Adverse Haematological Outcome and Environmental Lead Poisoning

www.nature.com/jea

Adverse haematological outcome and environmental lead poisoning

VINCENZO FONTANA,a ROBERTA BALDI,b MICHELA FRANCHINI,c PAOLA GRIDELLI,d ROBERTO NERI,e FRANCO PALMIERI,e RICCARDO PUNTONI,a UMBERTO RICCOf AND STEFANO PARODIa aDepartment of Environmental Epidemiology, National Cancer Research Institute (IST), Genoa, Italy bUnit of Public Health, Local Health Unit (ASL) 5, La Spezia, Italy cDepartment of Epidemiology & Health Service Research, National Research Council (CNR), Pisa, Italy dUnit of Public Health, Local Health Unit (ASL) 12, Viareggio, LU, Italy eSection of Environmental Toxicology, Environmental Protection Agency (ARPAL), La Spezia, Italy fUnit of Occupational Health, Local Health Unit (ASL) 5, La Spezia, Italy

The South-eastern Area (SA) of the Municipality of La Spezia (Liguria Region, Italy) is characterised by a heavy environmental lead (Pb) contamination, chiefly due to the emissions of a Pb-processing plant in operation since 1930. In order to assess the risk of Pb poisoning of residents of SA, and to estimate the degree of association between the blood Pb level (BLL) and haematocrit % (HCT), intended as a biomarker of early haematological dysfunction, we reanalysed data of 785 individuals collected in 1992 as a part of a larger national biological monitoring project. Multiple normal regression modelling was applied to estimate the role of residence on log-transformed BLL, and Median Ratio (MR) was used as an index of effect. The same statistical modelling was also applied to reveal the relationship between HCT and BLL. Allowing for several confounders (including occupational exposure to Pb), residents of SA showed a 14% increase (MR ¼ 1.14, 95% IC ¼ 1.06–1.23%) in the median BLL value compared to people living outside SA. The excess reached 27% (MR ¼ 1.27, 95% IC ¼ 1.14–1.41%) after 30 years of residence. Parallel results were also obtained in a subgroup composed only of pupils (o18 years, non-smokers, non-drinkers). Finally, regression analysis highlighted a statistically significant parabolic trend in HCT in relation to BLL. The non-linear dose–response relationship, which attests to an adverse effect on the erythrocytic function of BLLs at least over 17.00 mg/dl, is in agreement with the findings of other authors and consistent with the results of an excess occurrence of self-reported anaemia obtained from a previous comparative survey carried out on the same population. Journal of Exposure Analysis and Environmental Epidemiology (2004) 14, 188–193. doi:10.1038/sj.jea.7500318

Keywords: environmental lead pollution, blood lead level, adverse haematological effects.

