Article Text
Abstract
Background Observational evidence shows that maternal (and paternal) health behaviours around pregnancy are associated with offspring health. It is often difficult to infer whether associations represent causal effects or arise due to confounding. We have conducted a multi-cohort, multi-exposure, multi-outcome study, applying several approaches that can strengthen causal inference. We have developed a web-app to make our results available to the research community.
Methods In a reduced phenome-wide association study (PheWAS), we separately regressed over 200 child health outcomes on up to four parental health behaviour classes at several timepoints, or polygenic risk scores (PRS) for these health behaviours. To increase power, we ran the PheWAS in four cohorts and meta-analysed results (maximum N=106,396). Finally, we developed a web-app (EPoCH) to search, visualize and download results.
Results Our EPoCH web-app facilitates triangulation of approaches to strengthen causal inference. Users are shown associations (effect estimates and 95% confidence intervals): 1) adjusted for various confounders; 2) from different cohorts, allowing cross-context comparisons; 3) for (i) maternal and paternal exposures, and (ii) pre- and post-natal exposures, enabling negative control comparisons; 4) for different pregnancy exposure timepoints, enabling exploration of the importance of exposure timing; 5) for different doses of ordinal exposures, enabling exploration of dose-response effects; 6) for PRS, facilitating Mendelian Randomisation analyses.
Conclusion The EPoCH app provides a useful resource for improving causal inference in parental prenatal effects research. Findings can help identify the most appropriate prenatal targets for more effective interventions to improve child health.