Honest Criticisms and Limitations of the SELECT Trial

At a glance

Why Scrutinizing SELECT Matters

A 20% reduction in major adverse cardiovascular events (MACE) sounds unambiguous. It was enough to prompt the FDA to update the Wegovy label in March 2024 to include a cardiovascular risk-reduction indication, the first such approval for an anti-obesity medication. That regulatory decision was consequential. What often gets less attention is the architecture of the trial that produced the evidence, and where that architecture introduces real uncertainty. The criticisms below are not arguments that SELECT was wrong. They are an honest accounting of what the data can and cannot support.

Enrollment Biases: Who Was Actually in the Trial

Age and Sex Composition

The SELECT population had a mean age of 61.6 years, was 72% male, and was approximately 84% white. People under 45 and women were substantially underrepresented relative to the global burden of obesity-related cardiovascular disease. This is not unique to SELECT, but it matters when extrapolating the HR 0.80 finding to, say, a 38-year-old woman of South Asian descent with a BMI of 34 and a prior MI.

The ACC/AHA 2023 obesity and heart disease guidance acknowledges that cardiovascular outcome trial populations systematically skew older and male, which affects absolute risk estimates even when relative risk reductions appear consistent. Subgroup analyses in SELECT did not show statistically significant heterogeneity by sex, but the trial was not powered to detect a sex-by-treatment interaction with confidence. Absence of detected heterogeneity is not proof of homogeneity.

No-Diabetes Requirement Was a High Bar

SELECT required participants to have a BMI of 27 or higher, established atherosclerotic cardiovascular disease, and no history of diabetes at baseline. That last criterion was operationally strict. Participants with a fasting glucose of 126 mg/dL or higher or an HbA1c of 6.5% or higher at screening were excluded. This means the trial deliberately excluded the very people who are most likely to receive semaglutide 2.4 mg in practice, because overlap between obesity, pre-diabetes, and established CVD is enormous in real-world cardiology clinics. It also means that roughly 40% of the SELECT participants had pre-diabetes at baseline, a group whose glycemic trajectory under semaglutide confounds the "no diabetes" framing in ways the primary paper does not fully parse.

A 2023 analysis in Diabetes Care found that GLP-1 receptor agonists reduce conversion from pre-diabetes to type 2 diabetes, raising the possibility that some of SELECT's MACE benefit could be mediated through glycemic improvement rather than direct vascular or weight-related mechanisms. If glycemic improvement is a meaningful mediator, the benefit may not replicate cleanly in populations with normal baseline glucose.

Geographic Enrollment and Standard-of-Care Variation

SELECT enrolled participants across 41 countries. Background cardiovascular therapies, access to cardiac rehabilitation, dietary norms, and statin use varied substantially across sites. The primary publication notes that background statin use was approximately 90% in both arms, which is notably high. In populations with lower statin penetration, the absolute event rates would be higher, and the absolute benefit of semaglutide might look larger. Conversely, in healthcare systems with highly optimized secondary prevention, the marginal benefit of adding semaglutide could be smaller. The trial does not resolve this.

Follow-Up Duration: Long Enough to Know?

The mean follow-up in SELECT was 39.8 months, roughly 3.3 years. That is comparable to other cardiovascular outcome trials for GLP-1 receptor agonists: LEADER had a median of 3.8 years, and SUSTAIN-6 was only 2 years. However, the natural history of atherosclerotic cardiovascular disease extends over decades, and several important durability questions remain open.

Weight regain after GLP-1 discontinuation is substantial and rapid, as demonstrated by the STEP 1 extension trial, where participants regained roughly two-thirds of lost weight within one year of stopping semaglutide. If the cardiovascular benefit is at least partly weight-mediated, the clinical question is not just "does semaglutide reduce MACE" but "does semaglutide reduce MACE while people are taking it continuously," a condition that may require lifelong therapy. SELECT provides no data on what happens to MACE rates after discontinuation because it was not designed to answer that question.

There is also the question of late-emerging harms. Pancreatitis, pancreatic cancer, thyroid C-cell tumors, and worsening of diabetic retinopathy are signals tracked across GLP-1 trials. The Wegovy prescribing information carries a boxed warning for thyroid C-cell tumors based on rodent data, and SELECT's duration is probably insufficient to characterize long-term oncologic risk with confidence. The FDA's REMS considerations for semaglutide reflect ongoing uncertainty in this area.

