Insufficient maternal intake of vitamins has been linked to higher rates of neural tube defects and other categories of birth defects (Werler et al. 1999; Botto, Olney, and Erickson 2004; Canfield et al. 2005). Further, diarrhea in pregnant women may lead to malabsorption of nutrients, alterations in the intestinal microbiome, and altered food or vitamin intake (Thomas, Andrew, and Gerald 1972; Rosenberg, Solomons, and Schneider 1977; Camilo et al. 1996; The and Le 2022). As diarrhea is a common complaint (Singh et al. 2018), it is important to determine whether it is an independent risk factor for birth defects.

A previous study observed an association between maternal reports of diarrhea during early pregnancy and increased odds of neural tube defects among their offspring (Felkner et al. 2003). Two additional studies observed associations between diagnoses of gastrointestinal infections during early pregnancy and an increased odds of selected birth defects (Acs et al. 2010) and all birth defects in aggregate (Rumbo et al. 2021). However, many types of birth defects were not assessed by these studies. And, most of the previous studies were limited by small sample sizes.

Our primary goal was to use data from the National Birth Defects Prevention Study (NBDPS) to conduct a comprehensive study of the association of maternal diarrhea and 32 well defined categories of birth defects, most of which have not been assessed in previous studies. Our secondary goal was to determine whether associations between maternal diarrhea and birth defects were modified by maternal use of multivitamins during the periconceptional period.

The NBDPS was a population-based, case–control study that collected data on pregnant women with estimated dates of delivery between 1997 and 2011. Because the study questionnaire did not include a question on diarrhea until 2006, the current analysis includes only NBDPS data for births with estimated dates of delivery from 2006 to 2011. Cases were identified from birth defect surveillance programs at nine sites (Arkansas, California, Georgia, Iowa, Massachusetts, New York, North Carolina, Texas, and Utah). Controls were live-born infants without major birth defects, selected from birth certificates or birth hospital logs to be representative of the same populations from which the cases were sampled.

Cases were diagnosed by physical examination, imaging, autopsy, or pathology reports from surgical procedures and include live births, fetal deaths, and pregnancy terminations. Cases were aggregated across study sites and reviewed by clinical geneticists affiliated with the NBDPS, according to established guidelines (Rasmussen et al. 2003; Reefhuis et al. 2015). “Geneticists classified cases as having isolated defects if there was one major birth defect, two or more major birth defects affecting only one organ system, or one major birth defect with a sequence of related defects. Cases were classified as having multiple birth defects if there were two or more major unrelated defects in different organ systems.” (Rasmussen et al. 2003; Reefhuis et al. 2015). Birth defects that were strongly suspected or known to have been caused by single-gene disorders or chromosomal abnormalities were excluded from the NBDPS. Maternal interviews were conducted using a standardized computer-assisted telephone interview, in either English or Spanish, between 6 weeks and 24 months from the date of delivery. The frequency of participation was 63.0% for controls and 59.4% for cases. Interviews were completed within an average of 1 year from the estimated date of delivery for case mothers and 9 months for control mothers.

Mothers were asked whether they had “three or more unusually loose stools in one day” any time between 90 days before conception to 90 days after conception. If they answered yes, they were asked, “On about how many days did you have diarrhea?”. Mothers who responded that they did not have diarrhea during this period were coded as unexposed. Hereafter, the period between 90 days before and 90 days after the estimated date of conception is referred to as the “periconceptional period” for maternal exposure to diarrhea.

We included all NBDPS birth defects among subjects that were born or delivered during 2006 to 2011 and had 125 cases or more in the NBDPS database during this period. Mothers with missing information for the initial question on diarrhea or missing information for one or more of the final covariates (described below) were excluded. For analyses of hypospadias, controls were restricted to pregnancies that had male infants.

Using logistic regression, we estimated adjusted odds ratios (aORs) and 95% confidence intervals (CI) for associations between maternal report of diarrhea and 32 categories of birth defects (18 non-cardiac and 14 cardiac). Because previous studies suggested that acute infectious diarrhea was associated with birth defects (Acs et al. 2010; Rumbo et al. 2021) and the average duration of infectious diarrhea is two to four days (Bellido-Blasco and Arnedo-Pena 2011), we conducted separate analyses for diarrhea reported as lasting one to five days and diarrhea reported as lasting six days or more. The referent group for both categories of maternal diarrhea was comprised of mothers who stated they had no diarrhea during the periconceptional period. Adjusted odds ratios were only calculated for birth defects if there were three or more exposed cases.

