Abstract Body (Do not enter title and authors here): Background: Diet is a major modulator of blood metabolites linked to the risk of type 2 diabetes (T2D) and coronary heart disease (CHD). However, no study has comprehensively assessed dietary factors that jointly predict diet-metabolite-disease relations.
Hypothesis: We aimed to derive food-based Dietary Metabolite Indexes (DMIs) from metabolomic risk scores (MRSs) that capture associations between diet variability and plasma metabolites linked to the risk of incident T2D and CHD, and evaluated whether these DMIs significantly predicted the incidence of T2D and CHD.
Methods: Using a two-stage approach, we first analyzed relations between nutrient intake variability from 7-day diet records and metabolites in the Women’s Lifestyle Validation Study (WLVS). MRSs for T2D (n=1250) and CHD (n=1524) were trained using the diet-related metabolites and tested in nested case-control studies within the Nurses’ Health Study (NHS). DMIs were constructed in the WLVS (n=687) using validated food frequency questionnaires to predict each MRS, with coefficients derived via elastic net models. These DMIs were then applied to a parental NHS cohort of participants initially free of T2D and cardiovascular disease (n=61195).
Results: We identified 171 metabolites significantly associated with both the levels and variability of nutrient intakes. Two novel MRSs comprising 156 metabolites for T2D and 158 metabolites for CHD were associated with incident T2D (relative risk [RR] per 1 SD: 2.77 [95% CI: 1.81, 4.24]) and CHD (RR 1.32 [1.02, 1.70]). For DMI construction, 18 and 24 food items were selected for T2D and CHD, respectively. The CHD-DMI included inverse loadings for fish, fruits, nuts, red/orange vegetables, and whole grains, and direct loadings for processed/red meats, artificially sweetened beverages, and Western-style foods. The T2D-DMI included inverse loadings for citrus fruits, nuts, dark-green vegetables, coffee, and direct loadings for high-fat foods, sweetened beverages, and animal foods, including processed meats. Both DMIs significantly predicted the incidence of T2D (hazard ratio [HR] per 1 SD: 1.15 [1.12, 1.18]) and CHD (HR: 1.06 [1.02, 1.11]) after adjusting for traditional risk factors.
Conclusion: Our study identified key dietary factors associated with dietary variability and plasma metabolomic profiles linked to the risk of incident T2D and CHD. Food-based DMIs hold great promise for precision nutrition and prevention strategies for cardiometabolic diseases.