Early Development: Original Article
Nutrient Pathways and Neural Tube Defects: A Semi-Bayesian Hierarchical Analysis
Carmichael, Suzan L.a; Witte, John S.b; Shaw, Gary M.a
From the aMarch of Dimes Foundation, California Research Division, Oakland; and bDepartment of Epidemiology and Biostatistics and Institute for Human Genetics, University of California, San Francisco, CA.
Submitted 20 February 2008; accepted 24 September 2008.
Supported by funds from the Centers for Disease Control and Prevention, Center of Excellence Award U50/CCU913241, NIH/NINDS R01 NS050249, and NIH/NICHD R01 HD 42538-03.
Supplemental material for this article is available with the online version of the journal at www.epidem.com; click on “Article Plus.”
Correspondence: Suzan Carmichael, March of Dimes Foundation, Children’s Hospital Oakland Research Institute, 5700 Martin Luther King Jr. Way, Oakland, CA 94609. E-mail: firstname.lastname@example.org.
Background: We used conventional and hierarchical logistic regression to examine the association of neural tube defects (NTDs) with intake of 26 nutrients that contribute to the mechanistic pathways of methylation, glycemic control, and oxidative stress, all of which have been implicated in NTD etiology. The hierarchical approach produces more plausible, more stable estimates than the conventional approach, while adjusting for potential confounding by other nutrients.
Methods: Analyses included 386 cases and 408 nonmalformed controls with complete data on nutrients and potential confounders (race/ethnicity, education, obesity, and intake of vitamin supplements) from a population-based case-control study of deliveries in California from 1989 to 1991. Nutrients were specified as continuous, and their units were standardized to have a mean of zero and standard deviation (SD) of 1 for comparability of units across pathways. ORs reflect a 1-SD increase in the corresponding nutrient.
Results: Among women who took vitamin supplements, semi-Bayesian hierarchical modeling results suggested no associations between nutrient intake and NTDs. Among women who did not take supplements, both conventional and hierarchical models (HM) suggested an inverse association between lutein intake and NTD risk (HM odds ratio [OR] = 0.6; 95% confidence interval = 0.5–0.9) and a positive association with sucrose (HM OR 1.4; 1.1–1.8) and glycemic index (HM OR 1.3; 1.0–1.6).
Conclusions: Our findings for lutein, glycemic index, and sucrose suggest that further study of NTDs and the glycemic control and oxidative stress pathways is warranted.
Click on the links below to access all the ArticlePlus for this article.
Please note that ArticlePlus files may launch a viewer application outside of your web browser.
The association of neural tube defects (NTDs) with folic acid is well-established.1–6 An association of NTDs with a variety of other nutritional factors has also been reported, for example, other B vitamins, glycemic index, and zinc.7–12 Most previous studies have examined 1 nutrient at a time. This approach ignores the fact that various nutrients may contribute to similar underlying mechanisms and that many nutrients are highly correlated. We propose to expand on previous findings in 2 ways: (1) by developing an analytic framework tied to underlying biologic pathways, and (2) by using hierarchical modeling, which takes into consideration the high correlations across exposure variables.
Several nutrition-related pathways are likely to contribute to NTD etiology. For example, 1 mechanistic pathway proposed to explain the association of NTDs with folic acid is methylation, which is important to many cellular processes, including gene expression, posttranslational modification of proteins, and detoxification of xenobiotics.13 Other nutrients that contribute to methylation—vitamin B12, methionine, choline—are also associated with the risk of NTDs.7–10,14–19 Another mechanistic pathway important to NTD etiology is glycemic control, as evidenced by the association of NTDs with diabetes and glucose levels11,20–22 and with dietary factors such as the average glycemic index of the diet and intake of simple sugars.12,23,24 A third nutrition-related pathway that seems to contribute to the risk of NTDs is oxidative stress. Evidence supporting a role for this pathway includes recent studies showing that glutathione or its metabolites (as markers of oxidative stress) are higher in mothers who delivered offspring with NTDs,16,25 and earlier studies suggesting preventive effects of vitamins E and C.26–29
Intakes of many nutrients are often highly correlated. This can complicate the analysis of their impact on NTDs, in part because simultaneously analyzing multiple nutrients may result in overparameterized models and unrealistically large and imprecise effect estimates. We can reduce these models using procedures such as stepwise regression, but such procedures tend to rely on arbitrary criteria such as specific levels of statistical significance and can lead to underestimated variances.30 An alternate approach that may overcome these challenges is hierarchical modeling, which can give more precise effect estimates and a reduction in the likelihood of false-positive results via shrinkage estimation based on existing knowledge.31,32 In addition, hierarchical modeling allows the examination of the combined contribution of selected groupings of exposures to risk of disease, such as nutrients that contribute to a common physiologic pathway.33
Here, we use conventional and hierarchical approaches to examine the association of NTD risk with intake of nutrients that contribute to the mechanistic pathways of methylation, glycemic control, and oxidative stress by using data from a population-based case-control study. The analyses estimate independent effects of individual nutrients that contribute to these pathways.
