Most current nutrition policies and dietary recommendations still reflect decades of research addressing the mechanism of action or health risks of individual nutrients. Yet, most high-income countries including the UK are far from reaching the dietary intakes which are recommended for good health. Food-based dietary patterns (DPs) can help target specific combinations of foods that are associated with disease risk, recognising the coexistence of multiple nutrients within foods and their potential synergistic effects. Reduced rank regression (RRR) has emerged as a useful exploratory approach which uses a priori knowledge of the pathway from diet to disease to help identify DPs which are associated with disease risk in a particular population. Here we reviewed the literature with a focus on longitudinal cohort studies using RRR to derive DPs and reporting associations with non-communicable disease risk. We also illustrated the application of the RRR approach using data from the UK Biobank study, where we derived DPs that explained high variability in a set of nutrient response variables. The main DP was characterised by high intakes of chocolate and confectionery, butter and low-fibre bread, and low intakes of fresh fruit and vegetables and showed particularly strong associations with CVD, type 2 diabetes and all-cause mortality, which is consistent with previous studies that derived 'Western' or unhealthy DPs. These recent studies conducted in the UK Biobank population together with evidence from previous cohort studies contribute to the emerging evidence base to underpin food-based dietary advice for non-communicable disease prevention.
Keywords: CVD; Diabetes; Dietary patterns; Reduced rank regression.