One more comment on the relevance of time and Kaplan-Meier curves to the NNT and its interpretationl

Two trial examples. EMPA-REG OUTCOMES and REDUCE-IT. If you don't have access to the NEJM articles, here is a REDUCE-IT trial document from Amarin and here's a good summary presenation from EASD 2015 for EMPA-REG OUTCOMES. Both should be freely available. Notice how the placebo and treatment arms didn't really start to separate from one another for the 3-point MACE secondary outcome until after 1.5 years of treatment with Vascepa in REDUCE-IT, but incidence of 3-point MACE in the empaglozin group in EMPA-REG OUTCOMES started separating from the placebo group after only ~ 3 months. One needs to look at the data over time in a Kaplan-Meier curve analysis in order to appreciate this. 

This greatly influences the interpretation of the NNT. Even though the median duration of treatment in EMPA-REG OUTCOMES was 2.6 years, benefits were observed very early on in the first few months. However, in REDUCE-IT patients had to be treated for much longer (~1.5 years) before the beneficial effect was observed (median treatment period 4.9 years). NNTs are often expressed at the time of median dosing. However, in the case of EMPA-REG OUTCOME beneficial effects were observed much earlier than the 2.6 years of median dosing. Furthermore, the medical experts that look at this trial data can often infer what physiological process(es) may have been affected to reduced disease incidence based upon how quickly the placebo vs. treatment curves separate.

BearDownAZ