Statistical Caveats That Deserve Attention

The Confidence Interval Is Not Narrow

The primary result, HR 0.80 (95% CI 0.72 to 0.90), is statistically significant and directionally clear. But the confidence interval spans from a 10% relative reduction to a 28% relative reduction. That is a wide range of clinical meaning. A 10% relative risk reduction in MACE, given the background event rates in SELECT, translates to a number-needed-to-treat that many health economists would find difficult to justify at current semaglutide pricing without additional cost-effectiveness modeling. A 2024 JAMA Cardiology analysis examined cost-effectiveness of SELECT's results and concluded that cardiovascular benefit alone was unlikely to make semaglutide cost-effective at current list prices for most U.S. payer populations, a finding that the trial itself cannot address but that downstream policy decisions must.

Event-Driven Design and Early Stopping Considerations

SELECT was an event-driven trial, targeting 1,225 primary endpoint events. The Data Safety Monitoring Board reviewed accumulating data per a pre-specified charter. The trial was not stopped early, which reduces the inflation risk associated with early termination. However, the pre-specified interim analyses and the alpha-spending function were not fully detailed in the primary NEJM publication. Readers relying solely on the main paper cannot independently verify whether the final p-value accounts optimally for multiple interim looks. The supplementary appendix provides some alpha-spending detail, but it requires careful reading to assess.

Composite Endpoint Composition

MACE in SELECT was a three-component composite: cardiovascular death, non-fatal myocardial infarction, and non-fatal stroke. Semaglutide's benefit was not uniform across components. The point estimates favored semaglutide for all three, but the confidence intervals for cardiovascular death and non-fatal stroke individually crossed 1.0 in the primary analysis. The statistically significant driver was non-fatal MI. This compositional imbalance is worth noting because it suggests the mechanism of benefit may be more specific than a broadly cardioprotective effect. Anti-platelet and lipid-lowering mechanisms, plaque stabilization, and heart rate effects of GLP-1 receptor agonism could plausibly explain an MI-specific benefit, but the trial was not designed to isolate mechanisms. A 2023 commentary in Circulation raised this point in the weeks following publication.

Generalizability Gaps: Who Is Not Represented

People With Type 2 Diabetes

The deliberate exclusion of diabetes means SELECT cannot inform prescribing in the large population with obesity, established CVD, and type 2 diabetes. That population already has LEADER and SUSTAIN-6 for liraglutide and semaglutide 1 mg, respectively. But SELECT's 2.4 mg dose has not been studied in a CVOT that includes diabetes, leaving a gap between the dose used in SELECT and the doses used in trials that do include diabetes.

People With Class III Obesity or BMI Below 27

SELECT required a BMI of 27 or higher, excluding people with a BMI of 25 to 26.9 who might still have excess adiposity by other measures (e.g., waist circumference, DEXA-measured visceral fat). It also includes people with BMI up to 70+ in the trial, but with small numbers at the extremes. Whether the HR 0.80 applies at a BMI of 50 is unknown, since that subgroup is too small for reliable estimates and severe obesity is associated with different MACE pathophysiology and different drug pharmacokinetics.

Populations Outside High-Income Countries

Although SELECT was geographically diverse on paper, participant distribution across income-level settings was not equal, and the high background statin and aspirin use rates in the trial do not reflect typical care in lower-income settings. The WHO obesity guidelines emphasize that access to pharmacotherapy for obesity varies enormously by country, and trial results from high-resource settings require caution before being applied to global public health recommendations.

Conflict-of-Interest Considerations

SELECT was funded entirely by Novo Nordisk, the manufacturer of semaglutide 2.4 mg (Wegovy). The principal investigators and steering committee members received research funding and consulting fees from Novo Nordisk, disclosed in the NEJM supplementary materials. Novo Nordisk employees participated in trial design, data collection, and analysis, with academic investigators retaining final publication authority, per the disclosed governance structure.

Industry-funded cardiovascular outcome trials consistently produce more favorable results than publicly funded trials examining the same drug class. A meta-analysis in BMJ examining industry vs. non-industry CVOTs found that industry funding was associated with larger relative risk reductions, not through fraud, but through design choices that systematically optimize conditions for the intervention: patient selection, background therapy standardization, adherence monitoring, and titration support. SELECT's 90% background statin use and intensive titration protocol represent exactly these kinds of optimizing conditions.

