Accelerometer-Derived and Genetically Inferred Physical Activity and Human Disease

Abstract Physical activity is favorable to health but the relations with human disease and causal effects are poorly quantified. Previous studies have largely relied on self-reported estimates 1–3 which are subject to recall bias, confounding, and reverse causality. Using wrist-worn accelerometer measurements from the UK Biobank, we tested associations between moderate-to-vigorous physical activity (MVPA) – both total MVPA minutes and whether MVPA was above a guideline-based threshold of ≥150 minutes/week 4–6 – and incidence of over 1,200 diseases. In 96,466 adults who wore accelerometers for one week (mean age 62±8 years), MVPA was associated with 401 (33%) tested diseases at a false discovery rate of 1% during a median of 6.2 years follow-up. Greater MVPA was overwhelmingly associated with reduced disease risk (98% of associations) with hazard ratios (HRs) ranging from 0.10-0.95 per standard deviation (SD) and associations spanning all 18 disease categories tested. A similar pattern of associations was observed when assessing the guideline-based threshold of ≥150 MVPA minutes/week. We examined a polygenic risk score for MVPA as an instrumental variable for activity within a separate UK Biobank sample (N=392,058, mean age 57±8 years). Greater genetically inferred MVPA was associated with reduced risk of 182 (14%) incident diseases (HR range 0.74-0.97 per 1 SD), and included strong associations conferring reduced risks of obstructive bronchitis, peripheral vascular disease, type 2 diabetes, and gastritis/duodenitis. Objective physical activity is broadly associated with lower disease incidence and many associations are consistent with a causal effect.