Randomised studies--useful for whom?