Semaglutide’s COVID Signal in SELECT Needs Scrutiny

Semaglutide cut COVID-19 deaths in SELECT. It did not stop people from getting COVID.

This finding deserves attention, and it is also the one most likely to get oversold. In the 2024 SELECT mortality analysis, people assigned to weekly semaglutide 2.4 mg had fewer COVID-19 deaths than people assigned to placebo: 43 versus 65, for a hazard ratio of 0.66 (95% CI 0.44-0.96). Serious COVID-19 adverse events were also lower, 232 versus 277, with an unadjusted P value of 0.04. But COVID-19 infection incidence was similar between groups.

If a drug lowers death after infection without lowering infection itself, the obvious explanation is not better immunity in the usual sense. It points instead to something happening later, after SARS-CoV-2 has already gained a foothold: less inflammatory escalation, less cardiometabolic collapse, less progression to fatal disease. That is a real biological question. The trial was not built to answer it.

What SELECT actually proved

The core result from SELECT is much firmer than the COVID-19 signal. In the primary 2023 SELECT publication, 17,604 adults with established cardiovascular disease, obesity or overweight with BMI at least 27 kg/m2, and no diabetes were randomized to weekly subcutaneous semaglutide 2.4 mg or placebo and followed for a mean of 3.3 years. The prespecified primary endpoint, major adverse cardiovascular events, fell with semaglutide: hazard ratio 0.80 (95% CI 0.72-0.90).

That sets the evidentiary hierarchy. SELECT was powered and designed to test cardiovascular outcomes. On that question, semaglutide delivered. The COVID-19 result came later, in a mortality-focused secondary analysis of the same trial population. Randomization still helps, which gives the finding more credibility than a loose observational claim, but it is still exploratory.

The same 2024 secondary analysis reported broader mortality reductions as well. All-cause mortality fell with semaglutide, hazard ratio 0.81 (95% CI 0.71-0.93). Non-cardiovascular death also fell, hazard ratio 0.77 (95% CI 0.62-0.95). Infection-related deaths were 62 with semaglutide and 87 with placebo, hazard ratio 0.71 (95% CI 0.51-0.98). Those numbers extend beyond the coronary artery story. They do not, by themselves, prove a general anti-infective effect.

The class effect is real. The infection story is not settled.

The broader GLP-1 receptor agonist literature makes the cardiovascular benefit easy to believe. A 2025 meta-analysis of 99,599 patients found that GLP-1 receptor agonists as a class reduce major adverse cardiovascular events, cardiovascular mortality, and all-cause mortality. Another 2025 meta-analysis in type 2 diabetes reached a similar conclusion for injectable and oral GLP-1 receptor agonists across cardiovascular, kidney, and mortality outcomes.

That creates a strong prior for semaglutide improving hard outcomes in high-risk patients. It does not answer the COVID-19 question, because those meta-analyses did not separate infectious disease mortality. The same gap appears in broader obesity-drug reviews, including a 2025 network meta-analysis in Nature Medicine and a 2025 Cochrane review of semaglutide for obesity. They support weight-loss efficacy and overall cardiometabolic benefit. They do not tell us whether semaglutide changes the odds of dying from an infection.

That absence matters. When a striking result appears in one trial and nowhere else, caution is not cynicism. It is basic hygiene.

A plausible mechanism exists, but it is still a hypothesis

The most plausible explanation runs through inflammation and adiposity, not viral defense. GLP-1 receptor agonists reduce visceral fat, improve metabolic function, and lower inflammatory biomarkers. In a 2025 phase 2 trial of orforglipron, another GLP-1 receptor agonist, treatment lowered C-reactive protein and other cardiovascular risk biomarkers in participants with type 2 diabetes or obesity without diabetes. In the 2024 STEP-HFpEF program, semaglutide also improved CRP in obese patients with heart failure with preserved ejection fraction.

A 2025 meta-analysis in obese chronic heart failure provides the broader frame: excess adiposity drives chronic systemic inflammation, endothelial dysfunction, and organ stress. That pathway could make severe COVID-19 more dangerous in people who are already metabolically and cardiovascularly fragile. If semaglutide lowers that inflammatory background state, it could, in theory, reduce the odds that infection tips into respiratory failure or multi-organ trouble.

That theory fits the SELECT paradox better than an immunity story. If semaglutide improved mucosal defense, innate viral clearance, or exposure resistance, infection rates should have dropped. They did not. The drug seems, at most, to have affected severity after infection was established.

Still, biomarker improvements are not mechanistic proof. Lower CRP is evidence of an anti-inflammatory effect, not proof of direct GLP-1 signaling on the specific immune cells that drive fatal COVID-19. Human data in this evidence base do not resolve whether the effect is direct, indirect through weight loss and metabolic change, or a statistical fluke.

Why the COVID-19 result could be real, and why it could disappear

There are two respectable ways to read the SELECT COVID-19 finding.

The optimistic reading is that the signal fits the rest of the drug's profile. Semaglutide lowers cardiovascular risk, improves weight, reduces inflammatory biomarkers, and appears to reduce all-cause mortality in this high-risk population. Severe COVID-19 is partly a cardiometabolic stress test. A drug that improves the host response could plausibly reduce deaths without preventing infections.

The skeptical reading is that this is exactly the kind of result secondary analyses generate when enough outcomes get inspected. The COVID-19 mortality result rests on 108 total deaths across 17,604 participants. The confidence interval, 0.44 to 0.96, is compatible with a large benefit but also close enough to the null that a modest shift in events could erase statistical significance. The serious adverse event result used an unadjusted P value of 0.04, and the analysis was not adjusted for multiplicity across SELECT's larger subanalysis program.

There is also a pandemic-specific problem. SELECT ran from 2018 through 2023, spanning very different phases of COVID-19: pre-vaccine waves, vaccine rollout, Delta, Omicron, and large changes in background fatality risk. The published mortality analysis does not report stratification by vaccination status, pandemic phase, or dominant variant period. If the two groups happened to differ even slightly in timing of exposure or vaccine uptake, that could distort a small event-count mortality outcome.

The internal pattern raises another caution. Infection-related deaths were lower overall, but once COVID-19 deaths are removed, the remaining non-COVID infection counts are tiny: roughly 19 versus 22. That is nowhere near enough to support a broad claim that semaglutide protects against fatal infections in general.

What biohackers should take from this

If you meet SELECT-like criteria, semaglutide's cardiovascular benefit is evidence-based. The strongest claim here is the boring one, which is rude but common in medicine: the well-powered primary endpoint matters more than the side story. In non-diabetic adults with obesity and established cardiovascular disease, semaglutide 2.4 mg reduced major cardiovascular events and was associated with lower all-cause mortality in the SELECT population.

The COVID-19 signal belongs in a different box. It is intriguing, biologically plausible, and far from settled. It should not drive clinical decisions about COVID-19 protection. It does not prove semaglutide boosts immunity. It does not establish a post-infection rescue mechanism. It does not yet have external replication from other trials, real-world datasets, or pharmacoepidemiology.

Best current reading: semaglutide clearly helps high-risk cardiometabolic patients; it may also reduce the odds that COVID-19 turns fatal in that same population, but the evidence for that second claim is still preliminary.

That is enough to justify watching the signal closely. It is not enough to market semaglutide as a COVID shield. The data support one conclusion with confidence and another with curiosity.