NNT Calculator — Number Needed to Treat

The number needed to treat (NNT) is the number of patients you need to treat to prevent one additional bad outcome (such as death, stroke, or heart attack). For example, an NNT of 5 means you have to treat 5 patients with the drug for one additional patient to avoid the bad outcome. It is a standard evidence-based medicine measure introduced by Laupacis and colleagues in 1988.

First compute the absolute risk reduction: ARR = CER − EER, where CER is the control event rate and EER is the experimental event rate. Then NNT = 1 / ARR. For example, if the control event rate is 50% and the treatment event rate is 30%, ARR = 0.5 − 0.3 = 0.2, so NNT = 1 / 0.2 = 5. NNTs are always rounded up to the nearest whole person.

NNT counts whole people, and rounding down would understate how many patients must be treated. An NNT of 4.2 means 4 patients are not quite enough to prevent one bad outcome on average, so you round up to 5. The Centre for Evidence-Based Medicine states that NNTs are always rounded up to the nearest whole number.

NNT (number needed to treat) describes benefit: how many patients must be treated to prevent one additional bad outcome. NNH (number needed to harm) describes harm: how many patients must be treated for the treatment to cause one additional bad outcome. If the experimental event rate is higher than the control rate, the absolute risk increase (ARI = EER − CER) is inverted instead: NNH = 1 / ARI.

ARR is the absolute difference in event rates (CER − EER), while RRR is the proportional reduction relative to the control group ((CER − EER) / CER). The same RRR can mean very different things: a 71% relative risk reduction with a baseline risk of 96% gives NNT = 2, but the same 71% with a baseline risk of 0.96% gives NNT = 147. NNT combines the treatment effect with the patient's baseline risk, which is why it is more clinically useful.

Lower NNT means a stronger effect: an NNT of 1 is the theoretical best (every treated patient benefits). Roughly, NNT 1–2 is a very large effect, 3–5 a large effect, 5–10 a moderate effect, 10–20 a modest effect, and above 50 a very small effect. But NNT always depends on the time horizon and outcome: NNT = 15 over 5 years for preventing death is different from NNT = 15 over 3 months for curing a headache.

Event rates accumulate over time, so NNT is meaningless without a follow-up period. An NNT of 25 to prevent one stroke over 5 years of treatment is not comparable to an NNT of 25 over 6 months. Always report NNT together with the treatment duration, for example: '25 patients must be treated for 5 years to prevent one stroke.'

The 95% confidence interval for NNT is obtained by inverting the confidence interval of the ARR. If the ARR interval includes zero, the result is not statistically significant: the interval runs from an NNT on the benefit side, through infinity, to an NNH on the harm side. This calculator reports such cases honestly instead of showing a single misleading number.

No. NNT is a population-level measure: it describes the average effect across a group of patients like those in the trial. It does not predict whether a specific individual will benefit, because individual baseline risks differ. This calculator is for education in evidence-based medicine and is not medical advice.

The concept was introduced by Laupacis, Sackett, and Roberts in 'An assessment of clinically useful measures of the consequences of treatment' (New England Journal of Medicine, 1988). The worked example in this calculator — intensive vs usual insulin therapy in the Diabetes Control and Complications Trial (CER 9.6%, EER 2.8%, NNT 15) — is the derivation table reproduced in the Centre for Evidence-Based Medicine's teaching materials.