Introduction exposed to environmental Pb deserve to be monitored in order to detect early reversible abnormalities and life- Over the past decade, industrialised countries have been threatening disorders and, as a corollary result, to improve subject to much lower levels of environmental lead (Pb) the scientific knowledge on the biological impairments contamination owing to the legal restrictions on industrial attributable to Pb burdens. emissions and the widespread use of unleaded petrol (Fur- It is well-known that Pb, which is mainly absorbed into the man and Laleli, 2001). Although a sustained and sizeable blood plasma via gastrointestinal and respiratory tracts decrease in human body Pb burden has been observed at (Mahaffey et al., 1992), is able to affect the erythrocytic the same time (Gottlieb, 1998), scientific interest and function through a reduction of haeme biosynthesis and concerns among public health researchers and epidemiolo- increased demolition rate of red blood cells, as well as to gists over the long-term negative potential of Pb contamina- induce a sub-clinical anaemic condition at least in individuals tion have not abated. In particular, attention has focused on with higher blood Pb level (BLL) (Hillman, 1998). In this reproductive, genotoxic, carcinogenic, and neurological context, haematocrit % (HCT) might be used both as a effects (Fu and Boffetta, 1995; Apostoli et al., 2000; specific biomarker of early reversible adverse effects and as an Hertz-Picciotto, 2000; Lorente et al., 2000; Ojajarvi et al., intermediate end point of more severe health outcomes. 2000; Pesch et al., 2000; Restrepo et al., 2000; Schwartz et al., The environmental setting under observation is represented 2000; Silbergeld et al., 2000). For these reasons, populations by the south-eastern area (SA) of the Municipality of La Spezia (Liguria Region, Italy), which is characterised by heavy ground pollution of Pb dust emitted mainly by a Pb- 1. Address all correspondence to: Vincenzo Fontana, DSc, Department of processing plant (PbO) that has been active since 1930. Environmental Epidemiology, National Cancer Research Institute (IST), Unpublished measurements of Pb in lichens carried out in L.go R.Benzi, 10, 16132, Genoa, Italy. Tel.: þ 39-010-56-00-808. Fax: þ 39-010-56-00-501. E-mail: [email protected] 1994 by the local Environmental Protection Agency Received 18 April 2003; accepted 29 August 2003 (ARPAL) revealed an average concentration of almost Lead poisoning effects Fontana et al.

200.0 mg/g dry weight, which was the highest value registered 785 individuals (84.7%). These were geo-referenced accord- in the whole Municipality (Figure 1). ing to their residence in 1992 and classified as residents or During 1992, the Province of La Spezia (nearly 190,000 non-residents in SA. Finally, a job-exposure matrix, defined inhabitants) was included in a national biological monitoring in the INIH campaign protocol, was used to classify subjects campaign of the general population against the risk of Pb as professionally exposed or non-exposed to Pb. poisoning carried out by the Italian National Institute of In order to estimate the effect of residence on BLL, and Health (INIH campaign) (Neri and Palmieri, 1998). The BLL on HCT, adjusted for the possible confounding effect of survey was designed to estimate the prevalence of poisoned a number of covariates, multiple regression modelling individuals, regardless of the Pb source (traffic, factories, (Kleinbaum et al., 1998) was applied to the response mining and recycling plants, etc.). variables (BLL or HCT). Final models were obtained by a By using information from the INIH campaign, we sought backward hierarchical non-automatic selection of confoun- both to assess the risk of Pb poisoning of residents in SA ders, including power and interaction terms. Covariates were compared to that of subjects residing elsewhere in the same removed from the model on the basis of their influence on the Province, and to evaluate the presence of early haematolo- estimates of the study predictors (residence or BLL), and gical impairments plausibly attributable to BLL. their statistical significance (two-tailed partial F-test). Lever- age, influence, and residual measures were used as goodness- of-fit indices to evaluate the adequacy of each model. In Methods particular, a non-parametric smoother (lowess) (Cleveland, 1979) was used to highlight specific patterns in the residual The INIH campaign data included characteristics on life- scatter-plots. For each parameter estimate, 95% confidence style, demographics, occupation, and health, collected intervals (95% CI) were calculated, and to show the through a structured questionnaire on 927 individuals explanatory power of each model the coefficient R2 was residing in the Province of La Spezia. During the interview, computed. In view of the right-skewed BLL distribution, a blood sample was also taken from respondents. Among the data were log-transformed. In the regression analysis, the haematological parameters, HCT was the only biomarker result of this transformation leads to estimate a measure of useful for analysis. Complete information was obtained for effect referred to as the Median Ratio (MR) (Miettinen, 1985): for a categorical predictor, MR represents the expected median of BLL in a category divided by the same index of the reference category. Data were analysed using SAS statistical software (SAS Institute Inc., 1997).

