The telos puts pharmacotherapy in scope because a drug is often the realistic alternative to a lifestyle change — and for the biggest lever in an obese stratum (weight), semaglutide is now the best-evidenced drug route. The SELECT trial (Lincoff, NEJM 2023) is the landmark: the first demonstration that treating obesity pharmacologically reduces hard cardiovascular events. But it proved this in one population only, and the population is not the one most of the wiki’s low-risk strata belong to. Both halves of that sentence are the finding.
What SELECT proved, and the population it proved it in
A single large event-driven RCT (n=17,604), on top of standard secondary-prevention care (90% on statins, 86% on antiplatelets):
primary MACE (CV death, non-fatal MI, non-fatal stroke): «569 of the 8803 patients (6.5%)» on semaglutide vs «701 of the 8801 patients (8.0%)» on placebo — «hazard ratio, 0.80; 95% confidence interval, 0.72 to 0.90; P<0.001», a reduction «by 20%» over a mean 39.8 months.
Absolute risk reduction is 1.5 percentage points over ~3.3 years -> NNT ~67 [INFERRED (Lincoff - Semaglutide and Cardiovascular Outcomes SELECT 2023) — arithmetic on the two event rates; the trial states no NNT]. That is a real, incremental benefit on top of guideline therapy — not instead of it.
Who was enrolled fixes what the number means: «patients 45 years of age or older who had preexisting cardiovascular disease and a body-mass index… of 27 or greater but no history of diabetes». This is a secondary-prevention population (prior MI, stroke, or symptomatic PAD), non-diabetic, with a placebo event rate of 8% over 3.3 years (~2.4%/yr).
The primary-prevention gap — stated by the trialists themselves
This is the load-bearing limit, and it is not the wiki’s inference — it is the paper’s own:
«An important limitation of this trial is that we included only patients with preexisting cardiovascular disease. The effects of semaglutide on primary prevention of cardiovascular events in persons with overweight or obesity but without previous atherosclerotic disease were not studied.»
So for an obese person without established atherosclerotic disease — low 10-year risk, or a zero coronary-calcium score — the hard-outcome benefit of semaglutide is not proven. This is the insufficient-evidence state, not the no-effect state (the two are kept distinct): SELECT gives no reason to think the drug stops working in primary prevention, only that it was never tested there.
The two trials bracket the hole precisely. STEP-1 is a primary-prevention population (no CVD required) — but its endpoint is weight and cardiometabolic surrogates, not events. SELECT is a hard-outcome trial — but only in secondary prevention. So: primary prevention has strong surrogate evidence and zero hard-outcome evidence; secondary prevention has the hard-outcome evidence. No trial delivers a hard-outcome benefit in primary prevention — the gap is not for want of looking at the drug, it is that the one relevant question has never been the endpoint of a trial in this population.
Two reasons the benefit would be smaller there even if the relative effect transported — see Baseline Risk and the Relative-Absolute Split and SCORE2 Baseline Risk and the ESC Treatment Thresholds:
- Absolute benefit scales with baseline risk. A primary-prevention person at, say, 5% 10-year risk (~0.5%/yr) sits roughly 5x below SELECT’s placebo event rate. Apply the same HR 0.80 and the absolute reduction shrinks ~5-fold — from ~1.5 pp/3.3 yr toward a fraction of a percentage point. The relative effect is the fragile assumption; the absolute one is small by construction.
- The relative effect itself may not transport. Mechanism (plaque already present vs not) differs between secondary and primary prevention; SELECT cannot speak to it.
Confirmed vs widely-repeated — the hierarchical-testing catch
Only the primary composite is a confirmatory result. The pre-specified gatekeeping hierarchy tested CV death next — «hazard ratio, 0.85; 95% CI, 0.71 to 1.01; P = 0.07» — which failed the gate, so «superiority testing was not performed for the remaining confirmatory secondary end points». Everything downstream is a point estimate, not a confirmed effect:
- All-cause mortality «0.81 (0.71 to 0.93)» and heart-failure composite «0.82 (0.71 to 0.96)» — directionally favourable, statistically non-confirmatory. Semaglutide reduces mortality is a common but unlicensed reading of this trial.
