Surrogate Endpoint Biomarkers for Cervical Cancer Chemoprevention Trials

Journal of Cellular Biochemistry, Supplement 23:113-124 (1 995)

Mack T. Ruffin IV, MD, MPH: Mohammed S. Qgaily, MD: Carolyn M. Johnston, MD? Lucie Gregoire, PhD: Wayne D. Lancaster, PhD; and Dean E. Brenner, MD6

Department of Family Practice, University of Michigan Medical Center, Ann Arbor, MI 48109-0708 Department of Internal Medicine, Division of Hematology and Oncology, Simpson Memorial Institute, Ann Arbor, MI 48109-0724 Department of Obstetrics and Gynecology, Division of Gynecologic Oncology, Ann Arbor, MI 48109-0718 Department of Obstetrics and Gynecology, Wayne State University School of Medicine, Detroit, MI 48201 Department of Obstetrics and Gynecology, Center for Molecular Medicine and Genetics, Detroit, MI 48201 Department of Internal Medicine, Division of Hematology and Oncology, Simpson Memorial Institute, Ann Arbor, MI 48109-0724

Abstract Cervical intraepithelial neoplasia (CIN) represents a spectrum of epithelial changes that provide an excellent model for developing chemopreventive interventions for cervical cancer. Possible drug effect surrogate endpoint biomarkers are dependent on the agent under investigation. Published and preliminary clinical reports suggest retinoids and carotenoids are effective chemopreventive agents for CIN. Determination of plasma and tissue pharmacology of these agents and their metabolites could serve as drug effect intermediate endpoints. In addition, retinoic acid receptors could serve as both drug and biological effect intermediate endpoints. Possible biological effect surrogate endpoint biomarkers include cytomorphological parameters, proliferation markers, genomic markers, regulatory markers, and differentiation. Given the demonstrated causality of human papillomavirus (HPV) for cervical cancer, establishing the relationship to HPV will be an essential component of any biological intermediate endpoint biomarker. The pathologic effect surrogate endpoint biomarker for cervical cancer is CIN, used clinically for years. The desired effect for chemopreventive trials is complete regression or prevention of progression. In planning chemopreventive trials, investigators need to consider spontaneous regression rates, the subjective nature of detecting CIN, and the impact of biopsy on regression. If intermediate endpoint biomarkers that met the above criteria were available for cervical cancer, then new chemopreventive agents could be rapidly explored. The efficacy of these new agents could be determined with a moderate number of subjects exposed to minimal risk over an acceptable amount of time. The impacts on health care for women would be significant. 0 1995 Wiley-Liss, Inc.

Send correspondence to Dean E. Brenner, MD, Department of Internal Medicine, Division of Hematology and Oncol- ogy, Simpson Memorial Institute, 102 Observatory Street, Ann Arbor, MI 48109-0724. 0 1995 Wiley-Liss, Inc. 114 Ruffin et al.

Key words: Aneuploidy, cervical intraepithelial neoplasia (CIN), cytophotometry dysplasia, HPV morphometry, oncogenes, tumor markers

CLINICAL IMPACT OF CERVICAL CANCER great majority of all grades of CW can be attributed to HPV infection and that HPV infection Cervix carcinoma is an important and feasible meets epidemiologic criteria as a causal agent for target for chemoprevention efforts. This common cervical cancer. malignancy [I], while not lethal at early stages of the disease in most cases, occurs in young, child- bearing, sexually active women. Carcinoma is ISSUES FOR CHEMOPREVENTION TRIALS thought to result from the progression of cervical intraepithelial neoplasia (CIN). Cancer chemoprevention trials pose unique Cervix carcinoma is an important health prob- problems in design. Chemoprevention agents are lem world-wide [l].In the U.S., 15,800 new cases aimed at a healthy population. Therefore, toxicity of invasive cervix cancer are predicted in 1995 is intolerable. An effective dose must be deter- with 4,800 deaths attributable to this disease mined that is not toxic. The dose must be toler- (1.8% of all cancer-related deaths in women) C23. ated on a chronic basis over a long period These figures do not include the more than (years). Furthermore, in many premalignant con- 50,000 cases of carcinoma in situ (CIS) and many ditions or high-risk individuals, there is no easily times that number of cases of cervical dysplasia identifiable therapeutic endpoint. Endpoints for for which we have no estimates for incidence or cancer clinical trials are usually a measured re- prevalence. The age-adjusted death rate for cervi- duction in tumor size or a statistically measured cal carcinoma in the U.S. is 3.2 per 100,000 and survival in a population whose survival is lim- remains level [3]. Anecdotally, we have observed ited. For most chemopreventive interventions, no in our clinics an increased number of younger such easily measured endpoints exist. Currently, women with invasive cervical carcinomas and we must rely upon events that happen years dysplasia. This shift to younger age groups has later, such as the development of cancer. Com- been attributed to the growing spread of human plex biostatistical and epidemiologic tools are papillomavirus (HPV) infections in the popula- necessary in massive study populations to prove tion [41. clinical efficacy. Another approach is to develop A number of important epidemiologic risk new, surrogate endpoint biomarkers. factors have been identified for developing CW and invasive cervical cancer. Early sexual experi- Criteria for Surrogate Endpoint Biomarkers ences, the number of sexual partners, and male partner factors (number of sexual partners, his- An optimal intermediate biological endpoint tory of venereal disease, early sexual experience) for an at-risk population will be readily ex- are important risk factors L5-81. Smoking in- pressed in plasma or in tissues accessible to bi- creases the risk of developing cervical cancer [9- opsy, related in some way to the process of neo- 111. Of key importance is infection with human plastic transformation, may be easily measured papillomavirus (HPV). HPV DNA sequences can from small quantities of tissue, quantifiable as a be recovered from greater than 90% of cases of continuous variable, and may be expected to be invasive cervix carcinoma. Whether the links modulated by a chemopreventive intervention. between sexual history, smoking, and squamous The following review describes some of the his- intraepithelial lesions influence risk primarily tological and molecular changes that occur before through HPV or whether they are independent and/or during the malignant transformation pro- risk factors suggesting other contributing causes cess. Many of these changes are potential surro- of invasive carcinoma remains controversial. gate endpoint biomarkers. Our goal is to identify However, the recent case control trial by Schiff- the most likely candidates based on the evidence man et al. [12] using PCR assays found that the available to satisfy the above criteria. Cervix and Surrogate Endpoints 115

