Semantic sMatter #3

There are a plethora of terms that describe statistical outcomes out there in medical literature land. One of the more practical and clinically helpful is the Number Needed to Treat, or NNT.

Do you know what it means?

[su_tabs]
[su_tab title=”Terminology”]

What is meant by “Number Needed to Treat” and what are the considerations when interpreting it?

[/su_tab]
[su_tab title=”Explanation”]

The Number Needed to Treat (NNT) is an expression of the average number of patients that are needed to receive an intervention in order to prevent one additional adverse outcome, or, the number of patients that are needed to receive an intervention to have one extra patient benefit compared to a control group of similar patients. (The corollary, number needed to harm is the number of patients that are needed to receive an intervention in order to prevent one additional beneficial outcome.)

The ideal NNT is 1 => Everyone benefits with the intervention; no one benefits from being in the control group.

The NNT is the inverse of the absolute risk reduction and so is calculated by NNT = 1/Absolute risk reduction. So watch out for trials that only quote the relative risk reduction and do not at least provide the numbers to work out the absolute RR.

The NNT is both time and population dependant:

If the NNT for an intervention is 10 over 5 years, then it will be 50 over 1 year and 2.5 over 20 years
The NNT is what it is for the population it was calculated for. So if the outcome being measured is mortality and your population’s mortality is different from that of the trial’s control population, then you cannot be sure what the NNT for your population will be.

[/su_tab]
[su_tab title=”References”]