Results

Blood Lead The univariate analysis of the BLL distribution (Table 1) revealed that residents of SA displayed a BLL geometric mean value (9.13 mg/dl) higher than that of the other subjects (7.73 mg/dl). In particular, when the duration of residence was greater than 30 years, the BLL geometric mean was 11.62 mg/dl. Also, 37.1% of the whole sample had a value higher than 10.00 mg/dl, which is currently considered as the threshold of biochemical and sub-clinical abnormalities (Markowitz, 2000). BLLs appreciably higher than threshold were also observed in subjects professionally exposed to Pb for more than 13 years, those aged 50 years or more, and those who were heavier alcohol drinkers (420 g/day). The univariate relationships were conﬁrmed by multiple log- normal regression analysis. Speciﬁcally, after allowing for confounding and using non-residents as the reference category, residents of SA showed a 14% increase in the Figure 1. Municipality of La Spezia (Liguria Region, Italy). median BLL value (MR ¼ 1.14, 95% CI ¼ 1.06–1.23) (data Geographic distribution of Pb concentration in lichens (mg/g dry weight) in 1992 and location of the Pb-processing plant (PbO) within not shown), which reached 27% after 30 years of residence the South-eastern Area (circled). (MR ¼ 1.27, 95% CI ¼ 1.14–1.41) (Table 2). Occupational

Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(2) 189 Fontanaetal. Lead poisoning effects

Ta bl e 1 . Distribution of blood lead levels (mg/dl) of 785 residents in Time spent in trafﬁcked roads (min/day) the Province of La Spezia (Liguria Region, Italy) in 1992 according to r30 202 7.65 1.77 2.30–33.50 factors taken into consideration for analyses. 31–75 195 8.49 1.79 1.60–44.50 76–140 186 8.62 1.75 1.70–42.55 Factors and categories No. GM GSD Range 4140 202 8.61 1.75 2.40–50.75