This study included pregnancies ending between 1989 and 1991 among mothers residing in selected California counties.5,34 Eligible cases comprised 624 infants/fetuses diagnosed with anencephaly, spina bifida cystica, craniorachischisis, or iniencephaly. Cases were ascertained by reviewing medical records at all hospitals and genetic clinics. The control group was composed of 612 live-born infants without reportable birth defects. They were randomly selected from area hospitals in proportion to the total population of live-born infants. We excluded 11 case mothers and 1 control mother who had a previous NTD-affected pregnancy. Interviews were completed for 549 (88%) case and 540 (88%) control mothers. Interviews were conducted in English (74%) or Spanish (26%), primarily in person (95%), on average 4.9 months for cases and 4.6 months for controls from the actual or estimated date of term delivery.
To assess nutrient intakes, we used data from a 100-item Health Habits and History Questionnaire that measured usual dietary intake during the 3 months before conception.35 The questionnaire was self-administered with an interviewer present to answer questions. After conducting error checks built into the analytic software,35 complete data on nutrient intakes were available for 454 cases and 462 controls.
We included 4 potential covariates in the analyses, based on their known association with NTD risk or nutritional status: maternal race/ethnicity (non-Hispanic white, US-born Hispanic, foreign-born Hispanic, other), education (less than, equal to, or greater than high school), obesity (body mass index <29 vs. ≥29 kg/m2), and intake of multivitamin/mineral supplements during the 3 months before conception (yes/no). We restricted analyses to the 386 mothers of cases and 408 mothers of controls for whom we had complete data on these covariates, who reported that they did not have type I or II diabetes, and who did not use food supplements such as SlimFast (SlimFast Foods Co., West Palm Beach, FL), because inadequate information was available to assess complete nutrient content of food supplements.
For each of the 3 main pathways of interest, we analyzed dietary intake of the following nutrients. (1) Methylation pathway: betaine, choline, folate, inositol, methionine riboflavin, vitamin B6, vitamin B12, and zinc. (2) Glycemic control pathway: fiber, fructose, glycemic index, glucose, and sucrose. (3) Antioxidant pathway: α-carotene, β-carotene, cryptoxanthin, cystine, glutathione, lutein, lycopene, oleic acid, provitamin A, vitamin C, vitamin E and zinc, and iron as a prooxidant. Details on the distribution of nutrient values among control mothers are provided in eTable (available in the online version of this article). In all analyses, nutrients were specified as continuous, and their units were standardized to have a mean of zero and SD of 1, for comparability of units across the hierarchical model pathways (described further later). All analyses included energy intake, race/ethnicity, education, and obesity as potential confounders. Separate analyses were conducted based on whether or not women took multivitamin/mineral supplements during the 3 months before conception.
We used 2 approaches to examine the association of nutrient intakes with NTD risk. First, we used conventional unconditional logistic regression to estimate odds ratios and 95% confidence intervals. Initially we ran a separate model for each nutrient and adjusted only for energy intake. Next, we ran a single model that simultaneously included all nutrients and energy intake and the other covariates mentioned previously:
where X is a matrix of all nutrients, β is a vector of regression coefficients that represent the effects of the nutrient variables on NTDs, and W is the matrix of covariates.
