting epidemiological data, 1994). This parameter has been shown to be the exposure history can be evaluated and other ratio of the relative risk of the factor in causing potential confounding factors can be controlled. marker-positive disease to the relative risk in caus- Since the issue of blood specimen banking has ing marker-negative disease. Moreover, it can be been discussed by Landi & Caporaso (this volume)

в .. Estimate (95% Cl) Design . Parameter Llnadjusted Adjusted$ 8(L4) 2.69 (1.23, 5.87) 3.61 (1.41, 9.29) Ρ Case-control А 2(t 1.62 (0.87, 3.00) 2.11 (0.98, 4.54) Ρ,( А ts /О (иИ =. y,() 1.66 (0.69, 4.01) . 1.71 (0.66, 4.43) . Case-case y!(и _ 1.66 (0.69, 4.01) 1.71 (0.63; 4.66)

$Adjusted tir age using iogislic regression tir the case-case approach, and potychotomous logistic regression for tie case control approach 6, = value positive; 02 = value negative. .

206 Methodological issues in the use of tumour markers

H+E'1

_ о _ IHC -'-'-.,-'. — о оmRNA DNA In situ — . .

СAS -200 Image analysis Protein WB NB SB NT RFLP PCR Sequencing

Epidemiological studies: risk factors Clinical studies: prognostic factors

Figure 1. Laboratory strategy for the use of tumour markers. in this monograph, we will briefly review here as well as Т1'53 point mutations measured by PСR- issues related to the storage of tumour and normal SSCP and sequencing techniques (Taylor et al., tissues. 1996). The development of assays that can use With the increasing demands for tumour tissue DNA extracted from farmalin-fixed, paraffiri- samples for biological, clinical оr epiderniological embedded tissue will enhance molecular epidemi- studies, tissue banking has become a very impor- ological investigation. There are some limitations tant practical issue (Lee et aI., 1995). Limited associated with the use of fornalin-fixed, paraffin- resources, untrained personnel and absence of embedded materials, mainly due to the fragmen- uniform protocols for tumour tissue banking have tation of DNA. Fresh or frozen samples are preferred, created obstacles for the proper and rapid colec- if available, for molecular assays at the DNA or tim, processing and storage of tumour tissues RNA level. Since the storage of formalin-fixed, (Grizzle, 1994). In addition, lack of information on paraffin-embedded tissues is a standard procedure diagnostic quality control, histological classifica- in medical centres or hospitals, we will focus iii tion (stage and grade), treatment and outcome discussion on the storage of frozen tissues. may further jeopardize the optimal usage of nor- In order to perform epidemiological studies and mal and tumour samples in cancer research. to share the specimen resource with other investi- There is a rapidly increasing ability to evaluate gators in biological and clinical sciences, a central- a range of tumour markers iп formalln-fixed, paraf- ized programme for research specimen banking is fin-embedded tissues, such as p53 nuclear protein needed in any large research medical centre, accumulation measured by immunohistochem- Biological specimens can be collected and evalu- istry (IHC) (Sarkis etal., 1993а, 1993b, 1994, 1995) ated for research suitability by trained personnel

207 Application of Biomarkers in Cancer Epidemiology

Progression inTis stage Progression in Ta stage Progression in Ti stage

Negative 15 patients 15 censored ---- Negatve 42 patients 37 censored --- Negative 33 palients 26 censored — Positive 15 patients 2 censored — Positive 12 patients 5 censored Postive 44 patients 12 censored Trek mark indicates last fellow up Tick rrrark indicates last follow up Tick mark indicates lost tollcw up m m Ф Ф O o ф ф

И И Ф Ф m col O O d Q- C C O o â ao 0 o r1 o- 0 2445 729t 1211144 24 46 72 96 120 144 158 192 216 0 24 45 72 96 120 144 168 192 218 Progression time (months) Progression time (months) Progression time (months)

Survival in Tis stage Survival in Ta stage Survival in T1 stage -- Negative 18 patients 15 censored ---- Negative 42 patients 39 censored --Negative 33 palienls 30 censored — Positive 15 patients 10 censored — Positive 12 patients 9 censored Posilive 44 patients 27 censored Tick mark indicates 1aef fallow up Tick mark indicates last follow up Tick mark indicates lsl follow up

a7 a)7 Ф7 Ф CO Cs C C C o o o o Ô O Q O_ d o o o 0 a o-

24 46 72 96 120 144 106 24 48 72 99 120 144 169 192 216 24 46 72 96 120 144 168 192 216

Survival time (months) Survival time (months) Survival time (months)

Figure 2. p53 overexpression and prognosis of superficial bladder cancer.

(nurses, technicians and pathologists) and can be At Memorial Sloan Kettering Cancer Center, the distributed among investigators. The materials can tumour banking programme assumes the respon- be collected and prepared to suit the individual sibility of delivering all routine specimens from the requirements of investigators, and collection and operating rooms to pathologists on a regular distribution can be centrally documented. The schedule. This ensures that specimens available for programme can relieve investigators of the burden, research are as fresh as possible. Pathologists are in terms of both time and cost, of procuring and responsible for determining whether diagnostic preparing a specimen on each occasion- The pro- requirements have been satisfied before giving the gramme needs to have an effective quality control portion of the residual tissue for banking. The system in order to evaluate all tissue specimens, specimen is then transferred in an iced specimen assuring investigators of properly diagnosed container. Normal and tumour tissues are stored in research material. plastic cassettes labelled with a coded number to

208 Methodological issues in the use of tumour markers protect patients' confidentiality. The tissue is sex, grade, stage and treatment need to be con- embedded in frozen section support media (OCT) trolled in the data analysis when assessing the and stored at -70°C. A clinical abstract of the effect of a tumour marker. patient's history with an accompanying pathology We have conducted a series of studies to evaluate report wil be provided for each specimen only if the association betweeпр53 nuclear overexpression IRB approval has been obtained to conduct such a and progression/survival in a group of patients study. In general, patients will be informed of the with superficial bladder cancer. p53 nuclear over- ongoing protocol and sign a informed consent expression was evaluated in tumours of 164 patients form. A data management system needs to be (T1= 77, Ta =54 and Tis = 33) with superficial blad- implemented and a relational database system der cancer by immunohistochemistry using the needs to be established so that tissue bank data- mouse monoclonal antibody РАЙ 1801 on deparaf- base can be linked to pathological, epidemiological finized tissue sections. Antibody 1801 detects both and clinical data. wild-type and mutant p53 proteins. Due to the prolonged half-life of the mutated p53 products, they Laboratory strategy for the use of tumour markers accumulate in the nucleus and can usually be A protocol for immunohistochernistry and for detected by immunohistochemical assays. We genetic analysis of formalin-fixed, paraffin-embedded studied 42 primary bladder tumour tissues to esti- tissues has kindly been provided by Drs William mate the sensitivity of immunohisto chemical Bennett and Curtis C. Harris, NCI Laboratory of (IHC) methods in the prediction of ТР53 muta- Human Carcinogenesis (see Appendix). A strategy tions (Coidon-Cardo et al., 1994). We found that has been suggested (Cordon-Cando et al., 1994) the highest sensitivity was reached when the cut- whereby, from a single tissue sample, different off (in terms of percentage of cells with nuclear techniques can be performed to examine irnmuno- immunoreactivity) was 20%, and so we have emp- phenotype and genotype. This strategy is i11us loyed 20% as the cut-off point for IHC results in trated in Fig. 1. Briefly, using consecutive tissue our analysis. The data were first correlated with sections cut at different thicknesses and deposited conventional prognostic parameters, including either on microslides or microtubes, one can: (1) stage, grade, vascular invasion, age and sex. Vari- evaluate morphology (i.e. haematoxylin and eosin ous univariate and multivariate analyses were (H&E) staining), (2) perform inimunohistochem- performed. In the study, none of the normal istry procedures, and (3) characterize molecular urothelial and stromal cells showed p53 nuclear alterations (i.e. Soufihern blot, restriction fragment overexpression. Patients with bladder tumours length polymorphism (RFLP), and PCR-ssCP and were stratified into two groups according to the sequencing). Antigen expression and/or its modu- percentage of cells with nuclear immunoreactivity. lation can be analysed by immunohistochemistry Ninety-three patients (56.7%) had none or less on tumour tissue samples. Finally, specimen iden- than 20% tumour cells with positive nuclear stain- tification can provide the correlation of laboratory ing (group A), while the remaining 71(43.3%) had data with epidemiological, pathological and clini- moie than 20% tumour cells with nuclear im- cal follow-up data (Fig, 1). munoreactivity (group B). separate analyses of progression and survival were performed for the three Issues concerning the use of tumour markers in stages of superficial bladder cancer (Tis, Ta, Ti) after prognostic studies adjusting for age, sex, grade, vascular invasion and The study of the prognostic value of tumour mark- adjuvant therapy. Patients in group B had consis- ers raises several special issues concerning subject tent significantly lower progression-free intervals selection and adjustment for other known prog- and survival (Р < 0.001) at all three stages (Fig. 2). nostic factors such as stage and grade of the disease. These results suggest that superficial bladder can- For example, study populations should usually be cers exhibiting p53 nuclear overexpression have a limited to incident cases diagnosed within a year higher rate of disease progression and short sur- in order to reduce selection bias from differential vival, and may be useful in selecting appropriate survival. Known prognostic factors such as age, therapy (Sarkis et al., 1993а, 1993b, 1994, 1995).

209 Application of 8iomarkers in Cancer Epidemiology

Appendix. Histology protocol: paraffin sections for immunohistochemical and genetic analyses

1. Background: genetic .analysis of archival fume hood. Point the bottle away from you when opening; human tissues internal pressure can lead to splattering. Access to archival human tissues provides many opportunities for both prospective and retrospective studies. 4. Paraffin section guidelines: RNAse Informative assays based on immunohistochemistry and precautions PCR technology are available. Specialized protocols are L Gloves are worn during microtomy. needed for these studies, and essential elements include ii. Disposable blades are used and replaced between confidentiality, proper selection of adhesive coatings and blocks. precautions for tissue carry-over in PCR-based studies. iii.The block holder is cleaned with xyfene between blocks. 2. Tissue specifications iv.DEPC-treated water is used in the tissue flotation bath. The optimal tissue sample is a paraffin block containing at v.Coated slides (i.e. silane or similar) are prepared with least 1 cm2 of tissue, including both tumour and non-tumour sterile DEPC water and handled only with gloved hands. tissue. The non-tumour tissue is used for an internal nega- vi. Cut 25 sections from each block according to the tive control and for germline analysis. Five-micron sections following specifications. Section nos 1-16 must be 5 are used for immunochemistry, and 20-micron sections for microns thick: nos 1 and 16 are stained by haematoxylin microdissection of non-necrotic tissues for genetic analysis. and eosin (H&E), nos 2-15 must be unstained. Section nos 17-25 must Ьé 20 microns thick and unstained; put 3. Glass slide specifications: silanation and two 20-micron-thick sections 0f each slide. Coated slides DEPC treatment and DEPC water must be used for all sections. Glass slides must be coated to promote tissue adherence, For small tissues, cut only 5-micron sections and place otherwise tissue sections will be lost during multiple wash- two 5-micron-thick sections on slide nos 17-25. ings and incubations of thé immunochemistry protocol. vii.If the tissue is exhausted during sectioning, make an Several coating agents are commonly used, including H&E stain of the last (or close to last) section. poly-L-Iysine, glue, silane and others; non-biological prepa- viii.Bake sections at no more than 60°C for no more than rations (i.e. silane) are less commonly contaminated with 2 h. DNase or 'Nase than poly-L-lysine and represent a better ix. Each slide will be labelled with the block number from choice for PCR-based studies. In addition, RNAse precau- the paraffin block, the sequential number of the section tions must be used for sections t глепдёд for microdissection, (i.e. 1-25) and the section thickness (i.e.5 or 20 microns). essential elements include DEPC treatment and baking to inactivate RNase and DNase. Specifications and protocols 5.TrouЫe-shooting guide for using this protocol are listed below: . . It is advisable to test the sections produced by a laboratory i.Coated or charged slides. Either commercial or locally using this protocol. A common problem is that tissue sec- prepared slides will be used. Commercially coated slides tions fall off the slide during microwave antigen-retrieval are available, specify RNase-free. procedures. Usually this is caused by inadequate or ii. DEPC treatment protocol. To destroy any RNase or improper silanation of sections, although some tissues are DNase attached to the glass slides, load the slides into a more likely than others to fall off the slides (i.e. tissues con- metal rack and place in water containing 0.1 % DEPC taining fat or bone). Therefore, investigators are encour- • (diethyl procarbonate) for 15-30 min; wrap the entire rack aged to send slides from the first 10-20 cases to a histol- in aluminum foil. and bake at 1В0°С for at least ogy laboratory, and to request a routine immunostain (i.e. 2 h. Cool to room temperature and store wrapped in foil p53 or Kî67) using 30 min of microwave antigen retrieval at room température indefinitely. (this is the longest interval commonly used). if tissue sec- Caution: DEPC is a potent protein denaturant and is a tions fall off the slides, then ask the laboratory to remake suspected carcinogen; it should be handled with care. their silane solutions and re-check their protocol, and then Wear gloves and safety glasses and work in a chemical repeat the pilot imrnunostairw until the problem is resolved.

210 Methodological issues in the use of tumour markers

Acknowledgements C.J., Petreili, N., Black, J., Zhang, Z.F., Satchidanand, The research was supported in part by grants S.K.S. & Asirwatham, J.Е. (1996b) Familial and nutritional risk factors for p53 overexpression in colorectal E5-06718 from the National Institute of cancer. Cancer EpidemioL Bioniarkers Fret., 5, 285-291 Environmental Health Services and СА-47538 from the National Cancer Institute, National Golomb, H.M., Alimena, G., Rowley, J.D„ Vardiman, Institute of Health, Department of Health and J.W., Testa, J.R. & Sovik, C. (1982) Correlation of occupation and karyotype in adults with acute nonlympho- Human Services. We are grateful to Drs Colin B. cytic leukemia. Blood, 60, 404-411 Begg, Barbara S. Hulka, H. Yamasaki and Anthony J. McMichael for their critical review, helpful com- Grizzle, W.E. (1994) Tissue resources in the detection and ments and suggestions, and to Ms Ming Sun for evaluation of markers. In: Srivastava, s., Lippman, S.M., Hong, W.K. & Mulshine, J.L., cris, Early Detection of her technical assistance. . Cancer: Molecular Markers, Armonk, NY, Futura Publishing Company, Inc., pp. 9-76 References Begg, C.B. & Zhang, Z.F. (1994) Statistical analysis if Harris, C.C. (1993) p53: At the crossтoads of molecular molecular epidemiology studies employing case-series. carcinogenesis and risk assessment. Science 262, Cancer Fpirfernioi. Biomarkers Prev., 3, 173-175 1980-1981 Brash, D.E., Rudolph, J.A., Simon, J.A., Lin, A., McKenna, Harris, C.C. & Hollstein, M.C. (1993) Clinical implica- G.J., Bodes, H.P. Halperin, A.J. & Ponten, J. (1991) A role tions of the p53 tumor-suppressor gene. N. J. EпgI. Med., for sunlight in skin cancer: UV-induced p53 mutations in 329,1318-1327 squamous cell carcinoma. Proc. Nat1 Acad. Sri., 88, Hollstein, M.C., Pen, L., Mandard, A.M., Welsh, J.A., 10 124-10 128 Montesano, R., Metcalf, А.А., Bak, M. & Harris, C.C. Bressac, B., Kew, M., Wands, J. & Ozlurk, M. (1991) (1991а) Genetic analysis of human esophageal tumors Selective G to T mutations of p53 gene in hepatoce11ular from two high incidence geographic areas: frequent p53 carcinoma from southern Africa. Nature, 350, 429-431 base substitutions and absence of ras mutations. Cancer Res., 51, 4102-4106 Chu, II. (1987) Biochemical markers for human cancer. In: Seifert, G. cd., Morphological Tumor Markers, New Hollstein, M., Sidrarlsky, D., Vogelstein, B. & Harris, C.C. York, springer-Verlag, pp. 19-46 (1991b) p53 mutations in human cancers. Science, 253, 49-53 Cordon-Cando, C., Dalbagni, G., Saez, G.Т., Oliva, M.R., Zhang, Z.F., Rosai, J., Reuter, V.E. & Pellicer, A. (1994) p53 Houldsworth, J. & Chaganti, R.S. (1994) Comparative mutations in human bladder cancer: genotypic versus genomic hybridization: an overview [comment). Am. J. PathoL, 145, 1253-1260 phenotypic patterns. lot. J. Cancer, 56, 347-353 Dubin, N. & Pasternak, B.S. (1986) Risk assessment for Hsu, I.C., Metcalf, R.A., Sun, T., Welsh, J.А., Wang, N.J. & case-control analysis by polychotamous logistic regres- Harris, C.C. (1991) Mutational hotspot in the p53 gene sion. Am. J. EpidemioL, 123, 1101-1117 in human heptocellular carcinoma. Nature, 350, 427-428 Fagioli, F., Cuneo, A., Piva, N., Carl!, M.G., Previati, R., Hulka, B.S. & Garrett, P (1996) Influence of biomarkers on the design, methods, and interpretation of epidemi- Balboni, M., Tomasi, P, Cariani, D., Scapoli, Г. & Castoldi, G. (1992) Distinct cytogenetic and clinico- ologic studies. J. Toxicol. Environ. Health, 40, 389 pathologic features in acute myeloid leukemia after occu- Hulka, B.S. & Margolin, B.H. (1992) Methodological pational exposure to pesticides and organic solvents. issues in epidecoiologic studies using biologic markers. Cancer, 70, 77-85 Am. J. EpidemioL, 135, 200-209 Field, J.K., Spandidos, D.A., Malliri, A., Gosney, J.R., Johansson, B., Mertens, F., Heim, 5., Kristoffersson, U. & Yiagnisis, M. & ste11, P.M. (1991) Elevated P53 expression Mitelman, F. (1991) Cytogenetics of secondary correlates with a history of heavy smoking in squamous cell myelodysplasia (sMDS) and acute nonlymphocytic carcinoma of the head and neck. Br. J. Cancer, 64, 573-577 leukemia (sANLL). Eur. J. Haernatol., 47, 17-27 Freedman, A.N., Michalek, A.M., Marshall, J.R., Mettin, Jones, RA., Buckley, J.D., Henderson, B.E., Ross, R.K. & C.J., Petrelli, N., Black, J., Zhang, Z.F., 5atchidanand, Pike, M.C. (1991) From gene to carcinogen: a rapidly S.K.S. & Asirwatham, J.E. (1996a) The relationship evolving field in molecular epidemiology. Cancer Res., between smoking and p53 overexpression in colorectal 51, 3617-3620 cancer. Br. J. Cancer, 73, 902-908 Kallioniemi, 0.R, Ka1lioniemi, A., Sudar, D., Rutovitz, D., Freedman, A.N., Michalek, A.M., Marshall, J.R., Mettlin, Gray, J. W., Waldman, F. & Pinkel, D. (1993) Comparative

211 Application of Biomarkers in Cancer Epidemiology

genomic hybridization: a rapid new method for detect- Patti, D., Blanchi, S., 17ecar1i, A., Cipriani, F., Avellini, C., ing and mapping DNA amplification in tumors. 5ет[пars Cocci, P., Falci i, F., Puntoni, R., Russo, A. & Vindigni, C. Cancer BioL, 4, 41-46 (1992) A case-control study of cancers of the gastric car- Kallioniemi, O.P., Kallioniemi, A., Piper, J., Isola, J., dia in Italy. Br. JCancer, 65, 263-266 Waldman, F.M., Gray, J.W. & Pinkel, D. (1994) Piegorsch, W.W., Weinberg, C.R. & Tауlоr, J.А. (1994) Optimizing comparative genomic hybridization for Non-hierarchical logistic models and case-only design analysis of DNA sequence copy number changes in solid for assessing susceptibilty in population-based case-con- tumors. Genes, Chromosomes & Cancer, 10, 231-243 trol studies. Stat. Med., 13, 153-162 Klavins, J.V. (1989) Search for specific tumor markers. In; Rothman, N., Stewart, W.F. & Schulte, P.A. (1995) Ting, S.W., Chen, J.S. & Schwartz, М.K., eds, Human Incorporating biomarkers into cancer epidemiology: А Tumor Markers. New York, Elsevier. pp. 101-104 matrix of biomarker and study design categories. Cancer kondo, K., Umemoto, A., Akimoto, S., Oyama, T., Epidеmiы. Biomarkers Prey, 4, 301-311 Hayashi, K,, Ohzishi, Y. & linden, У. (1992) Mutations Rowiey, J.D. (1983) Chromosome changes in leukemic in the P53 tumor suppressor gene in primary lung can- cells as indicators of mutageeic exposures. In: Rowley, cer in Japan. Biochem. Biophys. Res. Commun., 183, J.D. & Ultmann, J.E., eds, Chromosomes and Cancer, New 1139-1146 York, Academic Press. pp. 139-157 Kramer, B.S. & Srivastava, S. (1994) Molecular biology in Rowley, J.D. (1984) Biological implications of consistent lung cancer prevention: from the bench to the bedside. chromosome rearrangements in leukemia and lyin- In: Srivastava, 5., Lippman, S.M., Hong, W.K. & Mulshine, phoma. Cancer R es, 44, 3159-3168 J.L., eds, Early Detection of Cancer, Molecular Markers, Rowley, J.D., Golomb, H.M. & Vardiman, J.W. (1981) Armonk, NY Futura Publishing Company, Inc., pp. 1-12 Nonrandom chromosome abnormalities in acute Kreyberg, L. (1937) Lungecancer. Discussion (Norw.) leukemia and dysmyelopoietic syndromes in patients Detnorske medidnske Selskab. Norske Mag. f. Lagev, 98, with previously treated malignant disease. Blood, 58, 36-41 759-767 Kreyberg, L. (1952) One hundred consecutive primary Sandier, D.P. & Collman, G.W. (1987) Cytogenetic and epithelial lung tumors. Br. J. Cancer, 6, 112-119 environmental factors in the etiology of the acute Kreyberg, L. (1954) The significance of histological typ- leukemia in adults. Am. J. Epidemiol., 126, 1017-1032 ing in the study of the epidemiology of primary epithe- Sandler, D.P., Shore, D.L., Anderson, J.R., Davey, F.R., lial lung tumors. A study of 466 cases. Br. J. Cancer, 8, Arthur, D., Mayer, R.J., Silver, А.T., Weiss, R.B., Moore, 199--208 J.O., Schiffer, C.A. et cl. (1993) Cigarette smoking and Lee, L.W., Griffith, J., Zenick, H. & Hulka, B.S. (1995) risk of acute leukemia: associations with morphology Human tissue monitoring arid specimen banking: oppor- and cytogenetic abnormalities in bone marrow. J. Nat! tunities for exposure assessment, risk assessment, and Cancer Inst. 85, 1994-2003 epidemiologic research. Environ. Health Perspect., 103 Sarkis, A.S., Dalbagni, G., Cordon-Cardo, C., Zhang, Z.F., (Suppl. 3), 3-8 Sheinfeld, J., Fair, W.R., Herr, H.W. & Reuter, V.P. (1993a) McClay, E.F. (1989) F.pidemioiogy of bone and soft tissue Nuclear overexpression of p53 protein in transitional cell sarcomas. S oars m ii., 16, 264-272 bladder carcinoma: a marker for disease progression. J. ети с Nat! Cancer Inst., 85, 53-59 Mitelman, F., Nilsson, P.G., Brandt, L., Alimena, G., Gastakil, R. & Dallapiccola, B. (1981) Chromosome pat- Sarkis, А.S., Zhang, Z.F., Cordon-Cardo, C., Melarned, J., tern, occupation, and clinical features in patients with Dalbagni, G., Sheinfeld, J., Fair, W.R., Herr, H.W. & acute nonlymphocytic leukemia. Cancer Genet. Reuter, V.P. (1993Ь) P53 nuclear overexpression in Ta Cytogenet., 4, 197-214 non-invasive papillary transitional cell carcinoma of the bladder. lof. J. licol., 3, 355-360 Mitelman, F., Nilsson, P.G. & Brandt, L. (1984) Fourth International Workshop on Chromosomes in Leukemia Sarkis, A.S., Dalbagni, G., Cordon-Cardo, C., Melamed, 1982: Correlation of karyotype and occupational expo- J., Zhang, Z.F., Sheinfeld, J., Fair, W.R. & Herr, H.W. sure to potential mutagenicicarcinogenic agents in acute (1994) Association of p53 nuclear overexpression and nonIymphocytic leukemia. Cancer Genet. Cytogenet,, 11, tumor progression in carcinoma in situ of the bladder. J. 326-331 UroL, 152, 388.-392 Ozturk, M. (1991) p53 mutation in hepatoce11ular carci- Sarkis, As., Bajorin, D.P., Reuter, V.É., Herr, H.W., Netto, noma after aflatoxin exposure. Lencef, 338, 1356-1359 G., Zhang, Z.F., Schultz, RK., Cordori-Cardo, C. & Scher,

212 Methodological issues in the use of tumour markers

H.I. (1995) Prognostic value of p53 nuclear ovеrеxpres- Thompson, CT. & Gray, J.W. (1993) Cytogenedc profil- siеn in patients with invasive bladder cancer treated with ing using fluorescence in situ hybridization (FISH) and neoadjuvant MVAC. J. Clin. licol., 13, 1364-1390 comparative genomic hybridization (CGH). J. Cell. Schwartz, M.K. (1989) Tumor markers in diagnosis aid Biochem., 17G, 139-143 screening. In: Ting, S.W., Chen, J.S. & Schwartz, M.K., Vaughan, T.L., Davis, S., Kristal, A. & Thomas, D.B. eds, ionien Tumor Markers, New York, Elsevier, pp. 3-16 (1995) Obesity, alcohol, and tobacco as risk factors for cancers of the esophagus and gastric cardia: adenocarci- Sell, . (1992) Cancer markers of the 1990s. In: Sell, S., Ѕ noma versus squamous cell carcinoma. Cancer Epidemiol. ed., serological Cancer i'4 k , New Jersey, Humaria Press, агегѕ Biomarkers Prey., 4, 85-92 pp. 1-17 Vineis, P. & Caporaso, N. (1995) Tobacco and cancer: epi- Spruck, C.H.L, Rideout, W.M., O1umi, , Ohneseit, P.F., АК demiology and the laboratory. Environ, Health Perspect., Yang, A.S., Tsai, Y.С., Nichols, kw., Horn, T., Hermann, G.G., Steven, K., Ross, P.K., Yu, M.C. & Jones, P.A. (1993) 103, 156-160 Distinct pattern of p53 mutations in bladder cancer: rela- Zhang, Z.F., Aprikian, A., Sarkis, A.S., Zeg, Z,S., Pollack, tionship to tobacco usage. Cancer Res., 53, 1162-1166. D. Cordon-Cardo, C., Fair, W.R. & Begg, C.В. (1994a) (Erratum: Cancer Res. 1993, 53 (Supp1. 10), 2427) Factors associated with p53 accumulation in prostatic adenocarcinoma. Int. J. licol., 4, 897-901 Sturgeon, S.R., I-lartge, P., Silveгmaп, D.T., Kantor, A.F., Linehan, W.M., Lynch, C. & Hoover, A.N. (1994) Zhang, Z.F., Sarkis, A.S., Cordon-Cardo, C., Dalbagni, G., Association between bladder cancer risk factors arid Melamed, J., Aprikian, A., Pollack, D., Sheinfeld, J., Hen, tumor stage and grade at diagnosis. Epidemiology, 5, H.W., Fair, W.R., Reuter, V.E. & Begg, C.B. (1994b) 218-225 Tobacco smoking, occupation, arid p53 nuclear overexpression in early stage bladder cancer. Cancer Epidemiol.. Suzuki, H., Takahashi, T., Kuroishi, T., suyama, M., Bromarkers Prev., 3, 19-24 Ariyoshi, Y. & Ueda, R. (1992) p53 mutations in non-small cell lung cancer in Japan: assodation between Zhang, Z.F., karpeh, M.S., Lauwers, G.Y., Marrero, A.M., mutations and smoking. Cancer Res., 52, 734-736 Роllасk, D., Cordon-Cardo, C. & Begg, C.B. (1995a) Risk factors and p53 nuclear overexpression in early stage Taylor, J. . (1989) Oncogenes and their applications in А intestinal-type adenocarcinoma of the stomach. Cancer epidemiologic studies. Am. J. Epidemiol., 130, 6-13 Detection Prey., 19, 156-164 Taylor, J. ., Sandler, D.P., Bloomfield, C.D., Shore, D.L., А Zhang, Z.F., Zeng, Z.S., Sarkis, A.S., Klimstra, D.S., Ball, E.D., Neubauer, A., McIntyre, O.R. & Liu, E. (1992) Charytonowicz, E., Guillem, J.G., Cordon-Cardo, C., ras oncogene activation and occupational exposure in Cohen, A.M. & Begg, C. (199 Sb) Family history of cancer, acute myeloid leukemia. J. Nat[ Cancer Inst., 84, body weight and p53 nuclear over-expression in Duke's 1626-1632 C colorectal cancer. Br. J. Cancer, 71, 888-893 Taylor, J.A., Watson, M.A., Devereux, T.R., Michels, R Y., Saccomanno, G. & Anderson, M. (1994) p53 mutation hotspot in radon-associated lung cancer [see comments). Corresponding author Lancet, 343, 86-87 Z.-B. Zhang Taylor, J.A., Li, У., Не, M., Mason, T., Mettlin, C., Vogler, Department of Epidemiology and Biostatistics, Memorial Sloan Kettering Cancer Center, 1275 York W.J., Maygarden, Ѕ . & Liu, E. (1996) p53 mutations in bladder tumors from arylamine-exposed workers. Cancer Avenue, Box 44, New York, NY 10021, USA Res., 55, 294-298

213 Application of Biomarkera i9 Cancer Epidemiology To iolo, P., B п оFPetta, P. 5huker, D_E.G., Rolhmaп, N., HuIka, B. and Pearce, N., eds IARC Reienhilic PuhlrcaIiorri No. 142 International Agency for Research cm Caesar, Lyon, 1997

Quality control of biomarker measurement in epidemiology

D. Gompertz

The throughput and complexity of a biomarker assay will determine the amount of effort that can be expended on quality control and assurance. Clinical chemistry quality control procedures can be readily applied to simpler chemical analysis such as blood lead and cholesterol, but even complex cell-based biomarker techniques such as 'PRT mutation analysis and cytogenetics benefit from a formal quality control approach. Collaborative interlaboratory exercises are essential, especially when no certified reference material is available, and these can play a central role in the control of laboratory drift. Recommendations are made for the quality control of biomarker measurement based on clinical chemistry techniques. These include recommendations for coding samples so that the laboratory scientist is unaware of exposure status and for the use of formal laboratory protocols.

Incorporation of biomarker measurements into few dollars per sample (e.g. blood lead measure- epidemiological studies can greatly increase their ments) to labour-intensive methods that cost over power and the strength of associations that are $1000/sample, with a batch of a dozen taking sev- found. The measurement and control of error in eral days to process (e.g. dioxin and haemoglobin such studies are largely the province of the epi- addict measurements). High-throughput auto- demiologist (White, this volume), however, the mated analytical techniques are more easily quality of the laboratory measurements remains adapted to the quality control techniques that the responsibility of the laboratory scientist pro- have been so effective in improving clinical chern- viding the biomarker test (Armstrong et al., 1992; istry performance; however, low-capacity methods Aitio & Apostoli, 1995). The importance of `transi- requiring cell culture and other biological ap- tional' studies in determining intra-individual and proaches are more difficult to control in this way. interindividrial variability and the factors affecting Quality control procedures developed for the high- these has been stressed (Hulka, 1991; Schulte & throughput methods clearly will not be directly Perera, this volume) but the role of laboratory applicable to the more biological assays. There are; quality control in minimizing laboratory error howevеr, lessors that can be learnt from the clinical needs to be emphasised. Failure to control labora- chemistry approach to quality control and assur- tory drift over time and a lack of comparability ance that are applicable to all assays, and specific with other studies can cause major difficulties in recommendations can be made for biomarker mea- interpretation and analysis in epidemiological surements being used in epidemiological studies. In studies dependent on biomarker measurements this paper, a general approach to laboratory quality (Vineis etaL, 1993). control and assurance is described, followed by The range of biomarkers that might be incorpo- examples taken across the range of biomarkers rated into an epidemiological study requires a vari- available, i.e. blood lead, serum cholesterol, ety of techniques, extending from sophisticated dioxin, haemoglobin adducts, Э2Р-рostlabelling, analytical chemistry (atomic absorption spectrom- HPRT mutation assays and cytogenetic scoring. etry, isotope dilution mass spectrometry) to more As clinical chemistry developed, the precision biological methods dependent upon celt сиltите of an assay was monitored during an analytical and niolecular biology techniques. The through- run by observing the repeatability of duplicate put and complexity of a method can vary from the measurements and the repeatability of quality con- ability to analyse 100 samples a day at a cost of a trol standards analysed throughout the run. (For

215 Application of Bbmarkers in Cancer Epidemiology definitions of agreed terms used in laboratory qual- longitudinal control of laboratory performance ity assurance programmes, see Aitio & Apostoli, can be as important as in cohort studies involving 1995). Day-to-day reproducibility was assessed by multiple measurements on an individual at differ- repeated analysis of the quality control samples ent time periods. The control of long-term labora- and, at times, carry-over samples from a previous tory drift is a complex issue (Broughton et al., batch. Numerous texts have been written explain- 1986). A simple approach is to establish a pool of 11g these techniques of quality control quality control samples at the outset, but the use- (Whitehead, 1977; Aitio & Apostoli, 1995) and the fulness of this approach depends on the Iong-tеrт statistical approaches towards managing quality stability of these samples; it may be difficult to dis- control data, including statistical rules for the tinguish between laboratory drift and gradual dete- acceptability of quality control results (Westgard et rioration of quality control material. The number of г1., 1981). These statistical approaches have been quality control analyses performed over the course used to establish the amount of effort required to of a study may not necessarily be large enough to monitor precision throughout a batch, i.e. the establish statistically the extent of laboratory drift number of quality control replicates that are (Broughton et aI., 1986). However, the regular needed. This number depends on the length of a participation in interlaboratory quality assurance run that can be expected without significant drift, exercises combined with the use of in-house drift and as automated clinical chemistry instruments controls will give the laboratory scientist and the become more sophisticated, less attention to collaborating epidemiologists some reassurance. within-run drift is required. Some analyses such as There are special problems associated with Ыоод lead measurement (see below) still require longer-term epidemiological studies, during which frequent quality control samples to allow for there may be advances in analytical techniques or changing instrumental conditions. Strict protocols changes in instrumentation, reagents or scientific and automatic data handling can ensure tight con- staff. It is the responsibility of the analyst, when trit of day-to-day quality performance. Assay per- change 1s foreseeable, to establish the comparability formance can be compared with similar assays of the old and new techniques in terms of sensitiv- using coefficients of variation (CV) (within and ity (analytical), precision and bias. Over the short between batches); these are measures of analytical term, new and old methods should be run in paral- performance and do riot reflect intra-individual or lel using the same pool of quality control material or interindividual variation. samples. Methods that depend on visual scoring These quality control techniques can demon- (colonies, chromosomal aberrations) may require strate control of 'imprecision (reproducibility) considerable training and inter-reader comparison within any batch of samples and on a day-to-day periods to ensure that changing the technician does basis, but do not give any indication of the `true- not produce significant shifts in background от sen- ness' (or accuracy) of the result. This can only be sitivity. Interlaboratory comparisons before and obtained by reference to 'definitive' or 'reference' after any change are desirable to ensure that any methods or by use of certified reference material change in bias not detected by in-house control (Aitio & Apostol1, 1995). However, for newer assays methods is found and documented. There is evi- and more biological assays, such materials and dence from a number of fields that long-terms labo- methods are unlikely to be available arid the only ratory performance is best ensured by adherence to approach available to the laboratory scientist is to an agreed and detailed protocol and to the rule that establish comparability with other laboratories. individual analysts are not permitted to make mod- Interlaboratory assessment exercises have played a ifications without an assessment of the likely effects central role in establishing the performance of the on the performance characteristics of the assay. different analytical methods described below, and There is always the temptation for an originator of for some analyses there are ongoing interlabora- a new assay to incorporate improvements as experi- tory quality assurance schemes on a national or ence is gained, but this can cause major difficulties international level. during a study and its analysis. Cross-sectional epidemiological studies may, in In clinical laboratory science, it is accepted that practice, take considerable time to perform and thus quality assurance extends beyond analytical quality

216 Quality control of biomarker measurement in epidemiology control to selection of sampling materials (syringes, mean is a good reflection of the actual amount of tubes), maintaining sample integrity during lead present as measured by a definitive method transport and before analysis, control of contami- such as isotope dilution analysis (Bullock et aI., nation, sample handing and documentation. 1986). Stability of quality control material on long- Quality control of this pre-analytical phase can be term storage and the continuing availability of monitored by the field survey staff submitting interlaboratory comparison has enabled long-term duplicate 'blind' samples from the same individ- studies (10+ years) of the effect of environmental ual. It is now generally accepted that all 'exposed' change on population blood lead levels in the UK and `control samples from epidemiological studies to be performed (Delves et aI., 1996). should be coded and analysed without the labora- Blood lead is a well-established biomarker; the tory scientist knowing their status. It is the point analytical methods and their quality control have responsibility of the epidemiologist and the labo- matured over the last decade and so lead can be ratory scientist to establish a coding system that is used as a model for other toxic metal analyses and, effective and allows reliable tracking of a sample to a great extent, for other chemical contaminants. from receipt to reporting. Protocols for the collection and documentation Cholesterol of samples should be agreed between the labora- Serum cholesterol measurement is one of the usual tory scientist and epidemiologist. The timing of range of analyses offered by clinical chemistry lab- the sample in relation to recent exposures may be oratories with a possible throughput of hundreds critical for environmental contaminants with of samples per day. It is also part of the lipid bio- short elimination half-lives and should be agreed marker profile that is used in the epidemiology of as the protocol is developed, but it should always cardiovascular disease. It is probably this latter use be recorded. Similarly the time between collection that has been responsible for the attention that has and storage should be known. The number and been paid to the control of this measurement on origin of each batch of sample containers and a both a short- and long-term basis. In the USA, selection from each batch should be kept for future detailed recommendations have been produced for reference; this is especially important if there is the the measurement of total cholesterol in blood possibility of contamination of the containers [National Cholesterol Education Program (NCEP), before use. 1988]. More recently Westgard et al. (1991) have shown, for these recommendations, the number of Lead control measurements per run required and the Blood lead is an example of a biomarker analysis analytical coefficient of variation and bias that are that has been used successfully in the workplace for needed to produce a laboratory performance research and regulation and in population studies. deemed acceptable in practice, i.e. an intralabora- The analysis can use automated sampling equip- tory CV of less than 3%. More recently, Bachorik et ment and has a reasonable throughput (50-100 aI. (1995) have produced for NCEP recommenda- samples/day), and so classical clinical chemistry tions and criteria of analytical performance for quality control and assurance techniques can be measurements of low-density lipoprotein (LDL) used. Intra-batch precision (a typical CV of 3-5%) cholesterol. These authors discussed the problems is usually monitored by duplicate analysis of al the of maintaining linkage with existing epidemiolog- samples with quality control material analysed ical data as newer laboratory methods evolve. They after every six to eight actual samples. Inter- recommended that the most prudent course at laboratory quality assurance exercises are essential that time was to continue to measure LDL-choles- for maintaining laboratory performance. In the terol by methods similar to those used to establish United Kingdom (UK) scheme, a blind 'spiked' the epidemiological database on which the rela- sample is sent to contributing laboratories every tionships between cardiovascular risk and LDL- 2 weeks for analysis. Performance is assessed in cholesterol had been established. comparison with the mean result from the 100+ In longer-term epidemiological studies, during UK and overseas participants ('consensus mean'). which there may be considerable changes in pop- Earlier studies have shown that this consensus ulation mean levels of the biomarker of interest, it

217 Application of 8iomarkers in Cancer Epidemiology is important that laboratory drift is identified and tity of the individual samples. The study design controlled. Internal quality control procedures are resulted in approximately 4000 individual pieces not necessarily able to detect such drift and some of PCDD/PCDF data being generated by a variety external source of comparison is required. 5taпdard of analytical methods and by laboratories of dif- reference sera are available for cholesterol analyses fering experience. A number of statistical analyses but the availability of a regular proficiency testing of the data were performed. The simplest was the scheme, such as the National External Quality comparison of coefficients of variation for intra- Assessment Scheme (NÉQАS) in the UK, has the laboratory and interlaboratory variability. Eleven advantage that any laboratory is able to compare out of 16 laboratories produced similar values for its performance with a number of others. For the various ana1ytes in milk, with average intra- serum cholesterol, the current interlaboratory laboratory CVs of 20-30%. Nine of 15 laboratories coefficient of variation in this scheme is typically produced data for blood with average intralabora- 3.8% (n 438). tory CVs less than 30%. Several approaches were Although cholesterol measurements are used on used to identify the principal determinants of the a daily basis in a clinical setting and in population quality of data produced. The results showed that screening programmes, as well as in large-scale the experience of the laboratory was of more cohort studies, the maintenance of accuracy and importance than the actual method or sophistica- precision over a long period still presents difficul- tion of the instrumentation used. The authors con- ties, and considerable care is required to monitor cluded that laboratories that had well established performance using ongoing internal laboratory quality assurance programmes, which were fol- quality control procedures, external reference lowed carefully, produced superior results. material, stored sample pools and external quality The results from this study show that even these assurance schemes (Broughton et a1.,1986). complex and demanding assays with low throughput can be subjected to formal quality assurance Dioxins procedures and they also show the importance of Polychlorinated diberizodioxins (PCDDs) and interlaboratory comparison exercises and high dibenzofurans (PCDFs) occur as complex mixtures quality reference material in obtaining and main- in ultra-trace (picogram) amounts in biological flu- taining consistent results. ids, but in recent years the development of highly sensitive high-resolution gas-chromatography Carcinogen-protein adducts mass spectrometry has allowed them to be quanti- Protein adducts reflect the dose of an active meta- tated. The methods for measurement involve effec- bolite of the carcinogen produced in a target tissue tive sample clean-up and extraction techniques, and are thus useful surrogates for the shorter-lived are labour intensive and do not have a high equivalent DNA species (Skipper & Tannenbaum, throughput (Rappe et al., 1991). The considerable 1990; Wild & Pisani, this volume). Carcinogen- progress that has been made to ensure good qual- protein adducts have been described for ethylene ity control of these measurements has depended oxide, benzo[a]pyrene, 4-aminobiphenyl and afla- on the development of detailed in-house protocols toxin B~, among others. There has been little inter- (Patterson et "L, 1991), the use of certified refer- laboratory effort to document analytical perfor- ence materials (Rymen, 1994) and interlaboratory mance, although the technology (high-resolution proficiency testing exercises (Yrjanheikki, 1989; gas-chromatography mass spectrometry) is similar Stephens et "L, 1992). to that used for dioxin analysis. An extensive and impressive interlaboratory An interlaboratory study has been organised to exercise was initiated by WHO Europe in 1989 investigate performance in the measurement of (Stephens et at, 1992). In this study three samples N-(2-hydroxyеthyl)valine in human gloЫп, a each of human breast milk and blood were sent to haemoglobin adduct biomarker of ethylene oxide 19 participating laboratories from 14 countries. exposure (Tёrngvist et "L, 1992). Iп this study, Two of the three samples were spiked with known samples of globin from eight individuals were sent amounts of PCDDs and PCDFs; the third remained to four laboratories in different countries. The unspiked. None of the laboratories knew the iden- results varied from laboratory.to laboratory, and up

218 Quality control of biomarker measurement in epidemiology to 10-fold in extent for some samples. Some, but and the between-laboratory coefficients of varia- not all, of the differences were shown to be the tion ranged from 56-70% for the `standard' result of different internal standards. However, it method to 65-98% for the butanol-enrichment appeared that, although the absolute measures method. These high coefficients of variation are could not be relied upon, the ranking of the sam- not unusual in any laboratory discipline at the ples was reasonably consistent. early stages of method development and may This interlaboratory study contrasts with the reflect the wide range of conditions employed at dioxin study reported in the same year by Stephens each stage by the different laboratories. It is also et aI. (1992). Dioxin estimations were much more not totally clear whether the adducted samples well established, lessons had been learnt from pre- were completely stable during transit to the partic- vious round-robin exercises and standard material ipating laboratories. However, the results showed was becoming available. This adduct study illus- that there was a general qualitative agreement trates the problem of establishing interlaboratory between the laboratories and they also showed the studies for new assays when there are only a lim- possibilities for improvement in interlaboratory ited number of laboratories involved worldwide precision. The importance of stable and relevant and when it is difficult to produce standards of reference material is once again emphasized. adequate quality and in reasonable quantity. However, important lessons were learnt from the HPRT mutations in lymphocytes study, including the lack of suitability of this bio- Considerable effort has been made by more expe- marker at that time for large-scale epidemiology rienced laboratories to establish how HPRT muta- without considerably more development. tional frequencies vary from day to day in the same individual, between individuals and between 32 laboratories and with different assay techniques Р-postlabенing for DNA adduct analysis (Cole & skopek, 1994; Robinson et al., 1994). Thus Current 32Р-postlabеlling techniques for the detection and quantitation of the adducts of carcino- these laboratories have established much of the genic compounds with DNA have great sensitivity, 'transitional' information required when deciding how to use this end-point in an epidemiological being able to detect one bеnzо~я]pуrenediol ерох- ide residue per 10 nucleotides. Although this assay study, especially the extent of intra-individual and is still in its early stages of development, it has interindividual variability in control and exposed already shown its usefulness in studying exposure groups. to environmental and occupational pollutants Two laboratories (Universities of Vermont and (Perera et al., 1992; Nielsen et al., 1996). The assay Sussex; see Cole & 5kopek, 1994) have compared is complex, with extraction, digestion, labelling, split and repeat samples from the same individual. chromatography and counting stages, and indi These laboratories found similar mutation fre- vidual laboratories have developed their own vari- quencies although there were differences in exper- ants. This complexity and the amount of 3гР imental téchпiques. Both found that the observed required per assay limits the throughput possible mutant frequency for any indiviдиal donor may vary and thus the ability to incorporate a number of considerably between experimental determina- quality control samples in any batch. tions (two- to eightfold). It may be that differences An interlaboratory trial has been performed in laboratory protocols are of less importance than involving 15 laboratories in eight countries to each laboratory establishing and keeping to a well determine the extent of reproducibility of the defined protocol throughout a study. Robinson et assay as it was then being performed (Phillips & al. (1994) compared data sets from four laborato- Castegnaro, 1993). The laboratories were sent four ries and reported protocol variations, some of samples of extracted DNA—two were mouse DNA which may contribute to interlaboratory variation. after in-vivo treatment with polycyclic aromatic The examples of protocol variation that might be carcinogens, one sample was from lung tissue of a important were: preincubation of the mononuclear smoker and one was a control sample. The partici- cells in the presence of mitogen before cloning, pants were asked details of their individual proto- media, serum, inteтleukin-2 (IL-2) source and cols. Three methods were used by the laboratories concentration, type of feeder cells and irradiation

219 Application of Biomarkers in Cancer Epidemiology level, cell density per well, 6-thioguanine concen- exposure/control status. Experienced laboratories tration, and technique and level of experience of have routines for maintaining scoring standards each laboratory worker. The routine splitting of and training new scorers. One such laboratory blood samples or performing a larger experiment keeps a set of coded slides, which have been scored also varied between laboratories and was clearly an by experienced persons and show good agreement, important decision to make. to train recruits. This laboratory circulates coded Quality control of this assay is clearly difficult, slides amongst the experienced scorers and, in as the throughput does not allow a large number of experiments, prepares replicate slides from the replicate samples for control purposes. The use of same culture for scoring by several people (D.C. a strict protocol in any one laboratory is of impor- Lloyd, personal communication). tance, as is the investigation of the performance of Quality control of cytogenetic techniques new batches of sera, biochemicals arid other con- involving cell culture and scoring stages presents sumables. Albertini and Hayes (this volume) different problems from the control of the analysis emphasise that, for this, as for all culture systems, of a chemical biomarker. Three key recommenda- attention to reagents and conditidns is critical. The tions that can be made are as follows: (1) there maintenance of a full computerised database giv- should be development of a detailed laboratory ing details of all the experimental variables protocol, (2) there should be a scoring policy (including the laboratory workers or combinations which ensures that all slides are scored 'blind' and of laboratory workers involved, as well as media that consistency 1s maintained between different components such as serum batch) in addition to scorers by regular exchange of slides, and (3) where the raw data (plate counts) is particularly useful (1. possible, comparisons with other laboratories should Cole, personal communication). Albertini and be established on a regular basis. As newer methods, Hayes (this volume) note that the ability to cryop- such as the fluoresence in-situ hybridization (FISH) reserve the mononuclear cell fraction for subse- technique, are incorporated into population stud- quent testing provides a control for intralaboratory ies, laboratory intercomparison exercises will be drift and that such standards could become avail- necessary to establish an external validity. able for interlaboratory comparisons. Discussion Cytogenetics The examples described here show the consider- Although molecular methods for assessing cytoge- able efforts that laboratory scientists have made to netic damage are advancing rapidly, chromosome control the intralaboratory and interlaboratory aberration analysis has been used over the last 30 variability of their assay procedure. The quality years as a biological measure of radiation exposure control techniques used for the simpler and more (Tucker et al., this volume). There have been sev- routine chemical assays (lead and cholesterol) are eral interlaboratory collaborative exercises com- reflected in low interlaboratory CVs and an paring radiation dose estimation by cytogenetic increased confidence in their accuracy (trueness). analysis. These have covered either the whole assay As assay techniques become more sophisticated and analysis or just the scoring of aberrations. For and more demanding, and as the end-points are example, Lloyd et a2. (1987) dispatched aliquots of more biological, quality control procedures are whole irradiated blood to the participating labora- more difficult to incorporate. However, intralab- tories for culture and analysis, while Garcia et ai. oratory and interlaboratory comparisons do give (1995) sent unstained coded slides to their partici- confidence that an assay is performing reliably and pants. indicate the extent of its inherent variability. All cell culture systems need close attention to A lesson that comes from these differing labo- be paid to reagents and culture conditions ratory assays is that best performance is associated (Albertini & Hayes, this volume), and the mainte- with the development of a written laboratory pro- nance of quality performance depends greatly on tocol that includes appropriate control procedures. adherence to a well defined laboratory protocol. It As assays are developed and are then used in tran- 1s well established that the scoring must be of sitional studies, a stage occurs when the assay is coded samples, with the scorer 'blinded' to the sufficiently mature for it to be documented in such

220 Quality control of biomarker measurement in epidemiology a manner. Once this has been achieved, deviations 4. Incorporation of a biomarker assay into an from an agreed protocol are not acceptable with- epidemiological investigation requires close col- out formal investigation of the effects of change laboration between the epidemiologist and the on precision and, where possible, accuracy. This laboratory scientist, to take advantage of the discipline is essential for assays that need to be techniques that are available for minimizing maintained throughout a prospective cohort study and documenting laboratory error. with ririnlmum laboratory drift. The examples that have been discussed here Acknowledgements come from a range of laboratory disciplines, but it I thank those colleagues who have helped me to is clear that they ail reflect the need for the analyst! review quality assurance procedures in their scorer to be unaware of the status (exposed! specialist fields: Dr D. Bullock, Dr E. Bruner, Dr J. control) of any sample or of the value assigned to Cole, Dr H.T. Delves, Dr D.C. Lloyd, Dr D. 5huker, material circulated in interlaboratory comparison Dr J.R. Startin. 1 also thank the UK Department of studies. It may not be possible to disguise quality the Environment for support during this study. control material so that the laboratory scientist is unaware of its origin, but the submission of a References limited number of codedreplicate samples, which Aitio, A. & Apostoli, P. (1995) Quality assurance in bio- are not identifiable as such, will supplement in- marker measurement. Tozicol. Leif., 77, 195-204 house quality control procedures. Armstrong, B.К., White, E. & Sаracci, R. (1992) Principles The wide range of biomarkers that may be used of exposure measurement in epidemiology. Moпograрhs in cancer epidemiology makes it difficult to pre- Ергдетiы. Biosiadst., 21, 257-258 scribe rules for quality control and assurance in Bachorik, P.S. & Ross, J.W., for the National Cholesterol any detail. For some chemical assays, certified Eduction Program Working group on Lipoprotein reference material is available, while for some of Measurement (1995) National cholesterol education pro- the more biological assays it is difficult to see how gram recommendations for measurement of low-density such material could be produced. An interlabora lipoprotein cholesterol: executive summary. Clin. Chem., tory consensus can, however, often be agreed, 41, 1414-1420 using a pool of material, cryopreserved cells or at Broughton, EM. G., Holder, R. & Ashby, D. (1986) Long- least pre-prepared slides. It is clear that, although term trends in biochemical data obtained from two pop- the precision obtained by a laboratory may be ulation surveys. Ann. Clin. Biochem., 23, 474-486 good, any laboratory needs an external reference Bullock, D.G., Smith, N.J. & Whitehead, Т.P. (1986) to give it confidence that its measures have a valid- External quality assessment of assays of lead in blood. ity and are not drifting over the period of a study. CIin. Chem., 32, 1884-1889 Cole, J. & Skopek, Т.E. (1994) Somatic mutant frequency, Conclusions mutation rates and mutational spectra in the human There are a number of conclusions that emerge from population in vivo. Mutat. Res., 304, 33-105 this review of quality control of biomarker assays: Delves, H.T., Diaper, S.J., Oppert, 5., Prescott-Clarke, P., Periam,J., Dong, W., Colhoun, H. & Gompertz, D. (1996) 1. For a range of biomarkers, laboratory perfor- Blood lead concentrations in United Kingdom have mance depends on the use of an agreed written fallen substantially since 1984. Вт. Med. J., 313, 883-884 protocol that specifies details of reagents, sam- S.S., Garcia, О.F., Ramalho, A.T., Di Giorgio, M., Mit, pling and analytical procedures, calibration and Espinoza, M.E., Manzano, J., Nasazzi, N. & Lopez I. (1995) quality control methods. Intercomparison in cytogenetic dosimetry among five 2. Laboratories should take advantage of inter- laboratories from Latin America. Mutai. Res., 327, 33-39 laboratory comparison exercises, whether the Hulka, B.S. (1991), Epidemiologic studies using biologi- assays are being newly developed or are mature cal markers: issues for epidemiologists. Cancer Fpidendoi. assays in routine use. Biomarkers Prey., 1, 13-19 The laboratory scientist should, as a general 3. Lloyd., D.C., Edwards, A.А., Prosser, J.S., Вarjaktarovic, rule be unaware of the exposure/control status N., Brown, J.K., Horvat, D., lsmail, S.R., Koteles, G.J., of the samples submitted. Almassy, Z., Krepinsky, A., Kucerova, M., Littlefield, L.G.,

221 Application of Biomarkers in Cancer Epidemiology

Mukherjee, U., Natarajan, A.T. & Sasaki, M.S. (1987) A Rymen, T., (1994) History of BCR work on dioxins. collaborative exercise on cytogenetic dosimetry for sim- Fresenius J. Anal. Chem., 348, 9-22 ulated whole and partial body accidentai irradiation. Skipper, P.L. & Tannenbaum, s.R. (1990) Protein adducts Mataf. Res., 179, 197-208 in the molecular dosimetry of chemical carcinogens. National Cholesterol Education Program (1988) Current Carcinogenesis, 11, 507-518 status of blood cholesterol measurement in clinical lab- Stephens, R.D., Rappe, C., Hayward, D.G., Nygren, M., oratories in the United States. Clin. Chem., 34, 193-201 Startin, J., Esbo11, A., Carle, J. & Yrjanheikki, Е.J. (1992) Nielsen, P.S., de Pater, N. Okkels, H. & Autrup, H. (1996) World Health Organization international intercalibra- Environmental air pollution and DNA adducts in flou study on dioxins and furans in human milk and Copenhagen bus drivers—effect of GSTM1 and NКг2 blood. Anal. Chem., 64, 3109-3117 genotypes on adduct levels. Carcinogenesis, 17, Tb 1021-1027 тngvist, M., Magnusson, A,-L., Farmer, P.B., Tang, Y.-S., Jeffrey, A.M., Wazneh, L., Beulink, G.D.T., van tier Waal, Patterson, D.G. Jr, Isaacs, S.G., Alexander, L.R., Turner, H. & van Sittert, N.J. (1992) Ring test for low levels of N- W.E., Hampton, L., Bernert, J.T. & Needham, L.L. (1991) (2-hydroxyethyl)valine in human hemoglobin. Anal. Method 6: determination of specific polychlorinated Biochem., 203, 357-360 dibenzo-p-dioxins and dibenzofurans i blood and adi- п Vineis, P., Schulte, P.A. & Vogt, R.F., Jr (1993) Technical pose tissue by isotope dilution-high resolution mass variability in laboratory data. I : Schulte, P. . & Perera, spectrometry. In: Rappe, C., Buses, H.R., Dodet, B. & п А F.P., eds, Molecular EpidemioIogy Prin ipies and Practices. O'Neill, LK., eds, Enviro r ent I Carcinogens, Methods с пп а San Diego, Academic Press of AпaIysis aid Exposure Measurement, Vol. 11, PoIy- chIoппated Dibenzodioxins and Dibenzo furaпs ('ARC Sсi- Westgard, J.O., Barry, P.L., Hunt, M.R. & Groth, T. (1981) eпtific Publications No. 108), Lyon, 'ARC, pp. 299-342 A multi-rule Shewart chart for quality control iii clinical chemistry. Clin. Cher., 27, 493-501 Perera, F.P., Herninlrski, K., Gryzbowska, E., Motykiewicz, G., Michalska, J., Santella, R.M., Young, T.-L., Dickey, C., Westgard, J.O., Petersen, P.H. & Wiebe, D.A. (1991) Brandt-Rauf, P., DeVivo, L, Blaner, W., Tsia, W.-Y. & Laboratory process specifications for assuring quality in Chorazy, M. (1992) Molecular and genetic damage in the US National Cholesterol Education Program. Clin. humans from environmental pollution in Poland. Chem., 37, 656-661 Nature, 360, 256-258 Whitehead, Т.P. (1977) Quality control in clinical chem- Phillips, D.H. & Castegnaro, M. (1993) Results of an istry. New York, John Wiley interlaboratory trial of 32P-postlabelling. In: Phillips, Yrjanheikki, E.J. (1989) Levels of PCBs, PCDDs and D.H., Castegnaro, M. & Bartsch, H., eds, PostlabeIling PCDFs in breast milk. Environmental Health Series, 34, Methods for Detection of DNA Adducts (IARC Scientific World health Organization, Regional Office for Europe. Publications No. 124), Lyon, ‚ARC, pp. 35-49 Copenhagen, FADL Publishers Rappe, C., Buser, H.R., Dodet, B. & O'Neill, Т.K., eds. (1991) Environmental Carcinogens, Methods o fAnalysis and Exposure Measurement, Vol. 11. Paiychiorinated Di еггzodioxins and Dibenzo furans (IARC Scientific Publications No. 108), Lyon, IARC Robinson, D.R., Goodal, K., Albertini, R.J., O'Neill, J.P., П. Gompertz Finette, B., Sala-Trepat, M., Moustacchi, E., Tates, А.D., MRC Institute for Environment and HealIb, University of Leicester, Beare, D.M., Green, М.H.L. & Cole, J. (1994) An analysis of in vivo hrpt mutant frequency in circulating T-lym- Leicester LE1 911N, UK phocytes in the normal human population: a comparison of four data sets. Mutat. Res., 313, 227-247

222 Applicalion of Biomerkern 1п Cancer Epidemiology Toniolc, P., Boffetla, P., 5huker, D.E.G., Rothman, N., Hidka, 6. апд Pearce, N., eds 'ARC tlnjerrOtc PuЫicatlons No. 142 International Agency for Research on Cancer, Lyon, 1997 5ampfe collection, processing and storage M.T. Land and N. Caporaso

We review issues related to the inclusion of biospecimens in epidemiological studies. Technical advances and the revolution in molecular biology have rendered the use of biomarkers increasingly feasible in epidemiological investigations, however the cost and complexity require interdisciplinary expertise and careful attention tо methodological detail in order to ensure validity.The widespread banking of biospecimens for long-term (cohort) studies requires special attention to be paid to these issues. Blood, urine and tumour tissue are in common use in medicine and at least some aspects of sample handling derives directly from this clinical experience, although special considerations apply in the epidemiological setting. An increasingly broad array of biospecimen types have been studied, including exhaled air, nail clippings, buccal cells, saliva, semen, faeces and breast milk. Relevant issues in the processing, storage, shipping, timing of collection and safety procedures are examined in terms of their potential to distort results. The role of carefully developed quality control protocols is emphasized. In order to take full advantage of the opportunities afforded by the use of biomarkers in epidemiological studies, careful attention to biospecimen processing, the stability of the biomarker and the precautions to he taken during transportation and storage of samples is necessary.

