Neural tube defects (NTDs) affect 3000 pregnancies annually in the United States (CDC, 2004). NTDs occur during neurulation, 21 to 28 days postconception, when the neural tube fails to close. Few risk factors have been consistently associated with NTDs, although both environmental (i.e., noninherited) and genetic factors are thought to play a role in their development (Detrait et al., 2005).

In particular, folic acid intake remains the most widely recognized environmental risk factor for NTDs, with randomized controlled trials showing that folic acid supplementation might prevent at least half of all cases (MRC, 1991; Czeizel and Dudas, 1992). Maternal prepregnancy type 1 or type 2 diabetes (Becerra et al., 1990) and maternal use of antiepileptic medications during the days of pregnancy when neurulation occurs (Lammer et al., 1987) have also been associated with NTDs. Additional environmental risk factors for NTDs are less well-studied in humans, and studies have shown mixed results.

One such risk factor is maternal pesticide exposure, which can occur in both the home and the workplace. Residential pesticide exposure can occur through air, water, or food contamination, as well as from home and yard/garden use. Occupational pesticide exposure can occur directly through mixing of chemicals, equipment loading, application, equipment clean-up/repair, or disposal of empty containers, or indirectly through handling of contaminated crops or foods. Occupational exposure can also occur, typically at relatively lower doses, in workplaces that are treated with pesticides. Generally, residential pesticide exposure is expected to be common and to occur at low doses, whereas occupational pesticide exposure is less common, but may occur at higher doses (Garcia, 1998). Positive associations with NTDs have been reported for maternal residential exposure to pesticides (White et al., 1988; Garry et al., 1996; Shaw et al., 1999; Rull et al., 2006; Brender et al., 2010; Yang et al., 2014) and among agricultural populations for conceptions that occurred during the growing season when pesticides are applied (Garry et al., 1996; Kristensen et al., 1997). The strength of reported associations between maternal occupational pesticide exposure and NTDs has been small to moderate, and associations have not consistently been observed (Zhang et al., 1992; Blatter et al., 1996; Blanco Munoz et al., 2005; Lacasana et al., 2006).

Previous studies of occupational pesticide exposure and NTDs have been limited by sample size (151 cases and 151 controls [Blanco Munoz et al., 2005 and Lacasana et al., 2006], 55 cases and 66 controls [Blatter et al., 1996], 101 cases and 1875 controls [Zhang et al., 1992]) and the exposure assessment. Due to sample size limitations, few analyses were stratified by NTD subtype. Because NTD subtypes may have different etiologies (Mitchell, 2005), combining NTD subtypes may mask underlying associations. Blanco Munoz et al. (2005) reported an odds ratio of 6.5 (95% confidence interval [CI], 1.4–29.6) for occupational pesticide exposure and anencephaly, an estimate considerably higher than those reported from other studies that combined NTD subtypes (Nurminen et al., 1995; Shaw et al., 1999). Different exposure assessment methods may also have contributed to inconsistent findings. Many studies used only selfreports of any exposure (yes/no) to occupational pesticides (Zhang et al., 1992; Blatter et al., 1996; Blanco Munoz et al., 2005; Lacasana et al., 2006).

Industrial hygienist review of occupational histories is thought to result in less exposure misclassification as compared to self-reports or use of job title alone (Fritschi et al., 1996; Bhatti et al., 2011; Peters et al., 2011). Only two studies of pesticide exposure and NTDs were identified that used maternal occupational histories reviewed by an industrial hygienist (Nurminen et al., 1995; Shaw et al., 1999); however, neither study had sufficient numbers of pesticide-exposed individuals to estimate risks precisely. Also, no studies of maternal occupational pesticide exposure were found that stratified exposure by specific pesticide type. It is likely that not all pesticides disrupt neural tube development; thus, testing for associations of all pesticides combined may bias findings toward the null. Because of the large number of pesticide formulations and overlapping exposure to pesticide types, it is difficult, particularly with a case–control design, to link exposure to a specific pesticide with NTDs; however, exposure assessment may be improved with stratification by pesticide class (e.g., insecticides, fungicides, or herbicides).

To remedy some of the limitations in previous work and extend the knowledge base, data from the National Birth Defects Prevention Study (NBDPS), a multisite population-based case–control study, were used to examine the relationships between maternal periconceptional occupational pesticide exposure and NTD subtypes in offspring.

Interview data were collected from mothers of 958 (68% of eligible) NTD case and 5008 (66% of eligible) control infants; 521 case and 2997 control mothers met the criterion of employment during all or part of the relevant periconceptional period. Of these, 18 case and 47 control interviews were excluded as follows: incomplete maternal interviews (cases n = 1; controls n = 4), maternal diagnosis of type 1 or type 2 diabetes before or during the index pregnancy (cases n = 7; controls n = 15), or maternal periconceptional exposure to folic acid antagonists (cases n = 10; controls n = 28). To improve homogeneity of NTD subtypes, the maternal interview for one NTD case diagnosed with multiple NTD subtypes was also excluded, leaving 502 NTD cases (126 with anencephaly, 5 with craniorachischisis, 310 with spina bifida, and 61 with encephalocele) and 2950 controls eligible for analysis. Where pesticide exposure information was missing, inconsistent, or could not be evaluated based on the information provided (cases = 6; controls = 20), mothers were excluded from calculations of the odds ratio estimates.

