To review the practice of evidence-based medicine with respect to drug treatment given to medical in-patients. Retrospective study. Teaching hospital, Hong Kong. Medical records of 129 consecutive patients who were admitted to the acute adult general medical ward from 1 September 1998 to 30 September 1998 were reviewed. Primary diagnoses, drug treatments prescribed, and the level of evidence (based on a literature search of randomised controlled trials and relevant studies) that supported the treatment given. For the 129 patients studied, 91 drug interventions had been prescribed on 312 occasions. Treatment that was supported by randomised controlled trials was prescribed in 162 (52.9%) cases. In 121 (38.8%) cases, patients were given standard and commonly used drugs that were not supported by evidence from clinical trials, and in 29 (9.3%) cases, the treatments given had no substantial supporting evidence. The management of some frequently encountered medical conditions was not based on trial data, because the relevant studies had not been conducted. Basing treatment on comparative efficacy results is a worthwhile goal, but there are limitations in conducting literature searches to identify relevant trials and studies. Evidence-based medical practice is not applicable in a large number of commonly encountered conditions.