Lower sperm counts following prenatal tobacco exposure

Abstract

BACKGROUND: Recent findings have indicated that prenatal exposure to tobacco smoke may cause lower sperm concentration in ejaculates of adult men. To extend the research on this hypothesis we investigated the dose-dependency of the association, controlling for other prenatal exposures. METHODS: From 1987 to 1996, four separate occupational semen studies were conducted at three centres in Denmark. A total of 945 men provided semen and blood samples, and information on reproductive and lifestyle factors. In 2004, we collected data on the maternal smoking habits during pregnancy from 522 mothers of the participating men. RESULTS: Adjusting for study subgroup, abstinence time and other factors, we found statistically non-significant differences in mean sperm concentrations: 65.0×10⁶/ml [95% confidence interval (CI) 51, 81] among sons of non-smokers; 59.1×10⁶/ml (95% CI 46, 75) among sons of mothers who had smoked 1–10 cigarettes/day; and 57.7×10⁶/ml (95% CI 40, 81) among those whose mothers had smoked >10 cigarettes/day. The former group had a higher odds ratio (OR) for oligozoospermia (sperm concentration ≤20×10⁶/ml) of 1.5 (95% CI 0.9, 2.8), the latter group an OR of 2.6 (95% CI 1.2, 5.8). CONCLUSION: We observed a dose-dependent association between prenatal tobacco exposure, lower sperm concentration and higher risk of oligozoospermia.

Introduction

MacKenzie and Angevine (1981) showed markedly lower fertility in mice exposed in utero to benzo(a)pyrene, a mutagenic polycyclic aromatic hydrocarbon and constituent of tobacco smoke (International Agency for Research on Cancer, 1986). This finding prompted investigations in human populations, and Baird and Wilcox (1986) reported unaltered fecundity (probability of conception during one menstrual cycle) following prenatal exposure to tobacco smoke in a study of 678 couples. Ratcliffe et al. (1992) subsequently investigated the effect of maternal smoking during pregnancy on semen characteristics in adult sons, whose mothers had participated in a randomized clinical trial of diethylstilbestrol use during pregnancy. No significant effect on semen characteristics was reported, and the two ‘negative studies’ indicated no need for further research on the ‘smoking hypothesis’.

Jensen et al. (1998) reported a lower fecundity where the males had experienced prenatal exposure to tobacco smoke in a study of 430 Danish couples. A more recent study of 265 male twins and singleton brothers (Storgaard et al., 2003) reported a 48% lower sperm concentration (×10⁶ sperm per ml semen) among sons exposed to smoking >10 cigarettes/day during pregnancy. Finally, in a study of 1770 army conscripts from five European countries, Jensen et al. (2004) reported a 20% lower sperm concentration among sons of pregnancy-smokers.

This study was carried out to further investigate the dose-dependent association between maternal smoking during pregnancy and sperm concentration in ejaculates of adult offspring. Storgaard et al. (2003) also reported decreased serum inhibin B levels and increased serum levels of FSH in men exposed to smoking >10 cigarettes/day during pregnancy; as a secondary aim we wanted to see if we could corroborate these findings in our data.

Materials and methods

Population

From 1987 to 1996 four separate occupational semen studies were conducted at three centres in Denmark, where a total of 945 men provided semen and blood samples, and information on reproductive and lifestyle factors. The First Pregnancy Planner study (Planners) was conducted in 1992–1995 at the Aarhus and Copenhagen centres (Bonde et al., 1998). A study of semen quality and sexual hormones in greenhouse workers (Gardeners) was conducted in 1994 at the Aarhus centre (Abell et al., 2000). A study of semen quality and sex hormones in steel welders (metalworkers) was conducted in 1987 at the Aalborg centre (Bonde, 1990) and a study of semen quality in pesticide-spraying farmers (Farmers) was conducted in 1995–1996 at the Aarhus Centre (Larsen et al., 1998). The original study populations and the drop-outs during the recruitment to the present study are presented in Table I.

