Comparing genome-wide significant and chemosensory variants as instruments for dietary patterns in Mendelian randomization.

Abstract

BACKGROUND: Diet is a modifiable risk factor for cardiometabolic disease, yet establishing causality remains challenging. Mendelian randomization (MR) leverages genetic variants as instrumental variables (IVs) to enable causal inference. METHOD: Using two-sample MR, we assessed the causal effects of four principal component-derived dietary patterns (DPs)-Unhealthy, Healthy, Meat-based, Pescatarian-on cardiometabolic outcomes including body mass index, coronary artery disease, blood lipids, blood pressures, type 2 diabetes, fasting glucose and insulin, and glycated haemoglobin. Two sets of IVs were employed: conventional genome-wide significant variants associated with each DP, filtered for pleiotropy and directionality; and biologically informed variants in chemosensory receptor genes, given the role of taste and smell perception in food choice. RESULTS: Using conventional IVs, the Pescatarian DP was associated with reduced fasting insulin (βIVW = -0.10 pmol/L per SD increase in the Pescatarian DP score, 95% confidence interval -0.15, -0.04; P = 1.19 × 10-3), surviving multiple sensitivity analyses. Associations between the Unhealthy DP and elevated blood pressure and glycated haemoglobin should be interpreted cautiously; one of the two filtered IVs was strongly associated with caffeine intake, limiting the attribution of these findings to the DP itself. Chemosensory Receptor IVs yielded null findings, reflecting insufficient power. CONCLUSION: Evidence for causal effects of DPs on cardiometabolic traits was limited, with the strongest support for a protective effect of the Pescatarian DP on fasting insulin. Chemosensory IVs demonstrated limited utility for DPs, likely reflecting the heterogeneous and complex sensory profiles of overall diets. Future efforts should consider guideline-based dietary indices to facilitate interpretability and translation.