Perinatal & Pediatric
Associations between diphenhydramine and specific birth defects stratified by imputed indication, National Birth Defects Prevention Study and Birth Defects Study to Evaluate Pregnancy exposureS Eva Williford* Meredith Howley Eva M Williford Sarah C Fisher Martha M Werler Julie M Petersen Suzanne M Gilboa Craig Hansen Elizabeth C Ailes Mollie Wood Marilyn L. Browne
Diphenhydramine is used for various indications, including nausea and vomiting in pregnancy, allergy relief, and as a sleep aid. While diphenhydramine has been associated with increased risk for a small number of specific birth defects, findings from National Birth Defects Prevention Study (NBDPS; 1997–2011) were largely null. Yet, confounding by indication may persist as NBDPS did not collect indication. Data from Birth Defects Study to Evaluate Pregnancy exposureS (BD-STEPS; 2014–2019), a study conducted subsequent to NBDPS in a subset of sites, included questions on indication. Thus, we sought to use BD-STEPS data to impute missing indication in NBDPS data and generate estimates of the association between diphenhydramine and 12 birth defects in the combined data stratified by indication.
Cases included infants, stillbirths, or terminations with a birth defect. Controls included live born infants without a birth defect. Mothers of cases and controls reported early pregnancy medication use and BD-STEPS participants reported indication. We used a sequential regression multivariate imputation approach and created 50 datasets imputing missing values, including indication. We used Firth’s logistic regression with propensity scores to calculate adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for the association between diphenhydramine and each birth defect stratified by indication. We pooled ORs across the 50 imputed datasets using Rubin’s rules.
In NBDPS and BD-STEPS, 461 cases (2–3% of each birth defect) and 341 controls (2%) reported early pregnancy diphenhydramine use; 66% were missing indication, which were all NBDPS participants. The pooled ORs for the 12 defects across the three indications ranged from 0.6 to 2.0; all CIs included the null. Our findings confirm previous NBDPS null findings and suggest that confounding by indication is likely not a large source of bias in studies of early pregnancy diphenhydramine use and the 12 birth defects analyzed.
