Original article: is the protective association between hyperemesis gravidarum and birth defects biased by pregnancy...

We assessed whether the protective association between hyperemesis gravidarum and birth defects could be due to selection bias from exclusion of pregnancy terminations.

We designed a cohort study of 2,115,581 live births in Canada, 1990-2016. The main exposure measure was hyperemesis gravidarum. The main outcome measure included any birth defect at delivery. We estimated risk ratios (RR) and 95% confidence intervals (CI) for the association of hyperemesis gravidarum with birth defects in log-binomial regression models, and assessed the extent of selection bias through correction factors.

Hyperemesis gravidarum was associated with 0.88 times the risk of birth defects in models not corrected for bias (95% CI 0.82-0.94). Correction for selection bias suggested that if screening for birth defects was associated with 1.33 times the chance of detecting birth defects and having a pregnancy termination, there would be no association with hyperemesis gravidarum. If ultrasound was associated with 2.00 times the chance of detecting birth defects and 1.50 times the risk of pregnancy termination, hyperemesis gravidarum would be associated with 1.27 times the risk of birth defects (95% CI 1.18-1.35).

The protective association between hyperemesis gravidarum and birth defects in previous studies may be due to selection bias.