Résumé :
|
[BDSP. Notice produite par INIST r0M1IR0x. Diffusion soumise à autorisation]. The need to evaluate drugs'effects in real clinical practice is increasingly important. Randomized clinical trials (RCTs) and database analyses (DBA) are the two main methods to assess treatments effectiveness. RCTs remain the "gold standard" for comparing altemative treatments. However, they are conducted under strict, protocol-driven conditions that may limit their generalizability. Advantages of new high quality clinical databases, on the other hand, include the simple and economic access to large number and range of cases, and the ability to capture all aspects of actual medical practice. The main potential limitation of DBA is the potential for comparison bias due to the lack of randomization. Despite the efforts to design naturalistic trials and to use sophisticated statistical techniques to minimize selection bias, the inherent limitations of both methods (problems of extemal and intemal validity, respectively) have not been completely solved. Thus, the actual challenge is the development of some new strategy capable of generating results with an acceptable balance between internal and external validity. As randomization is essential to minimize comparison bias, we point out the possibility to include randomization modules in computer-based patient records. The theoretical foundation of these "randomized database studies" is the simultaneous use of both experimental and observational methods in the assessment of drugs'effectiveness. (...)
|