Assessing causality in associations between maternal adiposity and obstetric and perinatal outcomes: A Mendelian randomization study

Abstract

Background Observational studies consistently suggest that maternal obesity is related to a number of adverse perinatal outcomes. However, which of these are causal or due to confounding is unclear. We aimed to carry out Mendelian randomization (MR), which uses genetic variants robustly associated with the exposure of interest as instrumental variables (IV), to determine the causal effect of maternal BMI on clinically important obstetric and perinatal outcomes. Methods Data from White European individuals in nine cohorts (ALSPAC, BiB, DNBC-PTB, HAPO, EFSOCH, Generation R, GOYA, Project Viva and INMA) were used (n = 5230 to 24,875 for different outcomes). First, multivariable regression associations between maternal BMI (pre-pregnancy or at the start of pregnancy) and 23 outcomes were examined adjusting for a range of potential confounders. Second, we combined 32 or 97 (depending on availability in cohorts) established genetic variants that have been shown to be robustly associated with BMI in a weighted allele score and used this in IV analyses to obtain causal estimates (allele score Wald ratios) of maternal BMI on each of the outcomes. We used MR-Egger, inverse variance weighting and weighted median IV analysis in sensitivity analysis to explore potential bias due to horizontal pleiotropy. When available, MR analyses for outcomes that could be influenced by offspring genetic predisposition to higher BMI were adjusted for offspring genotype. Results In standard multivariable analysis, a 1 standard deviation (SD, ∼ 4.95 kg/m2) higher maternal BMI was associated with a 0.12 SD (95% CI 0.11–0.14) higher offspring birthweight, 1.73 (95% CI 1.64–1.84) odds ratio (OR) of gestational hypertension and 1.30 (95% CI 1.17–1.93) OR of induced labor. In MR analyses, a 1 standard deviation higher maternal BMI was associated with a 0.11 SD (95% CI 0.06–0.17) higher offspring birthweight, 1.51 (95% CI 1.17–1.93) OR of gestational hypertension and 1.30 (95% CI 1.01–1.72) OR of induced labour. We also found positive associations of maternal BMI with birth length, caesarean delivery, any hypertensive disorder of pregnancy and membrane rupture before onset of contractions in multivariable regression analyses. MR point estimates for effects of BMI on these outcomes were in the same direction and of a similar magnitude to multivariable results, but the 95% confidence intervals were wide and included the null. Lastly, in multivariable regression analyses we found a positive association of BMI with gestational diabetes, which was not supported by MR results. Results from MR Egger regression, inverse variance weighted and weighted median methods were generally consistent and did not suggest strong bias by pleiotropy. Conclusions Our findings support a causal effect of increased maternal BMI on higher birthweight and greater odds of gestational hypertension and induction of labor. MR causal estimates were consistent with associations for many other outcomes. We are currently working with more studies to provide more precise estimates.