The appropriate collection, processing and storage time and resource investment warrant careful of biospecimens form an essential but often over- planning to ensure efficient, safe and focused looked component to any study that includes bio- procedures (Kaaks et a?., 1994) markers. If the type or the quantity of the biologi- It is critical to collect information linked to sam- cal substrate is not adequate for the marker to be ples in order to interpret the markers оptiтаlly: analysed, or if it is less than optimally processed at time and date of collection, recent diet and sup- any stage in the sequence between collection (han- plement use, reproductive information (i.e. men- dling, labelling, processing, aliquoting, storage) strual cycle), recent smoking, current medication and assay, analyses and subsequent epiderniologi- use, recent medical illness, storage conditions, etc. саl inference will suffer. Contamination, volatiliza- can be crucial far later interpretation of results. For tion (e.g. rnethylethylketone), the short half-life of example, information on current medication use some substances (e.g. nicotine) and the potential may be crucial to the interpretation of a metabolic that the collection procedure itself influences the phenotype, i.e. quinidine use and the debrisoquine substance to be measured (e.g. chromium levels phenotype. from stainless steel needles used for venepuncture) Overall quality assurance involves the system- are design issues that must be anticipated in atic application of optimum procedures to ensure advance (Bernard, 1995). Since inaccuracy or valid, reproducible and accurate results. The most imprecision can result in increased variation, mis- fundamental component is probably the adoption dassification and loss of power, this is a major con- of standardized operating procedures for each cern to epidemiologists. Iп case -control studies, if aspect of biospedmen handling. As part of a qual- case samples are handled differently from control ity assurance plan, stored specimens could be samples, differential rпisgаssшсаtioп may occur. tested on a regular basis to detect sample deterio- These issues take on special importance in cohort ration. Another aspect of quality assurance studies, where large numbers of samples and the involves aliquoting material into multiple small

223 Application of Biomarkers in Cancer Epidemiology vials and storing each person's specimen in at least tissue. The usefulness of these collections is greatly two different physical locations to reduce the like- enhanced if there are associated epidemiological lihood of loss of а large volume of specimen as a data. For example, a series of lung tumour speci- result of accidental thawing due to freezer failure. mens would be of much greater interest if age, gen- When samples are used later, they should be der, smoking, occupational exposure and clini- selected from specimens that received the same cal/survival information were also readily avail- treatment throughout the storage process or the able. saine variations in handling; this should be con- In general, biospecimens that can be collected trolled for in the statistical analysis. safely with little discomfort to the donor hold the Careful records of disbursements are also greatest promise for use in large-scale studies. A list needed, i.e. which specimens, how much material of biospecimens suitable for epidemiological stud- remains, and documentation on factors such as ies follows. This Iist, although far from being thawing which could influence future uses of the exhaustive, can provide an insight into adequate material. Automatic edit checks (which can only procedures for sample collection, processing and be overridden by specific authorization of the prin- storage. cipal investigator) may be used to block disburse- ment of the last specimen from an individual. Blood In this work, we highlight some relevant issues Collection of blood provides material that can give regarding the collection of biological specimens physiopathological and genetic information, can in epidemiological studies. The work does not reflect the biological rhythms (circadian, men- attempt to provide an encyclopaedic review of all strual, etc.) of the host, and can reveal recent or maтkeтs, specimen types or collection techniques, remote exposures or their sequelae. Detailed but rather attempts to identify some central issues descriptions of techniques relevant to clinical stud- along with examples. ies are available in clinical laboratory manuals and so this treatment will only highlight major issues Types of biospecimens of concern to the epidemiologist. The use of skilled Various types of biological materials can be col- technicians and precise procedures when perform- lected for epidemiological studies, depending on ing phlebotomy are important because painful, the study design, the markers of interest and the prolonged or repeated attempts at venepuncture availability of an assay. With the rapid develop- can cause patient discomfort or injury and result in ment of molecular biology techniques, it has less than optimum quality or quantity of sample. become critical to collect samples, planning not As an example, prolonged venepuncture can only for the main bioniarkers of interest, but also induce the release of prolactin, or increase white to process and store material in a way that allows blood cells (Stаtlarrd et al., 1978). Obtaining blood for new biomarkers to be tested in the future. through too narrow a needle can result in haemol- Several prospective studies have considered ysis, with distortion of cell counts and electrolytes blood specimens stored at low temperature in bio- (especially potassium). The donor's position while logical banks as a source of information. Biological providing the specimen may also influence analyte markers from samples collected before the onset of levels; e.g. serum cholesterol may be higher in disease and stored until clinical expression may standing subjects compared with supine subjects provide essential information on exposure to because of orthostatic decreases in plasma volume endogenous and exogenous factors not biased by (Kieldsen et al., 1983). Prolonged use of a tourni- the metabolic effects of ilIness (Winn et al., 1994). quet (i.e. greater than 1 min) can result in haemo- Other biological specimen banks have been estaЬ- concentration. Any of these factors can result in Iished to store specimens from persons who have pre-analytic variation in measurements. Thus, already developed a disease, to characterize the his- careful review of general and specific factors that tory of the disease. Many collections of pathologi- may impact on a specific assay is mandatory before cal material exist, and have increasing value with study initiation. the availability of improved techniques to extract Blood is a great source of materials that can be and characterize DNA from tumour and normal collected for different purposes. Plasma/serum,

224 Sample collection, processing and storage lymphocytes, monocytes, erythrocytes, granulo- arid is used cliпicaцy in haematology studies, but cytes and platelets can be obtained through can be influenced by Mgz+concentration. It is well venepuncture and appropriate separation of blood suited to DNA-based assays but has problems for components. cytogenetic analyses (increases of sister chromatid To avoid multiple venepunctures when a test exchange, decreases of mitotic index, etc.). requires a large blood sample, an evacuated tube Citrate also works by calcium chelation and is system with interchangeable glass tubes can be used in coagulation studies and blood barildng. It is used. Evacuated tubes are commercially prepared optimal for assays conducted on lymphocytes and with or without additives and with sufficient vac- DNA. uum to draw a predetermined blood volume (2-20 There have been a limited number of studies m1 per tube).The different tubes are generally indi- comparing various anticoagulants in investigation cated by their colour-coded stoppers: settings. For example, a study on the effects of time, temperature and additives on a functional assay of • Red-top tubes contain no additives. These Cl inhibitor (a component of the complement cas- tubes are used for tests performed on serum cade) showed that plasma containing heparin or samples. polybrene interfered with the functional assay; on • Lavender-top tubes contain EDTA, commonly the other hand, EDTA-treated or citrated plasma used clinically for complete blood counts (CBC). and serum kept at room temperature were ade- • Green-top tubes contain heparin (sodium, quate for the assay (Nielsen et ai., 1994). lithiu or ammonium). Although there are anecdotal reports of occa- • Blue-top tubes contain sodium citrate and cit- sional problems with heparin in PCA assays, stud- ric acid. Draw volume may be 2.7 or 4.5 m1. ies generally find that there are no major differ- • Black-top tubes contain sodium oxalate. Draw ences in the use of EDTA or heparin. For example, volume may be 2.7 or 4.5 ml. Storch et al. (1994) compared the two anticoagu- • Yellow-top tubes contain acid-citrate-dextrose lants for inhibitory effects on the detection of cyto- (ACD) solution. Draw volume is 12 ml. megalovirus from washed leukocytes in specimen • Grey-top tubes contain a glycolytic inhibitor transport tubes. Evaluation was made by the cen- (such as sodium fluoride, powdered oxalate or trifugation/shell vial culture techniques, the ррб5 glycolytic/microbial inhibitor). Draw volume atigenaemia assay and PCR. For each assay, the results with heparin and EDTA were equivalent. may be З to 10 ml. These tubes are used most often for glucose determinations in serum or There are other anticoagulants useful for special plasma samples. applications. If only serum is needed, there is no need for anticoagulants. To reduce contamination, Specific advantages and disadvantages of vari- it is best if serum is separated from other blood ous anticoagulants are described below. Nutrients components as soon as possible. and other substances often require special addi- Many newer methods have been described that tives to allow analyses, i.e. metaphosphoric acid to allow the collection of small quantities of blood measure vitamin C content. adequate for the characterization of DNA. These There are important differences between the dif- methods do not require venepunctuue or low- ferent anticoagulants. Heparm acts by binding to temperature conditions during collection, storage antithrombm III and thus accelerates the inactiva- and shipping. Examples of these are dried blood tion of thrombin and other clotting factors. The dis- specimens which are derived from whole blood advantages of heparin include the presence of spotted directly—or anticoagulated with EDTA impurities and the non-uniform position of sub- before spotting—onto clean slides and air dried at stituents, its ability to bind to many proteins, the room temperature. The slides with blood smears can potential to bind to platelets and cause agglutina- be transported or stored at room temperature, and tion, its potential loss of anticoagulant effect in serve as a good source of high-molecular-weight aged blood, and its influence on T-cell proliferation. DNA (Aggarwal et aL, 1992). Low-molecular-weight heparin avoids some of A quantity of 50 µ1 of dried blood can provide these problems. EDTA works by calcium chelation 0.5 µg DNA, sufficient for multiple PCR-based

225 Application of Biomarkers in Cancer Epidemiology

assays (McCabe et al., 1987). DNA is stable on cot- merging aliquots of specimens from persons ton cloth for at least 4 years (Gill et al., 1985). within a subgroup and testing the combined sam A study planned to verify the stability of dried рlё to obtain a group-specific average value. This blood spot specimens for the detection of human approach requires only a small number of labora- immunodeficiency virus DNA by PCR techniques tory tests to be performed, but yields only a mean showed that treatment at 37°С and 60% humidity value without a variance or information about the for 7 days, storage for 12 weeks at 22°C, followed distribution of results. by a freeze/thaw cycle had no adverse effect on DNA can be extracted from whole blood, leuko- PCR reactivity when compared with reference cytes and, as was recently demonstrated (BIomeke spots stored at -20° C. These findings suggest that et al., 1996), serum, plasma or paraffin-embedded dried blood spots are a powerful resource for test- tissues. Smа11 amounts of DNA from blood (suffi- ing for HIV by PCR, especially in remote areas cient for PCR-based assays) can be obtained where refrigeration and immediate sample pro- through finger-pricks (Guthrie cards), and the sub- cessing are unavailable (Cassol et al., 1992). jects can do the collection themselves at home. None the less, some discomfort is often felt with Blood components. From 10 ml of blood, the fol- this approach and the participation rate can be lowing quantities are obtained: reduced. Dried blood spots are very convenient for • plasma (or serum)-6 or 7 ml storage and shipping, and may be optimal for a • lymphocytes and mononuclear cells-10-20 large-scale epidemiological study. Whole blood x 106 cells/ml obtained directly from finger-pricks may have sev- • erythrocytes (RBCs) and other cells-5 x 106 eral applications, e.g. it can be used for analysing ce11s/µ1; 10-15 mg lb. complement activation using an ELISA enzyme immunoassay method (Chang aid Lister, 1993). Mononuclear leukocytes are the only cell type in blood capable of growth; they can be cryo- Processing. The sample processing depends on the preserved for the establishment of cell lines. marker needed. Investigators must design studies Cryopreservation permits cell viability (for tissue to fit the requirements of the critical biomarkers. culture-based mutation assays, transformation for For catecholamine measurements, for example, gene mapping, etc.) and can be the only source to blood should be centrifuged within 1 h of collec- measure RNA after PHA induction (RNA is difficult tion. Once plasma is prepared, catecholamines are to assay in quiescent cells because levels of activity stable for 1 day at 20°C, 2 days at 4°C, 1 month at are absent оr low). Granulocytes can serve as a -20°C, and up to 1 year at - 70°C (Boosma et al., source of DNA without sacrificing the lympho- 1993). A delay in blood processing may affect the cytes. Erythrocytes, stored after washing with assay results. For example, blood samples held for physiological saline, can be useful to study adducts 2 days at room temperature showed a mean 81% of haemoglobin. Plasma/serum can be used to decrease in cells positive for cytomegalovirus anti- measure microanalytes, diet components, vita- genaemia (Landry et al., 1995); on the contrary, a mins, xenobiotic exposures and so on. Plasma can delay of up to 24 hours in blood processing did not be obtained from an anticoagulated blood sample significantly change the specific activities of aryl through separation from cell components. It per- sulfatase A and cerebroside-beta-galactosidase in mits the analysis of coagulation factors, while both Ieukocytes and lymphocytes (Shah et al., serum permits better estimate of antibodies, nutri- 1995). The stability of assays in relation to time ents, and lipid and lipoprotein measurements. Loss and temperature of storage has not always been of plasma volume because of filtering after fibrin well documented, but should be considered in the formation is a disadvantage of plasma storage over context of specific studies. In one study, it was rec- serum storage. ommended that serum fatty acids be measured Pooled aliquots of serum specimens have been within 2 weeks at 4°C, within a few months at -20°C used (Wahrendarf et al., 1986) for nutritional or and within 1 year at -80°C to estimate the compo- other biochemical studies, e.g. 11V antibody test- sition of the major fatty acids (Umemura et al., ing (IcIahon et al., 1995). The approach requires 1991). Blood samples treated with Triton X-100,

226 Sаmple collection, processing and storage

ESTA and sodium fluoride may be stored up to 4 to appear following exposure than DNA-base weeks without aрpтeciаые effect on measured changes. blood lactate concentration (Hî11, 1995) It is critical to obtain information (in addition to the usual questionnaire typically administered) Storage. It is critical to maintain careful records of at the time of biospecimen collection to aid inter- the identity and location of all materials, with par- pretation. Typical information will be assay-spe- ticular .attention to storage history, occurrence of cific, but is likely to include, as a minimum, time temperature fluctuation and monitoring of stored and date of draw, volumes and type of specimen, control specimens, in order to check the effects of medical illness (currentlremote), last food con- storage duration. For example, samples stored on sumed (type aid time since collection), medica- top of a mechanical freezer may be exposed to tion use, reproductive information—i.e. time since more extreme temperature fluctuations then those last menstrual period—time since last cigarette, stored at the bottom. This cari lead to deteriora- and alcohol intake. Information will differ depend- tion of sample over time and can bias the analyses. ing on the study setting and the target population. In prospective studies based on storage of Some studies may require that specimens be col- biological samples at low temperature, a crucial lected from the same persons at several time question is to evaluate whether long-term preser- points; collection and related data should be stored vation of samples is able to affect the categorization with the study database. of the subjects involved. A recent study showed Specific Ыomaтkers will each have a precise that estradiol, free and total testosterone, and pro- time dependence on exposure; examples include lactin in serum and plasma samples maintained sister chromatid exchange (days), chromosome almost the same rank by hormonal concentration aberrations (around 6 weeks post-exposure), throughout a 3-yеаr period of cryoconservation at adducts Ivaries by type, e.g. DNA and albumin: -80°C (Bolelli et al., 1995). In another study, 25 days; lb: 120 days—see Environmental Health aliquots from 40 ml plasma pools preserved with Perspectives, Vol. 103 (Suppl. 3), 1995], and various metaphosphozic acid were assayed for their ascor- mutations (6 weeks to б months after exposure, bic acid values after 12, 24 and 42 months of storage but also dependent on repair activity). Similar con- at -70°C. Similarly, aliquots from 16 plasma pools siderations apply to nutrients and hormones, but were assayed for values of retinol, several carotenoids are less relevant to repeated or remote exposures. and two tocopherols for a period of storage at -70°C up to 4.3 years. No indications were found Urine of important losses of these antioxidant micronu Because urine is an ultrafiltrate of the plasma, it trients during storage (Comstock et al., 1995) can be used to evaluate and monitor body homeo- However, in a study on maternal screening for stasis, metabolic disease processes, exposure to three markers, immediate freezing of serum and xenobiotic agents, mutagenicity, exfoliated cells, subsequent thawing resulted in a significant increase DNA adducts, etc. Urine specimens are usually in beta-hCG and unconiugated E3 levels, but no readily obtainable, although sometimes its collec- change in AFP levels. AFP levels were influenced by tion is felt to be more inconvenient than blood centrifugation status, and all three analytes were collection. The type of specimen selected and the influenced by refrigeration status (Lantz et aг., 1995). collection procedure used depend on the tests to be performed. There are basically four types of Timing. The timing of sample acquisition can pro- urine specimens to be collected for epideniiologi- foundly influence the interpretation of results. cal studies: first morning, random, fractional and This is especially evident for those hormones timed. The ideal specimen is adequately concen- which have hourly, daily and/or monthly cycles. trated to ensure the detection of analyses of inter- Both timing of exposure and timing of biospecimen est (for a review, see Brunzel, 1994): collection need to be considered; i.e. mutations • To collect a first morning specimen, the sub- cannot be compared with chromosomal aberra- ject voids before going to sleep and, immedi- tions on the same sample, since, for example, ately upon rising, collects a urine specimen. The chromosomal changes may require a longer time specimen must be preserved if not delivered

227 Application oг Biоmaгkers in Cancer Epidemiology

within 2 h of collection. First morning speci- In epidemiological studies, the period of urine mens are ideal for substances that require con- collection used to measure excretion varies de- centration, and white cells, red cells and casts pending on the marker to be measured and the are more stable iп concentrated acidic urine aim of the study. For example, using overnight col- specimens. However, they are inadequate for lection to measure catecholamines may result in a cytology studies; in fact, the epithelial cells may much larger sample size requirement to achieve undergo degeneration, and salt crystallization the same power as a 24-h collection. However, 24- during processing may have adverse effects on h collections incorporate measurement errors that cytology studies. are not present in overnight collections (White et • Random urine specimens can be соцесtед at aI., 1995). In reference to optimal storage condi- any time. These specimens are usually satisfac- tions, a study by Boosma et a1. (1993) showed that tory for routine screening and for cytology stud- catecholamines are stable at 4°C for 1 month in ies. One method to increase the ce11ularity of unpreserved urine, and for 4 months in urine pre- the urine specimen is to have the subject drink served with EDTA and sodium metabisulfite. In a lot of water 2 h before collection and exercise acidified urine, catecholamines were nearly for 5 min by skipping or jumping up and down unchanged after 1 year at 4 and —20°C. prior to specimen collection. Urine processing can influence the results of • Fractional collections are used to compare the various tests. For example, freezing at —20°C and concentration of an analÿte in urine with its storing of urine samples prior to assessment of concentration и the blood. These specimens albumin concentration can affect the absolute vaI- are also termed 'double-voided specimens', ues obtained (Shield et al., 1995). Foi the quantita- because the first morning urine (containing tion of glutathione transferases by an ELISA proce- solutes and metabolites from the evening meal) dure, storage of urine samples requires the pres- is discarded, but the second urine excreted (fast- ence of low detergent concentrations, such as ing urine specimen) is collected. Tween 20, which both stabilizes the enzymes and • Timed collections, usually done over 12-24 h prevents their absorption by plastic test tubes periods (or even longer), eliminate the need to (sundberg et al., 1995). determine when excretion is optimal and allow The use of urine requires decisions to be made day-by-day comparison of excretion patterns. on the appropriate time frame for measurement of On day 1, at the start time, the subject empties numerous analyfes, microbial contamination, the his/her bladder. For the next 24 h, all subsequent cost of storing large volumes of material and the urine must be collected in the container. On paucity of studies on the effect of long-term storage day 2, at the end time (the same hour as the start on qualitative or quantitative analyte detection. time), the subject empties the bladder and includes this specimen in the collection. Only Tissues one first morning sample must be included. Tumour pathology samples are required to confirm Urine collections should be kept on ice or clinical diagnoses by histological analysis. Iп refrigerated throughout the duration of the recent years, there has been increasing interest in collection. examining tumour characteristics at the chromosomal and molecular levels. This generally requires Plastic or glass containers must be clean and collecting more material than is necessary for dry, and have a 50-3000 ml capacity, a wide mouth pathological evaluation. When possible, the tissue and a leak-proof screw cap. Depending on the ana- samples should contain tumour as well as normal lyte to be measured, a preservative maybe needed. tissue to permit investigation of the different char- The type of preservative may differ according to acteristics of the two tissues. For tissue blocks, col- test methodologies, time delay and transport con- lections of stained and unstained tissue samples ditions. In the laboratory, total volume must be from the same individual are ideal. recorded, the specimen well mixed to ensure In planning any human tissue banking effort, homogeneity, and aliquots removed for the appro- one should consider which tissues might already priate test. be, or will potentially be, available, such as dis-

228 sample collection, processing and storage carded tissue blocks or diagnostic specimens from atively stable deposit of triglyceride and fat-soluble surgery (informed consent may be required) Some substances, such as fat-soluble vitamins (i.e. vita- of these tissues may be less than ideal for specific mins A and D) and pesticides. Halogenated hydro- ends, e.g. forrualin-preserved samples have altered carbons may be measured in concentrations hun- chemical constituents and PCR assays are difficult. dreds of times greater than those in blood of the However, with development of new analytical same individuals. As a tissue, it represents the techniques, new uses may be found for any tissues greatest reservoir of carotenoids in the body, and that might be available. Several extraction tech- the tissue composition also reflects long-term niques have been proposed, e.g. the sonication dietary intake of essential fatty acids (for a review, method, which takes only 30 min from start to fin- see Kohlmeieт & Kohlrneier, 1995). Samples are ish to extract DNA from fresh, frozen and forma- collected from the upper outer quadrant of the buttock while the subject is lying face down. No 1in-fiк ed, paraffin-embedded tissue specimens (Heller et aI., 1992). Although available specimens local anaesthesia is needed. After disinfection of are often limited to formalin-fixed, paraffin- the skin with alcohol, the subject is asked to tense embedded blocks, frozen tissues have distinct the buttock to delineate muscle and fat. The upper advantages. Snap freezing of tissues is especially outer quadrant is grasped between the finger and suited for RNA extraction. Recently, a new method the thumb of one hand. A Luer-lock needle 1s has been developed for preparing normal or tumour inserted at a 450 angle, after which a vacuum tube tissue for RNA recovery; the tissue sample is placed is connected to the needle. The needle is then in a transparent bag which does not break when moved in and out of the adipose tissue, at which submerged in liquid nitrogen. While frozen, the tis- time some fat collects at the top of the Luer adapter sue is crushed with a hammer. After the specimen is between the needle and the tube. The adapter is completely lysed, RNA can be extracted even from capped and immediately frozen at —80°C very limited tissue samples (Gramza et a2., 1995) (Kohlmeier & Kohlmeier, 1995). As an example of the influence of storage temperature on a specific assay, gastric rnucosal biop- Bronciioalveolar lavage (BAL) sies stored at 4°C foi 1 and 2 weeks resulted in Broncho alveolar lavage fluid has been used for the recovery of 81% and 19% of Helicobactor pylori, many years in occupationally related epidemiolog- respectively. Storage at —20°C improved yields to ical studies. A typical example is the use of BAL to 100% and 57% after 4 and 12 weeks, respectively. assess and quantify asbestos exposure through Recovery improved still further at —70° C (Han et identification of asbestos bodies and cell content al., 1995). In general, it is best to use the lowest in the alveolar fluid (Orlowski et al., 1994). storage temperature practical, given cost and sam- Induced sputum samples and bronchoalveolar ple size constraints. lavage fluid (BALF) can also provide sufficient DNA A novel method has been developed for pro- for PCR-based assays. As an example, a study by cessing biopsy specimens for histochemical and Liesnard et al. (1994) showed that PCR can be immunohistochemical analyses by combining applied to BALF as a rapid method to detect freezing with low-temperature plastic embedding. cytomegalovirus infection. The method avoids the need for tissue fixation and Bronchoalveolar lavage is performed through a combines the superior morphological preservation fibre-optic bronchoscope inserted transnasally of fixed embedded tissue with the reactivity of crуо- after topical application of local anaesthesia. Sterile solution (200-250 ml) at 37°C is injected by 50 ml stаt sections. Biopsy specimens are stored at room temperature without loss of tissue-specific charac- bolus into a dependent subsegmental bronchus of teristics during storage (Murray & Ewen, 1991). the right middle lobe or lingula. The bronchoalveolar fluid is then recovered by mild Adipose tissue aspiration and can be ready for centrifugation, cell Adipose tissue aspirations maybe quite feasible for count, cell digestion, preparation of slides for the subject and involve low risk. Tissue samples asbestos bodies count through light or scanning can be analysed for assessment of prior exposures electron microscopic analyses, etc. (Bell et aй., in epidemiological studies. This tissue offers a rel- 1981; Roggli et al., 1994).

229 Application of Biomarkers in Cancer Epidemiology

Exhaled air Hair samples have been proven to be valid bio- Collection and analysis of exhaled air has been logical samples for nicotine measurement in esti- done for years to evaluate exposure to different mating average eпviroпmeпtaI tobacco smoke substances, particularly solvents such as benzene, (ETS) exposure in children. The hair nicotine lev- styrene and tetrachloroethylene. In addition, els were shown to be well correlated with cotinine exhaled air samples have been used as a source of creatinine ratios in urine from the same individu- exposure and susceptibility biomarkетs. An exam- als (Nafstad et al., 1995). ple of this is provided by the [3-13C-methyl] Moreover, hair and urine analyses are comple- caffeine breath test, developed to measure the mentary tests for establishing drug use. Hair analy- Р4501А2 activity, which can increase after expo- sis provides long-term information, from months sure to polybrominated biphenyls and dioxins to years, concerning both the severity and pattern (Lambert et al., 1990). After ingestion of a labelled of drug use. In contrast, urinalysis only indicates dose of 3 mg/kg caffeine, aliquots of end-tidal drug use that has occurred within the last 2-3 days. breath samples are collected at different times and External contamination of hair by drugs present in placed in vacationers for storage and transport. the environment (e.g. smoke) is the main problem Analysis by differential mass spectroscopy of the of hair analysis. The problem, however, can be samples gives an indirect measurement of the effectively avoided by washing the hair specimen, Р4501А2 activity. Other examples of tests that by kinetic analyses of the wash data and by exploit labelled compounds with detection of the metabolite measurement. The possibility of bias target compound in expired air include breath due to race and/or hair colour is avoided by the methane and 12 (fermentation), and breath urea exclusion of melanin from hair analysis (Du Pont (presence of urease positive organisms such as & Baumgartner, 1995) H. pylori). For collection of exhaled breath, various Finally, hair roots can be an optimal source of similar methods of sampling have been developed; DNA for PCR analysis and permit easy collection, for example, a method capable of measuring transport and storage and low overall costs sub-ppb levels of volatile organic compounds, (Thompson et aI., 1992). employing Tenax sorbents to collect breath samples from a Tedlar bag that has been filled by the subject Na1I clippings exhaling through a two-way mouthpiece. The Toenail or fingernail clippings are obtained in a subject inhales pure humidified and charcoal very easy and comfortable way, and they do not scrubbed air from a cylinder (Wallace & PeIlizzari, require elaborate processing, storage and shipping 1995). conditions and are thus suitable for large epidemiological studies (Garland et al., 1995). Toenail clip- Hair pings have been used for studies of trace elements Hair is an easily available biological tissue whose and for measurement of selenium levels (which typical morphology may reflect disease conditions reflect selenium intake) in order to investigate the within the body. It provides a permanent record of association between selenium status and cancer trace elements associated with normal and abnor- risk (Garland et al., 1993; van den Brandt et al., mal metabolism, as well as those assimilated from 1993). Another study showed that arsenic levels in the environment (Srivastava & Gupta, 1994). In fingernails can be a biological indicator of expo- fact, human hair analysis has proved to be a well- sure to arsenic (Agahain et aI., 1990) suited biological marker of occupational and envi- Nail clippings may be less likely to be contami- ronmental exposure to toxic metals (although, nated by environmental factors, handling procedures because of external contamination, they are not от cosmetics (Hambidge, 1982), but involve more useful for most nutrients). The method of hair complicated washing, specimen hydrolysis and analysis can be suitable for use in pilot prospective matux problems for the chemist (Winn et al., 1990). studies. If an excessive exposure is detected, analysis of conventional biological substrates, such as Buccal cells blood or urine, is recommended in order to verify As an alternative to using blood obtained by the exposure accurately (Bencko, 1995). venepuncture, cells may be obtained by an oral

230 sample collection, processing and storage rinse. Human cheek cells were first used as a non- brushes and female Dacron swabs, and storing invasive method for detecting tissue lipid profiles them at 4°C for 1 month, 2 weeks, 1 week and 3 in nutritional studies (Iclurchie et aL, 1984). days. DNA was subsequently prepared from these More recently, а'swish and spit' technique to coi- samples and amplified alongside freshly prepared led nucleated cells as a source for DNA has been re- buccal cell DNA. There was no significant differ- ported. For this technique, the subject vigorously ence in the yield of amplification products among swishes isotonic saline in the mouth and expecto- the different samples. To determine how stable the rates it into a collection container. DNA is ex- cheek cells would be when shipped under various tracted from the buccal cells and can provide weather conditions, samples were stored for 3 days excellent templates for PCR-based assays. In this in a —20°C freezer, a 37°C incubator and in the air- study, the integrity of the specimens was un- space of a 37°C water bath, in order to simulate the affected by storage at —20, 4, 25 or 37°C. Thus, this effect on buccal cell stability of very cold, hot or process may be extremely useful for large- humid conditions. Amplified product yields were scale studies that require DNA from subjects geo- equivalent to those of freshly collected and graphically distant from the research site (Hayney extracted samples. etal., 1995). Buccal mucosal transudate collected by sait- The oral rinse method is perhaps the most impregnated buccal swab was shown to provide a extensively used non-blood-based sampling tech- minimally uncomfortable measure of cocaine use nique. However, it involves liquid sample handling in a preliminary study of 44 subjects (Leonard and requires an additional centrifugation step to et aL 1994). Buccal cells can be used to count spin down the cells, increasing the time for DNA micronuclei. The proportion of exfoliated buccal preparation. Buccal cells may also be collected on cells with micronuclei offers the opportunity to cytology brushes and swabs for use in PCR-based assess sensitivity to gamma radiation and geno- assays. For cytology brushes, there is a variety of toxic compounds and, in addition, to monitor the collection approaches. Generally, just before spec- effectiveness of cancer intervention strategies imen collection, the subject should rinse the (Belief et ai., 1995). mouth with water to reduce bacteria interference. We have successfully used a technique involving Sагivа 30 s of gentle rubbing over the inner buccal surface Saliva can be an efficient, painless and relatively on each side of the mouth. This method would not inexpensive source of biological materials for cer- be appropriate in subjects with mucosal lesions. tain assays, and provides a useful tool for measur- Alternatively, a sterile metal spatula can be used to ing endogenous and xenobiotic compounds. scrape the buccal mucosa firmly but carefully so as Several devices, such as non-covered cotton roll, not to cause pain or bleeding; the buccal material polypropylene-covered polyether roll and paraffin on the spatula is then spread across the surface of wax chewing stimulation, have been developed a clean microscope slide. The prepared slide is to collect effective saliva samples. The different placed in fixative for approximately 1 h and then devices, which have been used to measure corti- allowed to dry at room temperature. Transport costeroids (Kathol et al., 1995), antibodies to may be done in an appropriate box at room tem- human immunodeficiency virus type 1 (I4IV-1) perature. Cheek cell samples and DNA prepared (Granade et al., 1995), and other compounds, from the swabs are highly stable. generate different data. Care should be taken when different devices are employed (Lenander- In а blind study comparing the analysis of 12 mutations responsible for cystic fibrosis in multi- Lumikari et al., 1995). Recently, a technique has plex products amplified with DNA from both been described to determine the СУР 1А2 phenot- blood and buccal cell samples from 464 individu- ype following a caffeine dose using saliva (Fuhr als, there was a 100% correlation of results for & Rost, 1994). blood and cheek cells collected on cytology Cotinine determination in saliva is considered a brushes for use in genetic testing (Richards et al., reliable marker for both environmental tobacco 1993). In this study, the stability of cheek cells was smoke exposure and active smoking (Samet et а!., evaluated by collecting cells on both cytology 1988).

231 Application of Biomarkers in Cancer Epidemiology

Breast milk exceeding 7, should precede the collection. The Several studies have been сопдисtед using breast specimen is collected by subjects through mastur- milk from lactating mothers to measure hormones, bation, and the entire ejaculate is collected in a epoxides of cholesterol (Wrensch et al., 1993), clean sterile plastic or glass container. Lubricants and exposure to chemicals, i.e. polychlorinated ordinary condoms should not be used because of biphenyïs (Becker et al., 1995, Schlaud etaL, 1995), their spermicidal properties. The collection container pesticides (Rogan & Ragan, 1994) and biological should be at room temperature or warmed before contaminants, i.e. aflatoxin M1 (E1-Nezami et a1., collection, should reach the laboratory within 1 h of 1995) or BorreIla burgdor fеri (Schmidt et al., 1995). collection, should be protected from extreme tem- In studies of the association between selenium peratures (i.e. maintained between 20 and 40°C), and levels and cancer risk (5anz & Diaz, 1995), breast should be labelled with the time of actual collec- milk has been used to estimate serum selenium lev- tion, a crucial factor in evaluating liquefaction and els, because it is well correlated to selenium intake. sperm motility (Woild Health Organization, 1992). Milk samples are collected when beast feeding is established. The samples are obtained by expres- Temperature sion, either manually or by vacuum pump. Before Specimen collection requires storage systems that expressing the milk, the subject should wash the are capable of maintaining the optimal tempera- nipples, and then the milk should be expressed ture for the diverse types of specimens: directly into autoclaved glass bottles, which are • • -20°C: certain items stable, i.e. urine opened and used for one expression only. After col- • -70°C: lection, milk can be kept in a domestic refrigerator —cell viability limited (not optimal) until it is sent to the laboratory (within 24 h), —DNA stable where it can be deep frozen for long-term storage. —serum stable —most hormones stable Faeces —most vitamins stable Certain ceI1s of interest (Nair et al., 1994), infectious • -120°C: hormones, carotenoids, other nutrients. markers (Ramamurthy et al., 1993), oncogenes (Celani et al., 1993; Mao et al., 1994), RNA Liquid nitrogen, present on the bottom of the (Davidson et al., 1995) and certain specific com- freezer, is at -196°C, while the area higher up in pounds may best be studied in faeces, e.g. the freezer, i.e. at the liquid/vapour interface is typ- faecapentaenes, which are potent mutagens ically around -120°C. If samples are stored at both excreted in faeces. hi one study, a 2-day stool spec- temperatures, analyses must be stratified and dif- imen was collected using a protocol that required fererrt storage procedures taken into account. subjects to place dry ice into a plastic container Most investigators use liquid nitrogen to niairi- held by a collection bonnet positioned on the tain cell viability. However liquid nitrogen can be toilet rim. The container was then placed into a expensive and labour-intensive on a large scale. Iп styrofoam dry ice chest. Subjects were instructed recent years, -140°C freezers with liquid nitrogen on how to avoid contamination with urine. Frozen back-up systems have been developed as an alter- stools underwent lyophilization prior to analysis. native to storage in liquid nitrogen. The scientific The proceduie was successful in that neither occult rationale for using these freezers as an alternative, bleeding, laboratory drift nor sample degradation is that, below -139°C for pure water (or about with storage (re-test after 2 years) influenced the -130°C for culture media), molecules still vibrate, assay. The correlation between first and replicate but do not move from one position to another, measures was greater than 0.90 (Schiffrnan et cL, thus preventing chemical reactions. Therefore, 1989). few, if any, changes that may affect cell viability should occur between -140 and -196°C. Sеmen Freezers may fail, leading to the necessity for 24- Semen specimens are collected to evaluate the h monitoring of the facility through a computerized effects of exposures on endocrine and reproductive alarm system to alert personnel and activate back- factors. 5exиal abstinence for at least 2 days, but not up equipment. For large-scale collections, systems

232 sample collection, processing and storage must be in place that monitor for the possibility of entrance), equipment (easy cleaning benchtops, fire, power loss, leakages, breaches of security etc. easily available handwashing and first aid such as An empty functioning freezer than can accept sam- eye washing fountains) and techniques (forbidding ples after a single failure is crucial. Monitoring mouth pipetting, puncture-resistant containers for should include active 'hands-on' manual and disposal of sharps, mandatory training, use of mechanical checks, since monitoring systems them- laboratory coats, etc.) have been made. Inocu- selves are prone to failure (e.g. a temperature gauge lation of all workers who may be involved with can become 'stuck' in place). Receipt and control human biospecimens with hepatitis B vaccine and procedures, a quality assurance programme includ- 'universal precautions' training should be stan- ing plans for routine preventive maintenance, and dard. a system for documenting the storage history of every specimen must be carefully planned. Procedures It is appropriate to conclude this survey by empha. Shippiпg sizing that epidemiological studies require stan- sample shipping requirements depend on the dardized approaches in order to ensure quality time, distance, climate, season, method of trans- control. All epidemiologists are familiar with edit port, applicable regulations, type of specimen and checks on data entry and similar measures used to marker(s) to be assessed. Usually, polyurethane ensure the integrity of data. A similar approach boxes containing dry ice are used to ship and trans- with regard to dealing with biospecimens is essen- port samples that require low temperatures. The tial. For example, our 'biospecimen collection quantity of dry ice should be carefully calculated, manual' contains numbered steps (each with a based on the estimated time of the trip (dry Ice rationale) for accomplishing the routine tasks evaporates with time, depending on the external required for pre-clinic, clinic and post-clinic рхо- temperature), the number of samples to be trans- cedures involved with biospecimen acquisition. ported in the boxes and an ample safety factor that Field trip preparation, packing and shipping takes into account likely delays. If the boxes have samples, etc. are dealt with in a similar manner. to be shipped by airline, it is suggested that they be Procedures are in place to standardize approaches placed in the hold of the plane, where the temper- at laboratories that perform routine tasks such as attire during the trip is very low. Bach box should DNA extraction and at repositories that receive, be accompanied by a typed chart describing in record, process, store, label and ship samples. Flow detail the contents of the box and the location of charts are often included to highlight decision each tube in the box. points, i.e. 'DNA to lab A', 'serum to lab B'. These For samples that require very low temperatures, procedures are invaluable in guiding day-to-day shipping in a liquid nitrogen container can be operations, assisting with the training of new per- optimal. If international shippings are planned, sonne!, serving as a starting point for the evalua- safety declarations for the foreign country's cus- tion of technical improvements, creating a frame- toms should be prepared. work for periodic reviews aid establishing a baseline for the evaluation of a response to untoward Safety events. No discussion of biospecimens is complete without a mention of biosafety concerns, although a com- References plete discussion goes beyond the scope of this brief Agahain, A., Lee, J.5., Nelson, J.H. & Johns, R.E. (1990) review. Historically, the earliest concern was to Aisenic levels in fingernails as biological indicator of protect the biospecirnen from contamination; exposure to arsenic. Are. Ind. Нуg. Assoc. J., 51, 646-651 however, in recent years, the appreciation of Aggarwal, R,K., Lang, J.W. & Singh, L. (1992) Isolation of potentially deadly blood-borne pathogens such as high-molecular-weight DNA from small samples of HIV and hepatitis has radically re-oriented this blood having nucleated erythrocytes, collected, trans- approach towards worker safety. Changes to labo- ported, and stored at room temperature. Genet. Anal. ratory design (e.g. air flow is 'once through' and Tech. App?., 9, 54-57 not recirculated, laminai flow hoods, restricted Becher, G., Skaare, JU., Polder, A., Sletten, B., Rossland,

233 Application of Biomarkers in Cancer Epidemiology

. ОК. Hansen, H.K. & Ptashekas, J. (1995) PCDDs, PCDFs, & Helzsouer, К.J. (1995) Stability of ascorbic acid, arid PCBs in human milk from different parts of Norway carotenoids, retiriol, and tocopherols irs plasma stored at and Lithuania. J. ToxicoL Environ. Health., 46, 133-148 -70°C for 4 years. Cancer Epidemiol. Biomarkers Prey., 4, Belien, J.A., Copper, M.P., Braakhuis, B.J., Snow, G.B. & 505-507 Baak, J.P. (1995) Standardization of counting micrопи- Davidson, L.A., Jiang, Y.H., Lupton, J.R. & Chapkin, R.S. c1ei; definition of a protocol to measure geno toxic dam- (1995) Noninvasive detection of putative biomarkers for age in human exfoliated cells. Carcinogencsis, 16, colon carcinoma using fecal messenger RNA. Cancercеr 2395-2400 Eidemiol. Biomarkers Prev., 4, 643-647 Bell, D.Y., Haseman, J.А., 5росk, A., McLennan, G. & DuPont, R.L. & Baumgartner, W.A. (1995) Drug testing Hook, G.E.А. (1981) Plasma proteins of the bron- by urine and hair analysis: complementary features and choalveolar surface of the lungs of smokers and non» scientific issues. Forensics Sri, Int. 70, 63-76 smoker s. Am. Rev. Respir. Dis.; 124, 72-79 E1-Nezami, H.S., Nicoletti, G., Neal, G.E., Donohue, D.C. Bencko, V. (1995) Use of human hair as a biomarker in & Ahokas, J.T. (1995) Aflatoxin M1 in human breast milk the assessment of exposure to pollutants in occupational samples from Victoria, Australia and Thailand. Food and environmental settings. Toxicology, 101, 29-39 Chem. ToxicoI., 33, 173-179 Bernard, A.M. (1995) Biokinetics and stability aspects of Fuhr, U. & Rost К.L. (1994) Simple and reliable С YP1А2 biomarkers: recommendations for application in popu- phenotyping by the paraxanthine/caffeine ratio in lation studies. Тoxiсыоgу, 101, 65-71 plasma and in saliva. Pharmacogenetics, 4, 109-116 Blomeke, B., Bennett, W.P., Harris, C.C. & Shields, P.G. Garland, M., Morris, J.S., Rosner, B.A., Stampfer, M.J., (1996) Serum, plasma, and paraffin embedded tisssues as Spate, V.L., Baskett, C.J., Willett, W.C. & Hunter, D.J. sources of DNAB for studying genetic susceptibility to (1993) Toenail trace element levels as biomarkers: repro- cancer. Proc. Am. Assoc, CaoncerRes., 37, 261 ducibility over a 6-year period. Cancer EрiдеmгоI. Biomarkers Prev., 2, 493-497 Bo1e1li, G., Muti, P., Micheli, A., Sciаjno, R., Franceschetti, F., Krogh, V., Pisani, P & Berrino, F. (1995) Garland, M., Morris, J.S., Stampfer, M.J., Colditz, G.A., Validity for epidemiological studies of long-term cryo- Spate, V.L., Baskett, C.K., Rosner, B., Speizer, EE., Willett, conservation of steroid and protein hormones in serum W.C. & Hunter, D.J. (1995) Prospective study of toenail and plasma. Cancer Epidemiol. Biomarkers Prev., 4, selenium levels and cancer among women. J. Nat! Cancer 509-513 Inst., 87, 497-505 Boоsma, F., Alberts, G., Van-Eijik, L., Ian-iii-'t-Veld, A.J. Gill, P., Jeffrey, A.J. & Werrett, D.J. (1985) Forensic appli- & 5chalekamp, M.A. (1993) Optimal collection and stor- cation of DNA "fingerprints". Nature, 318, 577-579 age conditions for catecholamine measurements in Gramza, A.W., Lucas, J. ., Mountain, R.E,, Schuller, D.E. human plasma and urine. Clin. Chem., 39, 2503--2508 М & Lang, J.C. (1995) Efficient method for preparing nor- Bruzel, N. (1994) Fundamentals of Urine and Body Fluids mal and tumor tissue for RNA extraction. BioTecnigues, Analysis, Philadelphia, PA, WB Saunders 18, 227-228 Casa], S., Salas, T., Gill, M.J., Montpetit, M., Rudnik, J., Granade, T.C., Phillips, S.K., Parekh, B., Pau, С .P. & 5y, C.T. & O'Shaugnessy, M.V. (1992) Stability of dried Georgge, J.R. (1995) Oral fluid as a specimen for detec- blood spot specimens for detection of human immun- tion and confirmatton of antibodies to human immunodeficiency virus DNA by polynverase chain reaction. J. odeficierncy virus type 1. AADE Ed, J,, 2, 395-399 Clin. Mi r oL, 30, 3039-3042 соЫ Hambidge, TCM. (1982) Flair analysis: worthless for vita- Celani, P., Berchtold, C.M., Mahbry, M., Carroll, M., mins, limited for minerals. Am. J. Clin. Nutr., 36, 943-949 Sidransky, D., Cesano, R.A., Jr & Lupo, R. (1993) Han, S.W., Flamm, R., Ichem, C. Induction of markers of normal differentiation in У., Kim, I."., Glarridge, J.E., Evans, D.G., Beyer, J., Drnec, J. & Graham, D.Y. human colon carcinoma cells by the v-ras-H oncogene. (1995) Transport of storage of Cell Growth Diff. 4, 341-347 Helicobacterpylori from gastric mucosal biopsies and clinical isolates. Eur J. Clin. Chang, T.M. & Lister, C.W. (1993) Use of finger-prick Microbiol. Infect. Dis., 14, 349-352 humai blood samples as a more convenient way for in- vitro screening of modified hemoglobin blood substitutes Hayney, M.S., Dimanling, P., Lipsky, J.J. & Poland, G.A. (1995) Utility of a "swish and spit" tecnique for the col- for complement activation: a preliminary report. Biornater. lection of buccal cells for TAP haplotype determination. Arti f. Cells Immobilisation BiotechnaI,, 21, 685-690 Mayo Clin. Proc., 70, 951-954 Comstock, G.W., Norkus, E.Р., Hoffman, S.C., Xu, M, -W Нi11 D.W. (1995) Effect of storage on measured blood lac-

234 Sample collection, processing and storage tate concentration: a research note. Int. J. sports Med., 16, screening. Hum. Genet., 75, 213-216 88-90 McMahon, К.J., Fang, C., Layug, L. & Sandler, S.G. (1995) Heller, M.J., Robinson, R.A., Burgart, L.J., TenEyck, С .J. & Pooling blood donor samples to reduce the cost of 11V- Wilke, W.W. (1992) DNA extraction by sonication; a 1 antibody testing. Viz Sang., 68, 215-219 comparison of fresh, frozen and paraffin-embedded tis- 1vicMurthie, E.J., Potter, J.D., Rohan, T.E. & Hetzel, S.S. sues extracted for use in polymerase chain reaction (1984) Human cheek cells: a non-invasive method for assays. Mod. PothoL, 5, 203-206 determininfg tissue lipid profiles in dietary and nutri- Kaaks, R., van der Twel, i., van Noord, P.A. & Riboh, E. tional studies. Nair. Rep. Iпt., 29, 519-526 (1994) Efficient use of biological banks for biochemical Mao, L., ‚ruban, R.H., Boyle, J.О., Tockman, M. & epidemiology: exploratory hypothesis testing by means Sidraпsky, D. (1994) Detection of oncogene mutations of a sequential t-test. Epidemiology, 5, 429-438 in. sputum precedes diagnosis of lung cancer. Cancer Res., Kathol, R.G., Poland, R.E., Stokes, P.E. & Wade, Ѕ. (1995) 54, 1634-1637 Relationship of 24-hour free cortisol to 4-hour salivary Murray, G.I. & Ewen, 5.W. (1991) A novel method for morning and afternoon cortisol and cortisone as mea- optimum biopsy specimen preservation for bistochemi- sured by a time-integrated oral diffusion sink. J. cal and immunohistochemical analysis. Am. J. Clin. Endcrbwl. Invest., 18, 374-377 Patlsol., 95, 131436 Kieidsen, S.E., Eide, I. & Leren, P. (1983) Effects of posture Nafstad, P., 8оtteп, G., 1-lagen, J.А.; Zahlsen, K., Nilsen, on serum cholesterol fractions, cholesterol ratio and O.G., Silsand, T. & Kongerud, J. (1995) Comparison of triglycerides. Scan. J. Clin. Lab. Invest., 43, 119-121 three methods for estimating environmental tobacco Kohlmeier, L. & Koh]rtreier, M. (1995) Adipose tissue as а smoke among children aged between 12 and 36 months. medium for epidemiologic exposure assessment. Environ. Int. J. Epidemiol., 24, 88-9 4 Health Perspect., 103 (suppl. ЗP), 99-106 Nair, G.Q., Shimada, T., Kurazono, H., Okuda, J., Pal, A., Kavasawa, T., Mihara, T., Uesaka, Y., Shirai, H. & Garg, 5. Lambert, G.H., 5choе1lеr, D., Humphrey, Н.E., Kotake, A.N., Lietz, H., Campbell, M., Kalow, W. & spielberg, S. (1994) Characterization of phenotype, serologic, and (1990) The caffeine breath test and caffeine urinary toxicologic traits of Vibsio Choleric 0139 bengal. J. Clin. metabolite ratio in the Michigan cohort exposed to poly- Microbiol., 32, 2775-2779 brominuted hiphenyls: a preliminary study. Environ. Nielsen, E.W., Johansen, H.T., Straume, B. flumes, Т.E. Health Perspect., 89, 175-181 (1994) Effects of time, temperature and additives on a Landry, M.L., Ferguson, D., Cohen, S., Huber, K. & functional assay of Cl inhibitor. J. immunoL Methods, Wetherill, P. (1995) Effect of delayed specimen process- 173, 245-251 ing on cytomegalovirus antigenemia test results. J. Clin. Orlowski E. Pairon, J.C., Ameille, J., Janson, X., Iwasubo, Microbгоl., 33, 257-259 Y., Dufour, J., Bignon, J. & Brochard, P. (1994) Pleural Lantz, M.E., O'Brien, W.F. & William, M.C. (1995) The plaques, asbestos exposure and asbestos bodies in bro- effect of sample preparation and storage on maternal choalveolar fluid. Am. J. Ind. Med., 26, 349-358 triple-marker screening. Obstet. Gynecol., 85, 919-923 Ramamurthy, T., Pat, A., Bag, PK., Bhattacharya, S.K., Lenander-Lumikari, M', Johansson, I., Vilja, P. & Nair, G.B., Kurazono, H., Yamasaki, S., Shirai, H., Takeda, Samaranayake, L.P. (1995) Newer saliva collection meth- T. & Vesaka, Y. (1993) Detection of choiera toxin gene in ods and saliva composition: a study of two sauvette kits. stool specimen by polymerase chain reaction: compari- AADEEd.J., 1,86-91 son with bead enzyme linked immunosorbent assay and culture method for laboratory diagnosis of cholera. J. Doit, M., Martin, C. & Ai th-Kindree, W. Leonard, J., т Clin. Microbiol, 31, 3068-3070 (1994) Correlation of buccal mucosal transudate collected with a buccal swab and urine levels of cocaine. J. Richards, B., Skoletsky, J., S%uber, А.P., Balfour, R., Stern, Addict. Dis., 13, 27-33 R.C., Dorkin, Н.~., Parad, R.B., Witt, D. & Ю inger, K.W. (1993) Multiplex PCR amplification from the CFTR gene Liesnard, C., De Wit, L., Motte, 5., Brancart, F. & Content, using DNA prepared from buccal brushes/swabs. Human J. (1994) Rapid diagnosis of cytomegalovirus lung infec- Mol. Genetics, 2, 159-1 63 tion by DNA amplification in bronchoalveolar lavages. Mol. Cci Probes, 8, 273-283 Rogan, W.J. & Ragan N.В. (1994) Chemical contaminants, pharmacokinetics, and the lactating mother. McCabe, E.R.B., Huang S.-Z, Seltzer, W.K. & Law, M.L. Environ. Health Perspect., 102 (suppl. 11P), 89-95 (1987) DNA microextraction from dried blood spots on filter paper blotters: potential applications to newborn Roggli, V.L., Coin, P.G., MacIntyre, N.A. & Bell D.Y

235 Application of Biomarkers in Cancer Ерidemiolagy

(1994) Asbetsos content of bronchoalveolar fluid. A com- Thompson, D.М., Brown, N.N. & Clague, А.E. (1992) parison of light and scanning electron microscopic Routine use of hair root or buccal swab specimens for analysis. Acta Cytologica, 38, 502-510 PCR analysis: advantages over using blood. Clin. С him. Acta, 207, 169-174 Samеt, J.M., Marbury, M.C. & spengler, J.D. (1988) Ѕtаtе of art: health effects and sources of indoor air pollution. Umemura, U., Koike, K.A., Sato, S., Iso, H., Shimamoto, Am. Rev. Respir. Dis., 137, 221-242 T. & Komachi, Y. (1991) The effect of storage on serum fatty acids. Nippon Eisegaku Zasshi, 46, 976-983 Sanz, A.M. & Diaz, R.C. (1995) Selenium in human lac- tation. Nutr. Rev., 53, 159-166 van den Brandt, P.A., Goldbohm, R.A., van't Veer, P., Schiffman, M.H., Van Tassell, R.L., Robinson, A., Smith, Bode, P., Hermus, R.J.J. & Sturmans, E (1993) Predictors of toenail selenium in men and women, L., Daniel, J., Hoover, R.N., Weil, R., Rosenthal, J., Nair, Cancer Epidemiol. Biomarkers Prey., 2, 107-112 P.1'., Schwartz, S., Pettigrew, H., Curiale, S., Baiist, G., Block, G. & Wilkins, T.D. (1989) Case-control study of Wahrendorf, J., Hanck, A.B. & Munoz, N. (1986) Vitamin colorectal cancer and fecapentaene excretion. Cancer measurements in pooled blood samples. Am. J. Res., 49, 1322-1326 EpidemioL, 123, 544-550 Schmidt, B.L., Aberer, E., Stockenhuber, C., Kiade, H., Wallace, L.A. & Pe1lizzari, E.D. (1995) Recent advances in Breier, F. & Luger, A. (1995) Detection of BorreIia bцrgdor- measuring exhaled breath and estimating exposure and fеri DNA polymeease chain reaction in the urine and body burden for volatile compounds (VOCs). Environ. breast milk of patients with Lyme borreliosis. Diagn. Health Perspect., 103 (suppl. 3), 95-98 Microbiol. Infect. Dis., 21, 121-128 White, I.R., Brurmer, E.J. & Barron, J.L. (1995) A com- Shah, S.N., Kohnson, R.C., Minn, K. & Schoenfeld, F. parison of overnight and 24 hour collection to measure (1995) Ary1su1fatase A and beta-galactosidase activities in urinary catecholamines. J. Clin. Epidemiol., 48, 263-267 leukocytes and lymphocytes from normal and psychi- Winn, D.M., Reichman, .E. & Gunter, E. (1990) atric subjects. Effects of blood processing delay aid inter- М Epidemiologic issues in the design and use of biological l ki -2 stimulation. Mol. Chem. NeuropathoL, 24, 43-52 еи п specimen banks. Epidemiol. Res., 12, 56-70 Shield, J.P., Hunt, L.P., Morgan, J.E. & Pennock, C. А. (1995) World Health Organization (1992) Laboratory Manual for Are frozen urine for estimating samples acceptable albumin the Examination o f Semen and S me -cervical Mucus excretion in research? Diabet. Mcd., 12, 713-716 Нитпп е п Interaction (3rd edn), New York, Cambridge University Srivastava, A.K., Gupta, B.N. (1994) The role of human Press, pp. 5-6 hairs in health and disease with special reference to envi- Wrensch, M., Petrakis, N.L., King, K.B. & Lee, M.M. ronmental exposures Vet. [lam. ToxicoL, 36, 556-560 (1993) Breast cancer risk associated with abnormal cytol- Statland, B.E., Winkel P. & Harris, S.C. (1978) Evaluation ogy in nipple aspirates of breast fluid and prior history of of biological sources of variation of leukocytes count and breast biopsy. Am. J. EpidemioL, 137, 829-833 other hematologic quantities using very precise automated analyzers, Am. J. Clin. PathoL, 69, 48-54 Storch, G.A., Gaudreault-Keener, M. & Welby, P.C. (1994) Comparison of heparin and EDTA transport tubes for Corresponding author detection of cytomegalovirus in leukocytes by shell vial N T. Landi assay, 65 aritigenemia assay, and PCR. J. Clin. рр Genetic Ерidemiology Branch, EPN 400 MIcrobial., 32, 2581-2583 National Cancer Institute, National Institutes of Health Sundberg, A.G., Nilsson, R., Appelkis 6130 Executive Blvd., Bethesda, 10 20892-7360, st, E.L. & Dallner, G. USA (1995) ELISА procedure for the quantitation of glutathione transferases in the urine. Kidney tnt., 48, 570-575

236 АррГicalion of Biomarkers in fiancer Epidemiology Tonia1o, P., Botfetia, P., Shukеr, D.E.G., Rothman, N. hike, В. aid Pearce N., eds 'ARC 5агепУifiс PuЫicatгoпs No 142 InlernaiionaI Agency for Research on Cancer, Lyon 1907

Issues involving biomarkers in the study of the genetics of human cancer

N. Caporaso and A. Goldstein

The investigation of hereditary factors in human cancer was suggested from kindreds that exhibited aggregations of cancer consistent with Mendelian inheritance. A subset of cancer that exhibits strong familial tendencies is due to single genes that `cause' cancer; more commonly, hereditary factors may influence tumorigenesis in a stepwise probabilistic rather than deterministic manner through a variety of mechanisms, e.g. influencing the disposition of carcinogens. The roles of both common susceptibility genes and rare `familial' cancer genes are receiving increasing attention in the general population. Population-based studies designed to examine more common genetic variants differ from linkage-based studies. Candidate susceptibility genes may be studied by phenotype or genotype approaches, and the relative advantages and disadvantages of each approach are considered.The issue of gene—environment interaction, implicit in the concept of susceptibility genes, is consideied.The influence of genetic factors on individual and attributable risk is addressed.

