Many women workers are exposed to organic solvents,1 which are widely used in a range of industrial products including paints, varnishes, inks and cleaning agents. Solvents belong to three main chemical families: oxygenated, petroleum and chlorinated solvents. Their main routes of exposure are dermal and respiratory. Despite their widespread use, their health effects on reproduction, especially their relation to congenital malformations, remain controversial.

Several solvents have been shown to be teratogenic for animals.2 3 In mice, for example, toluene and xylene (petroleum solvents) have been associated with the occurrence of cleft palate,2 and ethylene glycol monomethyl ether has been associated with the occurrence of neural tube defects.3

In humans, malformations and cytogenetic effects have been observed among the offspring of women exposed to glycol ethers during pregnancy.4 Some studies,5^–8 but not others,9 report an excess risk of spontaneous abortion among women occupationally exposed to solvents. A small prospective cohort,10 and a meta-analysis,11 performed by the same research group both report associations between maternal occupational exposure to solvents and major malformations. Two recent occupational cohort studies of women working in laboratories suggest similar results.12 13

Various case-control studies have shown relations between maternal occupational exposure to solvents and some subtypes of malformations, mostly oral clefts.14^–16 Some significant associations have also been reported between maternal exposure to solvents and cardiac malformations,14 17 and neural tube defects,14 18 but all the studies are not consistent.19

Because environmental factors are suspected of responsibility for the apparent increased incidence of male genital malformations,20 we think it is worthwhile to study their association with maternal occupational exposure to solvents.

Our prospective population-based cohort study was specifically designed to test the hypothesis of an impact of maternal exposure to solvents on the risk of congenital malformations. It sought to overcome several of the drawbacks of previous studies by using exposure data collected at the beginning of pregnancy.

METHODS

Study population

The French PELAGIE cohort (Perturbateurs endocriniens: Étude Longitudinale sur les Anomalies de la Grossesse, l’Infertilité et l’Enfance) began in 2002 in three districts of Brittany (north-western France): Ille-et-Vilaine, Côtes d’Armor, and Finistère. Women were recruited by gynaecologists, obstetricians or ultrasonographers at visits for prenatal care. They met the inclusion criteria if they completed the inclusion questionnaire before 19 weeks of gestation and were still pregnant at that time. The follow-up of the women included in the study began at 16 weeks of gestation and continued through the end of pregnancy. Participants provided informed consent for data collection, and the appropriate ethics committees approved the study procedures.

In the inclusion questionnaire, women reported their social, demographic and medical characteristics: obstetrical history, educational level, present occupation, nutrition and lifestyle factors (alcohol and tobacco consumption). They returned it to our laboratory with a urine sample (the first morning void) for measurement of exposure biomarkers (results of which will be the object of further publication). Midwives, paediatricians and hospital medical records provided medical information about pregnancy outcome (delivery or medical termination, length of gestation, measurements and health status of newborn at birth).

Three thousand four hundred and twenty-one women returned the inclusion questionnaire before 19 weeks of gestation (for an estimated participation rate of 80%) between April 2002 and December 2005. Information on pregnancy outcome was available for 3399 women (99.4%). The other 22 women were lost to follow-up.

Assessment of pregnancy outcome

The presence and type of congenital malformations in liveborn infants was validated by a paediatrician based on findings at the clinical examination before discharge from the maternity ward (or neonatology or paediatric critical care unit). Malformations were determined by pathology and karyotype examinations for pregnancies that did not result in live births (medically indicated abortions or fetal deaths). Paediatricians were asked specifically about oral clefts and male genital anomalies (hypospadias, cryptorchidism and micropenis). Mention of male genital anomaly was validated by a paediatric surgeon with 2 years of follow-up. Mention of heart murmur identified at the maternity ward was sought by checking with paediatric cardiologists in the region to detect potential additional cardiac malformation.

Malformations were coded according to the International Classification of Disease (10th revision of World Health Organization 1993), and grouped according to the EUROCAT recommendations,21 as major or minor malformations and by organ groups of malformations, in addition to genetic and chromosomal malformations considered as one group. We considered all male genital malformations that were repaired surgically, including cryptorchidism and glandular hypospadias, to be major malformations.

Solvent exposure assessment

Solvent exposure at work was defined by two independent methods: self-reported exposure to products that might contain solvents and a job-exposure matrix (JEM).

The questionnaire completed in early pregnancy asked women about the frequency of their contact (never, occasional, or regular) at work with 11 classes of products considered to contain solvents (paints, strippers, varnishes, dyes, inks, glues, gasoline, grease removers, detergents and cleaning agents, textile treatment agents, and cosmetics). Women were then classified as “not exposed” if they reported no exposure to any of these products. They were classified as “occasionally exposed” to solvents when they reported occasional exposure to at least one product and “regularly exposed” when they reported regular exposure to at least one product. To study malformations by organ subgroups, we have retained only two classes: “never or occasional exposure” and “regular exposure”.