Next we used a 2-stage hierarchical regression model that incorporates information on each nutrient and the pathways through which they may act on NTDs. The first stage of the hierarchical model was a conventional logistic regression that included all nutrients and covariates (Equation 1). The second-stage model was
where Z is the second-stage “design” matrix incorporating information about the nutrients on each higher-level covariate (ie, pathway), π is a vector of second-stage coefficients estimated from the model, and δ is a vector with elements that reflect the potential residual effect of each nutrient (ie, the effects not captured by Zπ), assumed normally distributed with zero mean and variance τ2. The columns of Z contained information about nutrients in the 3 pathways (ie, methylation, glycemic control, and antioxidants). Nutrients were scored as 0 if they were not involved in a pathway, as 1 if they were expected to have a positive effect on the pathway, and as negative 1 if they were expected to have a negative effect. The eTable shows the second-stage design matrix Z.
Equation (Uncited)Image Tools
Therefore, the second-stage model regresses the first-stage coefficient estimates for the 26 nutrients β on the second-stage covariates that group nutrients by common physiologic pathways (ie, columns of Z). Nutrients contributing to the same pathways were assumed to have been randomly sampled from a common underlying distribution and therefore have exchangeable parameters.36 The first- (β) and second-stage (Zπ) estimates are then combined using inverse variance weighting to give posterior estimates of nutrient effects on NTDs:
where I is the identity matrix, W is a weight matrix that determines how much the first-stage estimates are shrunk towards the second-stage estimates, V̂ is the first-stage covariance matrix, and S = [V̂ + τ2I]−1.31,33
Equation (Uncited)Image Tools
We used a semi-Bayesian approach, whereby the values in δ are assumed to be normally distributed with a fixed variance τ2. A large value for τ indicates that the second stage has little (or no, if τ = ∞) impact on the first-stage covariates, so the hierarchical estimates are essentially equal to those from the conventional logistic regression.1 In contrast, a small value of τ assumes the nutrient effects are driven primarily (or completely, if τ = 0) by the second-stage estimates. Based on the anticipated residual effects of nutrients on NTDs—after incorporating the second-stage covariates—we specified τ as 0.35. That is, we assumed with 95% certainty that the residual odds ratio for each nutrient exposure would fall within a 4-fold range, centered around 1 (ie, from 0.5 to 2, since exp (3.96 × 0.35) = 4).
Relative to controls, case mothers were more likely to be foreign-born Hispanic or obese and have less than a high school education (Table 1). Although vitamin supplement intake did not differ much by case-control status (18% of case mothers and 19% of control mothers took supplements), we still stratified by this variable because the impact of nutrients on NTDs may differ among supplement users compared with nonusers.
Among women who took supplements before pregnancy, most odds ratios from the conventional analyses of 1 nutrient at a time were in the expected direction, but only a single confidence interval excluded 1 (Table 2). Odds ratios from the multivariable conventional analysis that included all nutrients varied widely and were very imprecise. These results in part reflected the small sample size within this group. As expected, odds ratios from the hierarchical model were closer to 1, were within a more narrow range, and more precise (Table 2). The hierarchical modeling results suggested that among women taking supplements, there were no associations between dietary nutrient intake and NTDs.
Among women who did not take supplements, most odds ratios from the conventional analyses of 1 nutrient at a time were in the expected direction and many of the confidence intervals excluded 1. Odds ratios from the multivariable conventional analysis varied within a modest range (0.51 for methionine to 2.12 for cystine) (Table 2). One noteworthy result for both the conventional and hierarchical models was an inverse association between lutein intake and NTD risk (hierarchical model OR = 0.63; 95% CI = 0.47–0.85). Note that all odds ratios reflect a 1 SD increase in the corresponding nutrient. Sucrose was positively associated with NTDs (1.40; 1.08–1.82), as was glycemic index. For the latter, the conventional model gave an OR of 1.27 (0.97–1.66; P = 0.08), whereas the hierarchical model gave an OR of 1.25 (1.00–1.55; P = 0.05). Hence, the hierarchical model gave a more precise estimate of effect.
We conducted both hierarchical and conventional analyses of nutrients from multiple physiologic pathways. These analyses suggest that increased intake of lutein, reduced intake of sucrose, and lower glycemic index are associated with reduced NTD risk. This study substantially extends previous analyses of nutrient intake and NTDs by adjusting for potential confounding effects of multiple other nutrients, yet at the same time providing relatively stable estimates through the use of a hierarchical model and by using an analytic framework that focused on specific pathways.
