There is evidence that heavy exposure to inorganic lead is detrimental to semen quality. 1 Epidemiologic studies have found reduced fertility rates among the families of exposed men, but the evidence so far is inconclusive. Earlier 2 and more recent studies 3 also suggest that paternal exposure to lead is associated with adverse pregnancy outcome.
We have previously examined the occurrence of spontaneous abortion and congenital malformation among the families of men biologically monitored for exposure to inorganic lead. 4,5 The findings of these studies suggested an association between lead exposure and adverse pregnancy outcome. In the study on spontaneous abortion, the effect was modified by the age of the wife. 4 Paternal exposure to lead was associated with an increased risk of spontaneous abortion among younger wives (<27 years; odds ratio (OR) 2.2, 95% confidence interval (CI) = 1.2–4.1 for estimated blood lead (PbB) ≥1.0 μmol/L). The opposite was true in the older age group [wife ≥27 years; odds ratio (OR) 0.6, CI 0.3–1.2]. The tendency in the small malformation study 5 [adjusted ORs for paternal exposure at the level of PbB ≥1.0 μmol/L were 2.7 (CI = 0.8–9.2), 2.0 (CI = 0.5–8.6), and 1.9 (CI = 0.5–7.9) for wives of age <27, 27–30, ≥31 years, respectively] was not as clear as in the spontaneous abortion study.
The aim of this register-based study was to investigate whether paternal exposure to lead is associated with infertility. Because the study on time to pregnancy 6 provided limited support for an association between paternal exposure to lead and reduced fertility, we also examined whether a bias toward no association was introduced when the study was restricted to couples with pregnancies.
Subjects and Methods
We studied the occurrence of the first marital pregnancy among men biologically monitored for exposure to inorganic lead. Altogether 56,117 measurements were conducted for 19,349 men at the Finnish Institute of Occupational Health, from 1973 through 1983. Those monitored worked in many industries. The highest exposure levels were measured in lead smelting, car radiator repair, metal scraping, metal foundries, railroad equipment machine shop work, painting, manufacture of glass, and the storage battery industry. 7 We obtained information about the wives of the monitored men from the Finnish central population register (data collected on March 31, 1986). There were 15,988 marriages for 15,262 men. We identified the pregnancies of the wives from the nationwide database on medically diagnosed pregnancies, treated in hospitals between 1973–1983.
We used the following eligibility criteria: (1) the marriage had lasted at least 2 years within the study period from 1973 to 1983; (2) the date of marriage was between July 1, 1972 and December 31, 1981; (3) if the couple had divorced, the first blood lead measurement was performed before the year of divorce; and, (4) the wife was less than 40 years of age at the time of marriage. We included only one marriage for a man. In the case of multiple marriages, the marriage during which biological measurements had been made was given priority. Otherwise, we selected the first marriage. There were 4,146 marriages that met these criteria. We classified the pregnancies that occurred during the marriage and those that ended not more than 366 days before the date of a marriage as being fathered by the husband.
We assessed paternal exposure to lead on the basis of PbB measurements. For 59% of the men, there was only one measurement, and measurements during consecutive years were only available for a minority of men. According to the instructions during the study period, based on the Act on Labor Protection, all workers doing similar tasks should be monitored periodically (1–6 times a year) if the PbB of any worker exceeds 2 μmol/L. Most of the measurements were below this limit, so no additional measurements were needed in those work places. On the basis of our earlier studies for the job from which the measurement was taken (measurement job), we estimated the median length of employment to be about 6 to 7 years. 4,5
The median time between the date of marriage and the estimated date of conception was 274 days. To improve the accuracy of exposure assessment, we divided the exposed subjects into one probably exposed group, and two potentially exposed groups. This division was based on the timing of the PbB measurements in relation to the year of marriage. We defined exposure as follows: (1) probable (N = 2111) if there was a measurement (PbB ≥0.5 μmol/L) within a 5-year time period including a calendar year preceding the year of marriage and 4 consecutive years; and (2) potential if the worker had been monitored for the last time before (N = 664) or the first time after (N = 690) the 5-year period. We assessed the level of exposure to lead as a mean PbB for each exposure group. We used five exposure categories (PbB 0.5–0.9, 1.0–1.4, 1.5–1.9, 2.0–2.4, and ≥2.5 μmol/L, corresponding to approximately 10–20, 21–30, 31–40, 41–50, and ≥51 μg/dl) for probable exposure and two categories (0.5–1.4 and ≥1.5 μmol/L) for possible exposure. We used all 681 men whose mean PbB was <0.5 μmol/L as the reference category.