We adjusted our analyses for nine potential confounders: maternal age (< 18 years, 18–24 years, 25–29 years, 30–34 years and ≥ 35 years), maternal race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, and other race/ethnicity), maternal level of educational attainment (less than high school [≤ 11 years], completed high school [12 years], some college [13–15 years] and college graduate [≥ 16 years]), first live birth (yes, no), pre-pregnancy diabetes (yes, no), pre-pregnancy high blood pressure (yes, no), any maternal smoking between one month before conception and three months after conception (yes, no), and three levels of maternal use of multivitamins during the periconceptional period (optimal use [daily intake between 1 month before and 1 month after conception], suboptimal use [less than daily intake], inadequate use [no intake or intake beginning later than one month after conception]), and language of interview (English, or other language).

We described aORs for specific birth defects as elevated (or decreased) if they were statistically significant or borderline in significance (p < 0.10) in either the main analysis or one of the sensitivity analyses. When the analysis was restricted to isolated birth defects, any birth defect for which the aOR had been elevated in the main analysis but was subsequently both decreased by 10% or more and no longer elevated (p < 0.10), was deleted from the list of birth defects with elevated aORs.

We did not determine a minimally sufficient number of covariates because it would have entailed publishing separate sets of covariates for each of the 32 birth defects. Also, statistical power for this common exposure (maternal diarrhea) was more than adequate. Instead, we adjusted each birth defect for the same covariates that were included in previous NBDPS papers that assessed a range of different birth defects and then dropped two covariates whose inclusion in the models had a negligible effect on the results (see below). We also considered DAGS (directed acyclic graphs) to ensure that none of the covariates were colliders.

Variables for maternal pre-pregnancy body mass index (BMI) and study site were also initially included as covariates. High maternal BMI is a modest risk factor for many NBDPS birth defects (Waller et al. 2007). However, compared to the other covariates, it has a relatively large percentage of missing values (4.4%) and dropping it from our logistic regression models had a negligible effect on the aORs. Including study site in the model, decreased the precision of our models and when it was dropped it also had a negligible effect on the aORs. Therefore, both maternal BMI and study site were dropped from the final models.

Because maternal fever or antibiotic use may either precede or follow diarrhea, neither of these conditions met the definition of a confounder. Therefore, we did not adjust for them in our primary analyses. However, we addressed the question of whether our results were independent of these conditions in the sensitivity analyses described below.

Three sensitivity analyses were conducted. (1) To determine if associations between diarrhea and specific birth defects were independent of the presence of other birth defects, we repeated the main analysis restricted to isolated birth defects. To determine if our results were independent of cases of diarrhea associated with fever or use of antibiotics: (2) We repeated the main analysis restricted to mothers who did not report a fever during the periconceptional period and (3) We repeated the main analysis restricted to mothers who did not report use of antibiotics during the periconceptional period.

For those birth defects that had an elevated aOR in either the main analysis or one of the sensitivity analyses, we calculated the change in the aOR associated with each sensitivity analysis (aOR for the sensitivity analysis minus the aOR for the main results/aOR for the main results). If an aOR in a sensitivity analysis decreased by 10% or more (compared with the main analysis), it was considered to be an “important” change.

Finally, results were generated, stratified by the three categories of maternal use of multivitamins, described above. For each of the birth defects in the study, we employed weighted linear regression to evaluate whether there was a linear trend in the magnitude of the aORs across the three categories (Brownstein and Cai 2019). Weights were assigned based on the sample size used to calculate each aOR. Results were considered meaningful if the p value for the test for trend was < 0.10.

This study was approved by the Institutional Review Boards for each study site and the Centers for Disease Control and Prevention. All analyses were completed in SAS 9.4 (SAS Institute Inc.; Cary, NC, USA).

A total of 18,072 participating mothers and their infants were delivered within the study period (2006–2011). We excluded 670 (3.7%) of study participants who were missing information on the initial question regarding diarrhea. After this exclusion, there remained 17,402 participants. An additional 727 (4.2%) of the participants were excluded due to missing information on one or more of the covariates in the final model, leaving 16,675 participants in the analysis. Of these, 12,023 were case mothers and 4652 were control mothers.

A total of 1790 (10.7%) women reported having diarrhea. Of these, 908 mothers (50.7%) reported diarrhea lasting one to 5 days and 768 mothers (42.9%) reported diarrhea lasting 6 days or more during the periconceptional period. The median number of days reported for duration of diarrhea in these two groups was 2 days and 22 days respectively. There were 114 mothers (6.4%) who indicated that they had diarrhea during this period, but did not provide the number of days that it occurred.

