One large matched cohort says yes, and the size of the benefit is five fewer deaths per thousand infections. Thirty-day COVID mortality ran 0.3% among semaglutide users against 0.8% among matched metformin users, a relative risk of 0.39 [1]. Influenza ran 0.2% against 0.7%. The awkward part is that the benefit did not grow with the dose, which is the opposite of what a dose-dependent effect looks like.
What was compared
Adults with a documented COVID-19 or influenza infection were classified by whether they had been taking semaglutide beforehand or metformin without prior GLP-1 exposure, drawn from a 29-million-patient network. Matching covered demographics, body mass index, glycated hemoglobin, vaccination status, prior infection, antiviral use, healthcare utilization and baseline comorbidities, leaving 7,092 matched pairs for COVID and 1,920 for influenza.
That is a more thorough matching list than most studies of this kind manage, which is why the result is worth taking seriously rather than dismissing. It is the opposite end of the spectrum from a matched comparison that let follow-up differ by two hundred days.
The numbers, with their bases
COVID hospitalization ran 5.3% against 7.4%, and the composite severity outcome 8.2% against 10.6%. In influenza, hospitalization was 6.5% against 9.6%. Intensive care admission and mechanical ventilation were not significantly different in either cohort, and those are the outcomes least sensitive to who ends up coded as admitted.
The inflammation mechanism, and what it does not establish
The usual proposed mechanism is reduced inflammation. Nine randomized trials covering 6,239 adults with overweight or obesity and no diabetes found C-reactive protein fell 40.90% more than placebo, 95% CI 35.42 to 46.37 [2]. Body weight fell 12.04% more, 95% CI 11.00 to 13.08.
Those two percentages are not comparable and will be compared. The first is a share of a blood test result and the second is a share of the person. C-reactive protein varies by an order of magnitude between people, swings with any infection or injury, and is skewed hard to the low end. A large relative drop from a small starting value is a small absolute change. The analysis publishes no figure in milligrams per liter, and this page will not estimate one. The fuller reading sits in what happens to inflammation markers.
Who is taking these drugs anyway
Both readings of the infection result turn on who gets prescribed one. A nationwide Norwegian registry matched 190,818 people starting a weight-loss drug to population controls and read their diagnoses from the two preceding years [3]. Between 71% and 78% carried at least one recorded comorbidity, against 49% to 51% of controls, and 19% to 24% carried four or more.
Those ranges span five different drugs rather than expressing uncertainty. Opioids had been dispensed to 24% to 30% of users in the preceding year against 13% to 15% of controls, and antidepressants to 20% to 24% against 10% to 11%. A population starting one of these drugs is already carrying more illness than the people it gets compared with, which is the shape described in why people start.
What a buyer should take from it
Not a reason to buy, and not a reason to worry. The absolute risk of dying from either infection in these cohorts was under 1%, so a relative risk of 0.39 moves a small number to a smaller one. Nothing here is a substitute for a vaccine, and nobody has run a randomized trial of these drugs against infection outcomes.
What it adds is one more entry on the list of things the class appears to touch, and that list is long enough to be worth skepticism on its own. The widest observational view of it is in 175 outcomes across two million records.