The men were between 18 and 50 years of age, and the common exclusion criteria included vasectomy, known azoospermia and abnormalities of the reproductive organs. In the Planners study the participants had to be cohabiting with a person of the opposite sex, to be planning discontinuation of contraception within the study period and to have no prior knowledge of their fertility to fulfil the inclusion criteria. The three other studies used cross-sectional samples of current employees in well-defined trades or workplaces. A detailed description of the four populations is given in the respective papers (Bonde, 1990; Bonde et al., 1998; Larsen et al., 1998; Abell et al., 2000).

We were able to identify appropriate semen data (sperm concentration and semen volume) and the unique 10-digit personal identification code of The Danish Civil Registration System for 875 men (sons), which comprised the study population in the present study. We identified their mothers, from whom we wanted to collect information on exposures during their pregnancies with their sons. The identities, addresses and vital statuses of the 875 mothers were provided by The Danish Civil Registration System. Among these mothers, 165 were dead. Seventeen sons or mothers were resident outside Denmark, 26 sons had indicated unwillingness to participate in research (i.e. not wanting to contribute administrative data to research), 29 mothers were not registered (i.e. mothers' identity not available) and five sons were dead, leaving 633 pairs of mothers and sons as eligible. Among these, 522 pairs participated which yielded an overall participation rate of 82%.

Semen characteristics

The number of semen samples obtained from each son differed between the four studies, but we used only the first sample. In all studies, the men were asked to collect semen by masturbation. Samples were collected at home and examined in a mobile laboratory or at a hospital within 2 h (few exceptions) (Bonde et al., 1996). The men were instructed to keep the sample in a pocket close to the body during transportation to avoid cooling. The duration of sexual abstinence, spillage (if any), recent fever and current smoking were recorded in a self-completed questionnaire filled in when the men delivered the sample. The semen volume was measured in a graded tube with 0.1 ml accuracy. Formalin-fixed sperm were counted in either a Makler or Neubauer chamber (Aarhus) or a Bürger–Türk chamber (Copenhagen and Aalborg) using a phase-contrast microscope. It has previously been demonstrated that the use of different counting chambers gives similar results of sperm concentration (Auger et al., 2000). The total sperm count was computed by multiplying the sperm concentration by the semen volume. All the analyses were carried out by trained laboratory technicians in accordance with the guidelines of the World Health Organization (1992).

Reproductive hormones

Blood samples were drawn by venipuncture, when the semen samples were delivered, for measurement of FSH, LH, testosterone and sex hormone-binding globulin (SHBG). Inhibin B was measured in the Planners and Farmers studies only. The samples were stored at −20°C or −80°C until analysis. Testosterone was measured by radioimmunoassay, whereas FSH, LH, SHBG and inhibin B were measured by immunometric assay techniques as described in the respective papers (Bonde, 1990; Jensen et al., 1997; Larsen et al., 1998; Abell et al., 2000). According to the study hypothesis, only FSH and inhibin B were considered in the statistical analysis.

Assessment of prenatal tobacco exposure

We contacted the sons by letter, explaining our intention to send a questionnaire to their mothers, and 78 sons (12%) refused participation. The main reasons were illness or age-related weakness of the mother and a wish to keep private earlier participation in a semen study. In the spring of 2004, the remaining 555 mothers received a questionnaire by mail, and non-responders were re-contacted once with a new questionnaire. A total of 522 mothers (94%) responded to the first or second questionnaire. The questionnaire was made personal to the mothers by stating the son's name and birth date, which also helped multiparous women to think of the right pregnancy. The study approach was approved by the local ethics committee.

