Ian Roberts rightly highlights the misinterpretation of medical case studies that has encouraged some clinicians to use treatments which have not been evaluated (25 September, p 28). Unlike randomised ...