BIOCHEMICAL EFFECT SURROGATE (from a normal range of 270-747 ng/ml to 16- ENDPOINT BIOMARKERS 188 ng/ml). Retinol is easily measured in plasma using readily adapted high performance liquid We have defined a biochemical or drug effect chromatography methodology [13]. Cellular ret- surrogate endpoint biomarker as an intervention inol concentration in cervical epithelium has not which modulates a cell's normal biochemical been measured. Extraction procedures from tis- function. Drug effect surrogate endpoint bio- sue samples have not been validated, Given the markers address three important aspects of a accessibility of cervix epithelium, direct measure- chemoprevention trial's design. First, they serve ment of 4-HPR concentrations in target tissue is as reproducible, quantitative methods for deter- possible. mining whether a pharmacodynamically impor- Retinoids are known to exert their biological tant drug concentration has been delivered intra- effects by binding to specific nuclear receptor cellularly to the target site. Second, they can be proteins, which are members of the steroid re- used to define endpoints for Phase I chemopre- ceptor superfamily. These receptors regulate the vention dose searching trials. The use of a drug expression of specific target genes in a ligand- effect surrogate endpoint biomarker as an end- dependent manner. Two general classes of these point for a Phase I chemoprevention trial recog- receptors have been described, and are known as nizes that the therapeutic index differs for differ- retinoic acid receptors (RARs) and retinoid X ent drug use indications. Third, drug effect sur- receptors (RXRs). Phase I chemoprevention Wials rogate endpoint biomarkers can be used as mea- of new synthetic retinoids (e.g., 9-cis-retinoic acid, sures of adequate pharmacodynamic action at SR11203, SR11217, TTAB, TTNN) may be based the target issues in Phase I1 or I11 chemopreven- upon their specific binding of nuclear RARs or tion trials. Such data can be used as part of an RXRs. Thus, drug effect and optimal chemopre- adherence assessment. ventive doses in the future may be based upon An optimal drug effect surrogate endpoint specific drug binding to target sites. The poten- biomarker also serves as a biologic surrogate tial for retinoid regulation of cellular prolifera- endpoint biomarker. Thus, the drug effect sur- tion may permit the use of RARs and RXRs as rogate endpoint biomarker may be a cellular pro- biologic as well as drug effect surrogate interme- duct (protein, carbohydrate, gene expression) of diate biomarkers. a regulatory aspect of cellular growth control or a product reflecting changes in a cell that has BIOLOGICAL EFFECT functional but not morphologic transformational SURROGATE ENDPOINT BIOMARKERS changes. In many cases, the usefulness of drug effect surrogate endpoint biomarkers as biologic We define a biological effect surrogate end- endpoints is unclear. The drug effect surrogate point biomarker as one in which an intervention endpoint biomarker becomes one of a number of changes a cellular product. This cellular product potential biologic surrogate endpoint biomarkers must be linked to cancer risk, carcinogenic ex- that are assessed in Phase IIa and Phase IIb che- posure, carcinogenesis, or tumorigenesis. The moprevention trials. following are examples of biologic effect surrogate intermediate endpoints for cervical cancer. Potential Surrogates for Retinoid Chemoprevention Trials in Cervix