Compared with control mothers, case mothers were more likely (p < 0.05) to be younger, Hispanic, less educated, have 3 or more pregnancies, or to be obese; the proportion of NTD cases also differed among sites (Table 1). When stratified by subtype, differences between control mothers and spina bifida mothers tended to parallel those for all NTD cases combined; mothers of cases with anencephaly or encephalocele only differed from control mothers by the proportion of infants per site.

As shown in Table 2, 30% of control mothers and 33% of case mothers were evaluated as being occupationally exposed to any pesticide. Of these, 81% of control mothers and 80% of case mothers held jobs with a probability of exposure that was ≥67% (data not shown); most of this exposure was to insecticides only. Compared with control mothers, the estimated median cumulative exposure to insecticides was higher for mothers of anencephaly and spina bifida cases. Compared with control mothers, the estimated median exposure to fungicides was higher for mothers of spina bifida cases, whereas the estimated median cumulative exposure for each of the three pesticide classes was near equal or lower for mothers of encephalocele cases.

In crude analyses, statistically significant, positive associations were observed for insecticides+herbicides with all NTD cases combined and with spina bifida, as well as for insecticides+herbicides+fungicides with all NTD cases combined and with anencephaly (data not shown). The final adjusted regression model included maternal education, prepregnancy BMI, and site. The aORs for insecticides + herbicides tended to parallel those for crude analyses, although aORs for insecticides+herbicides+fungicides were positive for anencephaly or encephalocele but not for all NTDs combined (Table 3). A significantly increased aOR was only seen for spina bifida (insecticides+herbicides). Sample sizes precluded analyses of exposure to herbicides only; however, numbers allowed estimation of odds of exposure to each pesticide class if exposure to the other remaining classes was considered irrelevant (e.g., exposure to herbicides regardless of exposure to insecticides and/or fungicides). Using this approach, the aORs were near unity for mothers exposed to insecticides, fungicides, or herbicides (data not shown).

Examination of maternal cumulative exposure (0, >0 to <50%, ≥50%) to any pesticide and also to insecticides only produced aORs near unity for all NTD cases combined and for the NTD subtype groups examined. The aORs for maternal cumulative exposure to insecticides + herbicides were positive, but not significant, for all NTD cases combined and for spina bifida cases, as were the aORs for maternal cumulative exposure to insecticides + herbicides + fungicides for anencephaly and encephalocele. Dose-dependent effects were only observed for anencephaly. The majority of jobs assigned as exposed to insecticides + herbicides provided services to buildings (janitorial, landscaping, or pest control; 40.7%) or traveller accommodation (26.4%). Jobs considered exposed to insecticides + herbicides + fungicides were most commonly in food/drink service places (26.7%), grocery or specialty food stores (18.7%), and agriculture (crop or animal production, or support activities for these; 16.7%) (data not shown).

For each pesticide class examined, aORs for isolated NTDs were similar to those for all NTDs combined (isolated + multiple) (data not shown). Among mothers of NTD cases with multiple defects (n = 61), cumulative exposure at or above the median (≥84.375 mg) to any pesticide was positively associated (aOR: 2.1, 95% CI, 1.2–3.9) with an NTD compared with those with no exposure. In addition, sub-analyses by family history of an NTD (yes/no) produced no appreciable difference in the aORs. Furthermore, restriction of analyses to sites which included live births, stillbirths, and elective terminations also produced little change in the aORs (data not shown).

Maternal periconceptional occupational exposure to any pesticide (yes/no) or insecticides only was not associated with all isolated NTD cases combined or individual NTD subtypes. Generally, small increases or decreases in risk were observed in adjusted analyses. For mothers occupationally exposed to insecticides + herbicides, the aORs for all NTDs combined were increased, but not statistically significant; the aOR was significantly increased for spina bifida alone. The aORs were increased, but nonsignificant, for anencephaly alone and encephalocele alone in mothers exposed to insecticides + herbicides + fungicides; however, other than spina bifida, none of the associations exhibited an exposure-response relationship. Results for maternal cumulative exposure to a pesticide class were generally similar to those for any exposure to that class, and with the exception of anencephaly only, dose-dependent effects were not observed. Lack of dose-dependent effects for other subtypes may reflect the relatively small difference in doses between the lower and higher exposed groups; most jobs considered exposed in this population-based sample of women had very low intensity and/or frequency exposure to pesticides.

