Ageing drives progressive biological change and increasing risk of chronic disease and multimorbidity. Using UK Biobank baseline phenotypes and linked longitudinal outcomes, we will test whether integrating multiple modalities improves prediction of incident age-related diseases and informs prevention.
Research questions: (1) Predict incident disease using demographics, lifestyle/socioeconomic factors, medications and physical measures. (2) Quantify added value of blood/urine biomarkers and other assay-based measures (including proteomic/metabolomic panels where available). (3) Quantify added value of imaging-derived phenotypes (IDPs) and genetic variation (imputed genotypes and sequencing variant data where available) and identify key contributors to risk stratification.
Objectives: build a harmonised dataset; define incident outcomes from hospital, primary care (where available), death records and registries with baseline exclusions; develop and internally validate models with train/validation/test splits and sensitivity analyses; compare discrimination and calibration; and report interpretable, non-disclosive summaries.
Pre-specified outcomes will be analysed separately (secondary composite/multimorbidity analyses): type 2 diabetes; coronary heart disease/MI; stroke; heart failure; atrial fibrillation; dementia (all-cause, Alzheimer’s, vascular); chronic kidney disease; COPD/asthma; chronic liver disease/NAFLD; osteoarthritis; osteoporosis/fragility fracture; inflammatory/autoimmune diseases (e.g., rheumatoid arthritis, psoriasis, inflammatory bowel disease); gout; sensory disorders (cataract/AMD, hearing loss); and major cancers (breast, prostate, colorectal, lung).For each outcome, we will model incident (new-onset) disease after baseline, excluding prevalent cases at baseline, and ascertain events using linked hospital, primary care (where available), death records and registries with pre-specified code lists.
