Posters · Glossary
What is number needed to treat (NNT)?
The number needed to treat, or NNT, is how many people need to receive a treatment, over a stated time, for one extra person to benefit compared with the control. It is one divided by the absolute risk reduction. If 20 percent of untreated people have an event and 15 percent of treated people do, the NNT is 20.
Relative risk reductions sound impressive and say little about scale. The NNT turns a trial result into a number a clinician, a patient or a funder can weigh against cost and side effects.
Nuwan Madhusanka · Co-founder
4 min read · Published
| Measure | Control | Treatment | Result |
|---|---|---|---|
| People with an event | 80 of 400 | 60 of 400 | |
| Event rate | 20 percent | 15 percent | |
| Absolute risk reduction | 20% minus 15% = 5 percentage points | ||
| Relative risk reduction | 5 divided by 20 = 25 percent | ||
| Number needed to treat | 1 divided by 0.05 = 20 people for 12 months |
How to calculate it
Take the event rate in the control group, subtract the event rate in the treatment group to get the absolute risk reduction, and divide one by that difference expressed as a proportion. In the table, 20 percent minus 15 percent is 0.05, and one divided by 0.05 is 20. So 20 people need the program for a year to prevent one fall that would otherwise have happened. The Cochrane Handbook describes the same calculation and notes that an NNT is normally rounded up to a whole number.
Why it is more useful than a relative reduction
The same 25 percent relative reduction means very different things at different baseline risks. If the control event rate were 2 percent rather than 20, a 25 percent reduction would take it to 1.5 percent, an absolute reduction of half a percentage point and an NNT of 200. The relative figure is identical; the practical value is ten times smaller. That is why reporting guidance encourages presenting absolute effects, not only relative ones.
Number needed to harm
The same idea applied to adverse effects gives the number needed to harm: how many people need treatment for one extra person to be harmed. Weighing an NNT of 20 for preventing a fall against a number needed to harm of 50 for an exercise injury gives a clearer picture than either alone. Both should come with a time period and a confidence interval.
Reading an NNT on a poster
Check the time frame, because an NNT over five years is not comparable with one over a year. Look for the confidence interval, which for an NNT can be wide and awkward when the effect is uncertain. Check the baseline risk of the population studied, since an NNT from a high risk trial will not hold in a low risk clinic. And see whether the outcome prevented is serious enough for the number to matter.
Showing it on a poster board
A number needed to treat is a natural key number: 26 of the 40 poster boards carry stat blocks, which show a large figure with a short line beneath, room for the time frame and interval. A chart figure can show the event rates in each arm with one of the ten chart kinds, such as vertical bars. The clinical trial example shows the layout; include the counts in your brief, because the generator fills numbers only from it.
Questions people ask
What is a good number needed to treat?
It depends on the outcome, the time frame, the cost and the harms. An NNT of 50 can be excellent for preventing death with a cheap, safe drug, and poor for a minor symptom with an expensive treatment that has side effects. Compare it with the number needed to harm and the alternatives, not with a fixed threshold.
Can a number needed to treat be negative?
If the treatment group does worse than the control, the absolute risk reduction is negative and the calculation gives a negative number, which is better reported as a number needed to harm. Confidence intervals that cross zero difference are similarly awkward, spanning benefit and harm, and need careful explanation.
Why is the number needed to treat rounded up?
Because people come in whole numbers and rounding down would overstate the benefit. An NNT of 19.2 becomes 20, meaning about 20 people need treating for one to benefit over the stated period. The Cochrane Handbook recommends rounding up for this reason, and a poster should show the unrounded value too.
Is the number needed to treat the same in every population?
No. It depends on baseline risk. A treatment with the same relative effect has a lower NNT in a high risk group and a higher one in a low risk group. Applying an NNT from a trial to patients at different risk can mislead, so clinicians often recalculate it using their own patients' expected risk.
Can NNT be used for continuous outcomes such as pain scores?
Only after turning the outcome into a yes or no, such as the share of people achieving a clinically important improvement. The NNT is then calculated from those proportions. That step depends on the chosen threshold, so the threshold should be stated and justified on the poster.
How is the number needed to treat shown with a confidence interval?
Calculate the interval for the absolute risk reduction first, then take the reciprocal of each limit. An absolute reduction of 5 percentage points with an interval of 1 to 9 points gives an NNT of 20 with an interval of about 12 to 100. Intervals that include no difference need extra care in wording.
Make one with posters
The button opens the generator with this use case already described. Change the wording to match your own.
Create a poster with OneCraftRelated questions
- What is a rate ratio?What is a rate ratio: the incidence rate in one group divided by the rate in another, using person time. How it differs from a risk ratio and an odds ratio.
- What is an odds ratio?What is an odds ratio: the odds of an outcome in one group divided by the odds in another. A worked 2x2 example, how to read it, and why it is not a risk ratio.
- Sensitivity vs specificitySensitivity vs specificity: the share of cases a test catches and the share of healthy people it clears. Formulas, a worked 2x2 table and predictive values.
- What is intention to treat?What is intention to treat: analysing trial participants in the groups they were randomised to, whatever they received. Why it matters, and ITT vs per protocol.
Step by step in the builder: Add charts and diagrams.
Written and checked by the OneCraft team. Last checked .