We defined infertility as a non-occurrence of pregnancy or a delay of the first marital pregnancy. We estimated the relative risk of infertility for exposed men by applying a binomial regression with the SAS GENMOD procedure and a logit link. 8 For the analysis of pregnancy delay, follow-up ended at the time of conception, 1.5 years before divorce, at the end of study period or at 4 years, whichever occurred first. We grouped the follow-up time into five disjoint intervals according to the time since the date of marriage (cut-off points at 4 months and at 1 to 3 years). We handled the pregnancies that had started before the date of marriage as if they had started within 4 months. Because there were many tied events, we analyzed the follow-up data on pregnancies by using discrete proportional hazards regression with the SAS LOGISTIC procedure and a complementary log-log link. 8 We estimated success ratios (SR) for pregnancy for occupationally probably exposed men compared with unexposed or minimally exposed men (PbB <0.5 μmol/L). We calculated 95% confidence intervals using likelihood ratio statistics. Other factors that were related to infertility were older age of both spouses at the time of marriage, and a previous marriage for the husband. We also included a variable for a date of marriage before July 1, 1973 in the analysis because some of these couples may have had a pregnancy, fathered by the husband, before the study period.
A total of 523 (12%) couples had been divorced before the data collection. The prevalence of divorce was 7% among the unexposed and ranged from 13% to 23% among probably exposed men. Only 2% of the fertile couples divorced, whereas 42% of couples without pregnancies divorced. Thus, divorce rate was related to both exposure and fertility status. Nevertheless, Greenland et al.9 have pointed out that adjustment is inappropriate if a factor is affected by both exposure and the studied outcome. Furthermore, divorce is not an effect modifier in this study. Therefore, we ignored the divorce variable in the regression analysis and kept the divorced couples in our data set.
Paternal exposure to lead was associated with infertility. Overall, 26% of the couples did not have pregnancies during the study period from 1973 to 1983. This figure was 22% in the reference group, and ranged from 28% to 35% among the occupationally exposed men (Table 1). The risk ratios for infertility were consistently greater with increasing exposure to lead among men probably exposed to lead. Men monitored for the first time after the selected 5-year period also experienced an increased risk of infertility.
In the follow-up analysis among fertile and infertile couples, we observed a delayed pregnancy among men probably exposed to lead (Table 2). The results were similar when the group with probable exposure was restricted to men monitored for at least 4 years within the relevant 5-year time period (Table 3). Fertility was reduced in exposed storage battery workers but not among workers in lead smelting or metal foundry work. In analyses of the wife’s age at the time of marriage, the overall association was strongest in the oldest age group, although high levels of exposure were also related to reduced fertility in younger age groups. Excluding infertile couples reduced the association between lead exposure and fertility (Table 3).
Our results suggest that men occupationally exposed to inorganic lead had reduced fertility compared with those who were occupationally unexposed or minimally exposed. We saw no clear association between exposure to lead and reduced fertility in the analyses restricted to couples who had achieved a pregnancy. Thus, the association appears to be related to childlessness rather than to delay of pregnancy among fertile couples.