We also used a JEM.22^–24 This matrix assigned a probability of solvent exposure to a combination of two codes (one occupation code (International Standard Classification of Occupations of the International Labour Office 1968) and one code of industrial activity (International Standard Industrial Classification of all Economic Activities of the United Nations 1975)). This JEM originally classified exposure on five different levels: 10 — job-related exposure is not higher than for the general population; 20 — job may entail a cumulative exposure higher than for the general population; 3x — job may entail exposure to a level definitely higher than for the general population (31 — probability <1/3; 32 — probability 1/3–2/3; 33 — probability >2/3); 40 — job entails exposure to the specific agent at a level clearly higher than for the general population; and 50 — job entails exposure to the specific agent and exposure is known to be particularly high.

To have enough subjects in each category and according to the bimodal distribution of the initial levels, we combined these exposure categories into three groups: no exposure (level 10 (n = 2234)), medium exposure (levels 20 (n = 91) and 31 (n = 423)), and high exposure (levels 32 (n = 49), 33 (n = 22), 40 (n = 10) and 50 (n = 9)). For some rare outcomes, we could retain only two classes: no exposure (level 10) and exposure (levels 20, 31, 32, 33, 40 and 50).

Statistical analysis

Analyses were restricted to women who reported working at the beginning of pregnancy (n = 3005). We studied the relations between solvent exposure and congenital malformations separately for chromosomal and genetic malformations, other major malformations (excluding chromosomal and genetic), and minor malformations. Organ groups were also studied for other major malformations. We estimated odds ratios (ORs) and 95% CI with the babies with no malformations as the reference group.

Logistic regression was used to adjust for potential risk factors for congenital malformations cited in the literature. These variables were maternal age at inclusion (less than 25 years, 25–29 years, 30–34 years, 35 years and more), tobacco consumption at the beginning of pregnancy (non-smoker, at least one cigarette per day), alcohol consumption (0 drinks per week, at least one drink per week), educational level (primary or secondary, passed baccalaureate examination at the end of secondary school, 2 or more years of postsecondary education), and for studying male genital malformations, premature birth (born before 37 weeks of gestation, or at least at 37 weeks). Adjustment for these variables was made when it was possible (because of the number of cases) for studies of malformation subgroups. They were introduced into the model simultaneously and if some of them were at the origin of convergence default of the model, they were excluded from the analysis.

A trend test for categories of exposure was performed as appropriate. SAS software V8.2 was used for data analysis.

RESULTS

Population description and outcome

Of the women included in the PELAGIE study, 3005 worked at the beginning of their pregnancy (88%). Mean age at inclusion was 30.0 years (SD 5.0). An education level of at least 2 years of postsecondary education was reported by 61.4% (table 1).

Three thousand and five women represented 3041 pregnancies (36 women have twin pregnancies).

All but 33 pregnancies ended in a live birth; there were 14 elective abortions after prenatal diagnosis and 19 fetal deaths (after 16 weeks of gestation).

Concerning the 3041 infants, 26 infants had incomplete data for physicians’ examination, 2897 had no malformations and 118 infants were diagnosed with at least one congenital malformation: all 14 elective abortions and 104 of the live born babies. No malformations were diagnosed among fetal deaths. Eighty-four malformations were classified as major malformations (2.79% of the live births plus fetal deaths and elective abortions) and 34 as minor. The most frequent subgroups for major malformations involved limb (n = 20) and urinary malformations (n = 13) (table 2).

Occupational exposure to solvents

Women in the PELAGIE study cohort worked mainly in science-related (39.0%), clerical (23.2%) and service (15.6%) occupations (table 3).

Regular exposure to at least one product that may contain solvents was reported by 30.2% of our working population (table 3). Occupations entailing regular exposure were mainly hairdressers, cleaners and helpers, nurses’ aides, nurses and chemists/biologists. Of the women classified by self-report as regularly exposed at work, 73.3% reported regular exposure to detergents and cleaning agents, 15.5% to glue, 12.2% to grease removers, 11.0% to inks and 10.7% to cosmetics. Moreover, 32.0% of these women had more than one regular exposure.

The JEM placed 18.1% of the working women in the medium exposure category and 3.2% in the high exposure category (table 3). The occupations most often classified as highly exposed by the JEM were chemists/biologists and non-agricultural workers. Most nurses, nurses’ aides and hairdressers were assigned to the medium exposure category, whereas only 8.3% of the cleaners and helpers were assigned to this category.