The mothers were asked: ‘Have you ever smoked while pregnant with your son? (please also think of the very first months of the pregnancy)’. The questions on smoking during pregnancy had the following answering categories: ‘no smoking’, 1–5 cigarettes daily, 6–10, 11–15, 16–20, 21–25, >25 cigarettes daily, and ‘do not recall’. The same structure was applied to cheroots, but with intervals of one cheroot daily up to 6, and >6 cheroots daily. Pipe-smokers were asked about grams of tobacco smoked per week. Pipe and cheroot smoking were recoded as daily cigarette smoking according to estimated contents of tobacco (e.g. one cheroot corresponds to 4 cigarettes and 50 g of pipe tobacco per week corresponds to 1–10 cigarettes daily). We asked equivalent questions on the father's smoking habits while the mother was pregnant with the son. In addition, questions were asked about a number of other pregnancy-related exposures and events, as presented in Table II. The distribution of maternal smoking during pregnancy was as follows: non-smoker, 63%; 1–5 cigarettes/day, 13%; 6–10 cigarettes/day, 13%; 11–15 cigarettes/day, 5%; and 16–20 cigarettes/day, 2%. Other categories each contained <1% of the mothers. Forty-two percent of the mothers smoked before pregnancy, 5% stopped when becoming pregnant.

Statistical methods and analysis

The distribution of maternal smoking during pregnancy was divided into three categories, equivalent to: no smoking (63%), 1–10 cigarettes/day (28%) and >10 cigarettes/day (9%). The 1–5 and 6–10 cigarettes/day categories were combined because similar effects were observed in the two groups. Categories of >10 cigarettes/day were combined due to small numbers. In the following, 1–10 cigarettes/day will be referred to as light smoking and >10 cigarettes/day as heavy smoking.

The continuous outcome variables (sperm concentration, total sperm count, serum FSH and inhibin B) were transformed by the cubic root to normalize their distributions. We analysed these outcomes by multiple linear regression in a general linear model using the three categories of maternal smoking during pregnancy in a categorical explanatory variable, controlling for relevant confounders as described below. The adjusted results are presented as back-transformed least-square means with 95% confidence intervals (CI). We used the constructed smoking categories in a numerical explanatory variable in the model (1=‘non-smoker’; 2=‘1–10 cigarettes/day’; 3=‘>10 cigarettes/day’) to evaluate any linear association between maternal smoking during pregnancy and the outcome variables. The results are presented as regression coefficients (β) with 95% CI, and the unit is change in cubic root-transformed sperm concentration, total sperm count, FSH and inhibin B, respectively, per increase in smoking category.

We examined oligozoospermia (sperm concentration ≤20×10⁶/ml) as a dichotomous outcome in a logistic model, using either the three categories of maternal smoking during pregnancy or pregnancy-smoker versus non-smoker as explanatory variables, controlling for relevant confounders and presenting the results as odds ratios (OR) with 95% CI.

When appropriate, we stratified the analyses by study subgroup to check for homogeneity in effect measures. Potential confounding factors were identified and grouped into obligate confounders and potential confounders. The study subgroup (categorical variable) and the obligate confounders [period of sexual abstinence prior to semen sampling (continuous, log-transformed) and, in analysis of FSH and inhibin B, time of blood sampling (before 09:00, 09:00–12:00 and after 12:00)] were included in the models regardless of their effect. The potential confounders included: son's age at semen sampling (continuous), son's current smoking at semen sampling (yes/no), season at semen sampling (April–September/October–March), mother's age when giving birth to the son (continuous), mother's use of oral contraceptives prior to pregnancy (yes/no), mother's intake of alcohol during pregnancy (yes/no), mother's intake of coffee during pregnancy (yes/no) and duration of breastfeeding the son (0, ≤3 months, >3 months).

Among these, son's current smoking, duration of breastfeeding and use of oral contraceptives changed one or more estimates >5% and were retained in the models. The other potential confounders changed the estimates <5% and were not carried any further. The same three confounders were identified regardless of whether we used forward or backward selection in the models. Missing values in the breastfeeding (n=18) and oral contraceptive (n=14) variables were recoded to a missing category in each variable and included in the models. Re-analysis excluding observations with missing values showed essentially identical results to those presented.

Results

Characteristics of the participating mothers and sons are presented in Table II. Birthweight of the son, duration of breastfeeding and the mother's age were lower and intake of coffee, alcohol and oral contraceptives higher among maternal smokers. The prevalence of fathers smoking during pregnancy and sons smoking at sampling were higher if the mother smoked during pregnancy.

