The demographic profile of control women illustrates the low socioeconomic conditions prevalent on the border. More than a third reported a household income of < $10,000 per year, and only about half had completed high school (Table 1). Control women were leaner and more educated than were case women but comparable in age. Few women reported ever taking preconceptional folic acid supplements.

Nearly all women consumed corn tortillas (96% of case women and 93% of controls). As seen in Table 2, control women reported an average consumption of two tortillas per day during their first trimester of pregnancy (180 total during this 3-month period). That number was nearly identical to the intake reported during the 3 months before conception. Most women purchased rather than made corn tortillas.

After adjustment for BMI, serum B₁₂, and date of conception, moderate (301–400) compared with low (≤ 100) consumption of tortillas during the first trimester was associated with an increased OR of having an NTD-affected pregnancy (OR = 2.4, 95% CI, 1.1–5.3) (Table 3). Higher intakes (401–800 and > 800) were associated with either a slight decrease in occurrence (OR = 0.8 for 401–800) or no effect (OR = 1.0 for > 800). The type of tortilla usually consumed appeared to affect risk. NTD risk increased with exposure to homemade tortillas (OR = 2.9; 95% CI, 1.4–5.9) (Table 3). Among those who purchased tortillas, the effect estimates differed slightly by brand, but the CIs for individual brands were wide (data not shown). As shown in Table 3, increased exposure to fumonisins, based on postpartum sa:so ratio, was associated with an increased NTD occurrence except for the highest exposure category. The highest exposure category (sa:so > 0.35) was related to less frequent occurrence (OR = 0.7, 95% CI, 0.2–2.9) but was based on the fewest number of subjects and therefore had a wide CI. Results differed negligibly when serum folate was added to the model.

The mean fumonisin level measured in the 240 tortilla samples was 234 ng/g (range = 0–1,690 ng/g; SD = 256). When fumonisin levels by season were linked to dates of conception, the median daily fumonisin exposure (nanograms per day per kilogram of weight) was 172.6 for case women and 156.1 for control women. Using the imputed exposure levels for individual women, we also observed an inverted U-shape relation to risk, reflecting the tortilla consumption pattern (Table 4).

Our findings from the postpartum serum sa:so measure suggest that fumonisin exposure increases the risk of NTD, proportionate to dose, up to a threshold level. Data on the corn tortilla consumption and fumonisin ingestion appear to support an inverted U-shaped pattern of occurrence. This pattern may reflect the formation of NTDs up to a threshold of damage, at which point fetal death is more likely to occur. Because true biologic incidence is impossible to determine, the prevalence of birth defects is a function of embryologic incidence as well as intrauterine survival (Weinberg and Wilcox 1998). Additionally, experiments demonstrate that fetal resorption can occur in folate-deficient pregnant mice (Burgoon et al. 2002) as well as in hamsters and mice exposed to high doses of FB₁ (Floss et al. 1994; Gelineau-van Waes et al. 2005; Reddy et al. 1996).

We also observed a difference in risk effect between manufactured tortillas and homemade tortillas. The variations in small-scale tortilla preparation, especially the corn-to-lime ratio, results in wide variations in residual fumonisins (De La Campa et al. 2004). If tortillas made at home have a consistently lower concentration of lime or poorer quality corn is used, this could potentially explain some of the increased effect seen in homemade tortillas (De La Campa et al. 2004).

Alternative explanations for the effects that we observed include concern about the potential for recall bias in the estimate of tortillas consumed and lack of a true biomarker of fumonisins at the time of neural tube closure. Case and control women were asked to recall corn consumption as much as a year earlier. However, dietary recall over a much longer period has been shown to be generally reliable (Byers et al. 1987). Although it might seem intuitive that case women were more motivated to remember some events as they sought explanations for having a child with a birth defect, this has not proved to be a consistent bias in studies comparing prospective and retrospective measurement of exposures (Khoury et al. 1994; Mackenzie and Lippman 1989). It seems unlikely that an event as routine as eating corn tortillas would have been differentially recalled. Furthermore, recall bias would not easily account for the consistent observation of an inverted U-shaped relation between questionnaire and laboratory-measured exposures.

