Background: Although many physicians measure serum prostate specific antigen (PSA) during the follow-up of patients with hormone refractory prostate carcinoma (HRPC), little has been done to formalize the determination of how these serial values of PSA impact on prognosis. To understand HRPC fully, make decisions about choices of treatment as well as about clinical research on treatments for HRPC patients, and design appropriate measures of PSA response, it seems that first it would be necessary to understand how these serial measures of PSA relate to survival. The purpose of this study was to determine how repeated measurements of PSA impact on the probability of imminent death for patients with HRPC.
Methods: One hundred forty-eight men with HRPC were enrolled in Cancer and Leukemia Group B Study 9181, in which they were treated with either a low dose (160 mg/day) or a high dose (640 mg/day) of megestrol acetate (MA). Because preliminary data analysis indicated that these treatments had no effect on survival, the authors pooled the data to analyze the overall dynamics of PSA and survival during the follow-up period. The authors attempted to correlate initial and monthly PSA measurements, which were mandated by the study protocol, with the probability of death at any time during follow-up. For statistical analysis, the Cox proportional hazards model and the general linear model were used. In addition to the level of PSA, the authors used the relative velocity of PSA, which was defined as (dy/dt)/y, with "y" symbolizing serum PSA and "t" symbolizing time.
Results: Both log(PSA) and the average relative velocity of PSA (rva) were significantly correlated with survival time (P=0.0001 and P=0.0008, respectively), and the analysis performed with the Cox proportional hazards model yielded the following formula for a PSA hazard score: PSA hazard score =0.251*(log(PSA) - mean log(PSA)) + 24.5*(rva - mean rva) This hazard score tended to be higher for patients who were about to die. For example, there was a close correlation between the hazard score and the probability of death as the next observed event. Furthermore, the hazard score provided a dynamic measure of how PSA was affected by treatment.
Conclusions: The average relative velocity of PSA has been identified by the authors as a new measure of the dynamics of PSA in HRPC. It can be determined from sequential values of PSA. This average, together with the log(PSA), are significantly related to the probability of imminent death.