2.1 NTD higher risk groups

2.2 Statistical analysis

To characterize each higher risk group, we evaluated the distribution of the following covariates by case–control status: maternal race/ethnicity (white non-Hispanic, Black non-Hispanic, Asian non-Hispanic, Hispanic, other), maternal age (<25, 25–34, ≥35 years), maternal education (<12, 12, >12 years), planned pregnancy (yes/no), study center, and year of last menstrual period (1988–1997, 1998–2015).

We used unconditional logistic regression with Firth's penalized likelihood to estimate crude and adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for the association between FA intake and NTD risk within each higher risk group. Firth's penalized-likelihood is a standard penalization to reduce bias in maximum likelihood estimates of beta coefficients with small sample sizes (Firth, 1993). We did not report ORs for models that did not converge or exposure groups with only 1 case. We adjusted for maternal age, because it was hypothesized to be the strongest confounder a priori, and study center (Massachusetts, Philadelphia, Toronto, other) because center participation changed over the study period. We evaluated dose separately for FA supplementation only and for total folate intake because some clinical recommendations are restricted to supplements whereas others do not specify a preferred source. We descriptively evaluated patterns of use and dose among less than daily supplementers but did not compute ORs due to small numbers. We interpreted ORs as relative risks since NTDs are rare.

In sensitivity analyses, we first repeated the analysis adjusted for planned pregnancy instead of maternal age. Second, for all FA-containing products where dose was unknown, we uniformly assigned a dose of 400 μg. Third, we restricted the supplement analyses for NTD family history, AED use, and pregestational diabetes to 1993–2015 data when only non-malformed controls were used. Fourth, to assess if the patterns were different in pre- versus post-fortification years, we generated boxplots of continuous estimated total folate intake by case–control status and time period for each higher risk group. Fifth, we restricted the analysis to isolated NTD cases. Sixth, we conducted the NTD history analysis restricted to pregnancies with first-degree history. Lastly, we conducted a multidimensional quantitative bias analysis (Lash, Fox, & Fink, 2009) to evaluate the robustness of primary supplementation findings for the family history group to three scenarios of recall bias (see Supplemental Materials).

3.1 Sensitivity analyses

Results were similar when models were adjusted for pregnancy planning instead of maternal age, though some associations were attenuated among the pregestational diabetes group (Table 2).

Among women reporting FA supplementation, exact dose was unknown for at least one product in 18.5% of participants; among products with unknown dose, 31% were prescription prenatals, 30% were over-the-counter prenatals, and 41% were other multivitamins (non-exclusive). Results were similar when 400 μg was universally reassigned for these reports (instead of 1,000, 800, and 400 μg, respectively), except for the pregestational diabetes group, for which there were insufficient numbers to assess doses ≥1,000 μg (Table S2).

When we restricted the data to 1993 or later (when all controls had no major structural malformations), similar OR point estimates were observed except for supplementation ≥1,000 μg among the AED users, which approximated the null (Table 3).

The boxplots demonstrated that on average during post-fortification years women consumed more total folate (among controls in each higher risk group, pre-fortification median 261.9–369.3 μg and post-fortification 386.3–587.6 μg). Except AED users during post-fortification, controls consistently had higher estimated total folate compared to cases in both pre- and post-fortification years across all higher risk groups (refer to Figures S1–S4). Due to small sample size, we could not make this comparison in the pregestational diabetes group for pre-fortification years.

When we restricted to isolated NTD cases (Table S3), results were similar to the main analysis. Due to small sample size, we could not estimate associations in the pregestational diabetes group.

The first-degree family history group demonstrated slightly stronger results for daily supplementation relative to the broader family history group, but a dose–response effect could not be evaluated due to small numbers (Table S4). No mothers with first-degree family history reported supplementation with the recommended dose of 4,000 μg; though the subgroups were small, mothers who planned the pregnancy tended to report daily supplementation ≥1,000 μg (5 out of 15 pregnancy planners versus 2 out of 19 unplanned pregnancies).

The quantitative bias analysis for potential recall bias suggested that daily folic acid supplementation would be more beneficial than observed if the control mothers consistently reported less accurately than the case mothers or if the case mothers consistently overreported supplementation. If instead the control mothers overreported supplementation, our observed results would be an overestimate of the true association; however, based on our simulated scenarios, ≥30% of the mothers who did not actually supplement would have needed to misreport as daily supplementers for the attenuation to be substantial (see Supplemental Materials for more detail).

4.1 Family history

Among our participants with an NTD family history, only 4% of cases and 18% of controls reported taking ≥1,000 μg of FA daily and none supplemented with ≥4,000 μg/day, demonstrating low adherence to recommendations. It is unclear if this was because they were uninformed or because did not follow clinical guidance. The Medical Research Council trial (1983–1991) is the only study to date with sufficient numbers to evaluate doses as high as 4,000 μg (MRC, 1991). Among the mothers of 1,005 cases with NTDs and 8,239 controls in the National Birth Defects Prevention Study (NBDPS) (1997–2009), only 27 had a previous NTD-affected pregnancy (Arth et al., 2015). Periconceptional FA supplementation was 35% (10/17) among women with NTD recurrence compared to 80% (8/10) among those without recurrence (Arth et al., 2015), equating to a crude OR of 0.14. Our findings, using a broader definition of family history, were consistent with these NBDPS results.

