Randomized controlled trials (RCTs) in palliative cancer care often experience methodological problems. In this paper we discuss issues of major concern, including recruitment, patient attrition and compliance, arising from an RCT that compared comprehensive palliative care to conventional care. The main criteria for trial entry were incurable malignant disease and a survival expectancy of between 2 and 9 months. Patients' health-related quality of life (HRQL), self-assessed by multi-item questionnaires, was a defined endpoint. The planned number of patients was successfully recruited, although the patients were referred late in the course of their disease so that follow-up tended to be short. Compliance in completing HRQL questionnaires was good up to 1 month before the patient's death; but in the final weeks it was found to drop substantially. Based on our experience, recommendations are given for those planning similar research. Procedures for improving patient recruitment are suggested, stressing the need for local data management, repeated information to referral sources, extensive screening for potentially eligible patients and simple referral routines. Precise inclusion criteria, including prognostic factors other than physicians' estimates of life expectancy, should be used to ensure a sufficient follow-up period. For HRQL assessment, multi-item questionnaires can achieve excellent compliance up to 1 month before patients' death, but in order to evaluate the very final weeks of life we recommend the use of simpler methods.