Critical evaluation: Usefulness and NNT/NNH

Last updated: Monday, October 03, 2022

The number needed to treat (NNT) is an expression that literally describes the number of patients that we would need to treat with anotheraban for 2 years to prevent one stroke. It is calculated as shown below, remembering that if you’ve been using percentages throughout the calculation then use 100 as the numerator to make the maths work. NNTs are normally rounded up to whole numbers:

NNT = 1 ÷ ARR = 1 ÷ 0.02 or 100% ÷ 2% = 50

This means that 50 patients need to be treated with anotheraban for 2 years to prevent one stroke.

So what does this mean in terms of usefulness? In general the smaller the NNT the better the treatment, with 1 being the ideal NNT. But this doesn’t mean we should necessarily reject a drug with a high NNT. There are no rules about what an acceptable maximum NNT would be, and it will depend on several factors, such as the severity of the condition being treated, costs, side effects and individual values and preferences. We also need to realise that NNTs are open to interpretation. So some doctors may see an NNT of 50 over 2 years as a considerable benefit, whereas others may see the benefit as small.

Comparing NNTs can help us choose between drug treatments. So, if anotheraban has an NNT of 50 while yetanotheraban has an NNT of 35, we might choose yetanotheraban as it seems more effective.

However it is important to balance effectiveness or benefits of treatments against safety or potential harms. For this we can look at the number needed to harm (NNH).

In the same trial, 2 patients in the placebo group and 82 in the anotheraban suffer from life-threatening bleeding. We can describe how many patients we would need to treat for one to suffer from major bleeding (harm) using the number needed to harm (NNH). We first need to work out the absolute risk increase (ARI) of major bleeding:

ARI = AR anotheraban group - AR placebo group

ARI = (82 ÷ 2000) – (2 ÷ 2000) = 0.04 or 4%

Then the calculation is similar to that used for NNTs:

NNH = 1 ÷ ARI = 1 ÷ 0.04 or 100% ÷ 4% = 25

NNHs are normally rounded down. A higher NNH is generally more favourable, but as with NNT, there are no set rules about what an acceptable value might be.

Commissioners, or the people buying a service in which anotheraban was to be used, would need to consider these calculations carefully. Is it worth treating 50 patients for 2 years to prevent one stroke when for every 25 patients treated, one of them will have a major bleed?