A cardiovascular outcome trial is the most expensive sentence a drug can earn, and SELECT is the one that moved semaglutide out of the weight-management aisle and into a cardiology conversation. It ran for roughly four years, randomized 17,604 people, and produced a single relative figure that now appears on nearly every page selling the molecule, including pages for the compounded vials listed on the semaglutide boards. The trial is narrower than that figure suggests, and its most striking result is one no advertisement quotes: how much weight a participant lost did not predict whether the drug protected them.
What the trial was built to do
SELECT was a multicenter, double-blind, randomized, placebo-controlled, event-driven superiority trial. Eligibility required an age of 45 or older, preexisting cardiovascular disease and a body-mass index of 27 or greater, with any history of diabetes excluded. Assignment was 1:1, with 8,803 patients allocated to once-weekly subcutaneous semaglutide 2.4 mg and 8,801 to placebo, the active dose reached by escalation over 16 weeks.[1][5]
The primary cardiovascular endpoint was a composite of death from cardiovascular causes, nonfatal myocardial infarction and nonfatal stroke, counted as time to first event. Two durations are reported and they are not the same number: mean exposure to trial product was 34.2 ± 13.7 months, while mean follow-up was 39.8 ± 9.4 months.[1] The five-month gap between them is the treatment people stopped taking while remaining under observation, and it is the reason the result is an intention-to-treat statement about starting the drug rather than about staying on it — a distinction that matters to anyone reading the withdrawal evidence.
Who was actually enrolled
The baseline-characteristics paper describes a population that looks almost nothing like the typical cash-pay telehealth patient. Enrollment was 72.5% male, with a mean age of 61.6 years (SD 8.9) and a mean body-mass index of 33.34 (SD 5.04). The most common prior cardiovascular event was a myocardial infarction, recorded in 76.3% of participants, followed by stroke in 23.3% and peripheral artery disease in 8.6%; 24.3% carried a heart failure diagnosis. Two-thirds of participants, 66%, had an HbA1c in the prediabetes range of 5.7% to 6.4%.[2]
That is a secondary-prevention cohort: older, mostly male, and already injured. A 41-year-old woman with a body-mass index of 31 and no cardiac history was not eligible for SELECT, and the trial's hazard ratio makes no claim about her. The distance between those two people is the single most important thing to carry away from this page, and it is larger than the distance between any two molecules in the class.
The headline figure, converted into people
A primary endpoint event occurred in 569 of 8,803 patients on semaglutide (6.5%) and in 701 of 8,801 on placebo (8.0%), giving a hazard ratio of 0.80 (95% CI, 0.72 to 0.90; P < 0.001).[1]
Subtracting those two rates gives an absolute risk difference of 1.5 percentage points across a mean 39.8 months. The reciprocal of that difference is a number needed to treat of approximately 67 for that duration — that is, roughly 67 people with established cardiovascular disease took a weekly injection for over three years for one of them to avoid a first heart attack, stroke or cardiovascular death. For secondary prevention that is a real and useful result. It is also a quieter claim than “20% fewer events,” and both descriptions come from the same two counts.
Tolerability set a limit on how many people got to find out. Adverse events led to permanent discontinuation of the trial product in 16.6% of the semaglutide group against 8.2% on placebo.[1] The separate long-term analysis adds a detail with a practical edge: discontinuations rose as baseline body-mass index class fell, so the participants with the least excess weight were the most likely to leave.[3]
How much weight the trial produced
Weight loss continued over 65 weeks and was then sustained for up to four years. At 208 weeks, semaglutide was associated with a mean reduction in weight of −10.2% against −1.5% on placebo, a waist circumference change of −7.7 cm against −1.3 cm, and a waist-to-height ratio change of −6.9% against −1.0%, all at P < 0.0001.[3]
Ten percent over four years is materially less than the −14.9% recorded at 68 weeks in the pivotal obesity trial of the same dose,[6] which is worth holding next to the weight-loss figures before assuming a cardiology cohort behaves like a weight-management one. Serious adverse events were nonetheless less frequent on semaglutide in every body-mass index band, at 43.23, 43.54, 51.07 and 47.06 events per 100 years of observation against 50.48, 49.66, 52.73 and 60.85 on placebo.[3]
Did the protection follow the weight?
