OBJECTIVES: To construct and validate a prediction model of health utility (EQ-5D) for metastatic hormone-refractory prostate cancer (HRPCA) patients using cancer-specific health-related quality of life (HRQL) measures. METHODS: Data were obtained from a multicenter, multinational observational study of metastatic HRPCA patients conducted during 2002-2004. In addition to clinical and resource utilization, health utility (EQ-5D) and HRQL (Functional Assessment of Cancer Therapy – Prostate [FACT-P] and European Organization for Research and Treatment of Cancer Quality of Life Questionnaire [EORTC QLQ-C30]) data were collected. Predictive validity of ordinary least square (OLS), Tobit, and median regressions of various model specifications were tested using cross-validation samples. The selected specification was then further refined and tested for alternative model specifications and restrictions. RESULTS: OLS regression with both HRQL measures as individual components and patient demographics was the best performing model. It explained 58.2% of the observed EQ-5D variation in the validation sample. A model including only the prostate cancer–specific HRQL measure, FACT-P, explained 53.5% of the observed EQ-5D variation. Both models have good ability to distinguish patients with high health utility from those with low utility for cutoff points between the 20th and 80th percentiles of observed EQ-5D values. CONCLUSIONS: The prediction models developed have good predictive validity. These algorithms enable researchers to translate cancer–specific HRQL measures to health utility in metastatic HRPCA patients. The findings will help perform utility adjustments in cost-utility analyses.