Cancer and genetics and bladder cancer, and diethylstalbesterone and The general evidence in support of an important clear cell carcinoma of the vagina in daughters of role for genes in cancer derives from various inter- exposed women. Our view is that the development related areas: (1) the discovery of specific chromo- of cancer is a complex multistage process, with each somal defects associated with lymphoproliferative step involving a variable mix of environmental (and later other) cancers; (2) the universal obser- and genetic influences. Our knowledge of how vation of somatic mutations in tumours; (3) the these factors act in concert with genes to cause role of dominant oncogenes and tumour sup- malignancy is currently a patchwork. Three gen- presser genes in certain tumours; (4) the mapping eral areas where further work is needed arе: (1) eti- and cloning of genes specifically accounting for a ological factors responsible for certain cancers number of familial cancers; (5) the general obser- remain unknown (e.g. prostate, brain); (2) even for vation of increased tumour incidence in relatives cancers where the environmental agents are well of individuals with cancer compared to suitable established determinants of individual suscepti- controls; (6) the relationship between somatic and bility are not clear; (3) the mechanism of interac- hereditary mutations in specific tumours such as tion between exposures and genes is poorly under- retinoblastoma; (7) the increased incidence of stood. tumours associated with conditions that damage the genetic material, e.g. the 'chromosome insta- Approaches to the study of genetics and cancer и bility syndromes'; (8) the fact that many estab- human populations lished carcinogens are known to damage or disrupt We first briefly consider approaches that have been DNA. These lines of evidence can be considered to used to study highly penetrant `single' genes asso- have clearly established a central role for genetic ciated with familial cancers. We will then describe processes in neoplasia. population-based studies of low penetrant, but Likewise, environmental factors are considered more common, 'susceptibility genes, thought to causal for a variety of cancers. A few examples are contribute to complex disorders. Studies of somatic tobacco and lung cancer aromatic amine exposure gene findings will be briefly described.

237 Application of Biomarkers in Cancer Epidemiology

Table t Genes associated with familial cancer that have been mapped or cloned ..

RelinoЫastoma (retinoblastoma, AB1 13g14 Cavene et a1., 1983; Sparkes etal., osteosarcoma) 1983

Wilms' tumour (WAGR1) WГ2 (WT1) 11р13-14 Garwin etat, 1995 Neurofibromatosis type i NF7 17g11 Upadhya etal., 1995 (neurofibrosarcomas, others)

Neurofibromatosis type 2 (bilateral NF2 22g11-13 Seizinger et al., 1987; Wertalecki vestibular schwanпomas) et aL, 1988 Li-Fraumeni (breast, sarcoma, p53 Malkin et ai, 1990 leukaemia, brain)

Familial breast—ovary (breast, ovary) BRCA 1 Hall et aL, 1990; Albertsen et at, 1994; Б RСА2 Miki et а!.,1994; Wooster etaL,1994 Von Hi l—Lindau ррв (renal cell VIL 5eizingеr, 1988 carcinoma, haemangioblasloma)

CMI (melanoma) р16/CDKN2А 9р21 Kamb et a1., 1994; Hussussion st at, С DК4 12g13 1994; Zuo eta]. 1996

NBCC (basal cell, fibrosarcoma, РТСN 9g22 Hahn, 1996; Johnson et a1., 1996 medijlloblastoma)

Ataxia telangiectasia (lymphoma, ATM 11q22-q23 SavitSky et at, 1995 leukaemia, breast, others)

Familial adenomatous polyposisl APC 5g21 Bodmer et al., 1987; Kinzler et al., 1991; Gardner syndrome (colon) Nagase et at, 1992; Eckert et at., 1994

Xeroderma pigmentosa (squamous XР (A—G) many 9g34.1, others Tanaka et at., 1990. cell carcinomas, skin, adrenal) -

HNPCC (colon) h1SN2 2р22 Nicolaides et al., hiLN1 3р21 1994; Nystrom-Lahti of at, 1994; hРМS 1 2g31-33 Papadopolous etal., 1994 !УРMS2 7p22

Multiple endocrine neoplasia type 1 MEN1 11q 13 Larson, 1988; Thakker et at, 1993. (carcinoids, pancreas, parathyroid, pituitary)

Multiple endocrine neoplasia, type 2A RET 10g11.2 Muligan et at, 1993 (medullary thyroid, phaeochromocytoma), type 28 (same as 2А) aWAGR, 'Mirs' tumour, aniridia, genital, renal аЬюгтa]1iе5.

238 Biomarkers in the study of the genetics of human cancer

Familial cancer and `single gene traits Stцdies involving somatic genes Cancer families have been noted for centuries and Chromosome studies family aggregation studies have generally focused It was recognized in early studies that chromosome on three related questions. Does a significant aberrations accumulate with the age of a tumour excess of cancer exist? Is the cause consistent (Boveri, 1917). With the development of better with genetic or environmental causes, or both? If techniques to image chromosomes (i.e. staining genetic, what is the mechanism of inheritance? of metaphase spreads), it was appreciated that Recognizing rare syndromes that exhibit strong specific chromosome aberrations characterized familial aggregations implicated heritable elements certain severe multisystem birth defects such as in these entities. An initial step is often to test the Down syndrome (trisomy 21). In 1960, Nowell & transmission pattern of disease in families for a fit Hungerford (1960) discovered a consistent change with Mendelian patterns of inheritance (segrega- in the morphology of a chromosome in blood cells tion analysis). Once a genetic pattern is considered of patients with chronic myelocytic leukaemia likely, linkage studies or other approaches can (Philadelphia chromosome). We understand today be used to map the chromosomal location, to be that this translocation involves an exchange of followed by molecular genetic studies to identify pieces between chromosomes 9 and 22. The break the specific genes involved. For certain familial moves the obi oncogene from chromosome 9 to cancers (e.g. retinoblastoma), as well as various the vicinity of the bcr (breakpoint cluster region) inherited syndromes that often include cancer (e.g. gene on chromosome 22. When these two genes Beckwith-Wiedemann and Wilms' tumour), either are brought into juxtaposition, increased tyrosine the cliomosomal location has been identified or kinase activity results, and enhanced cell division the gene has been cloned, or both (Rowley, 1980; results in proliferation of cells with this clone. A Evans, 1993; Knudson, 1993) (Table 1). Gene map- variety of other translocations involved with ping depends on the availability of suitable multi- specific (typically haematopoietic) tumours in- generation families or sibships, accurate diagnosis, volve similar mechanisms. biospecimens from which DNA may be isolated, Chromosome findings have general importance and adequate informative markers (generally for cancer. The consistent and specific cytogenetic increasingly available). `Informative' in this con- changes observed with particular tumours aie an text means that the parent (at least one, depending important indication that the genetic material is upon the inheritance pattern) must be heterozy- fundamentally involved in the neoplastic process. gous at both marker and disease loci. DNA markers Chromosome aberrations in familial cancers can are characterized in the laboratory and suitable provide clues to the location of regions where analytical approaches, software and computational critical genes may be located. Although the early resources are applied to evaluate the 'recombina- techniques for studying chromosomes were diffi- tion fraction' (evidence of 'crossing over' between cult and time-consuming, early observations of a marker and disease locus). The resulting statisti- these consistent defects have led to a more precise cal evidence is expressed as a 'lid' score or log understanding of molecular pathology at the DNA of the odds that a given marker is 'linked' to the level. New approaches, such as fluorescent in-situ cancer, with scores of greater than 3 (1000:1) hybridization (FISH) technology, are revolutioniz- indicative of linkage (Ott, 1991). Once a suspect ing the field and rendering the recognition of syn- location has been identified by this approach, dromes characterized by chromosome findings positional cloning (also termed `reverse genetics') much simpler. can be used to precisely locate and clone the gene. Other approaches are possible. For instance, begin- Other somatic gene findings ning with a particular protein, the amino acid se- Somatic mutations are universally observed in quence is used to infer the base pair sequences. human cancer and are distinguished from hereditary Using in-situ hybridization, the chromosomal 'single or 'susceptibility' gene changes discussed location may be deduced. Table 1 indicates a previously in that they are observed in tumour tis- number of the major cancer genes identified by sue (as opposed to germline DNA) and they are not these approaches. transmitted to offspring.

239 Application of Biomarkers in Cancer Epidemiology

Loss of heterozygosity studies bave been used Second, there is the relationship of somatic to identify regions of gene loss to implicate mutations to inherited defects. Knudson's (1983) possible tumour suppresser gene loci for many findings in retinoblastoma are the paradigm for tumours, including chromosome 13 in retino- this type of relationship, with his observation that blastoma (13g14) and chromosome 3 (Зp) in retinoblastoma was caused by two mutational lung cancer (Yokota et al., 1987). These studies events, one on each allele of the same gene. In the complement chromosome studies and, together inherited form, one mutation is inherited via the with these, have helped to localize the chromo- germinal cells and a second occurs in a somatic cell some abnormalities that characterize various tu- due to an environmental insult or chance event. mours. For example, cytogenetic studies led to the In the sporadic form, both mutations occur in identification of Зp deletions in small cell lung somatic cells. The further observation that these cancer (Peng et aI., 1982), an area thought to in- events were accompanied by chromosome changes clude more than one tumour suppresser gene, consistent with a loss of genetic material suggested including the recently identified FlIT gene (Sozzi a gene whose absence was required for cancer, a et aL, 1996). phenomenon termed `anti-oncogene', but more More recent studies have emphasized the rela- commonly known today as tumour suppresser tionship of the type, number and specificity of gene today. Many of the hereditary cancer syn- these and other molecular findings to previous dromes listed in Table 1 have been shown to exposure history, hereditary factors, tumour type involve genes from this class, e.g. p53 and WT1. and aggressiveness. Tumour tissue mutations are Recently, certain hereditary genes have been characterized by special stains (i.e. immunostain- associated with somatic gene findings. Two recent ing for p53) or direct sequencing of DNA extracted examples include increased p53 mutations in from frozen tissue, or, as is increasingly the case, GSTMI null subjects with lung cancer (McGlynn from tumour blocks. Typically, slides cut from et al., 1995; Ryberg et al., 1994), and, in a series of adjacent material have undergone pathological patients with colon cancer, rapid acetylators verification. Often, microdissection is used to iso- (NAT2) were more likely to exhibit K-ras gene late tumour tissue for special study. mutations than intermediate or slow acetylators While molecular studies of human tumours (Ida et al., 1994). have proceeded since the advent of the techniques Finally, somatic mutations are thought to indi- the variable nature of many findings has led to cate something about the degree of aggressiveness efforts to examine hypotheses more rigorously of the underlying cancer and may be related to using the methods of epidemiology. Three exam- stage, grade or clinical behaviour of the tumour. As ples of these approaches are considered. observed by Boveri and many others, increased First, it is hypothesized that somatic mutations mutations are observed in advanced or poorly dif- may reflect specific exposures that caused the ferentiated tumours. The presence or absence of cancer. An example is the relationship between chromosomal abnormalities predicts survival in aflatox9n B1 and G to T tranversions at codon 249 chronic lymphocytic leukaemia (Criel et a1., 1997). of p53 in hepatoce11ular cancer from certain areas Aneuploidy and other molecular or chromosome of China or Africa (Soini et al., 1996). Less specifi- findings are related to bladder cancer aggressive- cally, and perhaps consistent with the multiple ness (Walmau et al., 1991; Esrig et al., 1994). An carcinogens present in tobacco smoke, G to elaborate example of somatic mutations and transversions predominate at the p53 locus in tumour progression was provided by Vogelstein et smoking-related tumours. UV light results in spe- ad. (1988) with the demonstration of a series of cific p53 changes (transitions) in skin tumours molecular events characterizing progression from (Nakazawa et al., 1994). Other genes may exhibit polyp to cancer to metastasis. these effects; for example, recently, methylation of It should be appreciated that there are many the estrogen receptor in lung tumours has been types of somatic `genetic' markers that involve the observed to be frequent in plutonium-related genetic material indirectly. A classic example tumours but low in tobacco-derived carcinogen- is sister chromatid exchange, while telomere related tumours (Issa et al., 1996). shortening (Sharma et al., 1996) is a more recent

240 Biomarkers in the study of the genetics of human cancer example. In each of these cases, study designs Early studies of low penetrarace gees and attempt to associate the finding with markers of cancer in the general population exposure, susceptibility or effect (disease type, pro- While the study of the highly penetrant 'single' gression) using biomarkers. Study design issues genes would proceed through the approaches and methodological consideration in this 'molec- described earlier in cancer families, investigators ular epidemiology' approach have been described looking for the more subtle effects of weaker 'susceptibility' genes would use different approaches, (Ретеrа, 1982; McMichael, 1994). based on investigations set in the general popula- Inborn errors of metabolism and the idea that tion. There were four studies that established an genetically controlled influences on metabolism interest in susceptibility to cancer and this class of can determine disease genetic traits between 1973 and 1984. The first was In 1908, Archibald Garrod described four rare the study of the phenotype of lymphocyte induci- recessively inherited conditions: albinism, alkap- bility of (CYPIA1-dependent) aryl hydrocarbon tonuria, cystinuria, and pentosuria. He postulated hydroxylase activity (Kellermann et al. (1973). that these disorders were due to genetically based Three other studies in the 1980s were crucial. defects in normal biochemistry. He termed these Ayesh (1984) studied debrisoquine metabolism 'inborn errors in metabolism' (Scriver et al., 1995). (CYР2D6) in relation to lung cancer. Lower et al. (1979) examined the N-acetylation phenotype This idea was а fertile concept for human disease etiology, initially for providing a genetic explaria- (NAT2) in aromatic amine-related bladder cancer tion for these conditions (perhaps the first descrip- using a sulfa probe drug, and seidegard et al. tion of an autosornal recessive disorder), but also (1986) examined glutathionе-S-transferase mu for suggesting that biochemical diversity is the activity (by measuring trans-stilbene oxidation in substrate upon which natural selection may act. erythrocytes) (GSTM2) in relation to lung cancer. Although the expression of genetic traits exhibits These studies had aspects that distinguished them variability due to both environmental factors and from both prior and more modern studies. As the genotype, the inborn errors are highly pene- described above, early cancer genetic studies were trait in that virtually all those who inherit the trait based on rare familial aggregations of cancer that exhibit features of the disease. This follows from were recognized climcaly. The clinical phenotype the fact that the biochemical disorder involves was used to identify affecteds in order to initiate endogenous metabolic processes, and therefore the gene mapping studies, sometimes aided by a cyto- consequence of the blocked pathway is unavoid- genetic abnormality that would offer a clue to a able given the ubiquitous presence of the endoge- gene location. In contrast, this new population- based approach selected a gene of interest (actually nous substrate (Scrivnеr et аL, 1995). A related class of phenomenon is distinguished from the inborn a phenotype, since there was no convenient errors in that an inherited defect involves the aber- method to identify the genotype at the time of rant metabolism of an extrinsic agent. The these studies), generally based on a mechanistic latter group are termed pharmacogenetic disorders hypothesis, i.e. bladder cancer should be more (Kalow, 1962). This class of disorders 1s distin- likely in individuals who are deficient in their guished in that no disturbance is present in an genetically determined ability to inactivate aryl individual with the trait until they are exposed to amines (i.e. bladder carcinogens). A population- the particular medication dependent on the vari- based approach was required because the traits of ant enzyme. Examples of this class of disorders are interest derived from low penetrance genes. Thus, glucose-6-phosphate dehydrogenase (G6РD) defi- large numbers of study subjects were required to ciency, the porphyrias and malignant hyperther- achieve the statistical power to detect the postu- rrlia. Since carcinogens both require metabolic acti- lated differences in risk between the gene types. vation and are also subject to metabolic processes The phenotype was recognized, not by clinical find- that facilitate elimination the existence of variant ings (as in the inborn errors of metabolism), but versions of the enzymes that control these rather through the use of a probe drug in two of processes might be expected to alter susceptibility the four studies (to determine a `metabolic pheno- to the particular cancers. type'), or in-vitro assay in the other two studies. In

241 Application of Biomarkers in Cancer Epidemiology

Consideration Phenotype Genotype

Advantages Historically validated approach Identifies heterolhomozygous subjects Reflects physiological, in-vivo disposition Technology evolving rapidly of drug Simple, requires only germline DNA sample Reflects impact of inducers, inhibitors Invariant to illness, diet, medications, etc. Can be performed with micro-quantities Non-invasive samples (i.e. mouth wash, paraffin) Disadvantages Potentially distorted by numerous factors, Functional status often requires study i.e. drug—drug interaction Some variants unknown Analysis typically more complex Ethical questions arise since DNA may be used More patient cooperation required for other tests Phenotyping protocols poorly adapted to Allelic heterogeneity field study Invasive, time-consuming nature of test . causes many exclusiohs .

contrast to the studies that characterized the to this unusually high degree of metabolic variation inborn errors of metabolism (based on the recog- are not limited to medications (Khoury et al., nition of a clinical phenotype), slow acetylators 1988). Chronic conditions are thought to occur with (e.g. of naphthylaniine, bladder carcinogen) or poor altered frequency based on exposures over time to metabolizers (of debrisoquine) exhibit no obvious specific agents subject to this type of variability. clinical findings, and it is only upon challenge Over the last 10 years, many other genes have with the appropriate agent that differences in drug been studied in relation to various tumours. These metabolism can be detected or clinical sequelae have been reviewed recently (D'Erricco etaL, 1996). recognized. The probe drugs used to characterize The genes of interest, mechanistic basis and the phenotype were of course innocuous non- phenotype/genotype approaches are listed in cardnogens. While the phenotyping method of Table 3. study is still used, modern studies increasingly depend upon direct identification of a genotype, Broadening the proposed mechanism of low arid the advantages and disadvantages of each penetrance genes beyond metabolism approach are summarized in Table 2. Early studies of cancer susceptibility focused on the In pharmacogenetic conditions, the phenotype idea that variation in enzymes made individuals does not always result in clinical sequelae, i.e. the activate or deactivate carcinogens differently, condition is not fully penetrant. In pharmacogenetic accounting for differences in susceptibility. For conditions, acute sequelae result from specific ex- instance, cancer etiology, development, progres- posures, typically to pharmaceutical agents (but also sion or prognosis may be influenced by hormones xenobiotics, carcinogens or endogenous compounds) (breast and estrogen; prostate and androgen), that have some aspect of their metabolismzt depen- infectious disease (Helicobactor pylori and gastric; dent upon the enzyme (Or receptor, immune factor hepatitis and liver cancer), and nutrient intake or or other element) that is subject to pharmacoge- immune factors. Examples of these mechanisms, netic variability. The term 'еcоgепеhсs' has been and the genes that may influence them are listed used to emphasize that agents that may be subject in Table 4.

242 Biomarkers ln the study of the genetics of human cancer

Methodological issues in studies of susceptibility phenotype approaches continue to be supplanted genes in the population by genotype investigations, many new phenotypes Phenotype/genotype issues. With regard to low with cancer associations (e.g. the bleomycin sensi- penetrance gene/cancer association studies, early tivity assay and lung cancer, DNA repair assay and phenotype-based work typically required the skin cancer) continue to be proposed (Cuis, administration of a probe drug followed by the col- 1996). It is therefore worthwhile to consider the lection of a timed urine or blood sample. Given advantages and disadvantages of each approach that it may be efficient to determine the status of (Table 2) and the types of studies needed to vali- a number of polymorphic traits at once, a'cocktail' date these approaches. approach (i.e. administration of multiple probe drugs) has been used as data have shown that cer- Phenotyping studies: methodological issues. The dif- tain combinations of probes aie safe and do not ficu[ties involved in establishing a phenotyping result in interference (Branch et al., 1995). While approach as appropriate for population-based

в a в в г в I1l в

Gene Mechanism Phenotype Genotype Hypothesized cancer

MspI, exon 7 Lung (Kellerman et ai, СУР IA1 Activates benzo[a]pyrene, AHI activity in lymphobIasts . role in estrogen polymorphisms 1973), Breast (Taint oral., metabolism (?) 1995) Under investigation Bladder, colon (Lang СУР1A2 Activates heteroсусцс Caffeine breath test, amuies, aryl amines caffeine urine, saliva et a1., 1994) metabolites

Debrisoquine,. Inactivating and Lung (Ayesh et al., СУР 2p6 Nicotine? dextromethorphan partially iеactivadng 1984) metabolism mutations exist

Polymorphisms Nasopharyngeal СУР 2E1 Activates low molecular weight nitrosamines, studied but function carcinoma (Hildesheim alcohol-inducible poorly characterized et al., 1995), lung (Uematsu etat, 1991)

GSTM1 Detoxification of epoxides Trans-stilbene oxide in RBCs Absent gene Lung (seidegard et at, . 1986), bladder (Bell et at, 1993)

G5TT1 Detoxitication of ethylene No Absent gene Myelodysplastic syndrome oxide, butadiene (Chen et a1., 1996)

NAT2 . Detoxification of aromatic Caffeine or sulfa Eviulations recognized Bladder (occupational . amines metabolites exposure-related) (Cartwright et al., 1982), breast (Ambrosone et al. 1996)

NAT1 Detoxification of aromatic ? Variants recognized Gastric, bladder colorectal amines (Bell et a1., 1995)

243 Application of Biomarkers in Cancer Epidemiology

Gene category Gene example Cancer site

Dominant oncogene ret, ras, rye Lung, others

Tumour suppresser genes p53, rb Lung, bladder, others (Wu et al., 1995)

Hormones Steroid metabolism Prostate, breast, endometrium (Carey etal., 1994)

Hormone receptor Estrogen, progesterone and Breast, endorneirium, (Fuqua et al., 1991), апдrодеп receptor prostate (Irvine etal., 1995)

Vitamins Vitamin D, bic acid, В12 Prostate (Taylor et al., 1996), other medical illnesses (Jacques et al., 1996)

Alcohol metabolism ADN, ALDH Oral cancer (also cirrhosis, alcoholism)

Nicotine metabolism С YP2D6, СУ P2В6 Smoking-related cancers (also COPD, others) Addiction Dopamine receptors Smoking- and alcohol-related cancers (Comings etal., 1994)

DNA repair XP, X-linked lymphoproliterative Skin cancers, Burkitt lymphoma, others syndrome, AT Hanawalt etal., 1996

Immune function 'LA Various rheumatological conditions

studies are substantial. As an example, we briefly ciation studies involving phenotypes is review some work with the debrisoquine pheno- 'effect–cause' bias—i.e. could disease itself (or type, а well-established example after two decades related treatment, nutritional status, etc.) alter the of work. It was discovered in the 1970s that after phenotype and bias study. Early studies showed administration of debrisoquine, certain individuals that the debrisoquine metabolic ratio was unre- (poor metabohzers) suffered prolonged hypoten- lated to stage, grade or performance status in 92 sive episodes. Population studies established that patients with lung cancer, and a later, more defin- 10% of the general population was deficient in the itive study showed that the metabolic ratio was enzyme required to hydroxylate the drug. A fre- invariant in individuals undergoing curative resec- quency histogram of phenotype values in a tion for lung cancer (5hаw et al., 1994). Western population exhibits a clear bimodal distribution. Family studies demonstrated that this Assigning the phenotype. The result of debrisoquine deficiency was inherited (Evans et al., 1980). The phenotypirng can be expressed as a ratio (the meta- safety and feasibility of administering a tracer dose bolic ratio, mr, is the molar ratio of parent drug to of debrisoquine had to be established in a clinical chief metabolite in the timed urine sample). This setting (Green-Gallo et al., 1991). Circadian varia- continuous variable must be assigned an ordinal tion, the precise time necessary to establish a sta- category to specify a phenotype. One approach ble phenotype (number of hours of urine collec- involves the use of a mixture model to analyse the tion), recent diet, the influence of concurrent med- overall distribution of mrs into phenotypic com- ications, and the comparability of day and night ponents. The resulting proportions may be com- protocols to perform phenotyping were examined pared with those expected under Hardy Weinberg (Shaw et al., 1990; Caporaso et cL, 1994). An issue equilibrium. Parameter estimates for one-mix and that often clouds the interpretation of cancer asso- two-mix models can be compared with a likeli-

244 Biomarkers in the study of the genetics of human cancer hood ratio x2 test. This approach can be applied to • Is the trait stable in the individual? (Intra- assign cut-points for any pharmacogenetic study individual variability should be small.) that involves the use of а probe drug (Caporaso et • Does the trait vary in the general population? aL, 1989). As it has become more common for (Interindividual variation should be large.) both phenotype arid genotype information to be • Can diet, medications, circadian variation, dis- available, receiver operator curve approaches are ease or nutritional status distort phenotyping? an alternative to assign cut-points (DeLeo, 1993). • Does the trait vary by ethnicity, age, gender, Early studies of pharmaсogеnеtic variation have smoking, alcohol consumption, body weight, relied exclusively on the phenotype, while ad- etc.? (Distorting factors should be understood vances in molecular biology have increasingly sup- so they can be controlled by design or adjusted planted this approach with direct determinations for in the analysis.) of the genotype. Attractive features of genotyping • Is the phenotype subject to induction or include the avoidance of effect-cause bias and the inhibition? If so, to what extent is the pheno- attendant difficulties of phenotyping, and the pos- type `fixed' (i.e. subject to hereditary control), sibility of identifying heterozygotes. While the and to what extent do exogenous factors con- relationship between the two study approaches is trol it? dynamic and evolving, some of the general • Is the trait genetically controlled? (Have family strengths and weaknesses of each approach are studies demonstrated a mode of inheritance?) indicated in Table 2. • Does the gene have a plausible mechanism for involvement with cancer? Some general methodological issues involved in the • Does the proposed substrate for the gene have selection of candidate low penetrance genes for any relation to cancer? (Is a procarcinogen acti- population studies vated, or cardnogen detoxified? Is there some A variety of factors can enter into the considera- other plausible mechanism, i.e. involving a tion of whether a proposed susceptibility factor is nutrient, oncogene, receptor, hormone, etc.?) a worthy candidate for study. A list of questions • Are gene variants known, and do they have should include the following: functional significance? Will all the important

^ в в 1

High penetrance Low penetrance

Role Tend to be 'causal', i.e. necessary Allele alters susceptibility, but is neither arid sufficient to result in disease necessary nor sufficient for disease causation

Example Brcai (breast/ovary) CYP1A1 (lung) (lung) АРС (polyposis coli) С YР2D6 RB (retinoblastoma) GSТ-M1 (lung, bladder) Gene type Mutation Polymorphism Study setting Family General population or epidemiological studies

Strength of association High (RR often з 200) Low to moderate (RR, 2-10) Relative or absolute risk High Low Population attributable risk Low High

Gene—environment interaction secondary arid variable Primary and implicit

Role of environmental exposure Secondary and variable Critical

245 Application of Biomarkers in Cancer Epidemiology

(Fig. 1). The calculations are based upon a method described by Khoury & James (1993) and are applied to a candidate susceptibility factor (GSTM1) studied in relation to lung cancer. PPV and PAR are calculated over a range of relative risks observed tri published studies and assuming the following: the prevalence of relevant exposure is 35/0 (cigarette smoking). The risk of disease (lung cancer) is assumed to be non-zero in the absence of the gene. 1 1.6 2.3 3 3.6 The relative risk of the exposure (tobacco smoking) Relative risk is arbitrarily estimated at 10. A 'type 2 interaction', i.e. the gene does not result in increased risk in the PPV absence of exposure (which is plausible, consider- PAR ing that lung cancer is rare in non-smokers), is assumed (Khoury & James, 1993). The following conclusions are suggested from Figure 1. Positive predictive value (PPV) population аttributаые risk (PAR)for a common' susceptibility gene. the graph. PAR is relatively high, while PPV is only modestly elevated, especially within the likely range of odds ratios, between 1 and 2. This sug- variants (polymorphisms) described in the gests that this gene will have no role in screening population be studied? or clinical testing and that clinical relevance to the • 1s the gene known to act in a relevant organ? individual is limited. The increase in the PAR, how- This may be the liver or the specific organ of the ever, is significant, even with a modestly increased primary tumour. odds ratio. Because both the gene and the disease • Have any studies in humans been performed? are common, even with the a 1.5 odds ratio, the Have epidemiological studies in humans been- burden of disease in the population, due to the carried out? If so, do the accumulated data sup- gene (in the presence of exposure) is substantial. port further study? References •Summary of contrasts between high and low Albertsen, H., Plaetke, R., Ballard, L., Fujimoto, L, penetrance gene studies Соппоlly, J., Lawrence, E., Rodriguez, P., Robertson, M., There are certain contrasts between genes identi- Bradley, P., Milner, B., l uhrman, D., Marks, A., Sargent, fied and studied in the context of families, i.e., the R., Cartwright, P., Matsunami, N. & Wыtе, R. (1994а) high penetrance 'single' genes listed in Table 1, Genetic mapping of the BRCAI region on chromosome 17g21. Am. J. Hum. lea et., 54, 516-525 and the low penetrance population-based studies suitable for the study of the susceptibility genes as Аmbrosone, C.S., Freudenheim, J.L., Graham, S., described in Tables 2 and 5. These contrasts are Marshall, J.R., Vena, J.E., Brasure, J.R., Michalek, A.M., summarized in Table 2. The 'rare' genes are associ- Laughlin, R., Nemoto, T, Gillenwater, K.A., Harrington, A.M., Shields, P.G. (1996) Cigarette smoking, N-acetyl- ated with high relative and absolute risks but low transferase 2 genetic polymorphisms, and breast cancer attributable risk, while the common genes exhibit risk. J. Am. Med. Assoc., 276, 1494-1501 the apposite qualities, i.e. modest relative and absolute risks but large attributable risk. Ayesh, R., Idle, J.R., Юс hie, J.С ., Crothers, M.J. & Hetzel, M.R. (1984) Metabolic oxidation phenotypes as markers There are contrasts between the implications of for susceptibility to lung r. Nature, 312, 169-170 'single' and 'susceptibility' genes for the individ- сапсе ual and for public health. These aie illustrated by a Bell, D.A., Stephans, E.A., Castranio, T., Umbach, D.M., consideration of the positive predictive value (PPV, Watson, M., Deakin, M., Elder, J., Hendrickse, C., Duncan, H. & Strong, R.C. (1995) Polyadenylation poly- individual risk given a positive test) and the popu- morphism in the acetyltransferase 1 gene (NAIl) lation attributable risk (or PAR, public health bur- increases risk of colorectal cancer. Cancer Res., 55, den due to the disease, based on the combination 353 7-35 42 of the at-risk genotype plus the relevant exposure)

246 Biomarkers in the study of the genetics of human cancer

Bell, D.A., Taylor, J.A., Paulson, D.F., Robertson, C.N., Bunan, D., Clifton, S.W., Crabtree J., Freeman, A., Iyer, K. Mohler, J.L. & Lucier, G.W. (1993) Genetic risk and car- & Jian, L. (1995) Sequence and analyses of the human cinogenesis exposure: common inherited defect of the ABL gene, the BCR gene, and region involved irn the carcinogen-metabolism gene glutathione S-transferase Philadelphia chromosome translocation. Genomics, 27, 67-82 11 (GSТМ1) that increases susceptiЫlity to bladder cancer. J. Nati Cancer Inst., 85, 1159-1164 Cloos, J. (1996) Mutagen sensitivity: a phenotype reflecting Bodmer, W.F., Bailey, C.J., Bodmer, J. (1987) Localization susceptibility to head and neck squamous cell camel noma. of the gene for familial adenomatous polyposis on chro- Amsterdam mosome 5. Nature, 328, 614-616 Comings, D.E., Muhleman, D., Ahn, C., Gysin, R. & Boveri, T.H., Zur Frage der Entstehung malignes Flanagan, S.D. (1994) The dopamine D2 receptor gene: a Tunioren. Jemma, 1914, pp. 25-41 genetic risk factor for substance abuse. Drug and alcohol dependence, 34, 175-180 Branch, R.A., Chern, H.D., Adedoyiri, A., Romkes-Sparks, M., Lesnick, T.G., Persad, R., Wilkinson, G.R., Fleming, Criel, A., Verhoef, G., Vlietinck, IL, Mecucci, C., Billet, J., C.M., Dickinson, A.J., Sibley, G. (1995) The procarcino- Michaux, L., Ieeus, P., Louwagie, A., Van Orshoven, A., gen hypotheses for bladder cancer: activities of indivjd- Van Hoof, A., Boogiertes, M., van den Berghe, H., & ual drug metabolizing enzymes as risk factors. DeWoIf-Peeters, C. (1977) Further characterization of Pharmacogenetics, 5, spec. No. 597-102 morphologically defined typical and atypical CLL; a clin- ica1, immunophenoxy cytogenetic and prognostic study Caporaso, N., Pickle, L.W., Bale, S., Ayesh, R., Hetzel, M. of 390 cases. Br. J. HaematoL, 97, 383-391 & Idle, J. (1969) The distribution of debrisoquine metabolic phenotypes and implications for the suggested D'Erricco, A., Taioli, E., Xiang, C. & Vineis, P. (1996) association with lung cancer risk. Genet. Epidemiol., 6, Genetic metabolic polymorphisms and the risk of can- 517-524 cer: a review of the literature. Biomarkers, 1, 149-1731 Caporaso, N., Whitehouse, J., Monkman, 5., Boustead, Eckert, W.A., Jung, C. & Wolff, G. (1994) Presymptomatic C., Issaq, H., Fox, S., Morse, M.A., Idle, J.R. & Chung, F.- diagnosis in families with adenomatosis polyposis using L. (1994) In vitro but not in vivo inhibition of CУP2D6 highly polymorphic dioucleotide CA repeat markers by phenethyl isothiocyaaate (PEITC), a constituent of flanking the Аr'С gene. J. Med. Genet., 31, 442-447 watercress. Pharmacogenetics, 4, 275-280 Esrig, D.E., Elamjian, D., Groshen, S., Freeman, J.A., Carey, A.H., Waterworth, D., Patel, K., White, D., Little, Stein, J.P., Chen, S.C., Nichols, P.W,, Skinner, D.G., Jones, J., Noveni, P., Franks, S. & Williamson, R. (1994) P.A. & Cote, R.J. (1994) Accumulation of nuclear p53 and Polycystic ovaries and premature male pattern baldness tumor progression fn bladder cancer. N. Eпgl. J. Med., are associated with one allele of the steroid metabolism 331, 1259-1264 gene СУР 17. Hum. Mol. Genet., 3(10), 1873-1876 Evans, H.J. (1993) Molecular genetic aspects of human Cartwright, R.A., Glashan, R.W., Rogers, H.J., Ahmad, cancers: The 1993 Frank Rose Lecture. Br. J. С aпсеr, 68, R.A., Barham-Hall, D., Higgins, E. & Kahn, M.A. (1982) 1051-1060 Role of N-acetyltransferase phenotypes in bladder car- Evans, D.A.P., Mahgoub, A., Sloan, T.P., Idle, J.R. & Smith, cinogenesis: a pharmacogenetic epidemiological R.L. (1980) A family and population study of the genetic approach to bladder cancer. Lancet, 2, 842.-845 polymorphism of debrisoquine oxidation in a white Cavenee, W.K., Dryja, T.P., Philips, R-A., Benedict, W.F., British population. J. Med. Genet., 17, 102-105 Godbout, R., Gallie, B.L., Murphee, A.L., Strong, L.C. & Fearon, E.R., Cho, K.R., Nigro, J.M., Kern, S.E., Simons, White, R.L. (1983) Expression of recessive alleles by chro- J.W., Ruppert, 3.1v1., Hamilton, S.R., Preisinger, A.C., mosomal mechanisms in retinoblastoma. Nature, 305, Thomas, B., Kinkier K.W. et al. (1990) Identification of a 779-784 chromosome 18q gene that is altered in colon cancer. Cavenee, W.K., Murphree, А.L., Shu11, M.M., Benedict, Science, 247, 49 W.F., Sparkes, R.S., Koch, E. & Nordenskjold, Zvi. (1986) Fuqua, S.A.W., Hî11, S.M., Chamness, G.C., Benedix, Prediction of familial predisposition to retiпоыastоma. M.G., Greene, G.L., O'Malley, B.W. & McGuire, W.L. New Engi. J. Med., 314, 1201-1207 (1991) Progesterone receptor gene restriction fragment Ctien, H., Sandier, D.P., Taylor, J.A., Shore, D.L., Liu, E., length polymorphism in human breast tumors. J. Nati Bloomfield, C.D. & Bell, D.А. (1996) Increased risk for CяncerInst., 83, 1157-1160 myelodyspIastic syndromes in individuals with glu- Gawin, B., Klamt, B., Konig, A., Thate, C., Le Paslier, D., tathione theta 1(GSTTI) gene defect. Lancet, 347, 295-297 Chumakov, I., Bhogal, R., Zehetne r, G., Bruns, G. & Chissoe, S.L., Bodenteich, A., Wang, Y.F., Wang, Y.P., Gessler, M. (1995) An integrated YAC clone for the

247 Application of Biomarkers in Cancer Epidemiology

WAGR region on human chromosome 11р13-р14.1. NA., Ding, W,, I-Iussey, C.,Tran, T. Miki, Y. & Weaver- Genornics, 30, 37-45 . Feld, . (1994) Analysis of p16 gene (CDKN2) as a candi- Green-Gallo, L.A, Buivgs, D.M., Fisher, K.L., Caporaso, date locus for the 9p melanoma susceptibility locus. Nature Genet., 8, 23-26 N., Slawson, R.G., Elias, G., Didolakar, M.S., Ivusich, W.I. & Resau, J.H. (1991) A protocol for the safe administra- Kellerman, G., Shaw, C.R. & Luyten-Kellerinan, M. tion of debrisoquine in biochemical epidemiologic (1973) Aryl hydrocarbon hydroxylase inducibility and research protocols for hospitalized patients. Cancer, 68, bronchogenlc carcinoma. New EngL J. Med., 289, 206-210 934-937

Haas, О.A., Argyriou-Tirita, A. & Lion, T. (1992) Parental Khoury, M.J., Adams, Jr, M.J. & Flanders, W.D. (1988) An origin of chromosomes involved in traпsloсапon t(9,22). epidemiologig approach to ecogenetecs. Am. J. Horn. Nature, 359, 414-416 Genet., 42, 89--95 Hall, J.M., Lee, M.K., Newman, B., Morrow, J.E., Khoury, M.J. & James, L.M. (1993) Population and famil- Anderson, L.А., Huey, B. & King, M.C. (1990) Linkage of ial relative risks of disease associated with environmen- early-onset beast cancer families to chromosome 17g21. tal factors in the presence of gene-environmental triter- Science, 250, 1684-1689 action. Am. J. Epidemiol., 137(11), 1241-1250 Hanawalt, P.C. (1996) Role of transcription-coupled DNA Kйzler, K.W., Nilbert, IC., Vogelstein, B., Bryan, T.M., repair in susceptibility to environmental carctuogenesis. Levy, D.S., Smith, K.J., Preisinger, А.C., Hamilton, S.R., Environ. Health Perspect., 104, Suppl. 3, 547-551 Hedge, P., Markham, A. et al. (1991) Identification of a Hilldesheim, A., Chen, C.J., Caporaso, N.E., Cheng, Y.J., gene at chromosome 5g21 that is mutated in colorectal cancer. Science, 251, 1366-1370 Hoover, R.N., Hsu, M.M., Levine, Р.~., Chen, I.14., Chen, J.Y., Yang, C.S., Daly, A.K. & Idle, J.R. (1995), Cytochrome Knudson A.G. (1993) All in the (cancer) family. Nat. Р4502Е1 genetic polymorphisms and risk of nasopha- Genet., 5, 103-104 . ryngeal carcinoma: results from a case-control study con- Knudson, A.G., Hethcote, H.W. & Brown, B.W. (1975) ducted in Taiwan. Cancer EpidemioL Biomarkers Prey., 4, 607-610 Mutation and childhood cancer: a probabilistic model for the incidence of retinoblastonsa. Proc. Soc. Natl Acad. Hussussron, C.J., Struewing, J.P., Goldstein, A.M., Sd. USA, 72, 5116 Higgins, P.A.T., Ally., D.S., Stratan, M.D., Clark, W.H., Jr, Knudson, .G. (1983) Model hereditary cancers of man. Tucker, M.A. & Dracopoli, N.C. (1994) Germline p16 А Prig. Nod. Acid Res. Mol. [oI., 29, 17-25. mutations in familial melanoma. Nature Genet, 8, 15-21 В Lang, N.P., Butler, M.A., Massengill, J., Lawson, M., Irvine, R.A., Yu, M.C., Ross, R.K. & Coetzee, G.A. (1995) Stotta, R. ., et . (1994) Rapid metabolic phenotypes for The CAG and GGC microsatellites of the androgen С ы acetyltransferase and cytochrome 4501 2 and putative receptor gene are in linkage disequilibrium in men with Р А exposure to food-borne heterocycic amines increase the prostate cancer. Cancer Res., 55, 1937--1940 risk for colorectal cancer or polyps. Cancer Epidem loi. Issa, J.-P.L, Baylin, S.B. & Belinsky, S.A. (1996) Biornarkers Prey., 3, 675-682 Methylation of the estrogen receptor CpG island in lung Lee, WI-I,, et al. (1979) Human retinoblastoma suscepti- tumors is related to the specific type of carcinogen expo- bility gene: cloning, identification, and sequence. sure. Cancer 11es., 56, 3655-3658 Science, 235, 1394 Jacques, .F., Bostom, A.G., Willaims, R.R., Ellison, R.C., Р Lohmann, D., Horsthemke, B., Gillessen-Kaesbach, G., Eckfeldt, J. . etal. (1996) Relation between folate status, Н Stefani, LI. & Hofler, H. (1992) Detection of small RB1 a common mutation in methylenetetrahydrofolate gene deletions in retinoblastoma by multiplex PCR and reductase, and plasma homocysteine concentrations high resolution gel electrophoresis. Hum . Genet., 89(1), (1996). Circulation, 93(1), 7-9 49-53 Johnson, R.L., Rothman, .L., Xie, J., Goodrich, L.V., А Lower, G. ., Nilsson, T., Nelson, .E., Wolf, H., Gamsky, Bose, J.W., Bordfas,J.M., Quinn, A.G., Myers, R.M., Cox, М С .E. & Bryan, G.T. (1979) N-Acetyltransferase phenotype DR., Epstein, E. . & Scott, M.P. (199 6) Human homolog Т Н and risk in urinary bladder cancer: Approaches in mole- of patched a candidate gene for the basal cell nevus syn- cular epidemiology. Preliminary results in Sweden and drome. Sde ce, 272, 1668-1671 п Denmark. Environ. Health Perspec., 29, 71-79 Kalow, W. (1962) Pharmacogenetics: heredity and Malkin, D., L1, F.P., Stong, L.C., Fraumeni, J.F. Jr, Nelson, response to drugs. Philadelphia, W.B. Saunders C.E., Khn, D.H., Kassel, 1J., Gryka, M.A., Bischoff, F.Z., & Kamb, A., Shattuck-Eidens, D., Eeles, R., Liu, Q., Gruis, Tainsky, M.A. (1990) Germline p53 mutations in a fami1-

248 Biomarkers in the study of the genetics of human cancer lai syndrome of breast cancer, sarcoma, and other neo- Teloinere shortening associated with disease patterns Cancer Res., 54, plasms. 5сјепсе, 250, 1233-1238 in myelodysplastic syndromes. 3557-3560 McGlynn, K.A., Rosvold, E.A., Lustbader, E.D., Hu, Y. et aL (1995) Susceptibility to hepatoce11ular carcinoma is Ott, J. (1991) Analysis o f human genetic Iinkage. Baltimore, associated with genetic variation in the enzymatic detox- MD, The Johns Hopkins University Press ification of aflatoxin B1. Proc. Nat! Asoc. ScL, USA, 92, Papadopoulos, N., Nicolaides, N.C., Wei, .P., Ruben, 2384-2387 У S.M., Carter, K.C., Rosen, C.А. Haseltine, \W.A., McMichael, A.J. (1994) "Molecular epidemiology" - new Fleishmann, R.D., Fraser, C.M. & Adams, M.D. (1994) pathway or new traveling companion. Am. J. EpidemioL, Mutation of a mutL homolog on hereditary colon cancer. 140, 1-11 Science, 263, 1625-1629 Miki, Y., Swensen, J., Shattuck-Eidens, D., Futreal, P.A., Perera F.P. (1982) Molecular cancer epidemiology: a new Harshman, K., Tavtigian, S., Liu, Q., Cochran, C., tool in cancer prevention. J. Nat! Cancer Inst., 78, Bennett, L.M. & Ding, W. et aL(1994) A strong candidate 887-898 for the breast and ovarian cancer susceptibility gene Rowley, J. Chromosome abnormalities in cancer. (1980) BRCA1. Science, 266, 66-71 Cancer Genet. Cyto., 2, 175 Mulligan, L.M., Kwok, J.B., Healey, С .S., Elsdon, M.J., Ryberg, D., Kure, E., Lystad, 5., 5kaug, V., Strageland, Eng., C., Gardner, Е., Love, D.R., Mole, S.E., Moore, J.K, Mercy, I., Borresen, A.L. & Haugen, A. (1994) p53 muta- Papi, L. (1993) Germ-line mutations of the RET proto- tions in lung tumors: relationship to putative suscepti- oncogene in multiple endocrine neoplasia type 2А. bility markets for cancer. Cancer Res., 54, 1551-1555 Nature, 363, 458-460 5avitsky, K., Bar-shira, A., Gilad, S., Rotman, G., Ziv, Y., Nagase, H., Miyoshi, Y, uni, A., Ogawa, M., Vanagaite, L., Tagle, D.A., Smith, S. Uzie1, T., Sfez, Ѕ. et al Utsunomiyi, J., Baba, S., SaSаzukf, T. & Nakamura, Y. (1995) A single ataxia telangiectasia gene with a product (1992) Correlation between the location of germline similar to PI-3 Kinase. Science, 268, 1749-1753 mutations in the APC gene and the number of colorec- ta1 polyps in familial adenomatosis polyposis patients. Scnivner, C.R., Beaudet, A., Valle, D. & Sly, W.S. (eds) Cancer 1! es., 52, 4055-4057 (1995)Тhе metabolic and molecular basis of inherited Nakazawa, H., English, D., Randell, P.L., Nakaema, K., disease, Vols I—IП, 7th edn. McGraw-Hill, NY. Matel, N., Armstrong, B.K., Yaroasaki, H. (1994) UV and Seidegard, J., Pero, R.W. Miller, D. & Beattie, Е.S. (1986) skin cancer: specific p53 mutations in normal skin as a A glutathione transferase in human leukocytes as a biologically relevant exposure measurement. Proc. Nat! marker for susceptibility to lung cancer. Carcinogenesis, 7, Acid. Sc!., USA, 91, 360-364 751-753 Nicolaides, N.C., Papadopoulos, N., Wei, Y.F. (1994) Selzinger, B.R., Rouleau, G., Ozelius, L.J., Lane, A.H., 5t Mutations of two PMS homologues in hereditary non- George-Hyslop, P., Huson, S., Gusella, J.F., Marhiza, R.L. polyposis colon cancer. Nature, 371, 649 (1987) Common pathologic mechanism for three tumor Nowell P.C. & Hungerford, D.A. (1960). A minute chro- types in bilateral acoustic neurofibromatosis. Science, mosome in human chronic granulocytic leukemia. 236,317-319 Science, 132, 1497 Sharma, H.W., Maltese J.-Y., Zhu, X., Kaiser, H.Е. & Nystrom-Lahti, M., Parsons, R., Sistonen, P., Pylkkanen, Narayanan, R. (1996) Telorneres, telomerase and cancer: 16, 511-516 L., Aaltonen, L.A., Leach, F.S., Hamilton, S.R., Watton, P., is theie a magic bullet? Anticancer Res., Bronson, E., Fusaro, R., et al. (1994) Mismatch repair Shaw, G.L., Falk, R.Т., Cagorasd, N.E., Issaq, H.J., Kale, genes on chromosome 2p and Зр account for a major R.G., Fox, S.D. & Tucker М.~. (1990) Effect of diurnal share of hereditary nonpolyposis colorectal cancer fami- variation on debnisoquine metabolic phenotyping. J. lies evaluable by linkage. Am. J. Hum. Genet., 55, 659-665 Nat? Cancer Inst., 82, 1573-1576 Ida, Y., Tanaka, M. & Nakanishi, L (1994) Relation Shaw, G.L., Falk, ~.T., Deslavriers, J., Nesbitt, J.C., Frame, between the occurrence of K-ras gene point mutations J.N. & Hoover, R.N. (1994) Lung tumor resection does and genotypes of polymorphic N-acetyltiansferase in not affect debrisoquine metabolism. Cancer Epidеmiol. human colorectal cancer carcinmomas. Carcinogenesis, Biomarkers Prev., 3, 141-144 15 (7) 1365-1369 Sоiпi, Y., Chia, S.C., Bennett, W.P., Groopman, 3.1. , thyashiki, ~.H., Ohyasbiki Km Fujimura, T., Kawakubo, Wang, J.S., DaBenedetti, V.M., Cawley, H., Welsh, J.А., K., shimamoto, T., Iwabuchi, A. & Toyama, К.(1994) Hasen, C., Bergasa, N.V., Jones, E.A., DisBisceglie, A.M.,

249 Application of Biomarkers fn Cancer Epidemiology

Trivers, G.L., Sandoval, C.A., Calderon, 1.E.. Munoz Walman, F.M., Carroll, P.R. & Kerschmann, R. (1991) Espinoza, L.E. & Harris, C.C. (.1996) An aflatoxin-associ- Centromeric copy number of chromosome 7 is strongly ated mutational hotspot at codon 249 in the p53 tumor- correlated with tumor grade and labeling index in spressor gene in ICC. Cancer, 17(5), 1007-1020 human bladder cancer. Cancer Res., 51, 3807-3813 Sozzi, G., Veronese, M., Negrliii, M., Baffa, R., Cottitelli, Wertalecki, W,, Rouleau, G.A., Superneav, D.W., M.G., Inoue, H., Tornelli, S., PIlitti, 5., De Gregorio, L., Forehand, L,W., Williams, J.P., Haines, J.L., Gusella, J.F. Pastorino, U., Pierotti, M.D., Obta, M., Huebner, K. & (1988) Neurofibromatosis 2: cцnical and DNA linkage Croce, C.M. (1996) The FlIT gene at 3р14.2 is abnormal studies of a large kindred. N. Eng1. J. Mcd., 319 (S), in lung cancer. Cell, 85, 17-26 27 8-283

Sparkes, R.S., Murphree, А.L. & Lingua, R.W. (1983) Gene Whang-Peng, J,W.P., Kao-Shan, C.S., Lee, E.G., & Bunn, for hereditary retinobla stoma assigned to human chro- PA., Carney, D.N., Gazdar, A.F. & Minna, J.D. (1982) mosome 13 linkage to esterose-D. 5сгеnсе, 219, 971-973 Specific chromosome defect associated with small-cell Taioli, E., Tracbman, Chart, X., Toniolo, P. & Garte, S. lung cancer: deletion of Зр(14-23). Science, 215, 181-182 (1995) A CYP1A1 restriction fragment length polymor- Wooster, R., Neuhausen, S.L., Mangion, J., Quirk, Y., phisrn is associated with breast cancer in African- Pond, D., Collins, N., Nguyen, K., Seal, S., Tran, T., Averill, American women. Cancer Res., 55, 3757-3758 D. et al. (1994) Localization of the breast cancer suscep- Tanaka, K., Maoyuki, lei., 5atokata, I., et aI. (1990) tibility gene, BRCA2 to chromosome 13g12-13. Science, 266, 2088-2090 Analysis of а human DNA excision repair gene involved in group A XP containing a zinc-finger domain. Nature, Wu, W. J„ Kakehi, Y., Habuchi, T., Kinoshita, H., Ogawa, 348, 73-76 O., Terachi, T., Huang, C-H,, Chiang, C.P. & Yoshida, 0. (1995) Allellic frequency of p53 gene codon 72 poly- Taylor, J.A., Hiivonen, А., Watson, M., Pittman, G., Mohler, J.L. & Bell, D.A. (1996) Association of prostate morphism in urologic cancers. Jpn J. Cancer Res., 86, cancer with vitamin D receptor gene polymorphism. 730-736 Cancer Res., 56, 4108-4110 Yokota, J., Wada, M., Shimosato, Y. et al. (1987) Loss of Thakker, R.V., Wooding, C. & Pang, J.T. (1993) Linkage heterozygosity on chromosomes, 3,13,17 in small cell analysis of 7 polymorphic markers at chromosome lung carcinoma and on chromosome З in adenocarci- IIpII.2-l1g13 in 27 multiple endocrine neoplasms type I noma of the lung. (1987) Proc. Nat1 Acad. Sd. USA, 84, families. Ann. Hum. Genet., 57, 17-25 9252-9256 Uematsu, E, Uckuchi, H., Motomiya, M., Abe, T., Sagami, Zoo, L., Weger, J., Yong, Q., Goldstein A.M., Tucker, I., Ohmachi, T. et a1. (1991) Association between restric- M.A., Walker, G.J., Hayward, N. & Dracopoli, N.C. (1996) tion fragment length polymorphism of the human Germilne mutations in the р16INХ Yд binding domain of cytochrome P450IIE1 gene and susceptibility to lung CDK4 in familial melanoma. Nature Genet. 12, 97-99 cancer. Jpn J. Cancer Res., 82, 254-256 Upadhya, M., Maynard, J., Osborn, M., Huson, S,M., Ponder, M., Ponder B.A., & Harper, P.S. (1995) Corresponding author Characterisation of germline mutations in the neuroti- N. Caporaso bromatosis type 1 (NF1) gene. J. Med. Genet. 32, 706-710 Genetic Epidemiology Branch Vogelstein B., Fearon, E.R. & Hamilton, S.R. (1988) National Cancer Institute, EPN 439 . Genetic alterations during colorectal tumor develop- 6130 Executive Blvd., F{ockville, MD 20692, USA ment. N. EпgI. J. Med., 319, 535-532

250 Application il Biomarkera h Cancer Epidemiology Tinde, P. Botfetla, Р. Phukor, D.E.G., Rothman, N. Hеlka, B, arid Pearce, N. eds 'ARC Scientific PuЫicatioпs No. 142 lHteгmHanat Agency for Research en Cancer, Lyon, 1997

Gene - environment interactions in the application of biomarkers of cancer susceptibility in epidemiology

S. Garte, C. Zocchetti and E. Taioii Metabolic susceptibility genes are important determinants of individual susceptibility to the effects of environmental carcinogens. These genes follow the form of `type 2' gene-environment interaction, whereby the polymorphic genetic risk factor functions only in the presence of an environmental exposure.Two different effects of carcinogen dose have been observed for these genes. Sometimes, increasing dose leads to a decreasing interaction, so that cases with the genetic risk factor have lower exposures than those cases without it. Other examples of a direct dose effect, whereby increasing exposure leads to increased interaction, have also been described. We propose a model based on multiple logistic regression to assess the nature of the dose effect iп this type of gene-environment interaction. This model allows for distinction between these two dose effects, and other effects such as protective or non-interactive effects of environmental and genetic risk factors.

