number needed to harm

**Definition:** NNH is the number of patients who need to receive an intervention (or be exposed to a risk factor) over a defined time period for one additional patient to be harmed, compared with controls.

**Formula:** NNH = 1 / ARI, where ARI (absolute risk increase) = risk in exposed group − risk in unexposed (control) group. It is expressed as a whole number, and by convention it is rounded up.

**Worked example:** Drug X causes a major bleed in 5% of treated patients and in 2% of controls. ARI = 0.05 − 0.02 = 0.03. NNH = 1/0.03 ≈ 33. About 33 patients must be treated for one additional major bleed to occur.

**Relationship to related measures:** - NNT (number needed to treat) = 1/ARR (absolute risk reduction). It measures benefit, so a lower NNT is better. - NNH = 1/ARI. It measures harm, so a higher NNH is better (safer). - Both are tied to a specific time frame, outcome, and population. They cannot be compared across studies with different durations or baseline risks. - Likelihood of being helped or harmed (LHH) = (1/NNT) / (1/NNH). An LHH greater than 1 means the intervention is more likely to help than harm.

**Interpretation and caveats:** - NNH depends on baseline risk. The same relative risk produces a smaller NNH (more harm) in high-baseline-risk patients. - Unlike relative risk (RR) or odds ratio (OR), NNH conveys absolute magnitude, so it is more useful for clinical decision-making and risk-benefit discussions. - If the intervention has no effect (ARI = 0), NNH is infinite. - In a cohort study or RCT, ARI is calculated directly. In case-control studies, risk can't be calculated directly, so NNH is less straightforward. - It is typically reported with a 95% confidence interval. If the CI for ARI crosses zero, the NNH CI is uninterpretable or spans infinity.

**Board-relevant points:** - Know the formula and be able to calculate it from a 2×2 table or from event rates. - Distinguish ARR (benefit, used for NNT) from ARI (harm, used for NNH). - A question may give percentages for adverse events in treatment and placebo groups. Subtract them, then take the reciprocal. - Trap: if the stem gives the relative risk increase, you must know the baseline (control) event rate to convert it to ARI. - Clinical examples: NNH for major bleeding with anticoagulants, for myopathy with statins, or for GI bleeding with NSAIDs.