A defence of the small clinical trial: evaluation of three gastroenterological studies.

Emphasis has been placed on the importance of ?, the type II error, in clinical trials and the corresponding necessity of including large numbers of subjects in a study.1 Large trials are in fashion and small trials?even when they yield statistically significant results?are regarded with suspicion; a small trial which does not show a conventional statistically significant difference is often thought worthless. Interpretation of the small trial has increasingly been recognised to be dangerous, and the argument that lack of a significant difference between two treatments must mean that they have equal efficacy is recognised to be false. The need for large trials is undeniable in diseases such as cancer and cardiovascular prob? lems where earlier trials have excluded large advantages for new treatments and small differences in outcome are still clinically and socially important. The unfortunate consequence of the emphasis on large trials in these settings is that small trials conducted in other diseases where only large differences in outcome are relevant have illogically become unacceptable in many people's minds. With appropriate statistical analysis, including calculation of the confidence limits for the likely "true" difference, small controlled trials can provide clinically useful information. We present three such small clinical trials?none of which show conventional