4.2 AED use and diabetes

We evaluated how FA is associated with NTD risk within diabetes and AED user subgroups, whereas prior studies have evaluated FA as a modifier of the associations between these risk factors and NTDs. Based on data distributions from these prior studies, we calculated crude ORs for FA and NTD risk to compare to our results. For AED users, among 196 women (1980–1996), any FA supplementation 5–12 weeks from last menstrual period was associated with an OR of 0.72 (Kjaer et al., 2008); among 19 valporic acid users specifically (1981–2007), periconceptional FA of 400 μg daily yielded a null association (Jentink, Bakker, Nijenhuis, Wilffert, & de Jong-van den Berg, 2010). Another study (1990–2013) observed 4 nervous system malformations among 1,259 pregnancies with first trimester AED exposure but none among the 66 women who supplemented with ≥5 mg of FA (Ban et al., 2015); information was not provided to compute ORs. Pregestational diabetes was reported among 43 NBDPS participants (1997–2004), any periconceptional FA supplementation was associated with an OR of 0.19 (Correa et al., 2012). Among 47 Slone Birth Defects Study participants with pregestational diabetes (1976–2011), periconceptional daily FA ≥400 μg was associated with an OR of 0.23 for spina bifida (Parker, Yazdy, Tinker, Mitchell, & Werler, 2013).

4.3 Obesity

A previous Slone Birth Defects Study analysis (1993–2011) found that, among women with obesity, spina bifida risk was reduced by approximately 15% with total folate ≥400 μg compared to <400 μg (Parker et al., 2013). In the present analysis, with expanded years and case definition, we observed a 35% reduction in overall NTD risk from daily FA supplementation at any dose, and the degree of benefit was similar for total folate from both diet and supplements ≥1,000 μg.

A recently study commented on the increase in NTD-affected pregnancies in Canada over the past decade and how these additional cases may be due to increased prevalence of several risk factors including pregestational diabetes and certain conditions which are sometimes treated with AEDs such as epilepsy and mood disorders (Liu et al., 2019). In addition, despite the high prevalence of obesity and its established association with NTD risk, research is lacking on an appropriate FA dose that might help to reduce, though perhaps not eliminate, NTD risk for this group. Our findings stress the need for further investigation into specific FA doses that may be needed to protect against NTDs in these groups of women at higher risk than the general population. Even a small decrease in NTD risk associated with FA would have an important impact on the general population because obesity continues to increase among women of childbearing age (Deputy, Dub, & Sharma, 2018).

The Slone Birth Defects Study has the benefit of spanning both pre- and post-fortification eras and included data on supplementation and medication use, family history of birth defects, demographic factors, and illness history. Unlike many prior studies, we were able to consider FA from both supplements and diet, and we were able to account for naturally-occurring food folate as well as synthetic FA from fortified foods. Despite the large size of the Slone Birth Defects Study, many of our findings were based on small numbers. For example, the higher risk groups were not mutually exclusive, except for women with prepregnancy obesity. While random errors in the dietary data are plausible, a validation study of the Willett food frequency questionnaire detected valid folate intake compared to biomarkers and 24-hour recall assessment (Yuan et al., 2018). We believe differential recall of the dietary data is unlikely because most U.S. consumers have poor health literacy and are not aware of micronutrient levels in food (Persoskie, Hennessy, & Nelson, 2017). In addition, this hypothesis was not known to the participants at the time of data collection. Based on our bias analysis, our primary findings for folic acid supplementation appear to be robust to certain scenarios of differential recall. We were unable to evaluate folic acid dose cutoffs higher than ≥1,000 μg; we did not collect information on reasons for not supplementing as recommended (e.g., unaware of recommendations, intolerance to vitamin supplements), although we do know that a portion of the pregnancies were unplanned. In the AED analysis, we were not able to evaluate associations with individual medications or account for indication or dose. The control definition changed in 1993 onward; patterns were generally similar in the sample restricted to 1993+. Residual confounding may be present since we could not adjust for all possible confounding factors; still, established risk factors for NTDs are only weak predictors in absolute terms (Agopian et al., 2013). The direction of the associations observed were largely in line with expectations and biologic plausibility, based on what is known regarding the association between FA and NTD risk among the general population (Wald, Law, Morris, & Wald, 2001). Misclassification of FA categories may have occurred due to recall issues; however, our sensitivity analysis with reassigned categories for those with unknown dose did not meaningfully change our conclusions.

In conclusion, our study supports the recommendation that all women of childbearing potential should consume at least 400 μg of FA daily, either through supplementation only or along with fortified foods, to protect against NTDs. The neural tube typically closes within the first 28 days after conception (Greene & Copp, 2014), often before pregnancy recognition (Branum & Ahrens, 2017), so supplementation should start before pregnancy to be effective. Given that many mothers in our study took less FA than current guidelines, this work highlights the need for clinical initiatives before pregnancy to improve FA intake. This work suggested that further benefit may be provided by higher doses for women at higher risk for NTDs, though the degree appears to vary by risk factor. Additional research is needed to inform clinical guidelines.