Number Needed to Treat (NNT)

The number of patients treated, over a defined period, for one of them to avoid one extra adverse event. NNT = 1 ÷ ARR (absolute risk reduction). Lower is better; NNT 1 means everyone benefits.

67 treated patients. Only the highlighted figure avoids the extra outcome because of treatment; the other 66 gain nothing from it.

Each figure is one patient treated for the trial’s duration. The single highlighted one is the patient whose bad outcome is prevented. The bigger the crowd needed to colour in one figure, the weaker the treatment’s real-world impact.

An impressive relative risk reduction can hide a tiny absolute benefit when events are rare. NNT re-anchors the effect to baseline risk — the fastest test of whether a “positive” trial should change ED practice.

• NNT = 1 ÷ ARR (always round UP)
• Always state the timeframe and the baseline risk
• NNH = the same maths, for harm

CRASH-2 (Lancet 2010) — 20,211 trauma patients with, or at risk of, significant haemorrhage, TXA vs placebo. 28-day all-cause mortality 16.0% → 14.5% (RR 0.91) → NNT ≈ 67. Time-critical: TXA helps if given ≤3 h; given >3 h the subgroup had more bleeding deaths — the NNT only holds for early treatment.

• Quoting an NNT with no timeframe/baseline risk.
• Confusing NNT (needs ARR) with relative risk reduction — classic SBA distractor.
• Forgetting NNT rises as baseline risk falls.