SELECT Subgroup Analyses: Who Responded Most and Least

At a glance

Parameter	Detail
Trial	SELECT (Semaglutide Effects on Cardiovascular Outcomes in People with Overweight or Obesity)
N	17,604
Intervention	Semaglutide 2.4 mg subcutaneous weekly
Comparator	Matched placebo
Duration	Mean follow-up 39.8 months (event-driven)
Population	Age ≥45, BMI ≥27, established atherosclerotic CVD, no diabetes
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 subgroup data matters here

The SELECT primary publication reported a population-level 20% relative risk reduction in three-point MACE. That number secured semaglutide's cardiovascular indication in obesity. But a single hazard ratio across 17,604 patients masks heterogeneity. Clinicians prescribing Wegovy for secondary cardiovascular prevention need to know: does the benefit hold in women (who were only 28% of enrollment)? Does BMI modify it? Does a patient's inflammatory status at baseline predict who gains most?

Pre-specified subgroup analyses in large cardiovascular outcome trials (CVOTs) are designed to answer exactly these questions. They carry more statistical weight than post-hoc cuts because the analysis plan, stratification, and multiplicity handling are locked before unblinding.

How SELECT structured its subgroup plan

The trial pre-specified 25 subgroups in the statistical analysis plan. Each used a Cox proportional-hazards model with a treatment-by-subgroup interaction term. The pre-specified factors included:

• Age (<65 vs. ≥65 years)
• Sex (male vs. female)
• Race (White vs. non-White)
• Ethnicity (Hispanic/Latino vs. not)
• Baseline BMI (<30 vs. 30, <35 vs. ≥35 kg/m²)
• Baseline eGFR (<60 vs. ≥60 mL/min/1.73 m²)
• Baseline HbA1c (<5.7% vs. ≥5.7%)
• Qualifying CV event type (MI, stroke, or PAD)
• Region (North America, Europe, other)
• Statin use at baseline (yes vs. no)

The interaction P values were reported without multiplicity adjustment, consistent with FDA guidance on subgroup analyses in confirmatory trials. This means individual subgroup P values should be interpreted as hypothesis-generating, not confirmatory. The overall trial was powered for the primary endpoint in the full cohort, not within any single subgroup.

Results by age

Interaction P = 0.65. The cardiovascular benefit was consistent regardless of age. The absolute risk reduction was numerically larger in the ≥65 group (2.3 percentage points vs. 1.2), which is expected given higher baseline event rates in older participants. This is clinically relevant: older patients with obesity and established CVD stand to gain more in absolute terms, a pattern previously observed with statins and now replicated with a GLP-1 receptor agonist.

Results by sex

Interaction P = 0.69. The point estimates were similar, but the female subgroup's confidence interval crossed 1.0. This does not mean semaglutide failed in women. It reflects the smaller sample size: only 28% of enrollees were female, yielding fewer events and wider confidence intervals. The interaction test showed no evidence of effect modification by sex.

This enrollment imbalance is a recognized limitation. Women are underrepresented in CVOTs across drug classes. The 2022 AHA scientific statement on sex differences in CVD trials has called for mandatory sex-stratified reporting and recruitment targets. SELECT met the reporting standard but not the recruitment one.

Results by baseline BMI

Interaction P = 0.87. The hazard ratios across BMI strata were remarkably stable, ranging from 0.78 to 0.82. The SELECT primary results explicitly noted this consistency, reinforcing that cardiovascular benefit was not confined to the highest-BMI patients.

This finding has prescribing implications. Some payers initially restricted Wegovy's cardiovascular indication to BMI ≥35. The trial data show the effect is present down to BMI 27, the lowest enrolled threshold. The FDA-approved Wegovy label accordingly sets the indication at BMI ≥27 with established CVD.

Results by baseline glycemic status

Participants were required to not have diabetes. But prediabetes (HbA1c 5.7 to 6.4%) was common.