Temporality of the blood sample collection is also an important concern. The sa:so ratio measures acute exposure to fumonisins, with levels returning to normal when the exposure is removed. In our sample, the sa:so ratio was measured 5–6 weeks postpartum and would reflect periconceptional levels at neural tube closure only if study participants were exposed to fumonisins at a constant level. We note, however, that in rats and mice, a subtoxic fumonisin dose will maintain elevated sphinganine when it follows exposure to a higher dose (Enongene et al. 2002; Wang et al. 1999). We have observed that tortilla consumption varies little with women’s pregnancy status, making it possible that this population is continuously exposed to some level of fumonisins. In addition, the biochemical mode of action gives no indication that this is a case of reverse causation, that is, the NTD (due to low folate or the shorter gestation) causing elevated sa:so ratios in the mother (Gelineau-van Waes et al. 2005).

Our imputation of daily fumonisin consumption also had several limitations. To provide an estimate of fumonisin consumption from corn tortillas, tortillas were collected from homes and local grocery stores, which may or may not indicate what was actually consumed by each woman. In addition, the collection of corn tortillas was not systematic, and the samples were not proportionately representative of region, brand, or season. Further, the daily dose of fumonisin exposure depended on the number of tortillas women recalled, which as we noted is subject to error. Despite these limitations, it is noteworthy that the estimated levels of fumonisin ingestion for some women (650–9,441 ng/kg body weight for women in the highest quartile) surpass the World Health Organization’s maximum tolerable daily intake of 2 μg/kg (2,000 ng/kg) of body weight of any combination of fumonisins (WHO 2002).

The higher proportion of refusals to participate among control women (27%) compared with case women (12%) is another potential concern. Analyses conducted by Suarez et al. (2000) validated that control women mirrored the demographic characteristics of all border Mexican-American women who gave birth during the study years, indicating the low likelihood that control women were seriously unrepresentative of the border population.

The evidence that fumonisins may play a role in the development of NTDs is slowly being assembled. Through our epidemiologic study, we have documented that a population with a high prevalence of NTDs also consumes large amounts of corn products (tortillas). In addition, we have demonstrated that the amount of fumonisins in corn used to make tortillas may be high. We have illustrated with laboratory data that, within a certain range, women with increasing sa:so ratios (a surrogate for fumonisin exposure) are increasingly likely to have had an NTD-affected pregnancy, independent of known NTD risk factors (folate, B₁₂, obesity, and other covariates). Previous work shows that this population obtains folate largely through diet and that the use of folic acid–containing vitamins is rare. The negligible effect of serum folate in the model reflects this lack of supplementation (Suarez et al. 2000). In fact, B₁₂ levels are a more important predictor of NTD risk than are folate levels in this population (Suarez et al. 2003), a risk factor included in the adjusted model for fumonisin exposure.

Recent laboratory experiments complement these epidemiologic findings by illuminating the biologic mechanisms through which fumonisins might increase risk for NTDs. Fumonisins have been shown to inhibit the biosynthesis of sphingolipids (Wang et al. 1991), which interferes with the uptake of 5-methyltetrahydrofolate and decreases total folate binding (Stevens and Tang 1997). Furthermore, mice embryos subjected to fumonisin exposure develop NTDs in vitro (Flynn et al. 1997; Sadler et al. 2002) and in vivo (Gelineau-van Waes et al. 2005). Administration of folate reverses this effect (Gelineau-van Waes et al. 2005; Sadler et al. 2002). Cumulatively, these biologic and epidemiologic findings support the hypothesis that fumonisin contamination of corn destined for human consumption poses a risk for NTDs through its impact on sphingolipid and, ultimately, folate metabolism.

Based on our findings, it is possible that the 1990–1991 Cameron County NTD outbreak shared an etiology with the ELEM epizootics that slightly preceded it. The discovery of an association between fumonisin exposure and NTDs may help to clarify both the etiologies of unexplained NTD outbreaks and the increased background prevalence observed in some populations (Moore et al. 1997; Mutchinick et al. 1999). In 2001, the FDA recommended that an evaluation of the dietary intake of corn products by specific population groups (e.g., Texas Hispanics) and the levels of fumonisins found in those corn products was needed to fully assess the potential health risks (U.S. FDA 2001). Future epidemiologic studies should focus on measuring individual fumonisin intake in specific high-risk populations and assessing its impact not only on the developing embryo but also on other outcomes such as impaired fecundity or pregnancy loss.