The sperm concentration was lower with increasing prenatal tobacco exposure in a dose-dependent manner, from 65.0×10⁶/ml (95% CI 51–81) among non-smokers to 57.7×10⁶/ml (95% CI 40, 81) among sons of heavy-smoking mothers (Table III). Likewise, the average total sperm count was lower with increasing prenatal tobacco exposure in a dose-dependent manner, from 165×10⁶ (95% CI 126, 211) among non-smokers to 149×10⁶ (95% CI 97, 217) among sons of heavy-smoking mothers. We observed a statistically non-significant negative linear trend in sperm concentration with increasing prenatal tobacco exposure in the complete dataset (β=–0.10, 95% CI −0.26, 0.07), and among Planners alone (β=–0.18, 95% CI −0.38, 0.03). However, no negative linear trend was demonstrated in a subgroup analysis of Gardeners, Farmers and Metalworkers combined (β=0.04, 95% CI −0.24, 0.32). The sons' serum levels of FSH and inhibin B were not significantly different across the smoking strata (Table III).

The crude estimates of sperm concentration for each subgroup are presented in Table IV. A marked decrease in mean and median sperm concentration with increasing prenatal tobacco exposure was observed in the Planners subgroup, but not in the three other subgroups. Apart from Planners, the >10 cigarettes/day strata contained too few observations to contribute valid information.

Seventy-five sons had oligozoospermia (Table V), and the adjusted OR for oligozoospermia among sons exposed to prenatal smoking of >10 cigarettes/day was 2.6 (95% CI 1.2, 5.8). For sons exposed to 1–10 cigarettes/day the OR was 1.5 (95% CI 0.9, 2.8). We fitted an adjusted logistic model with the explanatory smoking categories in a numerical variable and found an adjusted OR of 1.6 (95% CI 1.1, 2.3) for each increment. A stratified analysis showed similar crude OR for oligozoospermia (judged from the point estimates) in three of the four subgroups, with Farmers being deviant (Table V). However, very few farmers were exposed to pregnancy smoking of >10 cigarettes/day (Table IV).

We repeated the data analysis using paternal smoking during pregnancy as explanatory variable, controlling for maternal smoking during pregnancy, and no indication of an association between paternal smoking during pregnancy and oligozoospermia in the sons was observed. The reported associations between maternal smoking during pregnancy and oligozoospermia in the sons did not change if we included paternal smoking in the models.

We repeated the data analysis excluding sons with azoospermia (n=6), because this condition is more likely due to other factors than smoking (e.g. Sertoli cell-only syndrome, Klinefelter syndrome, congenital aplasia or obstruction of the duct system). All reported associations were strengthened, resulting in a more negative linear trend in sperm concentration with increasing maternal smoking during pregnancy in the complete dataset (β=−0.13, 95% CI −0.28, 0.03) and among Planners alone (β=−0.28, 95% CI −0.40, −0.02). The adjusted OR for oligozoospermia were 1.7 (95% CI 0.9, 3.0) for light maternal smoking during pregnancy and 3.0 (95% CI 1.3, 6.9) for heavy maternal smoking during pregnancy. The adjusted OR for one increment in smoking group was 1.7 (95% CI 1.2–2.5).

We compared the smoking information from the mothers with previously collected equivalent information from the sons (Table VI). We were able to do this for 265 Planners, and found a simple kappa coefficient of 0.8 (95% CI 0.7, 0.9), a sensitivity of 87% and a specificity of 93%, yielding a predictive value of a positive answer from the son of 88% and of a negative answer of 92%. By restricting the analysis to observations with complete agreement between son and mother (n=240), we found a crude OR for oligozoospermia of 1.9 (95% CI 1.0, 3.9) among sons exposed to maternal smoking during pregnancy. The equivalent crude OR for oligozoospermia was 1.6 (95% CI 0.8, 3.1) among the 265 Planners.

Discussion

Our data showed approximately 10% lower mean sperm concentrations and total sperm counts and higher OR for oligozoospermia, with increasing exposure to maternal smoking during pregnancy. The lower sperm concentrations and total sperm counts were not statistically significant. A dose-dependent association was observed for both sperm concentration and risk of oligozoospermia. No association between maternal smoking during pregnancy and sons' serum levels of FSH and inhibin B was observed.