- Non-fatal stroke «0.93 (0.74 to 1.15)» — not reduced. (An aggregator SR of this literature was seen reporting stroke as the strongest effect; reading the trial contradicts it.)
- Non-fatal MI «0.72 (0.61 to 0.85)» and coronary revascularization «0.77 (0.68 to 0.87)» carry most of the composite.
Weight loss: the drug route to the biggest lever
body weight «-9.39» % vs «-0.88» % (difference «-8.51 (-8.75 to -8.27)» percentage points at week 104); waist circumference difference «-6.53» cm.
~9% weight loss is the mechanism-adjacent reason this matters for Fatty Liver MASLD and Weight Loss and for Layer 1 - Ranking Interventions for a Stratum: it is a pharmacological route to the same weight lever a lifestyle programme targets. The trialists note that prior non-surgical approaches reached weight loss «substantially lower than the mean 9.4% decrease observed with semaglutide» and had «uniformly failed» to move CV outcomes — so the drug is not just another weight intervention, it is the first to clear the bar. The drug and the lifestyle change are not either/or — same lever, different route, combinable.
STEP-1: the primary-prevention arm — a bigger number, on a surrogate
SELECT’s 9.4% was measured in an older, established-CVD population. STEP-1 (Wilding, NEJM 2021) ran the same drug in the population the wiki’s obese strata actually resemble — «a body-mass index… of 30 or greater (≥27 in persons with ≥1 weight-related coexisting condition), who did not have diabetes», mean age 46, ~60% at BMI ≥35, only 2.5% with coronary artery disease — i.e. essentially primary prevention, as an adjunct to lifestyle (a 500-kcal deficit plus 150 min/week of activity). Its coprimary endpoint is weight, a surrogate, not events.
mean body-weight change to week 68: «−14.9%» vs «−2.4%», difference «−12.4 percentage points» (95% CI −13.4 to −11.5); responders losing ≥10%: «69.1%» vs «12.0%»; ≥15%: «50.5%» vs «4.9%»; ≥20%: «32.0%» vs «1.7%».
Half the treated group lost ≥15% and a third lost ≥20% — the latter «approaching that reported 1 to 3 years after bariatric surgery». Waist fell «–9.42 cm» placebo-adjusted, SBP «–5.10» mmHg, CRP more than halved, and fat mass fell more than lean mass.
Do NOT read the 14.9% (STEP-1) vs 9.4% (SELECT) gap as the drug works better in primary prevention. The two trials differ on exactly the axes that move weight response, so the comparison is confounded:
| Parameter | SELECT (Lincoff 2023) | STEP-1 (Wilding 2021) | Same quantity? |
|---|---|---|---|
| Mean % body-weight change (treatment-policy, ITT) | «−9.39%» | «−14.85%» | Yes — same construct |
| Follow-up at measurement | week 104 | week 68 | No — different window |
| Population | established CVD, age 61.6, 72% male | no CVD required, age 46, 74% female | No — older/sicker vs younger |
The honest read is ~15% in a primary-prevention obese adult, ~9% in an older secondary-prevention one — a population-and-window difference, not evidence of a prevention-setting effect on the drug.
Benefit separated «early after the initiation of treatment», before much weight was lost, so «more rapid treatment-induced physiological changes beyond the magnitude of body-weight loss may have mediated at least part of the cardiovascular benefit» — SBP fell «-3.31» mmHg and hsCRP «-37.82» % on top of high statin use. Mechanism is explicitly speculative in the source; do not treat any one surrogate as the transmission channel.
The adherence cost is part of the effect
An intervention not taken has no effect, and semaglutide has a real tolerability tax:
AE-driven discontinuation «1461 patients (16.6%)» vs «718 patients (8.2%)» (P<0.001), predominantly gastrointestinal («880 patients (10.0%)» vs «172 patients (2.0%)»).