Infertility caused by exposure to lead is a likely explanation for the findings. Studies on semen quality suggest adverse effects for lead on various semen parameters at PbB levels of >1.9 μmol/L. 1 Poor semen quality has been linked to reduced fertility in a recent Danish study. 10 Thus, it is biologically plausible that heavy exposure to lead is associated with reduced fertility. In addition to spermotoxic and endocrinological effects, there are other potential mechanisms through which lead can affect fertility. These include genetic or epigenetic effects, such as germ-cell mutations, alterations in chromatin stability, or changes in DNA methylation pattern. 11–13 Alterations in DNA methylation are related to gene activity and have been observed in rat liver cells exposed to lead. 13 Recently, male-mediated effects have been observed in genomic expression in 2-cell embryos fathered by male rats with blood lead levels as low as 15–25 μg/dl. 14 The findings suggest an effect on regulation of gene transcription or translation rather than genetic damage to the male germ cell.
Three previous studies have found reduced fertility rates among lead-exposed workers. 15–17 In two studies, effects were observed at PbB levels (mean 46.3 range 24–74 μg/dl, 16 and ≥50 μg/dl and duration of exposure ≥5 years 17) similar to those of the two highest exposure categories of the present study. Both of these studies indicated the importance of the duration of exposure in addition to the intensity of exposure. Unfortunately, we could not assess the effects of the duration of exposure. A third study 15 reported reduced standardized fertility ratios already at a PbB level of ≥25 μg/dl.
Two studies found no effect on fertility rates for exposure to lead at mean PbB levels of 35 μg/dl and 46.3 μg/dl, respectively. 18,19 In the French study, 18 the unexposed workers were selected from the same factory as the exposed ones. Blood lead was not measured for the reference group, and therefore the possibility cannot be excluded that the unexposed workers were moderately exposed to lead. The authors considered the diluting effect of this potential bias to be probably small. Their person-years analysis used a binomial model instead of the correct Poisson model, and the resulting model misspecification may have biased the results toward underestimation of the risk. Also, socioeconomic and cultural differences between exposed and unexposed workers may have biased the results away from an association. Eighty-five percent of the workers were North African, who generally have more children than French men.
A possible explanation for a negative finding in a person-years analysis is that reduced fecundity (biological ability to conceive) does not necessarily translate into reduced fertility (birth rate) if most of the couples plan the size of their family, as was stated by the Danish authors. 19 Our study focused on the couples and the time elapsed for the first pregnancy. This type of analysis is more powerful to detect an association with fertility and less prone to a bias caused by the dependence of fertility rate on socioeconomic factors than the person-years analysis.
Validity of Exposure Information
All the men were biologically monitored at least once between 1973 to 1983, but only a minority of men had been monitored in consecutive years. Therefore, misclassification of exposure is obviously present in our data. Because we used a 5-year time window for probable exposure that coincided with the 4-year follow-up, we estimate that about 75 to 80% of the probably exposed men were holding the measurement job during the follow-up.
The duration of employment in the most heavily exposed jobs may have been shorter than in the less exposed work tasks. Nevertheless, the results were similar in the analysis restricted to men monitored during consecutive years. Moreover, 65% of the annual means of PbB during the 5-year period equaled our exposure estimate among regularly monitored men (≥3 measurement years), 33% of the means were in an adjacent category, and only 2% of the means differed for two categories. These figures suggest that our results were not related to exposure misclassification. In some skilled occupations, such as car mechanics and welders, the change of work place did not necessarily mean a change in exposure level. The observation that fertility was reduced, especially among probably exposed workers with the lowest potential for misclassification of exposure, supports the study hypothesis.
Time-trend bias, a pitfall inherent in retrospective studies on time to pregnancy, can be avoided by doing a prospective study. 20 One strength of our study design is its lack of bias related to time trends in exposure, since we used a synthetic prospective approach in following the occurrence of pregnancies from marriage date.