Interaction P = 0.83. The benefit was effectively identical in both groups. This is a significant distinction from earlier GLP-1 CVOTs like SUSTAIN-6 and LEADER, which enrolled patients with established type 2 diabetes. SELECT proved the cardiovascular benefit of semaglutide exists independently of glucose lowering, a mechanistic insight that reframes GLP-1 receptor agonists as cardiovascular drugs rather than diabetes drugs with cardiac side benefits.

Results by inflammatory biomarker (hsCRP)

The most clinically provocative subgroup cut was by baseline high-sensitivity C-reactive protein (hsCRP).

Interaction P = 0.33. While not statistically significant for interaction, the numerical trend is consistent with a hypothesis that patients with higher baseline systemic inflammation derive greater benefit. Semaglutide reduced hsCRP by approximately 38% in the treatment arm, a magnitude comparable to the CRP reductions seen in the CANTOS trial of canakinumab, which targeted IL-1β directly.

This observation has generated considerable discussion. If semaglutide's cardiovascular mechanism is partly anti-inflammatory (through weight loss, direct receptor effects on macrophages, or both), then hsCRP might eventually serve as a treatment-selection biomarker. No guideline currently recommends checking hsCRP before starting semaglutide for cardiovascular prevention, but the data make this a plausible future direction.

Results by qualifying cardiovascular event

Interaction P = 0.96. The benefit was directionally consistent regardless of the qualifying event. Stroke and PAD subgroups had fewer events and correspondingly wider confidence intervals, but point estimates overlapped tightly. Patients who qualified via MI had the narrowest confidence interval simply because they constituted the largest subgroup.

Results by race and ethnicity

SELECT enrolled participants across 41 countries. The racial composition was approximately 84% White, 6% Black, 4% Asian, and 6% other or not reported. Hispanic/Latino ethnicity was reported by approximately 14% of participants.

Interaction P values were non-significant for both race and ethnicity. As with sex, the non-White and Hispanic/Latino subgroups had wider confidence intervals due to smaller sample sizes. The point estimates were nearly identical to the overall population, offering no signal of differential efficacy by race or ethnicity, though the power to detect a meaningful difference was limited.

Results by renal function

Interaction P = 0.53. Patients with moderate CKD (eGFR <60) showed a numerically larger benefit. Given the high cardiovascular risk burden in CKD and the known safety of semaglutide in renal impairment (per the FLOW trial), this subgroup finding adds confidence for prescribing in patients with overlapping kidney and cardiovascular disease.

What the subgroup data tell us about real-world prescribing

Three patterns emerge from the full subgroup forest plot:

1. No subgroup was harmed. Every pre-specified subgroup had a point estimate below 1.0. No interaction test was significant. This is unusually clean for a CVOT of this size.

2. Absolute benefit tracks with baseline risk. Older patients, those with CKD, and those with elevated hsCRP had higher baseline event rates and therefore larger absolute risk reductions. Number needed to treat (NNT) calculations favor high-risk subgroups.

3. Low enrollment limits conclusions in women and non-White patients. The trial cannot definitively confirm or exclude a smaller effect in these groups. It can only say the data are consistent with the overall result. Future real-world evidence studies will need to fill this gap.

Limitations of subgroup interpretation

Even well-conducted subgroup analyses carry inherent risks. The SELECT investigators acknowledged several:

• Multiplicity: With 25 subgroups tested, one or two nominally significant interaction terms would be expected by chance alone. None crossed that threshold, which actually strengthens the case for consistency.
• Post-hoc biomarker cuts: The hsCRP median split, while pre-specified in its binary form, invites continuous-variable analyses that were not part of the original plan. Dose-response relationships between hsCRP tertiles and treatment effect remain hypothesis-generating.
• Duration ceiling: Mean follow-up was 39.8 months. Whether subgroup-specific differences emerge or converge with longer treatment remains unknown.
• Generalizability: All patients had established atherosclerotic CVD. The subgroup findings may not extrapolate to primary prevention populations.

The bottom line for clinicians

The SELECT trial showed that semaglutide 2.4 mg reduced MACE by 20% across a broad range of patient characteristics. No clinically meaningful effect modification was found by age, sex, BMI, glycemic status, qualifying event type, race, or ethnicity. The signal that higher baseline inflammation may predict greater benefit warrants prospective validation but is not yet actionable in guidelines.

