Why did SELECT exclude patients with diabetes?

Prior GLP-1 receptor agonist CVOTs (SUSTAIN-6, PIONEER 6, LEADER) had already demonstrated cardiovascular benefit in type 2 diabetes populations. SELECT was designed to test whether semaglutide provides cardiovascular protection specifically in patients with obesity and established CVD but without diabetes, a question that had not been answered.

What is a treatment-policy estimand?

A treatment-policy estimand counts all events in the randomized groups regardless of whether patients continued taking the study drug. It reflects the real-world effect of prescribing a medication, including the impact of non-adherence and discontinuation. SELECT used this as its primary analysis framework.

How long did the SELECT trial last?

The trial used an event-driven design targeting 1,225 primary endpoint events. Median follow-up was 39.8 months. Some patients were followed for over 5 years depending on enrollment timing.

Was SELECT stopped early?

No. Two interim analyses were pre-planned with O'Brien-Fleming boundaries, but the trial was not stopped early at either look. The full planned number of events was observed, which reduces the risk of effect-size inflation sometimes seen with early stopping.

How much weight did participants lose in SELECT?

Participants on semaglutide 2.4 mg lost a mean of 9.4% of body weight compared to 0.9% in the placebo group. Whether the cardiovascular benefit was driven by weight loss, direct vascular GLP-1 effects, or systemic metabolic improvements (reduced CRP, improved lipids) cannot be determined from this trial alone.

Did nausea from semaglutide affect blinding in SELECT?

Nausea occurred in approximately 20% of semaglutide patients versus 7% on placebo. The trial did not report a formal blinding-index assessment. GI side effects may have functionally unblinded some participants and clinicians, though the adjudication committee remained blinded.

Does SELECT support using semaglutide for primary cardiovascular prevention?

No. SELECT enrolled only patients with established atherosclerotic cardiovascular disease (prior MI, stroke, or peripheral artery disease). Extending the findings to patients with cardiovascular risk factors but no established disease requires a separate trial.

What is the difference between the 20% and 28% risk reductions reported?

The 20% reduction (HR 0.80) comes from the treatment-policy estimand, which includes all events regardless of drug adherence. The 28% reduction (HR 0.72) comes from the trial-product estimand, which censors events occurring after treatment discontinuation. The 20% figure is the regulatory and clinical standard; the 28% reflects the on-treatment effect.

How diverse was the SELECT trial population?

Approximately 72% of participants were White. Black, Asian, and Hispanic populations were underrepresented relative to the global burden of obesity and cardiovascular disease. This limits generalizability across racial and ethnic groups.

Inside the SELECT Methodology: What Most Summaries Skip

Q: What was the primary endpoint of the SELECT trial?

The primary endpoint was time to first major adverse cardiovascular event (MACE), defined as cardiovascular death, nonfatal myocardial infarction, or nonfatal stroke. An independent blinded committee adjudicated all events using source documentation, ECGs, and biomarkers.

By HealthRX Medical Team

Published May 25, 2026Updated May 25, 2026Last reviewed May 25, 2026

At a glance

| Parameter | Detail | |---|---| | N | 17,604 | | Intervention | Subcutaneous semaglutide 2.4 mg once weekly | | Comparator | Matched placebo injection | | Duration | Median 39.8 months (event-driven) | | Primary endpoint | Time to first MACE (CV death, nonfatal MI, nonfatal stroke) | | Key result | HR 0.80; 95% CI 0.72, 0.90; P <0.001 |

Why the Design Matters More Than the Headline

A 20% relative risk reduction is a clean number. It travels well in press releases. But the clinical weight of that number depends entirely on who was enrolled, how endpoints were counted, what happened when patients stopped taking the drug, and which statistical lens the investigators chose. SELECT's published methodology contains decisions that are worth examining one by one.

Population Selection: Obesity Plus Established CVD, Minus Diabetes

SELECT enrolled adults aged 45 or older with a BMI of 27 kg/m² or higher and pre-existing atherosclerotic cardiovascular disease. The critical exclusion: anyone with type 1 or type 2 diabetes at screening. This was deliberate. The SUSTAIN-6 and PIONEER 6 trials had already shown cardiovascular benefit for semaglutide in patients with type 2 diabetes. SELECT needed to isolate a different question: does the drug protect the heart in people whose primary indication is excess weight, not hyperglycemia?

The BMI floor of 27 (not 30) aligned with the FDA label for Wegovy, which permits use at BMI 27 or above when a weight-related comorbidity is present. This labeling decision and the trial's inclusion threshold were not coincidental. Novo Nordisk designed SELECT to support a cardiovascular indication for a drug already approved for chronic weight management.