Drug effect surrogate endpoint biomarkers are There is compelling molecular and epidemi- defined by the drug being tested in a Phase I or ological evidence indicating that infection with I1 chemoprevention trial. Retinoids and carote- certain genital HPVs has a critical role in the noids have been the primary chemopreventive cellular changes that precede cervical cancer as agents tested to date in cervical cancer. Drug well as other genital cancers [12,14,151. Of the effect surrogate endpoint biomarkers for these more than 70 characterized HPV types, about agents would represent evidence of effect in 30% are associated with genital tract infections. plasma or at the target site, cervix epithelial cells. Whereas 90% of cervical cancers contain HPV Fenretinide (CHPR) suppresses retinol in plasma DNA sequences, only certain genital HPV types 116 Ruffin et al. are associated with cervical cancer. Lorincz et al. keratinocytes with virus DNA results in immor- [15], in their study of 2,624 women, detected talization of nontumorigenic cells [22,231. Immor- HPV DNA in 79.3% of women with confirmed talization is dependent on continued expression cervical neoplasia by Southern blot methods. In of E6 and E7 [241. addition, 23.7% of women with borderline atypia Knowing that HPV 16 and HPV 18 E6 bind to and 6.4% of women with a normal cervix were and promote the degradation of tumor sup- positive for HPV DNA. This study defined four presser gene product p53 provides insights into HPV risk groups based on associations with cer- the biochemical basis of cell transformation C25, vical cancer. The "low-risk'' group contained 261. Furthermore, the E7 product from these vi- HPV types 6, 11,42,43 and 44. DNA sequences ruses, the most abundant viral protein in cervical from these virus types were absent from all cer- cancer cells [27,281, binds to Rb105, another tu- vical cancers but were present in 20.2% of low- mor suppresser gene product [291. The low effi- grade squamous intraepithelial lesions (SIL). An ciency of keratinocyte immortalization by HPV 6 "intermediate risk" group consisted of HPV types and HPV 11 is correlated with the weak binding 31,33,35,51,52 and 58. Viral DNA was detected of their E6 and E7 gene products to p53 and in 23.8% of high-grade SIL but in only 10.5% of Rb105 [30,311. Expression of E6 may account for cervical cancers. The "high-risk HPV 16" group the accumulation of mutations in immortalized was defined by the presence of HPV 16; DNA cells; p53 is required for G, cell-cycle arrest after from this virus type was found in 47.1% of both DNA damage. Kessis et al. [32] have shown that high-grade SIL and cervical cancers. The "high- E6 expression perturbs G, arrest after DNA dam- risk HPV 18" group contained HPV 18, 45 and age, thus permitting DNA replication and fixa- 56; DNA from these virus types were found in tion of mutations. 26.8% of cervical cancers, but only 6.5% of high- The potential biologic effects of surrogate in- grade SIL. termediate endpoints related to HPV are the pre- In most cases of carcinoma, the HPV DNA is sence of viral DNA or the expression of early integrated into cellular genomes, while in CIN genes E2, E6 or E7. In a model system of normal lesions the viral DNA is extrachromosomal [161. human keratinocytes and several independently Integration appears to be random as far as the derived HPV 16 immortalized keratinocyte cell site of integration in the host chromosome is con- lines, Pirisi et al. [331 have recently demonstrated cerned. However, the viral genome integrates that the HPV 16 lines are more sensitive than near known oncogenes in some cell lines [173. normal keratinocyte cell lines to all-trans-retinoic Integration generally occurs in the El/E2 region acid. Incubation in all-trans-retinoic acid at of the viral genome, disrupting the E2 viral tran- M caused HPV 16 immortalized keratino- scription regulatory circuit "1. The E2 open cytes to express 2- to 4-fold less viral mRNA reading frame (OW) encodes a transcription reg- from E6 and E7 than untreated cells. The cells ulatory protein that is a DNA binding protein. did not terminally differentiate; when the reti- For HPV 16 and HPV 18, E2 appears to act as a noic acid was removed, proliferation resumed. repressor of the promoter from which the E6 and Future studies in women with CIN are needed to E7 genes are transcribed [19]. Thus, it is thought determine if E6 or E7 expression is related to that disruption of the E2 gene provides a selec- progression to invasive disease, or whether sup- tive advantage, leading to uncontrolled prolifera- pression of E6 or E7 expression is related to dis- tion of the cell due to deregulated expression of ease regression. The other criteria for surrogate E6 and E7 genes [20]. endpoint biomarkers have been met. In all cell lines examined thus far and in most cervical cancers, the E6 and E7 ORFs are pre- Cytomorphometry served and expressed as mRNAs and proteins. Continued expression of these proteins is critical The cellular morphology used to identify SIL for maintenance of the malignant phenotype of could serve as a surrogate endpoint biomarker. cervical cancer cells [211. In tissue cultures, HPV Nuclear and cytological parameters such as size, 16 and HPV 18 E6 and E7 can profoundly influ- density, nuclear/cytoplasm ratio, texture and ence the growth and differentiation of human geometry of nuclei have been examined to im- keratinocytes. Transfection of primary human prove the diagnosis of cervical cancer [34-391 Cervix and Surrogate Endpoints 117 and other diseases. However, the issue in chemo- dysplasia or CIS compared to CIN I or I1 [41461. prevention of cervical cancer is not only diagno- However, others have demonstrated no differ- sis, but prediction of progression or regression of ence in ploidy status between normal cervical preinvasive disease states. tissue and abnormal [471 and 100% aneuploid Few studies collect data that can address this status in CIN I [48]. The issue of progression issue. Rosenthal et af. [341 found that nuclear versus regression has been associated with mean optical density, cytoplasm mean optical ploidy status in two studies [41,491. HPV has density, and nuclear-cytoplasm area ratio, along been shown to alter the determination of DNA with autocorrelations of nuclear optical density, content or ploidy findings [42,46,48,501 suffici- were predictive of progression or regression in ently that HPV should always be measured. 23 cervical smears; 7 progressed and 16 regres- Most of the studies noted above failed to deter- sed without any intervention. Kwikkel ef aE. [35] mine the presence of HPV infection. Future found no features predictive of progression ver- studies of genomic markers need larger sample sus nonprogression in one set of women (n = 41). sizes of preinvasive disease states and consistent In another set of women (n = 20), they found cell assays of ploidy along with quantitative histo- area, shape of nucleus, and mean density of nu- pathology and determination of HPV status and cleus to correctly classify 80% of the specimens viral type before consideration as a surrogate as progression or nonprogression [35]. A unique endpoint biomarker. aspect of this study was the emphasis on visu- Measures of genomic instability may be useful ally normal intermediate cells rather than visu- surrogate endpoint biomarkers. The available ally abnormal cells as the basis for prediction data is minimal. Kaelbling and colleagues [51] [351. Thus, the available data on cytometric fea- demonstrated loss of heterozygosity of chromo- tures that predict progression or regression are some region 17~13in HPV-negative tumors. based on a small number of women and have Other studies revealed nonrandom structural inconsistent findings. changes in chromosomes 1, 3, 11, and 17p, with It remains unclear whether cellular or nuclear specific allelic losses on 3p reported in up to morphology will be useful surrogate endpoint 80-90% of tumors, and 30% on llq [52]. In con- biomarkers for cervical cancer. In contrast, some trast, Rader and colleagues [521 did not detect form of quantitative cytology and histopathology loss of heterozygosity in any of 15 early stage will be essential to assure accurate identification cervical cancers, all positive for HPV. As with and classification of patients in chemoprevention other genomic markers, a great deal more work trials. needs to be done before considering genomic instability as a surrogate endpoint biomarker.

Genomic Markers Proliferation Markers

The chromosomal karyotyping of cervical le- The carcinogenic transformation of normal sions has shown that the lesions can be diploid cells to cancer is clearly linked to significant or euploid, polyploid or aneuploid. Some au- changes in cellular kinetics. Proliferation markers thors believe ploidy is a good predictor of bio- and corresponding indices, such as mitotic index, logical behavior of cervical cancer [40], but the S-phase fraction, Ki-67, etc., have been used to available data are conflicting and inconsistent. In describe the cellular kinetics of normal cells, dys- addition, most studies have been done in women plastic cells, and cancerous cells. Brugal [53] has with invasive cervical cancer, which is not the argued that three types of markers are necessary target for chemoprevention. Any findings from to describe malignant population growth kin- these studies are likely not relevant to earlier etics: one measures the growth fraction (e.g., Ki- stages of disease. 67), a second evaluates cell cycle speed in arbi- Studies among women with earlier stages of trary units, and a third assesses the S-phase cell disease have some interesting but still inconsis- occurrence frequency (e.g., BrdU and PCNA). All tent findings about DNA content or ploidy sta- of these markers of cellular kinetics have the tus. There appears to be a trend of higher per- potential to serve as surrogate endpoint bio- centages of women with aneuploidy with severe markers. 118 Ruffin et al.

