Phil B.FontanarosaMD, Deputy EditorIndividualAuthorStephen J.LurieMD, PhD, Contributing EditorIndividualAuthor
Copyright 2000 American Medical Association. All Rights Reserved.
Applicable FARS/DFARS Restrictions Apply to Government Use.2000
To the Editor: Dr LaRosa and colleagues1 have calculated the number needed to treat (NNT)
from pooled estimates of the absolute risk differences from 5 trials of statins.
The NNT is commonly used to summarize the effects of treatment in a clinically
relevant way, taking into account both baseline risk without treatment and
the risk reduction achieved with treatment. However, the NNT derived from
pooled estimates of treatment effects may be seriously misleading.2 This is because the NNT is sensitive to factors
that change baseline risk, including the outcome considered, characteristics
of patients included in trials, secular trends in incidence and case-fatality,
and the clinical setting. In the statin trials included in the meta-analysis,
baseline coronary heart disease mortality risk, estimated from the control
group event rates, varies by more than an order of magnitude (Table 1), making the pooled NNT uninterpretable.
Ebrahim S, Smith GD. Statins and Risk of Coronary Heart Disease. JAMA. 2000;283(22):2935. doi:10.1001/jama.283.22.2931