Place of residence Dwelling floor Outside the Municipality 110 7.16 1.75 1.95–38.35 Ground 151 8.69 1.68 2.30–44.50 of La Spezia 1st 299 8.43 1.79 2.00–50.75 In the Municipality 326 7.94 1.73 1.60–39.85 2nd 143 8.30 1.84 1.70–42.55 but outside SA 42nd 192 7.91 1.76 1.60–43.45 In SA 349 9.13 1.79 1.70–50.75 r15 years 123 7.56 1.70 1.70–26.85 Traffic intensity in the dwelling road 16–30 years 113 8.81 1.77 3.00–50.75 Low 90 7.33 1.74 1.95–31.60 430 years 113 11.62 1.75 2.80–35.70 Moderate 175 8.17 1.82 3.00–35.70 High 520 8.57 1.75 1.60–50.75 Occupational exposure to Pb Non-exposed 627 7.93 1.77 1.60–445.00 Whole group 785 8.33 1.77 1.60–50.75 Exposed 158 10.12 1.69 3.35–507.50 No. number of subjects. r13 years 38 7.77 1.52 3.35–195.50 ¼ 413 years 45 12.20 1.77 3.95–507.50 GM ¼ geometric mean. Unknown 75 10.36 1.64 3.70–398.50 GSD ¼ geometric standard deviation. SA ¼ south-eastern area. Gender Female 319 6.81 1.74 1.60–398.50 Male 466 9.56 1.72 2.24–507.50 exposure to Pb was also a significant predictor of BLL: compared to non-exposed subjects, after 13 years spent in a Body mass index Pb-related activity, the estimated median excess in exposed r21.11 157 5.57 1.67 1.70–398.50 21.12–23.61 156 8.02 1.75 1.60–344.50 individuals was 32% (MR ¼ 1.32, 95% CI ¼ 1.16–1.51). 23.62–25.25 161 9.28 1.69 2.00–357.00 Remarkable contributions to the explanation of BLL 25.26–27.43 154 10.14 1.72 2.40–507.50 variability were also given by gender and smoking habit, 427.43 157 9.53 1.65 3.30–311.00 and steeply rising trends were observed for alcohol con- Age at blood sampling (years) sumption and age at blood sampling. r28 155 4.86 1.44 1.95–143.00 In order to avoid the effect of the possible misclassification 29–39 159 7.21 1.60 1.60–268.50 of subjects according to occupational exposure to Pb, 40–49 155 9.03 1.67 1.70–331.00 analysis focused on a subgroup of 86 pupils (36 males and 50–62 157 11.00 1.64 2.70–434.50 462 159 11.40 1.69 2.80–507.50 50 females), aged under 18 years, all of whom were non- drinkers and non-smokers. After adjustment for gender, age Month at blood sampling and month at blood sampling, body mass index, dwelling January–February 167 7.86 1.69 1.60–331.00 floor, and traffic intensity in the dwelling road, a parallel MR March 105 7.20 1.67 1.95–274.50 April 236 9.17 1.77 2.00–398.50 point estimate was found for this subgroup (MR ¼ 1.28, May-June 277 8.39 1.83 1.70–507.50 95% IC ¼ 1.06–1.55), with a slight increase after 8 years (median value) of residence in SA (Table 3). Alcohol consumption (g/day) Non-drinker 342 6.10 1.63 1.60–445.00 Haematocrit r20 221 9.11 1.64 1.70–278.50 21–45 151 12.18 1.65 3.90–507.50 Data breakdown by age and gender showed no significant 445 71 12.53 1.57 4.00–383.00 departures from the range considered normal in the haematological literature (Perkins, 1999). In particular, 5th Coffee consumption (cups/day) and 95th percentiles were 39–49 and 35–44 for males and Non-drinker 198 6.10 1.80 1.80–445.00 females, respectively. Although univariate analyses indicated 1–2 326 9.36 1.72 2.60–507.50 Z3 261 9.11 1.66 1.60–398.50 no particular pattern in HCT levels by any indicator of Milk consumption exposure to Pb (BLL, residence, occupation), after control- Always/seldom 501 7.84 1.76 1.70–50.75 ling for gender, age, and month at blood sampling, smoking Never 284 9.26 1.76 1.60–43.45 habit, milk consumption, health status, and body mass index, normal regression modelling allowed us to estimate a Smoking habit Never smoked 518 7.65 1.77 1.60–44.50 statistically significant parabolic dose–response relationship Ex-smoker 91 9.43 1.67 3.20–27.40 between BLL and HCT (Table 4, model 1). A similar result Current smoker 176 10.03 1.72 3.50–50.75 was obtained by replacing BLL with the years of residence in

190 Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(2) Lead poisoning effects Fontana et al.

Ta bl e 2 . Joint effect on blood lead level of the duration of residence in Table 3 . Effect of residence in the South-eastern Area of the the South-eastern Area of the Municipality of La Spezia (Liguria Municipality of La Spezia (Liguria Region, Italy) on blood lead level Region, Italy) and other covariates estimated through multiple log- estimated through multiple log-normal regression modelling in the normal regression modelling. subgroup of subjects classiﬁed as pupils (50 females and 36 males, o18 years). Factors and categories MR 95% CI F-test P-value Model Factors & categories MR 95% CI F-test P-value Constant 2.93a 2.43–3.54 FF Duration of residence 6.34 o0.001 1 Constant 4.01a 2.71–5.92 FF in SA (years) Non-resident 1.00 (Ref.) Residence in SA 6.44 0.013 r15 1.10 1.00–1.21 Non-resident 1.00 (Ref.) 16–30 1.11 1.00–1.23 Resident 1.28 1.06–1.55 430 1.27 1.14–1.41 2 Constant 4.01a 2.70–5.94 FF

Occupational exposure to Pb 6.71 o0.001 Duration of residence 3.18 0.047 Non-exposed 1.00 (Ref.) in SA (years) Exposed r13 years 1.16 1.01–1.34 Non-resident 1.00 (Ref.) Exposed 413 years 1.32 1.16–1.51 r8 1.27 1.02–1.60 Unknown 1.09 0.98–1.21 48 1.29 1.00–1.64