The general lack of significant associations and dose-response relationships may indicate that periconceptional maternal occupational exposure to pesticides at the level observed in this population does not increase the risk of NTDs. The positive associations observed were supported by some (Blatter et al., 1996; Blanco Munoz et al., 2005; Lacasana et al., 2006) but not all (Zhang et al., 1992; Nurminen et al., 1995; Shaw et al., 1999) previous studies in humans. Two of the three previous studies that reported significant associations (Blanco Munoz et al., 2005; Lacasana et al., 2006) were limited to cases diagnosed with anencephaly and analysis of job titles (comparing those employed in agriculture to those not employed in agriculture), adjusted for selected maternal covariables. The odds ratios reported in each study (aOR: 6.5, 95% CI, 1.4–29.6; aOR: 4.57, 95% CI, 1.1–20.0) were markedly higher than those reported in the current study. The third study (Blatter et al., 1996) was also restricted to mothers employed in agriculture and reported an association for spina bifida adjusted for selected maternal covariables (aOR: 3.4, 95% CI, 3.0–9.0). These increased odds may be due to higher or more frequent exposures, greater prevalence, or different pesticides than experienced by the study subjects in our population.

Mechanisms by which pesticides have been linked with NTDs are poorly understood, although animal studies provide some insights. Methyl carbamate, chlorpyrifos, and other organophosphate insecticides are cholinesterase inhibitors, and in animal studies, cholinesterase inhibition has been shown to alter cell proliferation and differentiation during neurulation (Slotkin, 2004). In particular, chlorpyrifos has also been shown to lead to excessive neuroepithelial cell death during neurulation in rat embryos (Roy et al., 1998), possibly resulting in too few cells for neural tube closure.

The current case–control study of maternal occupational exposure to pesticides and NTDs used data from one of the largest U.S. population-based studies of birth defects; comparison of selected characteristics of control mothers who participated in the NBDPS and all live births in the same geographic areas showed that NBDPS control participants tended to be similar to all live births (Cogswell et al., 2009). Also, assignment of maternal occupational pesticide exposure was based on a substantial literature review and industrial hygienist review of maternal reports of occupational histories. This approach is considered the “gold standard” for exposure assessment where direct monitoring data are unavailable, unlike previous studies which used maternal self-reports only or job title only. Exposure assessment through industrial hygienist review is expected to decrease the risk of exposure misclassification compared with use of self-report or job title alone (Fritschi et al., 1996). Two prior studies of NTDs and pesticide exposure that used industrial hygienist review to assess exposure (Nurminen et al., 1995; Shaw et al., 1999) were limited by sample size. In addition, no prior studies reported results from stratifying exposure by pesticide class. Lastly, NBDPS data permitted attempts to estimate aORs by phenotype, restricting to families without a family history of NTDs, and restricting to sites that included live births, fetal deaths, and elective terminations.

Despite the efforts made to improve exposure and outcome classification compared with previous studies, limitations remained. Our sample of exposed NTD case and control mothers was larger than previous studies, yet small sample sizes limited some analyses for NTD subtypes, phenotypes, pesticide classes, and exposure probability categories, resulting in either imprecise odds ratios or the inability to calculate odds ratios for specific exposure strata. Post-hoc power calculations, given the sample size and a 30% rate of maternal occupational pesticide exposure in controls, showed the minimum detectable odds ratio for an association between any maternal occupational exposure (yes/no) and all NTDs combined was 1.39 (90% power, alpha 0.05 two-sided test). Also, although NBDPS interview data permitted adjustment for several possible covariables, data on pesticide exposure outside the workplace—such as using pesticides at home, residing on a farm, or residing near land in crop production— were not collected and could not be considered analytically. Residential exposure to pesticides is generally expected to be at lower levels (Garcia, 1998) than occupational exposure; however, it is possible that residential exposure exceeded some of the occupational exposures reported here. For example, residential exposure to pesticides could have impacted the association being tested if exposure to the particular type of residential pesticides was associated with NTDs and with the occupational exposure (as might be expected, for example, of agricultural workers who also live near the fields where they work). Additionally, cumulative occupational exposure in this population was generally low; therefore, results may not be generalizable to those in occupations with relatively high exposure intensities, such as agricultural populations. Furthermore, we could not identify specific pesticides, and we did not evaluate paternal exposure to which the mothers may have been exposed. Lastly, overlap among exposure to pesticide classes was high, limiting analysis to individuals exposed to insecticides only, insecticides + herbicides, and insecticides + herbicides + fungicides.

In summary, the association between specific NTDs and maternal periconceptional occupational exposure to pesticides was investigated within a large, case–control study. Results showed slightly increased, but not statistically precise, associations between exposure to some pesticide classes and some NTD subtypes. Although some previous studies corroborate these results, and efforts were made to improve on the design of these studies, findings should be interpreted cautiously due to limitations in sample size and exposure classification. Also, a large number of associations were tested, and significant findings may be due to chance. Future studies should aim to increase the sample size, particularly in less prevalent subtypes, of mothers exposed to herbicides or fungicides. Investigators should consider characterizing associations in highly exposed populations, such as agricultural populations, rather than examining the general population as was done here. In addition, future studies should aim to use similar detailed exposure assessment methods, while also collecting data on exposure to pesticides outside of the workplace.