Risks of congenital malformations

We found a statistically significant relation between the risk of major malformations (non-chromosomal and non-genetic malformations) and maternal solvent exposure, assessed both by self-reported exposure (OR = 2.48, 95% CI 1.4 to 4.4, for regular exposure vs no exposure) and by the JEM (OR = 3.48, 95% CI 1.4 to 8.4, for high exposure vs no exposure) (table 4). A statistically significant dose–response relation was found with each assessment method (p values for trend, respectively, p = 0.002 and p = 0.005). No significant relation was found with either chromosomal and genetic malformations or minor malformations.

The subgroup analysis for major malformations found ORs of at least 2 for oro-facial clefts, urinary malformations, and male genital malformations with both assessment methods and statistically significant for at least one method (table 5).

DISCUSSION

Using a prospective design and two independent methods of assessing exposure to solvents during the appropriate window of susceptibility during pregnancy, we found statistically significant and dose-dependent relations between maternal exposure and the risk of major malformations. These relations were explained in part by the increased risk of several subgroups of malformations, specifically, oral clefts, urinary malformations and male genital malformations.

Our cohort was composed of pregnant women who saw one of the participating physicians and agreed to participate in our study. Although the estimated participation rate was high (80%), the need to recruit both physicians and patients resulted in selecting a highly educated population (slightly higher than the regional average), concerned by environmental problems and able to follow the instructions for completing the questionnaire and sending it back. This selection, reflected by the distribution of education level in this population, explains why the occupations classified as exposed to solvents in our population are mainly technical (nurses, chemists and biologists) or service workers rather than manufacturing labourers. These occupations may, however, account for most current women’s occupations that entail solvent exposure. We can hope that this limited number of occupations will facilitate our search for the specific solvent compounds and mixtures that might explain our findings.

We successfully determined the outcomes of more than 99% of the pregnancies, independently of exposure characterisation. This rules out any differential assessment of congenital malformations.

The major malformation rates in our study are consistent with French,25 and European registry data,26 except for cardiac abnormalities, which were less frequent than expected (0.20% of births and abortions vs 0.58%). Cardiac malformations, which are frequently not detected until later in life,27 were sought by checking with paediatric cardiologists in the region for babies with heart murmurs identified at the maternity ward. No additional cardiac malformations were ascertained. It is nonetheless possible that some non-symptomatic cardiac malformations without heart murmur have not been detected in our study. Other non-symptomatic and not externally visible at-birth malformations could also have not been detected in our study, for example, urinary malformations which are not detected until later in life. But, due to the performance of routine ultrasounds during pregnancy (at least three prenatal ultrasonography examinations, reimbursed by the national health insurance for all pregnancies in France) and the quality of these images, we assume that urinary malformations are now very unlikely not to be determined before delivery.28 For other subgroups of malformations, we think that our assessment strategy is unlikely to not detect them. If some urinary or cardiac malformations were not detected at birth (which is not very probable for the urinary malformations), the misclassification would not be differential, given that the failure to detect them is unrelated to the exposure assessment conducted in early pregnancy. Consequences on the results would be on average an underestimation of the association between solvent exposure and these malformations.

Because of the well-known difficulties in recording male genital malformations in newborns,29 30 and to limit classification error and select the most severe cases (as for other malformation subgroups), we considered only male genital malformations that were repaired surgically, including cryptorchidism and glandular hypospadias, to be major malformations.

To reduce any possible healthy worker effect, we chose before beginning our analysis, to restrict it to women who were working at the beginning of their pregnancy. The associations we found were not modified by the inclusion in the analysis of women who were not working (all of whom were classified as unexposed).

We used two indirect assessments of occupational exposure (self-report and JEM), both conducted before pregnancy outcomes were assessed. Chromosomal malformations were diagnosed by karyotype through amniocentesis (performed after 15 weeks of gestation, results available 2 weeks after), and non-chromosomal malformations were diagnosed at the second ultrasound (routinely performed in France between 20 and 23 weeks of gestation) or later during pregnancy or at birth. We could verify that in all cases, diagnosis was made after response to the inclusion questionnaire. We therefore assume that no differential recall bias is possible. Random misclassification is nonetheless possible. Self-report of the frequency of contact with compounds that may contain solvents is subjective, and JEM assessments reflect only an average of individual working circumstances. In addition, there were not enough cases to allow us to make use of the fine level of exposures originally proposed by the JEM. Solvent exposure status of some occupations may have changed since 1988 (year of this JEM publication). Classifications used to define exposure for this JEM are very precise especially for service workers and production workers. These classifications may be less precise for recent technical occupations (such as information technology worker), which are not exposed a priori. Exposure situations may have changed over the past 20 years, but we think that the JEM used in our study has selected the most exposed combinations (of occupational activities and industrial sectors). Consequences of classifying truly exposed as non-exposed by this method would be on average an underestimation of the relation between solvent exposure and the occurrence of malformations. For occupations considered to be exposed, agreement between the two methods was good, except for cleaners and helpers. The kappa measure of concordance between these two methods was 0.32 (95% CI 0.28 to 0.36) with a percentage of concordance of 74%. The associations found with the two methods are very concordant but probably represent attenuated estimates of the true associations.

