Central nervous system (CNS) tumours account for 20% of childhood cancers. Survivors often experience severe physical, neuropsychological and social sequelae of the disease and its treatment. Health status assessment in these individuals is an essential clinical outcome measure, yet little consensus exists regarding the optimum methodology. The influence of proxy respondents (parents, physiotherapists and doctors) and mode of administration (home and clinic) in which assessments is performed has been evaluated in a cohort of 37 survivors of childhood CNS tumours. A health-related quality of life (HRQOL) questionnaire, incorporating the Mark II and III Health Utilities Indices, was completed at home and in clinic by patients and parents. Doctors and physiotherapists completed this questionnaire plus Lansky Play-Performance and Karnofsky Performance scores. No significant differences between raters for single attribute scores occurred either at home or in clinic, although a wide range of agreement (kappa = 0.05-1.00, percentage agreement 53-100%) between observers was revealed. Most agreement occurred between parents and patients: this was greatest on home completion (kappa = 0.48-1.00, percentage agreement 53-100%). Doctors and physiotherapists agreed less on subjective attributes (emotion, cognition and pain). Better if responses were classified as normal and abnormal. Inter-observer agreement was greater for the HRQOL questionnaire than for Karnofsky and Lansky scores. Home completion of questionnaires provides a reliable, acceptable and convenient method of assessing health status.