Research Literacy
Relative vs. Absolute Risk: The Most Common Trick in Health Headlines
The short answer
In the SELECT trial, semaglutide reduced major cardiovascular events by 20% relative to placebo — the number most headlines led with. But using the trial's own raw event counts, 6.5% of the semaglutide group (569 of 8,803) had an event versus 8.0% of the placebo group (701 of 8,801), a difference of 1.5 percentage points. Framed as a relative risk reduction, the effect sounds sweeping. Framed as an absolute risk reduction, it means roughly 67 people would need to take semaglutide for about 3.3 years to prevent one event. Both numbers come from the same data. Neither is wrong. But they tell a reader very different things about how much benefit to expect.
Quick answers
- What is it?
- Relative risk reduction describes how much smaller one group's risk is compared to another as a percentage, while absolute risk reduction is the plain difference in how many people out of 100 were actually affected in each group.
- Why are people interested in it?
- Headlines often report only the relative-risk number because it sounds bigger, and readers need to know how to translate that into what it actually means for them.
- What has actually been studied?
- The SELECT trial, a randomized trial of 17,604 people with cardiovascular disease and obesity but no diabetes, comparing semaglutide to placebo on major cardiovascular events and treatment-discontinuing side effects.
- How strong is the evidence?
- A single large, well-controlled randomized trial (17,604 people): the headline 20% relative risk reduction and the underlying 1.5-percentage-point absolute reduction are both accurate, real, and derived from the same data.
- What don't we know yet?
- What the absolute risk reduction would look like in a lower-risk population, how the number needed to treat changes with longer or shorter treatment, and which framing a given reader should weight more heavily — a judgment the statistics alone can't resolve.
- Where can I read the original research?
- Sources at the bottom of this page: Lincoff AM, Brown-Frandsen K, Colhoun HM, et al., New England Journal of Medicine, 2023.
The short answer: In the SELECT trial, semaglutide reduced major cardiovascular events by 20% relative to placebo — the number most headlines led with. But using the trial's own raw event counts, 6.5% of the semaglutide group (569 of 8,803) had an event versus 8.0% of the placebo group (701 of 8,801), a difference of 1.5 percentage points. Framed as a relative risk reduction, the effect sounds sweeping. Framed as an absolute risk reduction, it means roughly 67 people would need to take semaglutide for about 3.3 years to prevent one event. Both numbers come from the same data. Neither is wrong. But they tell a reader very different things about how much benefit to expect.
What the research shows
The same trial, reported two different ways
According to PubMed, the SELECT trial, published in the New England Journal of Medicine in 2023, randomized 17,604 people with existing cardiovascular disease and overweight or obesity, but no diabetes, to weekly semaglutide (8,803 people) or placebo (8,801 people). Over a mean follow-up of 39.8 months, a primary cardiovascular event occurred in 569 semaglutide patients (6.5%) and 701 placebo patients (8.0%), yielding a hazard ratio of 0.80 (95% CI 0.72-0.90, P<0.001) — commonly summarized as a 20% relative reduction in risk. (DOI) That's the number that traveled furthest in coverage of the trial. But the trial's own raw percentages tell a second, quieter story: the absolute difference between groups was 8.0% minus 6.5%, or 1.5 percentage points.
Why the two framings sound so different
Relative risk reduction describes how much smaller one group's risk is compared to the other's — it's a ratio, and ratios can look large even when the underlying risks are both small. Absolute risk reduction describes the actual difference in how many people were affected in each group — a plain subtraction. Twenty percent sounds like a change that would matter to almost anyone. One and a half percentage points sounds smaller, because it is smaller in the sense that matters for predicting what happens to any given person: roughly 92% of the placebo group did not have a primary event during the trial, meaning most people in the trial — on either drug or placebo — never experienced the outcome being measured at all.
Turning the absolute number into something more usable
One way to make an absolute risk reduction concrete is the number needed to treat, or NNT — how many people have to receive the treatment for the length of the trial to prevent one additional event, calculated as 1 divided by the absolute risk reduction. Using SELECT's own numbers, 1 ÷ 0.015 comes out to about 67. That means roughly 67 people had to take semaglutide for a mean of about 3.3 years to prevent one cardiovascular death, heart attack, or stroke that would otherwise have occurred. That's a real, clinically meaningful effect in a high-risk population — it's not a reason to dismiss the finding. But it's a very different mental picture than "20% risk reduction" conjures on its own, and a reader who only sees the relative number has no way to reconstruct it.
Neither number tells the whole story alone
Relative risk reduction tends to stay fairly stable across different baseline risk levels, which is part of why researchers report it — it summarizes an effect in a way that generalizes reasonably well across similar populations. Absolute risk reduction, by contrast, depends heavily on the baseline risk of the population being studied: applying SELECT's 20% relative reduction to a much lower-risk group of people (say, without existing cardiovascular disease) would produce a much smaller absolute benefit, even though the relative number stayed the same. Reading a trial result well means asking for both figures, not settling for whichever one sounds more persuasive.
The same trick cuts the other way, on harms
Relative-versus-absolute framing isn't only a tool for making benefits sound bigger — it works just as well on risks, in either direction. In the same SELECT trial, adverse events leading to permanent discontinuation of the study drug occurred in 1,461 of 8,803 patients (16.6%) on semaglutide versus 718 of 8,801 patients (8.2%) on placebo (P<0.001). Framed relatively, that's more than double the rate — a 102% relative increase, which sounds severe. Framed absolutely, it's a gap of 8.4 percentage points: about 8 additional people out of every 100 treated stopped the drug because of a side effect, compared with placebo. Both descriptions are accurate. A reader evaluating the trial's trade-offs needs both the benefit side and the harm side expressed the same way — relative-to-relative or absolute-to-absolute — because mixing an absolute benefit against a relative harm, or vice versa, can make a trade-off look far more or less favorable than the raw numbers support.
What we don't know yet
What the absolute risk reduction would look like in a lower-risk population. Everyone enrolled in SELECT already had confirmed cardiovascular disease, so this 1.5-percentage-point effect can't be assumed to generalize to healthier people who haven't had a cardiovascular event.
How the number needed to treat changes with longer or shorter treatment. The 67-person estimate is tied specifically to a mean follow-up of 39.8 months; a different treatment duration could shift that figure in either direction, and this trial doesn't test that directly.
Which framing a given reader should weight more heavily. That's a judgment about values and context — how much a 1-in-67 chance of preventing a major event over three years matters to a specific person — that the statistics alone don't resolve.
Sources
- Lincoff AM, Brown-Frandsen K, Colhoun HM, et al. Semaglutide and Cardiovascular Outcomes in Obesity without Diabetes. New England Journal of Medicine, 2023;389(24):2221-2232. DOI
Article records maintained via PubMed. Peptide Proof reports what studies found. It does not provide medical advice, dosing guidance, or treatment recommendations.
| Evidence Index: semaglutide
Sources
- Semaglutide and Cardiovascular Outcomes in Obesity without Diabetes
- New England Journal of Medicine, 2023 · n=17604 · DOI
Get the evidence, not the hype.
New research briefs, plainly written and fully sourced. No dosing advice, ever.