Our finding of an association of NTD risk with glycemic index and sucrose but not with glucose or fructose among women who did not take supplements before pregnancy is consistent with a previous analysis from these study data, although the previous analysis did not stratify by supplement use; it adjusted for intake of folate but not any other nutrients.12 The current analysis expands on the previous by adjusting nutrients from the glycemic control pathway for each other and for other nutrients. The explanation for an association with sucrose but not glucose or fructose is unknown.
The current analysis also expands our previous analysis by examining several nutrients for the first time: α- and β-carotene, cryptoxanthin, fiber, glutathione, iron, lycopene, pro-α-carotenes, and vitamin E. With the exception of fiber, these nutrients are particularly important to the oxidative stress pathway. Another set of nutrients—cystine, lutein, oleic acid, and vitamin C, riboflavin, and vitamins B6 and B12—has been examined previously in this dataset, but only as part of our analysis that used conventional logistic regression techniques to examine multiple nutrients adjusted for the effects of each other.37 The current and previous analyses both found that among these nutrients, only lutein was associated with reduced risk of NTD, and the association was restricted to women who did not take supplements. Specifically, the odds ratio for intake in the lowest quartile compared with the higher 3 quartiles was 0.6 (95% CI = 0.4–0.8) in the previous analysis, after adjustment for a different set of nutrients. The current analysis expands on the prior one by adjusting for the additional nutrients from the antioxidant pathway mentioned previously and additional nutrients from the other 2 studied pathways. Lutein, a carotenoid, is found in a variety of fruits and vegetables. Increased intake has been associated with improved visual function, reduced risk of cardiovascular disease and certain cancers, and reduced levels of certain biomarkers of oxidative stress.38 We are unaware of other previous studies of lutein and NTDs.
We have conducted previous analyses of folate, methionine, and zinc from this dataset, both as part of our analysis of multiple nutrients37 and as separate analyses of each nutrient. Increased folate intake was associated with reduced NTD risk5 but not after adjustment for intake of multiple other nutrients,37 consistent with our current findings. Increased methionine and zinc intake were associated with reduced risk, with or without adjustment for folate or multiple nutrients.8,37,39 This is not consistent with the current analysis, in which there was no association. Our previous analyses of betaine and inositol did not adjust for multiple other nutrients and suggested that they were not associated with NTD risk7,40; this is consistent with the current findings. Our previous analysis of choline suggested that increased intake was associated with reduced NTD risk, even after adjustment for maternal characteristics, supplement intake, and dietary intake of folate, methionine, and energy.7 The current analysis did not suggest an association of choline with NTD risk. One explanation for these inconsistencies in findings is that our previous analyses adjusted for different sets of nutrients than the current analysis; for example, we did not include any of the nutrients from the glycemic pathway or most of the nutrients from the antioxidant pathway.37 Alternatively, previous results may have been affected by multicollinearity among nutrients. Another potential contributing factor is the specification of nutrient variables in previous analyses (eg, as quartiles rather than as continuous variables). In particular, previous analyses of choline and methionine showed decreased risks for women not in the lowest intake quartile, with no dose-dependent relationship.7,8 The current analyses would not have been sensitive to such a “threshold” association with intake.