In Table 1, we computed crude odds ratios (cORs) and 95% confidence intervals for associations between 11 maternal characteristics and maternal report of diarrhea (of any duration) during the periconceptional period. This analysis was restricted to mothers of controls. Mothers who were 25 to 29 years of age, had high pre-pregnancy blood pressure, were smokers, reported fever or reported use of antibiotics were more likely to report diarrhea (cORs of 1.29, 1.47, 1.61, 1.87 and 1.59, respectively) Mothers who were Hispanic, reported inadequate use of multivitamins, or did not interview in English were less likely to report diarrhea (cORs of 0.56, 0.78 and 0.65, respectively) (Table 1).

This is the first analysis of the association between maternal report of diarrhea and birth defects to include a wide range of different categories of cardiac and non-cardiac birth defects. We utilized data from the CDC sponsored NBDPS, one of the largest population-based studies of birth defects ever conducted. The large sample size of this dataset allowed us to make separate assessments for each birth defect, including the effect of excluding mothers who reported a fever or used antibiotics during the periconceptional period.

We also had sufficient statistical power to assess results for many (but not all) of the birth defects stratified by three levels of maternal use of multivitamins.

Among infants whose mothers were exposed to 1–5 days of diarrhea, after restricting to isolated birth defects, the results for two of the eleven birth defects (anorectal atresia/stenosis, longitudinal limb deficiency) that had elevated aORs in the main analysis, were decreased by more than 10% and were no longer associated with a p value < 0.10. Because the main results for these two birth defects (Table 2) may have been driven by the inclusion of other birth defects, we did not include them among the birth defects that we interpreted to be associated with maternal diarrhea. Of the 32 birth defects studied, there remain a total of nine birth defects that have elevated aORs for associations with maternal diarrhea lasting 1–5 days (hypoplastic left heart syndrome, secundum atrial septal defect, muscular ventricular septal defect (not simple), single ventricle complex, esophageal atresia, diaphragmatic hernia, omphalocele, gastroschisis and amniotic band syndrome).

Sensitivity analyses that excluded mothers who reported fever during the periconceptional period revealed that, for maternal diarrhea lasting 1–5 days, the elevated aORs for five birth defects; single ventricle complex, esophageal atresia, longitudinal limb, gastroschisis and amniotic band syndrome decreased by 21.5%, 9.6%, 7.8%, 12.2% and 25.2%, respectively. Similar results were not observed for mothers who reported diarrhea lasting 6 days or more. As acute gastrointestinal infections are usually of short duration and often accompanied by fever (Bellido-Blasco and Arnedo-Pena 2011), these results support the possibility that the large difference in our results for diarrhea lasting 1–5 days compared with diarrhea lasting six or more days, are due to a higher proportion of acute gastrointestinal infections (often with fever) in the group with a shorter duration of diarrhea. This explanation is also supported by previous studies that observed that maternal fever during early pregnancy is associated with a number of different types of cardiac and non-cardiac structural birth defects (Dreier, Andersen, and Berg-Beckhoff 2014; Botto et al. 2014; Waller et al. 2018; Mohan Dass et al. 2022).

For maternal diarrhea lasting 1–5 days, sensitivity analyses restricted to mothers who did not report using antibiotics during the periconceptional period revealed that the elevated aORs for two other birth defects; muscular ventricular septal defects (not simple) and diaphragmatic hernia, were decreased by 16.6% and 14.1%, respectively. These modest decreases in the effect of maternal diarrhea for two birth defects, suggest that maternal use of antibiotics is less likely to explain our results compared with maternal fever.

Three previous studies reported associations between maternal diarrhea (Felkner et al. 2003) or maternal acute gastrointestinal infections (Rumbo et al. 2021; Acs et al. 2010) and birth defects.

Felkner et al. (2003) conducted a case control study and observed that 9.9% of mothers reported one or more episodes of periconceptional diarrhea and these mothers had an increased risk of pregnancies affected by neural tube defects (anencephaly, spina bifida and encephalocele), aOR = 3.7, 95% CI = 1.8, 7.6. Using the same definition of maternal diarrhea as Felkner et al. (2003), we observed no association between maternal diarrhea and either of the two neural tube defects that we studied (anencephaly and spina bifida). This may be explained by a difference in the frequency of severe gastrointestinal infections between our study population and that of Felkner et al., which was restricted to Mexican American mothers living in border counties in Texas.

Rumbo et al. conducted a case control study of diagnoses of gastrointestinal infections and birth defects in Columbia (Rumbo et al. 2021). A diagnosis of a gastrointestinal infection during early pregnancy was associated with an increased risk for all birth defects in aggregate (crude OR = 2.06, 95% CI = 1.18, 3.59), but individual birth defects were not assessed (Rumbo et al. 2021).