Model 1: R2 ¼ 36.3%, F-test ¼ 5.48, P-value o0.001. Smoking habit 6.29 0.002 2 Never smoked 1.00 (Ref.) Model 2: R ¼ 36.5%, F-test ¼ 4.81, P-value o0.001. Ex-smoker 1.02 0.92–1.12 MR ¼ median ratio adjusted for gender, age and month at blood sampling, Current smoker 1.14 1.06–1.22 body mass index, dwelling floor, traffic intensity in the dwelling road. 95% IC ¼ 95% confidence interval of MR. F-test ¼ significance test. Alcohol consumption (g/day) 32.19 o0.001 Non-drinker 1.00 (Ref.) P-value ¼ significance level of F-test. Ref. ¼ reference category. r20 1.16 1.44–1.81 2 21–45 1.42 1.02–1.16 R ¼ coefficient of determination %. 445 1.61 1.29–1.48 SA ¼ south-eastern area. aEstimated baseline median value of blood lead level (mg/dl). Gender 86.58 o0.001 Female 1.00 (Ref.) Male 1.39 1.29–1.48 50

Age at blood sampling (years) 29.40 o0.001 r28 1.00 (Ref.) 45 29–39 1.16 1.05–1.28 40–49 1.33 1.20–1.47 50–62 1.59 1.43–1.77 462 1.75 1.56–1.96 40

Model R2 ¼ 53.9%, F-test ¼ 42.46, P-value ¼ 0.001. MR ¼ median ratio adjusted for milk consumption and month at blood 35

sampling. Haematocrit (%) 95% CI ¼ 95% confidence interval of MR. F-test ¼ significance test. 30 P-value ¼ significance level of F-test. Ref. ¼ reference category. 2 R ¼ coefficient of determination %. 25 SA ¼ south-eastern area. 0102030405060 a Estimated base-line median value of blood lead level (mg/dl). Blood lead level Figure 2. Estimated parabolic dose–response relationship between SA (Table 4, model 2). To illustrate, Figure 2 displays the haematocrit % and blood lead level (mg/dl), and relative 95% confidence interval. The fitted haematocrit % values derive from a non-linear dependence of HCT on BLL used as a continuous multivariate regression model with covariates set to their own average predictor. values.

residing outside SA?; and (2) were the observed BLLs able to Discussion induce early haematological impairments in the poisoned individuals? Although the study results seem to yield positive This analysis had to answer two speciﬁc questions: (1) did answers to both questions, some comments and reﬂections subjects living in SA show a higher BLL than subjects deserve to be made.

Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(2) 191 Fontanaetal. Lead poisoning effects

Ta bl e 4 . Assessment of dose–response relationship through multiple (Table 3) provided crucial indications on this matter. normal regression analysis. Model 1: haematocrit % versus blood lead Actually, because of their age (o18 years), these subjects level. Model 2: haematocrit % versus years of residence in the South- were not subject to any professional exposure, and their BLL eastern Area. may be attributable to environmental sources, among which the residence in SA resulted to have a prominent role. Model Predictor b 95% CI F-test P-value