Several lines of evidence provide the biologic foundation for our pathway-focused approach to nutrient intake. The preventive effect of folic acid on NTD risk is well established.1–6 One suggested mechanism for folate’s preventive role involves methyl group donors.41 Supportive evidence comes from studies showing that other nutrients related to the methyl donor pathway—not just folate—are associated with risks. For example, as noted previously, we observed that increased dietary intakes of choline and methionine were associated with reduced risks of NTD-affected pregnancies, independent of folate intakes.7,8 In addition, inhibiting choline uptake and metabolism in mouse embryos results in NTDs,42 and pregnant rabbits given anticholinergic drugs have higher frequencies of malformed offspring.43 Observational studies have also shown associations of NTD risks with other nutrients or analytes involved in methylation, including homocysteine,14–17 methionine,8,9 methylmalonic acid,18 and vitamin B12.10,19 NTDs in human offspring have also been associated with several indicators of aberrant glucose control, including overt diabetes,20 gestational diabetes,44,45 hyperglycemia,21 and obesity.46–48 One explanation for the association of aberrant glycemic control with NTD risk is oxidative stress, given that administering antioxidant nutrients blocks its harmful effects on neural tube development.49,50 Administering antioxidant nutrients also blocks the harmful effects of oxidative stress induced by mechanisms other than hyperglycemia.29 Previous epidemiologic studies also support the investigation of antioxidants.16,25–28 In addition, folic acid is an antioxidant.51–54 Given the recent experimental findings related to oxidative stress, coupled with the known association of folic acid in preventing NTDs,1–6 it has been recommended that this pathway should be studied in detail in humans, including whether nutrients such as vitamins C and E have protective effects similar to folic acid.29
The majority of studies of nutrient intakes and disease focus on 1 nutrient at a time or at most a few nutrients at a time. A hierarchical approach produces more plausible, more stable estimates than the conventional approach, while adjusting for potential confounding by other nutrients. It “smoothes” or “shrinks” estimates toward the center of a prior distribution based on current knowledge about interrelationships among the exposures, as specified in the Z matrix.55 Parameter estimates from the conventional approach that are unstable and extreme, such as those for dietary carotene among women who took carotene supplements, are given little weight in the hierarchical model. Instead, more weight is given to the second-stage estimates, which can lead to substantial changes in the final parameter estimates. Such results emphasize the value of hierarchical modeling, whereby unstable parameter estimates are improved by “borrowing strength” from more stable parameter estimates, resulting in posterior estimates that are closer to the truth and less likely to be false positives. Conventional analyses generally model each exposure separately, undertaking multiple one-at-a-time analyses and comparisons. In contrast, the hierarchical approach simultaneously analyzes all nutrients in a single model, giving results that are not subject to the same issues of multiple comparisons. This is yet another benefit of undertaking a hierarchical modeling approach. Our 0/1 scoring system for the Z matrix is relatively crude, but we believe this level of refinement is appropriate, given current knowledge about the individual nutrients and their potential contribution to each specified pathway. Misspecification of the Z matrix can result in biased estimates (ie, the estimates are shrunk toward an invalid mean), but simulation studies indicate that hierarchical estimates are preferable to maximum likelihood estimates only if the Z matrix is seriously misspecified.56
Strengths of this study include its population-based ascertainment of cases and controls, high maternal participation rate, relatively large sample size, and ability to control for relevant covariates. Notable limitations include the fact that the sample size for the analysis of women who took supplements before pregnancy resulted in imprecise estimates within that stratum. Also, dietary intake as a measure of exposure has limitations, given that it does not reflect factors such as bioavailability, maternal-fetal exchange, genetic variability in nutrient metabolism, or actual tissue levels of nutrients. In interpreting our findings, we also cannot exclude the possibility that observations were affected by recall bias. Although there is little concrete evidence that recall bias contributes substantially to the results of studies such as these,57 it is possible, for example, that case mothers over reported or control mothers under reported intake of vitamin supplements or certain foods.
In summary, our findings with respect to lutein, glycemic index, and sucrose, in conjunction with substantial evidence regarding the potential contribution of oxidative stress and glycemic control to NTD etiologies,11,20–22,29 suggest that further study of the these pathways, and lutein in particular, is warranted.