It has been known for many decades that human in 1946; the great geneticist Haldane (1946) dis- diseases are caused by some combination of envi- cussed gene-environment interactions in disease ronmental and genetic factors. The relative influ- causation. Since then, a substantial literature has ence of the two is obviously quite variable and been produced on the interaction between genetic forms a spectrum with certain highly infectious and environmental agents, and even a review of diseases at one end and severe genetic diseases of the review literature would not be feasible here. metabolism at the other. For the great majority of This paper will instead be limited to a discussion of human diseases (excluding extreme cases such as the models of gene-environment interactions that Ebola fever or Down syndrome), purely environ- are of primary importance in the area of individual mental or purely genetic etiologies are insufficient human susceptibility to the carcinogenic effects of to explain individual variability in occurrence, environmental carcinogens. With the introduction prognosis or outcome. This is especially the case of new biomarkers of cancei susceptibility, new with the most important chronic diseases of mod- paradigms for the classical terms of interaction and ern industrialized society, including heart disease confounding are needed (ffulka, 1991). The focus and cancer (Hirayama, 1989; Hegele, 1992; Lane will be on cancer as the disease end-point, and in et al., 1992; Hunt et al., 1993; Tiret et aL, 1993; particular on the most common (usually termed Brennan & 5ilmaп, 1994; Hayden et al., 1994; sporadic) category of cancer in the human popula- Hwang eiaL, 1995). For these categories of disease, tion, as opposed to cancers such as retinoblastonra, a great number of environmental and genetic risk Wilms' tumour and other examples of purely or factors have been identified, and it is probably safe mostly inherited cancers (which also include the to say that for all cases of cancer both types of fac- familial cancers of the breast, colon and brain). tors must play some role in disease causation. In There is a general consensus that the majority of general, when one assumes multiple causes, it is cases of human cancer (sporadic tumours of the useful to ask whether interactions between these lung, bladder, breast, etc.) are of largely environ- independent causes exist. The issue of interaction mental origin, where environment is defined between the environment and genetic factors is an broadly as including lifestyle and exposures to old one in biology, and in a classic paper entitled agents contained in cigarette smoke and diet, as `The interaction of nature and nurture' published well as occupational or environmental exposure to

251 Application of Biomarkers in Cancer Epidemiology

Metabolic genes DNA repair genes Carcinogen exposure Metabolism r DNA damage

Tumour suppressor Mutation- Oncogens genes Progression —Y Cancer

Figure 1. Mechanistic pathway to cancer. carcinogens. Evidence supporting a dominant radiation to cigarette smoke, interact either di- environmental etiology of human cancer comes rectly or indirectly with the structure and/or func- from decades of epidemiological research on migrant till of oncogenes and tumour suppressor genes. populations and other studies (Higginson, 1980; 5отаtiс mutations in these genes caused by expo- Doll & Peto, 1981; Garte 1992). Despite the impor- suie to chemicals and radiation are of critical tance of environmental exposure in human cancer, importance in the etiology of sporadic human can- the evidence for some form of genetic influence on cer. Such interactions may represent some of the almost all cancer etiology is quite compelling. most important mechanistic pathways in human Unlike most strains of laboratory animals used in carcinogenesis. The results of these interactions, carcinogenesis research, human beings are highly such as mutations or gene deletions, have been outbred and genetically diverse. They also show an proposed as bionvarkexs of biological effect for cer- extremely broad range of phenotypic responses to tain carcinogens, such as aflatoxtn (нolktеиn et ai., environmental stimuli. Foi example, there is а 1993) and ionizing radiation (Vahakagas et al., wide variation in human response to specific car- 1992). However, even in the absence of such inter- cinogenic exposures, cigarette smoking being one actions, these genes may still play an important role of several examples (Sellers et al., 1992). This in cancer causation if activating or inacti- interindividual variability in response is seen not vating mutations occur in the germline, such as in only in disease end-points but also when one uses the case of retinoblastoma or Li-Fraumeni syndrome biomarkers of exposure or biological effect such as (Friend et a2., 1988; Frebourg & Friend, 1992), or in DNA adducts or others (Marquis & Siеk, 1988; the case of rare H-ras alleles (Conway et al., 1995). Harris, 1989; Perera et aL 1991). While such vari- Another category of genes that are important in ability presents formidable obstacles to the devel- human carcinogenesis does not play any direct role opment of certain biomarkers, it also presents an in the mechanistic pathway leading from carcirno- exciting opportunity for research into the mecha- gen exposure to cancer; instead these genes have an nisms behind the variability and the development influence on events that occur on the pathway. of markers of susceptibility in individual humans. Genes involved in DNA repair (Lehman et al., 1992; Samson, 1992; Smith et ai., 1994а) are an example of Genes and human cancer this category, as are genes that mediate the conver- It is vitally important at this stage to distinguish sion of carcinogenic chemicals to their ultimate between different types of genes that are involved active forms. The distinction between these two in human carcinogenesis. One broad category of types of 'cancer genes' with respect to their position cancer genes includes both the dominant acting on the mechanistic pathway to cancer is illustrated oncogenes, such as ras and myc, and the recessive in Fig. 1. Genes responsible for the repair of DNA tumour suppressor genes such as p53 aid Rb. These damage by carcinogens have been intensively genes exert their effects as part of the biological studied because of their importance in carcino- pathway leading to tumorigenesis. There is ample genic mechanisms. They are certainly participants evidence that environmental carcinogens, from in one type of gene–environment interaction.

252 Gene-environment interactions in the application of biomarkers

However, they have not been widely used to date phenotype and later by genotype analysis) as bio- as biomarkers of either biological effect or suscep- markers of human cancer susceptibility. By defini- tibility. This situation is likely to change with the tion, these genes function only in the context of recent cloning of the gene for ataxia telangiectasia interaction with the environment, since the sub- (AT) (Savitsky et aL, 1995). Caniers of the homozy- strates of their gene products are xenobiotic chem- gins variant of this gene have greatly increased icals or their metabolites. sensitivity to the carcinogenic effects of ionizing The metabolism of carcinogens, like that of most radiation. While homozygotes (who show the toxic agents, generally proceeds through two phases. extreme symptoms of the disease) are very rare it (Garte & Kneip, 1988). In the first, unreactive, has been estimated that the frequency of het- non-polar compounds are converted, usually by erozygotes who may also have some degree of oxidation reactions, to electrophilic highly reactive increased susceptibility to radiation-induced car- intermediates. These are then able to form com- cinogenesis may be as high as 10% of the popula- plexes with conjugating molecules such as glucose tion. Further study of this and other repair-related or glutathione in phase 2 conjugation reactions. genes as biomarkers of cancer susceptibility, and in The complex is usually harmless and easily relation to interaction with environmental factors, excreted. However, the electrophilic metabolite should provide important and useful information may be able to react with other cellular nucleo- in the future. philes such as DNA before conjugation can occur. The focus of this report will be on a number of This is often the first step in the initiation of a car- genes that have been studied for many decades, cinogenic process. after the initial discovery by Е.C. and J.A. Miller Table 1 presents a summary of the most cons- (Convey et al., 1956) that most carcinogens monly studied metabolic susceptibility genes. The undergo complex metabolic reactions and that the cytochrome P450 enzymes, which represent a large actual carcinogenic compounds responsible for multigene family with differing substrate specifiсi DNA binding or damage are usually electrophilic ties, are important in phase 1 reactions. The metabolites of the parent compound to which the CYPIA1 gene product, aromatic hydrocarbon organism is exposed. As shown in Fig. 1, these hydroxylase (AHI), for example, catalyses the first genes act as effect modifiers on the carcinogenic oxidative step in the metabolism of polycyclic aro- pathway from exposure to cancer. matic hydrocarbons, such as those found in tobacco smoke, to carcinogens. An MspI restriction Metabolic genes fragment length polymorphism (RFLP) in the 3' A number of genes involved in the metabolism of non-coding region of CYP1A1 was found to be carcinogens have been shown to play a role in the associated with AHI enzymatic activity in a fam- risk of certain human cancers. In most cases, the ily study (Petersen et а1., 1991). A second polymor- putative biochemical mechanism by which such phism was found in the catalytic region (exon 7) of genetic factors exert their effects is fairly straight- the CYPIA1 gene, closely linked to the MspI RFLP forward and is related to the actual dose of the (Hayashi etal., 1991; Hirvonen etal., 1992; Cosma active carcinogenic metabolite that reaches the et aI., 1993b). Both polymorphisms have been genome in the target cell. Increased metabolism of associated with lung cancer in Japanese popula- a carcinogenic precursor to the ultimate carcinogen tions but not in Caucasian (Kawajiri et al., 1990; and loss of function of a conjugation mechanism Tefre et al., 1991; Hirvonen et al., 1992; Shields et for elimination of the active metabolite are exam- al., 1993) or African-American (Shields et aL, 1993; pies of how alterations in the activity of certain sugimura et al., 1994) populations. This discrep- gene products can affect the biologically relevant ancy may reflect the different frequencies of the dose in different individuals, even when exposures two polymorphisms between ethnic groups, as we are equivalent. Reviews of metabolic susceptibility previously reported (Calma et al., 1993b; Taioli et genes have been published (Idle et aI., 1992; Daly а!., 1995b). We have recently found a very strong et al., 1994; Hirvonen, 1995; Raunio et ad., 1995; association between breast cancer risk and the Rothman, 1995; Vineis, 1995), and there is a grow- homozygous MspI RFLP in AfrIcan-American women ing literature on the use of these genes (first by (Taioli etal., 1995с). We have also described a third

253 Application of Biomarkers in Cancer Epidemiology

Gene Metabolic pathway Cancer sites

GST14i Conjugation of organic opoxides with reduced Lung, bladder, colon, stomach, breast, liver glutathione

СУР 2D6 Hydroxylation of lipophnic xenobiotics, possibly NNK Lung, bladder, breast. NAT2 N-Acetylation of arylamines and N-hydroxylated Bladder, lung, colorectal, breast heterocyclic arylamines

CYPiAi Metabolism of polycyclic aromatic hydrocarbons, Lung, stomach, colon, breast TODD and estrogens

ÇУP2E7 Oxidation of N-nitrosamines, alcohol Lung, Ьаддег, colon

polymorphism in the human CYPIAI gene, an possibly other carcinogenic substances, and has Afriсап-American-specific MspI RFLP in intron 7 been associated with increased risk of lung cancer (Crofts et a2., 1993). This RFLP, which has not been in several studies of Caucasians and Asians. Another detected in over 300 Caucasians nor in Asians, gene first identified as a metabolic susceptibility occurs in 15% of African-Americans and is associ- marker through early phenotype analysis is ated with an increased risk of adenocarcinoma of N-acetyl transferase or NAT2 (Meyer 1994; Vineis et the lung (Taioh et ai., 1995а). аl., 1994; Yu et al., 1994). A homozygous polymor- Other phase 1 genes that have been identified as phism in this gene renders individuals slow acety- susceptibility factors are СУР 1А2 (Sinha et al., lators of certain substrate drugs such as isozionid. 1994; Catteau et aL, 1995), a gene that is mduced by The NAT2 gene participates in a complexmeta- heterocyclic aromatic hydrocarbons; CYP2D6, which bolic web, and individuals with the polymorphism is responsible for the metabolism of the drug (slow acetylators) may be at higher risk for bladder debrisoquine; and СУР 2E1 (Stephens et ai., 1994; cancer from carcinogenic exposure to arylarnines, but Kato etal., 1995; Watanabe etal., 1995). Phenotype at lower risk for colon cancer. This complication is analysis suggested a role for the debrisoquine роlу- not surprising given the tissue specific complexity morphism as a susceptibility factor in human lung of metabolic pathways and competing reactions cancer, but this has become less clear as genotype data catalysed by a number of genes. has been gathered. The specific carcinogen associ- As discussed by Caporaso (1996), there are ated with the CYР2D6 activity and the biochemi- major differences between metabolic gene poly- cal effect of the polymorphism are not known. morphisms and inherited mutations in cancer The СУР 2Е1 gene is inducible by ethanol, and is genes of the first category (such as BRCA1, for involved in the metabolism of such carcinogens as example). These differences have profound impli- butadiene, benzene and carbon tetтachloride. cations for screening, prevention, public health Two phase 2 genes have received wide attention and risk assessment. For example, polymorphisms as metabolic susceptibility markers. The gene cod- in metabolic genes tend to be much more common ing for one form of glutathione S-transferаsе, in the population (from 5 to 50%) than mutations in G5T11, is missing in about one-half of Caucasians cancer genes. At the same time, the increased cancer (Hirvоneп et al., 1993; Alexandrie et al., 1994; risks associated with metabolic gene polymorphisms Ichiba et a2., 1994; Kihara et aL, 1994; Nakajima et are usually on the order of twofold compared to аl., 1995). This null allele results in a lower level of very high odds ratios for inherited mutations in glutathione conjugation of РАН metabolites and tumour suppressor genes or oncogenes. Perhaps

254 Gene—environment interactions in the application of biomаrkers the most important difference relates to the clin!- make-up. It is certainly possible, however, that cal significance of these inherited genetic risk fac- future research will render this type of GEl highly tors. It is not dear how a physician сап counsel a significant for human cancer. The fourth type person who is found to contain a genetic predis- described by Khoury is common and important i position for cancer (such as is the case in Li- cancer, and occurs when both the exposure and the Fraumeni, etc.) independently of any exposure. On GRF carry some risk for disease, but the combina- the other hand, since metabolic gene polymor- tion is interactive and/or synergistic. The cancers phisms specifically confer increased sensitivity to associated with the DNA repair gene deficiencies the effects of environmental carcinogens, such as AT (see above) or xerodenma pigmerltosum increased surveillance and avoidance of such (a repair gene deficiency disease associated with agents may be an effective strategy for cancer pre- exposure to UV radiation) are examples. Most of vention for individuals carrying such polymor- the first category of cancer genes (c-rnус, р53, etc.) phisms. belong to this type of GET, since the gene mutations themselves carry an increased risk of cancer which Types of gene—environment interactions is exacerbated by exposure to environmental car- A number of authors have discussed various bio- cinogens (which are still carcinogenic in the absence logical forms of gene—environment interaction as of such gene mutations). The latter two types of applied to epidemiological studies. M.J. Khoury GEI described by Khoury refer to cases in which and his co-workers, stressing the population and the GRF is protective. family genetics aspect of such interactions and Ottman (1990, 1994, 1995) has also described their implications for public health, have described five similar types of gene—environment interac- six types of gene—environment interaction (GEI) tion. In the first, the disease may be caused by (Khouiy et cd., 1988, 1993, 1995; Khoury & James, either the genetic or the environmental agent, but 1993; Khoury & Wagener, 1995). In the first type, the genotype increases the expression of the agent. neither the environmental exposure nor the The second and third are the same as those genetic risk factor (GRF) have any effect by them- described by Khoury. In the fourth type, both envi- selves in the absence of the other, but when both ronmental and genetic risk factors must be present are present, interaction between them causes dis- to cause the disease, equivalent to Khoury's type 1. ease. The example given is that of phenylalanine In the final model, both factors influence risk by exposure (in the diet) and pheny1ketonuria geno- themselves, but with an interaction between them. type. This type of GEl is rare and not important in The critical point made by both groups is that human carcinogenesis. A type 2 GEI is defined as the term interaction covers a variety of biological one in which the GRF has no effect on disease in phenomena involving gene products aid xenobi- the absence of the relevant exposure, but which otics. The specific form of the gene—environment can function to exacerbate the effects of the expo- interaction is clearly as important as the fact that sure. In this type, the exposure by itself increases such interaction exists. Type 2 interaction (as used the risk of disease, even in the absence of the GRF. by both R. Ottman and M.J. Khoury) is the most This is the most important type of GEl for human relevant to GEl related to metabolic susceptibility carcinogenesis related to metabolic susceptibility genes and human carcinogenesis. In this type, the genes and we will return to it shortly. The third cancer is caused by exposure to an environmental type is the converse of the second, in that the GRF agent. If there is no exposure, then the presence or can produce disease in the absence of exposure; absence of the genetic risk factor is irrelevant for exposure mediates the effect of the GRF but, with- disease causation. Only when the dose of the envi- out the GRF, has no role in disease etiology. While ronmental agent is greater than 0 does the GRF this type of interaction may be important in cer- have any role, and then it is to modify the effect of tain cases of human carcinogenesis, not enough is the exposure. In the terminology of Khoury, Rg known as yet regarding the detailed mechanisms (defined as the relative risk of the GRF alone) is by which specific carcinogens act to be abbe to say defined as equal to 1. In a hypothetical 2 x 2 table that any environmental carcinogen has effects (Table 2) showing odds ratios (OR) of disease as a only on those people with a particular genetic function of both a genetic and an environmental

255 Application of Biomarkers in Cancer Epidemiology

doses than the proportion of cases without the GRF. The phenomenon occurs whén the multi- Table 2. Hypothetical table for type 2 plicative effect of being GRF+ (versus GRF—) on the gene—environment interaction disease—exposure odds ratio becomes less as the GRF exposure dose increases. If the end-point is not Exposure - •1- cancer but, for example, some marker of exposure such as adducts, then subjects with the polymorphism tend to have higher relative levels of lia 1.0 adducts at lower doses of exposure, while at high 2.0 5.0 exposure no difference in end-point will be observed between those with and without the GRE. a Reference cot. While at first, such a result may appear to be coin- terintuitive, it can be explained by the increased sensitivity (due to higher genetic susceptibility) of risk factor, the reference cell (OR =1) is that where individuals with the GRE to lower levels of expo- neither gene nor exposure is present. For type 2 sure. The decreasing interaction with dose has GEI, the сен of gene =1(present) and exposure = O been observed for smoking with CYPIA1 and lung (absent) also has an OR = 1. This h а fundamen- cancer (Nakachi et aL, 1993; Taioli eta?., 1995а), for tally different situation with respect to epidemio- NАT2 and haemoglobin adducts (Vineis et al., logical analysis from the case with those types of 1994), and for СУР1А2 and adducts (Landi et al., GEI where the GRF can, by itself, cause disease. personal communication). A direct dose effect of the gene is observed when Dose effects in type 2 gene—environment cases with the GRF have environmental exposure interaction doses that are higher than cases without the GRE; A common approach to the investigation of the higher the dose, the greater is the effect of genetic susceptibility towards environmental car- having the GRF on any other end-point (such as cinogens is to use a case—control or other study disease, adducts, etc.) This can be explained simply design to determine the odds ratio for disease (can- by stating that the greater gene—environment cer) associated with a particular genetic polymor- interaction seen at higher environmental doses phism. In many of these studies, environmental increases the probability of an individual becom- exposures such as cigarette smoking are taken into ing a case beyond that seen due to exposure alone. account and analysed for interaction by multiple This direct effect has been seen with GSTM1 and regression. Many authors have inappropriately lung cancer (Hirvonen et al., 1993; Kihara et al., considered such exposure to be confounding 1994). However, a study of GSTM1 aid asbestosis (London et al., 1995). For the metabolic genes, (Smith et a1., 1994b) showed a decreasing interac- whose activity is not independent but in fact com- tion with dose, suggesting that this phenomenon pletely dependent on exposure, it is an error to is not simply gene-specific, but must be related to analyse the effect of the gene correcting for expo- the mechanism of action of the gene product lead- sure as a confounder. If a true type 2 GET is present, ing to the end-point being measured. In Fig. 2, we then no effect of the GRF is expected in the have plotted the frequency of cases (for cancer as absence of exposure. А spurious effect of the GRE the end-point) or of subjects with adduct values will be found if data are adjusted for exposure con- above the median (for adduct levels as the end- sidered as a confounder. point) who are positive for a particular GRF as a When the dose of environmental exposure function of exposure level from six separate stud- (such as smoking) is analysed with respect to geno- ies of metabolic gene polymorphisms as biomark- type of a metabolic susceptibility gene, two appar- ers of susceptibility. The curves slope upwards in ently divergent patterns are seen. The first instance two cases (Hirvonen et a1., 1993; Kihara et al., 1994) could be described as a decreasing interaction with which both report on GSTM1 and lung cancer. For dose. This is seen when the proportion of cases with the other four examples, two using CYP1AI and the genetic risk factor (GRF) have lower exposure lung cancer (Nakachi et al., 1993; Taioi et a?.,

256 Gene-environment interactions in the application of biomarkers

1995а), one for NAT2 and haemoglobin adduct (Vineis et cL, 1994), and one for GST and asbestosis (Smith et cL, 1994b), the curves slope downwards as a function of dose.

GE1 dose effect in relation to carcinogenesis dose—response models Analysis of these two apparently divergent scenar- ios reveals that they may not be contradictory. One possible explanation is related to the position on the carcinogenesis dose-response curve at which Low Medium High the GRF is functioning. While the actual shape of Dose the carcinogenicity dose—response curve is unknown for any human cancer, and is the subject Nat Slow of considerable research and discussion, certain general principles may be agreed on. At low doses СУР 1А1 AA+ (leaving aside the question of a threshold by ----- CYPIA1 VAL!VAL assuming that we are speaking of a region beyond -- - GST-(Asbestosis) a putative threshold), we may say that increasing ----- Gsт dose leads to increasing response. Whether the ------GST shape of this part of the curve is linear or follows some other function is not important to this dis- Figure 2. GRFs and smoking doses. cussion. At very high doses, the response must begin to level off, because the incidence approaches either 100% ox some other biologically negatives, there is less effect of increasing dose in determined maximal value. As one approaches this the positives than in the negatives. Therefore, sub- maximal response, increasing the dose has less of jects with the GRF may exhibit disease (or high an effect on the disease probability than is seen at adduct levels, etc.) at lower doses of environmental lower doses. In all cases of cardnogenesis or any agent than those without the GRУ. other toxicological end-point, a saturating dose According to this model, genes such as CYP1A1 must always exist at which no further effect can be or NAT2 show an inverse dose effect with respect to seen at higher doses. We can assume that the effect lung cancer or haemoglobin adducts, because the of a genetic susceptibility factor is to increase the doses of carcinogen present in cigarette smoke are carcinogenic response at any particular dose, for so high. The effect of GSTM1 on lung cancer with example by causing increased enzymatic activity smoking is not as clear since the dose is also high. or by altering the metabolic profile of an agent. An One possibility is that the substrate for GST is far important assumption here is that the GRF has no from the saturable level even when the external effect on the maximal response, but instead shifts dose is very high, as in the case of smokers. the dose—response curve to the left. At lower dose According to this model, a direct dose effect would levels, when the risk of the outcome is a small frac- be seen for CYP1A1 polymorphisms with cancer tion of the maximal value, the curves will appear related to low exposure doses of carcinogenic almost linear, and the GRF leads to an increase in hydrocarbons, and an inverse dose effect might the slope. Therefore, in this region, subjects with be observed for GSTM1 (as is apparently the case the GRF should respond more to higher doses of for asbestosis) for the appropriate mechanistic environmental agents than subjects without the pathway. GRF. However, if one is working at dose levels that It is important to understand clearly the mean- produce disease (or other end-point) risks closer to ing of the inverse dose effect in type 2 GE1 to avoid the maximum level, than the converse will be true. a confusing message. For a GRF that exhibits an In this region, although the overall response in inverse dose effect associated with smoking and GRF-positive individuals is higher than in GRF lung cancer, the following may be said: individuals

257 Application of Biomarkers in Cancer Epidemiology

who are positive and smoke only a few cigarettes a This expression corresponds to the assuption day are at relatively greater risk, compared with GRF that the risk of disease is due only to the action of negatives, than if they smoke two packs a day. This the environmental exposure, and the only effect does not mean that such people should smoke of the GRF X2 is to modify the coefficient of the more, because their risk of cancer still increases exposure term. Now we can say that: with dose compared with non-smokers. In fact, an inverse dose effect implies that for GRF-positive b* = Ь, + bзХ г = b1(1+ 2) (4) individuals, even a low smoking dose is highly risky; people carrying the polymorphism are at where a = b3/b3. higher risk of cancer in comparison to the general For example, let us assume that the effect of population exposed to smoke. Only complete GRF such as the ILh to VAL polymorphism in the smoking cessation, as well as the avoidance of CYPIAI gene is to increase the enzymatic activity other relevant exposures, can lead to canceт pre- of the gene product, as has been shown (Cosma et vention in the susceptible group. rel., 1993a; Crofts et al., 1994; Landi et a1., 1994; Taioli et aL, 1995b). The result of having this GRF An approach for the analysis of type 2 GEI is an increased level of metabolism, presumably In order to proceed further with analysis of type 2 leading to an increased concentration of the ulti- gene—environment interactions for metabolic genes, mate carcinogen, given a particular exposure dose. it is necessary to use analytical statistical tools on While this scenario may represent an oversimplifi- case--control and other study designs (Hwang et a]., cation, it can be seen that the effect of the GRF is 1994). We propose an approach to the analysis of to quantitatively modify the effect of the exposure such studies that may have the ability to detect the term. This would be reflected in a value for the specific form of dose effect in a quantitative man- term a that is greater than 0. Note that if a is neg- ner. The common way to describe interactions ative, the genetic factor would be protective. If the between the effects of an environmental agent and GRF has no effect on the exposure (for example, if a genetic risk factor is to use both terms in a mul- the exposure is to an agent that is not a substrate tiple regression model, and to also include a term for the gene product), then a = 0. If the GRF is that multiplies the GRF by the environmental absent, then X2 = 0. In either case, the risk is a func- agent. The coefficient of this interactive term then tion of exposure only, with no contribution from determines whether interaction is present, as йlus- the gene. trated in equation (1): Now we can rewrite the regression model of equation (2) if there are data for the effects of mul- G(Y) = a + ЫХ1 + b)(2 + b3Х 1Х Z (1) tiple (n) levels of exposure (doses) as: where Y is the odds of disease, X1 is the environ- G(Y) = а + be,E, + be2Е2+ ...+ be,,En + b(eg)1GЕ1 mental exposure, and X2 1s the GRF. The coeffi- b(eg)ZGЕ2±...+ b(eg)2GЕп (5) cients b„ b2 and b3 are determined by regression analysis using an appropriate computer program. where bel = b for exposure level Е1, (termed X1 in If we accept by definition that in the absence of equation 2), b(eg)z = b for exposure level E,, and G environmental exposure the presence of the GRE stands for the GRF (X2), etc. Using this notation, Е, by itself has no effect on disease outcome (the def- stands for dose, and the term for bg (or b2 in equa- inition of a type 2 interaction), then b2 in the tion 1) is defined as 0 and does not appear. For regression model of equation (1) is defined as equal each dose level, from equation (4), the term a is to zero. Thus equation (1) becomes: equal to:

G(Y) = я + b,Х 1 + b3Х 1Х Z (2) ar = b(eg)r /be, (6) which can be written in a different way: If values of a are plotted against dose, several outcomes are possible. If the slope of this plot is G(Y) = + (b3 + b3 а Х 2)Х 1 = a + b*Х З (3) positive, then the gene -eпvironment interaction

258 Gene-environment interac1ians in the application of biomarkers

Exposure

Genotype GRF None Low Medium High GSTM1+ - ND 1.0 a 1.71 1.71 1.03 3.4 5.04 G57ffл1- (null aiele} + ND CYP1A1 'LE - 1.0a 2.83 18.6 33.1 CYF1A1 VAL + 0.93 22.2 40.0 40.1

aReference cell.

case—control studies can allow for a precise defini- follows а direct dose effect. If the slope is negative, then an inverse dose effect is operative. Another tion of this function. However, this approach is way to express this is to say that, if the term useful in the characterization of genetic risk fac- > 1, then the dose effect is direct; if tors in terms of their dose effects, and, especially, < 1, then there is an inverse dose effect. If а has the advantage of avoiding confusion between is less than 4 at any particular dose level, then an inverse dose effect and a protective effect. Two the genetic factor is protective at that level. Such a examples from the literature will be used to illus- scenario, whereby a particular genetic polymorph- trate the method. ism may be a risk factor at one level of exposure but protective at a different level, is possible given Examples of type 2 GEf analyses the highly complex web of interconnecting meta- We will use the data from Kihara et al. (1994) to bolic pathways that usually operate in carcinogenic illustrate a direct dose effect. Table 3 shows the mechanisms. odds ratios (ORs) for each category of smoking Clearly, we assume that а is some function of exposure and genotype. The first point is that the dose, but the particular function 1s likely to vary OR for the cell where the GRF is present but where for every GRF and for different specific chemical there is no exposure is equal to the reference OR exposures. It is unlikely that the limited amount of for the absence of gene and exposure. This marks dose—response data that is usually available from the interaction as a type 2 interaction. We see that,

Table 4. Coefficients from regressionanalysis for gene—environment interaction

1 be1 0.535 1.039 2 be2 0.535 2.92 З be3 - З. 2.06 1 b(еg)1 0.66 2 b(eg)2 1.05 0.76 3 b(eg)3 _ 0:190 1.98 1 аi 1.233 2 1.96 0.26 Э - 0.054

259 Application of Biomarkers in Cancer Epidemiology

as exposure level increases, the risk of disease related to the exposure to decide whether a type 2 increases, and that the increase is higher when the interaction is logical within a mechanistic context. GRF is present for each category of exposure level. Tabk 4 shows the coefficients Fbe, b(eg)j obtained 5ummaгy and conclusions from the multiple logistic regression model using The study of metabolic gene polymorphisms as the SAS statistical package Genr od. Also shown in cancer susceptibility genotypes is likely to expand the table are the values for а the interaction term, in the future, given the advances in PCR-based which increase directly as a function of dose. Thus, technology aid the expected advances in knowl- for this case, there is a direct dose effect. 5imilar edge of the human genome. In terms of gene—envi- results can be obtained using the data from other ronment interactions, it is critical to develop tools sources for this gene and smoking-related lung with which to define precisely the role of suscepti- cancer (Hirvonen et aL, 1993). bility biomarkers in cancer causation. Some such An example of an inverse dose effect is seen for markers may have little relevance to carcinogenic CYP1A1 ILE to VAL polymorphism in exon 7 of the exposures, while others may be involved in com- gene, as a GRF for smoking-induced lung cancer. plex associations that are difficult to unravel. We The data from Nakachi et al. (1991) are shown in have put forward the argument here that type 2 Tables 3 and 4. Here, although the OR for cancer GEl, for which the genotype alone, in the absence increases for both genotypes as a function of dose, of a relevant carcinogenic exposure, plays no role the ratio between the risks for the two genotypes in carcinogenesis, is the most important type of decreases at higher doses. The decrease of а with GEI for those genes that currently make up the increasing exposures illustrates this. We have also major category of cancer susceptibility biomarkers. observed an inverse dose effect for the association We have stressed the importance of exposure dose of the African-American-specific polymorphism in in the analysis of GEI for these genes, and discussed CYP1A1 with lung adenocarcinoma in smokers two different forms of dose effect. We have shown (Taioh etal., 1995а), aid other groups have reported how these forms, the direct dose effect and the in- similar findings using adducts as an end-point for verse dose effect, may be analysed and distinguished NАT2 and for С YРIА2 (Vineis etal., 1994; Landi et from each other and from other types of effects al., personal communication). (such as protection) by using regression analysts. In some cases, this analytical approach may be We have suggested that the apparent divergence used to determine whether a type 2 GEI is in fact between these two types of dose effect maybe due the correct model for a particular case study. This to the mechanistic position of the GRF on a sat- may be done by determining if the odds ratio for urable dose—response curve for cancer induction. the unexposed group with the GRF is close to 1. In We have not addressed certain issues related to the first example used for illustration (for GsT11), regression analysis of interaction. In general, ques- this was the case. Unfortunately, however, most tions of statistical inference using this analysis studies of this type use very small numbers of sub- have not been addressed here, but of course such jects, and the data may be of insufficient statistical questions are important in testing hypotheses and power to allow for an accurate determination of require careful investigation. The question of this odds ratio. This is especially true for studies of whether multiplicative versus additive models of lung cancer and smoking, where it is often difficult interaction should be used (Brown & Chu, 1959) to find sufficient numbers of non-exposed (non- has also not been addressed here. Although a mul- smokers) cases of both genotypes to have adequate tiplicative model was used to obtain the values power. In our second example (CYP1A1) there were shown in Table 4, we have found (not shown) that only two non-smokers among the cases? We there- the use of an additive model has an effect only on fore used hypothetical data, assuming a true type the magnitude and standard errors of the coeffi- 2 interaction with an odds ratio close to 1 for the cients and а-values, but not on the direction of the G = 1, E = 0 cell. Given the difficulty in proving a dose-dependent effect of the gene. type 2 interaction by the use of actual data] one In conclusion, we believe that studies investigat- alternative is to apply mechanistic knowledge ing the possibility of a metabolic gene polymor- regarding the mechanism of action of the gene phism as a GRF should always include as much

260 Gene-environment interactions in the application of biomarkers information as possible on relevant exposure levels. Cosma, G. Crofts, F., Currie, D. Toriiolo, P. & Garte, S.J. (1993a) The relationship between genotype and function Even if quantitative measures are unobtainable, of the human CYP lAi gene. J. Tas. Environ. Health, 40, questionnaire data giving approximate levels (such 315-322 as never, low or high) of exposure information can be of use in the analysis of the dose effect. Cosma, G., Crofts, F., Cunie, D., Wirgin, 1., Toniolo, E & Carte, S.J. (1993b) Racial differences in restriction frag- Comparison of a-values between studies and meta- ment length polymorphisms and messenger RNA analysis of a for specific gene-exposure combina- inducibility of the human CYP1A1 gene. Cancer tions may also prove valuable in the future. EpidemioI. Biomarkers Prey., 2, 53-57 It has been pointed out by several authors in the Crofts, F., Cosma, G.N., Taioli, E., Currie, D.C., Toniolo, field of biomarkers in epidemiology that the detec- P.Т. & Garte, S.J. (1993) A novel CYP1A1 gene роlymоr tion of cancer genetic susceptibility of this type has phism in African-Americans. Carcinogenesis, 14, profound positive public health implications for 1729-1731 cancer prevention. Detailed study of the interac- Crofts, F., Taioli, E., Trachman, J. Cosma, G.N., Currie, D., tions of these genes with environmental carcino- Toniolo, P. & Garte, 5.J. (1994) Functional significance of gens promises to allow epidemiologists to consider different human CYP1A1 genotypes. Carcinogenesis, 15, humans as collections of individuals of varying 2961-2963 sensitivity and responsiveness. While the entire Daly, A.K., Cholerton, S., Armstrong, M. & Idle, J.R. issue of genetic susceptibility differences among (1994) Genotyping for polymorphisms in xenobiotic people has important ethical, legal and political metabolism as a predictor of disease susceptibility. issues, we believe that increased knowledge in this Environ. Health Perspect., 102 (Suрp1. 9P), 55-61 area (such as the specific form of dose effect dis- Doll, R. & Peto, R. (1981) The causes of cancer: quantita- cussed here) will provide benefits in helping to tive estimates of avoidable risks of cancer in the United resolve these issues. States today. J. Nati Cancer Inst., 66, 1191-1308 Prebourg, T. & Friend, 5.1. (1992) Cancer risks from References germline p53 mutations. J. Clin. Invest., 90, 1637-1641 Alexandrie, A.K., Sundberg, 1.1., Seidegard, J., Torniing, G. & Rannug, A. (1994) Genetic susceptibility to lung cancer Friend, S.Н., Dyja, T.P. & Weinberg, R.A. (1988) with special emphasis on CYP1A1 and GST11: a study Oncogenes and tumor-suppressing genes. N. Eng[. J. on host factors in relation to age at onset, gender and Med., 318, 618-622 histological cancer types. Carcinogene.sis, 15, 1785-1790 Garte, 5.J. (1992) Environmental carcinogenesis, In: Environmental and Occupational Medicine. Brennan, Р. & Silman, A.J. (1994) An investigation of Rom, W, cd., gene-envixornnent interacdori iii the etiology of rheuma- Boston, Little Brown, pp. 105-124 toid arthritis. Am. J. EpidernioL, 140, 453-460 Garte, S.J. & Kneip, T.J. (1988) Metabolism. In: Kneip, T.J. Brown, C.C. & Chu, K.C. (1989) Additive aid multi- & Crable, J.V., eds, Methods for Biological Monitoring. plicative models and multistage carcinogenesis theory. Washington, American Public Health Association, pр.15-26 Risk Anal., 9, 99-105 Hаlдапе, J.B.S. (1946) The interaction of nature and nur- Caporaso, N. (1996) Genetic susceptibility and the com- ture. Ann. Eugenics, 13, 197--205 mon cancers. Biomarkers, 1, 174-177 Harris, C.C. (1989) Interindividual variation among Catteau, A., Bechtel, Y.C., Poisson, N., Bechtel, P.R. & humans in carcinogen metabolism, DNA adduct forma- Bonaiti-Pe11ie, C. (1995) A population and family study tion aid DNA repair. Carcinogen esis, 10, 1563-1566 of CYP1A2 using caffeine urinary metabolites. Eue. J. Clin. Hayashi, S.-1., Watanabe, J., Nakachi, K. & Kawajiri, K. PharmacoL, 47. 423-430 (1991) Genetic linkage of lung-cancer-associated Msp1 polymorphims with amino acid replacement in the Coney, А.H., Miller, E.G. & Miller, J.А. (1956) The metabolism of methylated aminoazo dyes. V. Evidence here binding region of the human cytochrome 110, 407-411 for inducliоп of enzyme synthesis in the rat by 3-methyl- P4501А1 gene. J. Biochem., cholanthrene. Cancer Res., 16, 450-460 Hayden, Ivf.R, Liu, 1.5 & Ma, Y. (1994) Gene environ- Conway, K., Edmiston, 5., Fried, D.B., Hulka, B.S., ment interaction and plasma triglyceride levels: the cru- Garrett, P.A. & Liu, E. (1995) Ha-ras rare alleles in breast cial role of lipoprotein lipase. Clin. Genet., 46, 15-18 cancer susceptibility. Breast CancerRes. Treat., 35, 97-104 Hegele, R.А. (1992) Gene-environment interactions in

261 Application of Biomarkers in Cancer Epidemiology

atherosclerosis. MoL Cell. Biochem., 113, 177-186 Hadidi, H. (1992) The pharmacogenetics of chemical car- Higginsorn, J. (1980) Importance of environmental and cinogenesis. Pharacogenetics, 2, 246-258 occupational factors in cancer. J. Todcol. Environ. Health, Kato, S., Shields, Р.G., Caporaso, N.Е., Sugimura, H., 6, 941-952 Trivers, G.E., Tucker, М.A., Trump, B.F., Weston, A. & Hirayama, T. (1989) Family history in childhood malignan- Harris, C.C. (1994) Analysis of cytochrome Р4502Е1 cies with special reference to genetic-environmental genetic polymorphisms in relation to human lung cancer. Cancer Epidemiol. Biomarkers Prey. 3, 515-518 interactions. In: Lynch, Н.T. & Hirayama, T., eds, Genetic Epidemiology of Cancer. Боса Raton, FL, CRC Press, pp. Kawajiri, K., Nakachi, K., Imai, K., Yoshi, A., Shinoda, N. 1111-1117 & Watanabe, J. (1990) Identification of genetically high Hirvonen, A. (1995) Genetic factors in individual risk individuals to lung cancer by DNA polymorphisms responses to environmental exposures. J. Occup. Environ' of the cytochrome р4501А1 gene. FЕBS Lett., 263, Med. 37, 37-43 131-133 Khoury, M.J., Adams, M.J. & Flanders, W.D. (1988) An Iirvonen, A., Pursiarien, K.H., Катjаlаiпеn, A., Anttila, S. & Vainio, H. (1992) Point mutational Msp1 and I1e-Val epidemiologic approach to ecogenetics. Am. J. Hum. polymorphism closely linked in the CYP1AI gene: lack Genet., 42, 89-95 of association with susceptibility to lung cancer in a Khoury, М.J., Beaty, Т.H. & Cohen, B.Н. (1993) Finnish study popuiation. Cancer Epidemiol. Biomarkers Fundamentals of Genetic Epidemiology. New York, Oxford Prev., 1, 485-489 University Press

Hirvonen, A., Husgafvel-Pursianen, K., Anttila, S. & Khoury, М.J. & James, L.M. (1993) Population and famil- Vainio, H. (1993) The GST11 null genotype as a potetial ial relative risks of disease associated with environmen- risk midifier for squamous cell carcinoma of the lung. tal factors in the presence of gene-environment interac- Carcinogenesis, 14, 1479-1481 tion. Am, J. Epidemiol., 137, 1241-1250 Holstein, М.C., Wild, C.P., Bleicher, F., Chutimataewin, Khoury, M.J., Beaty, T.H. & Hwang, S.J. (1995) Detection S., Harris, C.C., Srivafanalo!, P. & Montesano, R. (1993) of genotype-environment interaction in case-control p53 mutations and aflatoxin B.1 exposure in hepatocel- studies of birth defects: how big a sample size? Teratology, lular carcinoma patients from Thailand. fiL J. Cancer, 53, 51, 336 51-55 Kfioury, M.J. & Wagener, D.К. (1995) Epidemiological Hulka, B.S. (1991) Epidemiological studies using biolog- evaluation of the use of genetics to improve the predic- ical markers: issues for epidemiologists. Cancer Epidemiol. tive value of disease risk factors. Am. J. Hum. Genet., 56, Biomarkers Prey., 1, 13-19 835-844 I-luit, S.C., Hasstedt, S.J., Wu, L.L. & Williams, R.R. (1993) Kihara, M., Kihara, M. & Noda, K. (1994) Lung cancer A gene-environment interaction between inferred kallikrein risk of GSTI1 null genotype is dependent on the extent genotype and potassium. Hypertension, 22, 161-168 of tobacco smoke exposure. Carcinogenesis, 15, 415--418 Hwang, S.J., Bеаty, T.H., Liang, K.Y., Coresh, J. & Khoury, Landi, M.T., Bertazzi, PA., Shields, P.G., Clark, G., Lucier, M.J. (1994) Minimum sample size estimation to detect G.W., Garte, S.J., Cosma, G. & Caporaso, N.E. (1994) gene-environment interaction in case-control designs. Restriction fragment length polymorphism, mRNA Am. f. EpidemioL, 140, 1029-1037 expression and enzymatic product of CYP1A1 gene in humans. Phannacogenetics, 4, 242-246 Hwang, S.J., Beaty, Т.H., Panny, S.R., Street, N.A., Joseph, J.M., Gordon, 5., McIntosh, I. & Francomano, C.A. Lane A. Cruickshank J.K. Mitchell J. Henderson A. (1995) Association study of transforming growth factor Humphries S. Green, E (1992) Genetic aid environmen- alpha (TGF alpha) TagI polymorphism and oral clefts: tal determinants of factor VII coagulant activity in eth- indication of gene-environment interaction in a popuhi- nic groups at differing risk of coronary heart disease. tim-based sample of infants with birth defects. Am. J. Atherosclerosis, 94, 43-50 Epidemiol., 141, 629-636 Lehmann, A.R., Hoeijmakers, J.H., van Zeeland, A.А., Ichiba, M., Hagmar, L., Rannug, A., Hogstedt, B., Backendorf, C.M., Bridges, B.A., Collins, A., Fuchs, R.P., Alexandrie, A.K., Carstensen, U. & Hemniinki, K. (1994) Margison, G.P, Montesano, R. & Moustacchi, E. (1992) Aromatic DNA adducts, micronuclei and genetic poly- Workshop on DNA repair. Mutation Res. 273, 1-28 morphism for CYP1AI and GST1 in chimney sweeps. London, S.J., Daly, A.K., Fairbrother, K.S., Holmes, C., Carcinogenesis, 15, 1347-1352 Carpenter, C,L., Navidi, W.C. & Idle, J.R. (1995) Lung Idle, J.R., Armstrong, M., Biddy, A.V., Boustead, C., cancer risk iп African-Americans in relation to a race-spe- Cholerton, S., Cooper, J., Daly, A.K., Ellis, J., Gregory, W., cific CYP1A1 polymorphism. Cancer Res., 55, 6035-6037

262 Gene-environment interactions in the application of biomarkers

Marquis, J.K. & Siek, G.C. (1988) Sensitive populations Sellers, T.A., Potte, J.D., Bailey-Wilson, J.E., Rich, 5.5., and risk assessment in environmental policy-making. In: Rothschild, H. & Elston, R.C. (1992) Lung cancer detec- 5ахепа, J., ed., Hazard assessment of chemicals (Vol. 6). tioл and prevention: evidence for an interaction Washington, DC, Hemisphere Publishing between smoking and genetic predisposition. Cancer Res., 52, 26945-26975 Meyer (1994) Polymorphisms of human aсetyltrans- ferases. Environ. Health Penspect., 162 (Suppl. бР), 213-216 Shields, PG., Caporaso, N.E., Falk, R:1'., Sugimura, H., Trivers, G.E., Tiump, B.F., Hoover, R.N., Weston, A. & Imai, K., Hayashi, . Watanabe, J. & Nakachi, К., Ѕ Harris, C.C. (1993) Lung cancer, race, and a CYP1A1 Kawajiri, K. (1991) Genetic susceptibility to squamous genetic polymorphism. Cancer EpidemioI. Biomra kers cell carcinoma of the lung in relation to cigarette smok- т Prey., 2, 481-85 ing dose. Cancer Ret., 51, 5177-5181 Sinha, R., Rothman, N., Brown, E.D., Mark, S.D., Hoover, Imai, K., Hayashi, S. & Kawajiri, K. (1993) Nakachi, K., R.N., Caporaso, N.E., Levander, O.A., Knize, M.G., Lang, Polymorphisms of the CYP1A1 and glutathione S-trans- N.P. & Kadlubar, F.F. (1994) Pan-fried meat containing ferase genes associated with susceptibility to lung cancer high levels of heterocyclic aromatic amines but low levin relation to cigarette dose in a Japanese population. els of polycyclic aromatic hydrocarbons induces Cancer Res., 53, 2994-2999 cytochrome P.4501А2 activity in humans. Cancer Res., Nakajima, T., Elovaara, E., Anttila, S., Hirvonen, A., 54, 6154-6159 Camus, A.M., Hayes, J.D., Ketterer, B. & Vainio, H. (1995) Smith, M.L., Chen, I.T., Zhan, Q., Bae, I., Chen, .Y., Expression and polymorphism of g1utathione С Curer, T.M., Kastan, M.B., O'Connor, P.M. & Fornace S-transferase in human lungs: risk factors in smoking- A.J., Jr (1994a) Interaction of the p53-regulated protein related lung cancer. Carcinogenesis, 16, 707-711 Gadd45 with proliferating cell nuclear antigen. Science, Ottman, R. (1990) An epidemiologic approach to gene- 266, 1376-1380 environment interaction. Genet. EpidemioL, 7, 177-185 Smith, C.M., Kelsey, K.T., Wiencke, J.К., Leyden, K., Ottman, R. (1994) Epidemiologic analysis of gene-envi- Levin, S. & Christiani, D.C. (1994b) Inherited g1u- roпment interaction lii twins. Genet. Epidemiol., 11, 75-86 tathione-S-transferase defiency as risk factor for pul- Daman, R. (1995) Gene-environment interaction and monary asbestosis. Cancer Epidemiol. Biornarkers Prey., 3, public health [editorial; comment). Am. J. Hum. Genet., 471-477 56, 821-823 Stephens, E.A. Taylor,J.A., Kaplan, N., Yang, Cu., Hsieh, Perera, E, Mayer, J., Santella, R.M., Brenner, D., Jeffrey, L.L., Lucier, G.W. & Bell, D.A. (1994) Ethnic variation in A., Latriano, L., Smith, S., Warburton, D., Young, T.L., the CYPZEI gene: polymorphism analysis of 695 African- Tsai, W.Y., Hemminki, K. & Brandt-Rauf, P. (1991) Americans, European-Americans and Taiwanese. Biologic markers in risk assessment for environmental Pharmacogenetics, 4, 185-192 carcinogens. Environ. Health Perspect., 90, 247-254 Sugimura, H., Suzuki, I., Hamada, G.S., Iwase, T., Petersen, D.D., IcKinney, C.E., Lkeya, K., Smith, H.H., Takahashi, T., Nagura, K., Iwata, H., Watanabe, S., Kino, I. & Tsugane, S. (1994) Cytochrome 4501 1 genotype Bаlе, А.E., McBride, O.W. & D. W. Nebert. (1991) Human Р А CYP1A1 gene: cosegregation of the enzyme inducibility in lung cancer patients and controls in Rio de JaneiIo 3, 145-148 phenotype and an RFLP. Am. J. Hum. Genet., 48, 720-725 Brazil. Cancer ЕргдетiоI. Biomarkers Pieu, Raunio, H., Husgafvel-Pursiainen, K., Anttila, S., Taioli, E. Crofts, F. Demopoulos, R., Trachman, J. Toniolo, P. & Garte, S.J. (1995 ) An Atrican American specific Hietanen, E., Hirvonen, А. & Pelkonen, O. (1995) а Diagnosis of polymorphisms i carcinogen-activating CYPIA1 polymorphism is associated with adenocarc[- п noma of the lung. Cancer Res., 55, 472-473 and inactivating enzymes and cancer susceptibility а review. Gene, 159, 113-121 Taioli, E. Crofts, F., Trachman, J., Bayo, S., Bradlow, L., Rothman, N. (1995) Genetic susceptibility biomarkers in Tonlolo, P. & Garte, S.J. (1995b) Racial differences in studies of occupational and environmental cancer: CYP1a1 genotype and function. Toxicol. Lett., 77, 357-362 methodologic issues. Toxico!. Left. 77, 221-225 Samson, L. (1992) The suicidal DNA repair methyltrans- Taioli, E., Trachman, J., Chep, X., Toniolo, P. & Garte, S.J. ) CYPIA1 RFLP is associated with breast cancer in ferases of microbes. MoL Micro bio! ., 6, 825-831 (1995с African American women. Cancer Res., 55, 3757-3758 Savitsky K. Bar-Shira, A., Glad, S., Rotman, G., Ziv, Y., Vanagaite, L., Tagle, D.A., Smith, S., Uziel, T., & Sfez, S. Tefre, T., Ryberg, D., Haugen, A., Nebert, D.W., Skaug, V., (1995) A single ataxia telangiectasia gene with a product Brogger, A. & lorresen, A.L. (1991) Human CYPIA1 gene: lack of association between the 'spi restriction fragment similar to РI-3kiпase. Science, 268, 1749-1753

263 Application of Biomarkers in Cancer Epidemiology length polymorphism and incidence of lung cancer in a Watanabe, J., Yang, J.P., Eguchi, H., Hayashi, S., 1mai, K., Norwegian population. Phaгнщcоgeпehсs, 1, 20-27 Nakachi, K. & Kawajiri, K. (1995) An Rsa I polymorphism Tiret, L., Abel, L. & Rakotovao, R. (1993) Effect of ignor- in the С YР2Е1 gene does not affect lung cancer risk in а ing genotype-environment interaction on segregation Japanese population. Jpn J. Cancer Res., 86, 245-248 analysis of quantitative traits. Genet. EpiderwioI, 10, Yu, M.C., Skipper, P.L., Taghizadeh, K., Tannenbaum, 581-586 S.R., Chan, K.K., Hendeson, B.Е. & Ross, R.К. (1994) Acetylator phenotype, aminobiphenyl-hemoglobin Vahakangas, K.LL, 3атеt, J.M., Metcalf, R.A., Welsh, J.A., Bennett, W.P., Lane, D.P. & Harris, C.C. (1992) Mutations adduct Ievels, and bladder cancer risk in white, black, of p53 and ras genes in radon-associated lung cancer and Asian men in Los Angeles California. J. Nad Cancer from uranium miners. Lancet, 339, 576-580 Inst., 86, 712-716 Vineis, P. (1995) Use of biomarkers in epidemiology. The example of metabolic susceptibility to cancer. ToxicoI. Left., 77, 163-168 Corresponding author Vineis, P., Bartsch, H., Caporaso, N., Harrington, A., S. Garte Kadlubar, F.F., Landi, M.T., Malaveille, C., S1цe[ds, Л.G., Nelson Institute of Environmental Medicine and Skipper, P., Talaska, G. & Tannenbaum, S.R. (1994) Kaplan Cancer Center, Genetically based N-acetyltransferase metabolic puy- New York University Medical Center, morphism and low level environmental exposure to car- New York, NY, USA cinogens. Nature, 369, 154-156

264 Application il olumarkere in Cancer Epidemiology Torticic, P., Boffelta, P., 5huker, D.E.G., Rothman, N., Hulka, Band Pearce, N., edt 1AАC Scientific PIjblicationo No. 142 lnternaticnal Agency for Research on Cancer, Lyon, 1997

Using and interpreting surrogate end-points in cancer research

A. Schatzkin, LS. Freedman, J. Dorgan, L' McShane, М.Н. Schiffman and S.M. Dawsey

Researchers have proposed a broad range of molecular, cellular and histological markers as surrogate end-points for cancer (sECs).The effect of an intervention on a `valid' SЕС is concordant with its effect on cancer incidence. The validity of a potential SEC is determined primarily by the extent to which the marker is a necessary event on the causal pathway to cancer' Colorectal adenomatous polyp formation is an example of a reasonably valid SEC because these lesions are obligate precursors of most large bowel malignancies. However, the existence of a plausible major alternative causal pathway—one bypassing the potential SEC—weakens inferences from that marker to cancer. Moreover, unless the pathway to cancer operates nearly exclusively through the SEC, an SEC that is valid for one intervention or exposure may not be valid for another. Metabolic, ecological, observational epidemiological and 'intervention studies may yield data that are useful in revealing these causal interrelations of intervention (exposure), SЕС and cancer. Empirical studies of three questions are pertinent: (1) What is the relation of the SEC to cancer? (2) What is the relation of the intervention (exposure) to the SEC? (3) To what extent does the SЕС mediate the relation between the intervention (exposure) and cancer? Data on SEC measurement error are important in ascertaining the extent to which marker results have been attenuated by such error. It is essential to carry out these studies to evaluate potential SECS (such as epithelial cell hyperproliferation) with plausible major alternative pathways to cancer. At the present time, definitive evidence on etiology and prevention will emerge only from studies with cancer end-points or SECs that are, by and large, necessary steps on the causai pathway to malignant disease.

