A percentage can look complete while leaving its most important feature unstated: percentage of what? A relative comparison and an absolute difference describe the same research from different angles. A useful reading keeps both visible, together with the people and time period behind them.
The study in one compact record
SELECT randomized 17,604 adults aged at least 45 with established cardiovascular disease, BMI of at least 27 and no diabetes to weekly semaglutide 2.4 mg or placebo. Its composite outcome was cardiovascular death, nonfatal heart attack or nonfatal stroke. Events occurred in 6.5% versus 8.0%; the hazard ratio was 0.80 (95% confidence interval 0.72–0.90). Mean follow-up was 39.8 months. Permanent discontinuation following adverse events was 16.6% versus 8.2%. Novo Nordisk funded the trial. This account uses the published abstract, not a full-paper appraisal. [Original randomized trial; abstract reviewed]
Two quantities, two different questions
The rounded event proportions differ by 1.5 percentage points: subtract 6.5 from 8.0. This describes the observed proportions over the study's follow-up. It is not a 20-percentage-point difference. The hazard ratio compares event hazards through time; a value of 0.80 is commonly described as a 20% lower hazard. A time-to-event analysis contains information that a simple subtraction of final proportions does not preserve.
Percentage points describe a difference between percentages. A relative percentage describes a comparison with a reference quantity. A hazard ratio is a time-to-event measure. These terms are related, but are not interchangeable.
The population is part of the result
Before interpreting a headline for yourself, identify who could enter the study and who could not. A trial addressing an established disease population cannot automatically answer the same prevention question for people without that history. Age, background treatment and follow-up also shape the question the evidence answers. The boundary is a reason to ask for interpretation, not a reason to ignore the finding.
Benefits and treatment experience belong together
A primary outcome is selected to answer an important research question; it is not a complete inventory of everything a person might value. Discontinuation, adverse effects, burden and practical access can matter alongside a favorable efficacy result. Reading one outcome in isolation can hide this distinction. Different outcomes may also use different definitions and analysis methods.
A careful conclusion
The result supports an answer to a defined trial question. It does not forecast an individual outcome, prove every GLP-1 product equivalent or rate any telehealth provider. We do not calculate a number needed to treat from the rounded proportions here: that shortcut can obscure the time horizon and analysis. Bring the original source and a specific applicability question to the clinician interpreting your care.
Source notes
Checked October 5, 2026. Provider pages document advertised terms. Study summaries identify the scope of our review.
- Lincoff et al. — SELECT cardiovascular outcomes trialOriginal randomized trial; abstract reviewed