Obesity’s health risks shift with age and sex, new genetic study reveals

By Vijay Kumar MalesuReviewed by Susha Cheriyedath, M.Sc.Oct 27 2025

A massive UK Biobank analysis reveals that the health risks of obesity shift over time, peaking at different ages for men and women, and that midlife preventive care may blunt its cardiovascular damage.

Study: Time-resolved Mendelian randomization detects substantial variation in the detrimental effect of obesity throughout life. Image Credit: Billion Photos / Shutterstock

In a recent study published in the journal Science Advances, a group of researchers estimated how the cumulative effect of sustained high BMI on major diseases varies with age and sex, using a novel time-resolved MR framework applied to the United Kingdom (UK) Biobank.

Background

Nearly one billion adults worldwide live with obesity, making it a leading driver of diabetes, heart disease, and disability. People often ask, “At what age does extra weight do the most damage?” Yet most studies average risk across adulthood, masking when prevention could matter most. Obesity is usually quantified by BMI, but its impact likely changes as bodies, hormones, behavior, and clinical care evolve over life. Traditional analyses also struggle to separate cause from correlation. Genetics helps because MR can infer causality, but conventional approaches typically assume effects are time-fixed. This study introduces a method to track dynamic risk throughout life.

About the Study

The investigators analyzed UK Biobank participants with genetic and linked health-record data, restricting to unrelated adults of European ancestry after quality control. The final sample included 361,906 individuals (median end-of-follow-up age ~70 years). BMI measured at baseline was standardized within sex-by-age strata. Age in years was the primary time scale; first occurrences of T2DM, CAD, AF, and OA were derived from curated ICD-10 codes. To avoid sparse data and boundary effects at the oldest ages, follow-up was capped at 76 years. Covariates included sex, age at assessment, genotyping array, assessment center, and 25 genetic principal components.

To infer causality, the team used MR with polygenic scores (PGS) as instruments. GWAS for BMI were run in two independent half-samples (~180,953 each), identifying independent genome-wide significant SNPs. Disease-specific BMI PGS were built using Steiger filtering to exclude variants that explained more variation in the outcome than in the exposure, limiting reverse causation. Time-to-event models employed Aalen’s additive hazard model to estimate both cumulative (“life-course”) effects and momentary effects. The novel approach accounts for how genetic effects on BMI change with age. Sensitivity analyses probed selection bias, lipid-lowering treatment in CAD-free individuals at assessment, and SBP as an alternative exposure for AF and CAD.

Study Results

Across adulthood, higher BMI causally increased rates of T2DM, CAD, AF, and OA, but the magnitude and shape of the risks varied markedly by age and sex. For OA and AF, cumulative effects generally strengthened with age; however, BMI became a nontrivial OA risk more than 20 years earlier than it did for AF. In contrast, AF’s BMI-related risk ramped up steeply later in life. These differences suggest musculoskeletal pathways manifest earlier, whereas atrial substrate and comorbidity interactions amplify later.

For T2DM, the BMI effect rose from midlife but plateaued specifically around ages 60–70 years in the sex-combined analysis. In CAD, a striking trough appeared, with risk decreasing significantly from around age 50 and reaching near-null around 65–70 years, before rising again in older age. This U-shaped pattern was reproducible and not explained by retirement-age participation artifacts. In secondary analyses, the CAD trough was more pronounced among participants who reported lipid-lowering therapy at baseline and were free of prior CAD events, consistent with primary prevention (e.g., statins) dampening BMI-related coronary risk in that window. By contrast, BMI’s effect on AF lacked a trough, aligning with evidence that statins do not materially reduce AF risk. When the exposure was SBP instead of BMI, a moderate trough emerged for AF, plausibly reflecting midlife antihypertensive treatment, while SBP’s effect on CAD showed no clear trough.

Sex-stratified trends revealed generally stronger BMI effects in males for T2DM, CAD, and AF. An exception was OA, where effects were similar across sexes until roughly age 60, after which female trends hinted at a decrease, though with diverging confidence intervals introducing uncertainty. Notably, T2DM displayed a pronounced female-specific trough: women showed a temporary decline in BMI-related risk beginning around age 60 and lasting approximately 10 years, whereas men’s risk continued to climb. Analyses found this female trough was not explained by timing of menopause or by use of menopausal hormone therapy, raising the possibility of sex-differential engagement with prevention (for example, weight management or clinical monitoring).

Pathway-level clustering of BMI-associated SNPs suggested heterogeneous mechanisms: distinct genetic clusters conveyed different temporal risk signatures for CAD and T2DM, with “high-risk” clusters largely driving the CAD trough and sex differences in T2DM. Methodologically, the team confirmed genetic instrument effects on BMI decline with age, validating the need for time-resolved methods. Simulations confirmed their approach accurately recovered dynamic effects even with time-varying instruments. Importantly, correcting for potential selection into the cohort slightly reduced overall amplitudes but left key features, including troughs, intact.

Conclusions

Sustained high BMI causally elevates risk for diabetes, CAD, AF, and OA, but the “when” matters. Risk is not linear: specific age windows (e.g., 50–70 for CAD, 60–70 for T2DM in women) show attenuations that likely reflect preventive care (lipid-lowering or blood-pressure treatment), especially for CAD and AF. Women show a unique temporary midlife dip in BMI-related diabetes risk peaking around 60–70, unexplained by menopause or hormone therapy.

Clinically, the findings argue for timing prevention to life stages when it averts the most events, with sex-specific nuance. The time-resolved MR framework enables the detection of these dynamic risk patterns obscured in traditional analyses. The authors note limitations, including the model’s assumption of an immediate biological response to BMI changes and reduced precision for early-life effects due to the use of adult-focused genetic instruments.