This study found that after controlling for covariates, mothers of infants with CL/P and hypospadias had a reduced odds of experiencing NVP relative to controls. Treatment of NVP with a wide variety of different agents was relatively frequent - over 15% among women who reported NVP and about 10% of all women. The majority of medication groups and specific medications were not associated with the four birth defects studied, but numbers of exposed subjects were sometimes small. We did find, however, some positive associations with first trimester use of these medications: PPIs and hypospadias, steroids and hypospadias, ondansetron and CP. We also saw positive associations between PPIs and CP, and bismuth subsalicylate and NTDs; although the confidence intervals in the adjusted analyses included 1. The one inverse association was antacids and CL/P.
Our results show that pregnant women are currently taking a wide range of agents for treatment of NVP. While there is existing literature on the optimal management of NVP, much of the literature on the safety of the wide range of medications used by pregnant women is based on samples that are too small to assess risks for specific birth defects, and the large majority of teratogenic effects in humans have been shown to affect specific defects rather than increase the risk of birth defects overall. Thus, a major strength of the NBDPS case-control approach is our ability to evaluate the risks and safety of various exposures with respect to specific birth defects that have been clinically validated. The wide variety of medications reported resulted in relatively small cell sizes, limiting our ability to investigate the risk or safety of individual medications.
We observed an elevated risk for hypospadias and CP among women exposed to PPIs in the first trimester. PPI medications to which women were exposed included lansoprazole, omeprazole and esomerprazole magnesium. The number of subjects exposed to specific PPI medications was small, which made it difficult to consider relationships with the specific PPIs. While cohort studies (Lalkin et al., 1998
; Ruigomez et al., 1999
; Kallen 2001
; Kallen 1998
) and two meta-analyses (Nikfar et al., 2002
; Gill et al., 2009
) did not suggest an increased risk of birth defects overall among women who ingested PPIs in the first trimester, none had the power to rule out even modest increases in risk of specific defects. A recent study found no risk of major birth defects overall with use of PPIs in the first trimester, but did not have sufficient power to consider the risk of specific birth defects (Pasternak and Hviid, 2010
; Mitchell, 2010
Ondansetron is an anti-nausea medication primarily used to treat nausea and vomiting in patients receiving chemotherapy. However, women are also using this medication for treatment of NVP. Two previous studies (Asker et al, 2005
; Einarson et al, 2004
) found no association with use of this medication in pregnancy and major birth defects overall, but the sample sizes were quite small and insufficient to consider risks of specific defects. Our data were compatible with their findings except for CP, for which we observed a doubling in odds.
Bismuth subsalicylate, a combination of bismuth salts and sodium salicylate, can be a source of a large amount of salicylate (262 mg for original strength and 525 mg for maximum strength per tablespoon) and thus is not recommended for use in pregnancy (Black and Hill, 2003
). In previous human studies, salicylates have been associated with gastroschisis (Martinez-Frias et al., 1997
) and holoprosencephaly (Croen et al, 2000
), but our finding of an association of borderline significance with neural tube defects has not been seen before. In this analysis, steroids were associated with increased risk of hypospadias A previous analysis, which used NBDPS data to assess the odds of hypospadias after use of any corticosteroid through the 18th
week of pregnancy, found a weak association which decreased in magnitude (and had a lower 95% CI below 1.0) after adjustment for confounders (Carmichael et al, 2009
). Two previous case control studies reported a moderately increased risk for CL/P with use of steroids in the first trimester (Carmichael et al., 2007
; Pradat et al, 2003
) and a meta-analysis showed a greater than threefold risk of oral clefts with such exposure (Park-Wylllie et al., 2000). We did not find an increased risk of CL/P based on 6 exposed cases and 15 exposed controls; there were not enough exposed cases with CP to calculate a measure of association.
We found an odds ratio of 2.36 for metoclopramide for CP, based on 5 exposed cases and 18 exposed controls, with a lower confidence bound that did not exclude 1.0. While a recent study of over 3000 exposed pregnancies found no increase in risk of major birth defects overall, the study had insufficient power to consider risks associated with specific birth defects (Matok et al., 2009
A previous paper used NBDPS data to study the association between use of antihistamines in early pregnancy and risk of a spectrum of birth defects (Gilboa et al., 2009
). That study found associations between certain antihistamines and birth defects that we did not observe in this study. Specifically, they saw and increased risk for NTDs with use of diphenhydramine, doxylamine and promethazine; for CL/P with use of diphenhydramine; and, for CP with use of meclizine. Four of these five associations were weak to moderate (OR < 2.5) and one was stronger (OR > 6) but was imprecise. The present study differed from that study in that we used an additional year of NBDPS data and our study population was limited to women with NVP in the first trimester, a subset of women in the NBDPS.
Our study has a number of limitations including the potential for exposure misclassification due to incomplete recall or recall/reporting bias and selection bias as well as sparse data for some analyses. We obtained both NVP and NVP treatment information via a maternal interview conducted between six weeks and 24 months after the estimated date of delivery. Since on average, control mothers were interviewed three months sooner than cases, NVP and/or medication use could have been differentially recalled. If cases more often underreported their use of medication, the observed risk might be underestimated. We used as controls infants with no major birth defects. Studies have shown that recall of exposure information can depend on disease status (Khoury et al., 1994
). If mothers of cases recalled NVP and its treatment differently from controls, recall bias could result. To improve recall, we provided NBDPS subjects with a pregnancy calendar in advance of the interview to help them more accurately respond to questions about the exposure timing.
Interpretation is limited due to the small number of women exposed to specific medications or medication groups. Further, residual confounding could exist due to uncontrolled factors that differed between women who used and did not use NVP treatment (e.g. genetic factors). Of course, those factors would have to be associated with risk of the defects under study.
Since the analyses of the relationship between NVP medications and birth defects were limited to subjects with NVP who were either exposed or not exposed to treatments, confounding by NVP was unlikely to play a role; however, confounding by severity of NVP could have influenced the observed results. Though key analyses were repeated controlling for frequency of nausea and vomiting, this process is unlikely to control fully for confounding by indication.
Perhaps of most importance is the possibility of chance as an explanation for the statistically significant associations that we observed. There were 70 comparisons made which would suggest that three to four such associations would be expected by chance alone. Nevertheless, this is the first study with extensive medication and defect-specific data and the observed associations deserve further research.