Decent Newsroom
Sign In

Number needed to treat: turning trial results into something you can picture

By Research Methods Notes November 11, 2026
Clinical trial results are often reported as relative changes: "the drug reduced heart attacks by 30%." That sounds impressive, but on its own it doesn't tell you how many people actually benefit. The

Clinical trial results are often reported as relative changes: “the drug reduced heart attacks by 30%.” That sounds impressive, but on its own it doesn’t tell you how many people actually benefit. The number needed to treat (NNT) is a simple way to make the result concrete.

The calculation

Suppose that over five years:

  • 4% of people in the placebo group have a heart attack.
  • 2.8% of people in the treatment group do.

The absolute risk reduction is 4% − 2.8% = 1.2 percentage points, or 0.012.

The NNT is 1 / 0.012 ≈ 83. On average, about 83 people need to take the drug for five years to prevent one heart attack. The other 82 either would not have had a heart attack anyway or had one despite treatment.

The relative reduction here is 30% (1.2 / 4). Both numbers are correct; they just answer different questions.

Why NNT depends on baseline risk

The same relative effect gives very different NNTs in different populations:

  • If baseline risk is 20%, a 30% relative reduction lowers it to 14%: NNT ≈ 17.
  • If baseline risk is 1%, the same reduction lowers it to 0.7%: NNT ≈ 333.

That’s why a treatment can be clearly worthwhile for high-risk patients and hard to justify for low-risk ones, even when the trial reports a single relative effect.

Things to keep in mind

  • Always attach a time frame. “NNT of 83 over five years” is meaningful; “NNT of 83” is not.
  • Look at the confidence interval. NNTs computed from uncertain differences can have very wide, even unbounded, intervals.
  • Compare with the number needed to harm (NNH). If 1 in 50 people taking the drug gets a significant side effect, the trade-off looks different from the NNT alone.
  • NNT is an average. It doesn’t tell you which individual will benefit.

The takeaway

When you see a relative risk reduction, ask for the baseline risk and work out the absolute difference. NNT turns that difference into a number of people, which is often the most useful way to judge whether an effect matters.

More research explainers: https://www.crushsci.com

#statistics #research
Write a comment