Key facts
- Plain meaning: how many people you treat for one extra person to benefit.
- Formula: NNT = 1 ÷ absolute risk reduction (ARR).
- Direction: lower NNT = stronger effect, for the same outcome.
- Always paired with: a named outcome and a time frame (e.g. "over 3 years").
- Companion measure: number needed to harm (NNH) for adverse effects.
- Not a personal number: it describes a group in a trial, not any one reader.
What is number needed to treat?
Number needed to treat (NNT) is a way of expressing how much good a treatment does, in units people can picture. It answers a simple question: if you gave this treatment instead of the comparison to a group of people, how many would you have to treat for one extra person to benefit?
"Benefit" here almost always means avoiding a specific bad outcome — a heart attack, a stroke, a hospital admission. The idea was popularised as a clinically useful measure of treatment effect precisely because a raw percentage can feel abstract, whereas "you treat 25 people to prevent one event" is concrete. If a treatment has an NNT of 25 for a given outcome over a given time, then, on average, 24 of those 25 people would have had the same result with or without it, and one avoided the bad outcome because of it.
That framing is deliberately humble. NNT never claims to identify which person benefits — only that, across the whole group, treating that many produces one extra good result.
How NNT is calculated
NNT is built directly on the absolute risk reduction (ARR). The ARR is the difference between the rate of the outcome in the control group and the rate in the treatment group. NNT is simply its reciprocal:
- Absolute risk reduction = control event rate − treatment event rate.
- NNT = 1 ÷ absolute risk reduction (with rates written as proportions).
A worked, purely illustrative example makes the mechanics clear. Suppose a hypothetical treatment cut the rate of some outcome from 10% in the control group to 5% in the treatment group. The absolute risk reduction is 5 percentage points, or 0.05 as a proportion. Dividing one by 0.05 gives an NNT of 20: you would treat 20 people to prevent one event. If instead the outcome fell from 2% to 1%, the absolute reduction is only 0.01 and the NNT is 100 — the same halving of risk, ten times weaker in absolute terms, because the outcome was rarer to begin with.
By convention, NNT is rounded up to a whole number, since you cannot treat a fraction of a person. Those two numbers above are invented round figures chosen only to show the arithmetic; they are not results from any real study.
How to interpret an NNT
The first instinct — "lower is better" — is right, but incomplete. A smaller NNT does mean a treatment produces a benefit in more people per person treated. Yet an NNT is never good or bad in isolation; it only has meaning against three things:
- What outcome is being prevented. An NNT of 50 to prevent a death is a very different proposition from an NNT of 50 to prevent a mild, temporary symptom.
- Over what time frame. The same treatment usually has a smaller (better-looking) NNT the longer it is measured, because more events accumulate. An NNT "over 5 years" is not comparable to one "over 6 months."
- Against what harms. Every treatment carries downsides. The mirror-image measure, number needed to harm (NNH), estimates how many people must be treated for one extra person to experience a specific adverse effect. Weighing NNT against NNH is closer to how a clinician actually thinks.
Because of all this, comparing NNTs across different studies is a trap unless the outcome, population and time frame line up. Two NNTs are only fairly comparable when they answer the same question in the same kind of people over the same period.
A cardiovascular example
Cardiovascular prevention is where NNT is most at home, because trials there measure hard, countable events — heart attacks, strokes, cardiovascular deaths — over years of follow-up. Picture a purely hypothetical prevention trial in which the rate of a major cardiovascular event over several years is 8% in the placebo group and 6% in the treatment group. The absolute risk reduction is 2 percentage points; the reciprocal of 0.02 is 50, so the illustrative NNT is about 50 over that period. That single number captures something a relative figure hides: even a genuinely effective therapy may need to treat dozens of high-risk people for years to prevent one event, because most people in the group would not have had the event either way.
Real cardiovascular-outcomes trials report the same kind of absolute event rates that NNT is built from. Our summary of the SELECT cardiovascular outcomes trial, for example, walks through how a headline relative reduction sits alongside a smaller absolute difference in event rates — the exact tension NNT is designed to make visible. We deliberately do not compute an NNT for that trial here; the numbers above are a round illustration, not a finding from any study.
