Previously, we have shown that serial measurements of prostate-specific antigen (PSA) in hormone-refractory prostate cancer (HRPC) can be used to calculate an average relative velocity (rva) of PSA. Together, the level of PSA and the rva formed a two-variable model for survival time that worked at any time during the course of HRPC. Here, we have added serial measurements of hemoglobin and weight to test whether they improve the prior model based on PSA alone. Data from two Cancer and Leukemia Group B studies (9181 and 9182) on HRPC were combined to study the relationship between survival and serial measurements of PSA, serum hemoglobin, and patient weight. Altogether, there were 348 patients who could be evaluated. We used the Cox proportional hazard model for survival time with the interval censored method to accommodate time-dependent covariates, and tests for significance were two sided. Log (PSA), rva, log (hemoglobin), and log [weight (in kg)] were all significantly related to survival time during the course of HRPC (P < 3.0 x 10(-5)). Together, they formed a prognostic score based upon the relative hazard. Higher values of this score implied higher probability of death as the next observed event. Serial measurements of PSA, hemoglobin, and weight provide a prognostic score that can be applied continuously during the course of HRPC. Changes in the score may provide a reproducible measure of treatment effect.