The monoclonal antibody Ki-67 binds to a good supporting data and is feasible to collect nuclear antigen expressed by cycling cells of and measure, therefore, it is a strong candidate several human tissues and to the cytoplasm of for use as a surrogate endpoint biomarker. basal layer cells in squamous epithelia. Among It remains unclear whether one proliferative women with invasive squamous cell carcinoma index is enough to describe the cellular kinetics of the cervix, a range of 10-50% Ki-67 staining of a neoplasm. Given the available data, we rec- has been observed, indicating considerable vari- ommend incorporation of both Ki-67 and PCNA ation in tumor growth rates 1541. In addition, as possible surrogate endpoint biomarkers. there was no significant relationship between Ki-67 staining and conventional histological parameters [541, but a high Ki-67 score has been Regulatory Markers associated with early recurrence [55]. Rishi and colleagues [561 have used Ki-67 on 40 benign and The transformation of a normal cell into a can- malignant tumors, using cytological smears and cer cell is a complex multistep process resulting frozen tissues. The number of Ki-67 positive cells in a clone of cells that are no longer under nor- on cytological smears correlated well with Ki-67- mal regulatory control. Advances in several cells from corresponding tissue. These studies fields have provided several regulatory markers suggest that Ki-67 could be used on cytological to investigate the different steps in carcinogene- smears from the cervix instead of using biopsy sis related to regulatory control. The potential specimens. If the other criteria for surrogate end- regulatory markers include transcription factors, point biomarkers are established for Ki-67, then oncogenes, and tumor suppresser genes. this collection approach would be quite useful, Studies in oncogenes and cervical carcinoma With the availability of a monoclonal antibody have been done primarily in women with inva- to bromodeoxyuridine (BrdU), BrdU-containing sive disease. The expression of c-myc oncogene nuclei can be identified. Very few studies have has been reported to increase from normal tissue evaluated the use of BrdU in cervical tissue. to invasive disease [62-661, but the correlation Fukuda et al. [571 found that BrdU-positive cells with prognosis or relapse is inconsistent 165-701. were mainly located in the parabasal area of Other potential oncogenes include Ha-ras which normal epithelia and distributed throughout the is expressed primarily in high-grade or invasive epithelium with severe dysplasia and CIS. How- lesions [66,67]. In evaluating each of these onco- ever, marked intra-tumor heterogeneity has been genes as a potential surrogate endpoint bio- demonstrated [581 and the delivery method re- marker, the impact of HPV needs to be consid- quires involved intravenous administration or in- ered. HPV appears to have significantly more tracervical injections [57]. Thus BrdU does not impact on disease state and progression than currently look very feasible as a surrogate end- overexpression of any oncogene in the studies point biomarker. that have measured both HPV and other onco- Proliferating cell nuclear antigen (PCNA, genes [69,71-741. cyclin), first detected in 1978, has been suggested The expression rate of EGFR correlates with to be necessary for DNA replication and cellular growth properties in a squamous carcinoma cell proliferation. The pattern of PCNA staining in line. The presumption is that EGFR could be a normal cervical epithelium is localized to the useful surrogate endpoint biomarker for cervical parabasal area. Several studies of PCNA have cancer. EGFR expression has been noted in inter- demonstrated increased activity from normal epi- mediate and superficial areas in severely dys- thelia to condyloma with and without dysplasia plastic lesions and primarily in basal regions in to moderate and severe dysplasia, showing in- normal epithelium [75,761. Various clones of the creasing disruption of normal regulation of cell C4-I cervical cancer cell line demonstrated no proliferation [59,601. In addition, the loss of neg- correlation between growth rate, expression of ative growth control demonstrated by changes in EGFR, and level of HPV gene products [77]. PCNA appears to be linked to consequences of Among 97 HPV lesions of the cervix, EGFR and HPV infection, through either loss of p53 func- c-ubB-2 were not associated with specific HPV tion 1601 or epidermal growth factor receptor types, grade of CIN, or the clinical course [78]. (EGFR) regulatory mechanisms 1611. PCNA has Thus, the available data does not strongly sup- Cervix and Surrogate Endpoints 119 port the potential of EGFR as a surrogate end- logical epithelium 181,821. The orderly staining point biomarker for cervical cancer. pattern restricted to the outer layers of normal squamous cervical epithelium is generally lost in Differentiation Markers preinvasive and invasive cervical lesions. How- ever, there is a vast spectrum of staining patterns Cervical epithelium differentiates along the from normal epithelium to malignant lesions. squamous pathway in carcinogenesis. Thus, Recent studies suggest that involucrin appears markers related to this pathway could be surro- unable to reliably differentiate between benign gate endpoint biomarkers. Several different po- and neoplastic conditions [83,84]. Inconsistency tential surrogate endpoint biomarkers related to among available data could be improved if the differentiation include keratins and involucrin. technique for interpreting staining patterns could Intermediate filaments, characterized by keratins, be refined to a more objectively consistent pro- are a major cytoskeletal element present in vir- cess, such as computerized interpretatiodim- tually all epithelial cells. Keratins represent many aging. Involucrin is a potential surrogate end- different polypetides ranging in molecular point biomarker since it can be done on biopsy weight from 40-70 kD. They can be further sub- specimens, but the technical issues need to be divided into two groups according to molecular addressed and the relationship to neoplastic dis- weight and electrophoretic mobility. At least one ease needs to be clarified. keratin from each of these two subgroups is present in all epithelial cells. The ectocervical PATHOLOGICAL EFFECT epithelium contains type I1 keratins 1,4,5 and 6, SURROGATE ENDPOINT BIOMARKERS and type I keratins 13,14, 15, and 19 with some variable expression of keratins 2, 8, 10, 11, 16, We define a pathologic effect surrogate end- and 17 [79]. There is evidence to suggest that point biomarker as one in which a defined lesion changes take place in the keratin polypeptide such as CIN or low-grade/high-grade SIL re- distribution from normal cervical epithelium to verses after a defined treatment period. This CIN and cervical cancer [8OJ. This requires the pathologic endpoint permits the use of a "gold use of several monoclonal antibodies to detect standard" proof of surrogate biomarker efficacy the differences. Earlier studies, which relied and makes CIN an attractive human model to upon a few monoclonal antibodies such as identify surrogate endpoint biomarkers in cancer CAM 5.2, may have limited ability to detect chemoprevention [85]. changes in the various heterogeneous groups. Keratins may serve as a surrogate endpoint bio- Squamous Cell Carcinoma marker if the specific phenotype related to CIN as a Continuum of CIN and progression to invasive disease could be delineated. Squamous cell carcinoma usually arises from Involucrin is a cytoplasmic protein synthe- the squamous-columnar junction of the cervix, sized in stratified squamous epithelia. It is the and is preceded by, and thought to result from, major precursor to the crosslinked envelope the progression of cervical dysplasia and CIS formed immediately beneath the cellular plasma [86,87]. Further support for a continuum of dis- membrane of maturing squamous epithelial cells. ease is provided by the observation that cervical Involucrin is absent from the basal and sugra- dysplasia is most often diagnosed among women basal cell layers and only detected at the outer in their 20s, CIS in their 30s, and invasive cancer portion of the epithelium where it becomes after the age of 40 [%I. crosslinked. Therefore, involucrin could serve as Recent information challenges the assertion a specific marker of normal squamous differen- that invasive carcinoma of the cervix is part of a tiation and maturation. Involucrin expression is continuum from CIN. The alternative view is altered in preneoplastic and neoplastic conditions that CIN I and CIN 11-111 are distinct disease of numerous squamous epithelia, including the processes with CIN I being the consequence of a cervix. With respect to cervix, involucrin can be self-limited, sexually transmitted viral infection considered a highly specific marker of squamous and CIN 11-111 being a cervical cancer precursor cell differentiation in both normal and patho- lesion [891. Several observations support this as- 120 Ruffin et al.