Because the diagnosis of cancer is a relatively rare association of an exposure with the SEC is concor- event, clinical trials or observational epiderniolog- dant with its association with cancer incidence. If, ical studies with incident cancer end-points have for example, a large change in the SEC means a to be very large, lengthy and expensive. Studies large change in cancer incidence, then a small using surrogate end-points of malignant disease change in the SEC would mean a small change in can be smaller, shorter and cheaper than studies cancer incidence. If the SEC meets these condi- with incident cancer end-points. It is not surpris- tions, it can be considered a 'valid' surrogate for ing, then, that cancer researchers have long been that cancer. These conditions follow from the cri- interested in using these markers. teria proposed by Prentice (1989).

Definition of a surrogate end-point marker for Evaluating SAC validity: logical considerations cancer The validity of a potential 5ЕС depends on the We define a surrogate end-point marker for cancer extent to which it is a necessary step in carcino- (SEC) as follows: a surrogate for incident cancer genesis. The simplest causal pathway involving a yields a valid test of the null hypothesis of no as- potential SEС is shown inn Fig. 1. E1 represents an sociation between treatment and incident cancer. environmental or host factor. A change in E1 In other words, the effect of an intervention on the would alter SEC positivity and thereby modify the SEC is concordant with its effect on cancer incidence, incidence of cancer. The SEC is a valid surrogate or, for observational epidemiological studies, the for cancer.

265 Appliсatron of Biomarkers in Cancer Epidemiology

ber E1 SEC. CA of bad adenomas as a necessary consequence of п reducing the number of innocent adenomas. The intervention therefore lowers the incidence of Figure 1. Pathway to cancer (CA) with single exposure (El) colorectal cancer. and single marker (SEC). The existence of the flat dysplasia pathway complicates things. Our intervention has no effect Figure 2 depicts a more realistic picture of car- on pathway (b), even though it reduces observed cinogenesis. In this figure, E1 influences cancer adenomas via pathway (a). Tithe extent that path- through two alternative pathways, one through way (b) contributes to colorectal carcinogenesis, the potential SЕС, the other through Marker 2. adenoma development (as detected through (We assume, to simplify the discussion, that only colonoscopy) may not be a valid SEC for cancer. As one intermediate marker, SEC, resides on the pathway (a) becomes the less common of the two E1-SЕGcancer pathway. To the extent that E1 routes to cancer, an investigator observing fewer works through the alternative Marker 2 pathway, adenomas developing among intervention partic- we cannot be certain that SEC is a valid surrogate ipants could conclude that the intervention for studies involving E1. This is because E1 may reduces colorectal cancer incidence, when in fact affect Marker 2 in a way that offsets its influence the intervention might have a quite different effect on SEC; the final effect on cancer 1s unknown. If, on cancer occurrence through pathway (b). A large for example) E1 reduces SEC positivity but body of evidence, however, suggests that most col- increases Marker 2 positivity, E1 could increase orectal cancers do develop through pathway (a), cancer incidence. the adenoma--carcinoma sequence (Muto et al., 1975). Therefore, an intervention reducing adeno- Potential SECs illustrating these logical matous poiyp recurrence would be likely to reduce considerations colorectal cancer incidence. Adenoma recurrence Consider adenomatous polyps, an increasingly is a reasonably valid SEC. popular surrogate for colorectal cancer. Iп Fig. 3 HPV infection in cervical cancer appears to (pathway a), an event X is necessary for an adeno- be analogous to the adenoma colorectal cancer matous polyp to progress to colorectal cancer. example. The overwhelmingly large proportion Therefore, two types of polyiloid adenomas exist— of cervical cancer requires prior HPV infection those without X (innocent adenomas not pro- (Schiffman, 1992). The relatively rare X-event is gressing to cancer) and those with X (bad' adeno- whatever (still unknown) immunological deficit rias progressing to cancer). Both types are observ- leads to persistent HPV infection. HPV persistence able but indistinguishable through a colonoscope. results in inactivation by the E6 and E7 proteins of Furthermore, there exist flat areas of dysplasia with the HPV genome, of р53 aid pBh tumour suppres- X (not observable through the colonoscope) that sor gees, leading in turn to increasingly severe also progress to cancer (Н0, 1991). intra-epithelial neoplasia and eventually cancer. Suppose we have an intervention (a low-fat It is currently thought, however, that a small pro- eating plan) for example) which reduces E1 (some portion of cervical cancer can arise as a result faecal constituent) and thereby diminishes the of tumour suppressor gene product inactivation pool of adenomas susceptible to the relatively rare occurring, by mutation, in the absence of HPV X-events. This intervention thus reduces the пиm- infection, Because most cervical cancer does occur through HPV infection, an intervention that elim- iпates or reduces HPV infection would probably decrease cervical cancer incidence. Е1 SEC GA Inferences to cancer from other potential SECs, Marker 2 however, are considerably more problematic. Figure 4 depicts plausible causal pathways involving colorectal epithelial сеll proliferation (Baron Figure 2. single exposure (El) with pathways through two et cl., 1995) (the adenoma step depicted in Fig. 4 is alternative markers (SEC and Marker 2) to cancer (CA). simplified here). E1 is again an exposure amenable

266 surrogate end-points iri cancer research

(a)

Normal mucosa +E1 -- Polypoid non-X adenoma (+X) Polypoid X-adenoma . CA

(b) Normal mucosa +X Fiat X-dysplasia CA

Figure 3. Alternative pathways, with (a) and without (b) an adanomatous polyp step, to colorectal cancer (CA). E1 is exposure; X +s an event necessary for development of cancer. to modification. Tithe extent that other celular or Exposure dependence molecular events, such as diminished apoptosis or tri Fig. 5 we return to the simple, idealized scheme altered cellular adhesion factors, constitute an from Fig. 1, but now add another exposure, E2. Both important causal pathway from E1 to cancer, we E1 and E2 in Fig. 5 work through SEC on the path cannot be sure that El's relation to these other to cancer. Because SEC is a necessary precursor for events does not offset El's effect on cancer through cancer, the validity of this SEC is exposure-inde- proliferation. Cell proliferation is a problematic pendent, i.e. any other exposure, Е2, that influences SEC because the relative importance of the alter- cancer must operate through the SAC; the SAC is native pathways (through events other than pro- valid for studies of E2 as well as those of E1. liferation) is simply unknown. Figure 6 adds E2 to the more complex and real- Another Iogical consideration is as follows: a istic pathway depicted in Fig. 2. In Fig. 6, the exis- marker may not be directly on the causal pathway tence of a non-trivial alternative pathway (through to cancei but may be closely linked to a compo- Marker 2) means that the validity of the SEC is nent of that causal pathway such that it does make exposure-dependent. Even if E1 affects Marker 2 a reasonable SEC. One possible example of this are minimally (suggesting that SEC is reasonably valid micronuclei, which have been detected in epithe- for E1—cancer studies), we cannot assume that the lia1 cells from oral, oesophageal, bronchial and E2—Markеr 2—caпceт pathway also plays a minor large intestinal tissue (Garewal et al., 1993). Many role in the development of cancer. micronuc1eated cells are non-viable and therefore For example, a given agent might influence cannot be direct cellular precursors of a malignant colorectal carcinogenesis largely through its influ- tumor. The overall prevalence of micronucleated ence on cell proliferation (Fig. 4). In that case, cell cells, though, might strongly reflect microstruc- proliferation is a reasonably valid SEC for the first tural alterations in other cells that do eventually agent vis-à-vis colorectal cancer. A second agent, undergo malignant transformation and clonal however, might not affect cell proliferation but expansion. might increase apoptosis sufficiently to decrease

Normal mucosa+E1 Hyperprobteration Адепота and cancer

л Other events ' (e.g. apoptosis, celular adhesion)

Figure 4. Aiternative pathways, with and without mucosal hyperproliferation, to colorectal cancer. E1 is exposure.

267 Application of Biomarkers in Cancer Epidemiology

TabIe 1.2 к 2 tаые sr between SEС

Figure 5. Each of two exposures (El and Е2) leads through two markers (SЕС and Marker 2) to cancer (CA). cancer incidence (Bedi et ai., 1995). Focusing only on cell proliferation would give a falsely pes- simistic impression of the second agent's efficacy SEGcaпcer in inhibiting colorectal carcinogenesis. Observational epidemiological studies are important vehicles for examining this SEC-саncer ques- Investigating causal pathways involving SECs tion. In a recent case-control study, Schiffinan et al. We have argued that the causal structure underly- (1993) showed a markedly increased risk of severe ing the relationships between exposures, potential cervical neoplasia for those with HPV infection. SECS and cancer is critical in evaluating SECS. Data Tortiulo et al. (1995), in a case-control study nested that are helpful in revealing this structure can within a prospective cohort, observed a direct ida- emerge from investigations into three questions tionship between serum estrogens and breast can- (Schatzkin et al., 1990) (1) is the SEC associated cer. Observational cohort studies may also be nested with cancer (in particular, how large is the attrib- in trials. In the Polyp Prevention Trial (Sthatzkzn et utable proportion)? (2) is the intervention/expo- аi., 1996), for example, it wil be possible to relate sure associated with the SEC? and (3) does the SEC baseline proliferation indices to subsequent аде- mediate the relationship between the interven- пота recurrence. (We have referred here to studies tion/exposure and cancer? with rieop1ashc cancer precursor end-points, such Traditional epidemiological parameters are use- as cervical intra-epithelial poplasia (CIN) and аде- ful in carrying out these investigations. For simplicity, nomas. For purposes of discussion, we consider we refer in the following discussion to SЕCs that these here as proxies for cancer, although, as we are either positive or negative. However, the argu- have shown, the validity of these precursor end- ments offered here may be extended to encompass points is not absolute.) markers measured as continuous variables. Ecological studies may provide pertinent (if indi- Relative risk, a measure of cancer risk in relation rect) information, on the SEC-cancer question. to SEC positivity, is defined as [a!(a + b)] 1 [с1(с + d)] Researchers have examined, for example, mean pro- (Table 1). A relative risk of 1.0 indicates no associa- liferative indices in groups at different risks of соi- tion between SEC and cancer' The attributable pro- orectal cancer (Lipkin et al., 1984). In ecological portion (AP) represents the proportion of cancer studies, as opposed to observational studies with that is attributable to marker positivity: both marker and disease information on individuals, AР = S(1- 1/R), where R = relative risk and S = sen- the link between marker and disease is indirect; one sitivity, defined as a1(a + c). An AР of 1.0 means cannot be certain that those who are marker-posi- that marker positivity is necessary for the develop- tive are the ones with increased incidence of cancer. ment of cancer, i.e. the carcinogenic pathway must The AР is of great value here in evaluating the go through this positive marker. importance of alternative pathways. In idealized Fig. 1, the AР for SEC is 1.0. In the more realistic Fig. 2, however, with at least two pathways to cancer, E1 SEC CA AР < 1.0. If AР for SEC is high, however, even if it is E2 Marker 2 less than 1.0, it suggests that the alternative Marker 2 pathway plays a small role in the development of cancer. An substantially lower than 1.0 suggests Figure 6. Each of two exposures (El and Е2) can act through АР a number of pathways to affect markers (SEC and Marker 2) that one or more alternative pathways is indeed and lead to cancer (С A). operative.

268 surrogate eid-points in cancer research lпtеrvemiоп/expasure—SEC In a recent case—control study, for example, For a given SEC to be valid with respect to a Schiffman et al. (1994) examined the extent to given intervention, we need to demonstrate that which HPV infection mediated the relation the intervention results in а change in the SEC, or, between number of sекuаl partners and cervical in an observational setting, that the exposure of dysplasia. As Table 2 indicates, there was а strong interest is associated with marker positivity. direct relation between number of sexual partners We can address this question in small clinical and risk of cervical dysplasia. When the relation (metabolic) studies with the putative SEC as the between number of sexual partners and cervical end-point. Examples include studies of fat dysplasia was adjusted for the presence or absence (Prentice et al., 1990) от alcohol (Refchman et al., of HPV infection, the 11E for number of sexual part- 1993) consumption in relation to serum hormone ners dropped dramatically, suggesting that most of levels. We can also examine this question in a the relation between number of partners and dys- case—control or cohort study of, for example, the plasia relation is attributable to HPV infection. relations of reproductive risk factors to HPV infec- One can examine the mediating role of a poten- tion or breast сапсет risk factors to serum estrogen tial SЕС through stratified analyses or standard levels. An ecological study can examine, for exam- multiple regression techniques. In geneial, the ple, the mean proliferative index or degree of larger the intervention effect or exposure relation, epithelial cell DNA hypomethylation in popula- the fewer study participants are needed in a medi- tions with different (average) consumptions of ation analysis. Because exposure relative risks in dietary fat (Lipkin et al., 1985). observational epidemiological studies are often greater than the intervention effects observed in Iпterveпtioп/exposure—SEGcaлceг trials, mediation analyses are more likely to be suc- Suppose we have established that (1) the SEC is cessful in the observational epidemiological setting. causally connected to cancer, but AP < 1.0 and the Genetic mutations as exposures for cancer may route to cancer does not proceed exclusively prove to be a very fruitful source for mediation through the SEC; and (2) the intervention or expo- analyses of biochemical or cellular markers if they sure of interest is linked to the SEC. We would still demonstrate the very high ERs that are currently like to determine the relative importance of the predicted. intervention/exposure—SEC pathway, as opposed Mediation analyses may yield null results, i.e. to pathways operating through other SECS. To do adjusting for a potential SEC may have little iпflu- this, we examine the extent to which the expo- епсе on the relative risks for the intervention or sure/interventions's relation to cancer is mediated exposure. These null findings suggest that the by the sEC, i.e. we address whether SEC status potential SEC does not fuily mediate the relation- accounts for the observed intervention effect or ship between intervention/exposure and cancer. exposure-associated elevation in risk. This involves Even in the face of such null results, however, integrating SEC assays into either observational when there is an alternative pathway from the epidemiological studies or clinical trials. exposure (Ell) to cancer through a second marker,

Number of sexual partners

1 2 3-5 6-9 10+

Unadjusted 1.0 1.7 Э.1a 4.7a 4.4a . Adjusted for HPV Starus 1.0 1.0 1.1 1.5 1.6

a P<0.05

269 Applicalion of Biomагkers in Cancer Epidemiology

the SEC could still reside on the causal pathway to from each of the three types of studies discussed cancer. The degree to which the E1-cancer relation above. The intervention-marker and marker-cancer is attenuated after adjustment for SEC1 will relations will be attenuated by error in marker depend on the (probably unknown) extent to measurement; the marker-adjusted intervention which the E1-cancer relation is mediated by effect will be inflated. Marker 2 as well as SEC1 (Figs 2 and 6). Conclusion Interpreting the data on SECs: statistical Studies with surrogate end-points may give the considerations right answers about the effect of an intervention or All markers are measured with some error. Two sta- the association with an exposure. Positive results tistical caveats follow from this. First a potential from phase 2 studies with surrogate end-points SEC is useful only if it can discriminate among provide additional—but not incontrovertible— study participants, those in an intervention and support for moving on to the larger, more expen- control group or those in various categories of risk sive phase 3 studies with cancer end-points. factor exposure. Such discrimination is practically Merely being on the causal pathway to cancer possible only if the interparticipant variation in does not in itself constitute surrogate validity. It is the SEC values is not swamped by intra-individual the totality of causal connections that is critical. variation. (Intra-individual variation derives, for Only when the causal pathway goes predomi- example, from differences in markers obtained from nantly through that SEC can one reasonably make different tissue areas, measured at different time strong inferences from SEC findings to cancer. This points or read by multiple readers.) Statistically, appears to be the case for adenomas. this means that the intraclass correlation coeffi- When there exist major alternative pathways cient [ICC = interparticipant variationl(interpar- bypassing that SEC, as in cell proliferation and ticipant variation + intTaparticipant variation)] for many other potential SECS, inferences to cancer interparticipant variation (the proportion of all are problematic. This paper is irn part a plea to carry variation attributable to between-participant dif- out the studies—especially SEC-cancer and inter- ferences) is reasonably large (Fleiss, 19%). ventioпlexposuee-eЕС-cancer mediation studies— lntraparticipant variation may be reduced by necessary to evaluate those potential SECS with taking replicate samples (multiple biopsies from plausible major alternative pathways to cancer. different areas, multiple blood draws over time). Such studies are urgently needed if we are to know The reduction in iпtraparticipаnt variability in- how well we can generalize from SEC results to creases the relative contribution of the interpartic- cancer. The irony of the surrogate marker problem ipant variability, and thus the ICC. however, is that the Iarge, long, expensive Data on components of variance for potential studies required to evaluate these problematic SECs are very sparse. Few studies have provided SECS are those the markers were designed to data on SEC variability, particularly with respect to replace. Moreover, SEC evaluation is often inter- time-to-time variation. Notable exceptions are ventionlexposure-dependent: results of validation recent investigations attempting to determine the studies of a marker in relation to one exposure are number of estradiol measurements necessary to not necessarily transfzтгаые to the marker in rela- discriminate reasonably among individuals (Cauley tion to another exposurе/intеrvezuion. Thus, even et al., 1991; Toniolo et al., 1994; Hankinson et al., if we were to find out from some ongoing calcium 1995). Research into rectal mucosal proliferation intervention studies that cell proliferation is a rea- variability is also underway (Lyles et al., 1994). We sonably valid surrogate for colon cancer, we can- emphasize that quality control studies designed to not be certain that proliferation indices are equally capture information on marker variability are valid in aspirin or folate trials. essential if we wish to evaluate and subsequently At present, there appears to be no substitute for use a potential SEC. carrying out large-scale epidemiological studies and Second, and more generally, even if the ratio of clinical trials with cancer end-points or SECs that inter- to intraparticipant variation is acceptable, are, for the most part, necessary steps on the causal measurement error will tend to attenuate findings pathway to cancer.

270 Surrogate end-points in cancer research

References Prentice, R., Thompson, D., Clifford, C., Gorbach, S., Goldin, B. & Byar, D. (1990) Dietary fat reduction and Ваrоn, J.А., Wargovich, М.J., Tosteson, T.D., Sandier, R., Haile, R., Summers, R., van 5to]k, R„ Rothstein, R. & plasma estradiol concentration in healthy pre- Weiss, J. (1995) Epidemiological use of rectal proliferation menopausal women. J. Nat' Cancer Inst., 82, 129-134 measures. Cancer Epidemiol. Biomarkers Prey., 4, 57-61 Reichman, М.E., Judd, J.T., Longcope, C., 5chatzkin, A., Nair, P.P., Campbell, W.5., Clevidence, B. . & Taylor, P.R. Bedi, A. Pasrich, P.J., Akhtar, Al., Barber, J.P., Bedi, G. ., А С (1993) Effects of moderate alcohol consumption on Giardiello, F.M., Zehnbauer, B.А., Hamilton, S.R. & Jones, R]. (1995) Inhibition of apoptoses during development plasma and urinary hormone concentrations in pre- of colorectal cancer. Cancer Res., 55, 1811-1816 menopausal women. J. Nat' Cancer Inst., 85, 722-727 Cauley, J.A., Gutai, J.P., Kuiler, L.H. & Powell, J.G. (1991) Schatzkin, A., Freedman, L.S., Schiffman, M.H. & Reliability and interrelations among serum sex hor- Dawsey, S.М. (1990) The validation of intermediate end- mones in postmenopausal women. Am. J. Epidemiol., points in cancer research. J. Natl Cancer Inst., 82, 133, 50-57 1746-1752 Fleiss J.L. (1986) The Design and Analysis of Clinical Schatzkin, A., Lauza, E., Freedman, L.S., Tangrea, J., Experiments, New York, John Wiley & Sons, pp. 1-5 Cooper, M.R., Marshall, J.R., Murphy, P.A., Selby, J.V., Shike, I., Schade, R.R., Burt, R.W., Kikenda11, W. & Garewal, H.S., Ramsey, L., Kaugars, G. & Boyle, J. (1993) Cahill, J. (1996) The Polyp Prevention Trial; I. rationale, Clinical experience with the micronucleus assay. J. Cell design, recruitment, and baseline participant character- Biochem., 17 (Suppl. F), 206-212 istics. Cancer Epidemiol. Biomarkers Prey., 5, 375-384 Hankinson, S.E., Manson, J.E., Spiegelman, D., Willett, Schiffman, M.H. (1992) Recent progress in defining the W.C., Longcope, C. & speizer, F.E. (1995) Reproducibility epidemiology of human papillomavirus infection and of plasma hormone levels in postmenopausal women cervical neoplasia. J. Natl Cancer Inst., 84, 394-398 over a 2-3-year period. Cancer Epidemiol. Biomarkers Prev., 4, 649-54 Schiffman, М.H., Bauer, H.M., Hoover, R.N., Glass, А.G., Cadell, D.M., Rush, B.B., Scott, D,R., Sherman, М.E., Hi11, M J. (1991) From adenoma to carcinoma of the col- Kurman, R.J., Wacholder, S., Stanton, C.K. & Manos, orectum. Recent Results Cancer Res., 122, 71-84 М.М. (1993) Epidemiologic evidence showing that Lipkin, M., Blattner, W.A., Gardner, E.J., Burt R.W. Lynch, human papillomavirus infection causes most cervical LI., Deschner, E., Winawer, S. & Fraumeni, J.F. (1984) intraepithelial neoplasia. J. Nad Cancer Inst., 85, 958-964 Classification and risk assessment of individuals with Schiffman, M.H. & Schatzkin, A. (1994) Test reliability is familial polyposis, Gardner's syndrome, and familial critically important to molecular epidemiology: an non-polyposis colon cancer from [3Н]thymidine labeling example from studies of human papillomavirus infection patterns in colonic epithelial cells. Cancer Res., 44, and cervical neoplasia. Cancer Res., 54, 1944s-1947s 4201-4207 Toniolo, P., Koenig, K.L., Pasternack, B.S., Banетjee, S., Lipkin, M., Uehara, K., Winawer, S., Sanchez, A., Bauer, Rosenberg, C., Shore, R.E., Strax, P. & Levitz, M. (1994) C., Phillips, R., Lynch, H.T., Blattner, W.A. & Fraumeni, Reliability of measurements of total, protein-bound, and J.F., Jr (1985) Seventh-Day Adventist vegetarians have a unbound estradiol in serum. Cancer EpidemioI. Biomarkers quiescent proliferative activity in colonic mucosa. Cancer Prey., 3, 47-50 Left., 26, 139-144 Toniolo, P.G„ Levitz, M., Zeleniuch-Jacquotte, A., Lyles, С .M., Sandler, R.S., Keku, T.O., Kipper, L.L., Banerjee, S., Koenig, K.L., Shore, R.E., Strax, P. & Millikan, C., Murray, S.C., Bangdiwala, S.I. & Ulshen, Pasternack, B.S. (1995) A prospective study of eridoge- M.H. (1994) Reproducibility and variaibility of the rectal nous estrogens and breast cancer in postmenopausal mucosal proliferation index using proliferating cell women. J. Nail Cancer Inst., 87, 190-197 nuclear antigen immunohistochemistry. Cancer Epiderniol. Biomarkers Prey., 3, 597-605 Corresponding author Muto, T., Bussey, H J.R. & Morson, B.C. (1975) The evolu- A. Schatzkиn lion of cancer of the colon and rectum. Cancer, 36,2251-2270 Division of Cancer Prevention and Control, National . Prentice, R.L. (1989) Surrogate endpoints in clinical trials: Cancer Institute, Bethesda, MD 20892, USA definition and operational criteria. Stat. Med., 8, 431-440

271 Applicaüon ы 6iomarkers in Dancer Epidemiology iolo, P., Bofletta, P_, shcker, DEC., Rothman, N., Hulka, B- and Pearce, N., eda Toп 1ARC 5cientiнc Publicalionu No. 142 lnternalional Agency 1or Research on Cancer, Lyon, 1997

Biomarker end-points in cancer chemoprevention trials

C.W. Boone and G.J, Kelloff

Over the last decade, the Chemoprevention Branch, Division of Cancer Prevention and Control, National Cancer Institute, цSA, has been developing drugs that will slow or stop the progression to invasive cancer of precancerous (pre-invasive) lesions generally termed 'intraepithelial dysplasia' or dysplasia'. Over 40 short-term clinical trials are in progress, testing the following classes of agents on precancerous lesions in the different major organ systems: antimutagens (N-acetylcysteine, oltipraz), antiproliferatives (difluoromethylornithine, dehydroepiandrosterone, selenomethionine), antioxidants (vitamin E, curcumin), anti-inflammatories (aspirin, piroxicam, ibupro#en, sulindac sulfone) and hormonally active agents (tamoxifen in breast ductal carcinoma in situ and finasteride in prostatic intraepithelial neoplasia). Because of the strong practical need to keep so many clinical trials as short-term as possible, certain tissue changes known to be associated with high cancer risk were selected for use as biomarker end-points in the trials, such changes being quantitatively assayed by computer-assisted image analysis.These'surrogate end-point biomarkers' (SEBs) are based on the individual cellular morphological and functional changes universally used by histopathologists to diagnose the lesion of intraepithelial neoplasia (Riddell, 1984; Boone of aL, 1992; Wright of aL, 1994). High grades of this lesion precede invasive cancer in the great majority of cases, and therefore SEBs based on them are linked to high cancer risk.Table 1 summarizes some of the short-term clinical trials now being monitored by the Chemoprevention Branch. The SEBs abbreviated `PPNN' in the figure are: proliferative index (P); ploidy (DNA histogram) (Р); nuclear morphometry and chromatin texture (N); and nucleolar size and frequency (N). Computer-assisted image analysis 1s used to assay these features quantitatively, which gives the SEBs increased objectivity, reproducibility and sensitivity. Further details concerned with cancer chemoprevention trials using SEBs, and their relation to the field of cancer epidemiology, аrе given below.

Cancer chemoprevention in relation to cancer nosis and therapy of cancer, and particularly on epidemiology the conduct of randomized prospective interven- Rathman (J 986) describes clinical trials as a type of tion trials of therapeutic agents. Both areas of can- experimental epidemiology, stating that clinical cer epidemiology focus on cancer prevention, but trials, where the word 'trial' is used as a synonym from different perspectives. The environmental for 'experiment', are epidemiological studies of dif- cancer epidemiologist favours 'primary prevention' ferent treatments for patients who already have and seeks to proscribe elements in the environment the disease, and that 'the exposures in a clinical that are known to be associated with increased trial ... are preventives of the sequelae of the initial cancer risk, e.g. tobacco and alcohol use, exposure disease'. For this presentation, the specialized area to fossil fuel combustion products, and fatty diet. of experimental epidemiology concerned with The clinical cancer epidemiologist favours 'secon- clinical trials will be called 'clinical epidemiology' dary prevention' of already established pre-invasive and contrasted with what will be called traditional neoplastic disease and seeks to prescribe drugs that 'environmental epidemiology'. This contrast is will prevent progression to the stage of invasive- described in Table 2. ness. Although chemoprevention is characterized Whereas environmental cancer epidemiology here as a form of secondary prevention, in that it focuses on the cause and pathogenesis of cancer, prevents the further sequelae (i.e. invasive cancer) of clinical cancer epidemiology focuses on the diag- the 'primary disease' (i.e. neoplasia), it could be

273 Application of Biomarkers in Cancer Epidemiology

. в ' • • в в - • - - в • i - в I - в в ' в • • в в

Organ Chemoprevention Gross lesion SEBs Other SEBs (Microlesion = dysplasia)

Skin 4HPR Actinic keratoses PPNNa PCNA, RAR

Oral 4NPR Leukoplakia PPNN PCNA, RAR, EGFR, ТGF-[l, . involucrin

Breast 4NPR Mammogram: ductal carcinoma in situ PPNN PONA, р53, erbв-2

Lung 4HPR Chronic smoker PPNN PCNA, EGFR, р53, F1SH aneusomy

Bladder DFMO Focus of redness PPNN . ODC, EGFR, Lex

Colon Са2+ Adenomatous роlуp PPNN Ihy, BrdU, extLeк , keratins, intégrins

Prostate Proscar PINb PPNN

Cervix 4HPR Masaicism, punctation PPNN DNA aneuploidy, PCNA, K-ras

flPPNN: proliférative indвx; ploidp; nuclear morphometry; nucleolar morphometry' ЬP1N: prostatic intraepithelial neoplasia. argued that chemoprevention has elements of pd- produced by a carcinogen). The (BR) biomarker mary prevention because it involves arresting the may be considered as an early molecular lesion at neoplastic process before invasive cancer develops. the beginning of the sequence of neoplastic dis- Regardless of this point, it is clear that cancer ease progression [the precedent for calling an chemopreventiori is an activity that belongs to the abnormal molecular change a 'lesion' has been area of clinical cancer epidemiology. established by Ames (1995), who makes a useful distinction between DNA lesions, such as strand The relationship between cancer chemopreыeпtion breaks and carcinogen adducts (BED markers), and and molecular cancer epidemiology DNA mutations (BR markers), which require cell The rapidly expanding field of molecular cancer division for their production from DNA lesions]. epidemiology was developed using the termi- The fifth biomarker in this sequence has been nology of environmental epidemiology (Hu1ka et caцed the subclinical disease marker (subclinical in а1., 1990; Perera et al., 1982). Figure 1 illustrates the sense that the subject with the marker has no how the biomarkers identified along the carcino- signs or symptoms), and is defined by Hulka et al. gen exposure—neoplastic disease sequence were (1990) as `a measurable indicator of a stage of dis- defined: (1) the external (carcinogen) dose ease or a manifestation of the disease itself'. biomarker; (2) the internal (carcinogen) dose bio- Clearly, as shown in Table 2, the subclinical diag- marker; (3) the biologically effective dose nostic disease markers related to intraepithelial neo- (BED) biomarker (e.g. DNA and haemoglobin plasia (dysplasia) which are used as surrogate end- adducts to carcinogen); and (4) the biological point biomarkers (SEBs) in clinical trials of response (BR) biomarker (e.g. stable mutations chemopreventive agents are the same subclinical

274 Biomarker end-points in cancer chemoprevenbon trials disease markers as defined by Hulka et aI. (1990) Validation that reduction iп the extent of They are described more fully below. lntraepithelial neoplasia (dysplasia) is a SEB that predicts a reduction in cancer incidence lntraepithelial(pre-invasive) neoplastic disease: In epidemiological terms, intraepithelial neoplasia the target of cancer chemoprevention is a necessary component on the final common path- Multifocal intraepithelial (pre-invasive) neoplastic way to cancer. The concept is straightforward--if disease, a lesion of which is illustrated in Fig. 2, one were to give a chemopreventive drug that is the target lesion for the programme of the eradicated all of the multifocal lesions of intra- Chemoprevention Branch at the US National epithelial neoplasia (dysplasia), one could expect Cancer Institute. For most organ epithelia, in par- to have eliminated the possibility of the subse- ticular breast (Page etal., 1985), prostate (Bostwick, quent development of cancer (Fig. 2). Thus, it 1992), colon (The Multicentric Study of Colorectal appears reasonable that reduction of the multiple Adenomas Workgroup, 1995), cervix (Chanen, lesions of intraepithelial neoplasia, as determined 1990) and lung (Carter, 1985), multifocal lesions by visual inspection and multiple biopsies, may of the intraepithelial neoplasia arise in the epi- not be difficult to validate as a marker that predicts thelial compartment as monoclonal expansions a reduction in cancer incidence. Although multi- at multiple sites aid progress at different focal intraepithelial neoplasia (dysplasia) is a necessary condition permitting further progression to rates f от many years. Sortie intraepithelial neoplastic lesions remain stable or, in earlier stages, even cancer, it may not be a sufficient one, since the regress. In a small proportion of lesions, however, majority of dysplastic epithelial lesions do not the slowly growing intraepithelial neoplastic cell progress to invasion and may even regress. populations continue to progress to the point of invasion across the basement membrane, at which SE6s based on the diagnostic morphological time the lesion is, by definition, diagnosed as can- features of intraopithelial neoplasia (dysplasia) cer. Over the many years it takes intraepithelial measured by computer-assisted quantitative image neoplasia to progress, the goal of cancer chemo- analysis (COLA) prevention is to find drugs that will drastically Proliferative status. The continuous and abnormal slow, stop or bring about regression of the intraep- increase in the proliferative rate of intraepithelial ithelial neoplastic lesions while they are still con- neoplastic lesions is part of the process of clonal fined to the epithelial compartment, thereby pre- evolution, which is described as the continuous venting progression of any one of them to invasive production within the neoplastic population of cancer. genetic variants that are able to escape growth

Environmental cancer epidemiology Clinical cancer epidemiology Epidemiologists: toxicologists Oncologists, pathologists Focus: cause and pathogenesis of cancer Focus: diagnosis and therapy of cancer

Observational studies Prospective intervention trials Cancer prevention by proscription (primary prevention) Cancer prevention by prescription (secondary prevention) Biomarkers: Biomarkers: subclinical disease markers Biologically effective dose (BED) e.g. adducts Pap smear showing nuclear atypia Biological response (BR), e.g. mutations. Biopsy showing diagnostic disease markers for measured individually by computer-assisted dysplasia measured individually by computer-assisted quantitative image analysis quantitative image analysis

275

Application of Biomarkers in Cancer Epidemiology

inhibitory controls and undergo clonal expansion cancer, accurately predicting the recurrence-free at the expense of the more slowly growing back- survival and overall survival either by itself (van ground cells (Nowell, 1986; Boone et ai., 1992 Diest et aL, 1992) or as part of a commonly used (review), 1993). As selected clones in the popula- grading system (Bloom, 1950). In the coloredum, tion continue to arise and expand by virtue of their an upward shift of the zone of maximat prolifera- faster growth rate, the overall mean growth rate of tion towards the neck of the gland (`stage II anom- the neoplastic population increases. aly') has been shown to be an important predictor The proportion of tissue cells in the proliferative of increased cancer risk (Deschner & Haskens, cycle versus those in Go (not proliferating) can be 1982). accurately measured using CQIA. The computer measures the percentage of the total nuclear area in Ploidy status. The DNA content per cell nucleus a section that binds to a chromogen-Iabelled anti- measured by CQIA is displayed as a histogram, body probe, such as antibodies to PCNA, Ki-67, from which the type and degree of variation from (using MIB-1) or BrdU (technique described in diploidy may be quantitatively estimated. Esteban et a1. 1993). The proliferating fraction may AneupIoidy has been shown to occur during also be measured by tritiated-thymidine up- intraepithelial neoplasia in bladder (Tribukait, take/autoradiography or by mitotic counts. The 1987), prostate (Montironi et al., 1990; Tribukait, mitotic index (percentage of cells in mitosis) has 1991; Amih et aL, 1993), breast (Carpenter et proven to be a reliable prognostic factor in breast а1., 1987; Erhardt & Auer, 1987; Crissrnan eta?., 1990;

Molecular cancer epidemiology

Markers of exposure Markers of disease

Exposure -- Internal - Biologically—*- Early biological —~Subclinical -- Clinical —+- Prognostic dose of effective dose lesion preinvasive disease significance carcinogenic early neoplastic agent disease lesions

DNA adducts Gene mutation Intraepithelial Cancer neoplasia

Protein Allelic loss adducts or gain

Chromosome aberration

Markers of susceptibility Genetic/metabolic Nutritional status (Р45д, GST) Immunological status DNA repair Disease status

Figure 1. The position of intraepithelial neoplasia (dysplasia), shown as a subclinical disease biomarker in the carcinogen exposure-neoplastic disease sequence.

276

Bioniarker end-points in cancer chemoprevention trials

Phases of Neoplasia

lntraepitheliai Extraepithelial = preinvasive = cancer = precanceг = malignant = premalignant

I II IП N V Focal Dysplasia Latent Aberrant {Focal Dysplastic Phase Proliferation Proliferation} Microinvasion

Мeta- stasis

- Atypical 1A-lavrв _ Б -7оyrs Breast Hyp rplasia flC1S ~T в $ Cervix CIN 1 ) CIN III/CIS 1б-2Оуг Invasion cancer Colon `ѕ Adenoma ~t Y~

Prostate 20 `s PIN > io уг

Figure 2. Diagram illustrating the concept that if multifocal intraepithelial neoplasia is stopped by the administration of a chemopreventive drug, further progression to invasive cancer is prevented.

Visscher et ai., 1993), cervix, skin, oral teukoplakia, Nuclear morphometry (nuclear size, shape and pleo- larynx, lung, oesophagus, stomach and colorec- morphism). It is quite remarkable that, in a number tum (reviewed in Boone et aI., 1992). In one study of studies, alteration of nuclear shape alone proved of breast ductal caicirioma in situ, the cribriform to be a better predictor of mortality in stage A2 pro- histological pattern exhibited 38% aneuploidy static cancer than were the Gleason, Mostofi or whereas the comedo pattern exhibited 82% aneu- Johns Hopkins grading systems, or ploidy status ploidy (Crissman et al., 1990). In another study, (Epstein et L, 1984; Partiri et a1., 1989; Mohler et atypical hyperplasia of the breast also exhibited a7., 1992;). Multivariate analysis of up to 16 nuclear shape descriptors, measured by CQIA, including aneuploidy iп 4 of 13 cases (Carpenter etal.,1987). With regard to invasive neoplasia, ploidy status nuclear roundness factor, variance of roundness has proven to be a reliable prognostic factor of dis- factor and nuclear ellipticity, have accurately pre- ease-free survival and mortality in both breast (van dicted recurrence of cancer after surgery in 11 of 26 Diest & Baak, 1992) and prostate (Lieber, 1992) patients with renal cell carcinoma (Pound et ai., cancer. Concerning pre-invasive neoplasia (dyspla- 1993), 7/14 patients with transitional cell carci- sia), DNA aneuploidy determined by CQIA has noma of the bladder (Borland et a2., 1993), and proven to be an effective SEB that correlates with 17/27 patients with Wilms' tumour of the kidney the severity of intraepithelial neoplasia of breast (Gearhart et al., 1992). Nuclear pleomorphisin is (Beerman et a1., 1990) and prostate (Montlroni et an important concomitant of neoplasia that may aL, 1990; Lieber, 1992; Petein et al., 1992; Amin et be measured as the simple mean of the variances of size and shape. яl., 1993). 277 Application of Biomarkers in Cancer Epidemiology

Nuclear chromatin texture. `Chromatin clumping' is deviation units from the mean, and a critical value a semiquantitaiive nuclear feature long recognized diagnostic of neoplasia may be defined for each by pathologists as a correlate of the extent of neo- feature (Dr James Bacus, personal communica- plastic progression. Mildly dysp1astic cells, for tion). A composite biomarker made up of a num- instance, tend to exhibit less chromatin clumping ber of diagnostic features may then be constructed, than severely dysp1astic cells. With OQIA, the degree of cliromatin clumping may be measured Development of molecular SEBs precisely. The multitude of possible optical density The Chemoprevention Branch is also evaluating patterns transmitted through many hundreds of various candidate molecular SEВs as predictors of pixels over a given cell nucleus has been system- high cancer risk and for their efficacy in short-term atized into what is known as the Markovian texture clinical trials. Molecular SEBs being studied include feature matrix (Pressman, 1976). Selection may be those related to genomic instability, oncogeree made among dozens of different Markovian tex- amplification, tumour suppressor gene loss, aber- ture features for those which best contribute to a rant differentiation molecules and aberrant expres- multivariate classification function that will pre- sion of growth regulatory molecules and their dict the likelihood of intraepithelial neoplasia pro- receptors. There are screening programmes in the gressing to invasive cancer (Doudkine et al., 1995). planning stage that will allow the collection of Nuclear chromatin texture analysis by CQIA there- cohorts of subjects with activated oncogenes or fore has excellent potential for providing many mutated or lost tumour suppressor genes, от with features which are useful as SEBs. genetic markers of susceptibility to cancer, which may become the subject of chemoprevention trials. Nucleolar morphometry (number, size, shape, posi- The most desirable molecular SEBs may be those tion and pleomorphism). In a study of nine пucleo- that appear before the onset of intraepithelial neo- lar morphometrical features in breast cancer, sim- plasia in normal-appearing epithelium. The evalu- ple nucleolar frequency among nuclei (total num- ation of early genomic instability is an example. ber of nucleoli per 100 nuclei) was a significant The development of such early 'pre-dysplastic' predictor of recurrence-free survival (van Diest et SEBs, which is now in progress in the programme cl., 1990). Changes in nucleolar morphoinetry of the Chemoprevention Branch, will permit many have been reported to be a correlate of the extent of more months or even years of treatment with neoplastic progression in prostatic intraepithelial chemopreventive agents over a period when the neoplasia (Helpap, 1988; Montironi et al., 1991). molecular lesions may be more susceptible to sup- Argyrophilic nucleolar organizer region-associated pressive drugs. Molecular SEBs that are superim- proteins found within nucleoli form the basis for posed on neoplastic tissue changes already identi- the AgNOR stain. Using this stain, the mean num- fiable microscopically may offer considerable addi- ber of nucleolar organizer regions (NORs) clearly tional information in tissue biopsies, particularly as distinguish between tubular adenomas, villous technology involving microdissection of individ- adenomas with moderate nuclear atypia, villous ual cells followed by PCR and gel analysis becomes adenomas with severe nuclear atypia and colorec- established. tal adenocarcinorna (Yang et al., 1990). The AgNOR stain, especially if quantitated by CQIA, offers References good potential as a SEB. Ames, B.N. (1995) The causes and prevention of cancer. Prос, NatiAcad. Sci. USA, 92, 5258-5265 Continuous grading scale measured in standard devi- Amin, M.B., Schultz, D.S., Zarbo, R.J., Kubus, J. & ation units. The individual morphological and pro- Shaheen, C. (1993) Computerized static DNA ploidy liferative features of intraepithelial neoplasia dis- analysis of prostatic intraepithelial neoplasia. Arch. cussed above characteristically progress as a group Pathel. Lab. led,, 117, 794-798 towards greater aberrancy. The quantitative extent Beerman, H., K1uin, P.M., Hermans, J., Velde, С .J.Н. & of aberrant variation of each feature in the group, Cornelisse, C.J. (1990) Prognostic significance of DNA- measured by computer-assisted image analysis, ploidy in a series of 690 primary breast cancer patients. may be graded in terms of the number of standard JuL J. Cancer, 45, 3-39

278 Biomarker end-points in cancer chemopreventioh trials

Bloom, H.J.G. (1950) Prognosis in carcinoma of the Gearhart, J.P., Partin, A.W., Leventhal, B. & Beckwith, J.I. breast. Br. J. Cancer, 4, 259-288 (1992) The use of nuclear morphometry to predict response to therapy in "Wilms" tumor. Cancer, 69, 804-808 Bоoпе, С .B., Kelloff, G.J. & Steele, V.E. (1992) Natural history of intraepithelial neoplasia in humans with impli- Helpap, B. (1988) Observations on the number, size, and cations for cancer chemoprevention strategy. Cancer Res., localization of nucleoli in hyperplastic and neoplastic 52, 1651-1659 prostatic disease. Hгstopathology, 13, 203-211 Boone, C.W., Kelloff, G.J. & Freedman, L. (1993) Hulka, B.S., Wilcosky, T.C. & Griffith, J.D. (1990) Biologic Intraepithelial and postinvasive neoplasia as a stochastic letarkers in Epidemiology, New York, Oxford University continuum of clonal evolution, and its relationship to Press mechanisms of chemopreventive drug action. J. Cell Lieber, M.M. (1992) DNA ploidy: early malignant lesions. Biochem., 17G, 14-25 J. Cell Biochem., 16Н, 44-46 Borland, R.N., Partin, A.W., Epstein, J.I. & Brendler, С .B. Mohler, J.~., Partin, A.W., Epstein, J.I., Becker, R.L., (1993) The use of nuclear morphometry in predicting Mike1, U.V., Sesterhenn, I.A., Mostofi, F.K., Gleason, D.F., recurrence of transitional сеll carcinoma. J. UroL, 149, Sharief, Y. & Coffey, D.S. (1992) Prediction of prognosis 272-275 in untreated stage АZ prostatic carcinoma. Cancer, 69, Bostwick, D.G. (1992) Prostatic intraepithelial neoplasia 511-519 (PIN): current concepts. J. Cell Biochem., 16Н, 10-19 Montironi, R., Scarpelli, M., Sisti, S., Braccischi, A., Carpenter, R., Gibbs, N., Matthews, J. & Cooke, T (1987) Gusella, P., Pisani, E., Alberti, R. & Mariuzzi, G.M. (1990) Importance of cellular DNA content in pre-malignant Quantitative analysis of prostatic intraepithelial neopla- breast disease and pre-invasive carcinoma of the female sia on tissue sections. Anal. Quant. Cytol. listai., 12, breast. Br. J. surg., 74, 905-906 366-372 Carter, D. (1985) 5quamou5 cell carcinoma of the lung: Montironi, R., Braccischi, A., Matera, G., Scaгрelli, M. & an update. 5еmin. Diag. PathoL, 4, 226-234 Pisani, E. (1991) Quantltation of the prostatic intraepithelial neoplasia. Analysis of the nucleolar size, number, Chanen, W. (1990) The CIN saga, the biological and clin- and location. athoI. Res. Pract., 187, 307— 14 ical significance of cervical intraepithelial neoplasta. Л З Aust, Ni J. Obstet. GynaecoL, 30, 18-23 Nowell, P.C. (1986) Mechanisms of tumor progression. Cancer Ras., 46, 2203-2207 Crissman, J.D., Visscher, М.D. & Kubus, J. (1990) Image cytophotometric DNA analysis of atypical hуреrрlаsiа Page, D.L., Dupont, W.D., Rogers, L.W. & Rados, M.S. and intraductal carcinomas of the breast, Arch. PathoL (1985) Atypical hyperplastic lesions of the female breast: Lab. Merl., 114, 1249--1253 a long-term follow-up study. Cancer, 55, 2698-2708

Deschner, Е.Е. & Maskens, А.Р. (1982) Significance of the Partin, A.W., Walsh, A.C., Pitcock, R.V., Mohler, J.L., labeling index and labeling distribution as kinetic para- Epstein, J.1. & Coffey, D.S. (1989) Comparison of nuclear meters in colorectal mucosa of cancer patients and DMH morphometry and Gleason grade as predictor of prig-. treated animals. Cancer, 50, 1136-1141 nosis in Stage АZ prostate cancer: a critical analysis. J. Uni., 142, 1254-1258 Doudkirne, A., MacAulay, C., Poulin, N. & Palcic, B, (1995) Nuclear texture measurements in image cytometry. Perera, F.P. & Weinstein, I.B. (1982) Molecular epidemi- Pathologica, 87, 180-193 ology and carcinogen-DNA adduct detection: new approaches to studies of human cancer causation. J. Epstein, 3.1,, Berry, S.J. & Eggleston, J.C. (1984) Nuclear Chronic Dis., 35, 581600 roundness factor. A predictor of progression in untreated Stage А2 prostate cancer. Caner, 54, 1666-1671 Peteiп, M., Michel, P., van Velthoven, R., Pasteels, J.L., Brawer, .K., Davis, J.R., Nagle, R.B. & Kiss, R. (1991) Erhardt, K. & Auer, G.U. (1987) Mammary carcinoma. М Morphonuclear relationship tw prostatic intraep- Comparison of nuclear DNA content from in situ and Ье ееп ithelial neoplasia and cancer as assessed by digital cell infiltrative components. Anal. Quant. Cytoi. Flistol., 9, image analysis. Am. J. li . PathoL, 96, 628-634 263-267 С п Pound, C.R., Partin, A.W., Epstein, 3.1., Simons, J.W. & Esteban, J.M., Kandalaft, P.L., Mehta, P., Odom-Maryon, Marshall, F.F. (1993) Nuclear morphometry accurately T.L., Bacus, S. & Battifora, I. (1993) Improvement of the predicts recurrence in clinically localized renal cell carci- quantitation of estrogen and progesterone receptors in noma. Urology, 42, 243-248 paraffin-embedded tumors by image analysis. Am. J. Clin. PathoL, 99, 32-38 Pressman, N.J. (1976) Markovian analysis of cervical cell images. J. Histochem. Cytochem., 24, 138-144

279 Application of Biomarkers in Cancer Epidemiology

Riddell, R.H. (1984) Dysplasia and cancer in ulcerative Visscher, D.W., Micale, M.A. & Crissman, J.D. (1993) colitis: a soluble problem? Scaod. J. GаstroeпterоL, 104, Pathological and biological relevance of cytophotometric 137-149 DNA content to breast carcinoma genetic progression. J. Cell Biochem., 17G, 114-122 Rothman, K J. (1986) Modern ергдeтiоlоgy. Boston, Little, Brown. pp. 52x53 Wright, T.C., Kurman, R.J. & Ferenczy, A. (1994) The Multicenmc Study of Colorectal Adenomas (SMAC) Precancerous lesions of the cervix. In: Kurman, R.J., ed., Workgroup (1995) A multicenter study of colorectal ade- Blansteins Pаthology of the Female Genital Tract, New York, nomas: rationale, objectives, methods, and characteris- springer-Verlag tics of the study cohort. Tarif, 81, 157-163 Yang, P. Huang, G.S. & Zhu, X.S. (1990) Role of nuclei- Tribukait, B. (1987) Flow cytometry in assessing the ciii- lar organizer regions in differentiating malignant from ical aggressiveness of genito-urinary neoplasms. World J. benign tumours of the colon. J. Clin. PathoL, 43, 235-238 UroL, 5, 108-122 Tribukait, B, (1991) DNA flow cytometry in carcinoma of the prostate for diagnosis, prognosis, and study of tumor biology. Acta licol, 30 187-192 van Diest, P.J., Mouriquand, J., Schipper, NW. & Baak, J.P.A. (1990) Prognostic value of nudeolar morphometric variables in cytological breast cancer specimens. J. CГn. PathoL 43, 157-159

van Diest, P.J. & Baak, J.Р.А. (1992) Quantitative Cyto- and Histoprognosis in Breast Cancer, New York, Elsevier, pp. Corresponding author 55-62 C'W. Boone van Diest, P.J., Baak, J.P.A. Matze-Cok, P. & Wisse- Chemoprevention Branch, Division of Cancer Brekelmans, E.C.М. (1992) Reproducibility of mitosis Prevention and Control, National Cancer Institute, counting in 2,469 breast cancer specimens: results from National Institutes of Health, Bethesda, MD 20892, . the multicenter inorphometric mammary carcinoma USA project. 1-Ir' PatiwL, 23, 603--607

280 Application ы Biomarkers it Cancer Epidвmioiоgy Toniolo, P. Ralfatta, P., Shuker, DEC., Rothman, N. Hulka, B. and Pearce, N., edе IARC Scieпt9iе PuЫ iaations Na. 142 lпterпаtиоеаl Agency for Research on Cancer, Lyon, 1997

The use of biological markers as predictive early-outcome measures in epidemiological research

А.J. McMichael and А . Hall One of the possible uses of biomarkers in epidemiological research is as early-outcome measures to predict the occurrence of clinical disease and to elucidate the biological mechanism of pathogenesis. This use is conceptualIy less straightforward than the well established use of biomarkers to improve or extend exposure assessment or to study interindividual variations in disease susceptibility. In principle, this form of use could accelerate or otherwise facilitate etiological research. However, in practice, the recent review literature suggests that this mode of biomarker use, especially in cancer epidеmiolоgy, is the least clear- cut and the least well developed. The recurrent problem is identifying biomarkers that: (1) are on the causal pathway, (2) have a high probability of progression to clinical disease, and (3) account for all or most of the cases of the specified clinical outcome. Such biomarkers would be most useful if they conferred a long lead-time relative to clinical disease occurrence.