1 Constant 38.88a 38.03–39.73 FF Haematocrit It is well-known that HCT varies according to the kind and Blood lead level 8.12 o0.001 severity of a disorder, or owing to treatments received (mg/dl) (Hillman, 1998). Admittedly, the knowledge of subjects’ r5.00 0.00 (Ref.) 5.01–7.05 1.56 0.91–2.22 health conditions might have improved the validity of this 7.06–9.55 1.59 0.89–2.29 epidemiological investigation, by modifying to some extent 9.56–13.45 1.95 1.20–2.71 the estimated dose–response relationship. Nevertheless, the 413.45 1.90 1.11–2.73 exclusion from the analysis of individuals who reported a 2b Constant 39.14a 39.30–39.98 FF chronic disorder did not appreciably change this result (data Duration of residence 2.87 0.036 not shown). Furthermore, analysis of regression diagnostics in SA (years) highlighted a good statistical robustness of the estimated Non-resident 0.00 (Ref.) parabolic trend. In other words, subjects with the higher r15 0.44 À0.19–1.08 BLLs, for example, over 17.00 mg/dl (79, 10.1%), seem to 16–30 0.33 À0.38–1.03 430 À0.65 À1.40–À0.11 show a progressive and sensitive reduction in the number and/ or volume of red blood cells that might reflect a true 2 Model 1: R ¼ 43.5%, F-test ¼ 27.98, P-value ¼ 0.001. haematological dysfunction, such as a sub-clinical anaemic Model 2: R2 ¼ 42.4%, F-test ¼ 26.72, P-value ¼ 0.001. condition. It is noteworthy that this indication is consistent b ¼ regression coefficient adjusted for gender, age and month at blood sampling, smoking habit, milk consumption, health status, body mass with the results of a previous epidemiological survey index. conducted on the same area during 1998–1999, whose main 95% CI ¼ 95% confidence interval of b. finding was an excess prevalence and incidence of self-reported F-test ¼ significance test anaemia in residents of SA compared to the control group P-value ¼ significance level of F-test. (Fontana et al., 2000). Moreover, the non-linear trend Ref. ¼ reference category. R2 ¼ coefficient of determination %. estimated in HCT is in agreement with the results of other SA ¼ South-eastern Area. studies. In particular, on the basis of their investigations, some a Estimated base-line mean value of haematocrit %. Authors (Schwartz et al., 1990; Hense et al., 1992; Jacob bOccupational exposure to lead is included in the model. et al., 2000) pointed out or hypothesised a parabolic dose– response relationship similar to that found in this analysis. In conclusion, it can be stated that environmental Pb Blood Lead contamination in SA was one of the major determinants of Firstly, this analysis was based on data stemming from an BLL, which in turn appeared to be related to a haematolo- investigation (INIH campaign) whose aim was to describe Pb gical impairment. Preventive measures are therefore required poisoning in the general population, irrespective of Pb to reduce the environmental exposure, to lower the body sources. Therefore, the design features of the INIH campaign burden and to control and avoid further detrimental health might only partially fulfil the epidemiological requirements of outcomes. the current study. For this reason, a lack of specific information and the presence of selection factors cannot be Acknowledgments entirely ruled out. Residual confounding (i.e., incomplete confounding adjustment) and measurement error in the This work was supported through a grant from the occupational classification of subjects, performed according Municipality of La Spezia. to a job-exposure matrix, may have attenuated the effect of professional exposure to Pb on the study outcome and, accordingly, increased the predictive role of other covariates, References residence included. For instance (Table 2), the excess BLL Apostoli P., Bellini A., Porru S., and Bisanti L. The effect of lead on male fertility: calculated for males compared to females and the steep a time to pregnancy (TTP) study. Am J Ind Med 2000: 38: 310–315. positive age trend might be the result of this process. Finally, Cleveland W.S. Robust locally weighted regression and smoothing scatterplots. J incomplete information on the length of occupational Am Statist Assoc 1979: 74: 829–836. Fontana V., Baldi R., Franchini M., Gridelli P., Ceppi M., Magnoni U., and exposure to Pb (47.5% of missing data) might have further Puntoni R. Epidemiologic study of the residents of the southeastern area of the aggravated this drawback. Nonetheless, sub-group analysis Municipality of La Spezia. Epidemiol Prev 2000: 24: 172–179.

192 Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(2) Lead poisoning effects Fontana et al.

