We found that positive responses to a brief series of questions designed to measure food insecurity were associated with increased risks of certain birth defects, even after consideration of the potential confounding or modifying effects of maternal race-ethnicity, education, BMI, intake of folic acid-containing supplements, dietary intake of folate and energy, neighborhood crime, and stressful life events.
This study used 5 of 6 questions from a shortened, validated scale (1
). According to the scale developers, a score of 0-1 represents food security; 2-4, food insecurity without hunger, and 5-6, food insecurity with hunger. Rather than restrict our analysis to these 3 categories, we chose an alternative analytic approach (i.e. we did not collapse responses into these predefined categories) for the following reasons: given that we included only 5 of the 6 original questions, some women with a score of 1 may have been misclassified as food secure; a positive response to any of the questions suggests potential stress or nutritional compromise; the association of these questions with health outcomes has not been examined by many previous studies; and analyzing all possible scores allowed us to maximize our exploration and use of the data, rather than losing some potentially important information. The current analysis may be considered somewhat exploratory given that we chose this analytic approach and that no previous studies to our knowledge have examined the association of food insecurity with birth defect risks.
We observed evidence for effect modification in several of the final multivariable models. Results for anencephaly were stronger among women with higher scores on neighborhood crime, results for cleft palate and d-transposition of the great arteries were stronger among women who were underweight, and results for tetralogy of Fallot were stronger among women who did not take folic acid supplements during early pregnancy. We prefer not to speculate about potential explanations for these results, given the somewhat preliminary nature of the current study, but each of these observations is in the expected direction.
National estimates suggest that ~12% of U.S. households may have experienced food insecurity in the last year (38
). The current study found that 10% of control mothers had 2 or more affirmative responses to the food insecurity questions and would therefore be considered food insecure (1
). However, this percentage may be an underestimate of actual food insecurity in the past year, given that we used only 5 of the 6 original questions in the scale and our estimates were only for a 4-mo time span.
We hypothesized that food insecurity may be associated with risks of birth defects, because it is an indicator of increased stress or compromised nutrition. Maternal food restriction (defined as not eating for 13 h or more) and maternal stressful life events have been shown to be associated with elevated maternal corticotropin-releasing hormone and corticosteroid levels during pregnancy (39
). Corticosteroids are teratogenic in animal models for various organ systems (11
) and they represent 1 possible explanation for observed associations of maternal stress and food restriction with risks of birth defects. Corticosteroid medication use during pregnancy has been shown to be associated with increased risk of orofacial clefts (44
), but its association with other birth defects is less certain due to a lack of studies rather than a lack of evidence.
Strengths of this study include its comprehensive case ascertainment, detailed phenotypic review, population-based control selection, and satisfactory level of participation in maternal interviews. Given the relatively low frequency of the individual birth defects, we were limited to a retrospective study design and recall bias could have occurred. Although there is concern that mothers of malformed infants will overreport or more thoroughly report exposures than controls (52
), several studies suggest that recall bias is likely to be minimal for many exposures (53
), but none of them studied food insecurity. An additional concern is that the time between exposures and actual interview could have affected mothers’ ability to correctly recall exposures; it is unknown whether such limitations would be different for mothers of cases or controls. Other, longer instruments to assess food insecurity would have provided more detailed information that could have been useful to the analysis; however, the shortened version has been well validated (1
). Our approach of modeling the food insecurity score as a continuous variable is unique, but it is supported by statistical testing that compared the continuous approach with a more traditional, categorical approach. Given our somewhat provocative findings, as well as the limitation in our measurement of food insecurity, further studies of food insecurity and birth defects using more detailed measures of food insecurity may be useful. The current study controlled for potential confounding or modifying effects of several factors, including some aspects of stress (stressful life events, neighborhood crime), nutrition (intake of supplements and dietary intake of folic acid and energy), and socioeconomic level (maternal education), but residual confounding by unmeasured aspects of these factors may serve as an alternative explanation for our findings. Some women chose not to participate and a number of women were missing data on the studied covariates; the generalizability of our results beyond the mothers with complete data is unknown.
Food insecurity is a prevalent problem in the United States. It has negative consequences for a variety of nutritional and non-nutritional outcomes (57
). This study suggests that increased risks of certain birth defects may be included among them.