OBJECTIVES: Investigating and comparing the sources of prevalence data used in German drug reimbursement dossiers within the AMNOG-framework in order to explore causes for deviating prevalence estimates. METHODS: All publicly available dossiers on diabetes mellitus type II were retrieved from the Federal Joint Committee’s homepage (http://www.g-ba.de). We investigated the epidemiological sources in the dossiers on their representativity (i.e., population-level epidemiological data), actuality (year of data collection), method of data collection, data quality, and open access. RESULTS: As of early June 2014, 13 dossiers were published on diabetes mellitus type II. Overall, 23 different sources for quantifying the prevalence were used, which can be grouped into 5 categories: clinical trials, registries, health insurances’ claims data, commercial data-providers, and a governmental data-provider. Data was mostly collected 5 or more years ago (except for the commercial and governmental data-providers). Data was retrieved by means of physicians’ statements; patient surveys with voluntary disclosure; diagnostic tests (such as the oral glucose tolerance test); national registries; secondary data from health insurances, and (raw) secondary data of pharmacy prescriptions with physician diagnoses. Besides differences in diagnostic criteria, reasons for deviation among sources include: the sample size and making adjustments to improve representativity for Germany; the chosen approach to analyze secondary data; differentiation of type II from type I diabetes mellitus, and (not) considering undiagnosed or untreated cases. While accessing the data of publications and of the commercial providers requires a payment, health insurance data is not freely available in Germany (yet). The governmental data-provider is accessible without restrictions and free of charge. CONCLUSIONS: The need for accurate population-level data and access to it is highlighted by the present case-study. Sources for prevalence data in diabetes mellitus type II were characterized by a variety of differences in the methods applied to derive (the) data, leading to deviating estimates.