Acs et al. (2010) reported results of a Hungarian case–control study of infants of pregnant women diagnosed with infectious diarrhea. Exposed mothers had significantly increased risks of congenital heart defects (in aggregate) (aOR = 2.4; 95% CI = 1.0, 6.3), and limb deficiencies (in aggregate) (aOR = 14.1; 95% CI = 4.8, 41.4), which may be consistent with the associations we observed for narrower categories of congenital heart and limb defects. Acs et al. also observed that exposed mothers had increased risks of cleft lip with or without cleft palate (aOR = 12.3; 95% CI = 5.4, 28.0), a finding we did not observe. We suspect that their focus on diagnoses of infectious diarrhea may have represented a more severe condition than the exposure in our study (maternal report of diarrhea of any cause).

There are several ways by which diarrhea might cause lower maternal plasma levels of vitamins, which can then lead to deleterious effects in the developing embryo: (1) Diarrhea may disrupt bacterial species in the maternal microbiome that make important contributions to plasma vitamin levels (Camilo et al. 1996); (2) The increase in intestinal transit associated with diarrhea may allow insufficient time for absorption of vitamins from food; (3) And, mothers who are experiencing diarrhea may limit or change their food intake.

We did not find evidence to support our hypothesis that maternal use of multi-vitamins might modify associations between maternal diarrhea and birth defects. This may be explained if maternal diarrhea interfers not only with the absorbtion of vitamins from food but also with the absorbtion of multivitamins. And, for many birth defects, the statistical power to detect an effect of multivitamin use was limited, resulting in aORs that were imprecise or could not be calculated.

Differential recall bias, in which mothers of cases are more likely to recall exposures that could be deleterious to the embryo, is a potential concern when exposures are based on maternal interviews conducted after delivery. However, there has been limited research in this area and when Case et al. studied responses to the NBDPS interview question (What do you think causes birth defects?) (Case et al. 2014), no respondents mentioned diarrhea or gastrointestinal illnesses. Therefore, it is doubtful that mothers are aware that maternal diarrhea is a risk factor for birth defects. Thus, case mothers are probably not more likely to report diarrhea than control mothers.

To address the multiple statistical tests made in this study, we calculated the number of elevated aORs that we expected to observe under the null hypothesis that maternal report of diarrhea is not associated with any of the categories of birth defects in this study. And, because associations were assessed between diarrhea lasting 1–5 days and 32 birth defects and between diarrhea lasting 6 days or more and 30 birth defects, we made separate calculations for each of these independent exposures. These numbers were calculated based on a p value of 0.10 and the assumption that under the null hypothesis that 0.5 of all aORs would be elevated (32 × 0.10 × 0.50 = 1.60 and 30 × 0.10 × 0.50 = 1.50, respectively). By chance alone, for diarrhea, lasting 1–5 days, we would expect to observe approximately 1.6 birth defects with elevated aORs and for diarrhea lasting 6 days or more, we would expect to observe approximately 1.5 birth defects with elevated aORs. Thus, multiple statistical tests may explain the single elevated aOR that we observed for diarrhea lasting 6 days or more. However, multiple statistical tests are very unlikely to explain all nine of the birth defects with elevated aORs that we observed for diarrhea lasting 1–5 days.

Some misclassification of exposure (diarrhea) is likely to have occurred due to the wide periconceptional period (3 months before to 3 months after conception) that was used in the NBDPS question on diarrhea. Therefore, the question on diarrhea may have misclassified some mothers as exposed to diarrhea, who were exposed prior to conception; which is a time period when this exposure is unlikely to increase the risk of a birth defect. Also, the question on diarrhea asked about “unusually loose stools”, which may have been interpreted differently from one participant to the next. Thus, in this study, misclassification of diarrhea is likely to be non-differential, that is, occurring among cases and controls with a similar frequency. It follows that estimates from this study are most likely to have underestimated the magnitude of the increase in risk of birth defects associated with maternal diarrhea.

Another limitation of our study is that because information on the timing of maternal fever is also based on maternal recall, it is possible that some of the study participants did not accurately recall the time period in pregnancy during which a fever occurred.

This study provides evidence that short bouts of maternal diarrhea are associated with an increase in the frequency of approximately 28% (9 of 32) of the structural birth defects we examined. In addition, previous studies have observed that maternal gastrointestinal infections, which generally resolve in a few days (Bellido-Blasco and Arnedo-Pena 2011), have also been linked to an increased frequency of birth defects. Therefore, women who are planning to become pregnant or have recently become pregnant may benefit from measures to avoid infections, including those of gastrointestinal origin.

To date, only a few studies have addressed the possibility that maternal diarrhea or gastrointestinal infections are associated with birth defects. Additional research should be undertaken to confirm the associations we observed and to identify additional exposures that may modify them.

Supporting Information

Additional supporting information can be found online in the Supporting Information section.