Abstract
Clinical trials now often involve thousands of patients, and statisticians emphasize the importance of trial size in ensuring that 'correct' answers are obtained. However, when a good treatment appears for a disease that was hitherto untreatable - for example, oranges for scurvy or streptomycin for tuberculosis - only a small trial is needed. Large trials are only needed to demonstrate small effects. The meta-analysis of small trials is often misleading, and may hide undesirable effects of individual drugs. The concept of equivalence between treatments is important, and while a statistically adequate equivalence trial may have to be very large, many clinicians will question the need for extreme statistical propriety. Clinical trials often do not reflect 'real world' practice, and the clinical relevance of a trial is more important than its size.
