To minimize etiologic heterogeneity within the case group, cases with an associated chromosomal abnormality or other syndrome (n = 75) and those with a closed NTD (i.e., lipomyelomeningocele, n = 88) were excluded. Additionally, cases with missing geocoded maternal address were excluded (n = 109). After these exclusions, 533 spina bifida and 303 anencephaly cases were available for analysis. Of the 4,132 controls, 437 were excluded because of missing geocoded maternal address. The final control group consisted of 3,695 unaffected births for analysis. The proportion of case and control mothers missing address information was similar (11.5% and 10.5%, respectively), and differences between those with and without maternal address at delivery were minor (≤ 5%) on demographic factors (results not shown). Compared with controls, case mothers were more likely to be Hispanic, born in Mexico, young, and less educated (Table 1).

Scatterplots of benzene and each of the other HAPs (toluene, ethylbenzene, and xylene) are presented in Figure 1. Levels of BTEX were highly and significantly correlated (ρ̂ ≥ 0.97, p < 0.001) (data not shown). Because of the high correlation between these compounds, statistical models including multiple pollutants were not assessed.

Results from the final models assessing the associations between BTEX and NTDs are presented in Table 2. After adjusting for year of birth, maternal race/ethnicity, education, census tract poverty level, and parity, mothers who lived in census tracts with the highest benzene levels were more likely to have offspring with spina bifida [odds ratio (OR) = 2.30; 95% confidence interval (CI), 1.22–4.33]. The degree of confounding from all covariates was modest; that is, adjusted ORs differed from crude ORs by no more than 15%. There were also positive associations with the low-medium (OR = 1.77; 95% CI, 1.04–3.00), medium (OR = 1.90; 95% CI, 1.11–3.24), and medium-high benzene exposure groups (OR = 1.40; 95% CI, 0.82–2.38). When the reference group was defined as less than or equal to the 10th, 15th, or 20th percentile of exposure, the association between maternal residence in a census tract with the highest benzene levels relative to the referent group and the prevalence of spina bifida remained, although it was attenuated (OR_10th = 1.96; 95% CI, 1.17–3.28; OR_15th = 1.59; 95% CI, 1.00–2.54; and OR_20th = 1.57; 95% CI, 1.00–2.46).

Based on the likelihood ratio test between the adjusted model with cubic splines and the model without the spline terms, there was a significant nonlinear relationship between maternal benzene exposure and spina bifida prevalence (p = 0.03). To further illustrate the nonlinear trend between benzene and NTDs, the estimated logits (and 95% confidence bands) were plotted against increasing benzene levels (Figure 2). For spina bifida, the logit appears to steadily increase when benzene levels are ≥ 3 μg/m³ and becomes statistically significant after benzene levels are approximately > 5 μg/m³ (Figure 2A), whereas no such trend was seen with anencephaly (Figure 2B).

We found a significant association between the prevalence of spina bifida in offspring and maternal exposure to ambient levels of benzene as estimated from the 1999 U.S. EPA ASPEN model (U.S. EPA 2006). The association was greatest for those in the highest exposure group. Positive associations between benzene and spina bifida were also observed in lower exposure categories; however, there was no monotonic dose–response relationship. Our finding that the risk of having a spina bifida-affected infant more than doubled for mothers living in census tracts with estimated benzene levels of ≥ 3 μg/m³ is in keeping with a report classifying individuals living in areas with benzene levels > 3.4 μg/m³ as being at the greatest risk for adverse health effects (Sexton et al. 2007). There were also associations with toluene, ethylbenzene, and xylene and between BTEX and anencephaly; however, these associations were not statistically significant.

The association between benzene levels and spina bifida appears to be nonlinear. This is supported by studies reporting nonlinear associations between personal exposure to benzene and various biomarkers (i.e., urinary metabolites and albumin adducts) of exposure using data collected on occupationally and environmentally exposed individuals, whereby exposure-metabolite curves became steeper at higher exposure levels (Kim et al. 2006; Lin et al. 2007).

Despite the strong correlations between the BTEX compounds, a significant association with spina bifida was seen only with benzene. Scatterplots of benzene and each of the other HAPs (toluene, ethylbenzene, and xylene) indicate that the correlations between pollutants are not as great at higher levels (Figure 1). In addition, we found lower correlations between benzene and the other pollutants (toluene, ethylbenzene, and xylene) when we restricted the analyses to census tracts with the highest benzene levels (n = 119) (ρ̂ = 0.62, 0.71, and 0.77, respectively).

