OBJECTIVES: Population-based registries provide insights into quality of care and inform reimbursement decisions. This study aims to investigate whether registries are a suitable tool for outcomes research in assessing drug use and real-world cost-effectiveness in cancer. METHODS: We used four Dutch population-based registries to conduct outcomes research. Patients for the registries were included regardless of prognosis or treatment: 55% and 40% of all Dutch patients in metastatic renal cell cancer (mRCC) and three haematological cancers, respectively. Data were retrospectively collected at several points in time from medical records and hospital information systems on baseline characteristics, treatments, dosages, treatment response, survival, adverse events and resource use. All patients entered the registry at time of diagnosis. RESULTS: Our registries contained information of 615 mRCC and 3093 haematological cancer patients (non-Hodgkin, multiple myeloma, and chronic lymphocytic leukaemia). They provided important information about how patients, including those regularly excluded from clinical trials, are treated in daily practice. However, important data, including prognostic information, was commonly missing (e.g. 40-55% missing performance status). Furthermore, patients treated with the drug of interest were not comparable to patients not treated with this drug. Moreover, only small numbers of patients received the drug of interest (mRCC: N=34; non-Hodgkin: N=35), and many patients received different drugs in various combinations and treatment sequences in haematological cancers. This, in combination with the inability to fully correct for confounding, complicates the estimation of a real-world incremental cost-effectiveness estimate. CONCLUSIONS: Our registries provided important information to physicians and policymakers to enhance quality of care and facilitate evidence-based decision making. Although population-based registries include high numbers of patients, it remains a challenge to obtain sufficient numbers of similarly treated and comparable patients. Therefore, it is inevitable to use data synthesis in combination with comprehensive modelling techniques to obtain valid real-world incremental cost-effectiveness estimates.