Last updated: 2026-06-23

Number Needed to Treat (NNT)

The number needed to treat (NNT) summarizes how much a treatment helps in a single, intuitive figure: the average number of patients who must receive the treatment for one additional patient to benefit. Clinicians, evidence reviewers, and students use it to translate the statistics of a clinical trial into a practical sense of effort versus reward. An NNT of 10 means that, on average, ten people take the treatment so that one extra person avoids the outcome being studied; the other nine see no change in that outcome.

Absolute versus relative risk reduction

NNT is built on the absolute difference in risk between two groups, so it is worth separating two quantities that are easy to confuse. The control event rate (CER) is the proportion of untreated patients who experience the outcome, and the experimental or treatment event rate (EER) is the proportion among treated patients.

The absolute risk reduction is the plain difference between them:

$\text{ARR} = \text{CER} - \text{EER}$

The relative risk reduction scales that same difference against the baseline risk:

$\text{RRR} = \frac{\text{CER} - \text{EER}}{\text{CER}}$

Relative risk reduction often sounds more impressive because it ignores how common the outcome was to begin with. A treatment that cuts risk from 2% to 1% achieves a 50% relative reduction but only a one-percentage-point absolute reduction. NNT deliberately uses the absolute reduction so that the headline figure reflects real-world benefit rather than a flattering ratio.

The formula

Once the absolute risk reduction is known, the number needed to treat is simply its reciprocal:

$\text{NNT} = \frac{1}{\text{ARR}} = \frac{1}{\lvert \text{CER} - \text{EER} \rvert}$

The reciprocal follows directly from the definition. If the absolute risk reduction is 0.08, then each treated patient has an 8% chance of being the one who benefits, so on average it takes the reciprocal of that risk to accumulate one benefit:

$\text{NNT} = \frac{1}{0.08} = 12.5 \text{ patients}$

The absolute-value sign keeps the magnitude meaningful even when a treatment turns out to be harmful, which we return to below.

Why the result is rounded up

The raw NNT is usually a fraction. Because you cannot treat a fraction of a person, the convention is to round up to the next whole patient rather than to the nearest integer. Rounding up keeps the estimate conservative: it never claims a benefit from treating fewer patients than the arithmetic supports.

Worked example

Suppose a trial reports that 20% of control patients have a cardiovascular event over the study period, compared with 12% of treated patients. The event rates are CER = 0.20 and EER = 0.12.

Rounding up gives an NNT of 13: about thirteen patients must be treated for one additional cardiovascular event to be prevented, even though the relative risk reduction of 40% sounds substantial. Comparing the NNT against the burdens and costs of treatment is what turns the trial result into a decision.

When treatment causes harm

If the treatment event rate is higher than the control rate, the treatment increases the outcome rather than reducing it. The same reciprocal-of-the-difference calculation then yields the number needed to harm (NNH): the average number of patients treated for one additional patient to be harmed. The arithmetic is identical, so this calculator reports the magnitude either way and flags whether the figure represents a benefit or a harm.

Limits

An NNT is tied to the specific population, baseline risk, and time horizon of the study it came from. A treatment with an NNT of 50 over one year is not the same as an NNT of 50 over five years, and the figure can shift markedly when applied to patients whose baseline risk differs from the trial's. Confidence intervals around the NNT matter too, especially when the absolute risk reduction is small and uncertain. Treat the NNT as a communication aid alongside the underlying risks, not as a standalone verdict.

For the ratio-based companions to this measure, see the Relative Risk Calculator and the Odds Ratio Calculator.