Internal validity of trials is more important than generalisability

EDITOR,—Mike Pringle and Richard Churchill raise important issues relating to the practicalities of performing and interpreting the results of randomised controlled trials in general practice.1 They suggest that observational studies are acceptable when a randomised study is too difficult to perform or when its results will be clinically meaningless, reminding us of the complementary (rather than competing) roles that experimental and observational methods should have in the evaluation of health care interventions.2 We are concerned, …