Can treatment trial samples be representative?

A catchment area-based sample of patients recruited for an eating disorder treatment trial was compared with patients from the same geographical area seen in the 12 months before and after the trial. The three samples were very similar. The research sample was representative of the usual clinic sample from which it had been selected and thus the results could be extrapolated with some confidence to other similar clinical settings. It is concluded that whilst treatment trials, by their very nature, have explicit and implicit inclusion and exclusion criteria with appropriate designs they can be usefully representative.