The use of biological markers, molecular or other- effective dose'. This last integrates either multiple wise, in epidemiological research should be as a external sources of exposure or multiple internal means and not as an end (Hulka, 1991; McMichael, metabolic steps, or both. 1994). Therefore, the critical question to ask in any The second mode of biomarker use—the identi- particular study is: 'Has use of the biomarker in this fication of subgroups of study subjects with study yielded new information, or better informa- different levels of susceptibility to some particular tion, for understanding, and perhaps quantifying, exposure—often depends on measuring some the causation of the disease of interest?' biochemical or molecular genetic characteristic. The four major contexts in which biomarkers For example, one recent study indicated that the may be used in epidemiological research are: (1) to reduction in coronary heart disease mortality risk improve the assessment of exposure; (2) to identify associated with increasing levels of consumption of and take account of subgroups of persons of dif- alcohol, within the low-moderate range, depends fering susceptibility to the effects of the exposure; on whether the individual is Lewis blood group (3) as measures of early outcome with known (or positive or negative (Hein et a1., 1993). 5imilагly, presumed) predictive significance; and (4) as a the risk of colon cancer associated with increasing basis for differentiating disease subtypes with levels of meat consumption appears to depend on potentially different etiologies. whether an individual is a fast or a slow acetylator The first of these uses, improving the assess- (Roberts-Thomson et al., 1996)—a polymorphic ment of exposure, can be enhanced by use of a metabolic characteristic that is measurable directly wide range of biological markers, from blood cho- via genotype analysis of NAT2 alleles (Minchin et lesteroI aid urinary cotinine, to levels of DNA al., 1993; Roberts-Thomson et ad., 1996). adducts in surrogate and target tissue cells. Some These two uses of biomarkers can sometimes be such measures will identify component exposures combined to elucidate, or confirm, epidemiological within a complex mixture, others will confirm that relationships. For example, cigarette smokers are an internal exposure has indeed occurred, while known to be at increased risk of cancer of the urinary others yield an integrated measure of biologically bladder. The risk of this cancer is also higher in

281 Applicatron of Biomarkers in Cancer Epidemiology slow acetylators than in fast acetylators (Lower et In each of the first three of these uses, the pre- al., 1979; Hansen et cl., 1985; Cartwright et al., dictive ability of the biomarker is inferred from 1982). Since arylamines, present in cigarette smoke, knowledge of the natural history of the disease, are a known occupational cause of bladder cancer and is then applied in ways that save time in epi- (especially R-naрhthуlamine) and are metabolized demiologicaI research or that save time (and lives) by NAT2, the acetylation pathway, it has been in public health prevention. In the fourth use, one postulated that the carduogenic effect of certain examines whether the inferred predictivity of the arylamines might account for the epidemiological biomarker is also evident under the strict condi- findings. Studies of the levels of DNA adducts in tions of the controlled intervention—i.e. is there a white blood cells in different categories of persons strong within-individual correlation between have shed further light on those relationships: changes in the biomarker, as induced by the smokers have higher levels of 4-amino-bipherryl- intervention, and occurrence of the clinical éпд- haemoglobin adducts than non-smokers (Bryant et point (e.g. the Concorde Trial; see section on the cL, 1987), and in smokers (especially 'passive 'confirmation of the predictivity of biomarkers' smokers') the level is higher in slow than in fast below)? acetylators (Vineis et cL, 1994). The third possible use of bioniarkers—the use Routine screening for (immediately) pre-clinical of early-outcome measures to predict the occur- conditions rence of clinical disease—is the subject of this chap- There is a well established use of precursor lesions ter. It is conceptually less straightforward than the at the macroscopic level in clinical medicine for the previous two uses; its principal purposes appear to early detection of disease, e.g. precancerous polyps be to accelerate (or otherwise facilitate) etiological of the large bowel and hypertension-associated epidemiological research and to elaborate the arteriovenous 'nipping' within the retina. Similarly, intervening steps along the causal pathway. The the early detection of women likely to develop recent review literature suggests that this third cancer of the cervix is achieved by taking cell sam- mode of biomarker use, especially in cancer epi- ples from the cervical mucosa and seeking micro- demiology, is the least clear-cut and the least well scopic evidence of dysplasia. In general, this type developed (Perera & Santella, 1993; McMichael, of relatively late-stage precursor lesion is useful for 1994). Typically, there is a preference for identifying population screening programmes directed at the and using biomarkers that confer a long lead- secondary prevention of clinical disease. time relative to the appearance of clinical disease. A more novel cancer screening idea has recently Such biomarkers may be cellular, biochemical or arisen from the observation that persons with ade- molecular. nomatous polyps (which carry a high probability The fourth above-mentioned use of bioimrkers of progression to carcinoma) may be identified by is outside the scope of this paper. By differentiating the presence, in shed cells in faecal material, of the disease subtypes and thus increasing the specificity K-ras (oncogene) mutation (sidrausky et al. 1992). of the data analysis, this particular use should This mutation is one of the well known four muta- enhance the informativeness of a study (e.g. Taylor tions and is thought to be critical in the develop- et cL, 1992; Vahakangas et al., 1992). ment of colon carcinoma (Vogelsteias et cl., 1988; Fearon & Vogelstein,1990). The four mutations are Uses of predictive' biomarkers of the APC tumour suppressor gene (chromosome There are four main uses of biomarkers as predic- 5), the K-ras oncogene (chromosome 12), the DCC tors (or 'surrogate end-points'): (1) screening for tumour suppressor gene (chromosome 18), and the pre-clinical disease; (2) enhancement of conven- p53 tumour suppressor gene (chromosome 17). tional epidemiological studies of disease etiology; Note, however, that a prerequisite for any suc- (3) monitoring for variation or change in popu- cessful screening procedure is prior knowledge that lation health risk; and (4) confirmation, via con- the lesion is clearly predictive of the subsequent trolled intervention studies, of biomarkers that occurrence of clinical disease. There would be little could subsequently be applied in the evaluation point in detecting cervical cellular dysplasia if it and monitoring of primary prevention measures. were only weakly predictive of subsequent cervical

282 Biological markers as predictive early-outcome measures carcinoma. Such knowledge of the capacity to pre- Gaining earlier answers in prospective epidemiologi- dict prospectively can on]y cbrne from studies of the cal studies. Biomarkers could, in principle, help to natural history of the disease process. (Note that this gain both earlier and better answers to prospective is a different notion from that other central para- studies. Better answers may result by exclusion of meter of a screening test, `positive predictive value' ostensibly healthy persons who, via biomarker (PPV). The PPV is the probability that the presence screening, are assessed as already having the pre- of this lesion, as measured within a specified popu- clinical form of the disease. For example, from lation of screened persons, signifies the true pres- many studies over the past several decades, it has ence of the early, pre-cliпicaI, disease process.) been concluded that a low blood cholesterol con- The clinically oriented use of precursor lesions for centration is not a cause of increased risk of cancer screening, within either a clinical case-finding con- (with the possible exception of colon cancer), but text or a systematic population screening context, is is a metabolic manifestation of incipient cancer well established. This paper, however, addresses the (McMichael et al., 1984). It would therefore be use of biochemical and molecular biomarkers for possible to screen out from a newly recruited two other purposes: the direct enhancement of epi- cohort all persons with manifestly low blood demiological research into disease etiology and the cholesterol—or at least identify them for subsequent substantially earlier detection of altered health risks stratified analysis of cancer incidence. in monitored populations, in each case, there would Earlier answers may be obtained from epidemi- be a general preference for biomarkers as indices of ological studies if the use of a biomarker increases rather earlier stages in disease pathogenesis than is the statistical power of a study (McMichael 1994). the case with clinically oriented screening. For example, if instead of counting numbers of persons with cancer it were equally valid to count Direct enhancement of epiderniological research into the number of cells from standard-sized tissue disease etiology samples (e.g. 500 white blood cells per person) dis- The use of biomarkers as surrogate end-points in playing some specific precancerous predictive epidemiological research could, in principle, mutation, then statistical power would be greatly enhance the research in two main ways: increased (lattis & 5ilvеr, 1993). With respect to the task of gaining earlier 1. Such biomarkers could provide earlier answers from cohort studies, the sine qua non is answers in cohort studies by avoiding the need prior knowledge that the biochemical or molecular to await clinical disease end-points. Biomarker- marker is actually strongly predictive of the clini- based case–control studies could also yield cal disease, and that clinical disease never or rarely earlier answers (in relation to the advent of occurs without this antecedent marker, i.e. the some newly introduced exposure agent). This `attributable fraction' approaches unity. Note that would require identification of a random sam- this does not mean that the biomarker is necessar- ple of all persons ('cases') with the `predictive' ily on the causal pathway; the biomarker may sim- lesion. For example, if a new subtype of HPV ply be an outcome that is very highly correlated were suspected of causing cervical cancer, with the occurrence of the clinical disease. It is women with dysplastic cells (detected via only necessary, in a logical sense, that the bio- screening) could be compared with controls for marker can be used as a surrogate, antecedent, the presence of viral subtype DNA. event for the actual clinical disease event. This is nevertheless a demanding criterion: if the bio- 2. Such biomarkers could elucidate, or confirm, marker does not represent a stage iп the causal the underlying biological process of disease chain, there 1s the possibility that, while some par- pathogenesis. This, in turn, may assist in the ticular exposure agent to be studied does induce interpretation of epidemiological findings and the disease, it does not induce the biomarker in the drawing of causal inference. epiphenomenon (or vice versa). Even assuming that the chosen predictor does How well do these two main potential uses of indeed lie on the causal pathway, there may be predictive biomarkers work in practice? other late-stage risk factors that account for the

283 Application of 8iomarkers in Cancer Epidemiology

progression from early 'lesion' to clinical event. markers of cellular abnormality. For example, some Such factors would not be identified iп studies that studies have used the incidence of cytogenetic use the biomarker as the surrogate measure of clin- damage (e.g. micronucleus formation) to evaluate ical disease outcome. For example, whereas dietary the cancer-protective effect of certain interven- fat intake may substantially account for the genesis tions, such as the effect of micronutrient supple- of the underlying atherosclerotic arterial disease, mentation on the progression of oesophageal the actual formation of an occlusive blood clot may dysplasia (Munoz et ai., 1987). Are there other pos- be precipitated by other factors, such as cigarette sibilities? Would we, in carrying out a cohort study smoldng, deficient intake of certain micronutrients of diet and colon cancer, settle for the measure- (vitamin C, folic acid, etc.) or exposure to high ment of particular cancer-associated mutations in concentrations of airborne respirable particulates. faecally shed cells as the index of impending carci- There is an unavoidable dilemma here. To gain noma? Could a very specific mutation —for example, the most leverage in time, the biomarker should the third base pair mutation of codon 249 in the substantially antedate the clinical end-point. p53 gene, which, in liver tumour cells, is strongly However, the earlier the surrogate end-point, the less (but probably not exclusively) associated with alla- likely it is to be strongly predictive of the clinical toxin exposure (Bressac et al., 1991; Hsu et al., end-point. Two recent cohort studies have exam- 1991)—be measured in the DNA of white blood cells ined the predictivity of cytogenetic damage— as a surrogate index for the presumed occurrence of structural chromosomal aberrations, sister chro- the same mutation in the relevant target tissue? The matid exchanges and micronuclei (Hagmar et al., answer at the moment is that we do not yet have 1994; Bonassi et al., 1995). In both studies there is certain aid specific molecular biomarkers that evidence that a high level of chromosomal aberra- would be satisfactory alternative 'outcome' mea- tions in white blood cells foreshadows an increased sures in epidemiological studies of cancer etiology. risk of cancer. However, the risk increases are mod- Discussion of the use of early predictive indices est and there is no indication that such cytogenetic in epidemiological research tеnds.to focus on can- indices could be used as a substitute for cancer in cer epidemiology and the associated mutational etiological studies. events. The genotoxic model of carcinogenesis has It may be helpful also to consider several examples a clarity about it that is unusual in the iealm of from the non-cancer arena. A substantial scientific disease etiopathogenesis. The model assumes a literature indicates that coronary atherosclerosis, 'staged' mechanism that entails the sequential with manifest narrowing of coronary arteries, is cumulation of critical mutations. While genotoxi- predictive of ischaemic heart disease; that the rate city is only part of carcinogenesis, current knowl- of loss of lung function (e.g. FEV1) is quantitatively edge arid theory propose that it is an important part. predictive of life-shortening at older ages; and that Nevertheless, various pre-clinical chemical changes impaired glucose tolerance predicts the onset of (e.g. carcinoembryonic antigen (CEA) and alpha non-insulin-dependent diabetes mellitus (NIDDM). fetoprotein (AFP)) may be valuable as biomarkers, We might therefore use these subclinical disorders particularly since they probably reflect actual as 'outcomes' in etiological studies, seeking to cellular functional change. Although they tend to understand better the etiology of the clinical dis- occur at later stages in carcinogenesis than genetic ease. For example, in an ongoing follow-up study and chromosomal damage, and therefore offer less of a Swedish cohort, the occurrence of impaired lead-tune in etiological research, they may have glucose tolerance in mid-adult life is being used greater predictive power. as a surrogate for putative subsequent clinical Few other disease processes have the seemingly NIDDM (Lithell et al., 1996). `staged' character of carcinogenesis. In heart dis- There are, so far, very few good examples of the ease, diabetes, asthma and cerebral dementia, use of surrogate end-points as substitutes for the there are no mutation-like 'switches' thrown—or eventual clinical event. Sоше cancer epidemiology at least none that we know of. For non-cancer examples exist within the controlled trial domain. diseases, it is therefore much harder to imagine However, their focus is on cytogenetic events, what the critical non-tissue-based predictive bio- including specific mutations, rather than 'later' markers might be. (Examples are more evident

284 Biological markers as predictive early-outcome measures within clinical eрidеmiology, in the study of the or by affecting the immune response to hepatitis B factors that influence the опtсоте of disease—see virus. In relation to the latter possibility, for example, section below.) in one study the level of aflatoxin was found to be Overall, and until we have much more knowl- higher in children who were carriers of the virus edge about the significance and predictive power than in those who were not (Allen et aI., 1992). of these various biomarkers, it is clear that they The prevailing view, in line with the laboratory cannot simply be substituted for the clinical end- experimental evidence of the potent direct chemi- point of classical cancer epidemiological research. cal carcinogenicity of aflatoxiri, is that it probably Definitive answers, for the moment (perhaps acts as a genotoxin. Recent evidence that aflatoxin always?), must continue to depend on studies that may cause an unusually specific mutation of the use the occurrence or non-occurrence of cancer as important X53 gene—the 'hot spot' mutation at the outcome variable. codon 249 in which the third base pair undergoes G to T transversion—has therefore advanced dis- Elucidating, or confirming, the steps in the underlying cussion of this genetic option. Mutations of the biological process. A perennial discussion in epi- p53 gene, both 'hot spot' and otherwise, appear to demiology concerns the sufficiency of the 'black occur with equal frequency in tumours associated box' approach. Do we need to know more than that with persistent hepatitis B infection and in those a particular exposure or circumstances is empiri- not associated with this infection (Oda et al., cally associated with an increased risk of disease? 1992), which suggests that p53 mutation is not a For the progression of scientific ideas and the poten- result of hepatitis B virus infection. However, there tiation of future etiological research, knowledge of is a high frequency of this specific mutation of p53 the mediating mechanism is important. However, (45% of tumours) in the high-incidence areas of for the practical success of many public health inter- the world, such as Mozambique and Qidong, ventions, it is actually not necessary to know any- China (Ozturk etaL, 1991; 5heu et aL, 1992), where thing about the mediating biological mechanisms. exposure to aflatoxin also tends to be raised. This It may be enough, then, to know that smoking cig- ecological evidence has suggested that the mutation arettes causes lung cancer, or that drinking water may be specific to genotoxic aflatoxin exposure— from the Broad Street pump causes cholera. a notion that is further supported by the fact that this However, it is becoming increasingly clear to is not a 'hot spot' mutation in tumours other than cancer epidemiologists that an understanding of hepatoce11ular carcinoma within those populations. mechanism—e.g. which components of cigarette This evidence does not necessarily indicate that smoke cause lung cancer?—will allow further stud- the p53 mutation is the direct mechanism of alla- ies of interindividual variation in susceptibility. In toxin carcinogenicity since it may merely be acting the case of smoking, such knowledge would also as a surrogate marker of aflatoxin exposure in these assist the investigation and risk quantification asso- populations. There may still be a coexistent non- ciated with passive smoking. On a wider research genotoxic effect of aflatoxin, such as suppressing the front, epidemiologists now perceive that further immune defenses against HBV. However, the exam- understanding of the role of diet in many cancers, ple illustratés how the study of mutational spectra of alcohol in breast cancer, EMF in leukaemia and can yield an insight into plausible causal mecha- brain cancer, and ambient air pollution in lung nisms and, if the appropriate set of observations is cancer will require elaboration of the mediating made, can strengthen the basis for causal inference. mechanisms. The empirical epidemiological evi- Another unresolved diet-cancer issue is whether dence on its own 1s neither sufficiently strong nor antioxidants, especially fl-carotene, reduce the risk consistent to allow causal inference. It therefore of cancer, and if so, whether this accounts for the becomes attractive to study the sequence of prob- lowered risks of many cancers in association with able, or plausible, mediating biological events. above-average intake of fresh fruit and vegetables. The aflatoxin and liver cancer story provides an There are evident difficulties in carrying out suffi- instructive example (McMichael, 1994). Aflatoxin ciently large and long-term intervention trials, and may increase the risk of liver cancer either by direct inconsistent results have emerged recently from genotoxic action, by inducing oxidative damage, studies in China (Taylor et ai., 1994), Finland (ABC

285 Application of Biomarkers in Cancer Epidemiology

Study Group, 1994) and the USA (Hennekens et al., vicinity of the Chernobyl reactor site showed a rate 1996). In order to help clarify this situation, of mitochondrial genetic damage (base pair substi- Duthie and colleagues have carried out studies to lutions in the cytochiome b gene) that was several assess the extent to which, first, dietary antioxi- hundred times greater than in vole populations living dants as measured in blood are correlated with 32 km south-east of Chernobyl (Baker etaL, 1996). reduced reporter mutations in lymphocytes It is Iikely that this genetic damage was also sus- (Duthie et aI., 1995) and, second, antioxidant sup- tained over several years of low-dose-rate exposure. piementation (for 20 weeks) reduces oxidative Biomarkers mау also be used to clarify the role of DNA damage in lymphocytes (Duthie et al., 1996). exposure agents at different stages of the cancer The second study explicitly used bioniarkers of process. In particular, there has been long-standing DNA damage or mutation 'as indicators of car- interest in thе notion 0f 'initiators' and 'promotors', cinogenic risk'; the authors argued that 'DNA dam- or, less specifically, early-stage and late-stage effec- age, resulting in base change and mutation when tors. If biomarkers (e.g. biochemical markers such as replication occurs, is a very early evert in carcino- CEA and AFP) were able to identify the cancer stage genesis'. Subsequently, the authors concluded already reached, then the subsequent risk alteration (Duthie et al., 1996): associated with an exposure of interest would tell something about its stage (and mode?) of action. Herein, we show for the first time a highly significant moderating effect of long-term anti- Early detection of altered health risks in monitored oxidant supplementation on endogenous and populations exogenous oxidative DNA damage in lyrr pho- The above two sections refer to studies of etiology cytes, supporting the hypothesis that dietary and causal mechanism. In some circumstances, antioxidants may protect against cancer. where etiology and causation are not at issue but where there is a need to identify any (suspected) The discourse concerning radiation-induced change in population health risk, then biomarkers germlirie mutations as a possible cause of human may be used for monitoring purposes. cancer—highlighted by the study indicating an This approach is well known in the occupational increased risk of leukaemia in the children of radia- arena. Biological monitoring of exposed population-exposed Seiafield workers (Gardner et al., tions can provide early evidence that critical levels 1990)—has recently been extended by a study in of exposure have occurred and that the risk of clin- Belarus in the Mogilev district, north of Chernobyl ical disease has increased. For example workers in (Dubrova et al., 1996). The frequencies of lympho- lead-exposed jobs have long had their blood lead cyte DNA mutations (tandem repeat minisatellite (PbB) concentrations monitored. Lead enters the mutations, detected via five separate DNA probes) body from multiple environmental sources (food, were compared in samples of 79 Belarus сhildren and water and air), and, in the first instance, is measur- 105 UK children, all born in 1994. The mutation rate able in the blood. Because lead is concentrated in Belarus children was double that of UK children, within the red blood cells, PbB gives an integrated and thе rate in Belarus children showed a strong pos- measure of lead exposure over the preceding several itive association with level of local topsoil radiation months. It also integrates the exposure measure- (caesium-137) contamination. Those mutations ate ment across all exposure media. presumed to have been inherited from parents who Increasingly, various occupational exposures are were exposed in the aftermath of Chernobyl. being monitored via cytogenetic studies (Gun & This study of heritable radiation-associated McMichael, 1990). This includes ionizing radiation mutations thus adds some plausibility to the ear- and solvent exposures. The presumption here is that lier, necessarily inconclusive, epidemiological evidence of macroscopic damage to chromosomes research at Sеilafield. It also supports the general indicates an increased probability of mutational thesis that the dose rate is a critical determinant of events, which carry an increased risk of cancer. risk of genetic damage, wherein the unit-dose risks A mutation known to be particularly strongly are higher at lower dose rates. Concurrent research associated with a specific type of cancer may be in subterranean voles living in the immediate useful in monitoring changes in the risk status of

286 Biological markers as predictive early-outcome measures the population in relation to the future incidence Other considerations of that cancer. An interesting example arises in rela- The use of biomarkers, especially within a particu- tion to the prediction that the ongoing depletion of lar assumed mechanistic framework may suggest, stratospheric ozone will cause an increase in the mistakenly, that the mechanism is universal, incidence of skin cancers (Mаdrопich & de Gruijl, i.e. that it applies in all populations. However, 1994). Epidemiological monitoring of the inci- the significance—i.e. the predictive power—of bio- dence of skin cancers (basal се11, squamous cell, markers may vary considerably between popula- melanoma) would therefore be desirable in 'sen- tions. tinel populations over a range of latitudes. While This variation may reflect genetically based dif- data on melanoma can be captured by cancer reg- ferences in disease processes or rates of progres- istry data, basal cell and squamous cell cancers sion, or it may reflect culturally determined varia- may be less reliably reported. However, since these tions in those processes. Most non-acute disease clinically defined end-points may take years to processes have complex etiologies, with webs of become manifest, earlier predictive biomarkers are causation that differ among geographic, ethnic, needed. One possibility is to use certain direr- genetic and socio-economic groups. For example, forming mutations of the p53 gene in skin cells, the prevalence of chromosomal abnormalities in which appear to be induced specifically by UV white blood ceI1s in healthy population samples exposure (Ziegler et cL, 1993; Nakazawa et cL, was found to be considerably greater in West Africa 1994). It has also been suggested that some non- (Gambia) than in Europe, despite the fact that the human animals may be particularly useful as indi- former population has a much lower lifetime risk cator species—for instance, the exposed ear epithe- of cancer overall (Miele et cL, 1996). hum of sheep may be especially sensitive to UV- induced molecular biological damage. Conclusions The most widespread use of biomarkers in epi- Confirmation of the predictivly of biomarkers in con- demiology has been to improve the quality of trifled intervention studies exposure assessment. There is also now a growing Controlled intervention studies provide an oppor- usé of metabolic and genetic biomarkers to iden- tunity to test whether the predictive power of the tify population subgroups at differing susceptibil- biornarker is the same under the strict conditions ity to exposure-related disease. of controlled experimentation as has been inferred The application of biomarkers as predictors of from natural history observations. Within-individ- disease end-points is much less well-developed. ual changes in both the biomarker and the clinical This reflects the intrinsic difficulty in obtaining end-point are studied in response to the interven- the prospective evidence that would confirm the tion. Confirmation that an intervention-induced predictivity of the biomarker, and the uncertainties change in the biomarker is followed by a change in that persist so long as the causal role of the bio- the clinical end-point establishes the predictive marker (or its underlying phenomena) in the dis- significance of the biomarker (at least under those ease process under investigation is unclear. circumstances). Beyond their established role in screening for The recent results of the Concorde Trial of disease and their emerging role in monitoring zidovudine (AZT) treatment in HIVIAIDS under- high-risk populations, there is a need to evaluate score the need for caution in assuming that pre- critically the use of biomarkers in the enhance- sumed precursor indices of outcome can be used ment of epidemiological studies of disease etiology as a substitute measure for the outcome (Concorde and its primary prevention. Biomarkers can some- Coordinating Committee, 1994). The Concorde times improve the definition of disease entities. Trial found that drug-associated change in the They may also, in future, become increasingly CD4 count—which refers to a change in T-cell useful as early, surrogate, indices of clinical profile and is a widely recognized index of damage end-points, both for cancer and other diseases. In to the immune system—was predictive of neither general, however, much remains to be learnt the subsequent incidence of clinical AIDS nor about their predictive significance before that lat- survival. ter goal can be widely realized.

287 Application of Biomarkers in Cancer Epidemiology

References Gardner, M.J., Snee, М.Р., Hall, A.J., Powell, C.A., ABC (Alpha-Tocopherol, Beta-Carotene Cancer Downes, S. & Теrтеll, J. (1990) Results of case-control Prevention) Study Group (1994) The effect of vitamin E study of leukaemia and lymphoma among young people and beta carotene on the incidence of lung cancer and near Sellafield nuclear plant in West Cumbria. Br. Med. J., other cancers in male smokers. New Eng1. J. Med., 330, 300, 423-429 1029-1035 Gun, R.G. & McMichael, A.J. (1990) The expanding role of biological markers in occupational health research Allen, S.., Wild, S.P., Wheeler, J.G., Riley, E.М., Montesano, R., Bennett S., Whittle, H.C., Hall, A.J. & and monitoring. J. Occup. Health sa feiy (ANZ), 6, 413-417 Greenwood, B.М. (1992) Aflatoxin exposure, malaria aid Hagmar, L., Brogger, A., Hansteen, I.L., Heim, S., hepatitis B infection in rural Gambian children. Trans. Hogstedt, B., Knudsen, L. & Lambert, B. (1994) Cancer Royal Soc. Tropical Med. Hyg., 86, 426-430 risk in humans predicted by increased levels of chromo- Baker, R.J., Van Den Bussche, R.A., Wright, A.J., Wiggins, somal aberrations in lymphocytes: Nordic study group on the health risk of chromosome damage, Cancer Res., L.E., Hamilton, М.J., Reat, Е,P, Smith, M.H., Lomakin, M.D. & Chesser, R.K. (1996) High levels of genetic 54, 2919-2922 change in rodents of Chernobyl. Nature, 380, 707-708 Hansen, Н.P, Agarwal, D.P., Goedde, 1W., Bucher, H., Bonassi, S., Abbondandolo, A., Camurni, L., Dal Pra, L,, Huland, H., Brachmann, W. & Ovenbeck, R. (1985) De Ferrari, M., Degrassi, F. & Fini A. (1995) Are chro- Association of N-acetyl-transferase polymorphism and mosome aberrations in circulating lymphocytes predic- environmental factors with bladder carcinogenesis. Pur. tive of future cancer onset in humans? Preliminary Urd., 11, 263-266 results of an Italian cohort study. Cancer Genet. lattis, D. & Silver, K. (1993) Use of biomarkers in risk CytogeneL, 79, 133-135 assessment. In: Schulte, PА. & Perera, F.P., eds, Molecular Epidemiology: Principles and Practices, New York, Academic Brеssac, B., Kew, M., Wands, J, & Ozturk, M. (1991) Selective G to T mutations of p53 gene in hepatocellular Press, pp. 251-273 carcinoma from southern Africa. Nature, 350, 429-431 Hein, H.O., Sorensen, H., Snadicani, P. & Gyntelberg, F. (1993) Alcohol consumption, Lewis phenotypes, and risk Bryant, M.S., Skipper, P.L., Tannenbaurri S.R. & MaсLше, M. (1987) Hemoglobin adducts of 4-armnobiphmy1 in of ischaemic heart disease. Lancet, 341, 392-396 smokers and nonsmokers. Cancer Res., 47, 612-618 Hennekens, C.H., Buring, J.E., Manson, J.Е., StampFer, Cartwright, R.A., Glashan, R.W., Rogers, H,]., Ahmad, M., Rosner, B., Cook, N.R., Belanger, C., LaMotte, F., Gaziano, J.M., Ridker, P.M., Willett, W. & Peto, R. (1996) А.А., Barham-Hall, D., Higgins, E. & Kahn, M.A. (1982) Role of N-acetyltranferase phenotypes in bladder car- Lack of effect of long-term supplementation with beta cinogenesis: a pharinacogenetic epidemiological carotene on the incidence of malignant neoptasms and approach to bladder cancer. Lancet, 2, 842-645 cardiovascular disease. N. Engt. J. Мед., 334, 1145-1149 Concorde Coordinating Lommitteé (1994) Concorde: Hsu, I.C., Metcalf, R.A., Sun, Т., Welsh, J.А., Wang, N.J. & MRC/ANRs randomised double-blind controlled trial of Harris, C.C. (1991) Mutational hotspot in the p53 gene immediate and deferred zidovudine in symptom-free in human hepatocelutar carcinomas. Nature, 350, 4271128 HIV infection. Lancet, 343, 871--881 Hulka, S.S. (1991) Epidemiological studies using biolog- Dubrova, Y.E., Nesterov, V.N., Krouchinsky, N.G., ical markers: issues for epidemiologists. Cancer Epidemiol. Biouarkers Prey., 1, 13-19 Ostapenko, V.A., Neumann, R., Neil, D.L. & Jeffreys, А.J. (1996) Human mincate1lite mutation rate after the Lithell, H.O., McKeigue, P.М., Berglund, L., Mofisen, R., Chernobyl accident. Nature, 380, 683-686 Lithell, U. & Leon, D. (1996) Relationship of birtbweight Duthie, S.J., Ross, M. & Collins, A.R. (1995) The influ- and ponderal index to NIDDI and insulin response to ence of smoking and diet on the hypoxanthine phos- glucose challenge in men aged 50-60 years. Br. Med., J., phoribosyltransferase (hprt) mutant frequency in circu- 312, 406-410 lating T lymphocytes from a normal human population. Lower, G.M., Jr, Nilsson, T., Nelson, C.E., Wolf, H., Mut t. Res., 331, 55-64 а Gamsky, Т.E. & Bryan, G.T. (1979) N-Acety1transferase phenotype and risk in urinary bladder cancer: Duthie, S.J., Ia, A., Ross, М.A. & Collins, А.А. (1996) Antioxidant supplementation decreases oxidative DNA approaches in molecular epidemiology: preliminary damage in human lymphocytes. CancerRes., 56,1291-1295 results in Sweden and Denmark. Environ. Health Perspect., 29, 71-79 Fearon, E.R. & Vogelstein, B. (1990) A genetic model for colorectal tumorigenesis. Cel, 61, 759-767 McMichael, A.J. (1994) 'Molecular epidemiology': new

288 Biological markers as predictive early-outcome measures pathway or new travelling companion? Am. J. Epidemiol., Sidransky, D. Tokino, T., Hamilton, S.R., Kinzler, К.W., 140, 1-11 . Levin, B., Frost, P. & Vogelstein, B. (1992) Identification of ras oncogene mutations fn the stool of patients with Mclichael, A.J., Jensen, O.M., Parkin, D.M. & Zaridze, curable colorectal cancer. Science, 256, 102-105 D.G. (1984) Dietary and endogenous cholesterol and human cancei. Epidemiol. Rev., 6, 192r-210 Taylor, J.А., Sandier, D.Р., Bloomfield, C.D., Shore, D.L., Ball, E.D., Neubauer, A. Mcintyre, O.R. & Liu, E. (1992) Ras Madronich, Ѕ. & de Gruijl, F.R. (1994) Skin cancer and oncogene activation and occupational exposures in acute UV radiation. Nature, 366, 23 myeloid leukemia. J. NяtI Cancer Inst., 84, 1626-1632 Miele, M., Donato, E, Hall, A.J., Whittle, H., Chapot, B., Taylor, PR., Li, B., Dawsey, S.M., Li, J-Y., Yang, .S., Guo, Bonatti, S., De Ferrari, M., Artuso, M., Gallerani, E., С W. & Blot, W.J., (1994) Linxian Nutrition tervention .P. (1996) й Abbondandolo, A., Montesano, R. & Wild, С Trials Study Group. Prevention of oesophageal cancer: Aflatoxin exposure and cytogenetic alterations in indi- the nutrition intervention trials in Linxian, China. viduals from The Gambia, West Africa. Mutation Res., Cancer Res. (suppl. 54), 20295-2031s 349, 209-217 Vahakangas, K.H., samet, J.М., Metcalf, R.A., Welsh, J.A., Minchin, R.F., Kadlubar, FE & I1ett, K.F. (1993) Role of Bennett, W.P., Lane, D.P. & Harris, C.C. (1992) Mutations acetylation in colorectal cancer. MuLot. Res., 290, 35-42 of p53 aid ras genes in radon-associated lung cancer Munoz, N., Hayashi, M., Lu, J.B., Wahrendorf J, Crespi, from uranium miners. Lancet, 339, 576-580 . (1987) Effect of riboflavin, retinol, and M. & Bosch, F.К Vineis, P., Bartsch, Ii., Caporaso, N., harrirsgton, AM., zinc on micronuclei of buccal mucosa and of esophagus: F.F., Landi, M.T., Malaveille, C., Shields, P.G., a randomized double-bind intervention study in China. Kadlubar, Skipper, P., Talaska, G. & Tannebaum, S.R. (1994) J. Nat! Cancer Inst., 79, 687-691 Genetically based N-acetyltransferase metabolic poly- Nakazawa, H., English, D., Randell, P.L., Nakasawa, K., morphism and low-level environmental exposure to car- Martel, N., Armstrong, B.К. & Yamasaki, H. (1994) UV and cinogens. Nature, 369, 154-156 l skin as skin cancer specific p53 gene mutation in потта Vbgelstein, B., Fearon, E.R., Hamilton, S.R., Kern, S.E., biologically relevant exposure. Proc. Nat! Acad. Sd., 91, Preisinger, A.C., I eppert, I., Nakamura, Y., White, R., Smits, 360-364 A.M. & Bos, J.L. (1988) Genetic alterations during colorec- Oda, T., Tsuda, l-L, Scarpa, A., Sakamoto, M. & Hirohashi, tal tumor development. N. Еngl. J. Med., 319, 525-532 S. (1992) X53 Gene mutation spectrum in hepatocellular Ziegler, A.M., Leffell, D.J. & Kunala, S. (1993) Mutation carcinoma. Cancer Res., 52, 6358-6364 hotspots due to sunlight in the p53 gene of пап - Ozturk, M., Bressac, B., Puisieux, A. & 26 other collabo теlапота skin cancers. Proc. Nat! Acad. Sd, USA, 90, rators (1991) p53 Mutation in hepatocellular carcinoma 4216-4220 after aflatoxin exposure. Lancet, 338, 1356-1359 Perera, F. & Santelia, R. (1993) Carcinogenesis. In: β Schulte, P.A. & Perera, F.P., eds, MoIeculа r Ерiдетгоlоgу. Principles and Practices. New York, Academic Press. pp. 277-300 Roberts-Thomson, I., Ryan, P., Khoo, K.K., Hart, W.J., McMichael, А.J. & Butler, R.N. (1996) Diet, acetylator phenotype and risk of colorectal neoplasia. Lancet 347, Corresponding author 13721374 A.J. McMichael Sheu, J.-C., hang, G.-T., Lee, P-R, Chung, J.-C., Chou, H.-C., Department of Epidemiology and Population Sciences, Wang, J.-T., Lee, H.-S., Shih, L.-N., Yang, R-M., Chen, D.-S. London School of Hygiene and Tropical Medicine, Keppel Street, London WIlE 7HT, UK & Wang, T.-H. (1992) Mutation of the p53 gene in hepa- toceBular carcinoma iii Taiwan. Cancer Res., 52,6098-6100

289 АрR1kabon of 6iоmaгkeгs in Cancer Epidemiology Toniolo, P., nEetto, Р. Ehuker, D.E.G., Rothman, N., Hulka, B. and Pearce, N., edt 1ARС Sciеnkfгo PuЫXеliсјггs No, 142 lnteroaliaml Agency for Research on Cancer, Lyon, 1997

The use of biomarkers to study pathogenesis and mechanisms of cancer: oesophagus and skin cancer as models

R. Montesano, Р. Hainaut and J. lai

Recent advances in molecular biology have made it possible to use genetic alterations associated with cancer as biomarkers to study the pathogenesis and mechanisms of cancer. However, the lessons that can be drawn from the analysis of alterations in a particular cancer gene are extremely dependent upon the biological context in which they arise. In this article, we discuss the biological significance of alterations in the p53 tumour suppressor gene in cancers of the oesophagus and of the skin. In both tissues, different forms of cancer occur at high frequency (squamous-cell carcinoma and adenocarcinoma in the oesophagus; squamous-cell carcinoma, basal-cell carcinoma and melanoma in the skin). We show that specific patterns of р53 alteration occur in these various cancers and that analysis of these alterations is useful to make inferences about the etiopathogenesis of cancers of the oesophagus and of the skin.

In the last decade, considerable progress has been Hainaut et al., 1996). Comparison between these made 0f the identification of the genetic alter- different cancers reveals that there is no fixed rule ations whose accumulation leads to the develop- on the nature and sequence of genetic alterations ment of cancer. The genes concerned can be in cancer development. At the same time, it is divided into three groups: oncogenes, tumour sup- apparent that some of the genetic alterations, in pressor genes and genes involved in DNA repair. particular р53 mutations frequently occur in dis- The genetic alterations found include point mu- tinct tumour types and provide valuable insights tations, deletions, translocations, allelic losses, not only into the etiology but also into the patho- amplifications and deregulation of gene expression. genesis and clinical management (diagnosis and The nature of the genetic alterations and the type prognosis) of some cancers. of gene modified vary among different tumours. In this respect, the relevance of genetic alter- Activating ris mutations, for instance, frequently ations in oesophageal cancer and in cancer of the occur in pancreatic carcinonias (Berrozpe et al., skin is discussed in more detail. 1994), but are absent in cancer of the oesophagus (Holstein et aL, 1988). The temporal sequence of Molecular pathogenesis of oesophageal cancer occurrence of genetic alterations may also differ Oesophageal cancers comprise squamous-cell car- between tumour types, with certain changes occur- cinoma and Barrett's adenocarcinoma, two types of ring at an early stage оf cancer development in one tumour with different etiology and pathogenesis. type of tumour and at a late stage in another type. squamous-cell carcinoma (SCC) has been associated This is shown schematically in Fig. 1 for colon can- with exposure to tobacco, betel chewing, alcohol cer and glioЫastoma. and various dietary components, with a large The relevance of genetic alterations in the etiol- geographical variation in incidence (Munoz & ogy and pathogenesis of lung and breast cancers Crite11wgué, 1994). Adenocarcinomas (ADCs) of has been discussed in recent reviews (Gadzar et al., the oesophagus develop from Barrett's oesophagus, 1994; Greenblatt etaL, 1994; Hulka & Stark, 1995; a condition in which normal squamous epithelium

291 Application of Baomarkers in Cancer Epidemiology

Earliest change Latest change

Inactiivation of Inactivation Inactivation Colon cancer one copy of Activating Y of both copies of huh copies ras mutation АPC gene of DCC gene of p53 gene

Inactivation Recessive Inactivation Recessive of both copies Amplification Glioblastoma chan g e оп ~ ц of both copies change on of y interferon of rb-8 chromosome 17 е of р53 gene chromosome 10 gene cluster

Figure 1. Temporal sequence of genetic events during the progression of colon cancer and of glioblastoma (for explanation, see text); redrawn from Murray & Hunt, 1993 is replaced by metap1astic columnar epithelium. mosomes where the Rb, APC/MCC and DCC genes This condition is the result of a prolonged gastro- are located. However, АPC1 MCC or DCC genes may oesophageal reflux, and these patients have a not be principal targets for loss of heterozygosity more than 100-fold higher risk than the general (LOI) on 5q and 18q. In particular, mutations in population of developing oesophageal ADC, a the APC gene are very rarely detected (2 out of 163 cancer that now accounts for approximately 50% tumours) (Aoki et al., 1994; Powell et al., 1994; of all oesophageal cancers in the USA (Blot et cL, Shibagaki et al., 1994). It is reasonable to assume 1993). that other genes with tumour suppressor activity may be located in these chromosomal regions. No Prevalence of allelic losses mutations were detected in the Rb gene in tumours Table 1 shows the allelic losses in oesophageal can- with or without 13q allelic loss, although the Rb cers (SCC and ADC) that occur with high fre- protein was found to be absent in a significant frac- quency (>40% of informative cases), based on mul- tion of oesophageal SCC. tiple and reproducible studies. Putative candidate Another frequent allelic loss in both SCC and genes localized in the chromosomal areas of allelic ADC involves chromosome 9р2l-22, the region in losses and the occurrence of mutations in these which the gene MTS-1 is located. This gene genes are also indicated (for review see Montesano encodes the p16 protein, a specific cycliri-dependent et cL, 1996, and references therein). kinase (CDK) inhibitor checking G1- to S-phase No significant differences were detected in the transition in the ceI1 cycle. This function of p16 is prevalence of allelic losses at various loci in the two lost when the ITS-i gene is mutated. However, it types of oesophageal cancer, SCC and ADC. Allelic is still unclear whether the MTS-I gene alone or loss of chromosome 17р occurs very frequently in other gene(s) present in the same chromosomal oesophageal cancer, as in many human solid region are the target of allelic deletion. tumours. In addition, there is a high prevalence of mutations in the p53 tumour suppressor gene Alterations of cell cycle regulatory proteins (localized at 17р13) in tumours that retain two 17р Genetic alterations leading to the constitutive acti- alleles (Huang et al., 1993; Maesawa et aL, 1994). vation of the CDK/cyclin D1 pathway appear to be This suggests thatp53 mutation is an early event in common in 5CC of the oesophagus. These altera- oesophageal carcinogenesis which occurs prior to tions occur at several distinct levels, including allelic deletion in l7p. (a) inactivation of the p1$/MTS-1 by diverse mecha- Allelic loss in oesophageal cancer also occurs nisms, including deletion and missense mutation; with high prevalence in chromosomes 13q, 5q (b) amplification and overexpression of the cycIin and, possibly, 18q, and in regions of these chrо- Dl gene (on chromosome 11g13); and (с) deletion

292 Pathogenesis and mechanisms of cancer and inactivation of the retinoblastoma susceptibility both exon 1 and 2 of the gene (Esteve et al., 1996). gene Rb (on chromosome 13g14). Amplification Iп other tumours, recent evidence suggests that and overexpression of the cyclin DI gene is the the apparent loss of p16 activity could also result most common of these genetic events (for review from aberrant or strongly decreased p1б/MТ5-1 see Montesano et al., 1996). gene expression. The p16/MТS-1 locus has a com- In a study of 50 oesophageal tumours of varying plex structure and encodes two transcripts with geographical origin, amplification and overexpres- distinct protein coding potential that are differen- sion of the cyclin D1 gene was found in 32% of the tially regulated during the cell cycle. Deregulation tumours, and loss of Rb protein expression in 17. of p16/ITS-1 activity may thus result from an The tumours with cycin DI alterations exibited imbalance between the levels of expression of the normal levels of Rb expression, while those that two transcripts (Stone et "L, 1995). It is also possi- did not express Rb did not show amplification or ble that downregulation of the protein rather than overexpression of cydin DI (Jiang et al., 1993). point mutation of the gene may be a common Ρ inactivation (Sun et ai., Similaг findings have been observed in other mechanism of рI6/MТ5-1 tumour types. 1995). Aberrant expression of p16/M S-1, cyclin D1 or Seveтal studies have described missense mu- Т tations of the pi 6/1ТЅ-1 gene, but the reported fre- Rb may have similar functional consequences: quencies differ widely. Recent evidence suggests inactivation of the suppressor function of Rb and ZF proteins by that the prevalence of pi 6/МТЅ-i mutations is in promotion of the activation of Е fact quite low (10%), and that mutations occur in direct or indirect mechanisms (Fig. 2). This model

Tumours with aifelic loss! informative tumours (%) Chromosome sCC ADC Minimal area of loss Candidate genes Mutations hiLM-i Эр 25152 Зр21.3 (40-64)

5q 50194 34151 5g21.2 AРСIМСС . Very rare (36-80) (63-75)

5-i, !FNA 0 <>40 9р 55195 8117. 9р22 МТ (45-65) (47)

9q 11118 9g31 Е3s1 (60)

13q 66/134 28167 13g14.1 Rb No mutations (41-54) (36-43)

Nigh V/p 651124 47154 i7р13 p53 . . (43-65) (72-94)

17q 56191 10118 17g21.3 BCRA1, eгbB2, (62) (56) 17g11.2-q12 С 5F-3, NF-1, 1T84

18q 26179 (24) 18g23.3 DCC Rare . . (23-46)

SCC, squamous-cell carcinoma; ADC, adenocarcinoma

293 Application of Biomarkers in Cancer Epidemiology

Mutation spectrum of the р53 gene Among 240 cases of oesophageal ЅСС analysed in сусliп b 1 the literature, 110(45.8%) have been found to contain mutations that have been confirmed and р 16/MTSi identified by DNA sequencing. The prevalence of р53 mutations is even higher for ADC, with 46 cases of mutant р53 identified in 64 patients screened (71.8%) (see Montesano et al., 1996). дь phospho Rb 5CC and ADC of the oesophagus differ strikingly in the pattern of mutations inр53 (Fig. 3). р53 mutations in ADC show a very high frequency of transition at CpG dinucleotides (--63%). To date, this is the highest level of CpG transition found in any E2F S-phase cancer type (other cancer types with frequent CpG electors transitions are colon carcinoma and pancreatic cancer, with, respectively, - 46 and -36%). G:C to T:A transversions and mutations at A:T base pairs are comparatively rare (taken together, -14%). In contrast, in 5CC, CpG transitions aie less frequent Figure 2. Interplay between the products of the cyclin Di, than in most other tumour types (-18%), whereas i6/MTS-i and relnoblastoma susceptibility genes (Rb) in the р mutations at A:T base pairs account for 31% of all regulation of cell cycle progression from Gi to Ѕ. Cell progression from Gi to S requires, among other things, activation of specific mutations. The spectrum of p53 mutations in 5CC CDKs (CDК4 and CD Кб) in association with cyclin Di. The is thus indicative of the involvement of exogenous active CDK/cyclin D1 complex phosphoryfates sequentially the carcinogens, in agreement with epidemiological Rb prolein, thus releasing Rb-bound transcription factors of the data supporting the role of environmental agents, E2F family. Free E2Fs transactivate genes that are essential for in particular tobacco, nutritional components and entry into S апд DNA replication (S-phase effectors). CDK/cyclin Di activity is negatively regulated by binding of several cyclin alcohol. The high frequency of mutations at A:T kinases inhibitors, including p16/ITS-i. p16/MTS-1 is thought base pairs may reflect enhanced depurination of to be activated in response to growth control stimuli through DNA upon reaction of carcinogens with adenine pathways which are not fully elucidated. and/or exposure to DNA-reactive agents such as acetaldehyde, a metabolite of ethanol. The nature provides a paradigm explaining why tumours with of р53 mutations in ADC is more difficult to inter- indistinguishable biological characteristics show pret. Transitions at CpG dinudeotides are generally heterogeneity at the causative genetic level. considered as the hallmark of mutations occurring The high frequency of genetic aberrations in spontaneously by hydrolytic deamination of 5- pathways controlling G1/S transition may reflect a methylcytosine. Recent evidence indicates that the requirement to abrogate physiological G1/S con- mutability of CpG dinucleotides might also result trol to allow oesophageal cells to progress towards from enzymatic deamination and methylation by malignancy. Indeed, the homeostasis of the methyltransferases, which bind with high affinity oesophageal mucosa, as is the case with al surface to the premutagenlc DNA mismatches G:U and epithelia, is dependent upon a delicate equilibrium G:T, thereby preventing their efficient repair by between cell renewal, differentiation and death. mismatch repair proteins (Shen et aL, 1995; Yang Switching off any of the molecular controls bal et aI., 1995). Further studies are needed to identify anti g these three processes may compromise the and characterize such mechanisms in human cells normal life cycle of mucosal cells. Escape from and their possible role in the genesis of p53 mu- growth control, and reduced rates of cell differen- tations in ADC. tiation and cell death are critical steps towards the In many areas of the world, tobacco and alcohol acquisition of further genetic lésions and func- have been identified as major factors cooperating tional alterations that lead to the expression of in the etiopathogenesis of 5CC of the oesophagus. malignant phenotypes. Among 91 5CC patients for whom reLiable data on

294 Pathogenesis and mechanisms of cancer exposure to tobacco and/or alcohol are available, the distribution of p53 mutations reveals a strong 70 relationship with tobacco smoking (Montesano 60 et al., 1996). Only 20% of non-smoker 5CC patients ô 50 have 53 mutations, in contrast with 80% in р v 40 C patients who smoked more than 20 cigarettes/day. Ф Even lighter smoking (less than 20 cigarettes/day) 6 30 ш increases the p53 mutation frequency to a value up ј 20 to 50%. The relationship with alcohol consump- 10 tion is less clear. However, the relative impact of each risk factor is difficult to assess, as most of the GC:AT GC:AT GO:TA GC:TA A:T Ins/del patients with p53 mutations were exposed to both CpG tobacco and alcohol, and the contribution of other ED soc less well-defined risk factors is unknown. ADC Molecular pathogenesis of skin cancer A causal association between UV exposure, defective Figure 3. Spectrum of p53 mutations in squamous-cell carci- repair of UV-induced DNA photoproducts and noma (SCC) and adenocarcinoma (ADC) of the oesophagus. non-melanocytic skin cancer has long been postu- Data from the p53 mutation database (Holistein etas, 1996) are lated. This cancer is classified into two major his- expressed as a percentage of the total number of mutations tological types: basal-cell carcinoma (BCC) and found in SCC or in ADC. squamous-cell carcinoma (SCC), the former being the came from the work of Cleaver showing that cells commoner type in white populations. Descriptive from patients with xeroderma pigmentosum (XP), a studies in whites in North America, Australia and cancer-prone inherited disorder, are defective in the several other countries have also shown a positive excision repair of UV-induced pyrimidine dimers association between incidence and mortality from (for review, see Bootsma & Hoeijmakers, 1994). melanoma anд residence at lower latitudes. Studies of migrants suggest that the risk of melanoma is Alteration of DNS repair and risk of basal- and related to solar radiant exposure at the place of squamous-cell carcinoma residence in early life, and the results from Mammakan cells have a variety of repair mechanisms case—control studies are generally consistent with that will remove the DNA lesions produced by UV positive associations with residence in sunny irradiation. The pathway involved in the repair of environments throughout life, in early life and the commonest UV-induced DNA adducts by UVB even for short periods in early adult life and with and UVC is via the nucleotide excision repair measurements of cumulative sun damage. (NER) pathway. The molecular basis of this repair UVA, UVB aid UVC radiation produce measur- pathway, which eliminates a remarkable number of able DNA damage in human skin cells in vivo at different, structurally unrelated lesions, involves at doses commonly experienced by humans. Of the Ieast five distinct steps: damage recognition; mci- DNA photoproducts, the cyclobutarve-type pyrim- sian of the damaged strand on both sides of the idine dimers and the pyrimidine-pyrimidone 6-4 lesion at some distance from the lesion leaving the photoproducts have been shown to be cytotoxic non-damaged strand intact; removing the damage- and mutagenic in human cells, although the rela- containing oligorner; gap-filling DNA synthesis; and tive contribution of each type remains to be fully sealing of the DNA by DNA Iigase. In human cells, elucidated. Exposure to UV radiation has been it would appear that two NER subpathways are shown to increase the expression of various cellular operating, one dealing with the rapid and efficient genes and to have profound effects on the im- removal of lesions blocking ongoing transcription munological system which may contribute to the (transcription-coupled repair), and the second development of skin cancer. Evidence for the resulting in the slower and less-efficient removal of involvement of DNA photoproducts and their bulk DNA including the non-transcribed strand of repair in human skin carcinogenesis originally active genes (genome overall repair).

