Number Needed to Treat (NNT)

Mathematically, NNT is the reciprocal of the absolute risk reduction: NNT = 1 / ARR. Because reciprocals behave oddly near zero, confidence intervals for NNT can become very wide or unstable when the treatment effect is small, which is one reason methodologists recommend reporting the underlying absolute risk difference and its confidence interval too.

The number that sounds like a headcount but acts like a batting average

A drug ad might say it “cuts risk by 50%,” which sounds huge. But if the untreated risk was 2% and treatment lowers it to 1%, the absolute improvement is just 1 percentage point. That is where Number Needed to Treat steps in: it translates that gap into a more graspable question — how many people have to take this for one extra person to benefit?

Why NNT is almost never a whole person

Picture a baseball stat line. A batter can hit.333 even though no third of a hit exists; it is an average over many chances. NNT works the same way. If the absolute risk reduction is 10% (0.10), the number needed to treat formula is 1 ÷ 0.10 = 10. If the reduction is 3%, NNT is 1 ÷ 0.03 = 33.3. That does not mean you can treat one-third of a patient. It means that, on average, treating about 33 people leads to one extra good outcome compared with control.

That is the surprise: NNT is not a property of the drug alone. It is a property of the drug, the outcome, the follow-up time, and the baseline risk of the people studied. The same treatment can have a very different NNT in a high-risk group than in a low-risk group, because the absolute payoff changes even if the relative effect looks similar.

What counts as “good” depends on what you are buying

People often ask, What is a good NNT value? There is no universal cutoff. Lower is better because fewer people must be treated for one extra benefit. An NNT of 2 is a very large effect; an NNT of 100 is much smaller. But “good” also depends on the stakes, harms, cost, and inconvenience. A high NNT may still be worthwhile for a cheap, safe intervention preventing a severe outcome; a low NNT may still be unappealing if side effects are common.

This is why NNT should travel with its partner, Number Needed to Harm (NNH). Benefit without harm is incomplete math. A medication with an NNT of 25 for symptom relief and an NNH of 20 for a major side effect tells a very different story than the same NNT paired with an NNH of 500.

The one decision to make today

When you see an NNT in a paper, guideline, or supplement claim, do not ask whether the number is “good” in the abstract. Ask one concrete question instead: “Good for whom, over what time?” If the time frame is missing, or the population is not like you, the number is wearing someone else’s name tag.