- Design
- Counterfactual mediation analysis of a randomised trial
- Population
- SELECT: adults with CVD, BMI 27 or more, no diabetes
- Primary outcome
- Proportion of MACE reduction mediated by 24-month risk-factor change
- Effect
- Combined mediation 31.4% (95% CI −30.1 to 143.6); waist 64.0% (27.5 to 179.6)
A mediation analysis of the SELECT trial in the European Heart Journal (23 September 2026) asked how much of semaglutide's 20% reduction in major adverse cardiovascular events, in adults with cardiovascular disease and BMI of 27 or more without diabetes, could be explained by changes over 24 months in weight, waist, hsCRP, HbA1c, lipids, blood pressure, eGFR and albuminuria.
Semaglutide improved all of them. Estimated mediation was largest for waist circumference (64%) and hsCRP (42%), but every confidence interval was wide — waist 27.5% to 179.6%, weight −33% to 111%. All mediators combined explained an estimated 31.4% (95% CI −30.1% to 143.6%). The authors judged the weight and waist estimates unreliable.
The honest conclusion is that nobody yet knows how semaglutide protects the heart. What this does suggest is that the degree of weight loss is not a reliable guide to cardiovascular benefit in an individual.
- Do not judge semaglutide's cardiovascular value by how much weight a patient loses
- Continue semaglutide for secondary prevention in eligible patients even when weight loss is modest
- Keep treating blood pressure, lipids and glucose separately — none explains the benefit alone
- hsCRP fell on semaglutide; inflammation remains a plausible but unproven pathway
Why it matters
It undercuts the assumption that GLP-1 agonists protect the heart simply by making patients thinner.
Don't overread it
Mediation analysis is exploratory, and intervals this wide cannot confirm or exclude any single pathway.
The statistics, in plain English
Percent mediated is the share of the treatment effect statistically explained by a change in the measure. An interval that runs past 100% or below 0% — as all of these did — means the data cannot tell us whether the factor explains all, some or none of the effect.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for cardiology, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free