Number Needed to Treat (NNT)

The number needed to treat (NNT) is the number of patients who need to be treated with a specific intervention for one additional patient to achieve a favorable outcome compared to the control or alternative treatment. It is calculated as the inverse of the absolute risk reduction (ARR) and provides an intuitive, clinically meaningful measure of treatment effectiveness.

NNT is one of the most useful measures for communicating treatment effectiveness to clinicians, patients, and decision makers. It translates statistical results into a concrete, actionable number: how many patients must receive the treatment before one additional patient benefits. An NNT of 5 means that one additional patient benefits for every five treated, compared with the control, while an NNT of 100 means one additional patient benefits for every 100 treated: a much smaller absolute benefit for the same outcome and time frame. This simplicity makes NNT invaluable for clinical decision making, for comparing the effectiveness of different treatments, and for communicating evidence to patients during shared decision making. NNT is often reported in clinical guidelines, Cochrane reviews, and drug information databases.

NNT is calculated as 1 divided by the absolute risk reduction (ARR). The ARR is the difference between the event rate in the control group and the event rate in the treatment group: ARR = control event rate minus treatment event rate. For example, if 20% of control patients experience the outcome compared to 12% of treated patients, the ARR is 8% (or 0.08), and the NNT is 1/0.08 = 12.5, typically rounded up to 13. This means that for every 13 patients treated, one additional patient benefits.

The ideal NNT is 1, meaning that every patient treated benefits. In practice, NNTs for treatments considered effective range from the single digits (for example, some antibiotic treatments for bacterial infections) to the hundreds (statins for primary prevention of cardiovascular events in low risk populations). The NNT depends on the baseline risk: the same relative risk reduction produces a smaller NNT (more impressive benefit) when the baseline risk is higher. For example, a 50% relative risk reduction applied to a baseline risk of 20% yields NNT = 10, while the same relative risk reduction applied to a baseline risk of 2% yields NNT = 100.

Confidence intervals for NNT are derived from the confidence interval for the ARR. When the ARR is small or the study is underpowered, the confidence interval for the NNT can be very wide. When the confidence interval for the ARR includes zero, the confidence interval for the NNT includes infinity (which corresponds to an ARR of zero, no difference between the groups) and extends to negative values (which correspond to NNH, indicating that the treatment may cause harm). Altman (1998) recommended reporting such an interval in two parts, labeled NNTB (number needed to treat to benefit) and NNTH (number needed to treat to harm): an ARR confidence interval from −2% to 10%, for example, becomes NNTB 10 to ∞ to NNTH 50. This feature of NNT confidence intervals requires careful interpretation.

NNT should always be reported with the specific outcome, the comparator, the time frame, and the confidence interval. An NNT reported without these contextual details is incomplete and potentially misleading.

Report NNT whenever a clinical trial or meta analysis demonstrates a statistically significant absolute risk reduction for a dichotomous outcome. Use it for clinical decision making, guideline development, and communicating evidence to patients. Present it alongside the NNH for adverse events to provide a complete benefit risk profile. NNT is less appropriate for continuous outcomes (where mean differences or standardized mean differences are more suitable) or for studies that report only relative risk reductions without providing absolute rates.

A common mistake is calculating NNT from a relative risk reduction without considering the baseline risk, which can produce widely varying NNTs depending on the population. A 25% relative risk reduction has very different clinical significance in a high risk population (baseline risk 40%, NNT = 10) versus a low risk population (baseline risk 4%, NNT = 100). Another error is reporting NNT without the time frame, making it impossible to compare across studies. Some researchers round NNT down rather than up, which overstates the treatment benefit (convention is to round up to the next integer). Finally, comparing NNTs across different studies, outcomes, or populations without acknowledging the contextual differences can be misleading.

A large RCT of a new antiplatelet agent versus aspirin for secondary prevention of stroke in 8,000 patients (4,000 per group) finds a stroke recurrence rate of 9.2% at 2 years in the new agent group versus 12.5% in the aspirin group. The ARR is 3.3% (95% CI: 1.94% to 4.66%), and the NNT is 1/0.033 = 30.3, rounded up to 31 (95% CI: 22 to 52). This means that for every 31 patients treated with the new agent instead of aspirin for 2 years, one additional stroke is prevented. The confidence interval for the NNT, from 22 to 52, reflects uncertainty in the true treatment effect and is noted in the clinical guideline discussion.