sertion. First, the anatomic distribution of CIN I tive intervention. In addition, far more women is different (peripheral cervical lesions) from CIN suffer from the diagnosis of CIN I or CIN 11, so 11-111 (central cervical lesions) [90]. Second, the the impact in the U.S. will be less. mean age of women with CIN I and CIN 11-111 are comparable in incident case analysis [SI as SUMMARY distinguished from prevalent case analyses [911. Thus, the concept that younger women with CIN The list of potential surrogate endpoint bio- I progress to higher grades of CIN may be erron- markers for cervical cancer can be overwhelm- eous. Third, 61% of women with CIN 11-111 never ing. In this review, we have highlighted available had a smear showing CIN I [92]. Fourth, 50-90% data for more commonly noted candidates. We of CIN lesions most likely to regress are caused believe that our model of evaluating potential by HPV types other than "high-risk", whereas chemopreventive agents based on biochemical lesions most likely to progress to invasive carci- effect, biological effect and pathological effect noma are caused by high-risk HPV types [121. allows for a focused, systematic investigation. Finally, in vitro infection of human keratinocytes Clearly, the most critical issue is selecting valid with HPV 16 produces a stratified epithelium surrogate endpoint biomarkers. Currently, we that resembles CIN 11-111; infection with HPV 6 are focusing our efforts on retinoid-related che- results in a stratification resembling CIN I [931. mopreventive agents for CIN, so we have chosen A substantial proportion of CIN spontane- plasma levels and tissue levels of retinoids as a ously regresses [94]; CIN I regress more fre- biochemical effect. The RAR and RXR were cho- quently than CIN I1 or 111. The relative frequency sen for both biochemical and biological effects. of spontaneous regression varies widely from The other biologic effect surrogate endpoints are 11% to 44% [95-971 among published reports of E6 and E7 expression of HPV ,along with PCNA this phenomenon. Biopsies of the suspected and Ki-67. Finally, we include the pathological small lesions have either cured the dysplasia or effect of regression, progression or no change of have stimulated an inflammatory response about the histological diagnosis. the biopsy that destroys the residual CIN [983. Spontaneous regression of CIN I11 is rare [98]. REFERENCES Use of CIN as the Pathological Effect 1. Munoz N, Bosch F Epidemiology of cervical cancer. Surrogate Endpoint Biomarker In Munoz N, Bosch F, Jensen 0 (eds): "Human Papil- lomavirus." New York Oxford Press, 1989, pp 9-39. Using CIN or SIL as an endpoint for studies is 2. Wingo PA, Tong T, Bolden S: Cancer statistics, 1995. difficult because of high spontaneous regression CA Cancer J Clin 45:8-30, 1995. rates. However, the clinical significance of treat- 3. Devesa SS, Young J Jr, Brinton LA, Fraumeni .I Jr: ing a disease that is likely to regress spontane- Recent trends in cervix uteri cancer. Cancer M:2184- 2190, 1989. ously is in doubt. Therefore, some have limited 4. Barrasso R, DeBrux J, Croissant 0 High prevalence their chemopreventive studies to CIN II and 111 of papillomavirus-associated penile intraepithelial or high-grade SIL. Historically, the follow-up of neoplasia in sexual partners of women with cervical pregnant women with CIN I11 involves an initial intraepithelial neoplasia. N Engl J Med 317916-923, assessment with colposcopy and biopsy and cy- 1987. tological follow-up. Treatment is delayed until 5. Hulka BS: Risk factors for cervical cancer. J Chronic Dis 353-11, 1982. after delivery as long as the colposcopy is ade- 6. Miller AB, Barclay TH, Choi NW, Grace MG, Wall C, quate and the lesion and cytology do not change Plante M, Howe GR, Cinader B, Davis FG: A study during pregnancy. Therefore, chemoprevention of cancer, parity and age at first pregnancy. J studies using CIN 111 are possible and have been Chronic Dis 33:59545, 1980. done without risk to women as long as all sub- 7. Peters RK, Thomas D, Hagan DG, Mack TM, Hen- jects receive close observation and follow-up. derson BE: Risk factors for invasive cervical cancer This reduces the problem with spontaneous re- among Latinas and non-Latinas in Los Angeles County. J Natl Cancer Inst 771063-1077,1986. gression and reduces the number of women re- 8. Wright VC, Riopelle MA: Age at time of first quired. The compromise is that CIN 111 may rep- intercourse o. chronologic age as a basis for Pap resent a lesion that is too late for chemopreven- smear screening. Can Med Assoc J 127127-131,1982. Cervix and Surrogate Endpoints 121