The mothers' mean age was 66.2 years, and the mean recall time 40.2 years. Studies have reported high accuracy of recalled smoking habits and birthweight compared with medical records obtained on pregnancies 10–15 years earlier (Seidman et al., 1987; Yawn et al., 1998) and the reproducibility of long-term recall of smoking habits in the elderly also seems to be high (Cumming and Klineberg, 1994). To validate the accuracy of the exposure information we asked the mothers for the birthweight of the sons. The weight was available for 97% (n=498) of the population: compared with non-smokers, light maternal smoking during pregnancy was associated with a 175 g (95% CI 53, 296) lower birthweight for the sons, heavy maternal smoking during pregnancy with a 297 g (95% CI 101, 493) lower birthweight. This is equivalent to lower birthweights seen in larger prospective studies (Wilcox, 1993; England et al., 2001) and it augments the validity of the exposure information obtained.

Furthermore, in an earlier study on birthweight and semen characteristics (Olsen et al., 2000), the birthweights of 201 sons participating in this study were collected from midwives' records stored in regional public archives. At the time these sons were born, each midwife had her own protocol in which she consecutively recorded standardized health examinations performed shortly after the delivery, including measurement of birthweight: compared with non-smokers (n=121), light maternal smoking during pregnancy was associated with a 136 g (95% CI −22, 293) lower birthweight in the sons (n=60), heavy maternal smoking during pregnancy with a 249 g (95% CI 8, 490) lower birthweight (n=20). Hence, the use of prospectively recorded birthweights instead of recalled birthweights did not affect our conclusions regarding the validity of the exposure assessment.

Given the retrospective nature of the exposure information and the mothers' possible knowledge of their sons fertility status, the influence of recall bias has to be considered. We calculated the mean birthweights among sons with and without prenatal exposure to tobacco smoke, stratified by sons' oligozoospermia, to evaluate whether recall of exposure was dependent on sons' fertility status (oligozoospermia): compared with non-exposed, prenatal tobacco exposure was associated with a 225 g (95% CI −501, 52) lower birthweight among sons with oligozoospermia, and a 195 g (95% CI −318, −71) lower birthweight among sons without oligozoospermia. The same magnitude of lower birthweight in the two groups makes the presence of recall bias dependent on son's fertility status very unlikely.

We compared the information on prenatal smoking from 265 mothers with equivalent information collected from the sons 10 years earlier and before knowing their reproductive potential (Jensen et al., 1998). The sons were asked: ‘Did your mother smoke when she was pregnant with you?’. They were also asked to report whether this information was provided by their mother or not. Concordant with other findings (Sandler and Shore, 1986), high levels of agreement between information from the mothers and the sons was observed (Table VI). To evaluate the possibility of differential misclassification of exposure, we compared the agreement between information from the sons and the mothers, stratified by sons' oligozoospermia. No differences were found between the two groups, which indicates an unaltered reporting pattern in the oligozoospermia group before and after (possible) knowledge of the son's low reproductive capability. Furthermore, restriction of statistical analysis to observations with complete agreement between son and mother (n=240) strengthened the association compared with that found in the 265 Planners for whom we compared information. This indicates that any misclassification of exposure is non-differential, leading towards an underestimation of the actual adverse effect.

Some selection was made during recruitment, but we were able to characterize outcome measures in participants (n=522), non-participants (n=111) and non-eligible subjects (n=242) within this study. Mean values of sperm concentration, total sperm count, serum levels of FSH and inhibin B did not differ significantly between the three groups (i.e. no selection dependent on fertility within this study).

Our study was performed on men who volunteered for studies on semen quality, and we expect such a population to be selected towards lower fertility (Bonde et al., 1996, 1998). Should this selection explain the results, participation would, however, have been associated with both lower sperm concentration and higher prevalence of exposure to maternal smoking during pregnancy, but this is unlikely. Comparing the smoking prevalence among participants in this study (42% prior to pregnancy) with a smoking prevalence of ∼50% among women aged 20–29 years estimated in nationwide health surveys, conducted in 1953–1954 and 1986–1987, does not suggest strong selection dependent on smoking (Osler, 1992). In agreement with our observations, the smoking prevalence among rural residents (Farmers) was lower than among provincial and metropolitan residents in the health surveys (Osler, 1992). In summary, we do not believe that the observed associations are spurious either because of selection bias or recall bias.