1. MRC Vitamin Study Research Group. Prevention of neural tube defects: results of the Medical Research Council Vitamin Study. Lancet
2. Berry RJ, Li Z, Erickson JD, et al. Prevention of neural-tube defects with folic acid in China. New Engl J Med
3. Berry RJ, Li Z. Folic acid alone prevents neural tube defects: evidence from the China study. Epidemiology
4. Milunsky A, Jick H, Jick SS, et al. Multivitamin/folic acid supplementation in early pregnancy reduces the prevalence of neural tube defects. JAMA
5. Shaw GM, Schaffer D, Velie EM, et al. Periconceptional vitamin use, dietary folate, and the occurrence of neural tube defects. Epidemiology
6. Werler MM, Shapiro S, Mitchell AA. Periconceptional folic acid exposure and risk of occurrent neural tube defects. JAMA
7. Shaw GM, Carmichael SL, Yang W, et al. Periconceptional dietary intake of choline and betaine and neural tube defects in offspring. Am J Epidemiol
8. Shaw GM, Velie EM, Schaffer DM. Is dietary intake of methionine associated with a reduction in risk for neural tube defect-affected pregnancies? Teratology
9. Shoob HD, Sargent RG, Thompson SG, et al. Dietary methionine is involved in the etiology of neural tube defect-affected pregnancies in humans. J Nutr
10. Ray JG, Wyatt PR, Thompson MD, et al. Vitamin B12 and the risk of neural tube defects in a folic-acid-fortified population. Epidemiology
11. Groenen PM, Peer PG, Wevers RA, et al. Maternal myo-inositol, glucose, and zinc status is associated with the risk of offspring with spina bifida. Am J Obstet Gynecol
12. Shaw GM, Quach T, Nelson V, et al. Neural tube defects associated with maternal periconceptional dietary intake of simple sugars and glycemic index. Am J Clin Nutr
13. Blom HJ, Shaw GM, den Heijer M, et al. Neural tube defects and folate: case far from closed. Nat Rev Neurosci
14. Steegers-Theunissen RP, Boers GH, Trijbels FJ, et al. Maternal hyperhomocysteinemia: a risk factor for neural-tube defects? Metabolism
15. Mills JL, McPartlin JM, Kirke PN, et al. Homocysteine metabolism in pregnancies complicated by neural-tube defects. Lancet
16. Zhao W, Mosley BS, Cleves MA, et al. Neural tube defects and maternal biomarkers of folate, homocysteine, and glutathione metabolism. Birth Defects Res A Clin Mol Teratol
17. Van Der Put NM, van Straaten HW, Trijbels FJ, et al. Folate, homocysteine and neural tube defects: an overview. Exp Biol Med (Maywood)
18. Adams MJJr, Khoury MJ, Scanlon KS, et al. Elevated midtrimester serum methylmalonic acid levels as a risk factor for neural tube defects. Teratology
19. Ray JG, Blom HJ. Vitamin B12 insufficiency and the risk of fetal neural tube defects. QJM
20. Becerra JE, Khoury MJ, Cordero JF, et al. Diabetes mellitus during pregnancy and the risks for specific birth defects: a population-based case-control study. Pediatrics
21. Hendricks KA, Nuno OM, Suarez L, et al. Effects of hyperinsulinemia and obesity on risk of neural tube defects among Mexican Americans. Epidemiology
22. Groenen PM, Wevers RA, Janssen FS, et al. Are myo-inositol, glucose and zinc concentrations in amniotic fluid of fetuses with spina bifida different from controls? Early Hum Dev
23. Friel JK, Frecker M, Fraser FC. Nutritional patterns of mothers of children with neutral tube defects in Newfoundland. Am J Med Genet
24. Groenen PM, van Rooij IA, Peer PG, et al. Low maternal dietary intakes of iron, magnesium, and niacin are associated with spina bifida in the offspring. J Nutr
25. Martin I, Gibert MJ, Pintos C, et al. Oxidative stress in mothers who have conceived fetus with neural tube defects: the role of aminothiols and selenium. Clin Nutr
26. Verma K, Wei KD. Disorders of the developing nervous system of vitamin E-deficient rats. Acta Anat (Basel)
27. Smithells RW, Sheppard S, Schorah CJ. Vitamin deficiencies and neural tube defects. Arch Dis Child
28. Schorah CJ, Wild J, Hartley R, et al. The effect of periconceptional supplementation on blood vitamin concentrations in women at recurrence risk for neural tube defect. Br J Nutr
29. Loeken MR. Free radicals and birth defects. J Matern Fetal Neonatal Med
30. Freedman S, Pee D. Return to a note on screening regression equations. Am Stat
31. Witte JS, Greenland S, Haile RW, et al. Hierarchical regression analysis applied to a study of multiple dietary exposures and breast cancer. Epidemiology
32. Greenland S. Bayesian perspectives for epidemiological research. II. Regression analysis. Int J Epidemiol
