Semaglutide and Days Spent in Hospital

Cardiovascular trials report hazard ratios for events most people will never have. A prespecified analysis of the SELECT trial measured something more recognizable: how much time participants spent in hospital. Semaglutide cut it from 176.2 to 157.2 days per 100 patient-years.^[1]

What was measured

SELECT randomized 17,604 people with obesity and established cardiovascular disease but without diabetes. Its main result — fewer major cardiovascular events — we covered in the evidence-duration article. This analysis asked a different question of the same trial.

No heterogeneity appeared across BMI, age or sex subgroups — the reduction was consistent rather than concentrated in one group.^[1]

Why days in hospital is a better unit

A hazard ratio for a composite cardiovascular endpoint is hard to feel. Days in hospital is not. It captures severity as well as frequency — a shorter admission counts differently from a longer one — and it aggregates everything that put someone in a bed rather than only the events a trial was designed to count.

The cohort had a high rate of hospital admissions to begin with. That is what makes a 11% relative reduction worth stating in days.

It is also the outcome that translates most directly into cost, which is usually argued about separately from clinical benefit. Time in hospital is expensive to systems and disruptive to people, and it is the same number.

What kind of evidence this is

Prespecified and exploratory, which is a specific combination worth understanding. Prespecified means the analysis was planned before the data were seen, so it is not a result someone went looking for afterward. Exploratory means it was not the endpoint the trial was powered to answer, so it should be read as supportive rather than definitive.

What it has that most hospitalization findings do not is randomization. We have covered several observational studies reporting fewer hospitalizations on these drugs — in people with cancer, in infection risk — and every one carries the objection that healthier people get prescribed newer drugs. Randomization removes that objection entirely.