Calculate the number needed to treat (NNT) from 2×2 tables, odds ratios, risk ratios, or absolute risk reduction. Convert between clinical effect measures with 95% confidence intervals.
Enter the four cells of a 2×2 table: events and non-events for the treatment (intervention) and control groups.
EER = a/(a+b), CER = c/(c+d), ARR = CER − EER, NNT = 1/|ARR|
Select your data source: a 2×2 table, odds ratio with baseline risk, risk ratio with baseline risk, or a known ARR.
Fill in the required fields. For OR or RR inputs, you must also provide the control event rate (baseline risk) to compute absolute measures.
Review the NNT (exact and rounded), ARR with 95% CI, experimental and control event rates, and related measures like NNH.
Copy all computed values to your clipboard in a formatted text block ready for your manuscript, report, or meta-analysis.
The same odds ratio or risk ratio produces very different NNTs depending on how common the outcome is in the control group. Always report the baseline risk alongside the NNT so readers can assess applicability to their population.
A statistically significant p-value does not guarantee a clinically meaningful treatment effect. NNT bridges this gap by translating relative measures into absolute patient-level impact, making it easier to judge whether an effect is worth pursuing.
Without a reliable estimate of the baseline risk, converting from OR or RR to NNT is misleading. Use the CER from the control arm of your trial, or from epidemiological data for the target population.
The number needed to harm (NNH) quantifies adverse effects using the same framework. A treatment with NNT = 10 and NNH = 100 means 1 in 10 patients benefits while 1 in 100 is harmed. Comparing NNT and NNH supports balanced treatment decisions.
The NNT calculator translates relative effect measures into absolute clinical impact — the single most important step in bridging meta-analytic evidence and bedside decision-making. When a meta-analysis reports a pooled odds ratio of 0.75, the clinical meaning depends entirely on the baseline risk: in a high-risk population (CER = 40%), this OR corresponds to an NNT of approximately 13, while in a low-risk population (CER = 5%), the same OR yields an NNT of approximately 91. This baseline-dependent relationship is why the number needed to treat calculator requires both the effect estimate and the control event rate as inputs — a point emphasized in the Cochrane Handbook (Higgins et al., 2023, Chapter 15). The assumed control risk (ACR) can be sourced from the trial control arm for trial-specific NNT or from population-level epidemiological data when the goal is to project treatment impact onto a real-world clinical setting. NNT is also inherently time-horizon dependent: an NNT of 50 over one year corresponds to an NNT of approximately 10 over five years for a constant hazard, so the follow-up period must always accompany the reported NNT.
The ARR to NNT converter computes the reciprocal transformation: NNT = 1/|ARR|, where ARR (absolute risk reduction) equals the control event rate minus the experimental event rate. Confidence intervals for NNT are derived from the ARR confidence interval by inversion, which creates an interpretive challenge when the ARR interval crosses zero — the NNT confidence interval passes through infinity, splitting into a "numbers needed to treat" range and a "numbers needed to harm" range. Altman (1998) described this behavior and recommended presenting the full interval rather than truncating at clinically implausible values, as the discontinuity itself conveys important information about treatment uncertainty. Visual Rx (Cates plots) offer a patient-friendly way to present NNT by depicting 100 person icons coloured to show expected outcomes with and without treatment, making absolute risk differences immediately intuitive for shared decision-making.
Converting from odds ratio to NNT or risk ratio to NNT follows established formulas that this tool implements with exact confidence intervals. For odds ratios: EER = (CER × OR) / (1 − CER + CER × OR). For risk ratios: EER = CER × RR. Both conversions require specifying the baseline risk from the control arm of the trial or from epidemiological data for the target population. When computing pooled effect sizes from individual study data, use our effect size calculator for odds ratios, risk ratios, and standardized mean differences with variance estimates suitable for meta-analytic pooling.
