Time-resolved Mendelian randomization detects substantial variation in the detrimental effect of obesity throughout life.

The global burden of disease attributable to obesity continues to rise. The disease incidence is substantially higher in elderly populations, but how obesity affects disease risk across a lifetime is largely unknown. To quantify the long-term temporal impact of obesity, access to large-scale longitudinal cohorts spanning many decades would typically be required. However, these longitudinal studies are rare and may be heavily biased by the presence of unaccountable confounding. Here, we develop a method-time-resolved Mendelian randomization-to estimate the cumulative effect of body mass index on disease risk at different ages. Using the UK Biobank, we find strong age-varying patterns for type 2 diabetes mellitus, coronary artery disease, and atrial fibrillation, as well as for osteoarthritis. We demonstrate that some of the most notable temporal characteristics are sex specific, while other features are shared between sexes. Analyses suggest that these features can be manifestations of primary prevention strategies.