These two methods did not allow us to identify more specific chemical classes of solvents. “Modern” JEMs specific for chlorinated, petroleum-based and oxygenated solvents are currently under construction in France and will be applied to our data as soon as they are available. However, solvents often occur in mixtures, and identifying a single compound or class of compounds remains a challenge.

Domestic exposure to solvents has not been considered in this analysis, even though it may concern a large proportion of women from time to time and may occur during a vulnerable period. This type of exposure is hard to objectify, and we considered that it was on average less frequent and less intensive than occupational exposures.

There are very few known risk factors for congenital malformations: ethnicity, maternal age, specific diseases or medication use during pregnancy, and tobacco and alcohol use. We found associations with none of them except maternal exposure to solvents. Our population is very homogeneous in terms of both geographical origin (98.8% of European origin) and socioeconomic status, and the prevalence of risk behaviours (heavy smoking or alcohol use) is very low.

Despite the rarity of these pregnancy outcomes we found a statistically significant association between maternal exposure to solvents and the risk of major malformations, and a more specific association with the subgroups of oral clefts, urinary malformations and male genital malformations. Few prospective studies have previously been conducted in general populations. A Canadian cohort study comparing 125 women occupationally exposed to solvents with 125 women exposed at work to other non-teratogenic products found significant associations (relative risk = 13, 95% CI 1.8 to 99.5).10 This study has been criticised for its small size and highly selected recruitment.31

Two occupational cohort studies of women working in laboratories found a relation between maternal exposure to solvents and major malformations.12 13 Wennborg et al found a significant relation (OR = 2.5, 95% CI 1.0 to 6.0 for solvents in general and OR = 2.1, 95% CI 0.9 to 4.9 for benzene).12 This study has several particular limitations in its exposure assessment, including its retrospective nature and the high non-response rate (33%). Zhu et al,13 using a JEM assessment for laboratory technicians (n = 991), found a non-significant association between maternal solvent exposure and major malformations (hazard ratio  =  2.0, 95% CI 0.7 to 5.7). An occupational cohort study compared hairdressers with a control group in Sweden and reported a significant relation between hairdressing and major malformations (OR = 1.3, 95% CI 1.1 to 1.6).32 But the specific association with solvent exposure has not been explored.

This is the first time, to our knowledge, that a prospective cohort in the general population has found an association between oral clefts and maternal solvent exposure. Testing this specific hypothesis was one of the initial objectives of this study. Our results are indeed consistent with several previous case-control studies.14^–16 33^–36 All but two of those retrospective studies,33 34 used blinded assessment by industrial hygienists to define exposure. This method allows solvent exposure to be defined more precisely.37 Most of these studies suggested significant risk associated with oxygenated solvents,14^–16 35 36 notably glycol ethers.14^–16 Other studies have shown significant relations between maternal exposure to petroleum and chlorinated solvents and oral clefts.14 15 Because of the multiple exposures, none of these studies could report conclusions for specific solvent exposures.

One case-control study about male genital malformations showed a non-significant relation between hypospadias or epispadias and mothers’ work in leather industries (OR = 4.05, 95% CI 0.77 to 21.44).34 The association between maternal exposure to solvents and hypospadias was not explored. A case report suggested an association between maternal exposure to ethylene glycol monomethyl ether acetate and hypospadias.38

Our study also suggests a relation between maternal exposure to solvents and urinary malformations. A case-control study and a case report have already suggested similar results.39 40

Previous case-control studies have reported associations between maternal solvent exposure and central nervous system malformations,14 18 especially neural tube defects.14 Another case-control study with assessment methods,19 similar to the studies of Cordier et al did not find a relation with neural tube defects. In our study the slight association observed with central nervous system defects may not be related to neural tube defects: in this subgroup, one malformation is a spina bifida and the other three are hydrocephalus.

CONCLUSION

Our prospective population-based study provides additional evidence of an association between maternal exposure to solvents and major malformations. Exposure to solvents at the beginning of pregnancy was assessed by two independent methods and before pregnancy outcome assessment.

A large proportion of working women performed the occupational activities involved, and the exposures we considered are very likely still present in the workplace. Additional investigations are underway in order to identify the specific chemicals involved. They include measurements of urinary biomarkers of exposure to glycol ethers and chlorinated solvents (in a nested case-control study) and detailed description of products used at work for the main occupations involved. These results will be the object of subsequent publications.