On one marker, substantially. In a prespecified SELECT substudy, semaglutide reduced high-sensitivity C-reactive protein by 37.8% at 104 weeks against placebo, from baseline geometric means of 1.96 and 1.91 mg/L [1]. The fall was already evident at weeks 4 and 8, before major weight loss, and it appeared among participants who lost no weight at all. What that does not establish is that lowering inflammation is how the drug works, and the authors are careful to say inflammation contributes in part rather than being the mechanism.
The obvious explanation for a weight-loss drug preventing heart attacks is the weight loss. A prespecified substudy of SELECT tested that by following an inflammation marker, and found a pattern the weight explanation does not fit. The same trial, cut a different way, is covered in the frailty analysis.
Baseline hsCRP was almost identical in the two arms, at a geometric mean of 1.96 mg/L on semaglutide and 1.91 mg/L on placebo, and it predicted what happened next. Event risk rose across the three baseline strata of below 2, 2 to below 10, and 10 mg/L or above, including significant associations with cardiovascular and all-cause death. Semaglutide then reduced hsCRP by 37.8% at 104 weeks, and reduced event risk in every one of those strata.
The timing is the part that matters. Larger falls in hsCRP went with larger weight loss, which on its own would suggest the marker was simply tracking the scale. But the fall was already evident at 4 and 8 weeks, before major weight loss had happened, and it occurred among participants who did not lose weight at all. It was also independent of LDL cholesterol, of statin use, and of which cardiovascular entry criterion had let someone into the trial.
What the blood vessels did, measured two ways
A 32-week randomized trial gave 120 people with type 2 diabetes semaglutide, empagliflozin, both, or placebo, and measured how well blood vessels widen in response to demand [2]. In the semaglutide group that index rose 0.11 from each participant’s own starting value, 95% CI 0.008 to 0.21, p=0.03. Against placebo it also rose 0.11, on an interval from −0.04 to 0.24, p=0.16.
Nothing about the measurement changed between those two sentences. What changed is what it was held up against, and once the placebo group absorbed the ordinary improvement that happens inside a trial, the effect stopped being distinguishable from noise. The adhesion molecules disagreed with each other too, with E-Selectin falling against placebo and VCAM-1 rising. The full reading is in the endothelial function analysis.
The study where the treated quantity was equalized
A matched cohort compared these drugs against mineralocorticoid receptor antagonists as fourth-line therapy for resistant hypertension, leaving 4,153 patients in each group followed a median of 1.4 years [3]. Systolic pressure fell 5.7 mmHg on the GLP-1 and 6.3 mmHg on the comparator at twelve weeks, with overlapping intervals.
Every other outcome differed sharply: major adverse cardiovascular events at a hazard ratio of 0.63, kidney events at 0.64, and all-cause mortality at 0.34. On the thing both drugs were prescribed to do, the groups were the same, so something other than the treated quantity is separating them. A 66% lower risk of death over 1.4 years is larger than any plausible pharmacological explanation, and what probably separates the groups is set out in the resistant hypertension cohort.
What the inflammation finding is good for
It establishes that the cardiovascular benefit is not purely a function of pounds lost, which matters to anyone weighing whether a smaller response on the scale means a smaller cardiovascular return. The comparison with the other intervention that produces large sustained weight loss is drawn in surgery against a GLP-1 on cardiovascular outcomes, and the older drug whose cardiovascular case rested on different reasoning is set out in where metformin ranks.
One limit deserves stating plainly. SELECT enrolled people with established atherosclerotic disease, overweight or obesity, and no diabetes, most of them already well treated with statins, so these numbers describe a specific and fairly narrow population. Whether they extend to people starting the drug for weight alone, with no cardiovascular diagnosis, is not something this substudy can answer. The gap between who gets studied and who gets prescribed is a running problem examined in uptake after a heart attack or stroke.
Nothing here is a reason to buy a drug for inflammation, and no seller on this roster offers an anti-inflammatory indication. What the pattern is useful for is reading a claim: ask which marker moved, what it was compared against, and whether anything happened to a person.