OBJECTIVES The use of self-report questionnaires for the assessment of health-related quality of life (HRQOL) is increasingly common in clinical research. This method of data collection may be less suitable for patient groups who suffer from cognitive impairment, however, such as patients with brain cancer. In such cases, one can consider employing the patients' significant others as proxy raters of the patients' health-related quality of life. The authors examined the response agreement between patients with brain cancer and their significant others on a health-related quality of life instrument commonly used in cancer clinical trials, the EORTC QLQ-C30, and on a brain cancer-specific questionnaire module, the QLQ-BCM. METHODS The study sample consisted of 103 pairs of patients, with either recently diagnosed or recurrent brain cancer, and their significant others (75% spouses, 22% relatives, and 3% friends). Patients and proxies independently completed the EORTC QLQ-C30 and the QLQ-BCM at three different times. RESULTS Approximately 60% of the patient and proxy scores were in exact agreement, with more than 90% of scores being within one response category of each other. For most HRQOL dimensions assessed, moderate to good agreement was found. Statistically significant differences in mean scores were noted for several dimensions, with proxies tending to rate the patients as having a lower quality of life than the patients themselves. With the exception of fatigue ratings, this response bias was of a limited magnitude. Less agreement and a more pronounced response bias was observed for the more impaired patients, and particularly for patients exhibiting mental confusion. This finding was confirmed by longitudinal analyses, which indicated lower levels of patient-proxy agreement at follow-up for those patients whose physical or neurologic condition had deteriorated over time. CONCLUSIONS In general, patients and their significant others provide similar ratings of the patients' quality of life. Lower levels of agreement and more biased ratings can be expected among those patients for whom the need for proxies is most salient. It is argued, however, that discrepancies between patient-proxy ratings should not be interpreted, a priori, as evidence of the inaccuracy or biased nature of proxy-generated data. Future studies are needed to examine the relative validity and reliability of patient-versus proxy-generated health-related quality of life scores.