Can you compute number needed to treat (NNT) in a cohort study, and what information is required?

• A. NNT = 1 / (risk difference between treated and untreated); requires absolute risk reduction, i.e., event probabilities in exposed vs unexposed over a specified time.
• B. NNT = RR / RD
• C. NNT cannot be computed from cohort data
• D. NNT equals the incidence in treated minus untreated

Answer: (A)

The key idea is that NNT tells you how many people need a treatment to prevent one additional adverse event over a defined time. To get it from a cohort study, you need the absolute risks in both groups over that same time period. Specifically, you estimate the risk of the outcome in the treated group and the risk in the untreated (or control) group. The absolute risk reduction is the amount by which the untreated risk exceeds the treated risk (positive for a beneficial treatment). Then you take the reciprocal of that ARR: NNT = 1 / ARR. This relies on having actual probabilities (risks), not just relative measures, and it requires specifying the follow-up period.

So, from cohort data you compute NNT by using the difference in event probabilities between treated and untreated over a defined time, then taking the reciprocal. The other options aren’t correct because they either rely on an incorrect formula, misstate what can be derived from cohort data, or use a different measure from what defines NNT.