Number Needed to Treat Calculator
NNT and NNH from two event rates, with absolute risk reduction and a plain interpretation.
What it does
Number needed to treat turns an absolute risk reduction into a count of people. If a treatment lowers the event rate by 8 percentage points, about 13 people must be treated for one extra person to avoid the event.
The calculation starts with the event rate in each group, not with a relative risk. That distinction matters because a large relative reduction can still leave a small absolute gain when the event is rare.
The defaults use 20 events in 100 control patients and 12 events in 100 treated patients. The risks are 20.0% and 12.0%, the absolute risk reduction is 8.0 percentage points, and NNT is 12.5, which is reported as about 13.
When the treatment group does worse, the same arithmetic produces a number needed to harm. An NNT of 12.5 and an NNH of 12.5 are not equivalent claims: one counts people treated to prevent an event, the other counts people treated to cause one.
How to use
- Enter the number of events in the control group.
- Enter the total number of people in the control group.
- Enter the number of events in the treatment group.
- Enter the total number of people in the treatment group.
- Press Calculate to see both risks and the absolute risk reduction.
- Read NNT or NNH, then round up to the next whole person.
Formula
- ARR
- Absolute risk reduction, the difference between the control and treatment event rates.
- NNT
- People treated for one additional person to avoid the event.
- NNH
- People treated for one additional person to have the event.
If the two risks are equal, the reciprocal is infinite and the tool says there is no difference. A negative ARR means the treatment had more events, so the result is reported as NNH. Always keep the time horizon with the number, because a one-year NNT is not a ten-year NNT.
Worked example
Example: 20 control events and 12 treatment events per 100
Defaults: control 20 of 100, treatment 12 of 100.
- Control risk: 20 / 100 = 20.0 %.
- Treatment risk: 12 / 100 = 12.0 %.
- Absolute risk reduction: 20.0 % - 12.0 % = +8.0 percentage points.
- NNT = 1 / 0.08 = 12.5.
- Round up: treat about 13 people to prevent one event.
- NNH is not applicable because the treatment lowered the risk.
Reference table
| Measure | Value | Meaning |
|---|---|---|
| Control risk | 20.0 % | Events per 100 control patients |
| Treatment risk | 12.0 % | Events per 100 treated patients |
| Absolute risk reduction | +8.0 percentage points | The actual gap between the groups |
| NNT | 12.5 | Reciprocal of the 0.08 risk reduction |
| Reported NNT | About 13 | Rounded up so the claim is not optimistic |
Limits
- The result applies to the same time horizon as the study data.
- Baseline risk matters: the same relative effect can give very different NNT values.
- A single study gives a point estimate, not a guarantee for a different population.
- NNT does not include side effects unless the event you count is the side effect.
Common mistakes
- Using the relative risk reduction instead of the absolute risk reduction.
- Rounding 12.5 down to 12 and making the treatment look stronger.
- Swapping the control and treatment groups, which flips the sign.
- Reporting NNT without the duration of treatment.
- Ignoring harm or cost when the NNT looks favorable.
Try it yourself
Quick checks. Answers sit tucked under each question.
Control has 30 events in 100 and treatment has 20 in 100. What are ARR and NNT?
ARR is +10.0 percentage points and NNT is 10.0, reported as about 10 people treated to prevent one event.
Control has 10 events in 100 and treatment has 15 in 100. What changes?
ARR is -5.0 percentage points, so the result is NNH 20.0, about 20 people treated to cause one extra event.
Both groups have 10 events in 100. What is the NNT?
The risks are equal, ARR is 0, and the tool reports an infinite NNT with no difference between groups.
Questions
What does an NNT of 13 mean?
About 13 people must be treated for one additional person to avoid the event, compared with the control treatment.
Why is NNT rounded up?
You cannot treat half a person. Rounding up keeps the claim conservative instead of overstating the benefit.
What is number needed to harm?
The same reciprocal when the treatment increases the event rate. It estimates how many people are treated for one extra event.
Can NNT be negative?
The reciprocal is reported as NNH when the treatment does worse. A negative NNT is not meaningful in a report.
Does a small NNT always mean a good treatment?
No. Consider side effects, cost, and the severity of the event. A small NNT for a mild outcome may matter less than a larger NNT for a serious one.
Why does the time horizon matter?
Event rates accumulate over time. An NNT for one year cannot be applied to ten years without new data.
Sources
- Monaghan et al., odds ratios, relative risk, absolute risk, and number needed to treat
- Smart Health Choices, relative and absolute risk reduction and number needed to treat