Scattered throughout medical literature is the metric known as number needed to treat or NNT.
It has been claimed that in trials in which there are varying follow up times the simple NNT measure can be misleading - see here for a statistical discussion of that point from the NEJM commentary that suggests Kaplan Meier approach be used to account for varying followup time.
NNT is typically presented without a confidence interval or an error term.but apparently there are at least two methods for determining the CI of a NNT ( see here ) but according to this critique the two methods can yield markedly different results which I believe leads to the conclusion that one cannot compare NNT from one trial with a NNT from another trial.
It seems that one of the factors taken into account in the PHSTFP recommendations was the difference in the NNT from women 40-49 versus women over 50 in regard to mammograms.
No comments:
Post a Comment