Number Needed to Treat (NNT) Calculator

Enter the number of people with and without the outcome in the treatment and control groups to get the number needed to treat (NNT), or the number needed to harm (NNH) when the treatment increases the risk. The calculator gives Altman’s confidence interval, including the case where it runs through infinity, together with the absolute and relative risk reduction.

Counts

2 × 2 table of counts
Total
100
100
Total45155200

Put the exposed or treated group in the first row and the outcome of interest (the event) in the first column. Type over the labels to rename them.

Confidence level

Load a worked example

Number needed to treat (NNTB)

7exact 6.7

Treat 7 people in row 1 instead of row 2 for one fewer event. 95% CI 3.8 to 29.3.

NNT to benefit
6.67
95% CI 3.8 to 29.3
Risk difference (row 1 − row 2)
−15.00 points
95% CI −26.15 to −3.41
Relative risk
0.500
95% CI 0.287 to 0.870
Relative risk reduction
50.0%
95% CI 13.0% to 71.3%
Odds ratio
0.412
95% CI 0.205 to 0.826
Risk in each row
15.0% vs 30.0%
p (log odds ratio, z test)
0.0124
p (Pearson χ²)
0.0111
χ² = 6.452, df 1
p (Fisher’s exact)
0.0171
  • The outcome is common (over 10%), so the odds ratio is noticeably further from 1 than the relative risk. Do not describe this odds ratio as a risk ratio.

Report it like this

The risk was 15.0% (15/100) in Treatment and 30.0% (30/100) in Control; NNTB = 7 (absolute risk reduction 15.0 percentage points), 95% CI 3.8 to 29.3.

How to use the calculator

  1. Put the treatment group in the first row and the control group in the second.
  2. Put the outcome in the first column. Usually this is the bad outcome the treatment should prevent (death, relapse, infection).
  3. Read the NNT. If the treatment lowers the risk, the calculator shows the number needed to treat for one person to benefit (NNTB). If it raises the risk, it shows the number needed to harm (NNTH).
  4. Check the interval. If it runs through infinity, the data are compatible with both benefit and harm.

If your outcome is a good one (such as recovery), the risk is higher with treatment and the calculator will label the result NNTH. In that case read it as the number needed to treat for one extra recovery, or put the bad outcome in the first column instead.

NNT formula and confidence interval

The absolute risk reduction (ARR) is the risk in the control group minus the risk in the treatment group. The NNT is its reciprocal (Laupacis et al. 1988):

ARR = c/(c + d) − a/(a + b),   NNT = 1 / ARR

By convention the NNT is rounded up to the next whole number: an NNT of 6.7 is reported as 7, because you cannot treat part of a patient. The calculator shows the unrounded value as well.

The confidence interval is the reciprocal of the interval for the risk difference, here the Newcombe score interval. When that interval excludes zero, the NNT interval is simply 1/upper to 1/lower. When it includes zero, the NNT interval has two parts and passes through infinity, which Altman (1998) recommended reporting as, for example, “NNTB 3.8 to ∞ to NNTH 14.3”.

Worked examples

  • Benefit: 15 of 100 patients on treatment and 30 of 100 on control had the event (the default table). ARR = 30% − 15% = 15 percentage points (95% CI 3.4 to 26.1). NNT = 1/0.15 = 6.7, rounded up to 7; 95% CI 1/0.261 to 1/0.034 = 3.8 to 29.3.
  • Harm: 24 of 120 exposed and 12 of 130 unexposed people developed the disease. The risk rises by 10.8 percentage points, so NNTH = 9.3, rounded up to 10 (95% CI 5.1 to 49.7): one extra case for every 10 people exposed.
  • Interval through infinity: 10 of 50 on treatment and 15 of 50 on control had the event. ARR = 10 points, NNTB = 10, but the risk difference interval runs from a 26.3-point reduction to a 7.0-point increase, so the NNT interval is NNTB 3.8 to ∞ to NNTH 14.3. The trial cannot rule out harm.

Why the NNT depends on baseline risk

A treatment that halves the risk (relative risk 0.5) has very different NNTs in different patients:

  • Baseline risk 30%: risk falls to 15%, ARR 15 points, NNT 7.
  • Baseline risk 4%: risk falls to 2%, ARR 2 points, NNT 50.
  • Baseline risk 0.4%: risk falls to 0.2%, ARR 0.2 points, NNT 500.

This is why an NNT should always come with the time period and the population it applies to (for example, “NNT 50 over 5 years in adults with diabetes”), and why NNTs from trials with different baseline risks should not be compared directly. The relative risk calculator explains the relative measures in more detail.

How to report the NNT

Give the risks in each group, the NNT as a whole number, its confidence interval, the outcome and the follow-up time:

The event occurred in 15.0% of treated and 30.0% of control patients over 12 months (absolute risk reduction 15.0 percentage points, 95% CI 3.4 to 26.1), giving a number needed to treat of 7 (95% CI 3.8 to 29.3).

The Cochrane style guide asks for NNTB and NNTH rather than plain NNT, and for every NNT to be a whole number rounded up. For the odds ratio from the same table, see the odds ratio calculator.

Questions

How do I calculate the number needed to treat?
Subtract the risk in the treatment group from the risk in the control group to get the absolute risk reduction, then divide 1 by it. If 30% of controls and 15% of treated patients have the event, the absolute risk reduction is 0.15 and the NNT is 1 ÷ 0.15 = 6.7, rounded up to 7.
What is a good NNT?
There is no single cut-off. An NNT of 2 to 5 is a large effect, while preventive treatments often have NNTs of 50 or more and are still worthwhile when the outcome is serious and the treatment is cheap and safe. Compare the NNT with the number needed to harm for the treatment’s side effects.
What is the number needed to harm (NNH)?
The number needed to harm is the number of people who need to receive a treatment or exposure for one extra person to have a harmful outcome. It is 1 divided by the absolute risk increase. The calculator reports it when the risk is higher in the first row.
Why does the NNT confidence interval include infinity?
When the confidence interval for the risk difference includes zero, its reciprocal passes through infinity: one end means benefit (NNTB) and the other harm (NNTH). Report both parts, for example NNTB 3.8 to ∞ to NNTH 14.3, as Altman recommended.
Should the NNT be rounded up or down?
Round up to the next whole number. An NNT of 6.2 is reported as 7. The Cochrane style guide asks for all NNTs, including NNTH, to be whole numbers with decimals rounded up.
Can I calculate the NNT from an odds ratio or relative risk?
Only if you also know the baseline risk in the control group. With a control risk p and relative risk RR, NNT = 1 / (p × (1 − RR)). The same relative effect gives a small NNT in high-risk patients and a large one in low-risk patients.

Sources