Counter-weight: serious adverse events were lower on semaglutide («33.4%» vs «36.4%»). And the weight benefit reverses on stopping — so this is a sustained commitment, the same return-on-investment framing Lifetime Benefit - The Frame for Younger Adults attaches to lifelong preventive therapy: a longer horizon is a longer treatment duration, not only a larger gain.
Diabetes prevention — a secondary signal that fits the prediabetic stratum
Two-thirds of SELECT was prediabetic. Progression to diabetes was cut sharply: progression to HbA1c >=6.5% «0.27 (0.24 to 0.31)». STEP-1 shows the same in its younger primary-prevention population: among prediabetics, «84.1%» on semaglutide reverted to normoglycemia by week 68 vs «47.8%» on placebo. For an obese, impaired-fasting-glucose person this is a distinct, plausibly-relevant benefit — though on a surrogate (glycemic threshold), not a patient-important outcome, and whether it lowers hard diabetes-related outcomes in primary prevention inherits the same caveat.
Provenance and limits
- Industry origin: both trials were «Funded by Novo Nordisk» and sponsor-designed («The sponsor (Novo Nordisk) designed the trial and oversaw its conduct» — STEP-1). A COI flag on both, not a refutation — but the symmetric-standards rule applies, and two sponsor-run trials are not two independent lines of evidence.
- Two trials, one drug, one sponsor. SELECT + STEP-1 are both semaglutide 2.4 mg. Tirzepatide (dual agonist, larger weight loss) has no dedicated CV-outcome trial; oral formulations are not here. Do not generalize GLP-1 class reduces CV events in obesity from one agent.
- Net-safety direction is population-dependent, not a contradiction. Serious AEs were lower on semaglutide in SELECT («33.4%» vs «36.4%») but higher in STEP-1 («9.8%» vs «6.4%»). Not a tension: SELECT’s established-CVD patients carry a high background serious-event rate the drug’s CV benefit offsets, whereas STEP-1’s healthy population has little to offset, so the drug’s own GI/hepatobiliary serious events dominate. The GI tolerability tax is consistent across both (discontinuation «7.0%» vs «3.1%» in STEP-1).
- Narrow demographics: SELECT 27.7% women / 3.8% Black; STEP-1 «preponderance of women and White participants» and «relatively short duration».
- Cross-population comparison is not confirmation: the T2D GLP-1 meta-analysis the discussion cites (Sattar 2021, «0.86; 95% CI, 0.80 to 0.93») is a different population reaching a similar HR — an echo, not independent backing of the obesity result (and Sattar is not held here).
Decision relevance
- Obese + established CVD + non-diabetic: semaglutide is a strong incremental option — NNT ~67 for MACE over ~3.3 years, on top of statins/antiplatelets, with a net-favourable serious-AE profile.
- Obese + primary prevention (low 10-year risk, CAC=0): the hard-outcome benefit is unproven and would be small in absolute terms even if it transported. What is well-evidenced there (STEP-1) is large and reliable: ~15% weight loss (half lose ≥15%), waist/BP/CRP improvement, and diabetes prevention (84% revert from prediabetes) — all real, all surrogate. Rank the drug for such a person on the weight lever and diabetes prevention, not on a promised CV-event reduction.
- Against lifestyle: not a substitution question with a fixed answer — same lever, and the swing factors are adherence, cost, GI tolerability, and reversibility on stopping, all elicited at layer 3.
Gaps this opens
- No primary-prevention CV-outcome trial for any obesity drug — the single largest hole for applying this to a low-risk stratum. G.
- No head-to-head of drug vs sustained lifestyle vs bariatric surgery on hard outcomes.
- Weight loss on hard outcomes as an exposure in its own right (independent of the drug) is still unheld -> AWAITS Weight Loss and Mortality.