33. Hung RJ, Brennan P, Malaveille C, et al. Using hierarchical modeling in genetic association studies with multiple markers: application to a case-control study of bladder cancer. Cancer Epidemiol Biomarkers Prev
34. Croen LA, Shaw GM, Jensvold NJ, et al. Birth defects monitoring in California: a resource for epidemiologic research. Paediatr Perinat Epidemiol
35. Block G, Hartman AM, Dresser CM, et al. A data-based approach to diet questionnaire design and testing. Am J Epidemiol
36. De Roos AJ, Poole C, Teschke K, et al. An application of hierarchical regression in the investigation of multiple paternal occupational exposures and neuroblastoma in offspring. Am J Ind Med
37. Shaw GM, Todoroff K, Schaffer D, et al. Periconceptional nutrient intake and risk for neural tube defect-affected pregnancies. Epidemiology
38. Granado F, Olmedilla B, Blanco I. Nutritional and clinical relevance of lutein in human health. Br J Nutr
39. Velie EM, Block G, Shaw GM, et al. Maternal supplemental and dietary zinc intake and the occurrence of neural tube defects in California. Am J Epidemiol
40. Shaw GM, Carmichael SL, Yang W, et al. Periconceptional dietary intake of myo-inositol and neural tube defects in offspring. Birth Defects Res A Clin Mol Teratol
41. Van den Veyver IB. Genetic effects of methylation diets. Annu Rev Nutr
42. Fisher MC, Zeisel SH, Mar MH, et al. Inhibitors of choline uptake and metabolism cause developmental abnormalities in neurulating mouse embryos. Teratology
43. McBride WG, Hicks LJ. Acetylcholine and choline levels in rabbit fetuses exposed to anticholinergics. Int J Dev Neurosci
44. Aberg A, Westbom L, Kallen B. Congenital malformations among infants whose mothers had gestational diabetes or preexisting diabetes. Early Hum Dev
45. Anderson JL, Waller DK, Canfield MA, et al. Maternal obesity, gestational diabetes, and central nervous system birth defects. Epidemiology
46. Shaw GM, Velie EM, Schaffer D. Risk of neural tube defect-affected pregnancies among obese women. JAMA
47. Shaw GM, Todoroff K, Schaffer DM, et al. Maternal height and prepregnancy body mass index as risk factors for selected congenital anomalies. Paediatr Perinat Epidemiol
48. Werler MM, Louik C, Shapiro S, et al. Prepregnant weight in relation to risk of neural tube defects. JAMA
49. Wentzel P, Gareskog M, Eriksson UJ. Folic acid supplementation diminishes diabetes- and glucose-induced dysmorphogenesis in rat embryos in vivo and in vitro. Diabetes
50. Li R, Chase M, Jung SK, et al. Hypoxic stress in diabetic pregnancy contributes to impaired embryo gene expression and defective development by inducing oxidative stress. Am J Physiol Endocrinol Metab
51. Nakano E, Higgins JA, Powers HJ. Folate protects against oxidative modification of human LDL. Br J Nutr
52. Rosenquist TH, Ratashak SA, Selhub J. Homocysteine induces congenital defects of the heart and neural tube: effect of folic acid. Proc Natl Acad Sci U S A
53. Joshi R, Adhikari S, Patro BS, et al. Free radical scavenging behavior of folic acid: evidence for possible antioxidant activity. Free Radic Biol Med
54. Ho PI, Ashline D, Dhitavat S, et al. Folate deprivation induces neurodegeneration: roles of oxidative stress and increased homocysteine. Neurobiol Dis
55. MacLehose RF, Dunson DB, Herring AH, et al. Bayesian methods for highly correlated exposure data. Epidemiology
56. Greenland S. Methods for epidemiologic analyses of multiple exposures: a review and comparative study of maximum-likelihood, preliminary-testing, and empirical-Bayes regression. Stat Med
57. Swan SH, Shaw GM, Shulman J. Reporting and selection bias in case-control studies of congenital malformations. Epidemiology
Supplemental Digital Content
© 2009 Lippincott Williams & Wilkins, Inc.
What does "Remember me" mean?
By checking this box, you'll stay logged in until you logout. You'll get easier access to your articles, collections,
media, and all your other content, even if you close your browser or shut down your
To protect your most sensitive data and activities (like changing your password),
we'll ask you to re-enter your password when you access these services.
What if I'm on a computer that I share with others?
If you're using a public computer or you share this computer with others, we recommend
that you uncheck the "Remember me" box.
Data is temporarily unavailable. Please try again soon.