A prespecified analysis published in 2025 asked that question directly, examining baseline adiposity, treatment-induced change in adiposity and subsequent event risk. The benefit was consistent across every baseline weight and waist circumference category. Then the landmark analysis: events occurring after week 20 were compared across participants grouped by how much adiposity they had shed in the first 20 weeks, and in the semaglutide group no linear trend linked week-20 weight loss to subsequent event risk.[4]
Waist circumference behaved slightly differently. A greater waist reduction by week 20 was associated with lower subsequent risk, and waist reduction by week 104 with lower in-trial risk. Formal mediation put a number on it: an estimated 33% of the observed benefit ran through waist reduction, leaving a hazard ratio of 0.86 (95% CI, 0.77 to 0.97) after adjustment for time-varying waist changes.[4]Two-thirds of the effect is therefore unexplained by the thing the drug is sold for. The authors' own conclusion is that the cardioprotective effect was independent of baseline adiposity and weight loss, with mechanisms beyond adiposity reduction.
The placebo arm points the other way
The comparison that complicates every simple reading sits in the control group. Among participants receiving semaglutide, a lower baseline bodyweight was associated with lower event risk, at an average 4% reduction per 5 kg (hazard ratio 0.96; 95% CI, 0.94 to 0.99; P = 0.001) and per 5 cm of waist (0.96; 95% CI, 0.93 to 0.99; P = 0.004). Among participants receiving placebo, lower baseline waist circumference predicted lower risk (0.96; 95% CI, 0.94 to 0.99; P = 0.007) but baseline bodyweight did not (0.99; 95% CI, 0.97 to 1.01; P = 0.28) — and in that arm weight loss was associated with increased event risk.[4]
Unintended weight loss in older people with vascular disease is a marker of illness, not an achievement, and the same scale reading means opposite things depending on what produced it. A page that promises a healthier heart as a consequence of the pounds lost has the arrow pointing backward: in SELECT the drug protected people whether or not the scale moved much.
What “early” can and cannot be claimed
A second claim travels alongside the first — that the curves separated before any meaningful weight was lost — and the analyses above do not establish it. A landmark analysis starting at week 20 describes what happened after week 20; it does not date the point at which the event curves parted, and neither the primary report nor the prespecified adiposity analysis states such a date in what it publishes.[1][4]
So the defensible version is the narrower one. That the benefit was not proportional to weight lost is a prespecified, quantified finding. That it arrived before the weight did is a plausible inference from the same data that nothing cited here measures. The two are often quoted as a single sentence, and only the first half carries a confidence interval.
Where the effect did and did not vary
A prespecified heart failure analysis split the cohort by history at enrollment. Of 17,604 patients, 4,286 (24.3%) had investigator-defined heart failure — 2,273 with preserved ejection fraction, 1,347 with reduced ejection fraction and 666 unclassified. In that group semaglutide produced a MACE hazard ratio of 0.72 (95% CI, 0.60 to 0.87), a heart failure composite of 0.79 (95% CI, 0.64 to 0.98), cardiovascular death of 0.76 (95% CI, 0.59 to 0.97) and all-cause death of 0.81 (95% CI, 0.66 to 1.00), with every interaction P value above 0.19 against those without heart failure.[5] The benefit held in both the reduced (0.65; 95% CI, 0.49 to 0.87) and preserved (0.69; 95% CI, 0.51 to 0.91) ejection fraction subgroups.
The same trial's kidney endpoints, and the much larger kidney trial run in diabetes, are set out in the FLOW article.
What SELECT does not decide
It does not establish primary prevention. Every participant had documented cardiovascular disease, so the result describes lowering risk in people already carrying an event, not conferring protection on someone whose only finding is a body-mass index. It does not speak to people with type 2 diabetes, who were excluded by design and are covered by a different set of trials.
It also studied branded, FDA-approved product at a labeled dose under trial supervision. Compounded semaglutide is not FDA-approved, and compounded drugs are not reviewed by the FDA for safety, efficacy or quality before they are dispensed — the difference is the subject of the compounded-versus-brand article, and a hazard ratio earned across 17,604 supervised patients does not transfer to a vial by sharing a molecule name. How each figure on this page was established is described in the methodology.