For the prescribing clinician, the practical message is straightforward: if a patient meets the indication (BMI ≥27, established CVD, no diabetes), the cardiovascular benefit of semaglutide is supported regardless of which demographic or clinical subgroup they fall into.

Frequently asked questions

›Did semaglutide work better in men than women in SELECT?

Point estimates were similar (HR 0.79 in men, 0.84 in women) with no significant interaction. The wider confidence interval in women reflects lower enrollment (28%), not a biologically different response. The trial data are consistent with benefit in both sexes.

›Does a higher BMI predict a bigger cardiovascular benefit from semaglutide?

No. The hazard ratios were nearly identical across BMI 27, 30, 30, 35, and ≥35 kg/m² (range 0.78, 0.82). Cardiovascular benefit did not scale with the degree of obesity in SELECT.

›Should I check hsCRP before starting semaglutide for heart protection?

No guideline currently recommends this. The SELECT data showed a numerically larger benefit in patients with above-median hsCRP, but the interaction was not statistically significant. It remains a hypothesis-generating observation, not a treatment-selection tool.

›Did SELECT include enough non-White patients to draw conclusions?

Approximately 16% of participants were non-White. Point estimates were consistent with the overall result, but confidence intervals were wide. The trial was not powered to detect race-specific differences. Real-world evidence is needed.

›Is semaglutide cardiovascular benefit independent of blood sugar control?

Yes. SELECT excluded patients with diabetes, and the benefit was identical in those with normal HbA1c vs. prediabetes. This establishes that the cardiovascular effect operates through non-glycemic mechanisms.

›Did patients with kidney disease benefit from semaglutide in SELECT?

Patients with eGFR <60 mL/min/1.73 m² had a numerically larger benefit (HR 0.74) compared to those with preserved renal function (HR 0.81), though the interaction was not significant. Combined with the FLOW trial data, this supports use in CKD.

›Were the SELECT subgroup analyses pre-specified or post-hoc?

The 25 primary subgroup analyses were pre-specified in the statistical analysis plan before unblinding. Interaction P values were reported without multiplicity adjustment, consistent with FDA guidance for exploratory subgroup evaluation.

›What was the number needed to treat in high-risk subgroups?

Exact NNTs vary by subgroup event rates. Older patients (≥65) had an absolute risk reduction of approximately 2.3 percentage points over 40 months, corresponding to an NNT of roughly 43. Higher-risk subgroups consistently had lower (better) NNTs.

›How does SELECT compare to GLP-1 cardiovascular trials in diabetes?

SUSTAIN-6 and LEADER showed cardiovascular benefit of semaglutide and liraglutide in type 2 diabetes populations. SELECT extended this to patients without diabetes, demonstrating the cardiovascular effect is not mediated solely through glucose improvement.

›Did the type of prior cardiovascular event affect semaglutide's benefit?

No. Patients qualifying via prior MI, stroke, or peripheral arterial disease all showed consistent hazard ratios (0.79, 0.82) with no significant interaction. The benefit was not restricted to any single qualifying event type.

References

1. 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
2. Wegovy (semaglutide) prescribing information. Novo Nordisk. Revised 2024. FDA Label
3. Marso SP, Bain SC, Consoli A, et al. Semaglutide and cardiovascular outcomes in patients with type 2 diabetes (SUSTAIN-6). N Engl J Med. 2016;375(19):1834-1844. PubMed
4. Ridker PM, Everett BM, Thuren T, et al. Antiinflammatory therapy with canakinumab for atherosclerotic disease (CANTOS). N Engl J Med. 2017;377(12):1119-1131. PubMed
5. Perkovic V, Tuttle KR, Rossing P, et al. Effects of semaglutide on chronic kidney disease in patients with type 2 diabetes (FLOW). N Engl J Med. 2024;391(2):109-121. PubMed
6. Cholesterol Treatment Trialists' Collaboration. Efficacy and safety of statin therapy in older people: a meta-analysis. Lancet. 2019;393(10170):407-415. PubMed