Background: Meta-analysis is now widely used in order to increase the
power of individual clinical studies. Important far-reaching conclusio
ns have been derived by pooling the results of studies which in isolat
ion would not be large enough to reach definitive conclusions. While t
he statistical power is thereby amplified, so is the potential for err
or. Objective: To assess the potential and the pitfalls of meta-analys
is as a guide to clinical practice. Conclusions: Conclusions derived f
rom meta-analysis may be influenced by unrecognized selection bias and
heterogeneity of studies included. Publication of the results may be
in a form which does not lend itself readily to critical analysis and
misleading results may therefore be accepted.