9. Slattery ML, Robison LM, Schuman KL, French TK, cytes with human papillomavirus type 16 DNA. Abbott TM, Overall J Jr, Gardner JW. Cigarette J Virol 61:1061-1066, 1987. smoking and exposure to passive smoke are risk dac- 24. Munger K, Phelps W, Bubb V, Howley P, Schlegel R: tors for cervical cancer. JAMA 261:1593-1598,1989. The E6 and E7 genes of the human papillomavirus 10. Marshall JR, Graham S, Byers T, Swanson M, Brasure type 16 together are necessary and sufficient for J: Diet and smoking in the epidemiology of cancer of transformation of primary human keratinocytes. the cervix. J Natl Cancer Inst 70:847-851, 1983. J Virol 63:44174421, 1989. 11. Licciardone JC, Brownson RC, Chang JC, Wilkins JR: 25. Werness B, Levine A, Howley P: Association of Uterine cervical cancer risk in cigarette smokers: A human papillomavirus types 16 and 18 E6 proteins meta-analytic study. Am J Prev Med 6:274-281,1990. with p53. Science 248:7&79, 1990. 12. Schiffinan MH, Bauer HM, Hoover RN, Glass AG, 26. Scheffner M, Werness B, Huibregtse J, Levine A, Cadell DM, Rush BB, Scott DR, Sherman ME, Kur- Howley P: The E6 oncoprotein encoded by human man RJ, Wacholder S, Stanton CK, Manos MM: papillomavirus types 16 and 18 promotes the de- Epidemiologic evidence showing that human papillo- gradation of p53. Cell 63:1129-1136, 1990. mavirus infection causes most cervical intraepithelial 27. Smotkin D, Wettstein F Transcription of human neoplasia. J Natl Cancer Inst 85:958-964,1993. papillomavirus-type 16 early genes in a cervical 13. Peng YM, Dalton WS, Alberts DS, Xu MJ, Lim H, cancer and cancer-derived cell line and indenti- Meyskens F Jr: Pharmacokinetics of N-4-hydroxy- fication of the E7 protein. Proc Natl Acad Sci USA phenyl-retinamide and the effect of its oral 83:46804684, 1986. administration on plasma retinol concentrations in 28. Crook T, Morgenstern J, Crawford L, Banks L: cancer patients. Int J Cancer 43:22-26, 1989. Continued expression of HPV-16 E7 protein is 14. Munoz N, Bosch F HPV and cervical neoplasia: required for maintenance of the transformed Review of case control and cohort studies. IARC Sci phenotype of cells co-transformed by HPV-16 plus Pub1 119:251-261,1992. EJ-YUS.EMBO J 8:513-519, 1989. 15. Lorincz AT, Reid R, Jenson AB, Greenberg MD, 29. Munger K, Werness B, Dyson N, Phelps W, Howley Lancaster W, Kurman RJ: Human papillomavirus P: Complex formation of human papillomavirus E7 infection of the cervix: Relative risk associations of 15 proteins with the retinoblastoma tumor suppressor common anogenital types. Obstet Gynecol 79:328- gene product. EMBO 8:40994105, 1989. 337, 1992. 30. Gage J, Meyers C, Wettstein F The E7 protein of the 16. Durst M, Kleinholz A, Hotz M, Gissman E: The nononcogenic human papillomavirus type 6b (HPV- physical state of human papillomavirus type 16 DNA 6b) and of the oncogenic HPV 16 differ in retinoblas- in benign and malignant tumors. J Gen Viral 66: toma protein bininding. J Virol 64:723-730, 1990. 1515-1522, 1985. 31. Munger K, Yee C, Phelps W, Pientenpol J, Moses H, 17. Durst M, Croce C, Gissmann L, Schwarz E, Huebner Howley P: Biochemical and biological differences K Papillomavirus sequences integrate near cellular between E7 oncoproteins of the high- and low-risk oncogenes in some cervical carcinomas. Proc Nat human papillomavirus types are determined by Acad Sci USA 84:1070-1074, 1987. amino terminal sequences. J Vim1 65:3943-3948,1991. 18. Howley P: Structural and transcriptional analysis of 32. Kessis T, Slebos R, Nelson W, Kastan M, Plunkett B, human papillomavirus type 16 sequences in cervical Han S, Lorincz A, Hedrick L, Cho K Human papi%- carcinoma cell lines. J Virol 61:962-971, 1987. lomavirus 16 E6 expression disrupts the p53-medi- 19. McBride A, Spalholz B, Lambert P, Howlet P: Func- ated cellular response to DNA damage. Proc Natl tional domains of papillomavirus E2 proteins. In Acad Sci USA 90:3988-3992,1993. Villarreal L (ed): "Common Mechanism of Transfor- 33. Pirisi L, Batova A, Jenkins G, Hodam J# Creek K: mation of Small DNA Tumor Viruses." Washington Increased sensitivity of human keratinocytes immor- DC: Am SOCfor Microbiol, 1989, pp 115-126. talized by human papilloma virus type 16 DNA to 20. Bernard B, Pailly C, Lenoir M, Darmon M, Thierry F, growth control by retinoids. Cancer Res 52:187-193, Yaniv M: The human papillomavirus type 18 (HPV 1992. 18) E2 gene product is a repressor of the HPV 18 34. Rosenthal D, Philippe A, Hall T, Harami S, Missirlian regulatory region in human keratinocytes. J Virol N, Suffin S: Prognosis of moderate dysplasia: Pre- 62:2994-3002, 1989. dictive value of selected markers in routinely pre- 21. DiMaio D: Transforming activity of bovine and hu- pared cervical smears. Anal Quant Cytol Histol 9: man papillomavirus in cultured cells. Adv Cancer 165168,1987. Res 56:133-159, 1991. 35. Kwikkel H, Timmers T, Boon M, van Rijswijk M, 22. Durst M, Dzarlieva-Petrusevska R, Boukamp P, Stolk J: Relation of quantitative features of visually Fusenig N, Gissman L Molecular and cytogenetic normal intermediate cells in cervical intraepithelial analysis of immortalized human primary keratino- neoplasia I and I1 smears to progression or nonpro- cytes obtained after transfection with human papil- gression of the lesion. Anal Quant Cytol Histol 9: lomavirus type 16 DNA. Oncogene 1:251-256, 1987. 405410,1987. 23. Pirisi E, Yasumoto S, Feller M, Doniger J, DiPaolo J: 36. Hanselaar AG, Vooijs GP, Van't Hof-Grootenboer A, Transformation of human fibroblasts and keratino- Pahlplatz MM: Cytophotometric analysis of cervical 122 Ruffin et al.