NNT versus relative risk
Much of the confusion around treatment effects comes from mixing up relative and absolute measures. A relative risk reduction — say, "20% fewer events" — describes the proportional change and stays the same whether the underlying risk is large or small. That is why relative figures can sound dramatic even when the real-world impact is modest.
NNT, being the reciprocal of the absolute risk reduction, refuses to flatter a result. When the baseline risk is low, a big relative reduction still prevents few events, and the NNT is correspondingly large. Reading the two side by side is the honest approach: the relative figure tells you how much the risk moved in proportion, and the NNT tells you how many people share in that movement. Neither is the whole story alone, which is one reason careful write-ups of phase 2 versus phase 3 trials report absolute numbers, not just relative ones.
What NNT does not tell you
NNT is a compression of a lot of detail into one figure, so it inevitably leaves things out:
- It is not a personal probability. An NNT of 40 does not mean any given patient has a 1-in-40 chance of benefiting; it is a group average that depends heavily on baseline risk. A higher-risk person may stand to gain more than a lower-risk one from the very same treatment.
- It hides uncertainty. Like any estimate from a trial, an NNT has a confidence interval around it. A tidy point estimate can sit on top of a wide range, especially in smaller studies.
- It ignores the size and timing of benefit. Preventing an event says nothing about quality of life, how long the benefit lasts, or how bad the event would have been.
- It is tied to one trial's population. An NNT measured in older adults with established disease — the kind of group studied in conditions such as type 2 diabetes or established heart disease — may not transfer to healthier or younger people.
Used well, NNT is a translator: it turns a probability into a headcount you can reason about. Used carelessly — quoted without its outcome, time frame, or harms — it becomes just another number that sounds precise while meaning less than it seems. That balance matters across every drug we summarise, including the newer peptides studied for weight loss, where relative and absolute framing frequently diverge.
Frequently asked questions
What does number needed to treat mean in plain English?
Number needed to treat, or NNT, is the number of people who have to receive a treatment instead of the comparison for one extra person to benefit — usually meaning one additional bad outcome is prevented. A lower NNT means the treatment helps more people for every person treated, so an NNT of 20 is more impressive than an NNT of 200 for the same kind of outcome.
How is NNT calculated?
NNT is one divided by the absolute risk reduction, where the absolute risk reduction is the difference between the event rate in the control group and the event rate in the treatment group, expressed as a proportion. Because it depends on the absolute difference, NNT changes with how common the outcome is and over what length of time it is measured, so it is only meaningful when the outcome and time frame are stated.
Is a lower or higher NNT better?
For a treatment that prevents a harmful outcome, a lower NNT is generally better, because fewer people need to be treated for one to benefit. But NNT is never good or bad on its own. It has to be weighed against how serious the prevented outcome is, how long treatment lasts, cost, and the harms of treatment, which are sometimes summarised as the number needed to harm.
Why can relative risk look more impressive than NNT?
Relative measures, such as a 20% relative risk reduction, describe the proportional change and stay the same whether the underlying risk is high or low. NNT reflects the absolute difference, so when the baseline risk is small the same relative reduction prevents few events and the NNT is large. Reading both together keeps a result the right size instead of overstating or understating it.
Does a good NNT mean a treatment is right for me?
No. NNT is a population-level summary from a specific trial in a specific group of people over a specific time. It does not tell any single person whether they will be the one who benefits, and it says nothing about individual history, other conditions, or side effects. Whether a treatment is appropriate for you is a clinical decision made with a licensed clinician. This page is educational and not medical advice.
Sources
This explainer draws on the foundational description of NNT and a general reference on drug-trial statistics:
- Cook RJ, Sackett DL. The number needed to treat: a clinically useful measure of treatment effect — BMJ, 1995; PubMed 7873954.
- David S, et al. Drug Trials — StatPearls; PubMed 31536202.
- General reference record: number needed to treat.