Skip to content
This GLP
← Research
Evidence

What a trial result would do to a population that never took the drug

Applying SELECT's measured changes to 200,012 survey records drops projected cardiovascular events by 2.99 points. Every figure is a model output.

Dana Sullivan7 min read
Projected 10-year cardiovascular events13.82%observed10.83%modeled(nobody took a drug)A gap of 2.99 points, produced by arithmetic rather than treatment.

SELECT showed semaglutide reducing cardiovascular events in people who already had cardiovascular disease, and its inflammation substudy is covered in inflammation fell before the weight. The question this addresses is what that would mean for the far larger group who are at risk but not yet diagnosed [1].

The method is worth describing precisely, because it determines what the numbers are. A risk equation predicting ten-year cardiovascular events was fitted using five factors — BMI, HbA1c, systolic blood pressure, high-sensitivity CRP and non-HDL cholesterol. It was then applied to survey participants twice: once with their real values, and once with those five values shifted by the placebo-adjusted amounts SELECT recorded. Among 21,720 people with a BMI of 27 or above and elevated baseline risk, projected incidence fell from 13.82% to 10.83%.

The load-bearing assumption is that the drug’s benefit runs entirely through those five risk factors. There is direct evidence from SELECT itself that it does not: inflammation fell before major weight loss and in participants who lost none, which means at least part of the mechanism is not captured by moving BMI and cholesterol on a spreadsheet. That cuts in the model’s favor — if some benefit arrives by another route, this underestimates it. The opposite risk, that a risk equation fitted in one population behaves differently in another, cuts the other way.

The absolute figure is the one to carry. Three fewer cardiovascular events per hundred people over a decade, in a group at elevated but not established risk, is a meaningful public health quantity and a modest personal one. Absolute reductions were larger in men (3.14 points) than women (2.70), with near-identical relative reductions — the same relative-versus-absolute distinction set out in the semaglutide and tirzepatide comparison and the FLOW numbers needed to treat.

What the analysis usefully establishes is scale. If these effects hold in primary prevention, the number of people who would need treating to realize them is enormous, and the cost question that follows is not a clinical one. The secondary analyses accounting for reduced compliance are the most honest part of the paper, given that most people stop within a year — a figure this site tracks alongside the network meta-analysis of nineteen drugs.

Frequently asked

Did anyone in this study take semaglutide?
No. It is a modeling exercise applying the risk factor changes measured in the SELECT trial to survey records, then recalculating predicted risk.
How large is the projected benefit?
Projected ten-year cardiovascular incidence fell from 13.82% to 10.83%, an absolute reduction of 2.99 percentage points — roughly three fewer events per hundred people per decade.
What is the main assumption?
That the drug's benefit works entirely through the five modeled risk factors. SELECT's own inflammation substudy suggests part of the effect arrives by another route, which this model cannot capture.

Sources

  1. [1] Schrage B, Lackner MK, Hoshiyar A, de Lemos JA, et al. (2026). Emulated Effects of Glucagon-Like Peptide 1 Receptor Agonist Therapy in the General Population Journal of the American College of Cardiology. PMID 41811277

Where to get it

Best GLP-1 injections

Every injectable seller we can verify, with the price each one publishes and an honest read of what the trials measured.

Compare providers →

More in Evidence