Dietary preference and susceptibility to type 2 diabetes mellitus: a wide-angle Mendelian randomization study

Abstract

ABSTRACTBackgroundSusceptibility to type 2 diabetes mellitus (T2D) is driven by genetic and environmental risk factors. Dietary preferences are a modifiable and largely environmental risk factor for T2D. The role of diet in disease liability has been limited to observational and epidemiologic studies with mixed findings.ObjectiveTo clarify the role of diet on susceptibility to T2D using genetic variants associated dietary preferences.MethodsWe collected genome-wide association data for 38 dietary preference traits plus T2D and 21 related cardiometabolic traits. We performed Mendelian randomization (MR) using genetic variants to test causal hypotheses between diet as the exposure and T2D or cardiometabolic traits as outcomes using univariable and multivariable methods along with the MR Robust Adjusted Profile Score (MR-RAPS) approach to increase power. We performed mediation analyses to evaluate the effects of dietary preferences on T2D to elucidate potential causal graphs and estimate the effects of dietary preferences mediated by potential mediators.ResultsWe report 17 significant relationships between dietary preferences and T2D or a cardiometabolic risk factor (Bonferroni-corrected P < 5.99 x 10-5), including that higher intake of cheese, dried fruit, muesli, or fat-based spreads protected against T2D. We detected 7 additional associations (Bonferroni-corrected P < 1 x 10-4), with inclusion of additional genetic variants in MR-RAPS analysis. In multivariable MR, we discovered that body mass index (BMI) was a common, shared mediator for many of these observed associations. In mediation analysis, we confirmed that substantial proportions of the protective effects of cheese, dried fruit and muesli intakes on T2D were mediated by BMI. We further observed that educational attainment was an additional mediator exclusively for muesli intake-T2D association.ConclusionsOur results provide genetic evidence supporting a link between diet and body weight, and are in line with observation of obesity and T2D in individuals and their specific preferences for food.