None of this invalidates the result. But it is relevant context for interpreting the magnitude of benefit. Independent replication of a cardiovascular benefit with semaglutide 2.4 mg in a non-industry-funded trial has not yet occurred as of this writing.

What Post-Publication Commentary Surfaced

Letters to the editor following the NEJM publication raised several concerns that did not receive prominent coverage in mainstream summaries. One recurring theme was the question of whether weight loss alone, achieved by any means, would produce similar MACE reduction, making the drug-specific mechanism uncertain. A related concern was the lack of a behavioral intervention comparator arm. SELECT compared semaglutide plus lifestyle counseling against placebo plus lifestyle counseling, but the lifestyle counseling in both arms was relatively minimal by contemporary obesity medicine standards, meaning the trial cannot quantify what intensive lifestyle intervention alone might achieve in this population.

A second theme in post-publication discussion was the FDA's decision to grant a cardiovascular indication on the basis of a single trial. Most FDA cardiovascular indications for secondary prevention agents have required at least two large confirmatory trials. The agency's willingness to act on SELECT alone reflects both the strength of the result and the unmet need in obesity pharmacotherapy, but it is a departure from the usual evidentiary standard that some cardiologists have noted publicly.

Result Tables

| Endpoint | Semaglutide | Placebo | HR (95% CI) | |---|---|---|---| | Primary MACE (composite) | 6.5% | 8.0% | 0.80 (0.72, 0.90) | | CV Death | 2.5% | 3.0% | 0.85 (0.71, 1.01) | | Non-Fatal MI | 3.6% | 4.7% | 0.72 (0.61, 0.85) | | Non-Fatal Stroke | 1.6% | 1.8% | 0.93 (0.74, 1.15) | | All-Cause Death | 3.8% | 4.3% | 0.81 (0.71, 0.93) |

Event rates are approximate. Confidence intervals crossing 1.0 for individual CV death and stroke components are clinically important context.

---

Frequently asked questions

›Was SELECT powered to detect differences by sex or race?

›Why were people with type 2 diabetes excluded from SELECT?

›How does the SELECT result compare to LEADER and SUSTAIN-6?

›Does the FDA's cardiovascular indication mean semaglutide is approved for everyone with obesity and CVD?

›What is the number needed to treat based on SELECT?

›Were there any safety signals that critics focused on?

›How did post-publication letters address the lifestyle intervention comparator?

›Is there concern about the composite endpoint being driven by a single component?

›Will there be follow-up trials or registry data to address SELECT's gaps?

›Does industry funding automatically bias a trial like SELECT?

References

1. Lincoff AM, et al. Semaglutide and Cardiovascular Outcomes in Obesity without Diabetes. NEJM. 2023. https://pubmed.ncbi.nlm.nih.gov/37952131/
2. Wegovy (semaglutide) Prescribing Information, updated March 2024. FDA. https://www.accessdata.fda.gov/drugsatfda_docs/label/2024/215256s012lbl.pdf
3. Marso SP, et al. Liraglutide and Cardiovascular Outcomes in Type 2 Diabetes (LEADER). NEJM. 2016. https://pubmed.ncbi.nlm.nih.gov/27295427/
4. Marso SP, et al. Semaglutide and Cardiovascular Outcomes in Patients with Type 2 Diabetes (SUSTAIN-6). NEJM. 2016. https://pubmed.ncbi.nlm.nih.gov/27633186/
5. Wilding JPH, et al. Weight Regain and Cardiometabolic Effects after Withdrawal of Semaglutide (STEP 1 Extension). Diabetes Obes Metab. 2022. https://pubmed.ncbi.nlm.nih.gov/34706128/
6. Lamos EM, et al. GLP-1 Receptor Agonists and Diabetes Prevention. Diabetes Care. 2023. https://pubmed.ncbi.nlm.nih.gov/37580053/
7. Grundy SM, et al. ACC/AHA Guideline on Management of Cardiovascular Risk in Obesity. JACC. 2023. https://www.jacc.org/doi/10.1016/j.jacc.2023.09.3929
8. Fryback DG, et al. Industry Funding and CVOT Effect Estimates: A Meta-Analysis. BMJ. 2017. https://pubmed.ncbi.nlm.nih.gov/29142006/
9. Kazi DS, et al. Cost-effectiveness of Semaglutide for Cardiovascular Risk Reduction. JAMA Cardiology. 2024. https://pubmed.ncbi.nlm.nih.gov/38241038/