Cambridge University Press, Palliative and Supportive Care, 03(7), p. 289
DOI: 10.1017/s1478951509990216
Full text: Download
AbstractObjective:Self-reported health-related quality of life (HRQoL) is an important predictor of survival alongside clinical variables and physicians' prediction. This study assessed whether better prediction is achieved using generic (SF-36) HRQoL measures or cancer-specific (EORTC QLQ-C30) measures that include symptoms.Method:Fifty-four lung and 46 colorectal patients comprised the sample. Ninety-four died before study conclusion. EORTC QLQ-C30 and SF-36 scores and demographic and clinical information were collected at baseline. Follow-up was 5 years. Deaths were flagged by the Office of National Statistics. Cox regression survival analyses were conducted. Surviving cases were censored in the analysis.Results:Univariate analyses showed that survival was significantly associated with better EORTC QLQ-C30 physical functioning, role functioning, and global health and less dyspnea and appetite loss. For the SF-36, survival was significantly associated with better emotional role functioning, general health, energy/vitality, and social functioning. The SF-36 summary score for mental health was significantly related to better survival, whereas the SF-36 summary score for physical health was not. In the multivariate analysis, only the SF-36 mental health summary score remained an independent, significant predictor, mainly due to considerable intercorrelations between HRQoL scales. However, models combining the SF-36 mental health summary score with diagnosis explained a similar amount of variance (12%–13%) as models combining diagnosis with single scale SF-36 Energy/Vitality or EORTC QLQ-C30 Appetite Loss.Significance of results:HRQoL contributes significantly to prediction of survival. Generic measures are at least as useful as disease-specific measures including symptoms. Intercorrelations between HRQoL variables and between HRQoL and clinical variables makes it difficult to identify prime predictors. We need to identify variables that are as independent of each other as possible to maximize predictive power and produce more consistent results.