Patients who developed diabetes during the trial were not withdrawn. They continued in their randomized group. This matters because roughly 2 to 3% of each arm did develop diabetes over the follow-up period, and their events still counted. The intention-to-treat frame kept them in.

Randomization and Blinding

Randomization was 1:1 to semaglutide 2.4 mg or placebo, stratified by region (to account for differences in background cardiovascular care). The dose escalation schedule mirrored the commercial Wegovy protocol: 0.25 mg weekly for 4 weeks, stepping up every 4 weeks through 0.5, 1.0, and 1.7 mg before reaching the target 2.4 mg dose at week 16.

Blinding was maintained with visually identical prefilled pens. But semaglutide's side effect profile creates a practical challenge. Nausea, which occurred in roughly 20% of the semaglutide group versus 7% of placebo patients, is a functional unblinding signal. The trial's primary publication does not report a formal blinding-index analysis. Whether GI side effects influenced patient or clinician behavior (exercise, diet changes, medication adherence) in ways that contributed to the cardiovascular benefit remains an open question.

The Primary Endpoint: Three-Point MACE

The composite primary endpoint was time to first occurrence of cardiovascular death, nonfatal myocardial infarction, or nonfatal stroke. This three-point MACE is the standard FDA-required composite for cardiovascular outcome trials (CVOTs), established by the 2008 FDA guidance for diabetes drugs and since adopted for obesity pharmacotherapy as well.

An independent Clinical Events Committee (CEC), blinded to treatment assignment, adjudicated every suspected event. This centralized adjudication reduces bias from site-level variability in event classification. The CEC reviewed source documents, ECGs, cardiac biomarkers, and imaging. Their adjudication rate for the primary endpoint was not reported as a standalone metric, but the use of an independent committee is standard in trials of this scale.

One subtlety: "nonfatal MI" included both spontaneous (Type 1) and procedure-related (Type 4a, 4b) infarctions. Because the semaglutide group lost more weight and might therefore have undergone fewer weight-related procedures, the inclusion of periprocedural MI in the composite is worth noting, though the numbers involved were small.

The HealthRX Estimand Interpretation Framework

Most trial summaries skip the estimand entirely. SELECT pre-specified two, and the difference between them changes what the result means in practice.

Treatment-policy estimand (primary). This counts all first MACE events regardless of whether the patient was still taking the study drug. If a patient randomized to semaglutide stopped injections at month 6 and had an MI at month 30, that MI counted against semaglutide. This is the intention-to-treat philosophy: it answers "what happens if we prescribe this drug," not "what happens while patients are actively taking it."

Trial-product estimand (supportive). This censored events that occurred after permanent treatment discontinuation plus a washout window. It approximates the on-treatment effect and answers a different question: "what happens while patients are actually on the drug?" The hazard ratio for this estimand was 0.72 (95% CI 0.64, 0.82), a 28% relative reduction compared to the 20% in the primary analysis.

The gap between 20% and 28% is not trivial. It reflects the roughly 17% of semaglutide patients who discontinued treatment (versus 16% on placebo), many due to gastrointestinal side effects. The treatment-policy estimand is more conservative and more generalizable to real-world prescribing, where adherence is imperfect. The trial-product estimand hints at a larger benefit for patients who tolerate the drug long-term.

| Estimand | HR | 95% CI | Interpretation | |---|---|---|---| | Treatment-policy (primary) | 0.80 | 0.72, 0.90 | Effect of prescribing semaglutide | | Trial-product (supportive) | 0.72 | 0.64, 0.82 | Effect while on semaglutide |

Clinicians deciding whether to start a patient on semaglutide for cardiovascular protection should anchor on the 20% figure. Clinicians counseling a patient who tolerates the drug well and asks "how much is this helping me?" can reasonably reference the 28% figure, with the caveat that it comes from a supportive analysis.

Event-Driven Design and Follow-Up Duration

SELECT was event-driven, meaning it was designed to continue until a pre-specified number of primary endpoint events (1,225) had been observed. The actual number of confirmed first MACE events was 1,270. The median follow-up was 39.8 months, with a maximum exceeding 5 years for the earliest enrollees.

This design has a specific statistical advantage: it ensures adequate power regardless of fluctuations in the event rate. But it also means the follow-up duration was not fixed. Some patients contributed 5 years of data. Others contributed 2. The Kaplan-Meier curves in the primary publication show separation beginning at approximately 12 months and widening progressively, which suggests the benefit is not a one-time early effect but accumulates with continued exposure.