intraepithelial neoplasia grade 111, with and without DNA analysis in moderate dysplasia of the uterine synchronous invasive squamous cell carcinoma. cervix. Correlation to the progression and regression Cytometry 11:901-906, 1990. of the lesion. Acta Oncol 29:755-759, 1990. 37. Mariuzzi G, Santinelli A, Valli M, Sisti S, Montironi 50. Ohta M, Ikeda M: The significance of HPV 16, 18 R, Mariuzzi L, Alberti R, Pisani E: Cytometric evi- infection and the DNA ploidy associated with the dence that cervical intraepithelial neoplasia 1 and I1 progression of uterine cervical dysplasia. Nippon are dysplasias rather than true neoplasia. An image Sanka Fujinka Gakkai Zasshi 45:540-546,1993. analysis study of factors involved in the progression 51. Kaelbling M, Burk RD, Atkin NB, Johnson AB, of cervical lesions. Anal Quant Cytol Histol 14:137- Klinger HP: Loss of heterozygosity on chromosome 147, 1992. 17p and mutant p53 in HPV-negative cervical carci- 38. Artacho-Perula E, Roldan-Villalobos R, Salas-Molina nomas. Lancet 340:140-142,1992. J, Vaamonde-Lemos R Multivariate discriminant 52. Rader JS, Huettner P, Karnarasova T, Shah D, Ger- analysis of normal, intraepithelial neoplasia and hu- hard D: Molecular characterization of invasive cervi- man papillomavirus infection of the uterine cervix cal tumors. Proc Annu Meet Am Assoc Cancer Res samples. Histol Histopathol 9:135-140, 1994. 35:A1396, 1994 (abstract). 39. Artacho-Perula E, Roldan-Villalobos R, Salas-Molina 53. Brugal G: Interpretation of proliferation markers. J, Vaamonde-Lemos R Histomorphomeh-y of normal Anal Quant Cytol Histol 16:3940, 1994. and abnormal cervical samples. Anal Quant Cytol 54. Brown DC, Cole D, Gatter KC, Mason DY Carci- Histol 15:29&297, 1993. noma of the cervix uteri: An assessment of tumour 40. Mitchell M, Hittelman W, Hong W, Lotan R, Schot- proliferation using the monoclonal antibody Ki-67. Br tenfeld D: The natural history of cervical intraepithe- J Cancer 57178-181, 1988. lial neoplasia: An argument for intermediate end- 55. Chung TK, Cheung TH, Wong FW,Wong YF: E-67 point biomarkers. Cancer Epidemiol Biomarkers Frev and AgNORs staining in squamous cell carcinoma of 3:619426, 1994. the cervix: A comparison. Gynecol Obstet Invest 37: 41. Bibbo M, Dytch HE, Alenghat E, Bartels PH, Wied 127-129, 1994. GL: DNA ploidy profiles as prognostic indicators in 56. Rishi M, Schwarting R, Kovatich AJ, Ehya H: CIN lesions. Am J Clin Pathol 92261-265, 1989. Detection of growth fraction in tumors by Ki-67 42. De Vita R, Calugi A, Maggi F, Mauro F, Montevecchi monoclonal antibody in cytologic smears: A pro- L, Vecchione A: Flow cytometric DNA analysis of the spective study of 40 cases. Diagn Cytopathol9:52-56, human cervix affected by human papillomavirus 1993. and /or intraepithelial neoplasia. Anal Quant Cytol 57. Fukuda K, Iwasaka T, Hachisuga T, Sugimori H, Histol 12:306-313, 1990. Tsugitomi H, Mutoh F Immunocytochemical detec- 43. Hanselaar AG, Vooijs GP, Oud PS, PahlpPatz MM, tion of S-phase cells in normal and neoplastic cer- Beck JL: DNA ploidy patterns in cervical intraepi- vical epithelium by anti-BrdU monoclonal antibody. thelial neoplasia grade 111, with and without synchro- Anal Quant Cytol Histol 12135-138, 1990. nous invasive squamous cell carcinoma. Measure- 58. Bolger BS, Cooke TG, Symonds RP, MacLean AB, ments in nuclei isolated from paraffin-embedded Stanton PD: Measurement of cell kinetics in cervical tissue. Cancer 622537-2545, 1988. tumours using bmmodeoxyuridine. Br J Cancer 68: 44. Yokosuka K, Sato K, Izutsu T, Kagabu T, Nishiya I, 166-171, 1993. Wied GL: A nuclear DNA study of uterine cervical 59. Mittal K, Demopoulos R, Goswami S: Proliferating dysplasia with reference to its prognostic signifi- cell nuclear antigen (cyclin) expression in normal and cance. Nippon Sanka Fujinka Gakkai Zasshi 401760- abnormal cervical squamous epithelia. Am J Surg 1766, 1988. Pathol 17117-122, 1993. 45. Dudzinski MR, Haskill SJ, Fowler WC, Cunie JL, 60. Cardillo MR, Stamp GW, Pignatelli MN, Lalani EN Walton LA: DNA content in cervical neoplasia and Immunohistochemical analysis of p53 oncoprotein its relationship to prognosis. Obstet Gynecol 69:373- and proliferating cell nuclear antigen (PCNA) in the 377,1987. cervix uteri. Eur J Gynaecol Oncol 14:484490, 1993. 46. Perticarari S, Presani G, Michelutti A, Facca MC, 61. Mathevet P, Mitchell MF, Tortolero-Luna G, Silva E, Alberico S, Mandruzzato GP: Flow cytometric analy- Hittelman WN: Early steps in proliferative dysregul- sis of DNA content in cervical lesions. Pathol Res ation in HPV -associated cervical carcinogenesis.Proc Pract 185:686488, 1989. Am Assoc Cancer Res 35:A942, 1994 (abstract). 47. Hughes RG, Neil1 WA, Norval M: Nuclear DNA 62. Riou G, Barrois M, Le MG, George M, Le Doussal V, analysis of koilocytic and premalignant lesions of the Haie C: c-myc proto-oncogene expression and grog- uterine cervix. Br Med J 294:267-269, 1987. nosis in early carcinoma of the uterine cervix. Lancet 48. Watts KC, Campion MJ, Butler EB, Jenkins D, Singer 1~761-763, 1987. A, Husain OA: Quantitative deoxyribonucleic acid 63. Riou GF, Bourhis J, Le MG: The c-myc proto-onco- analysis of patients with mild cervical atypia: A po- gene in invasive carcinomas of the uterine cervix: tentially malignant lesion? Obstet Gynecol 70205- Clinical relevance of overexpression in early stages 207, 1987. of the cancer. Anticancer Res 10:1225-1231, 1990. 49. Kashyap V, Das DK, Luthra UK: Microphotometric 64. Riou G, Sheng ZM, Zhou D, Lusinchi A, Le Doussal Cervix and Surrogate Endpoints 123

