Evaluering av medisinsk behandling: Hva vet vi, og hvordan bør kunnskapen håndteres?

Abstract

Data on the effects of medical therapies are available from clinical studies. Well planned and executed randomised, placebo-controlled studies are given the highest level of evidence when study results are reviewed. The available body of data on different therapies lend themselves to synthesis in clinical overviews or therapy recommendations that regularly achieve status as normative for clinical practice. This warrants caution and rigour as far as the underlying procedures are concerned. The current review encompasses a brief discussion of previous failures of pharmacovigilance, including the thalidomide disaster and the COX-2 debacle. The quality criteria for data evaluation and the methodological basis for meta-analyses are presented. Possible sources of error in meta-analyses, such as different reporting of effects, publication bias, heterogeneity, conflicts of interest and problems with evaluation of side effects, are discussed. The advantages of meta-analysis are exemplified with data on therapy of lower back pain, which recently has been subjected to clinical practice guidelines that, in our view, have several shortcomings. Meta-analyses have significant advantages over less formalised methods for the evaluation of the current state of knowledge of therapeutic alternatives.