Semaglutide's Heart Benefit Is Not Just About Weight Loss

In 2023, the SELECT trial showed that weekly semaglutide cut major cardiovascular events by 20% in 17,604 adults with heart disease and overweight or obesity but no diabetes. In absolute terms, 6.5% of people on the drug had a heart attack, stroke or cardiovascular death over a mean follow up of 39.8 months, against 8.0% on placebo. The intuitive explanation was the scale: people lost weight, so their hearts did better. Two analyses of the same trial published in September 2026 make that explanation much harder to hold.

The first, in the European Heart Journal on September 23, tried to measure how much of the benefit flowed through known risk factors. The second, in JAMA Cardiology on September 16, asked whether the drug still helped the frailest patients, the group doctors worry about most when a medicine makes people lose weight. Meanwhile, on August 28 the FDA added a cardiovascular indication to tirzepatide (Mounjaro) for adults with type 2 diabetes, and a Danish analysis presented at the EASD meeting in Milan argued that eligibility rules based on body mass index miss people at real cardiac risk.

What the mediation analysis measured

Mediation analysis asks a counterfactual question: if people on semaglutide had experienced the same changes in a risk factor as people on placebo, how much of the benefit would remain? Helen Colhoun of the University of Edinburgh and colleagues ran that calculation for 12 candidates measured over two years, including body weight, waist circumference, blood pressure, cholesterol, HbA1c, kidney markers and high sensitivity C reactive protein (hsCRP), a blood marker of inflammation.

Semaglutide improved all 12. None of them explained the effect with any confidence. Taken one at a time, waist circumference had the largest point estimate at 64%, then hsCRP at 42.1%, HbA1c at 29% and body weight at just 19.5%. Every estimate came with a wide confidence interval: for body weight it ran from minus 33% to 110.7%. When the authors modeled weight and waist together, the estimate fell to 12.3%. A model with the other factors but without weight and waist reached 46%, and adding weight and waist back lowered it to 31.4%.

That last pattern is the oddest result in the paper, and the authors offer a plausible reason. In an earlier SELECT analysis, losing more than 5% of body weight by week 20 in the placebo group was linked to more cardiovascular events, not fewer. Weight lost to illness or frailty is a warning sign, and it muddies any attempt to read weight change as a protective pathway. The European Society of Cardiology summarized it plainly: no more than half of the reduction in heart disease could be explained by changes in the risk factors measured.

The frailty question

The JAMA Cardiology paper, led by John Ostrominski of Brigham and Women's Hospital, sorted the same 17,604 participants with a 31 item frailty index built from accumulated health deficits. Thirty one percent were classed as not frail, 47% as more frail and 22% as most frail.

Cardiovascular events rose steeply with frailty, which is why the subgroup matters: the frailest people have the most to gain in absolute terms and, in theory, the most to lose from shedding muscle along with fat. The hazard ratios for the primary outcome were 0.84 in the least frail group, 0.70 in the middle group and 0.92 in the most frail group, with an interaction test of P = .09, meaning the trial could not detect a real difference between groups. Taken alone, the most frail group's confidence interval (0.76 to 1.10) included no effect, so the reassurance rests on consistency across groups, not on proof of benefit in that group by itself. Semaglutide also reduced heart failure events, hospitalization for any cause and death from any cause regardless of frailty category.

Two secondary results stand out. Quality of life scores improved more as frailty rose (P = .02 for interaction). And over two years, people on semaglutide were more likely to move to a less frail category (odds ratio 2.46) and less likely to move to a more frail one (odds ratio 0.47). The excess of treatment discontinuations due to adverse events, compared with placebo, was smaller among frailer participants.

The BMI 25 argument

At EASD, Karen Hvid of Copenhagen University Hospital presented an analysis of 313,145 adults with a BMI of 25 or above from the UK Biobank and the Copenhagen General Population Study, followed for more than ten years. About two thirds met current criteria for GLP-1 weight loss drugs: a BMI of 30 or more, or 27 or more with a weight related condition. Among those who did not qualify, more than 90% had a BMI between 25 and 26.9, and roughly half of that group (44% in the Danish cohort, 48% in the UK Biobank) had high remnant cholesterol, low grade inflammation or both.

Those people developed coronary heart disease at rates similar to people who qualified. With both markers elevated, risk was 41% higher in the Danish cohort and 47% higher in the UK Biobank than in ineligible people with healthy levels, compared with 41% and 35% higher for people who met the indication. Hvid's own conclusion was measured: clinical trials are now needed. The study shows that a group carries similar risk. It does not show that a drug would lower it.

Tirzepatide's new label

The FDA decision for Mounjaro rests on SURPASS-CVOT, which enrolled 13,299 adults with type 2 diabetes and established atherosclerotic disease and compared tirzepatide not with placebo but with dulaglutide (Trulicity), a GLP-1 drug with proven cardiovascular benefit. Over a median of 210.1 weeks, tirzepatide had an 8% lower rate of cardiovascular death, heart attack or stroke (hazard ratio 0.92, 95% CI 0.83 to 1.01). That met the bar for noninferiority; superiority was not established. The label now covers adults with type 2 diabetes at high cardiovascular risk.

Put together, these papers move semaglutide further from the diet aisle and closer to the statin shelf. The mediation result is not proof of a hidden mechanism, and its intervals are too wide for that. It is a strong argument against judging the drug by the scale alone, and the frailty data remove one of the main reasons to withhold it from older, sicker patients. The BMI 25 debate is where enthusiasm should slow down: similar risk is a reason to run a trial, not to write a prescription. For why subgroup and mediation results deserve caution, see our guide to reading a health study in five minutes, and for the other big lever on heart risk, our piece on the first oral PCSK9 pill.