Food-level dietary factors reveal cardiometabolic risk hidden by diet quality scores.

Non-negative matrix factorization of 148 foods in 87,162 UK Biobank participants identified eight intake factors, with four linked to incident T2D, CKD, and CAD over 9.4 years.

• Why it matters: Diet quality scores may hide food-level variation that is prognostically relevant, and this has not been tested at large scale for cardiometabolic outcomes.
• What they did: Intake patterns were derived with NMF from 148 foods and beverages, then tested in Cox proportional hazards models for incident CAD, T2D, and CKD over a median 9.4-year follow-up, with added evaluation via stepwise models, net reclassification improvement, and MWAS of 249 metabolites.
• The result: A high sugar and processed foods factor predicted higher T2D (HR 1.19; 95% CI 1.11–1.27) and CKD (HR 1.15; 1.07–1.23), while plant-based and fish predicted lower CKD (HR 0.90; 0.84–0.96) and cheese and whole grains predicted lower CAD (HR 0.94; 0.90–0.98); factors improved discrimination beyond diet quality scores (ΔC up to +0.0049) and reclassified 8–16% more, with negligible gains when adding scores to factor models.

AI summary of the abstract · may contain errors · not medical advice · read the paper.