NNT is particularly powerful when presented alongside the number needed to harm (NNH), enabling a direct benefit-harm comparison. A treatment with NNT = 10 and NNH = 50 means one patient benefits for every five harmed — a ratio that many clinicians and patients would find acceptable. GRADE guidelines (Guyatt et al., 2011) recommend presenting absolute effects — including NNT when applicable — in Summary of Findings tables; GRADEpro software formats these tables with NNT prominently displayed alongside the certainty of evidence rating for each outcome. In network meta-analysis, NNT can be derived from indirect effect estimates by applying the same OR-to-NNT or RR-to-NNT conversion using an assumed control risk, enabling treatment ranking that incorporates absolute clinical impact. You can generate Summary of Findings tables using our GRADE certainty of evidence tool. For statistical significance testing of the underlying 2×2 table, our chi-square calculator provides Pearson's chi-square and Fisher's exact test. When test-and-treat decisions depend on diagnostic accuracy, our diagnostic accuracy calculator computes sensitivity, specificity, and likelihood ratios from the same 2×2 structure.
The number needed to treat (NNT) is the average number of patients who need to receive a treatment for one additional patient to benefit compared to a control. An NNT of 1 means every patient benefits; an NNT of 10 means you need to treat 10 patients for 1 to benefit. Lower NNTs indicate more effective treatments. NNT is always rounded up to the next whole number for clinical interpretation because you cannot treat a fraction of a patient.
To convert an odds ratio (OR) to NNT, you need a baseline risk (the control event rate, CER). First compute the experimental event rate: EER = (CER × OR) / (1 – CER + CER × OR). Then calculate the absolute risk reduction: ARR = CER – EER. Finally, NNT = 1 / |ARR|. Without a baseline risk, the conversion is impossible because the same OR implies different absolute differences at different baseline rates.
Both convey the same information in different forms. ARR (absolute risk reduction) is a proportion, while NNT is its reciprocal. ARR is easier to use in calculations and meta-analytic pooling because it has well-defined standard errors. NNT is more intuitive for clinicians and patients: “we need to treat 20 people for 1 to benefit” is easier to grasp than “the absolute risk reduction is 5%.” Many guidelines recommend reporting both.
There is no universal threshold. Context matters enormously. An NNT of 2–5 is excellent for most interventions. Statins for secondary prevention of cardiovascular events have an NNT around 25–40 over 5 years, which is considered worthwhile given the severity of the outcome. For preventive interventions in low-risk populations, NNTs of 50–100+ may still be acceptable if the treatment is safe and inexpensive.
NNT has several limitations: (1) It depends on the baseline risk, so the same treatment can have very different NNTs in different populations. (2) Confidence intervals for NNT can be difficult to interpret when the ARR CI crosses zero (the interval passes through infinity). (3) NNT does not account for the magnitude of benefit or the severity of harms. (4) NNT assumes a fixed treatment duration, which may not apply to chronic conditions. Always report NNT alongside the baseline risk and follow-up period.
From a 2×2 table, calculate the experimental event rate (EER = events in treatment / total in treatment) and control event rate (CER = events in control / total in control). Then ARR = CER – EER, and NNT = 1 / |ARR|. Round up to the next whole number. If EER > CER, the result is a number needed to harm (NNH) instead.
NNT itself is always a positive whole number, but when the treatment increases rather than decreases the event rate (EER > CER), the ARR becomes negative. In this case, the reciprocal is reported as the number needed to harm (NNH) — the number of patients treated for one additional adverse outcome compared to control. Negative ARR indicates harm, not benefit.
When the 95% CI for the ARR crosses zero, the NNT confidence interval passes through infinity. This occurs because NNT = 1/ARR, and as ARR approaches zero, NNT approaches infinity. The interval splits into an NNT range (benefit) and an NNH range (harm), with infinity in between. This indicates uncertainty about whether the treatment helps or harms.
Our biostatisticians can handle effect size extraction, NNT calculations across subgroups, heterogeneity assessment, and publication-ready forest plots for your entire systematic review.