295 Application of Biomarkers in Cancer Epidemiology

1ndividuals lacking components of the NER subjects with family history, the P-value was pathway have been found to have increased sun 0,047). They also found that DRC declined with sensitivity and, in some cases, increased risk of age in both cases and controls (Wei et al., 1993). skin cancer. This is observed in XP patients who In a second study by Hall et al. (1994), the DRC show sun sensitivity and increased cancer risk in was measured in subjects involved in a population- addition to cutaneous manifestations, including based study of the incidence and prevalence neurological degeneration and pigmentation of non-metanocytic skin cancer in Geraldton, abnormalities. Western Australia. Subjects with SCC and BCC Within the XP phenotype, large variations in were considered separately, and for each case one repair activity are noted (ranging from 2 to 80% of control was chosen, matched by age and sex. The 'normal' levels). If КP is considered to represent the mean levels of CAT activity in cells transfected lower range of repair capabilities, those individuals with plasmids irradiated at 700 J/mz were found to expressing a reduced repair response may be at an be generally higher in cases than in controls, and increased risk of cancer. It could be imagined that, there was little difference between cases and con- due to the multiprotein complexes involved in the trois in the DRC capacity according to the UV. The NER pathway, minor alterations in protein struc- DRC showed little association with age, sex and ture might produce dramatic alterations in protein— viability of the lymphocytes, but was positively protein interactions aid subsequent alterations in correlated with blastogenic rate (P = 0.055). Thus, the DNA-repair capacity (DRC). Сие to the multi- no evidence was found that subjects with non- step nature of the NER pathway, an assay that melanocytic skin cancer had lower repair capaci- reflects the overall DRC and that can be carried out ties than the controls, nor was any statistically on easily available biological material, such as significant difference in repair activity detected lymphocytes, would be ideal for population-based between the two groups. Only in the young substudies to assess the role of DRC in modulating jects with SСС was there any suggestion that cases cancer risk. 5uсh an assay was developed by had lower repair capacities, although this differ- Grossman and his collaborators (Athas et аl., 1990) ence could be due to chance (P = 0.53). and has subsequently been used in two published The differences between the results of these two studies designed to evaluate the role of DRC as a studies may be explained by several factors. Firstly, risk factor for non-melanocytic skin cancer. In this the age range of the subjects in the Baltimore study assay, blood lymphocytes from subjects are cul- was larger than that in the Australian study, with tured and transfected with either control or UV- 21.6% of the BCC cases under the age of 40 at the irradiated plasmids containing a reporter gene (the time of sample collection. As the apparent differ- chlorаmphenicol-acеtyltrаnsferase gene CAT), and ence in DRC was most marked in early-onset cases, the repair capacity is determined by measuring the possibility of detecting a difference between CAT gene expression in protein extracts prepared cases and controls might be reduced due to the from transfected cells. older age of the Australian population. A second Grossman and his collaborators applied the major difference was in the mean CAT activity assay in a study population in Baltimore, Maryland. measured iп the assay, which was substantially This population was made up of Caucasians (aged higher in the Australian study (13% in BCC cases, 20-60 years) who had lived in Baltimore or its sub- 5D = 6.2%; 12.2% in SCC cases SD = 7.1%; and 12% urbs for most of their lives. The DItC was measured in controls, SD = 5.6%). This may indicate a greater in 88 patients with a history of one or more histo- degree of random error in the laboratory measure- logically confirmed BCCs, and in 135 control ments. The methods used in the studies were not subi ects. The mean CAT activity with plasmids UV- identical, mainly for logistic reasons. The handling of irradiated at 700 J1m2 was 8.00% (5D = 2.2) in con- the blood samples is one example. Iп the Australian trols without family history of actinic keratosis, study, lymphocyte purification could only be car- 7.28% (SD = 2.2) in controls with a family history ried out after overnight transport of the samples of BCC or a previous history of actinic keratosis, whereas Wei et al. (1993) processed the samples and 7.35% (SD = 2.0) in cancer cases (Л = 0.103; locally on the same day. Since subsequent cell however, when cancer cases were compared with viability was lower in the Australian study, the pis-

296 Pathogenesis and mechanisms of cancer siЫlitу that the surviving cells were not truly Prevalence of alelic loss representative of the individual cannot be The pattern of chromosome loss has been studied in excluded. skin tumours from patients with 11CC, SCC, melan- Another possible explanation relates to the oma and Ferguson-Smith syndrome (a syndrome higher ambient level of solar radiation iп Western in which the lesions are histologically indisting- Australia. Exposure to UV radiation has been uishable from SCCs) using microsatellite markers. shown to increase the expression of various genes. A high frequency of chromosome 9q loss (up to Short-term effects of UV on the ability to repair 6% in informative tumours) has been observed in radiation-induced DNA damage have been noted, BCC, within a region where both the naevoid 11CC but little is known about the long-term effects. It is syndrome and Ferguson-Smith syndrome have also possible that the result of high environmental been mapped (9g22.3-q31). This suggests that the exposures to UV radiation is to mask the DRC dif- two conditions may reflect mutations in the same ferences between cases and controls obsérved by gene. A relatively high frequency of LOI (14%) Wei et ad. (1993). Recent epidemiological evidence was also observed for chromosome 1q, suggesting suggests that beyond a certain sun exposure, the that this region may play a role in the develop- risk of BCC does not increase further and that a ment of BCC. particular amount of sun exposure delivered in In sCC, loss of chromosome 9 has been found infrequent, probably intense, increments will in up to 33% of tumours (16149). However, the pat- increase the risk of BCC. This finding may support tern of chromosomal alterations in SCC is different the possibility that in cases of extreme sun expo- from that observed in BCC, with LOI at 9p (41%) sure, DRC is no longer a major risk factor for 11CC. 13q (46%), 17р (33%) and Зp (23%) (Quinn et al., In summary, the use of the measurement of 1994). In primary melanomas, LOI at 9p was the DRC in cells to identify individuals at increased most frequent (47% of informative cases), and it risk of UV carcinogenesis and to improve the esti- has been suggested that a melanoma susceptibility mation of the effects of sun exposure is appealing. or tumour suppressor gene may be localized The assay developed by Grossman and colleagues at 9p2l-23, a region that contains the cyclin- is potentially applicable to molecular epidemiol- dependent kinase inhibitor gene р16/M7`S-1. LOI ogy. However, its use in such studies has, to date, on chromosome arms Зp, бq, 10q, 11q and 17p produced conflicting data, which may reflect dif- was also relatively frequent. LOI at Зp and 10q ferences in experimental approach as well as dif- was found in lesions of 4.5 rum in depth or less, ferences between the populations studied. while LOI at бq, 11q aid 17p was only detected There is limited evidence, apart from that com- in more invasive tumouis. These results suggest that ing from XP patients and that obtained by Wei et loss of chromosome 9p occurs before loss of other a2. (1993), that DRC is altered in patients with skin chromosome arms in sporadic cutaneous melanoma. cancer. A number of studies have examined the ability to repair pyrimidine dhners induced by Activatim of oncogenes single exposure to stimulated solar radiation, A high variability in ras gene mutations has been although many of these studies are limited by reported in melanomas (from 5 to 27%) and may small populations (IARC, 1992). The impact of be explained by the results of Jafari et aL (1995). reduced DRC as a risk factor for BCC is potentially They demonstrated that the predominate mu- significant, especially when the occurrence of tation was at codon 61 of the N-ras gene and exclu- increased ambient UV irradiation, due to alter- sively in modular malignant melanomas (31% ations in ozone levels, is taken into consideration. mutation frequency). No ras mutations were re- However, in order to fully assess the contribution ported in superficial spreading melanomas or in of DRC to skin cancer susceptibility, additional lentigo malignant melanomas, demonstrating the studies need to be completed in which the effects need for consideration of the specific pathology of age, family history and sun exposure can be when carrying out this type of analysis. Whether evaluated and linked to other important factors in ras gene mutations play an important role in non- the carcinogenic process, such as the mutagenic melanocytic skin cancer remains to be fully estab- alterations in cellular DNA. lished. Spencer et ai. (1995) have found activating

297

Application of Biomarkers in Cancer Epidemiology

mutations at codon 12 of K-ras and at codons 12, р53 mutations in sCC and BCC to be between 12 and 13 and 61 of the H-ras gene in 16% of actinic ker- 58%, and the analysis of the mutations has revealed atoses and in 12% of SCCs. Daya-Grosjean et ai. thé presence of UV- specific mutations targeted to (1993) have shown that there 1s a twofold increase dipyrimidfne sequences, with the majority being in mutation frequency of the ras genes in XP C to T transitions. Ziegler et al. (1994) have demon- tumours compared with control tumours. The strated that UV-induced p53 mutations are already majority of these mutations were at codon 12 in all detectable in actinic keratosis, with mutations three ras genes, and were located opposite dipyrim- being detected in 60% of 45 cases of actinic kera- idine sequences. This is consistent with a funda- tosis examined from 24 patients. The base changes mental role of unrepabed DNA damage as an initi- found implicated sunlight as the mutagen: 89% ating lesion in skin carcinogenesis. occurred at adjacent pyrimidines and most were C to T substitutions or CC to TT double base changes. Mutation spectrum of р53 gene With the exception of cutaneous melanoma, skin Jmrnunological response cancer is often associated with р53 mutations. Over the past few years, it has been realized that Recent studies have estimated the incidence of exposing the skin to sunlight has profound effects

Normal _ cpо5еd.` HyperplaStic:: Dÿsplastic :: Cancer

Tobaccolalcohol р53 mutations Basal-cell SCC others (A or T) hyperplasia

lesophageal Barrett's р53 overexpression р53 mutations ADS mucosa metaрlasia (CT CPG)

Chronic reflux

р53 sun, UVB mutations Sun, UVB Actinic I SCC (СС -ТТ) \ keratosis Skin jl DNA cyclobutane genetic alteration dimers

Figure 4. 53 mutations in the temporal sequences of events leading to cancer of oesophagus [squamous-cell carcinoma (5 р СС ) and adеnocarсiпoma (ADС )] or of the skin. The progression of the disease is represented from left (normal tissue) to right (cancer). The occurrence of genetic lesions is shown with respect to specific pathological stages. This model illustrates that р53 alterations occur at different stages and in different contexts in both cancer of the oesophagus and cancer of the skin. In S С C of the oesophagus, mutation of р53 is thought to be a very early event that occurs in exposed tissues prior to the formation of a dysplastic lesion. In contrast, in AUC, р53 mutations occur later in the development of the lesion, probably in late dysplasia. In SСС of the skin, mutations occur in normal cells, but the propagation of a clone of cells carrying a р53 mutation is dependent upon continuous exposure to UV

298 Pathogenesis and mechanisms of cancer on the immunological system and that these induced by chronic gastric reflux. Subsequent inac- immunological changes can contribute to the tivation of p53 by mutation and/or allelic result is development of skin cancer and alter resistance to an abrogation of cell cycle control at the G1/S tran- infectious diseases. Studies on the nature and sition. Consequently, subpopulations of aneuploid mechanism of the immunological alterations fol- cells frequently develop during the later stages of lowing exposure to UV radiation suggest that UV- carcinogenesis, with increased proportion of cells induced DNA damage triggers а cascade of events in the S-phase and G2-phase. Mutations in the X53 leading to a state of antigen-specific, systemic T- gene are more common in ADC than in 5CC, but lymphocyte-mediated immunosuppression. It is are not found in non-dysplastic Barrett's metaplasia, proposed that with low doses of UV radiation, anti- suggesting that they occur later at the transition gen-presenting cells in the. irradiated skin are mod- from high-grade dysplasia to carcinoma. ified, either directly by the action of UV or lidi- These findings at the cellular and molecular rectly as a consequence of an influx of inflamma- levels underscore the different etiology and patho- tory cells or by the local release of cytokines. Larger genesis of SCC compared with ADC, and suggest doses of UV may bring about the release of that the genetic alterations observed may represent cytokines systemically, causing an alteration tri the molecular fingerprints of critical risk factors activity of antigen-presenting cells. Cytokine involved in the development of these two cancers. imbalances in individuals unable to repair UV- induced DNA damage may explain some of the Cancer of the skin. The model for human skin car- additional clinical symptoms observed in XP and cinogenesis presents some uncertainties as far as other sun-sensitive syndromes. the progression of genetic events is concerned. A role for p53 has been clearly demonstrated by Conclusion Ziegler et aL (1994). p53 mutations are found in Temporal sequence of genetic alterations in the actinic keratosis, and inactivation of X53 in mouse progression of cancer skin reduces the appearance of sunburn cells, Figure 4 summarizes the role of p53 mutation in apoptotic keratinocytes generated by overexposure the temporal sequence of events leading to cancers to UV. Skin thus appears to have a p53-dependent of the oesophagus or to cancer of the skin. mechanism which removes `precancerous' UV- damaged cells via apoptosis. If this response is Cancer of the oesophagus. In SCC, р53 overexpres- reduced or altered by p53 mutation, sunburn sin or accumulation of the X53 protein due to might act as a selection pressure favouring the sur- prolonged half-life has been detected in normal, vival of mutated cells. Whereas UV-damaged nor- non-dysplastic oesophageal cells and, iп particular, mal cells will die as sunburn cells, a proportion of in the nuclei of proliferative cells in the basal layer the damaged cells with mutant p53 will be resis- of the mucosa. In a number of cases, p53 muta- tant to apoptosis. Continued sun exposure, which tions are already present in the p53-positive cells in produces alterations in immunological surveil- normal mucosa. Moreover, in a given patient, mul- lance and gene expression, can act as a tumour tiple and distinct patches of p53-positive cells with promoter and results in the clonal expansion of different mutations can be detected. This indicates mutated cells to form an actinic keratosis which that multiple independent pre-neoplastic foci or usually regresses in the absence of sun exposure tumours develop in the oesophageal squamous but progressively enlarges if exposure continues. mucosa, and it is of interest to elucidate which of these foci have selective growth advantage in can- Importance of biological context in interpreting the cer development. During tumour development, significance of bromarkers frequent additional genetic alterations include The two examples discussed here indicate that the 3g21.3 and 9g31 allelic losses. detailed analysis of specific biomarkers may reveal In ADC, overexpression of p53 protein and important information for the understanding increased proportions of cells with a GO/G1 DNA of both the etiology and pathogenesis of human content are detected in metaplastic Barrett's cancer. In particular, the analysis of p53 gene mu- epithelium, possibly as a result of DNA damage tation and protein expression is useful to make

299 Application of Biomarkers in Cancer Epidemiology

inferences about the nature of the etiologic agents of skin tumors isolated from DNA repair deficient involved in these cancers. In addition, in both xeroderma pigmentosum patients. Cancer Res., 53, 1625-1629 oesophagus and skin cancers, alteration of р53 Esteve, A., Martel- appears to represent а central event in tumour pro- Planche, G., Sy11a, B.S., Hollstein, M. gression. However, the interpretation of the signif- Hainaut, P. & Montesano, R. (1996) Low frequency of icance of p53 alteration as a blomarker is depen- р16/CDKN2 gene mutations in esophageal carcinomas. dent upon the biological context. In cancers of the lit. J. Cancer, 66, 301--304 oesophagus, the main consequence of p53 alter- Gadzar, A.F., Bader, 5., lung, J., Kishimoto, Y., Sekido, Y., ations is likely to be the disruption of G1/S cell sugio, K., ViriTiani, A., Fleming, J., Carbone, D.P. & cycle control, giving rise to populations of actively Minna, J.D. (1994) Molecular genetic changes found in growing cells at high risk of becoming precursors human lung cancer and its precursor lesions. In: The Molecular Genetics of Cancer, Vol. 59, Cold Spring Harbor, to neoplastic lesions. In skin cancers, loss of USA, Cold Spring Harbor Laboratory Press, pp. 565-572 р53 function may abrogate a normal mechanism which removes UV-altered cells via apoptoses, thus Greenblatt, M.S., Bennett, W.P., lollstein, M. & Harris, facilitating the survival of cens containing UV- С .C. (1994) Mutations in the p53 tumour suppressor gene: clues to cancer etiology and molecular pathogene- damaged DNA. Clonai expansion of mutated cells sis. Cancer Res., 54, 4855-4878 depends upon continuous exposure to the sun until further genetic alterations give rise to squa- Hall, J., English, D.R., Artuso, M., Armstrong, ~.K. & mous cell carcinoma. These two examples empha- Winter, M. (1994) DNA repair capacity as a risk factor for non-melanocytic skin cancer—a molecular epidemiolog- size that comparable genetic events may have dif - ical study. Int. J. Cancer, 58, 179-184 ferent impacts depending upon the tissue, cellular and molecular context in which they arise. Hollstein, M.C., Smits, A.M., Galiana, C., Yamasaki, H., Bos, J.L., Mandard, A., Partensky, C. & Montesano, R. (1988) Amplification of EGF receptor gene but no evi- Acknowledgements dence of ras mutations in primary esophageal cancers. We gratefully acknowledge the financial support Cancer Res., 48, 5119-5123 received from the Commission of the European Communities (grant nos EV5V 92-0199 and CT Hollstein, M., Shomer, B., Greenblatt, M., Soussi, T., СТ Hovig, E., Montesano, R. & Harris, C.C. (1996) Somatic 94-0581). We thank Mrs Magali Maillol for typing point mutations in the p53 gene of human tumors and the manuscript. cell lines: updated compilation. Nucleic Acids Res., 24, 141-146 References Huang, Y., Meltzer, S.J., Yin, J., Tong, Y., Chang, E. ., Aoki, T., Mori, T., Du, X., Nisihira, T., Matsubara, 1: & Н Srivastava, S., McDaniel, T., Boynton, R.F. & Zou, Z.Q. Nakamura, Y. (1994) Allelotype study of esophageal car- (1993) Altered messenger RNA and unique mutational dnoma. Genes mosom. Cancer, 10, 177-182 Сбго profiles of p53 and Rb in human esophageal carcinomas. Athas, W.F., Hedayati, M.A., Matanoski, G.M., Farmer, Cancer Res., 53, 1889-1894 E.R. & Grossman, L. (1990) Developement and field-test Hulka, B.S. & Stark, A.T. (1995) Breast cancer: cause and validation of an assay for DNA repair in circulating lym- prevention. Lancet, 346, 883-87 phocytes. Cancer Res., 51, 5786-5793 IARC (1992) IARC Monograph on the Evaluation of Berrozpe, G., Schaeffer, J., Peinado, M.A., Real, F.X. & Carcinogenic Risks to Humans: Solar and UItraviolet Pemcho, M. (1994) Comparative analysis of mutations Radiation, Vol. 55, Lyon, international Agency for in the p53 aid K-ras genes in pancreatic cancer. Int. J. Research on Cancer Cancer, 58, 185-91 Jafari, M., Papp, T., Kirchner, 5., Diener, U., HenschleL D., Blot, W.J., Devesa, S.S. & Fraumeni, J.F. (1993) Burg, G. & Schiffmannn, D. (1995) Analysis of ras muta- Continuing climb in rates of esophageal adenocarci- tions in human melanocytic lesions: activation of the ras noma: an update [letter]. J. Am. Med. Assoc., 270, 1320 gene seems to be associated with the nodular type of Bootsma, D. & Hoeijmakers, J.H J. (1994) The molecular human malignant melanoma. J. Cancer Res. Clin. OncoL, basis of nucleotide excision repair syndromes. Mutation 121, 23-30 Res., 307, 15-23 Jiang, W., Zhang, Y.J., Kahn, 5.1., Hollstein, M.C., Daya-Grosjean, L., Robert, C., Drougard, C., Suarez, H. & Santella, R.М., Lu, S.H., Harris, C.C., Montesano, R. & Sarasin, A. (1993) High mutation frequency in ras genes Weinstein, LB. (1993) Altered expression of the cyclin

300 Pathogenesis and mechanisms of cancer

D1 and retinoblastoma genes й human esophageal can- Stone, S., Jiang, P., Dayanantk, P., Tavtigian, S.V., cer. Proc. Natl Асад. 5e!., USA., 90, 9026-9030 Katcher, H., Parry, D., Peters, G. & Kamb, A. (1995) Maesawa, C., Tamura, G., Suzuki, Y., Ogasawara, S., Ishida, Complex structure and regulation of the P16 (MTS1) K., Saito, K. & Satodate, R. (1994) Aberrations of tumor- locus. Cancer Res., 55, 2988-2994 suppressor genes (p53, арс, mcc and Rb) in esophageal Sun, Y., Hildesheim, A., Lanier, A.E., Cao, Y., Yao, K.T., squamous-cell carcinoma. fit. J. Caner., 57, 21-25 Raab, Traub, N. & Yang, C.S. (1995) Ni point mutation Montesano, R., Holistein, M. & Hainaut, P. (1996) but decreased expression of the p16IMTS1 tumor suppressor gene in nasopharyngeal carcinomas. Genetic alterations i oesophageal cancer and its rele- Oncogene, п 10, 785-788 vance to etiology and pathogenesis: a review. Int. J. Cancer Fred. OncoL, 69, 225-235 Wei, Q., Matanoski, G.M., Farmer, E.R., Hedayati, M.A. & Grossman, L. (1993) DNA repair and aging in basal cell Munoz, N. & C stellsagué, X. (1994) Epidemiology of а carcinoma: a molecular epidemiology study. Proc. Nat[ oesophageal cancer. Eur. J. Gas troenterol, t ., 6, Нерцы Acad. 5 i., 90, 1614-1618 649-655 с ., Zingg, J.M,, Mi, , . Murray, A. & Hunt, T. (1993) The Cell Cycle, Oxford, Yang, A.S., Shen, J.С S. & Jones, РА (1995) HhaI and aII DNA methyltransferases bind Oxford University Press Нр DNA mismatches, methylate uracil and block DNA Powell, S.М., Papadopoulos, N., Kinzler, K.W., repair. Nucleic Acids Res., 23, 1380-1387 Sinoldnski, K.N. & Meltzer, S.J. (1994) APC gene muta- Ziegler, AM., Jonason, A.S., Leffell, D.J., Simon, J.A., tions in the mutation cluster region are rare in Sharma, H.W., KimmeIman, J., Remington, L., Jacks, T. & esophageal cancers. Gastroenterology, 107, 1759-1763 Brash, D.E. (1994) Sunburn and skin cancer in the onset Quinn, А.G., Sikkir'k, S. & Rees, J.L. (1994) Basal cell car- of skin cancer. Nature, 372, 773-775 cinoma and squamous cell carcinoma of human skin show distinct patterns of chromosome loss. Cancer Res., 54, 4756-4759

Shen, J.C., Zingg, J.М., Yang, A.S., Schmuttе, C. & Jones, PA. (1995) А mutant Нра11 methyltransferase functions as a mutatoI enzyme. Nuclek Acids Res., 23, 4275-4282 5hibagaki, I., 5himada, Y., Wagata, T., Ikenaga, M., Imaroura, M, & Ishizaki, K. (1994) Allelotype analysis of Corresponding. author esophageal squamous cell carcinoma. Cancer Res., 54, R. Montesano 2996-3000 Unit of Mechanisms of Carcinogeпésis, Spencer, J.M., Kahn, S.M., Jiang, W., DeLeo, V.A. & International Agency for Research on Cancer, Weinstein, I.B. (1995) Activated ras genes occur in 150 cours Albert Thomas, 69372 Lyon Cedex 08, France human actinic keratoses, premalignant precursors to squamous cell carcinomas. Arch. Dermatol., 131, 796-800

301 Application of Biomarkero iri Cancer Epiбemiology Toniolа, Р., BoPefte, R, 5huker, D.E.G., Rothman, N., Huike, B. and Pearno, N., edo ]ARC 5cienhilic PuЫications Nc. 142 1r,Lerriakcna1 Agency for Research on Cancer, Lyon, 1997

Comparing measurements of biomarkers with other measurements of exposure

R. Saracci

The issue of the relative merit of biomarkers and alternative measures of exposure arises most commonly in the context of epidemiological studies aimed at hazard detection and quantification. When exposures are from biological agents, biomarkers are usually the first and often the only justifiabIe choice. In general, however, the relative merit of different types of exposure measurements need to be evaluated on a case-by-case basis. Biomarkers may be affected by random errors, time-related sampling errors, physiological confounding and disease- induced differential error, all of which need to be explicitly evaluated before embarking on the use of a biomarker in a full-scale epidemiological study. Random errors affecting biomarkers may be reduced by replication or combination of measurements, or both. Alternative measurements of exposure can be evaluated against a biomarker when there is adequate evidence for regarding the marker as the true measure of a biologically relevant exposure.

Biomarker measurements: for what purpose? of biological variables; and measurements of vari The term exposure' can have, and in the past has ous kinds extractable from existing records. had, diverse meanings, ranging from very narrow When comparing the relative merits of biomarker to very broad: in epidemiology it has become cus- measurements and other types of measurement, the tomary to regard an exposure as 'any of a subject's following four purposes for which biomarker mea- attributes or any agent with which he or she may surements are performed should be considered: come in contact that may be relevant to his or her health' (Armstrong et a1., 1992, p. 4). This all- Control of exposure level embracing definition incorporates variables exter- Biomonitoring is a long established practice, essen- nal to the human body (e.g. concentration of an tially within the industrial setting, which acts as a agent in air, social class, a drug), as well as variables complement or an alternative to environmental internal to the organism (e.g. sex, blood level of a measurements. In this application, a biomarker is hormone, a gene), the key feature common to both regarded (based on evidence or assumptions, or kinds of variable being that they are considered as both) as a biologically relevant effect of external potential causes of biological and health effects. In agents, and research usually centres on the tech- the area of exposure measurement for epidemio- nology required to achieve the desired level of con- logical research, the fastest growing sector is repre- trol of the agent in the environment. To the sented by biomarker measurements, underpinned extent, however, that the biological relevance of by current developments in biology, particularly at the marker is not yet adequately established—at cellular, subce11ular and molecular levels. In prind. least for the specific purpose of exposure monitor- plc, measurements of biomarkers using up-to-date ing—the same issues arise as are considered in the techniques can potentially favourably replace next paragraph. many other types of measurement, in particular direct (physical, chemical or biological) or indirect Hazard detection and risk quantification. (questionnaire and record) measurements of envi- For this research purpose, a genuine question arises ronmental variables; questionnaire measurements as to whether a biomarker or another type of mea-

303 Application of Biomarkers in Cancer Epidemiology

surement is preferable. A biomarker measurement to disease—obviously for these studies there are no may prove superior in six respects: (1) individual- alternatives to the use of biomarkers measure- ization, as the measurement is now performed on ments. A key consideration, however, is that epi- the subjects in the study or on biological speci- demiological studies can only be—in general and mens taken from them; (2) objectivity, to the extent exceptions apart—ancillary tools for pathogenesis that the measurement is independent of the studies, not because of a lack of pertinent bio- observed subjects' perceptions and substantially markers but because of their obseivational nature independent of the observer if instrumental or lab- within the complex settings of human life. oratory methods are used; (3) quantitative speciticity Mechanistic insights are gained, for instance, by to the exposure of interest, to the extent that the analysing in human tissue specimens the genetic method of measurement responds only to the changes characteristic of the different histological agent of interest with no, or a minimal, response to conditions of the colonic mucosa (normal tissue, other agents; (4) quantitative sensitivity, to the adenoma, carcinoma). However, to elucidate and extent that the method of measurement responds demonstrate the causal succession and intercon- in a quantitative way to the agent down to a very nection of genetic and non-genetic events requires low limit of detection (meaning by 'very low' a exploration and testing in systems accessible to concentration well below the minimum likely to direct experiment within simplified and controlled induce a detectable biological effect); (5) Ыolо iсal settings, e.g. animal models like transgenic and sigпi!lсапсe, to the extent that the biomarker mea- knockout mice. In the same way that descriptive surement may not only measure exposure but also epidemiology data provide etiological insights and contribute at least some information on the mech- suggestions to be fully explored and tested in ana- anisms through which an agent may induce a dis- lytical epidemiological studies, epidemiology in ease,, adding plausibility to an observed expo- general can provide mechanistic insights and sug- sure-response relationship; (6) cost, to the extent gestions to be fully analysed and established in that biomarker measurements, especially when experimental studies. performed with automated systems capable of combining high accuracy aid precision with low Prognostic studies. cost per unit measurement, can be less expensive In prognostic studies of outcome, the population than alternatives. The worth of a biological marker investigated is composed of patients rather than measurement in these respects, taken together, `healthy' people. Again, as a general rule, biomarker may make it pтeferaЫe to the alternatives, in that measurements dominate the field and questions of it is better capable (a11 other things being equal) of alternatives seldom arise. In addition to measure- providing a higher estimate of the risk associated ments of biomarkers, however, other kinds of mea- with an exposure and a steeper exposure-response surement may address different issues and may be relationship; от, while providing similar estimates, necessary as well; for example, the measurement of it is less costly. This, however, should never be clinical performance status by a questionnaire in taken for granted: while for individualization a addition to biomarkers such as the histological type biomarker measurement performed concurrently of a tumour or the amplification of an oncogene. within a study is always superior, for objectivity, specificity, sensitivity, biological significance and These four considerations highlight one point: cost, the relative merits of the biomarker measure- in essence, it is in the area of hazard detection and ment and of an alternative measurement can only risk quantification that issues of merit comparison be evaluated case by case. between biomaikers measurements and other types of measurements arise. The comparison can Pathogenesis studies, be cast in the systematic terms of evaluating the These studies focus not on the 'black box' rela- properties of measurements methods, in particular tionship between the primary causal factors, envi- their error elements, as done both for categorical ronmental or hereditary (inputs), and disease (out- and continuous variables by White (this volume), put), but on the interrelated sequences of biologi- or can be centred on signalling potential pitfalls cal events which lead from such etiological factors and strengths in the application of biomarkers

304 Comparing measurements of biomarkers with other measurements of exposure measurements compared to alternatives, as in the respect to smoking status, reducing both the sensi- following seven sections of this article. tivity and the specificity of the biomarker to 95% (from 100% in the case of the perfect measure- Random error ments of Table 1). However, the lifetime smoking Consider a 'true' situation, perfectly known with- status as assessed by the questionnaire measure- out errors, involving two dichotomous exposure ment is 'robust' under a variety of circumstances variables: lifetime smoking habit, evaluated by and assumed to remain error-free. Table 2 shows questionnaire as S+ and S—, and a marker of bio- that: logically relevant exposure (BS f- or BS—) reflecting Measuring exposure via the biomarker and the sum of all Ыologically effective lung carcinogens 1. in tobacco smoke, as all pathogenetic paths of via the questionnaire are now nearly equiva- tobacco lung carcinogens transit through it. Three lent, the former yielding a relative risk of 8 and features of what can happen in these circum- the latter yielding a relative risk of 7.4. stances (Table 1) are worth noticing: 2. Once the biomarker status is known, assess- tug smoking status by questionnaire yields addi- 1. subjects positive for the marker have a risk of tional information as indicated by the fact that, lung cancer 11 times higher than subjects neg- after stratification for ВЅ, the relative risk for ative for the marker. smoking is now 2.4 (reciprocally determining Ρ 2. In the absence of information on the marker, В s would add to an initial knowledge of smok- knowledge of smoking status as obtainable by ing status). questionnaire gives instead a lower relative risk З. This extra information has, however, a dis- for lung cancer (7.4). This is because some 2.4% turbing aspect: if controlling for a variable like Ρ of the BS± subjects are positive because of general В S, intermediate in the sequence between exter- environmental exposures to lung carcinogens, nal exposure to tobacco smoke and lung cancer, but are lifetime non-smokers; and, conversely, [n does not account for all of the effect of tobacco 20% of smokers, the marker remains negative, smoke, the common interpretation would be for physiological (metabolic) conditions. that such an effect is also mediated by pathogenic paths other than the one going through 3. Once the Ыomатker status of a subject is known, no further information on risk is added ВS. In fact, this is not the case. The appearance by the knowledge of smoking status: the relative of an alternative path is simply due to an error risk for lung cancer conditional on BS+ оr В5— in the measurement of BS. Hence, unless we status is lin both strata. know the size of this error, we may be unable to decide which of the two interpretations is likely In Table 2 a random error has been introduced to be correct. Actual measurements of any vari- in the biomarker measurement, mon-differential in able are unavoidably affected by random errors.

вs+ (п = 820) 8s-- (►т =1180) St (n = 800) S— (n - 20) S~ (п = 200) S— (n = 980) Lung cancer 80 2 2 . 10 No lung cancer 720 18 198 970

Relative risk (s+/s—) = 1.0 Relative risk (S+/Ѕ—) =1.0 . Relative risk (BS+1B5—) = 11.0

Relative risk (S+/s—) = 7.4

305 Application of Biomarkers in Cancer Epгdomiology

вs+ (n' - 838) вs- (п =1162) S+ (п'= 77р) Ѕ– (п'= 68) s+ (n'= 230) s- ( '= 932) л Lung cancer 76 3 6 9 No lung cancer 694 65 224 923 Relative risk (s+/s—) = 2.4 Relative risk (S+1S—) = 2.4 Relative risk (88+/Bs—) = 8.0 Relative risk (s+/S—) = 7.4

It follows from the simple example just presented categorizing the subjects of a population through that the relative sizes of the errors of the measured the measurement of a dichotomous variable. More variables will determine which pattern of rela- often, the error wil1 be expressed as one compo- tionship is observable; and in turn the iii terpreta- nent of the total variance of the measurements of till of the pattern will depend to a large extent a continuous variable in a population of subjects. on knowledge of the magnitude of these errors. Among Iaboratory analysts, it has instead been The issue is not restricted to the comparison of (and still 1s) customary to express the analytical biomarkers with alternative measures of exposure, error of measurement in the form of the propor- but bears, much more generally, 6f the possibil- tional standard deviation (coefficient of variation) ity of correctly disentangling in an observational of replicated measurements performed on the study the net effects of individual variables, inter- same samples)—from one or preferably from sev- correlated and measured with different degrees of eral subjects—in the same run of analyses, in dif- precision (Philips & Davey Smith, 1991). ferent runs or in different laboratories. The practice of using the coefficient of variation appears to be The error of measurement has been expressed based on the fact that very often the measurement here through the sensitivity and the specificity in variability increases with the magnitude of the measurement, yielding a constant coefficient. Table 3 depicts the attenuation of a true relative risk (odds ratio) resulting from a method of measurement affected by errors expressed by three different coefficients of variation. For instance, with a coefficient of variation of 5%, a true odds ratio of 5.0 will be observed as a value of 4.56. Several True odds ratio Coefficient of variation assumptions have been made in constructing the of laboratory random table, the most relevant being that it refers to a error variable (e.g. plasma total cholesterol) whose indi- 3% 5% 10% vidual observed values in a population may exhibit a coefficient of variation of 15% around the popu- 15 1.48 1.46 1.35 lation mean. For variables with a higher propor- 2 1.97 1.92 1.68 tional dispersion the attenuation would be less, 3 2.93 2.82 2.27 and for variables with a narrower proportional range of values in the population, the attenuation 5 4.82 4.56 3.32 would be even more marked than it is in the table. 10 9.51 8.77 5.67 Unless the coefficient of variation for the laboratory error is kept, say, within 5%, a material attenuation

306 Comparing measurements of biomarkers with other measurements of exposure

(а) 0J+ (n =1000) 0"- (n =1000) Lung cancer 60 60 No lung cancer 840 840 Relative risk (0.1+!0 --) = 1.0

(b) S+ (n =1000) 5- (n =1000)

о.1+ (п = 500) 0'l- (n = 500) о'l+ (n = 500) о.- (n = 500) Vitamin C low Vitamin C very low Vitamin C high Vitamin C low Lung cancer 50 50 10 10 450 450 490 490 No lung cancer -i-l0')--)=1 .0 Relative risk (OJ+1оJ-) = 1.0 Relative risk (0.l of the true odds ratio can ensue from this variation smoke is such a well known source of metaboli- alone, to which further attenuation must be cally active chemicals that information on it is added, chiefly from intrasubject variability. very rarely omitted in epidemiological studies involving biomarkers, and correction for its con- Metabolic confounding founding effect is feasible. Sti11, one may need A second aspect in need of thorough consideration better than only gross information on smoking concerns metabolic paths or, more generally, bio- habits; dose may matter, as may time, as some logical transformation processes (for instance, cell metabolic effects are long-term and others short- maturation processes) affecting a Ыomarkeх of term. It may be less obvious that it is necessary to exposure but not an external (questionnaire or correct for the influences of other agents (drugs, environmental) measurement of exposure. Table nutrients), particularly as some metabolic chains, 4(a) shows the lung cancer figures for 1000 con- such as the enzymes of the P450 superfamily, may sumers of orange juice (providing vitamin C sup- be interfered with by a vast number of xenobiotics, piementation) and 1000 non-consumers, the risk not all of which are known ог measurable. Iп gen- being the same in both categories. When stratified eral, confounders of an exposure biomarker need by smoking habit (Table 4b), the relative risk in the not be the same as confounders of an alternative two categories of orange juice consumers remains 1; measurement of the same exposure. This is an however, because of the enhancing effect of orange important consideration, since, in the overall plan juice on vitamin C plasma levels (a biomarker), of what needs to be measured in a study, one has and the often reported lowering effect of smoking to include not only the exposure(s) of interest but on the same biomarker, the measured levels of vit- key confounders as well. amin C are likely to turn out as schematized in the headings of Table 4(b) (high, low, very low). If one Disease-induced differential error had categorized the biomarker levels without This potential source of error is mentioned here for taking smoking habits into account, the exposure- the sake of completeness, as it is usually—unlike risk relationship of Table 5 would have resulted, other sources—well acknowledged in studies in- showing a clear increase of risk with decreasing volving biomarkers. In case-control and cross- levels of plasma vitamin C. Of course, tobacco sectional investigations, the presence of disease,

307

Application of 8iomarkers in Cancer Epidemiology

Table 5. Lung cancer occurrence in subjects classified by Vitamin C plasma levels

Vitamin C high (n = 500) Vitamin C low (n -1040) Vitamin C very low (n 500)

Lung cancer 10 60 50 No lung cancer 490 940 450 Relative risks 1.0 3.1 5.4 (Reference category)

e.g, colon cancer, niay in fact be the cause of interest is of the same order of magnitude, a range changes in a biomarker, e.g. blood cholesterol level, of values of the marker from —30 to +30% of the which will show up as an association between the average will be observed merely because of biomarker and the disease. This association, due to differences in the time of sampling, inducing a sub- differential error, could be wrongly interpreted as stantial misclassification of subjects by level of expo- indicating that the Ыотлarker precedes the disease sure. Alongside short-term (e.g. day-to-day or week- in time and can therefore predict the disease to-week) and medium-term (e.g. month-to-month) occurrence. The same may apply in cohort studies, variations, longer term variations in a biomarker as an unrecognizable, still subclinicaI, disease may can occur over a period of years. For example, vita- induce а change in a biomarker in those subjects min C intake could vary greatly over a person's life who are bound, later on, to move on to clinically and it may be that intake over the 20 years before recognizable disease. 5tratificatiоn of subjects by diagnosis influences lung cancer. In that case, a food disease stages may be one way (often the only one) frequency measure of vitamin C intake, which of getting a clue as to whether disease-induced dif- would have substantial error, could be a better mea- ferential error has occurred. An association which sure of the relevant exposure than an accurate bio- becomes stronger with advancing stages of the dis- marker of vitamin C that measures only recent ease points to the possibility of a disease-induced intake. Whereas the adverse effects of short-term differential error, although the absence of such a and medium-term variation can be reduced by the gradient represents only weak evidence against use of multiple measurements, long-term variation such bias, particularly if treatment has altered the in an exposure, and in the corresponding bio- natural course of the disease. marker, seriously limits the use of the marker and can only be overcome if a marker of cumulative Sampling time variability exposure becomes available. One well-recognized and critical aspect in the com- In the case of internal exposure, hormones are parative evaluation of a biomarker with respect to known to exhibit sizeable daily variation (for an alternative measure of exposure is the time of instance, in the evening, levels of cortisol may be appearance, persistence and disappearance of the less than half that in the morning) and again biomarker in relation to the time of the external casual sampling during the day may misclassify exposure or the fluctuations in time of an internal subjects and reduce the detectability of risks. exposure (e.g. hormones). Time differences in The general implication is that when assessing sampling may contribute substantially to subject the total intrasubject and intersubject variability of misclassification in respect to exposure, obscuring a biomarker, a good deal can be gained by knowledge the detection of associated risks and reducing the of its kinetics, which determines the magnitude of superiority of the biomarker compared to alterna- the differences related to sampling time. tive measurements. For example, if blood is taken from subjects during a survey lasting 3 months after Biomarkers of biological agents a relevant exposure has occurred, for instance in an In discussing the relative merits of biomarkers and accident, and the half time of the biomarker of alternative measurements of exposure, a basic truism

308 Comparing measurements of bionarkers with other measurements il exposure should not be forgotten: biomarkers are, by their second, even fewer of their relatives have it; and very nature, at their best when measuring biologi- third, the general population contains subjects cal agents (exposures), provided the agent leaves— bôth with and without that genotype. Now that as is often the case because of the specific biological molecular genetics techniques make it possible to interactions with the host—a characteristic long- determine directly a subject genotype, these term trace in the body. For example, when, in a sources of dilution have become virtually irrele- large cohort of more than 20 000 subjects, exposure vant, as comparisons of risks in persons with dif- to hepatitis B was assessed through ananinestic ferent genotypes, and different expressions of reporting of clinical hepatitis, a relative risk for sub- them, can be made. There is little question that sequent primary hepatocellular cardnoma of about genetic biomarkers, for use in both linkage and 4 was found; when, however, the HBAgs (Australia association studies, represent the most radical antigen) biomarker, indicative of a subject carrier methodological advance for etiological studies in status of the virus, was measured, the risk in sub- recent years. jects positive for the antigen was found to be more than 200 times higher than in subjects negative for Biomarkers as the reference the antigen, effectively ruling out confounding and Comparisons between a biomarker and alternative making bias unlikely (Beasley et aL,1981). Similarly, measurements of exposure are made in three increased risks of cervical cancer associated with a rather different circumstances: history of sexual pramiscuity, of the order of two- to threefold, were found in several studies (Cramer, Quantitative comparison 1982), leading to the hypothesis that a sexually This comparison stands on the hypothesis that the transmissible biological agent was causally implied. two measurements measure the same quantity Measuring exposure to papkioma viruses (particu- (exposure) and that, ideally, a correlation of 1.0 larly HPV 16) with increasingly reliable techniques, should exist between the two measurements. prevalence odds ratio of 20 and more have been Divergence from this perfect correlation derives shown in a large number of case-control studies, from errors in both measurements and can be supporting the conclusion of a causal link between quantified in the form of a reliability coefficient, the virus and cervical cancer (IARC, 1995). which, besides being of interest in itself, allows the This advantage of biorriarkers applies not only validity coefficient to be estimated, at least at its to the class of environmental biological agents but upper and lower bounds (Armstrong et a2., 1992, also to the other class of etiologically relevant bio- pp. 82-83). In the much less common situation in logical agents, the genes. Indirectly assessing the which sound evidence is available that the bio- exposure to a gene through family aggregation of marker can be regarded as the ultimate and true cancer cases may cause an enormous attenuation of the true relative risk associated with a gene increasing the risk of cancer. Table б (from Peto, 1980) refers to the case of a gene with a dominant allele confer- ring an increased risk of cancer and having a population frequency of 5%. When those homo- and heterozygotes for the allele (95.7% of the population) are compared with those without the allele, a Relative risk in relatives true relative risk of 100 becomes an observable rel- of cancer patients ative risk of less than five if measured—as has been True relative risk done for decades in epidemiology—by actually (XX, OX versus 00) Twin sibling Child contrasting the cancer risk among siblings of can- 2 1.07 1.04 1.04 cer cases with the risk in the general population. 10 3.02 2.00 1.99 This dilution, probably one of the strongest one 100 8.60 4.75 4.70 cari find in exposure assessment, derives from 1000 10.10 5.48 5.42 three sources: first, not all the cases of the investigated cancer have the risk-enhancing genotype;

309 Application of Вi maгkers in Cancer Epidemiology

s+, s+ (n = seal s+, s-; s-, 5+ (л 532) B5—, s-- ( = 788) в в в п Lung cancer 64 22 g Ni lung cancer 616 510 780 ReЭative risks 10.1 4.2 1.0

(Reference category)

measure of the exposure, both the validity and reli- archaeological techniques, he destroyed as many ability of the alternative measure can be directly `markers' as he revealed. From that moment, based determined with respect to it. on a limited validating evidence, the whole series of events and their interrelationship became regarded Qualitative validation as the true history of the Troy war. In epidemiology, This type of validation assumes the biomarker as a `Troy validation' of very indirect exposure mea- the reference and validates the alternative measurements against the few feasible measurements surement (obtained by questionnaire or environ- of a meaningful biomarker may be formally unsat- mental measurement) by looking for a correlation isfactory but substantively informative. with the Ыоmarker, the underlying and implicit hypothesis being that the correlation may indeed Analogic extrapolation be zero, given the very indirect nature of the alter- Whatever relationship is found between a bio- native measurement. This type of qualitative vali- marker and an alternative measure of exposure dation, in which finding a statistically significant may be extrapolated, by analogy and with caution, correlation is even more important than the mag- to hold in circumstances different from those nitude of the correlation, is particularly relevant under which the relationship was found. For when dealing with past exposures. For instance, instance, in the short term (i.e. a week), reported that the levels of dioxin (TCDD) are higher tri fat exposure to environmental tobacco smoke in non- samples available from only a few pesticide smoking women correlates well with urinary coti- sprayers than in the general population represents nine levels, which reflects actual exposure to circumstantial evidence of the exposure to TCDD tobacco smoke (Riboli et al., 1990). This offers sup- or to substances contaminated by TCDD. This, in port, by anаlоgy to the contention that in the long 0m, lends support to an interpretation in terms of term (i.e. years), self-reported exposure to ЕТ5 occupational exposures of the relationship which reflects actual exposure to ЕTS. may have been found between reconstructed work histories and disease end-points in the whole Composite markers cohort of sprayers. One way in which biomarker measurements, even if This recalls the story of the ancient city of Troy, imperfect, may be valuable is through combination in Asia Minor, as told in the epic poems of Homer. into composite variables, formed by several bio- It was commonly held among scholars for genera- markers or by biomarkers and other measurements tills that the events described by Homer, as they of exposure. This is particularly advantageous appeared so well detailed, never took place at all when some of the measurements can be performed and were pure legend. However, in 1870, the cheaply. Referring back to the case of a random German merchant aid archaeologist Heinrich error affecting the measurement of a biomarker BS Schliemann went and dug where Homer said Troy (Tables 1 and 2), assume now that the loss of sen- was and found the site of the city (Lister, 1967). sitivity and specificity is sizeable, both being Incidentally, as he was using rather primitive reduced to 80%. In these conditions, the relative

310 Comparing measurements of biomarkers with other measurements of exposure risk for the biomarker of tobacco smoke is just 3.4, When evidence allows a biomarker to be regarded while the relative risk assessed via a tobacco smok- as the true measure of exposure, the validity and ing questionnaire remains 7.4. Once the latter is reliability of alternative measurements can be known, the relative risk for the biomarker is, directly determined against it. A less elegant and within the strata of tobacco smoking habits only ambitious, but no less valuable and common, use 1.8; it may be tempting (particularly if the bio- is in confirming, on the basis of whatever mea- marker measurement is costly) to dismiss it as surements may be feasible (`Troy validation'), that worthless. This may in fact be wrong, as the bio- indeed other indirect measurements of exposure marker measurement contributes information bear at least some relation to a biologically rele- when considered jointly with the questionnaire vant exposure. measurement. An odds ratio of 10.1 is found in the Two final remarks are apposite. First, there is no highest category of a new variable created by coil- need to have biomarkers, as opposed to other pounding the questionnaire information and the exposure variables, measured in every epidemio- biomarker information, which shows (Table 7) a logical study, and epidemiologists ought to make clear gradient of risk with increasing levels of the clear and stress to colleagues from other disci- variable. This elementary combination of variables plines, including those sitting in peer review and underlines the important principle that when dealing grant awarding committees, that what qualifies a with correlated variables purporting to measure the good or a bad epidemiological study is not the mere same underlying, but not directly accessible, vari- presence or absence of measurements of some of able, it may definitely be advantageous actually to the latest available Ыomarkers. Second, the flow of measure several of them, even if they are individ- newly measurable biomarkers appears to be such ually affected by sizeable non-random errors, and that it is reasonable to expect that biomarkers of to combine them in a composite variable exposure which are capable of meeting the exact- (Armstrong et aI., 1992, pp. 115-125). Indeed, this ing requirements demanded by their use in epi- may be one of the main ways of overcoming the demiology will soon be forthcoming. inherent imperfection of exposure measurements in epidemiology. . References Armstrong, B.K., White, E. & Saracci, R. (1992) Principles Conclusion of Exposure Measurement in Epidemiology. Oxford, Oxford Biomarkers of exposure are enriching the epidemi- University Press ologist's armamentariuni for evaluating exposures, Beasley, R.P, Hwang, L.Y., Lin, C.C. & Chien, C.S. (1981) present and past. The issue of the relative merits of I-lepatoceilular carcinoma and hepatitis B virus. A bioniarkers compared to alternative measures of prospective study of 22707 men in Taiwan. Lacet, ii, exposure arises particularly in the context of stud- 1129-1133 ies aimed at hazard detection and quantification. Cramer, D.W. (1982) Uterine cervix. In: Sclsottenfeld, D. In general, the relative merits of the different types & Fraumeni, J.F., eds, Cancer Epidemiology and Prevention, of measurement of exposure need to be evaluated Philadelphia, W.B. 5aundexs, pp. 881-900 on a case-by-case basis. However, when the exposure International Agency for Research on Cancer (1995) is represented by a biological agent, biomarkers are IARC Monographs on the Evaluation of Carcinogenic Risks to the first, and often the only justifiable, choice. Humans, Vol. 64, Human PapгlIomaviruses, Lyon, Whatever the type of study and the type of expo- International Agency For Research on Cancer sures involved, biomarkers may be affected by ran- Lister, R.P. (1967) Turkey Observed, London, Eyre & dom errors, time-related sampling errors, physio- Spoinswoode logical confounding and disease-induced differen- Peto, J. (1980) Genetic predisposition to cancer. Iд: tial error; fui assessment of these aspects is needed Cairns, J. Lyon, J.L. & Skilrük, M., eds, Cancer Incidence before embarking on the use of a biomarker in a in Defined Populations, Cold Spring Harbor, Cold spring full-scale epidemiological study. Random errors Harbor Laboratory, pp. 203-213 affecting biomarkers may be reduced by replica- Phillips, A.N. & Davey Smith, G. (1991) How indepen- tion оr combination of measurements, or both. dent are "independent" effects? Relative risk estimation

311 Application of Biomarkers in Cancer Epidemiology

when correlated exposures are measured imprecisely. Cl . Epidemiol., 44, 1223-1231 Ј. гп Riboli, E.., Preston-Martin, S., Saracci, R., Haley, N.J., Trichopoulos, D., Becher, H., Burch, J.D., Fontham, E.T.Н., Gai, Y.T., Jiпдаl, S.1C., Ku L.С ., Le Marchand, L.L., segnan, N., Shimizu, H., Stanta, G., Wu-Wiцiаms, A.H. & Zatonski W. (1990) Exposure of nonsmoking women to environmental tobacco smoke: а 10-соипtтy coilaborative study. Cancer Causes Control, 1, 243-252

R. Saraccэ National Research Council, Via Trieste, 41 56100 Pisa, Italy

312 Application it Biomarkers in Carrcer Epidemiology Toniolo, P., Bottelle, Р, Ehuker, D.E.G., Rothman, N., Rulka, B. and Pearce, N. ads iARC ScienIiIic Publicali01s No. 142 1ntemauunj Agency for Research on Cancer, Lyon, 1997

Ethical and social issues in the use of biomarkers in epidemiological research

Р.А. Schulte, D. Hurter and N. Rothman The use of biomarkers in epidemiological research may raise ethical and social issues. These issues stem from the belief that research participants have `rights' to appropriate information before, during and after studies so that they can make informed decisions. Ethical issues can arise during protocol development, obtaining participation, and in the interpretation and notification of text and study results. Additionally, there are ethical cpnsiderations concerning the use of biological specimens collected and stored for one purpose and subsequently used for other research purposes. A major ethical issue is the maintenance of participants privacy and the confidentiality 0f their test and study results. Ethics committees need to be well- informed about the scope, limitations and expectations of biomarker research in order to be able to respond to social and scientific developments in the use of biomarkers.

Epidemiological research potentially raises many erаl, cross-sectional studies of healthy workers are ethical questions and issues. The use of biomarkers completed in a short period of time, with the in such research may raise further issues, because expectation that the biowarkers under study may biomarkers are obtained from the individual per- provide some insight into the potential risk of an son and have the potential for providing impor- exposed group as a whole, or possibly into an indi- tant information about exposures, biological vidual's risk of subsequent cancer development. effects of exposures and susceptibility to disease for Notifying workers of their results in these studies is that individual (Grandjean, 1991; Schulte, 1992; common. By contrast, the case-control study Van Damme et aL, 1995). At the same time, there involves subjects who are already sick, along with is the widespread misconception that biological randomly selected controls who are not definable information is always more valid than other infor- in an а priori sense to be at risk. These subjects are mation, such as that obtained from questionnaires, generally not notified of results. Finally, in the environmental monitoring or record review. None third example, everyone is healthy and samples are the less, the potential contribution of biomarkers provided with the expectation that results will not to enhancing determination of carcinogen expo- be available for a relatively long time. sure-disease associations, identifying disease In the following pages, we will generalize about earlier, or identifying particular etiological sub- these issues, but the appropriateness may vary by groups riiakes the use of biomarkers desirable and study design or detail and the ethical issues should inevitable. be addressed on a case-by-case basis. There is increasing recognition that many of the In this paper, we will use the steps in the issues related to recruiting and informing subjects research process as the organizing theme and dis- of test and study results have varied depending on cuss ethical and social issues fox each step. Where study design. Consider three examples: (1) a cross- there are issues that differ according to the type of sectional study involving occupational exposure marker or the use of a marker, these aspects will be and a biomarker of early effect (e.g. cytogenetic identified. Finally, we will discuss the use of stored effects); (2) a cancer case-control study evaluating specimens in biomarker research. the impact of common polymorphisms of me- A premise of this paper is that ethical use of bio- tabolizing enzymes; and (3) a prospective cohort markers in research involves attention to the study with banked biological specimens. in gen- rights' of subjects to appropriate information

313 Application of Biomarkers in Cancer Epidemiology

before, during and after studies, so that they can hal subjects are told about the study and whether make informed decisions. Failure to plan or budget they can truly give informed consent. If subjects adequately for these efforts can lead to these rights are deceived or coerced into participating in a not being met. We would note, however, that there study, or are given false expectations (e.g. 'we can is some difference of opinion about when and tell if you are sick or well') with respect to the value what to tel research participants, and this will be of the study to the participant, ethical principles discussed in later sections. are violated. For example, a researcher could coerce a potential subject directly (e.g. 'you may lose your Protocol development and study design job if you don't participate') or by implication. Ethical issues come into play from the moment Communicating false expectations or using pres- biomarkers are considered for a study. Why is the sure is patently dishonest and unethical. It is biomarker being considered? Biomarkers are usu- unlikely that such deception or coercion would be ally more resource- and labour-intensive than overt; rather, it would be more subtle and difficult other measures of exposure, outcome or risk. The to detect. Abroad spectrum of opinion exists about use of scarce resources to develop, validate or apply what obtaining informed consent entails and a biomarker can be wasteful or inefficient if there when it is achieved. Some believe that for markers is not a good rationale. Essential to the design of whose meaning is not known at the time of the transitional, etiological or applied studies is the study, a subject or worker in an occupational study need to identify the driving scientific or public cannot give truly informed consent (Samuels, health questions and to determine whether they 1994). This implies a much higher standard of could be answered by some other approach interpretation for biomarker information than for (Rothman, 1993; Rothman et al., 1995). Тhis may other information routinely obtained by question- be Iеss of an issue for laboratory studies where bio- naires, environment monitoring or record linkage. marker work is the defining activity. It may be more In studies to validate markers of exposure, the level critical when considering using biomarkers as of understanding of the meaning of the marker is independent or dependent variables in epidemio- similar to that from classical exposure sources. logical studies or for public health applications Frequently, airborne exposure, levels in blood, or such as screening, monitoring or in risk assessment frequency of DNA or protein adducts are part of (Office of Technology Assessment, 1990; Perera, the same exposure paradigm. Markers of effect or 1987; Schulte & HaIperirn, 1987; Rudiger, 1994). susceptibility are different. Until there is determi- Ethical and social problems may also arise from nation of predictive value and course in the nat- a failure of researchers to anticipate and plan the ural history, such markers are clearly only research actions required for dealing with the more extreme variables with no clinical meaning, and partici- biomarker assay results. This may include repeat pants should be made aware of this. If a marker has testing, counselling or diagnostic evaluations. For been validated (i.e. quantitatively linked to risk of transitional studies in which the characteristics of a disease at the group or individuaI level), then a marker are being determined, and where there are clear description of it should be given to potential clearly no associated clinical findings, prognostic research participants. With regard to informing significance or meaning, the needs of subjects may participants of risks, general practice has been to be different from those situations, such as screening identify only medical risks; however, it has been or biological monitoring, where a marker can have argued that truly informed consent should include implications for individual risk or for disease. With reference to non-medical risks that might affect markers of susceptibility, it may be important to participants. For example, a study subject may be consider the impact of the research not only on informed that they carry a genetic mutation that individual participants, but also on their families. puts them at a high risk of subsequently developing cancer. In the extreme case, the mere acknowl- Obtaining participation edgement on an employment or insurance appli- How subjects are recruited into studies can involve cation that they have had a biological or genetic serious ethical and social issues (Schulte & test may result in denial of employment or insur- Sweeney, 1995). These issues hinge on what poten- ance. Another variation on this scenario is that

314 Ethical aid social issues

misinterpretation of a biomarker assay result could in deciding when biomarkers indicate that early occur and have the same impact. warning steps should be taken. These may include Participants consent to provide the specimens efforts to control exposures (in occupational or and corollary demographic and risk factor infor- environmental settings), the need for subsequent mation, and hence cooperate in the specified testing, ongoing monitoring, or simply, and often research. The subject generally does not consent оr most importantly, counselling and a demonstra- imply consent to distribution of the data in а way tion of caring. that identifies him от her individually to any other Interpretation of biomarker data can be diffi- parties, such as employers, unions, insurers, credit cult. For example, in cross-sectionaI studies of agencies, lawyers, family members, public health populations with occupational or environmental agencies, etc. . exposure, evaluating the relationship between Dissemination or revelation of results beyond exposure and markers of early biological effect, the explicit purposes for which specimens were biomarkers will not be indicators of risk per se, collected intrudes on subjects privacy. Studies but of exposure, susceptibility given exposure, оr where biological specimens and DNA are banked biological changes that could be homeostatic for future use may require informed consent about responses to an exposure (Ashford, 1986). The this future use. In this respect, questions are raised investigator needs to sort out these changes against about whether specimens collected for one pur- a background of extensive intra-individual aid pose can be used for different research purposes interindividual variзbшty in biomarkers. The current and about the responsibility for conveying results technological capabilities offer investigators and back to the subjects (Schulte & Sweeney, 1995). practitioners the opportunity to utilize techniques Also related to this is the ownership of specimens. with heightened sensitivity for detecting changes Who owns them—the subject, the researcher, the at cellular and molecular levels, and for detecting sponsoring agency or others? Although this has exposures to minute amounts of a xenobiotic. At the been adjudicated in the case of a clinician who same time, at these levels, inherited and acquired profited from a hairy cell leukaemia line derived host factors aid other confounding factors can be from cells taken from a patient (Cooper, 1985; strong causes of wide variability in biomarker Office of Technology Assessment, 1987), we have results unrelated to the exposure of interest. found no references (except Clayton et aL, 1995, The results of studies of biomarkers of suscepti- see later discussion) to the issue as it pertains to bility can lead to findings that might be misun- epidemiological research with stored specimens. derstood or abused (Lappe, 1983; Ashford, 1986; Nelkin & Tancredi, 1989). For example, some genes Interpretation and notification of test and study (such as those that are commonly occurring, that results confer low relative risk and that require a specific Biomarker research yields individual test (assay) exposure or other genes to increase risk of disease) and study results (Schulte & 5ingal, 1989). (see Caporaso & Goldstein, this volume) do not Research participants may want, or have, a right provide unambiguous information, but various to these results and an interpretation of them. groups in society may start using such genotype Interpretation of these results is the responsibility information as if it represented 'diagnoses' rather of investigators. Some institutions require investi- than risk factors (Wagener, 1995). gators to provide individual test results to subjects In some studies, multiple biomarkers will be as well as overall study results, while others may assessed, and researchers have a responsibility to advise them not to communicate results of assays consider whether issues of multiple comparisons can that have no clinical relevance. Attendant to these lead to inappropriate selection of significance levels. efforts is the provision of an interpretation as far as Association of biomarkers not included in original is possible. Even though participants are told that hypotheses should be evaluated at more rigorous tests may be purely for research purposes and have levels of statistical significance, and subsequent no clinical value, they still ultimately want to interpretations should be considered in that light. know if they are 'aIl right'. Investigators and prac- One area of interpretation that is problematic titioners face ethical issues in interpreting tests and is what is called `individual risk assessment'.

