Abstract: Problems arising from treatment comparisons within non-randomized observational studies are well known, and the conduct of randomized clinical trials (RCTs) is the widely accepted gold standard in this area. In systematic reviews performed by the Cochrane Collaboration it is current practice to include only results of RCTs. Since, for various reasons, randomization of patients to treatment groups is not always possible, the question arises under which circumstances evidence gained from non-randomized patients might also be included into systematic reviews. In situations where consent to randomization is expected to be a problem the Comprehensive Cohort Study (CCS) design may be used [1,2]. In a CCS all patients fulfilling the clinical eligibility criteria are recruited. Patients not accepting randomization choose their preferred treatment and are treated and followed according to the study protocol in the same way as the randomized patients. This design was adopted in three clinical trials of the German Breast Cancer Study Group (GBSG) including more than 2000 patients. Our experiences with this design will be described and it will be demonstrated how the evidence from results based only on randomized patients may be extended by incorporating data from non-randomized patients. The CCS design may be considered as a first step towards the inclusion of information from non-randomized patients into systematic reviews. References: 1. Schmoor C, Olschewski M, Schumacher M: Randomized and non-randomized patients in clinical trials: Experiences with comprehensive cohort studies. Statistics in Medicine 15, 263-271, 1996. 2. Silverman WA, Altman DG: Patients' preferences and randomized trials. Lancet 347, 171-174, 1996.