[How evidence-based is medicine? A systematic literature review]

In the early 1990es, it was supposed that only 10-15 per cent of medical interventions were based on results from randomised controlled studies. A systematic review of available empirical studies on the topic was performed in order to elucidate to what extent interventions in different medical specialities are evidence-based.

Literature search in MEDLINE for the period 1995-2002 retrieved 15 published studies, which investigated to what extent interventions in individual patients were based on randomised controlled studies. The retrieved papers were read critically in order to extract data concerning the patients, their clinical problems, and the interventions performed. Finally, the classification of the clinical scientific methods, which was the evidence-base of the interventions, were extracted.

In general internal medicine over 50 per cent of the interventions seems warranted by randomised controlled studies, whereas the figures are somewhat lower for general practice, surgery, and dermatology. A small study shows that psychiatry may candidate for the top score with 65 per cent of the interventions based on randomised controlled studies. However, more studies are requested to confirm this finding.

It is suggested to perform regular studies of the evidence-base for clinical interventions in different specialities, preferably with assistance from colleagues from other specialities. There should be sufficient access to literature databases and databases of "digested" evidence, which was not always the case in the reviewed studies. It is important that the indication for an intervention has been evaluated in randomised controlled studies, which compared a possible intervention to observation without treatment or treatment with placebo--a mere comparison in randomised controlled studies of two active interventions may not be sufficient evidence. Authors are encouraged to agree on a comparable classification scale for grading the evidence in publications. It is concluded that there is a need for this kind of regular audit in order to safeguard that patients are offered interventions based on the best evidence.