Any trial gives you three numbers. The relative one is the headline: a drug that takes heart-attack rate from 20 in 100 down to 15 cuts risk "by a quarter," ignoring the underlying rate. The absolute one is the real drop, 20% to 15%, five percentage points. Run that same quarter-reduction where only 2 in 100 were at risk and you get half a point. Same drug, same biology, different decision.
The number-needed-to-treat (NNT) is that absolute drop flipped over: five points means 1 in 20 treated people avoided the event, half a point means 1 in 200 1.
This isn't niche. Across 35 studies, most patients overshot the benefit of statins, mammograms, colonoscopies, and blood-pressure drugs by tenfold or more, and undershot the side effects 2. The relative number is why: a result feels bigger shown as a relative reduction than as an absolute one or an NNT, and printing both together doesn't fix it 3.
Anchors: a statin after a heart attack, 1 in 30 avoid an early death over five years 4; the same drug in low-risk people, about 1 in 95 5; an SSRI for real depression, 1 in 7 6. Same statin, NNT 30 to 95, purely from who's taking it.
Any treatment claim gets the same three questions — a drug ad, a screening letter, a prescription your doctor floats.
With a clinician it collapses to one sentence: "What's my absolute chance of this outcome over the next few years if I do nothing, and how much does this change it?" It's a question they're trained to answer and one nobody asks; handing patients the real numbers lowers regret without raising anxiety 7. thennt.com lists published NNTs and NNHs for a figure first.
The cost lands at both ends. Some swallow a daily pill for years whose real benefit they'd have declined if they'd seen it. Others hear "statins cut risk by 25%," shrug, and walk away from a one-in-thirty lifesaver for them specifically 4. The actual number tells you how much to care.
The fine print — when to skip it, and what people get wrong
"50% reduction means half of us benefit." No — it means the treated group's event rate is half the untreated group's; often 1 in 100 gains. "The headline applies to me." NNT says 1 in N benefit, not which one.
NNT is a summary, not a verdict. A huge NNT can still be a yes when the outcome is catastrophic and the fix cheap. It hides subgroups, so use stratified figures where given, and don't multiply it across untested horizons.
- 1Laupacis A, Sackett DL, Roberts RS (1988). An assessment of clinically useful measures of the consequences of treatment. New England Journal of Medicine. link
- 2Hoffmann TC, Del Mar C (2015). Patients' expectations of the benefits and harms of treatments, screening, and tests: a systematic review. JAMA Internal Medicine. link
- 3Akl EA, Oxman AD, Herrin J, et al. (2011). Using alternative statistical formats for presenting risks and risk reductions. Cochrane Database of Systematic Reviews. link
- 4Scandinavian Simvastatin Survival Study Group (1994). Randomised trial of cholesterol lowering in 4444 patients with coronary heart disease: the Scandinavian Simvastatin Survival Study (4S). The Lancet. link
- 5Ridker PM, Danielson E, Fonseca FA, et al. (2008). Rosuvastatin to prevent vascular events in men and women with elevated C-reactive protein. New England Journal of Medicine. link
- 6Cipriani A, Furukawa TA, Salanti G, et al. (2018). Comparative efficacy and acceptability of 21 antidepressant drugs for the acute treatment of adults with major depressive disorder: a systematic review and network meta-analysis. The Lancet. link
- 7Stacey D, Légaré F, Lewis K, et al. (2017). Decision aids for people facing health treatment or screening decisions. Cochrane Database of Systematic Reviews. link
დაკავშირებული სახელმძღვანელოში (9)
- — The PREVENT number is the kind of risk figure these literacy skills help you interpret.
- — These drugs are a case study in reading a modest benefit honestly against real risk and cost.
- — Breast screening's headline benefit is a perfect case for reading the absolute numbers, not just the percentage.
- — Reading the real benefit numbers is exactly what an annual medication review needs: it's how you spot the pills no longer earning their place.
- — Reading your own labs pairs with knowing how to read a treatment's real benefit — together they make you a sharper patient.
- — The multi-cancer blood test is a perfect case for running the numbers — impressive detection claims, unproven benefit, real false-positive harm.
- — PSA is the textbook case for reading the actual numbers — one death prevented per 600 men, harms up front.
- — Asking your doctor for the number-needed-to-treat is one of the most useful questions you can bring to the visit.
- — Knowing the real number-needed-to-treat is exactly the kind of question a second opinion should answer about a proposed treatment.