BACKGROUND.
The pretreatment serum prostate specific antigen level (PSAL) is the most significant predictor of biochemical and local failure in patients treated with definitive radiotherapy. The role of prostate specific antigen (PSA) density (PSAD) relative to PSAL for such patients is controversial. In this article, we describe a comparative analysis of the prognostic value of PSAL and PSAD.
METHODS.
The study was comprised of 365 patients who were treated with external beam radiotherapy for regionally localized (Tl-T4, Nx, MO) adenocarcinoma of the prostate between 1987-1993 and in whom PSAL and pretreatment prostate volume by transrectal ultrasound were available. The mean and median doses were 66.8 Gy and 66 Gy. Median follow-up for those patients living was 27 months. The mean PSAL was 12.7 ng/mL with a median of 9.1 ng/mL. PSAII was calculated by dividing the PSAL by the pretreatment prostate transrectail ultrasound volume (in cc). The mean PSAD was 0.44 and the median was 0.31. Biochemical failure was defined as two consecutive increases in follow-up PSAs, one increase by a factor of greater than 1.5 or an absolute increase of greater than 1 ng/mL.
RESULTS.
The distributions of PSAD and PSAL were similar and were positively skewed. When log-transformed, the distributions of both parameters were nor - malized and linear regression revealed a high correlation ( P < 0.001). PSAD was significantly associated with several potential prognostic factors, including stage, Gleason score, PSAL, and pretreatment prostatic acid phosphatase. Univariate analyses of PSAD and PSAL revealed that these were the most significant correlates of local and biochemical control. By contrast, stage and Gleason score were the only factors predictive of freedom from distant metastasis. Multivariate analyses using Cox proportional hazards models were then performed to determine if PSAD provided prognostic information independent of PSAL. The manner in which PSAD and PSAL were categorized (four, three, or two groups) dramatically influenced the significance of these covariates. Optimization of these groups for significance showed grouped PSAD to be the stronger predictor of local control and grouped PSAL to be the stronger predictor of biochemical control. However, when PSAD and PSAL were used as continuous variables, PSAL was the only independently significant prognostic factor for local or biochemical control.
CONCLUSIONS.
PSAD was highly correlated with actuarial patient outcome in the univariate analyses and appeared to provide independent information when PSAL was between 4 and 20 ng/mL. However, the differences based on PSAD were relatively small and, therefore, not clinically useful.