Quality of life (QL) assessment is an increasingly important component of clinical research, especially with cancer patients. The literature strongly supports the view that QL should be assessed by the patient rather than the clinician. While clinical parameters such as performance status or toxicity ratings may bear some relationship to QL, they are not a substitute for its measurement. In spite of these observations, clinicians have been reluctant to accept the need for patient-rated measures of QL. In this paper, data from a sample of 109 newly-diagnosed breast cancer patients were used to examine the relationship between expert-rated measures and a patient-rated measure of QL; to determine whether the Cancer Rehabilitation Evaluation System (CARES), an instrument for assessing the rehabilitation needs of cancer patients, is a measure of QL; to explore whether there are any medical, social or demographic variables which the clinician can use to predict how patients assess their QL; and to determine which variables (expert-rated scales, medical, social or demographic variables, or rehabilitation needs) have the most effect on how patients evaluate their QL. In this sample, patient ratings of QL were widely distributed and were only moderately correlated with the expert-rated Karnofsky Performance Status (r = 0.53) and Global Adjustment to Illness Scale (r = 0.59). In addition, there were no significant correlations between important clinical variables (axillary node status, type of surgery, receipt of chemotherapy) and patient-rated QL. Among the clinical variables and instruments studied, the Global CARES score demonstrated the best correlation (r = -0.74) with the patient-rated assessment of QL. A stepwise multiple linear regression procedure was performed with QL as the dependent variable in order to identify which factors accounted for the most variance in patient assessment of QL. The potential predictor variables used in this procedure were chosen from among those that would be available to a clinician. The Global CARES score was the best single predictor of QL, accounting for 55% of the variance, followed by Karnofsky Performance Status, the Medical Interaction and Sexual summary scales of the CARES, and the patient's educational status. Data from the CARES provided additional descriptive information about the type and frequency of rehabilitation problems experienced by these patients in relation to their ratings of QL.(ABSTRACT TRUNCATED AT 400 WORDS)