Fu H., and Boffetta P. Cancer, occupational exposure to inorganic lead Miettinen O.S. Theoretical Epidemiology: Principles of Occurrence Research in compound: a meta-analysis of published data. Occup Environ Med 1995: 52: Medicine. John Wiley & Sons, New York, 1985, pp. 237–238. 73–81. Neri R., and Palmieri F. Blood lead levels in the population of La Spezia. Ann Ist Furman A., and Laleli M. Maternal. umbelical cord blood lead levels: an Istanbul Super Sanita` 1998: 34: 51–57. Study. Arch Environ Health 2001: 56: 26–28. Ojajarvi I.A., Partanen T.J., Ahlbom A., Boffetta P., Hakulinen T., Jourenkova Gottlieb S. Sustained fall in UK blood lead level. BMJ 1998: 317: 99. N., Kauppinen T.P., Kogevinas M., Porta M., Vainio H.U., Weiderpass E., Hense H.W., Filipiak B., Novak L., and Stoeppler M. Nonoccupational and Wesseling C.H. Occupational exposures and pancreatic cancer: a meta- determinants of blood lead concentrations in a general population. Int J analysis. Occup Environ Med 2000: 57: 316–324. Epidemiol 1992: 21: 753–762. Perkins L.S. Normal blood and bone marrow values in humans. In: Lee G. (Ed.). Hertz-Picciotto I. The evidence that lead increases the risk for spontaneous Wintrobe’s Clinical Hematology. 10th edn. Lippincott Williams & Wilkins, abortion. Am J Ind Med 2000: 38: 300–309. Baltimore, 1999, pp. 2738–2750. Hillman R.S. Anemia. In: Fauci A.S. (ed). Harrison’s Principles Pesch B., Haerting J., Ranft U., Klimpel A., Oelschlagel B., and Schill W. of Internal Medicine. McGraw Hill, New York, 1998, pp. 334–339. Occupational risk factors for renal cell carcinoma: agent-speciﬁc results from a Jacob B., Ritz B., Heinrich J., Hoelscher B., and Wichmann H.E. The effect of case–control study in Germany. MURC Study Group. Int J Epidemiol 2000: low-level blood lead on hematologic parameters in children. Environ Res 2000: 29: 1014–1024. 82: 150–159. Restrepo H.G., Sicard D., and Torres M.M. DNA damage and repair in cells of Kleinbaum D.G., Kupper L.L., Muller K.E., and Nizam A. Applied Regression lead exposure people. Am J Ind Med 2000: 38: 330–334. Analysis and Other Multivariable Methods. 3rd edn. PWS-Kent Publishing SAS/STAT Software. Changes and Enhancements through Release 6.12.SAS Company, Boston, 1998. Institute Inc., Cary, 1997. Lorente C., Cordier S., Bergeret A., De Walle H.E., Goujard J., Ayme S., Knill- Schwartz B.S., Stewart W.F., Bolla K.I., Simon P.D., Bandeen-Roche K., Jones R., Calzolari E., and Bianchi F. Maternal occupational risk factors for Gordon P.B., Links J.M., and Todd A.C. Past adult lead exposure is oral clefts. Occupational Exposure and Congenital Malformation Working associated with longitudinal decline in cognitive function. Neurology 2000: 55: Group. Scand J Work Environ Health 2000: 46: 137–145. 1144–1150. Mahaffey K.R., McKinney J., and Reigart J.R. Lead and compounds. In: Schwartz J., Landrigan P.J., Baker E.L., Orenstein W.A., and van Lindern I.H. Lippmann M. (Ed.). Environmental Toxicants. Human Exposures and Their Lead-induced anemia: dose-response relationships and evidence for a thresh- Health Effects. Van Nostrand Reinhold , New York, 1992, pp. 360–391. old. Am J Public Health 1990: 80: 165–168. Markowitz M. Lead poisoning: a disease for the next millennium. Curr Probl Silbergeld E.K., Waalkes M., and Rice J.M. Lead as a carcinogen: experimental Pediatr 2000: 30: 62–70. evidence and mechanisms of action. Am J Ind Med 2000: 38: 316–323.

Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(2) 193