Benzene is known to cross the placenta and has been found in cord blood at levels equal to or higher than maternal blood [Agency for Toxic Substances and Disease Registry (ATSDR) 2007]. Moreover, benzene can lead to genetic toxicity by covalently binding to DNA and forming DNA adducts, which, if not repaired, disrupt the microenvironment of the cell, leading to inhibition of important enzymes, cell death, and alteration of other cells (ATSDR 2007; Kim et al. 2006; Lan et al. 2004). If this occurs during the critical window of development, the complex cellular processes involved in neurulation (e.g., folate metabolism, cell proliferation, cellular adhesion, and vascular development) may be disturbed, resulting in NTDs.

Oxidative stress could also play a role in the teratogenic effect of benzene. Reactive oxygen species (ROS) formed after benzene exposure lead to DNA strand breakage and fragmentation leading to cell mutation (Hansen 2006; Xia et al. 2004). The importance of oxidative stress as a mechanism of teratogenesis is suggested by several animal studies (Fantel 1996). Treatment of pregnant rabbits and mice with ROS inhibitors diminished the effect of teratogens and reduced the amount of DNA oxidation (Liu and Wells 1995; Parman et al. 1999; Wells et al. 1997). One study conducted in rats demonstrated that increased embryonic oxidation resulted in failure of neural tube closure (Morriss and New 1979).

Positive associations between maternal occupational exposures to organic solvents and congenital malformations have been reported. One study assessing maternal occupational exposure to benzene reported an OR of 5.3 (95% CI, 1.4–21.1) for neural crest malformations (including NTDs) (Wennborg et al. 2005). In addition, among Mexican Americans, mothers occupationally exposed to solvents were 2.5 times as likely (95% CI, 1.3–4.7) to have NTD-affected pregnancies than control mothers (Brender et al. 2002). In a meta-analysis of five studies (not including the two previously discussed), mothers who were occupationally exposed to organic solvents had 1.6 times greater odds (95% CI, 1.2–2.3) of having an infant with a birth defect (including NTDs) (McMartin et al. 1998).

A potential limitation of this study is related to the exposure assessment, which relied on modeled predictions of ambient air levels of BTEX (i.e., the ASPEN model) and may have resulted in misclassification. Personal exposure is a function of outdoor and indoor pollutant levels, as well as individual behavior (i.e., time spent outdoors vs. indoors) (Lee et al. 2004). However, it has been shown that for benzene, the ASPEN model is a good surrogate for exposure measures based on personal monitoring (Payne-Sturges et al. 2004). The fact that ASPEN data were available only for 1999 and not for the entire study period is an additional potential limitation. This may be a suitable surrogate for other years, because the sources of HAPs (e.g., emissions from roadways and industrial facilities) were unlikely to change during the study period (Grant et al. 2007; Sexton et al. 2007; Whitworth et al. 2008). Additionally, information on maternal periconceptional use of folic acid and/or multivitamins (a potential confounder) was not available. However, this population represents pregnancies conceived after mandatory folic acid fortification (January 1998), and a recent study found little evidence of an association between NTDs and maternal folic acid intake or multivitamin use since fortification (Mosley et al. 2009). Finally, exposure misclassification due to use of maternal address at time of delivery is also a potential source of bias in this study. Because NTDs develop within the first 4 weeks after conception, address at delivery may be different than address during the critical window of exposure (Selevan et al. 2000). However, our own analyses using cases and controls from Texas included in the National Birth Defects Prevention Study, with complete residential information during pregnancy, suggest there was no significant change in benzene exposure assignment when using address at delivery versus address at conception (Lupo et al. 2010a).

Strengths of this study include the use of a population-based birth defects registry that employs an active surveillance system to ascertain cases throughout the state of Texas. This should limit the potential for selection bias. Furthermore, the Texas Birth Defects Registry includes information on pregnancy terminations, reducing any potential bias due to the exclusion of these cases. An additional strength was the use of a relatively small (census tract–level) measure of exposure. Using larger geographic units to estimate exposure (e.g., counties) may not capture the spatial variability of benzene (Pratt et al. 2004). Furthermore, separate analyses were conducted for spina bifida and anencephaly, as opposed to combining the groups into a single phenotype. This is important, as the effects of some exposures appear to be heterogeneous across the subtypes of NTDs (Lupo et al. 2010b; Mitchell 2005).

This study provides the first assessment of the relationship between maternal exposure to ambient levels of BTEX and the prevalence of NTDs in offspring. Our analyses suggest that maternal exposure to ambient levels of benzene is associated with the prevalence of spina bifida among offspring. We believe that future investigations of air pollutants and NTDs should include additional measures of exposure (e.g., air pollutant monitoring and biomarker data) and additional covariate information (e.g., genotypes and nutrient status).