A headline says a GLP-1 medicine was linked to a “20% lower risk” of a health event. That sounds straightforward until you ask: 20% of what?
It doesn’t automatically mean 20 fewer people out of every 100 had the event. The headline is describing a comparison, and the starting number matters. Here’s a way to make sense of it without needing a statistics textbook.
Start with a simple example
Imagine two groups of 100 people are followed for the same length of time. In the first group, 10 people experience a particular event. In the second, eight do.
The difference is two people per 100, or two percentage points: 10% minus 8%. But two is also 20% of the first group’s 10 events. So, in this made-up example, you could describe the second group’s result as 20% lower relative to the first.
Both numbers describe the same comparison from different angles. The relative figure tells you how large the change is compared with the starting figure. The percentage-point difference shows the gap between the two percentages.
Now imagine the first group had one event per 100 people and the second had 0.8 per 100. That is still a 20% relative difference, but the gap is 0.2 percentage points. The starting point changes what the headline looks like in everyday numbers.
These examples are just arithmetic. They don’t predict what will happen to you or describe a particular treatment.
A real GLP-1 research example
One study you may hear about is SELECT. It compared weekly injectable semaglutide with a placebo alongside usual care in adults aged 45 or older who had established cardiovascular disease and overweight or obesity, but did not have diabetes.
Researchers counted a combined outcome: cardiovascular death, nonfatal heart attack or nonfatal stroke. During the study, at least one of those events was reported in 569 of 8,803 people assigned semaglutide, or 6.5%, and 701 of 8,801 assigned placebo, or 8.0%.
Those rounded proportions differ by 1.5 percentage points. That’s a useful companion to a headline about relative change. It’s also a group result from this particular study, not a promise that an individual will or won’t have an event.
The study also reported a figure called a hazard ratio of 0.80. This is a time-to-event comparison often summarized as about a 20% lower relative rate. It isn’t simply the result of subtracting the rounded 6.5% and 8.0% figures. Different summaries use the study data in different ways, so don’t expect the arithmetic from a headline to reproduce every reported result exactly.
Ask what was counted, and for how long
A “heart event” in one study might mean something different in another. SELECT combined three types of events in its main outcome; its numbers don’t say that each type fell by the same amount. The study followed people for an average of about 40 months, so the percentages shouldn’t be read as one-year results.
Also check who took part. Someone without established cardiovascular disease may have a different starting risk from the participants in SELECT. A result for one medication, formulation and group doesn’t automatically transfer to every GLP-1 treatment or every person.
What if the headline mentions weight instead?
Percentages can describe different things. A study might report an average percentage change in body weight, the proportion of people who reached a weight-loss target, or a relative difference in a health outcome. Those are three different measurements.
Before comparing numbers, ask whether the studies measured the same thing, included similar people and followed them for a similar length of time. Our guide to comparing GLP-1 trial results walks through those questions.
Take the useful question to your visit
You don’t have to calculate your own risk from a news story. Bring the article or study name to your clinician and ask, “What were the actual numbers in each group, and how relevant are the people in the study to me?” Ask about potential downsides as well as benefits.
A good headline can start the conversation. The details help you decide what the headline really says.