Other occupational exposures, such as other metals or organic solvents, may have had an effect on fertility. One half of the men were categorized as occupationally exposed to organic solvents in the study on spontaneous abortion. 4 Exposure to solvents was most prevalent at PbB levels of 0.5 to 1.4 μmol/L (eg, metal painters and car mechanics). Thus, the reduction in fertility at low exposure levels could be explained by occupational exposure to solvents. In the time-to-pregnancy study among men exposed to organic solvents, however, occupational exposure to these agents was only suggestively associated with reduced fecundability. 21 In addition, no confounding effect was observed for paternal exposure to solvents in the time-to-pregnancy study among men exposed to lead. 6
Exposure to other metals was observed to be most common at high exposure levels (eg, in brass foundry work). 4 As in the case of solvents, however, the inclusion of a variable for exposure to other metals did not change the parameter estimates for paternal exposure to lead in the study on time to pregnancy. 6 Exposure to lead was associated with reduced fertility among storage battery workers whereas no clear association was found among metal foundry or lead smelting workers, who are likely exposed to other metals. Similarly, exposure to cadmium or manganese was not related to fertility in the Belgian study, which found a decrease in fertility rates among lead-exposed workers. 16 These observations do not indicate confounding by exposure to other occupational agents.
Validity of Pregnancy Information
The pregnancies of the wives were identified from the medical records in 1973–1983. This database covers 94% of all officially recorded births and 85% of all recorded induced abortions in Finland. 22 The coverage is about 80 to 90% for spontaneous abortions (depending on the gestational age). It is not likely that these deficiencies in our database on pregnancies could have any important effect on the results.
We do not know whether the childless couples had planned for pregnancies. Part of the observed association may be explained by the tendency of selecting exposing jobs among voluntarily childless men. The observations that younger (17–25 years) and unmarried men (data not shown) had a higher prevalence of heavy exposure than older and married men, respectively, are compatible with this explanation. Because only registered information was available, this shortcoming cannot be completely controlled. Also, the crude outcome measure in the study on pregnancy delay might have weakened the findings.
Impact of the Findings for the Studies on Time to Pregnancy and Pregnancy Outcome
For epidemiologic studies on reproductive health, it has been hypothesized that a phenomenon called “dose-response fallacy” may occur when only recognized pregnancies are studied. 23 Congenital malformations, spontaneous abortions, subclinical abortions, and infertility fall on a continuum according to biological severity. If the risk of a less severe health outcome falls due to the increasing risk of more severe health outcomes, along with increasing exposure, a non-traditional dose-response relation may be observed in a study of one type of health outcome.
For retrospective studies on time to pregnancy in particular, it has been stated that “if the exposure is associated with sterility or long times to pregnancies, the selection of subjects with pregnancies only will bias the results towards the null.”24 Although the presence of this bias is intuitively obvious when the exposure under study is detrimental to fertility, it can seldom be verified. Our data, although limited, suggest that a substantial bias toward the null may have been introduced into our study on time to pregnancy by restriction to couples with pregnancies. 6
Age is known to be related to fertility. Also, the susceptibility to toxic substances may vary according to age. The effects of lead on pregnancy outcome seemed to be modified by the age of the wife in our study on spontaneous abortion. 4 In the present study, the association between lead exposure and reduced fertility seemed to be stronger in older age groups. A similar finding was made in the Danish study. 19 In that study, exposure to lead was associated with reduced fertility among older men but not among younger men (cutoff point at 30 years of age) in a subset of battery workers with at least one PbB >20 μg/dl. (Jens Peter Bonde, personal communication, 1999). The dose-response fallacy can be a potential explanation for the low risk of spontaneous abortion for exposed men among the older wives in our previous study. 4
We thank Sven Hernberg for his valuable comments on the draft of this paper. We are indebted to the staff of the Laboratory of Biochemistry at the Institute for their skillful technical assistance.
1. Apostoli P, Kiss P, Porru S, Bonde JP, Vanhoorne M, and the ASCLEPIOS study group. Male reproductive toxicity of lead in animals and humans. Occup Environ Med 1998; 55:364–374.
2. Uzych L. Teratogenesis and mutagenesis associated with the exposure of human males to lead: a review. Yale J Biol Med 1985; 58:9–17.
3. Anttila A, Sallmén M. Effects of parental occupational exposure to lead and other metals on spontaneous abortion. J Occup Environ Med 1995; 37:915–921.
4. Lindbohm M-L, Sallmén M, Anttila A, Taskinen H, Hemminki K. Paternal occupational lead exposure and spontaneous abortion. Scand J Work Environ Health 1991; 17:95–103.
