“How many people need a treatment for one more person to benefit?” That's the question behind a research measure called number needed to treat, often shortened to NNT.
It can make a study's average result easier to picture. But the number only makes sense when you know the outcome, the comparison group and the period of time being studied. It isn't a promise about what will happen to you personally.
Start with a fictional example
Imagine a study follows two groups of 100 people for one year. In the usual-care group, 10 people experience a particular health event. In the treatment group, six people experience it.
The difference is four people per 100, or four percentage points. If the estimate represents the treatment effect, about 25 people would need to receive the treatment instead of usual care for one additional person to avoid that event over one year. That's because one divided by 0.04 is 25.
The numbers are invented. They don't describe a real medicine, event or personal risk. Their purpose is to show how the measure relates to an absolute difference between groups. For an actual GLP-1 comparison, our SELECT trial guide looks at the event counts and follow-up time that give a headline context.
What “one additional person” means
An NNT of 25 does not mean researchers can identify which person benefited. It also doesn't mean the other 24 people received no possible benefit or that the 25th person was guaranteed a particular result. The measure summarizes a group comparison for one specified outcome.
Some people in both groups may avoid the event. Some people receiving treatment may experience it. Individuals can also have side effects or benefits that aren't captured in the chosen outcome. NNT focuses on the additional difference estimated between groups, not the complete experience of each person.
The comparison matters. A number calculated against usual care may differ from one calculated against another active treatment. Ask what the alternative actually was and whether both groups received other support.
The clock is part of the number
“NNT of 25” is incomplete without the time period. Preventing one additional event over one year is a different statement from preventing one over five years. A longer follow-up may observe more events, but that doesn't mean you can simply multiply a short-term result to predict a long-term one.
The outcome matters just as much. An NNT for avoiding a hospital admission answers a different question from an NNT for achieving a particular change in a laboratory measure. Look for the exact event and how researchers measured it.
If the outcome is a combination of several events, check what the components are. One part may account for more of the combined result than another, and the components may differ in importance to patients.
Why baseline risk changes the picture
The absolute difference depends partly on how common an event is without the treatment. A similar relative effect can translate into different numbers of events avoided in groups with different starting risks.
For example, reducing an event rate from 10 in 100 to six in 100 is a four-point absolute reduction. Reducing it from five in 100 to three in 100 is a two-point reduction. The relative reduction is the same in this illustration, but the NNT changes from 25 to 50 for the same follow-up period.
That doesn't make one group more deserving of care. It shows why a study's number shouldn't be applied automatically to someone whose health history and baseline risk differ from those of participants.
Don't leave uncertainty or downsides out
The NNT is calculated from an estimated difference, and that estimate has uncertainty. If the confidence interval includes little benefit, a possible harm or no difference, one precise-looking NNT can be misleading. Ask about the underlying event rates and interval rather than relying on a single rounded number.
Benefits are only part of a decision. Consider side effects, inconvenience, cost, monitoring and the possibility that an outcome important to you wasn't measured. Researchers may also report a number needed to harm for a particular adverse outcome, but that number has its own time frame, comparison and uncertainty.
A treatment with a lower NNT for one outcome isn't automatically the best choice for every person. The importance of the event and the treatment's burdens matter.
If the estimate comes from a review combining several studies, check whether the comparison groups and follow-up periods were similar enough to make a single number useful. The review may use an assumed baseline risk to translate a relative effect into an absolute difference. Ask whether that assumption fits the population you're discussing.
Questions to ask
When you see an NNT in a health story, ask: “Over how long? Compared with what? Which event? What were the actual numbers in both groups? How certain is the estimate?” Then discuss whether the study population resembles you and what alternatives are available.
Those questions turn a statistic into something more useful. NNT can help describe an average difference, but it works best alongside the underlying numbers and a conversation about your own priorities.