315 Application of Biomarkers in Cancer Epidemiology

Generally, epidemiological studies (with or with- to start a new study and ask the relevant questions, out biomarkers) yield group results. The risk per- a new hypothesis using a biomarker can often be tains to the group as a whole and not necessarily to tested using specimens from previous studies. 1f it is individual members of the group. It is possible to desirable to have prospectively сопесtед specimens, compute an individual risk using a risk function for instance if the biomarker level may be biased by equation (Truett et al., 1969); however, if the disease, then available specimen banks with follow- marker being used has not been validated for dis- up data will be the preferred resource for testing the ease, the calculation will be meaningless. Thus far, new hypothesis. Otherwise, it might take many for the current generation of molecular biomarkers years to develop a new specimen bank with suffi- used in cancer research, there are practically no cient outcomes and follow-up to test a hypothesis. markers, with the exception of a few genetic muta- Ethical issues for stored specimens relate to rions linked to high risk of disease in cancer fam- (1) whether consent for use of the specimens in ily syndromes, for which an individual risk can be research was originally given, and (2) whether this determined based on the level of the marker. consent was generic or specific to the hypothesis to All of these characteristics of biocoaiker data be tested, and whether the consent obtained when may lead an investigator to conclude that а partic- the specimens were collected stil meets the stan- ular biomarker is of uncertain meaning with regard dards of informed consent. to risk. None the less, the investigator has the Many specimens stored for research purposes obligation to portray accurately the degree of would have been collected after informed consent uncertainty in test and study results. There are a to research was given; however, some types of spec- range of opinions about communicating results of imens, particularly clinical specimens initially used biomarker tests on individuals or groups if there is for diagnostic or prognostic purposes, may have no clinical meaning, such as usually occurs in tran- been stored without consent or even without the sitional studies to validate markers. Some believe patient's knowledge. Frequently in clinical set- the autonomy of participants is not honoured if tings, a wide variety of tests are ordered without they do not receive the information, while others any consultation with the patient, although clear believe that the information is meaningless to par- exceptions exist, such as HIV testing, for which ticipants. The latter view has the appearance of being consent is usually mandatory. It has long been paternalistic, but may be viewed as doing no harm. held as ethically acceptable practice to conduct Other ethical issues involved in notification are some types of research on 'discarded' blood or tis- the importance of communicating information in sues, i.e. specimens left over after the clinical tests a timely fashion and the evaluation of the impact are performed. Access to these tissues has been crit- of notification efforts. The timeliness of notifica- ical to the development of new clinical markers tion is mainly an issue when results indicate an such as histological or immunocbemical markers action that could reduce exposure or risk, or effect of cancer prognosis, in which hundreds or thou- timely treatment. Evaluating the impact of notifi- sands of uniformly collected specimens are fre- cations may not need to be a routine matter, but quently needed to establish a new test as being since the impact of notification cannot always be informative. It would seem a natural extension of anticipated, it may be useful to have included in this tradition that new biomarkers of genetic sus- the notification an opportunity for the participant ceptibility or prognosis would also be evaluated in to obtain more information or provide feedback this way. However, because of the potential high about the results. predictive value of some of these tests, as well as the implications for family members, this tradition Use of stored specimens in biomarker research is being challenged, and a lively debate is currently Biomarker research is qualitatively different from underway about the ethics of using these tissues. A most other epidemiological research, because tech- recent statement from a working group of the riical developments make new assays feasible on Ethical, Legal, and Social Implications of the stored specimens long after the original consent is Human Genome Project suggested that informed obtained. Unlike questionnaire-based research, in consent should usually be obtained before testing which the response to a new hypothesis is usually for genetic susceptibility on clinical specimens

316 Ethical and social issues

(Clayton et al., 1995), although the statement subjects, in which the first rule is that virtually all acknowledged that research involving `minimal such research must be approved and reviewed by risk', and for which re-consenting subjects would an appropriate ethics committee but relatively few be impracticable, could be exempted. The defini- types of research are absolutely proscribed or highly tion of 'minimal risk' and the determination of regulated. Some have proposed that research what constitutes `impracticability' are at the centre involving genetic susceptibility is qualitatively dif- of much of the current uncertainty and debate. ferent from other research, and that much stricter Even in a research study, the original consent standards of informed consent should apply form can only be as thorough as the original aims (Armas, 1995); while others have argued that the of the study and the state of knowledge at the time level of consent or notification should be com- permit. Samples from participants in a study of mensurate with the degree of risk involved, and cancer risk factors, for instance, may subsequently thus less stringent procedures may be appropriate be useful in a study of cardiovascular disease or for low risk, relatively common polymorphisms psychiatric illness. Even the best designed and (e.g. Р450 genes) than for high-risk genotypes (e.g. informed consent process in a study of genetic sus- BRCA1, BRСА 2). In the USA, the possibility that ceptibility to cancer may be outdated with the biomarkers of susceptibility could be used to dis- discovery of a new susceptibility gene or a new criminate in the context of health insurance or prognostic implication of an 'old' gene. A major employment is a major concern, which may dilemma in current biomarker research is whether expose research participants to potential economic the generic consent originally given by a partici- harm. On the other hand, epidemiology has a good pant to do research on a specimen is adequate con- track record in protecting participants from loss of sent to conduct a specific test which may not even confidentiality in many studies over many years have been envisaged at the start of the study. The which have included Ыgh1y sensitive question- obvious strategy of obtaining fresh consent has at naire-based data. Although some unique issues are least three major problems: (1) subjects may be raised by biomarker research, most issues are simi- very difficult to contact if follow-up has not been lar to those encountered in other types of research, maintained, or they may have died; (2) a high pro- and can be overseen by appropriately constituted portion of non-consent, due either to inability to ethics committees who are in the best position to re-contact or to refusal, may bias the study; (3) for be aware of the local and particular aspects of any certain especially valuable specimens, such as proposed biomarker research. Especially close those from cohort studies, multiple genes may be scrutiny should be given to any proposal using a of interest and a process of very specific informed biomarker with likely high predictive value. Ethics consent would generate an almost continuous committees also need to be well informed about the stream of consent requests to the participant. scope, limitations and implications of biomarker Failure to obtain a new informed consent may research, as the ethical climate in this field may expose the researcher to allegations of unethical change quite rapidly as scientific developments behaviour, or may create a difficult situation if the occur and society responds to these developments. biomarker information is of clinical relevance to the participant and yet the participant was not Confidentiality of data pre-test counselled about the test. The nature and Investigators need to maintain the confiden- force of these problems will be very different tiality of biomarker data because of the potential according to the predictive quality of the bio- for misuse or abuse leading to discrimination, marker and its clinical implications, and the social labelling and stigmatization. This can be increas- and cultural setting of the research. ingly difficult because ownership of stored speci- Owing to the heterogeneity of study settings, mens may be in question and various investigators and of social norms and responses, it is likely to be may request the use of specimens for research, liti- impossible to draft uniform rules on what consti- gation or commercial enterprise. In some cases, lutes ethical behaviour in every application of bio- where specimens are identifiable or are capable of marker research and every situation. This is cur- being linked to databases where identification is rently the case with research involving human possible, it may be difficult to assure confidentiality.

317 Application of Biomarkers in Cancer Epidemiology

Informatics and the ability to link disparate data- Rddiger, H.W. (1994) Changing emphasis irs occupa- bases are progressing at a rapid pace. In some coun- tional toxicology: are we ready? Tot. Arch. Occup. Env. Health, 66, 69 tries, there may be a need for further legislation to prohibit unauthorized access to, or use of, sped- Samuels, 5.W. (1994) A moral history of the evolution of a men results. The challenge will be to assure the caste of workers. Presentation at Workshop on the rights of study participants while providing for a Development and Applications of Biomarkers, Santa Fe, broad range of research opportunities. NM, April 26-29 Schulte, ~.А. (1992) Biomarkers in epidemiology: References Scientific issues and ethical implications. Environ. Health Armas, G.J. (1995) Editorial: Genetic prophecy and Perspect., 98, 143-147 genetic privacy—can we prevent the dream from becom- Schulte, P.A. & Halperin, WE. (1987) Genetic screening trig a nightmare? Am.'. PuNic Health, 85, 1196-1197 and monitoring for workers. In: Harrington, J.М., cd., Ashford, N.A. (1986) Medical screening in the workplace: RecentAdvances in Occupational Health, Vol. 3, Edinburgh, legal and ethical considerations. Serein. Occup. Mcd', 1, Churchill Livingstone, pp. 135-154 67-79 Schulte, P.A. & Singal, M. (1989) Interpretation and com- Clayton, E.W., Steinberg, K.K., Khoury, M.J., Thomson, munication of the results of medical field investigations. E., Andrews, L, Kahn, M.E., Kopelrnan, L.M. & Weiss, J. Occup. Med., 31, 5898-5894 J.O. (1995) Informed consent for genetic research on Schulte, P.A. & Sweeney, M.H. (1995) Ethical considera- stored tissue samples. J. Ara. Mcd. Assoc., 274, 1786-1792 tions, confidentiality issues, rights of human subjects, Cooper, J.P. (1985). Biotec[mology and the Law, New York, aid uses of monitoring data in research and regulation. Clark Boardman Ltd Environ. Health Perspect., 103 (suppl. 3), 69-74 Grand jean, P. (1991) Ethical aspects of genetic predispo- Truett, J., Cornfield, J. & Kannel, W. (1967) A multivari- sition to disease. In: Grandi can, P., ed, Ecogenetics, ate analysis of coronary heart disease in Framingham. J. London, Chapman and Hall, pp. 237-251 Chrom. Dis., 20, 511-524 Zappe, M. (1983) Ethical issues in testing for differential Van Damme, K,, Caste1eyn, L., Heseltine, E., Huici, A., sensitivity to occupational hazards. J. Occup. Mcd., 25, Sorsa, M., Larebeko, N. & Vineis, P. (1995) Individual sus- 797-808 ceptibility and prevention of occupational diseases: scientific and ethical issues. J. Ocuip. Environ. Med., 37, Nelkin, D. & Tancredi, L. (1989) Dangетоиs Diagnostics. 91-99 The Sосiаl Power o fВiоlоgiсаl information, New York, Basic Books, Inc Wagener, D.K. (1995) Ethical considerations in the design and execution of the national and Hispanic Office of Technology Assessment (1987) New Develop- Health and Nutrition Examination survey (HANES). ments in Biotechnology, Vol. 1, Ownership a fHuman Tissues Environ. Health Perspect., 103 (suppL 3), 75-80 and Cells (ОТА-BA-337), Washington, US Government Printing Office Perera, F.P. (1987) The potential usefulness of biological markers in risk assessment. Environ. Health Perspect., 76, 141-145

Rithman, N. (1993) Epilogue. In: Schulte, P.A. & Регега, F,V., eds, Molecular Epidemiology: Pr nciples and Practices, г Corresponding author San Diego, Academic Press, pp. 565-569 P,A. Schulte Rothman N. Stewart, W.F. & Schulte, P.A. (1995) National Institute for Occupational Saiаiy and Health, Incorporating biomarkers into cancer epidemiology: a 4676 Columbia Parkway, C-14, Cincinnati, OH 45226, matrix of biomaxker and study design categories. Caner 115А Epidemiol.Biomarkers Prev., 4, 301-311

318 'ARC Monographs on the Evaluation of Carcinogenic Risks to Humans

Volume i Volatile Anaesthetics Volume 23 Some Inorganic Substances, 1976; 306 pages; ISBN 92 832 1211 8 Some Metals aid Metallic Compounds Chlorinated Hydrocarbons, Aromatic (out of print) 1980; 438 pages; ISBN 92 632 1523 0 Amines, N-Nïtroso Compounds, and (out of print) Natural Products Volume 12 1972; 184 pages; ISBN 92 832 1201 0 Some Carbamates,Thiocarbamates Volume 24 (out of print) and Carbazides Some Pharmaceutical Drugs 1976; 282 pages; ISBN 92 832 12126 1980; 337 pages; ISBN 92 832 1524 9 Volume 2 Some Inorganic and Organometallic Volume 13 Volume 25 Compounds Some Miscellaneous Pharmaceutical Wood, Leather and Some Associated 1973; 181 pages; ISBN 92 832 1202 9 Substances Industries (out of print) 1977; 255 pages; ISBN 92 632 12134 1981;412 pages; ISBN 92 832 1525 7

Volume 3 Volume 14 Volume 26 Certain Polycyclic Aromatic Asbestos Some Antineoplastic and Hydrocarbons and Heterocyclic 1977; 106 pages; ISBN 92 832 12142 lmmunosuppressive Agents Compounds (out of print) 1981;411 pages; ISBN 92 832 1526 5 1973; 271 pages; ISBN 92 832 12037 (out of print) Volume 15 Volume 27 Some Fumigants, the Herbicides Some Aromatic Amines, Anthraquinones Volume 4 2,4-D and 2,4,5-T, Chlorinated aid Nitroso Compounds, and Some Aromatic Amines, Hydrazine Dibenzodioxins and Miscellaneous Inorganic Fluorides Used fn Drinking and Related Substances, N-Nitroso Industrial Chemicals Waler and Dental Preparations Compounds and Miscellaneous 1977; 354 pages; ISBN 92 832 12150 1982; 341 pages; ISBN 92 832 1527 3 Alkylating Agents (out of print) 1974; 286 pages; ISBN 92832 12045 Volume 28 Volume l б The Rubber Industry Volume 5 Some Aromatic Amines and Related 1982; 486 pages; ISBN 92 832 1528 1 Some Organochiorine Pesticides Nitro Compounds — Hair Dyes, 1974; 241 pages; ISBN 92 832 12053 Colouring Agents aid Miscellaneous Volume 29 (out of print) Industrial Chemicals Some industrial Chemicals and 1978; 400 pages; ISBN 92 832 1216 9 Dyestuffs Volume 6 1982; 416 pages; ISBN 92 832 1529 X Sex Hormones Volume 77 1974; 243 pages; ISBN 92 832 1206 1 Some N-Nitroso Compounds Volume 30 (out of print) 1978; 365 pages; ISBN 92 832 1217 7 Miscellaneous Pesticides 1983; 424 pages; ISBN 92 832 15303 Volume 7 Volume lB Some Anti-Thyroid and Related Polychlorinated Biphenyis and Volume 31 Substances, Nitrofurans and Polybrominated Biphenyls Some Food Additives, Feed Additives Industrial Chemicals 1978; 140 pages; ISBN 92 832 1218 5 and Naturally Occurring Substances 1974; 326 pages; ISBN 92 832 1207 X 1983; 314 pages; ISBN 92 832 1531 1 (out of print) Volume 19 Some Monomers, Plastics and Volume 32 Volume 8 Synthetic Elastomers, and Polynuclear Aromatic Compounds, Some Aromatic Azo Compounds Acrolein Part 1: Chemical, Environmental and 1975; 357 pages; ISBN 92 832 1208 8 1979; 513 pages; ISBN 92 832 1219 3 Experimental Data (out of print) 1983; 477 pages; ISBN 92 832 1532 X Volume 9 Some Aziridines, N-, S- and Volume 20 Volume 33 0-Mustards and Selenium Some Halogenated Hydrocarbons Polynuclear Aromatic Compounds, 1975; 268 pages; ISBN 92 832 1209 6 1979; 609 pages; ISBN 92 832 12207 Part 2. Carbon Blacks, Mineral Oils (out of print) and Some Nitroarenes Volume i0 1984; 245 pages; ISBN 92832 15338 Some Naturally Occurring Substances Volume 21 (out of print) 1976; 353 pages; ISBN 92 832 1210 X Sex Hormones (Il) (out of print) 1979; 583 pages; ISBN 92 832 1521 4 Volume 34 PoIynucIear Aromatic Compounds, Volume 11 Volume 22 Part 3: Industrial Exposures in Aluminium Cadmium, Nickel, Some Epoxides, Some Non-Nutritive Sweetening Production, Coal Gasification, Coke Misce1Ianeous Industrial Chemicals Agents Production, and Iron and Steel Founding and General Considerations on 1980; 208 pages; ISBN 92 832 1522 2 1964; 219 pages; ISBN 92 832 1534 6 Volume 35 in Pain1 Manufacture and Painting Volume 59 Polynuclear Aromatic Compounds: 1989; 535 pages; ISBN 92 832 1247 9 Hepatitis Viruses Part 4: Bitumens, Coal-Tars and Derived Products, Shale-Oils and Soots Volume 48 1994; 286 pages; ISBN 92 832 12592 1985; 271 pages; ISBN 92 832 1535 4 Some Flame Retardants arid Textile Volume 60 Chemicals, and Exposures in the Same Industrial Chemicals Volume 36 Textile Manufacturing Industry 1994; 560 pages; ISBN 92 832 1260 6 Ally] Compounds, Aldehydes, 1990; 345 pages; ISBN: 92 832 12487 Epoxides and Peroxides Volume 61 1985; 369 pages; ISBN 92 832 1536 2 Volume 49 Schistosomes, Liver Flukes and Chromium, Nickel and Welding Meiicobacter pylon Volume 37 1990; 677 pages; ISBN: 92 832 1249 5 1994; 280 pages; ISBN 92 832 1261 4 Tobacco Habits Other than Smoking; Betel-0uid and Areca-Nut Chewing; Volume 50 Volume 62 and Some Related Nitrosamines Some Pharmaceutical Drugs Wood Dusts and Formaldehyde 1985; 291 pages; ISBN 92 832 1537 0 1990; 415 pages; ISBN: 9283212599 1995; 405 pages; ISBN 92 832 1262 2

Volume 38 Volume 51 Volume 63 Tobacco Smoking Coffee, Tea, Mate, Methyixanthines Dry cleaning, Somв Chlorinated 1986; 421 pages; ISBN 92 832 1538 9 and Methylglyoxal Solvents and Other Industrial 1991; 513 pages; ISBN: 92 832 1251 7 Chemicals Volume 39 1995; 558 pages; ISBN 92 832 1263 0 Some Chemicals Used in Plastics Volume 52 and Elastomers Chlorinated Drinking-Water; Volume 64 1986; 403 pages; ISBN 92 832 12398 Chlorination By-products; some Human Papillomaviruses other Halogenated Compounds; 1995; 409 pages; ISBN 92 832 1264 9 Volume 40 Cobalt and Cobalt Compounds Some Naturally Occurring and 1991; 544 pages; ISBN: 92 832 1252 5 Volume 65 Synthetic Food Components, Printing Processes, Printing Inks, Furocoumarins and Ultraviolet Volume 53 Carbon Blacks and Some Nitro Radiation Occupational Exposures in Compounds 1986; 444 pages; ISBN 92 832 1240 1 Insecticide Application, and Some 1996; 578 pages; ISBN 92 832 12657 Pesticides Volume 41 1991; 612 pages; ISBN 92 832 12533 Volume 66 Some Halogenated Hydrocarbons Some Pharmaceutical Drug arid Pesticide Exposures Volume 54 1996; 514 pages; ISBN 92 832 12665 1986; 434 pages; ISBN 92 832 1243 X Occupational Exposures to Mists and Vapours from Strong Inorganic Acids; Volume 67 Volume 42 and other lndustnaI Chemicals Human Immunodeficiency Viruses arid Silica and Some Silicates 1992; 336 pages; ISBN 92 832 1254 1 Human T-ceIl Lymphotropic Viruses 1987; 289 pages; ISBN 92 632 1242 8 1996; 424 pages; ISBN 92 832 12673 Volume 55 Volume 43 Solar and Ultraviolet Radiation Volume 68 Man-Made Mineral Fibres and Radin 1992; 316 pages; ISBN 92 832 1255 X Silica, Some Silicates, Coal Dust end 1988; 300 pages; ISBN 92 832 12436 para-Aramid Fibrils Volume 56 1997; 506 pages; ISBN 92 832 1268 1 Volume 44 Some Naturally Occurring Alcohol Drinking Substances: Food Items and Volume 69 1988; 416 pages; ISBN 92 832 12444 Constituents, Heterocyclic Aromatic Polychlorinated Dibenzo-para-dioxins Amines and Mycotoxins and Dibenzoturans Volume 45 1993; 600 pages; ISBN 92 832 1256 8 1997; c. 600 pages; ISBN 92 832 1269 X Occupational Exposures in Petroleum Refining; Crude Oil and Major Volume 57 Supplements Petroleum Fuels Occupational Exposures of Hairdressers Suррlеmeпt Na.1 1989; 322 pages; ISBN 92 832 1245 2 and Barbers and Personal Use of Chemicals and Industrial Processes Hair Colourants; Some Hair Dyes, Associated with Cancer in Humans Volume 46 Cosmetic Colourants, Industrial ('ARC Monographs, Volumes 1 to 20) Diesel and Gasoline Engine Exhausts Dyestuffs and Aromatic Amines 1979; 71 pages; ISBN 92 832 14048 and Some Nitroarenes 1993; 428 pages; ISBN 92 832 1257 6 (out of print) 1989; 458 pages; ISBN 92 832 1246 0 Volume 58 supplement No. 2 Volume 47 Beryllium, Cadmium, Mercury and Long-Term and Short-Term Some Organic Solvents, Resin Exposures in the Glass Screening Assays for Carcinogens: A Monomers and Related Compounds, Manufacturing Industry Critical Appraisal Pigments and Occupational Exposures 1994; 444 pages; ISBN 92 832 12584 1980; 426 pages; ISBN 92 832 1404 8 Supplement No. 3 Supplement No. 5 Supplement No. 7 Cross Index of Synonyms and Trade Cross Index of Synonyms and Trade Overall Evaluations of Names in Volumes 1 to 26 Names in Volumes 1to 36 Carcinogenicity: An Updating 1982; 199 pages; ISBN 92 832 14056 1985; 259 pages; ISBN 92 832 14080 of 'ARC Monographs (out of print) (out of print) Volumes 1 to 42 1987; 440 pages; ISBN 92 832 1411 0 Supplement No.4 Supplement No. 6 Chemicals, Industrial Processes and Genetic and Related Effects: Supplement No. 8 Industries Associated with Cancer in An Updating of Selected 'ARC Cross Index of Synonyms Humans (Volumes 1 to 29) Monographs from and Trade Names in Volumes 1 to 46 1982:292 pages; 1SВN 92 832 14072 Volumes Ito 42 (out of print) 1987; 729 pages; ISBN 92 832 1409 9 1989; 346 pages; ISBN 92 832 1417 X

'ARC Scientific Publications

No. 1 No. 11 Methods of Analysis. Volume 1: Analysis Liver Cancer Oncogenesis and Herpes-viruses Il of Volatile Nitrosamines iп Food 1971; 176 pages; ISBN 0 19 723000 8 Edited by G. de-Thé, M.A. Epstein and Editor-in-Chief: H. Egan H.zur Hausen 1978:212 pages; ISBN 0 197230172 No. 2 1975; Two volumes, 511 pages and Oncogwesis and Herpesviruses 403 pages; ISBN 0 19 723010 5 No. 19 Edited by P.M. Biggs, G. de Thé and Environmental Aspects of N-Nitroso L.N. Payne No. 12 Compounds 1972; 515 pages; ISBN 0 19 723001 6 screening Tests in Chemical Edited by E.A. Walker, M. Castegnaro, Carcirsogenesis L. Griciute and R.E. Lyle No. 3 Edited by R. Montesano, H. Bartsch and 1978; 561 pages; ISBN 0 19 723018 0 N-Nitroso Compounds: Analysis and L. Tomatis Formation 1976; 666 pages; ISBN 0 19 723051 2 No. 20 Edited by P Bogovski, R. Preussman Nasopharyngeal Carcinoma: Etiology and E.A. Walker No. 13 and Control 1972; 140 pages; ISBN 0 19 723002 4 Environmental Pollution and Edited by G. de Thé and Y. Ito Carcinogenic Risks 1978; 606 pages; ISBN 0 19 723019 9 No. 4 Edited by C. Rosenfeld and W. Davis Transplacental Carcinogenesis 1975; 441 pages; ISBN 0197230121 No. 21 Edited by LTomatis and U. Mohr Cancer Registration and its 1973; 181 pages; ISBN 0 1 9 7230032 No. 14 Techniques Environmental N-Nifroso Compounds. Edited by R. MacLennan, C. Muir, No. 6/6 Analysis aid Formation R. Steinitz and A. Winkler Pathology of Tumours in Laboratory Edited by E.A. Walker, P. Bogovski and 1978; 235 pages; ISBN 0 19 7230202 Animals. Volume 1: Tumours of the Rat L. Griciute Edited by V.S. iurusov 1976; 512 pages; ISBN 0 19723013 X No. 22 197311976; 533 pagis; ISBN 92 832 14102 Environmental Carcinogens: No. 15 selected Methods of Analysis. No. 7 Cancer Incidence in Five Continents, Volume 2: Methods for the Host Environment Interactions in the Volume III Measurement of Vinyl Chloride in Etiology of Cancer in Man Edited by J.A.H. Waterhouse, Pofy(vinyl chloride), Air, Water and Edited by Ft. Doll and I. Vodopija C. Muir, R Correa and J. Powell Foodstuffs 1973; 464 pages; ISBN 0 19 723006 7 1976; 584 pages; ISBN 0 19 723014 8 EdItor-in-Chief: H. Egan 1978; 142 pages; ISBN 0 19 723021 0 No. 8 No. 16 Biological Effects of Asbestos Air Pollution and Cancer in Man No. 23 Edited by P. Bogovski, J.C. Gilson, Edited by U. Mohr, D. Schmdhl and Pathology of Tumours in Laboratory V. Timbrell and J.C. Wagner L. Tomatis Animais. 1973; 346 pages; ISBN 0 197230075 1977; 328 pages; ISBN 0 19 723015 6 Volume II: Tumours of the Mouse Nn. 9 No. 17 Editor-in-Chief: V.S. Turusov N-Nitroso Compounds in the Environment Directory of On-Going Research in 1979; 669 pages; ISBN 0 19 723022 9 Edited by P. Bogovski and E.A. Walker Cancer Epidemiology 1977 1974; 243 pages; ISBN 0 19 723008 3 Edited by C.S. Muir and G. Wagner No. 24 1977:599 pages; ISBN 92 832 11170 Oncogenesis and Herpesviruses Ill No. 10 (out of print) Edited by G. de-Thé, W. lenin and Chemical Carcinogenesis Essays F. 'app Edited by R. Montesano and L. Tomatis No. 18 1978; Part I: 580 pages, Part II: 512 pages; ISBN 0 197230237 1974; 230 pages; ISВN 0 19723009 1 Environmental Carcinogens. Selected xi No. 25 No. 34 Edited by J. Waterhouse, C. Muir, Carcinogenic Risk: Strategies for Pathology of Tumours in Laboratory K. Shanmugaratnam Intervention Animals. Volume III: Tumours of the and J. Powell Edited by W. Davis and C. Rosenfeld Hamster 1982; 811 pages; ISBN 0 19 723042 3 1979; 280 pages; ISBN 0 19 723025 3 Editor-in-Chief: V.S. Turusov 1982; 461 pages; ISBN 0 19 723034 2 No. 43 No, 26 Laboratory Decontamination and Directory of On-going No. 35 Destruction of Carcinogens in Research in Cancer Epidemiology Directory of On-going Research in Laboratory Wastes: Some 1978 Cancer Epidemiology 1980 N-Nitrosamines Edited by C.S. Muir and G. Wagner Edited by C.S. Muir and G. Wagner Edited by M. Castegnaro, 1978; 550 pages; ISBN 0 19 723026 1980; 660 pages; ISBN 0197230350 G. Eisonbrand, G. Ellen, L. Keefer, (out of print) (out of print) D. Klein, E.B. Sansone, D. Spincer, G. Telling and K. Webb No.27 No. 36 1982; 73 pages; 1SBN 0 19 723043 Molecular and Cellular Aspects of Cancer Mortality by Occupation and Carcinogen Screening Tests Social Class 1851--1971 No. 44 Edited by R. Montesano, H. Bartsch and Edited by W.P.D. Logan Environmental Carcinogens: Selected L. Tomatis 1982; 253 pages; ISBN 0 19 723036 9 Methods of Analysis. 1980; 372 pages; ISBN 0 19 723027 X Volume 5: Some Mycotoxins No. 37 Edited by L. Stoloff, M. Castegnaro, No.28 Laboratory Decontamination and P. Scott, 1.K. O'Neill and H. Bartsch Directory of On-going Research in Destruction of Af[atoxins B1, B2, G1, 1983; 455 pages; [SBN 0 19 723044 X Cancer Epidemiology 1979 G2 in Laboratory Wastes Edited by C.S. Muir and G. Wagner Edited by M. Castegnaro, D.C. Hunt, No. 45 1979; 672 pages; ISBN 92 832 11286 E.B. Sansone, P.L. Schul!er, Environmental Carcinogens: Selected (out of print) M.G. Siriwardana, G.M. Telling, Methods of Analysis. H.P. van Egmond and E.A. Walker Volume 6: N-Nitroso Compounds No. 29 1980; 56 pages; ISBN 0 19 7230377 Edited by R. Preussmann, Environmental Carcinogens. Selected I.K. O'Neill, G. Eisenbrand, Methods of Analysis. Volume 3: Analysis Ni. 38 B. SpiegeihaIder and H. Bartsch of Polycyclic Aromatic Hydrocarbons Directory of On-going Research in 1983; 508 pages; ISBN 0 19 723045 8 in Environmental Samples Cancer Epidemiology 1981 Editor-in-Chief: H. Egan Edited by C.S. Muir arid G. Wagner No. 46 1979; 240 pages; ISBN 0197230288 1981; 696 pages; ISBN 0 19 723038 5 Directory of On-going Research in (out of print) Cancer Epidemiology 1982 No. 30 Edited by 0.5. Muir and G. Wagner Biological Effects of Mineral Fibres No. 39 1982; 722 pages; ISBN 0 19 723046 6 Editor-in-Chief: J.C. Wagner Host Factors in Human (out of print) 1980; Two volumes, 494 pages & Carcinogenesis 513 pages; ISBN 0 19 723030 X Edited by H. Bartsch and B. Armstrong No. 47 1982; 583 pages; Cancer Incidence in Singapore 1968-1977 No. З1 ISBN 0 19 723039 3 Edited by K. Shanmugaratnam, N-Nitroso Compounds: Analysis, H.P. Lee and N.E. Day Formation and Occurrence No. 40 1983; 171 pages; ISBN 0 19 7230474 Edited by E.A. Walker, L. Griciute, Environmental Carcinogens: M. Castegnaro and M. Bbrzsbnyi Selected Methods of Analysis. No. 48 1980; 835 pages; ISBN 0 19 723031 8 Volume 4: Some Aromatic Cancer Incidence in 'he USSR (2nd Amines and Azo Dyes in the Revised Edition) No. 32 General and Industrial Edited by N.P. Napalkov, G.F Tserkovпц Statistical Methods in Cancer Environment V.M. Merabishvili, D.М. Parkin, Research.Volume 1: The Analysis of Edited by L. Fishbein, M. Castegnard, M. Smans and C.S. Muir Case-control Studies I.K. O'Neill and H. Bartsch 1983; 75 pages; ISBN 0 19 723048 2 By N.E. Broslow and N.Е. Day 1981; 347 pages; ISBN 0 19 723040 7 1980; 338 pages; ISBN 92 832 0132 9 No. 49 Ni. 41 Laboratory Decontamination No. 33 N-Nilroso Compounds: Occurrence and Destruction of Carcinogens Handling Chemical Carcinogens in and Biological Effects in Laboratory Wastes: the Laboratory Edited by H. Bartsch, K. O'Neill, Some Polycyclic Aromatic Edited by R. Montesano, H. Bartsch, M. Castegnaro and M. Okada Hydrocarbons E. Boyland, G. Della Porta, L. Fishbeio, 982;755 pages; ISBN 0 19 723041 5 Edited by M. Castegnaro, G. Grimmer, R.A. Griesemer, A.B. Swan and О. Hutzinger, W. Karchвr, H. Kunte, L. Tomatis No. 42 M. Lafontaine, H.C. Van der P1as, 1979:32 pages; ISBN 0 19 723033 4 Cancer Incidence in Five Continents E.B. Sansone and S.P. Tucker (out of print) Volume IV 1983; 87 pages; ISBN 0 19 723049 0 No. 50 R. Montesano No. 67 Directory of On-going Research in 1985; 288 pages; ISBN 92 832 11588 Transformation Assay of Established Cancer Epidemiology 1983 Cell Lines: Mechanisms and Edited by C.S. Muir and G. Wagner No. 59 Application 1983; 731 pages; ISBN 0 19 723050 4 Monitoring Human Exposure to Edited by T Kakunaga and H. Yamasaki (out of print) Carcinogenic aid Mutagenic Agents 1985; 225 pages; ISBN 92 832 11677 Edited by A. Berlin, M. Draper, No. 51 K. Hemminki and H. Valuo No. 68 Modulators of Experimental 1984; 457 pages; ISBN 0 19 723056 3 Environmental Carcinogens: Selected Carcinogenesis Methods of Analysis. Volume 7: Some Edited by V. Turusov and R. Montesano No. 60 Volatile Halogenated Hydrocarbons 1983; 307 pages; ISBN 0 19 723060 1 Burkitt's Lymphoma: A Human Edited by L. Fishbein and I.K. l'Neii Cancer Model 1985; 479 pages; ISBN 92 832 11685 No.52 Edited by G. Lenoir, G. O'Coпor and Second Cancers in Relation to Radiation C.L.M. Olweny No. 69 Treatment for Cervical Cancer: Results 1985; 484 pages; ISBN 0 19 723057 1 Directory of On-going Research in of a Cancer Registry Collaboralion Cancer Epidemiology 1985 Edited by N.E. Day and J.C. Bocce, Jr No. 61 Edited by C.S. Muir and G. Wagner 1984; 207 pages; ISBN 0 19 723052 0 Laboratory Decontamination 1985; 745 pages; ISBN 92 823 11693 and Destruction of Carcinogens in (out of print) No. 53 Laboratory Wastes: Some Haloethers Nickel in the Human Environment Edited by M. Castegnaro, M. Alvarez, M. No. 70 Editor-in-Chief: F.W. Sunderman, Jr lovu, E.B. Sansone, G.M. Telling and The Role of Cyclic Nucleic Acid 1984; 529 pages; ISBN 0 19 723059 8 D.Т. WIHiama Adducts in Carcinogenesis and 1985; 55 pages; ISBN 0 19 723061 X Mutagenesis No. 54 Edited by B. Singer and H. Bartsch Laboratory Decontamination and No. 62 1986; 467 pages; ISBN 92832 11707 Destruction of Carcinogens in Directory of On-going Research in Laboratory Wastes: some Hydrazines Cancer Epidemiology 1984 No. 71 Edited by M. Castegnaro, G. Ellen, Edited by C.S. Muir and G. Wagner Environmental Carcinogens: Selected M. Lafontaine, H.C. van der Fias, 1984; 717 pages; ISBN 0197230628 Methods of Analysis. Volume 8: Some E.B. Sansone and S.P. Tucker (out of print) Metals: As, Be, Cd, Cr, Ni, Pb, Se, Zn 1983; 87 pages; ISBN 0 19 723053 Edited by I.K. O'Neill, R Schuller and No. 63 L. Fishbein No. 55 Virus-associated Cancers in Africa 1986; 485 pages; ISBN 92 832 1171 5 Laboratory Decontamination and Edited by A.O. Williams, G.T. l'Conor, Destruction of Carcinogens in G.B. de Thé and C.A. Johnson No. 72 Laboratory Wastes: Some 1984; 773 pages; ISBN 0 19 723063 6 Atlas of Cancer in Scotland,197Cx19B0: N-Nftrosamides Incidence and Epidemiological Edited by M. Castegnaro, No. 64 Perspective M. Bernard, L.W. van Broekhoven, Laboratory Decontamination Edited by I. Kemp, P. Boyle, M. Smans D. Fine, R. Massey, E.B. Sansone, and Destruction of Carcinogens and C.S. Muir P.L.R. Smith, B. Spiegelhalder, in Laboratory Wastes: 1985; 285 pages; ISBN 92 832 1172 3 A. Stacchini, G. Telling and J.J. Vallon Some Aromatic Amines and 4- 1984; 66 pages; ISBN 0 19 7230547 Nitrobiphenyl No. 73 Edited by M. Castegnaro, J. Barek, Laboratory Decontamination No. 56 J. Dennis, G. Ellen, M. Klibanov, and Destruction of Carcinogens Models, Mechanisms and Etiology of M. Lafontaine, R. Mitchum, in Laboratory Wastes: Tumour Promotion P. van Roosmalen, E.B. Sansone, Some Antineoplastic Agents Edited by M. Bбrzsôпуi, N.E. Day, L.A. Sternson and M. Vah1 Edited by M. Castegnaro, J. Adams, K. Lapis and H. Yamasaki 1985; 84 pages; ISBN: 92 832 11642 M.A. Armour, J. Barek, J. Benvenuto, 1984; 532 pages; ISBN 019 723058 X C. Confalonieri, U. Golf, G. Telling No. 65 1985; 163 pages; ISBN 92 832 1173 No. 57 Interpretation of Negative N-Nitroso Compounds: Occurrence, Epidemiological Evidence for No. 74 Biological Effects and Relevance to Carcinogenicity Tobacco: A Маjoг Inlerniational Health Human Cancer Edited by N.J. Wald and R. Doll Hazard Edited by I.K. O'Neill, R.C. von Borstel, 1985; 232 pages; ISBN 92 832 11650 Edited by D. Zaridze aid R. Peto C.T. Miller, J. Long and H. Bartsch 1986; 324 pages; ISBN 92 832 1174 X 1984; 1013 pages; ISBN 0 19 723055 5 No. 66 The Role of the Registry in Cancer No. 75 No 58 Control Cancer Occurrence in Developing Age-related Factors in Edited by D.M. Parkin, G. Wagner and Countries Carcinogenesis C.S. Muir Edited by D.M. Parkin Edited by A. Likhachev, V. Anisimov and 1985; 152 pages; ISBN 92 832 0166 3 1986; 339 pages; ЭSBN 92 832 11758 No. 76 No. 85 No. 94 screening for Cancer of the Uterine Environmental Carcinogens: Methods Human Papillomavirus and Cervical Cervix of Analysis and Exposure Cancer Edited by M. Hakama, А.B. Miller and Measurement. Volume 10: Benzene Edited by N. Munoz, EX. Bosch and N.E. Day and Aikylated Benzenes O.M. Jensen 1986; 315 pages; ISBN 9283211766 Edited by L. Fishbein and I.K. O'Neill 1989; 154 pages; ISBN 92 832 11944 1988; 327 pages; ISBN 92 832 11855 No. 77 No. 95 Hexachlorobenzene: Proceedings of No. 86 Cancer Registration: Principles and an International Symposium Directory of On-going Research in Methods Edited by C.R. Morris and J.R.P. Cabra Cancer Epidemiology 1987 Edited by O.M. Jensen, D.M. Parkin, 1986; 668 pages; 1SВN 92 832 11774 Edited by D.M. Parkin and J. Wahrendort R. MacLennan, C.S. Muir and R. Skeet 1987; 685 pages; ISBN: 92 832 11863 1991;296 pages; ISBN 92 832 1195 2 No. 78 (out of print) Carcinogenicity of A1kylating No. 96 Cyloslatic Drugs No. 87 Pen natal and Multigeneration Edited by D. Schmhl and J.M. Kaldor International Incidence of Childhood Carcinogenesis 1986; 337 pages; ISBN 9283211782 Cancer Edited by N.P. Napalkov, J.M. Rice, Edited by D.M. Parkin, C.A. Stiller, L. Tomatis and H. Yamasaki No. 79 C.А. Bieber, G.J. Draper. B. Terracini and 1989; 436 pages; 1S8N 92 832 11960 Statistical Methods in Cancer Research. J.L. Young Volume III: The Design and Analysis 1988; 401 page; 1SBN 92 832 1187 1 No. 97 of Long-term Animal Experiments (out o1 print) Occupational Exposure to Silica and By J.J. Gart, D. Krewski, RN. Lee, Cancer Risk R.E.Ta one and J. Wahrendorf г No. 88 Edited by L. Simonаtо, A.C. Fletcher, 1986; 213 pages; ISBN 92 832 11790 Cancer Incidence in Five Continents, R. Saracci and T. Thomas Volume V 1990; 124 pages; ISBN 9283211979 No. 80 Edited by C. Muir, J. Waterhouse, Directory of On-going Research in T. Mack, J. Powell and S. Whelan No. 98 Cancer Epidemiology 1966 1987; 1004 pages; ISBN 92 832 1186 X Cancer Incidence in Jewish Migrants Edited by C.S. Muir and G. Wagner 10 Israel, 1961-1981 1986; 805 pages; ISBN 92 832 1180 4 No. 89 Edited by R. Sleinitz, D.М. Parkin, (out of print) Methods for Detecting DNA Damaging J.L. Young, C.A. Bieber and L. Katz Agents in Humans: Applicatïons in 1989; 320 pages; ISBN 92 832 11987 No. 8i Cancer Epidemiology and Prevention Environmental Carcinogens: Methods Edited by H. Bartsch, K. Hemminki and No. 99 of Analysis and Exposure LK. O'Neill Pathology of Tumours in Measurement. Volume 9: Passive 1988; 518 pages; ISBN 92 832 11898 Laboratory Animals, Second Edition, Smoking (out of print) Volume 1,Tumours of the Rat Edited by I.K. O'Neill, K.D. Brunnemani, Edited by V.S. Turusov B. Dodet and D. Hoffmann Ni. 90 and U. Mohr 1987; 383 pages; ISBN 92 В32 1181 2 Non-occupational Exposure to 1990; 740 pages; ISBN 92 832 11995 Mineral Fibres For Volumes 2 and 3 No. 82 Edited by J. Bignon, J. Peto and R. Saracci (Tumours of the Mouse and Tumours of Statistical Methods in Cancer 1989; 500 pages; ISBN 92 832 1190 1 the Hamster), see Research. Volume 11: The Design and lАRC ScieпЁЮс Publications Analysis of Cohort Studies No. 9 i Nos. III ап d 126. By N.E. Breslow and N.E. Day Trends in Cancer Incidence in 1987; 404 pages; ISBN 92 832 0182 5 Singapore 1966-1982 No. 100 Edited by H.E Lee, N.E. Day and Cancer: Causes, Occurrence and No. 83 K. Shanmugaratnam Control Long-term and Short-term Assays 1988; 160 pages; ISBN 92 832 1191 X Editor-in-Chief: L. Tornatis for Carcinogens: 1990; 352 pages; ISBN 92 832 01108 A Critical Appraisal No. 92 Edited by R. Montesano, H. Bartsch, Cell Differentiation, Genes and Cancer No. 101 H. Vainio, J. Wilbourn and H. Yamasaki Edited by T. Kakunaga, T. Sugimura, Direclory of On-going Research in 1986; 575 pages; ISBN 92 832 11839 L.Torahs and H. Yamasaki Cancer Epidemiology 1989-1990 1986; 204 pages; ISBN 92 832 11928 Edited by M. Coleman and J. Wahrendorf No. 84 1989; 828 pages; ISBN 92 В32 2101 X The Relevance of N-Nitroso No. 93 Compounds 10 Human Cancer: Directory of On-going Research in No. 102 Exposure and Mechanisms Cancer Epidemiology 1988 Patterns of Cancer in Five Conlinents Edited by H. Bartsch, I.K. O'Neill and Edited by M. Coleman and J. Wahrendorf Edited by S.L. Whelan, D.M. Parkin and R. Schulte-Hermалп 1988; 662 pages; ISBN 92 832 11936 E. Masuyer 1987; 671 pages; ISBN 92 832 1 184 7 (out of print) 1990; 160 pages; ISBN 9263221028 xiv No. 103 No. 112 Y.T Gao, J. Ferlay and J. Powell Evaluating Effectiveness of Primary Autopsy in Epidemiology and 1992;1020 pages; ISBN 9283221206 Prevention of Cancer Medical Research Edited by M. Hakama, V. Beral, Edited by E. Riboli and M. Delendi No. 121 J.W. Cullen and D.М. Parkin 1991;288 pages; ISBN 92 832 2112 5 Time Trends in Cancer Incidence and 1990; 206 pages; ISBN 92 632 21036 Mortality No. 113 By M. Coleman, J. Esiévе, P. 6amiecki, Na. 104 Laboratory Decontamination and A. Arslan and H. Renard Complex Mixtures and Cancer Risk Destruction of Carcinogens ïn 1993; 820 pages; ISBN 92 832 2121 4 Edited by H. Vainio, M. 5orsa and Laboratory Wastes: Some Mycotoxins A.J. McMichael Edited by M. Castegnaro, J. Barek, No. 122 1996; 441 pages; ISBN 92 832 2104 4 J.M. Frémy, M. Lafontaine, М. MiragIia, International Classification of Rodent E.B. Sansone and G.M. Telling Tumours. No. 105 1991; 63 pages; ISBN 92 832 21 13 3 Part 1. The Rat Relevance to Human Cancer of Editor-in-Chief: U. Mohr N-Nitroso Compounds, Tobacco No. 114 1992-1996; 10 fascicles of 60-100 Smoke and Mycotoxins Laboratory Decontamination and pages; ISBN 92 832 2122 2 Edited by 1.K.O'Nei1I, J. Chen and H. Bartsch Destruction of Carcinogens in 1991; 614 pages; ISBN 92 832 2105 2 Laboratory Wastes: some Polycyclic No. 123 Heterocyclic Hydrocarbons Cancer in Italian Migrant Populations No. 106 Edited by M. Castegnaro, J. Barek, J. Jacob, Edited by M. Geddes, D.М. Parkin, Atlas of Cancer Incidence In the U. Kirso, M. Lafontaine, E.B. Sansone, M. Khlat, D. Balzi and E. Вuiattl Former German Democratic G.M. Telling aid T. Vu Duc 1993; 292 pages; ISBN 92 832 2123 0 Republic 1991; 50 pages; ISBN 92 832 2114 1 Edited by W.H. Mehnert, M. Smans, No. 124 C.S. Muir, M. Mühner and D. Schhn No. 115 Posgabешng Methods for the 1992; 384 pages; ISBN 92 832 21060 Mycotoxins, Endemic Nephropathy Detection of DNA Damage and Urinary Tract Tumours Edited by D.H. Phillips, No. 107 Edited by M. Castegnaro, R. Plestina, G. M. Castegnaro and H. Bartsch Atlas of Cancer Mortality in the Dirheimer, I.N. Chernozemsky and 1993; 392 pages; ISBN 92 832 2124 9 European Economic Community H. Bartsch No. 125 Edited by M. Smans, C. Мuir and P. Boyle 1991; 340 pages; ISBN 92 832 2115 X 1992; 213 pages + 44 coloured maps; DNA Adducts: Identification and ISBN 92 832 2107 9 No. 116 Biological significance Mechanisms of Carcinogenesis in Edited by K. Hem minki, A. Dipple, No. 108 Risk Identification D.E.G. Shuker, F.F. Kadlubar, Environmental Carcinogens: Edited by H. Vain ii, R Magee, D. Segerb5ck and H. Bartsch Methods of Analysis aid Exposure D. McGregor and A.J. McMichael 1994; 478 pages; ISBN 92 832 2125 7 Measurement. Volume 11: 1992; 615 pages; ISBN 92 832 21 16 8 Polychlorinated Dioxins and No. 126 Dibenzofurans No. 117 Pathology of Tumours in Laboratory Edited by C. Rappe, H.R. Ruser, Directory of On-going Research in Animals, second Edition. Volume З: B. Dodet and 1K. O'Neill Cancer Epidemiology 1992 Tumours of the Hamster 1991; 400 pages; ISBN 92 832 2108 7 Edited by M. Coleman, E. Demaret and Edited by V Turosov and U. Mohr J. Wahrendorr 1996; 464 pages; ISBN 92 832 2126 5 No. 109 1992; 773 pages; ISBN 92 932 2117 6 Environmental Carcinogens: Methods No. 127 of Analysis and Exposure No. 118 Butadiene and Styrene: Assessment Measurement. Volume 12: Indoor Air Cadmium in the Human Environment: of Health Hazards Edited by B. Seifert, H. van de Wie1, Toxicity and Carcinogenicity Edited by M. Sorsa, K. Peltonen, H. B. Dodet and I.K. O'Neill Edited by G.F Nordberg, R.F.М. Herber Vainio and K. Hemminki 1993; 385 pages; ISBN 92 832 2109 5 and L. Alessio 1993; 412 pages; ISBN 92 832 2127 3 1992; 470 pages; ISBN 92 832 2118 4 No. 110 No. 128 Directory of On-going Research in No. 119 Statistical Methods in Cancer Research. Cancer Epidemiology 1991 The Epidemiology of Cervical Cancer Volume IV. Descriptive Epidemiology Estève, E. Benhamou and L. Raymond Edited by М.R Coleman aid J. Wahrendo i and Human Papillomavirus By J. 1991; 753 pages; ISBN 9263221109 Edited by N. Munoz, F.X. Bosch, 1994; 302 pages; ISBN 92 832 2128 1 K.V. Shah and A. Meheus No. 111 1992; 288 pages; ISBN 9283221192 No. 129 Pathology of Tumours in Laboratory Occupalional Cancer in Developing Animals, Second Edition. Volume 2: No. 120 Countries Tumours of the Mouse Cancer Incidence in Five Continents, Edited by N. Pearce, E. Matos, Edited by V. Turusov and U. Mohr Vol. VI H. Vainio, P. Boffetta and M. Kogevinas 1994; 800 pages; ISBN 92 832 2111 1 Edited by D.M. Parkin, C.S. Muir, S.L. Whelan, 1994;191 pages; ISBN 92 832 2129 X

xv No. 130 No. 136 No, 140 Directory of On-going Research in Chemoprovention in Cancer Control Mechanisms of Fibre Cancer Epidemiology 1994 Edited by M. Hakama, V. Beral, Carcinogenesis Edited by R. 5ankaranarayanan, E. Buiatti, J. Faivre and D.M. Parkin Edited by A.B. Kane, P Bofletta, R. J. Wahrendorf and E. Détuaret 1996; 160 pages; ISBN 92 832 2136 2 Saracci and J.D. Wilbourn 1994; 800 pages; ISBN 92 832 21303 1996; 135 pages; ISBN 92 832 21 40 0 No, 137 No. 132 Directory of On-going Research in No. 142 Survival of Cancer Patients in Cancer Epidemiology 1996 Application of Biomarkers to Europe. The EUBOCARE Study Edited by R. San karanarayan, J. Cancer Epidemiology Edited by F. Berrino, M. Sant, 'Naroridorf and E. Démaret Edited by P. Toniolo, R Boffelta, D. A. Verdecchia, R. Capocaccia, 1996; 810 pages; ISBN 92 832 21370 Shuker, N. Rothman, B. Hulka T. Hakulinen and J. Estève and N. Pearce 1995; 463 pages; ISBN 92 832 2132 X No. 138 1997; 336 pages; ISBN 92 832 2142 Social Inequalities and Cancer No. 134 Edited by M. Kogevinas, N. Pearce, M. No. 143 Atlas of Cancer Mortality in Susser and P. Boffetta Cancer Incidence in Five Central Europe 1997; 412 pages; ISBN 92 832 2138 9 Continents, Vol. VIII W. Zatonski, J. Estéve, м . Smala, Edited by D.M. Parkin, S.L. J. Tyczynski and P Boyle No. 139 Whelan, J. Ferlay, L. Raymond 1996; 300 pages; ISBN 92 832 21 34 6 Principles of Chemoprevention and J. Young Edited by B.W. Stewart, D. McGregor 1997; 1350 pages; ISBN 92 832 2143 5 No. 135 and P. Kleihues Methods for Investigating Localized 1996; 358 pages; ISBN 92 83221397 Clustering of Disease Edited by ~.Е. Alexander and P. Boyle 1996; 235 pages; ISBN 92 832 21354 IARC Technical Reports

No. 1 No. 6 No. 11 Cancer in Costa Rica La genèse du Centre international de Nitroso Compounds: Biological Edited by R. Sierra, R. Barrantes, recherche sur le cancer Mechanisms, Exposures and Cancer G. Munoz Leiva, D.M. Parkin, By R. 5оhier and A.G.B. Sutherland Etiology C.A. Bieber and N. Munoz Calero 1990, 102 pages; ISBN 92 832 14188 Edited by I. O'Neill and H. Bartsch 1988; 124 pages; 1992; 150 pages; ISBN 92 832 1425 X ISBN 92 832 1412 9 No. 7 Epidémио1ogïe du cancer dans les No, 12 No. 2 pays de langue latine Epidëmiologiе du cancer dans les SEARCH: A Computer Package to 1990, 292 pages; ISBN 92 832 1419 6 pays de langue latine Assist the Statistical Analysis of 1992; 375 pages; ISBN 92 832 14269 Case-Control Studies No. 8 Edited by G.J. Macfarlane, P. Boye and Comparative Study of Anti-smoking No. 13 P Maisonneuve Legislation in Countries of the Health, Solar UV Radiation and 1991; 80 pages; ISBN 9283214137 European Economic Community Environmental Change By A. J. Sasco, R Dalla-Vorgia and By A. Kricker, B.K. Armstrong, No. 3 R Van der Flat М.E. Jones and R.C. Burton Cancer Registration in the European 1992; 82 pages; ISBN: 92 832 1421 8 1993; 213 pages; ISBN 92 832 1427 7 Economic Community Etude comparative des Législations Edited by M.P. Coleman and E. Démaret de Contrôle du Tabagisme dans les No. 14 1988; 188 pages; ISBN 92 832 14145 Pays de la Communauté ëconomique Еpidémiologie du cancer dans les еигорёеппе pays de langue latine No. 4 1995; 82 pages; ISBN 92 832 2402 7 1993; 400 pages; ISBN 92 832 14285 Diet, Hormones and Cancer: Methodological Issues for No. 9 No. 15 Prospective Studies Epidémiologie du cancer dans les Cancer in the African Population of Edited by E. Riboli and R. Saracci pays de langue latine Bulawayo Zimbabwe, 1963-1977 1988; 156 pages; ISBN 92 832 1415 3 1991; 346 pages; ISBN 92832 14234 By M.E.G. Skinner, D.M. Parkin, А.P. Vizcaino and A. Ndhlovu No. 5 No. 10 1993; 120 pages; ISBN 92 832 14293 Cancer in the Philippines Manual for Cancer Registry Personne' Edited by A.V. Laudtco, D. Esteban and Edited by D. Esteban, S. Whelan, Ni. 16 D.M. Parkin A. Laudico and D.M. Parkin Cancer in Thailand 1984-1891 1989; 186 pages; ISBN 92 832 1416 1 1995; 400 pages; ISBN 92 832 14242 By V. Vatanasapt, N. Martn,

Xvi H. Sriplung, K. Chindavi)ak, S. sontipong, Cancer in Tianjin No. 26 S. Sriamporn, D.M. Parkin and J. Ferlay By Q.S. Wang, P. Boffetta, Mortalité par Cancer des lmigrés en 1993; 164 pages; 15BN 92 832 14307 M. Kogevinas and D.M. Parkin France, 1979-1985 1994; 96 pages; ISBN 92 832 1433 1 By C. Bouchardy, M. Khlat, i . Wanner No. 18 and D.M. Parkin Intervention Trials tar Cancer Prevention No. 23 1997; 150 pages; ISBN 92 832 2404 3 By E. Buiatti An Evaluation Programme for Cancer 1994:52 pages; ISBN 92 832 14323 Preventive Agents No. 27 By Bernard W. Stewart Cancer in Three Generations of Young No. 19 1995; 40 pages; ISBN 92 832 14382 Israelis Comparability and Quality Control in By J. lscovich and D.M. Parkin Cancer Registration No. 24 1997; 150 pages; ISBN 92 832 2441 2 By D.M. Parkin, V.W. Chen, J. Ferlay, Peroxisome Proliferation and its Role J. Galceran, H.H. Storm and S.L. Whelan in Carcinogenesis No. 29 1994; 110 pages plus diskette; 1995; 85 pages; ISBN 92 832 1439 0 International Classification of Childhood ISBN 92 832 1433 1 Cancer 1996 Ni. 25 By E. Kramarova, C.A. Stiller, J. Ferlay, No. 20 Combined Analysis of Cancer D.M. Parkin, G.J. Draper, J. Michaelis, J. Neglia and S. Qurechi Ep1dërmolоgie du cancer Mortality in Nuclear Workers in dans les pays de langue Canada, the United Kingdom and the 1996:48 pages + diskette; latine United States of America ISBN 92 832 14439 1994; 346 pages; ISBN 92 832 1434 X By E. Cardis, E.S. Gilbert, L. Carpenter, G. Howe, I. Kato, J. Fix, L. No. 21 Salmon, G. Cowper, B.К. Armstrong, CD Conversion Programs for Cancer V. Beral, A. Douglas, S.A. Fry, J. Kaldor, By J. Ferlay C. Lavé, P.G. Smith, G. Voelz and 1994; 24 pages plus diskette; L. Wiggs ISBN 92 832 1435 8 1995; 160 pages; ISBN 92 832 1440 4 No. 22

'ARC CaecerBase

No. 1 No. 2 EUCAN9о: Cancer in the European CI5VI1 Electronic Database of Cancer Union (Electronic Database with Graphic Incidence in Five Continents. Vol.I1 Display) By J. Ferlay, R.J. Black, S.L. Whelan By J. Ferlay, R.J. Black, P. Pisani, M.T. and D.M. Parkin Valdivieso and D.M. Parkin 1997; Computer sot iware on 3.5" IBM 1996; Computer software on 3.5° IBM diskettes + user's guide (c. 48 pages); diskette + user's guide (50 pages); ISBN 92 832 1449 8 ISBN 92 832 1450 1

All 'ARC Publications are available directly from IARCPress,150 Cours Albert Thomas, F-69372 Lyon cedex 08, France (Fax: +33 4 72 73 83 02; E-mail: [email protected]).

'ARC Monographs and Technical Reports are also available from the World Health Organization Distribution and Salвs, CH-1211 Geneva 27 (Fax: +41 22 791 4857) and from WHO Sales Agents worldwide.

'ARC 5c1entifk Publications are also available from Oxford University Press, Walton Street, Oxford, UK 0X2 GDP (Fax: X44 1865 267782).