Calculate NNT, NNH, ARR, RRR, relative risk, and odds ratio from clinical trial data with confidence intervals and cost-effectiveness analysis.
The Number Needed to Treat (NNT) is one of the most intuitive measures of treatment effect in evidence-based medicine. It answers a simple question: "How many patients need to receive this treatment for one additional patient to benefit?" An NNT of 10 means you would need to treat 10 patients for one to have the outcome prevented by treatment, while the other 9 would be treated without direct benefit.
NNT is derived from the absolute risk reduction (ARR) — the difference between the event rates in the control and treatment groups. Unlike relative measures (relative risk, odds ratio), NNT incorporates baseline risk, making it more clinically meaningful for shared decision-making. A treatment that reduces relative risk by 50% could have an NNT of 4 (if baseline risk is 40%) or an NNT of 200 (if baseline risk is 1%).
This calculator computes NNT from event rates or raw trial data, generates confidence intervals, calculates related measures (ARR, RRR, RR, OR), performs basic cost-effectiveness analysis, and provides visual representations of treatment effect. When the experimental event rate exceeds control, the calculator automatically reports Number Needed to Harm (NNH).
NNT translates abstract treatment effects into concrete, patient-facing numbers that facilitate shared decision-making. Instead of telling patients "this drug reduces your risk by 25%," you can say "we would need to treat 40 people like you for 5 years for one to avoid a heart attack." This transparency builds trust and enables informed consent.
NNT = 1 / ARR = 1 / (CER − EER). ARR = CER − EER. RRR = (CER − EER) / CER × 100. RR = EER / CER. When EER > CER, result is NNH (Number Needed to Harm). Cost to prevent = NNT × cost per treatment.
Result: NNT = 48; ARR = 2.10%; RRR = 21.0%; Cost to prevent one event = $24,000
With a control event rate of 10% and treatment event rate of 7.9%, the absolute risk reduction is 2.1%, yielding an NNT of 48. This means 48 patients need to be treated to prevent one event. At $500 per treatment, preventing one event costs $24,000.
Major clinical guidelines increasingly report NNT alongside traditional effect measures. The AHA/ACC cardiovascular prevention guidelines report NNT per 100 patients treated for 10 years for statin therapy by risk category. The USPSTF reports NNT for screening recommendations. Understanding NNT helps clinicians translate guideline recommendations into individualized patient care. When a guideline recommends a treatment for a broad population, calculating the NNT for your specific patient's risk level reveals whether the benefit justifies the burden.
The NNT Group maintains a curated database of NNTs for common medical interventions, grading them with color-coded recommendations. This free resource provides pre-calculated NNTs from landmark trials, making it easy to compare treatment options during clinical encounters. Their framework categorizes interventions as "Green" (clearly helpful), "Yellow" (unclear), "Red" (clearly harmful), and "Black" (no benefit), helping clinicians quickly assess the strength of evidence for everyday clinical decisions.
Several mistakes are common when using NNT. First, comparing NNTs across different conditions or time frames is invalid — NNT 10 for a treatment that prevents death is more valuable than NNT 10 that prevents a headache. Second, NNT derived from relative risk reduction applied to a different baseline risk than the original study may not be reliable if the relative risk reduction is not constant across risk strata. Third, NNT for composite outcomes may mask heterogeneity — a treatment with NNT 20 for "cardiovascular events" may have NNT 100 for death but NNT 15 for nonfatal MI. Always understand what outcome the NNT represents.
Relative risk can be misleading without baseline risk context. A drug that reduces heart attack risk by 50% (RRR) sounds impressive, but if baseline risk is only 2%, the ARR is just 1% (NNT=100). The same 50% RRR applied to a 40% baseline risk gives ARR of 20% (NNT=5). NNT directly communicates the number of patients impacted, making it ideal for shared decision-making with patients. It answers "what does this treatment mean for ME?" rather than "what fraction of risk does it remove?"
NNT acceptability depends on the condition and treatment risks. For a fatal condition with a safe treatment, NNT of 50-100 may be acceptable (statins for cardiovascular prevention). For a treatment with significant side effects, NNT should be much lower. Perfect NNT is 1 (every patient benefits). Antibiotics for bacterial infection often have NNT of 2-5. Surgical interventions may need NNT <10 to justify operative risk. Cancer screening often has NNT >100 per year screened. Always consider NNT alongside NNH for side effects.
NNT (Number Needed to Treat) applies when treatment reduces events — it is the number needed to prevent one bad outcome. NNH (Number Needed to Harm) applies when treatment increases events — it is the number treated per one additional patient harmed. Ideally, NNT for benefit should be much lower than NNH for side effects. For example, warfarin for atrial fibrillation may have NNT of 25 for stroke prevention but NNH of 50 for major bleeding, making the benefit-to-harm ratio favorable.
Only if your patient shares similar baseline risk. NNT is specific to the study population's baseline risk. A statin trial in high-risk patients (CER=15%) may yield NNT=30, but applying the same relative risk reduction to a low-risk patient (CER=3%) gives NNT=150. You must estimate your patient's baseline risk (using Framingham or ASCVD calculators for cardiovascular risk, for example) and then apply the study's relative risk reduction to calculate a patient-specific NNT.
NNT confidence intervals can be counterintuitive. When the ARR confidence interval crosses zero (includes both benefit and harm), the NNT CI technically passes through infinity and wraps around. For example, an NNT of 15 with an ARR CI of -1% to 5% means the CI includes NNT=15 to NNT=∞ (no effect) and also NNH=100 to NNH=∞. This occurs when the study result is not statistically significant. A narrow NNT CI entirely on the benefit side (e.g., NNT 8 to 25) provides confidence in treatment efficacy.
Cost per event prevented = NNT × cost per treatment course. This is a simplified cost-effectiveness analysis. If preventing one stroke costs $50,000 (NNT=50 × $1,000/year), and a stroke costs $250,000 in healthcare and lost productivity, the treatment is cost-effective. However, true cost-effectiveness analysis also considers quality-adjusted life years (QALYs), time horizons, discounting, and indirect costs. Still, NNT × cost is a useful first approximation for bedside decision-making.