Two interim analyses were planned, with O'Brien-Fleming spending function boundaries to control the overall type I error rate at 5%. The trial was not stopped early for efficacy at either interim look, which means the full planned information fraction was observed. This avoids the overestimation of treatment effect that sometimes occurs with early stopping.

The Comparator Choice: Placebo, Not Active Treatment

SELECT compared semaglutide to placebo, not to an active weight-loss intervention or another cardiovascular drug. Both arms received standard-of-care cardiovascular treatment. The 2023 AHA/ACC guidelines for chronic coronary disease were contemporary with the trial, and baseline medication use reflected high-quality secondary prevention: over 90% on statins, approximately 85% on antiplatelet agents, and roughly 80% on antihypertensives.

The placebo comparison means SELECT answers the question "does adding semaglutide to optimized background therapy reduce MACE?" It does not answer whether semaglutide is superior to bariatric surgery, intensive lifestyle intervention, or other GLP-1 receptor agonists like tirzepatide or liraglutide for cardiovascular protection in this specific non-diabetic population.

Weight Loss: Secondary Outcome, Primary Confounder?

The semaglutide group lost a mean of 9.4% of baseline body weight versus 0.9% in the placebo group. This difference raises a question the trial cannot fully resolve: is the cardiovascular benefit driven by weight loss per se, by direct vascular effects of GLP-1 receptor agonism, or by metabolic improvements (reduced inflammation, lower hsCRP, improved lipid profile) that accompany both?

The investigators reported significant reductions in C-reactive protein (37.8% decrease), waist circumference, and systolic blood pressure in the semaglutide arm. But because these changes co-occur, mediation analysis cannot cleanly attribute the MACE reduction to any single pathway. Prior cardiovascular outcome trials with other weight-loss drugs (rimonabant, sibutramine) did not show benefit and in some cases showed harm, which argues against weight loss alone as the mechanism.

Limitations the Authors Acknowledged

The trial publication lists several limitations directly:

Generalizability. The population had established CVD. Patients with obesity but only cardiovascular risk factors (primary prevention) were excluded. Extrapolating the result to a primary-prevention population requires a separate trial.
Duration. While nearly 40 months of median follow-up is substantial, the durability of benefit beyond 5 years and the effect of drug discontinuation remain unknown.
Diabetes exclusion. Because diabetic patients were excluded, the result applies to a specific subgroup of the broader population eligible for semaglutide.
Diversity. Approximately 72% of enrolled participants were White. Representation of Black, Asian, and Hispanic populations was limited relative to the global burden of obesity and cardiovascular disease.
No active comparator. As noted above, the placebo comparison does not position semaglutide against bariatric surgery or other pharmacotherapies.

What This Means for Practice

The FDA added a cardiovascular indication to the Wegovy label in March 2024, making semaglutide 2.4 mg the first anti-obesity medication approved specifically to reduce cardiovascular risk. This label change was based directly on SELECT. But the label specifies the benefit was shown in patients with established CVD, not in primary prevention, and not in patients with diabetes (who are covered by separate CVOT data from SUSTAIN-6).

For prescribers, the methodology has practical implications. The treatment-policy estimand means the 20% benefit estimate already accounts for real-world non-adherence rates. The event-driven design and lack of early stopping mean the effect size is unlikely to be inflated. And the exclusion of diabetes means a clinician prescribing semaglutide to an obese patient with coronary artery disease and type 2 diabetes is extrapolating beyond the trial's direct evidence, even though other data (SUSTAIN-6) supports cardiovascular benefit in that population.

Frequently asked questions

›

References

Lincoff AM, Brown-Frandsen K, Colhoun HM, et al. Semaglutide and cardiovascular outcomes in obesity without diabetes. N Engl J Med. 2023;389(24):2221-2232. PubMed
FDA. Wegovy (semaglutide) prescribing information. Revised 2024. FDA Label
Virani SS, Newby LK, Arnold SV, et al. 2023 AHA/ACC/ACCP/ASPC/NLA/PCNA guideline for the management of patients with chronic coronary disease. Circulation. 2023;148(24):e218-e318. PubMed
Marso SP, Bain SC, Consoli A, et al. Semaglutide and cardiovascular outcomes in patients with type 2 diabetes. N Engl J Med. 2016;375(19):1834-1844. PubMed
FDA. Guidance for industry: diabetes mellitus, evaluating cardiovascular risk in new antidiabetic therapies to treat type 2 diabetes. 2008. FDA Guidance