V, Barrois M: c-myc and c-Ha-rus proto-oncogenes in factor receptor expression. Cancer Res 50:373@3736, cervical cancer: Prognostic value. Bull Cancer 7734- 1990. 347, 1990. 78. Tervahauta A, Syrjanen S, Syrjanen K: Epidermal 65. Sagae S, Kuzumaki N, Hisada T, Mugikura Yp Kudo growth factor receptor, c-erbB-2 proto-oncogene and R, Hashimoto M: rus oncogene expression and prog- estrogen receptor expression in human papilloma- nosis of invasive squamous cell carcinomas of the virus lesions of the uterine cervix. Int J Gynecol uterine cervix. Cancer 63:1577-1582, 1989. Pathol 13:234-240, 1994. 66. Sagae S, Kudo R, Kuzumaki N, Hisada T, Mugikura 79. Czemobilsky B, Moll R, Franke W: Intermediate fila- Y, Nihei T, Takeda T, Hashimoto M: rus oncogene ments of normal and neoplastic tissues of the female expression and progression in intraepithelial neopla- genital tract with emphasis on problems of differ- sia of the uterine cervix. Cancer 66:295-301, 1990. ential tumor diagnosis. Path Res Pract 179:31-37, 67. Pinion SB, Kennedy JH, Miller RW, MacLean AB: 1984. Oncogene expression in cervical intraepithelial neo- 80. Smedts F, Ramaekers F, Robben H, Pruszczynski M, plasia and invasive cancer of cervix. Lancet 337819- van Muijen G, Lane B, Leigh I, Vooijs P: Changing 820, 1991. patterns of keratin expression during progression of 68. Hendy-Ibbs P, Cox H, Evan GI, Watson JV:Flow cervical intraepithelial neoplasia. Am J Pathol 136: cytometric quantitation of DNA and c-myc oncopro- 657-668, 1990. tein in archival biopsies of uterine cervix neoplasia. 81. Dekmezian R, Chen X, Kuo T, Ordonez N, Katz RE: Br J Cancer 55:275-282, 1987. DNA hybridization for human papillomavirus (HPV) 69. Choo KB, Chong KY, Chou HF, Liew LN, Liou CC: in cervical lesions. Relationship of the presence of Analysis of the structure and expression of the c-myc various viral subtypes to expression of HPV struc- oncogene in cervical tumor and in cervical tumor- tural proteins, involucrin, and carcinoembryonic anti- derived cell lines. Biochem Biophys Res Commun gen. Arch Pathol Lab Med 111:22-27, 1987. 1583334-340, 1989. 82. van Bommel PF, Kenemans P, Helmerhorst TJ, Gallee 70. Symonds RP, Habeshaw T, Paul J, Kerr DJ, Darling MP, Ivanyi D: Expression of cytokeratin 10, 13, and A, Burnett RA, Sotsiou F, Linardopoulos S, Span- involucrin as prognostic factors in low stage didos DA: No correlation between rus, c-myc and squamous cell carcinoma of the uterine cervix. c-jun proto-oncogene expression and prognosis in Cancer 742314-2320, 1994. advanced carcinoma of cervix. Eur J Cancer 10:1615- 83. Serra V, Ramirez AA, Lara C, Marzo C, Castells A, 1617, 1992. Bonilla-Musoles F Distribution of involucrin in nor- 71. Mitra AB, Murty VV, Pratap M, Sodhani P, Chaganti mal and pathological human uterine cervix. Gynecol RS: ERBB2 (HER-%/neu) oncogene is frequently Oncol 36:3442, 1990. amplified in squamous cell carcinoma of the uterine 84. Elsayed A, Richart RM, Crum CP: Involucrin expres- cervix. Cancer Res 54637439, 1994. sion in cervical intraepithelial neoplasia: A critical 72. Ocadiz R, Sauceda R, Salcedo M, Ortega V, evaluation. Gynecol Oncol 262-34, 1987. Rodriguez H, Gordillo C, Chavez P, Gariglio P: 85. Boone C, Kelloff G, Steele V Natural history of intra- Occurrence of human papillomavirus type 16 DNA epithelial neoplasia in humans with implications for sequences and c-myc oncogene alterations in uterine- cancer chemoprevention strategy. Cancer Res 52: cervix carcinoma. Arch Invest Med 20:355-362,1989. 165-1659, 1992. 73. Willis G, Jennings B, Ball RY, New NE, Gibson I: 86. Peterson 0 Spontaneous course of cervical precan- Analysis of rus point mutations and human papillo- cerous conditions. Am J Obstet Gynecol 72:1063, mavirus 16 and 18 in cervical carcinomata and their 1956. metastases. Gynecol Oncol49:359-364, 1993. 87. Richard R: Natural history of cervical intraepithelial 74. Van Le L, Stoerker J, Rinehart CA, Fowler WC: H-rus neoplasm. Clin Obstet Gynecol 105383, 1967. codon 12 mutation in cervical dysplasia. Gynecol 88. Canadian Task Force: Cervical cancer screening pro- Oncol49:181-184, 1993. grams: Epidemiology and natural history of carci- 75. Yamasaki M, Maruo T, Akahori T, Mochizuki M: noma of the cervix. Can Med Assoc J 114:1003-1031, Immunohistochemical studies on epidermal growth 1976. factor receptor and the rnyc oncogene product in 89. Kiviat NB, Koutsky LA: Specific human papilloma- squamous cell carcinoma of the uterine cervix. Nip- virus types as the causal agents of most cervical pon Sanka Fujinka Gakkai Zasshi 4051-58, 1988. intraepithelial neoplasia: Implications for current 76. Mittal K, Pearson J, Demopoulos R Patterns of views and treatment. J Natl Cancer Inst 85:934-935, mRNA for epidermal growth factor receptor and 1993. keratin B-2 in normal cervical epithelium and in cer- 90. Burghardt E, Ostor AG: Site and origin of squamous vical intraepithelial neoplasia. Gynecol Oncol38:224- cervical cancer: A histomorphologic study. Obstet 229, 1990. Gynecol 62:117-127, 1983. 77. von Knebel-Doeberitz M, Gissmann L, zur Hausen 91. Koss L (ed): "Diagnostic Cytopathology," 3rd ed. H: Growth-regulating functions of human papilloma- Philadelphia: J.B. Lippincott, 1979. virus early gene products in cervical cancer cells 92. Koutsky L, Holmes K, Critchlow C, Stevens C, acting dominant over enhanced epidermal growth Paavonnen J, Beckman A, DeRouen T, Galloway D, 124 Ruffin et al.

Vernon D, Kiviat N: A cohort study of the risk of 95. Kinlen LJ, Spriggs AI: Women with positive cervical cervical intraepithelial neoplasia grade 2 or 3 in smears but without surgical intervention. A follow- relation to papillomavirus infection. N Engl J Med up study. Lancet 2:463-465, 1978. 3271272-1278, 1992. 96. Campion MJ, McCance DJ, Cuzick J, Singer A: Pro- 93. Merrick DT, Blanton RA, Gown AM, McDougall JK: gressive potential of mild cervical atypia: Prospective Altered expression of proliferation and differentia- cytological, colposcopic, and virological study Lancet tion markers in human papillomavirus 16 and 18 2~237-240, 1986. immortalized epithelial cells grown in organotypic 97. Carmichael JA, Maskens PD: Cervical dysplasia and culture. Am J Pathol 140167-177, 1992. human papillomavirus. Am J Obstet Gynecol 160: 94. Richart RM, Barron BA: A follow-up study of pa- 916918, 1989. tients with cervical dysplasia. Am J Obstet Gynecol 98. Spriggs AI: Natural history of cervical dysplasia. Clin 105:38&393, 1969. Obstet Gynaecol 8:65-79, 1981.