SELECT Trial Mediation Analysis Finds Semaglutide's Heart-Protective Effect Not Fully Explained by Weight Loss or Other Known Risk Factors

The European Heart Journal published a mediation analysis of the SELECT cardiovascular outcomes trial on September 23, 2026, addressing a question left open by the trial's original results: what actually drives semaglutide's reduction in heart attacks and strokes. SELECT enrolled 17,604 adults with overweight or obesity and established atherosclerotic cardiovascular disease but not diabetes, and previously reported that semaglutide reduced major adverse cardiovascular events (MACE) by 20% relative to placebo. The new analysis was led by Helen M. Colhoun of the University of Edinburgh, with co-authors including several employees of Novo Nordisk, which manufactures semaglutide and funded the study.

Researchers used the Vansteelandt repeated regression method, a counterfactual statistical approach, to examine changes from randomization to 24 months in body weight, waist circumference, high-sensitivity C-reactive protein (hsCRP), glycated hemoglobin (HbA1c), lipids, blood pressure, estimated glomerular filtration rate, and urinary albumin-to-creatinine ratio. Each measure was tested individually, and then in combination, as a potential statistical mediator of the trial's cardiovascular risk reduction.

Semaglutide significantly improved every mediator studied (p<0.05). Examined separately, point estimates for the percentage of the cardiovascular benefit attributable to each factor were largest for waist circumference (64.0%, 95% confidence interval 27.5 to 179.6), hsCRP (42.1%, 95% CI 17.9 to 110.5), HbA1c (29.0%, 95% CI -20.7 to 120.5), and body weight (19.5%, 95% CI -33.0 to 110.7). Every estimate carried a wide confidence interval, and supplementary analyses suggested the body-weight and waist-circumference figures specifically were unreliable, potentially because the relationship between weight or waist change and cardiovascular events differed between the semaglutide and placebo groups.

In a multivariable analysis combining all potential mediators, the joint estimate for percent mediation was 31.4% (95% CI -30.1 to 143.6). When body weight and waist circumference were excluded and only the remaining variables, such as hsCRP, HbA1c, lipids, blood pressure, and kidney measures, were considered together, the estimate rose to 46.0% (95% CI -6.0 to 161.0). In both cases the confidence intervals were wide enough to include implausible values, reflecting substantial statistical uncertainty.

The authors concluded that mediation analysis could not fully or reliably attribute semaglutide's cardiovascular benefit in SELECT to the known risk factors they examined, and wrote that further investigation of additional biomarkers and mechanisms is warranted. The finding does not overturn SELECT's headline result, a 20% reduction in major cardiovascular events, but it leaves open, more than two years after that result was first published, how much of the effect stems from weight loss, from other measured risk-factor improvements such as reduced inflammation, or from mechanisms the trial did not measure.

Semaglutide is already approved by the FDA and the European Medicines Agency and registered with Australia's Therapeutic Goods Administration, marketed as Wegovy for chronic weight management and cardiovascular risk reduction and as Ozempic and Rybelsus for type 2 diabetes. This mediation analysis does not change that regulatory status in any market. It is an exploratory statistical analysis of existing trial data, not a new clinical trial or a controlled test of any specific mechanism, and is not treatment or dosing guidance; cardiovascular risk management decisions should be made with a qualified clinician.