Population-based cancer registries collect basic information on incident cases, whereas clinical databases hold a broad range of variables that make them attractive sources in epidemiological research. Quality, completeness and representativeness of the clinical database in the Danish Breast Cancer Cooperative group (DBCG) were compared with those in a complete population-based database holding detailed information on patients with breast cancer. The study included 1765 patients diagnosed in Aarhus County between 1983 and 1989. Data on tumour characteristics were of high quality. Fifty percent of patients were included in DBCG trials and a total of 83% were notified in the DBCG register, the rest having significantly poorer stage distribution and survival probabilities. However, for patients younger than 70 years of age, the difference in overall survival between all patients and those notified in the DBCG register was only 2%. Follow-up was incomplete in patients treated outside the programme, and this had a major impact on recurrence-free survival. The DBCG database holds high-quality data, but as a population-based database, it is neither comprehensive nor representative.