5. Sallmén M, Lindbohm M-L, Anttila A, Taskinen H, Hemminki K. Paternal occupational lead exposure and congenital malformations. J Epidemiol Community Health 1992; 46:519–522.
6. Sallmén M, Lindbohm M-L, Anttila A, Taskinen H, Hemminki K. Time to pregnancy among the wives of men occupationally exposed to lead. Epidemiology 2000; 11:141–147.
7. Anttila A. Occupational exposure to lead and risk of cancer. Acta Universitatis Tamperensis ser A vol. 417, Tampere, Finland. Academic dissertation.
8. SAS Institute Inc. SAS/STAT®
Software: Changes and Enhancements through Release 6.11, Cary, NC: SAS Institute Inc.
9. Greenland S, Pearl J, Robins JM. Causal diagrams for epidemiologic research. Epidemiology 1999; 10:37–48.
10. Bonde JPE, Ernst E, Kold Jensen T, Hjollund NHI, Kolstad H, Hendriksen TB, Scheike T, Givercman A, Olsen J, Skakkebæk NE. Relation between semen quality and fertility: a population–based study on 430 first-pregnancy planners. Lancet 1998; 352:1172–1177.
11. Olshan AF, Faustman EM. Male-mediated developmental toxicity. Ann Rev Publ Health 1993; 14:159–181.
12. Johansson L, Pelliciari CE. Lead-induced changes in the stabilization of the mouse sperm chromatin. Toxicology 1988; 51:11–24.
13. Kanduc D, Rossiello MR, Aresta A, Cavazza A, Quagliariello E, Farber E. Transitory DNA hypomethylation during liver cell proliferation induced by a single dose of lead nitrate. Arch Biochem Biophys 1991; 286:212–216.
14. Gandley R, Anderson L, Silbergeld EK. Lead: male-mediated effects on reproduction and development in the rat. Environ Res 1999; 80:355–363.
15. Selevan SG, Hornung R, Kissling GE, Cottrill C, Leffingwell SG. Reproductive outcomes in wives of lead exposed workers. Cincinnati, OH: US National Institute for Occupational Safety and Health, Department of Health and Human Services; 1984(PB85–220879):1–42.
16. Gennart J-P, Buchet J-P, Roels H, Ghyselen P, Ceulemans E, Lauwerys R. Fertility of male workers exposed to cadmium, lead, or manganese. Am J Epidemiol 1992; 135:1208–1219.
17. Lin S, Hwang S-A, Marshall EG, Stone R, Chen J. Fertility rates among lead workers and professional drivers: a comparative study. Ann Epidemiol 1996; 6:201–208.
18. Coste J, Manderau L, Pessione F, Bregu M, Faye C, Hemon D, Spira A. Lead-exposed workmen and fertility: a cohort study on 354 subjects. Eur J Epidemiol 1991; 7:154–158.
19. Bonde J, Kolstad H. Fertility of Danish battery workers exposed to lead. Int J Epidemiol 1997; 26:1281–1288.
20. Weinberg CR, Baird DD, Wilcox AJ. Sources of bias in studies of time to pregnancy. Stat Med 1994; 13:671–681.
21. Sallmén M, Lindbohm M-L, Anttila A, Kyyrönen P, Taskinen H, Nykyri E, Hemminki K. Time to pregnancy among the wives of men exposed to organic solvents. Occup Environ Med 1998; 55:24–30.
22. Lindbohm M-L, Hemminki K. Nationwide data base on medically diagnosed spontaneous abortions in Finland. Int J Epidemiol 1988; 17:568–573.
23. Selevan SG, Lemasters KL. The dose-response fallacy in human reproductive studies of toxic exposures. J Occup Med 1987; 29:451–454.
24. Baird DD, Wilcox AJ, Weinberg CR. Use of time to pregnancy to study environmental exposures. Am J Epidemiol 1986; 124:470–480.
© 2000 Lippincott Williams & Wilkins, Inc.