We were able to control for other prenatal exposures, such as alcohol and coffee intake, and none of these seemed to confound the results. Confounding by illness was not suspected given the distribution across smoking strata (Table II). Confounding from an uneven distribution of men with azoospermia was not present. It was not possible to control for the potential confounding from socioeconomics, but the sons were originally recruited from specific occupations or trade unions, making the span in socioeconomic differentials small. The finding of lower sperm concentrations among younger army conscripts exposed to maternal smoking during pregnancy works against the idea of confounding by exposures in adulthood (Jensen et al., 2004). Furthermore, if the observed association is really caused by lifestyle or dietary habits or any other unknown factor associated with smoking, we would expect maternal and paternal smoking during pregnancy to be equally good proxy measures of this factor. The lack of association between paternal smoking during pregnancy and oligozoospermia in the sons does not support this.

Including sons' birthweights in the statistical models only marginally reduced the strength of the associations, indicating that only a small part of the suggested effect of maternal smoking during pregnancy on sperm concentration, if any, is mediated through lower birthweight. Jensen et al. (2004) suggested birthweight as an intermediary factor, accounting for <25% of the difference in lower sperm concentration and smaller testis size. Previous studies on the association between birthweight and semen characteristics have been conflicting (Francois et al., 1997; Olsen et al., 2000) but it seems reasonable to suggest that specific adverse effects to the male embryonal gonads are responsible for the largest part of the observed association.

Storgaard et al. (2003) found no dose-dependent association, but only reported lower sperm concentration among sons exposed to maternal smoking during pregnancy of >10 cigarettes/day. This discrepancy is, however, within the limits of random variation. Storgaard et al. also reported higher FSH (statistically non-significant) and lower inhibin B among the heavily exposed sons. We were not able to corroborate these findings, and whether or not prenatal exposure to tobacco smoke constituents is a mechanism affecting FSH and inhibin B levels in adulthood remains to be demonstrated. The fact that serum measurements of FSH and inhibin B were not available for all participants in this study should be kept in mind when evaluating our findings.

Ratcliffe et al. (1992) found no differences in sperm concentration but the possible adverse effect of prenatal tobacco exposure may have been concealed by lower sperm concentrations among diethylstilbestrol-exposed sons (Gill et al., 1979; Schumacher et al., 1981). In fact, an increased prevalence of oligozoospermia among sons exposed in utero to tobacco smoke was found when the analysis was confined to subjects without prenatal exposure to diethylstilbestrol.

In this study only 9% of the participating sons were exposed to maternal smoking of >10 cigarettes/day during pregnancy, and it provided insufficient statistical power to elucidate the effects of exposure to >15 cigarettes/day. Hopefully future studies using balanced sampling by maternal smoking during pregnancy will contribute information on adverse effects of higher exposure levels.

In conclusion, our data showed that sons exposed in utero to tobacco smoke had lower sperm concentrations and increased risk for oligozoospermia. A dose-dependent association was observed for both sperm concentration and oligozoospermia, but the sperm concentrations were not significantly lower. In our judgement the data provide support to the hypothesis that maternal smoking during pregnancy is associated with lower spermatogenic capacity in the adult sons, though inference as to causality remains premature. No association between exposure to maternal smoking during pregnancy and sons' serum levels of FSH and inhibin B was observed.

We are indebted to professor Jørn Olsen, The Danish Epidemiology Science Centre, Aarhus University, and to the doctors: Solveig Brixen Larsen, Tina Kold Jensen, Niels Henrik I Hjøllund, Anette Abell and Tine Brink Henriksen, for their great effort in the planning and conduct of the original semen studies. This study was supported by grant number 2004b137 from The Health Insurance Foundation (Sygekassernes Helsefond).

References