Number Needed to Harm (NNH)

The number of patients treated, over a defined period, for one extra patient to suffer a specific adverse outcome. NNH = 1 ÷ ARI (absolute risk increase) — the same arithmetic as NNT, but for harm. A smaller NNH means a more harmful treatment.

A group of treated patients in which a single highlighted figure is the one extra person harmed by the treatment. The fewer figures needed to colour in one, the more harmful the drug or intervention — the mirror image of the NNT pictograph.

Each figure is one patient treated for the defined period. The red figure is the patient who suffers the harm because of treatment. A small NNH (few figures per harm) is bad; a large NNH (a big crowd per harm) is reassuring. Always read NNH next to the NNT for benefit and judge the two together.

Net benefit is the balance of good against harm. A treatment with NNT 20 for benefit but NNH 15 for a serious harm is doing more damage than good. Harms are also frequently under-reported in trials, so a missing NNH is a red flag, not reassurance.

• NNH = 1 ÷ ARI (absolute risk increase; round UP)
• Smaller NNH = more harmful (opposite of NNT)
• Must share the same timeframe & baseline as the NNT to compare

Aspirin for primary prevention (Zheng & Roddick, JAMA 2019 meta-analysis, 164,225 patients) — aspirin cut major cardiovascular events only modestly while increasing major bleeding: absolute risk increase 0.47% → NNH ≈ 210 for a major bleed. The NNT for the cardiovascular benefit is of a similar order, so in low-risk people aspirin can harm about as often as it helps — which is why primary-prevention aspirin fell out of routine guidance.

• Quoting an NNT for benefit and ignoring the harm / NNH entirely.
• Comparing an NNH and NNT measured over different timeframes or baseline risks.
• Treating a missing or large NNH as safety — adverse events are often under-ascertained.