A systematic review of 3 randomized trials with highly varied NNTs should
not result in a single NNT with a narrow confidence interval.
Assuming the 3 trials are proper to combine in the first place (similar
treatment, similar patient population, similar duration, etc.), the
differences between the 3 trials are likely differences in baseline risk or
time course of the underlying disease at initiation of the trial.
In this case, the relative risks may be similar enough across trials that
one could do a systematic review and define the relative risk with greater
certainty (narrower confidence intervals). Then that relative risk could be
used to define NNTs at different levels of baseline risk/absolute risk.
But it would not be appropriate to combine disparate NNTs/absolute
risk/baseline risk into a single meta-analysis and conclude a single
NNT/absolute risk difference across the spectrum represented in the trials.
Because of the clinical and EBM push to get to NNT and use absolute risk
instead of relative risk (a useful concept when applied appropriately), it
may drive improper meta-analysis.
Brian S. Alper MD, MSPH
Editor-in-Chief, DynaMed (http://www.DynamicMedical.com)
Medical Director, DynaMed, LLC
10 Estes St.
Ipswich, MA 01938
office (978) 356-6500 extension 749
cell (978) 874-0719
fax (978) 356-6565
home (978) 356-3266
"It only takes a pebble to start an avalanche."
|