Losing Weight Predicted More Heart Attacks

A prespecified analysis of 17,604 people in the SELECT trial asked whether semaglutide’s protection against heart attacks came from the weight it removed. The answer was mostly no.^[1] And in the placebo group, the same analysis found that losing weight was associated with more cardiovascular events — which is the most instructive sentence published about this drug class this year.

The benefit did not follow the weight

Everyone in SELECT was at least 45, carried a BMI of 27 or more, and had heart disease already diagnosed; none had diabetes. Across 41 countries, half drew weekly semaglutide at 2.4 mg and half a placebo. This analysis looked inside that result to ask what the protection tracked.

• Benefit appeared across every baseline category of body weight and waist circumference. Heavier and lighter participants gained similarly.
• Weight lost by week 20 showed no linear relationship to heart attacks and strokes afterward.
• Waist reduction did track it — both at week 20 and over 104 weeks.
• An estimated 33% of the benefit was mediated through waist reduction (hazard ratio 0.86, 95% CI 0.77–0.97 after adjusting for changes over time).

The authors’ own interpretation is that the cardioprotective effect was independent of baseline adiposity and of weight loss, with only a small association with waist circumference — pointing to mechanisms beyond fat reduction.

Two-thirds of the heart benefit is not explained by anything the scale or the tape measure recorded.

That waist matters while weight does not is itself informative. Waist circumference tracks visceral fat — the metabolically active fat around the organs — while body weight lumps that together with muscle, bone and water. Two people losing the same weight can lose very different tissue, which is a recurring theme in this drug class.

We reached a similar conclusion from a much weaker analysis, a post-hoc mediation study in sleep apnea, in is it just the weight loss. This is the same question answered prespecified, in 17,604 people, against a hard endpoint.

The sentence in the placebo column

The explanation is not mysterious and it is worth knowing well. When nobody has given you a drug that causes weight loss, the commonest reasons to lose weight are illness — developing cancer, heart failure, or the general decline that precedes a cardiac event. Weight falls because something is wrong. The weight loss is a symptom, and the cardiac event follows for the same underlying reason.

Epidemiologists call this reverse causation, and it is why decades of observational studies linking weight loss to worse outcomes were so hard to interpret. What makes this instance valuable is where it appeared: inside a randomized controlled trial, in the control arm, where it cannot be waved away as a defect of observational data.

That is worth carrying beyond this trial. Any study that reports weight change as an outcome or a predictor, without establishing what caused it, is reporting an ambiguous quantity.

What it means practically

• The cardiovascular benefit is not a reward for losing weight. In SELECT, people who lost less were protected about as much as people who lost more.
• That is reassuring for slow responders, who form a substantial minority in every trial and who often conclude the drug is not working.
• Waist may be the better thing to watch than weight, and it is the cheaper measurement.
• Unexplained weight loss without a drug is a reason to see a doctor, not a success. That is the whole placebo finding in one line.

Two limits belong with all of it. The trial was funded by Novo Nordisk, which makes semaglutide — ordinary for a trial of this kind and worth stating. And the 33% mediation figure is a modeling estimate that rests on assumptions trial data cannot verify, the same caution that applies to every mediation analysis.