Statistics - explanations and formulas

Number Needed to Treat - including how to calculate

What is a “good” NNT?  What is an “acceptable” NNH?  There are no absolute answers to these questions.  For any NNT, the magnitude of the treatment effect needs to be taken into consideration as well as any associated side effects.  For any NNH, the number of patients exposed and the severity of the hazard both need to be taken into account.

When it comes to an NNT, smaller is going to be better.  However, even an NNT in the thousands may make an intervention worthwhile if there are few associated side effects, the cost is relatively low, and the potential benefit is great.  Similarly, for an NNH, larger is always going to be better.  However, an NNH in the single digits might be acceptable if the harm is negligible.  Alternatively, even a large NNH may not be acceptable if the severity of the harm is great relative to the expected benefit, or a very large number of patients will be exposed.

For any intervention, the associated NNTs and NNHs must be weighed against each other, taking into account how many individuals will be exposed plus the frequency and magnitude of all potential benefits and harms.  Only after weighing all of these factors in aggregate can a clinically relevant decision be made.


The number of patients over a given time period that need to be treated with the experimental treatment compared to the control treatment in order to prevent one bad outcome.

Absolute difference = Budesonide Event Rate (CER) - Acetazolamide Event Rate (EER) = 73% - 43% = 30%

NNT = 1/.30 = 3.3 (round up to 4)

NNT=1/absolute difference