Comparison of Preoperative Prostate Specific Antigen Density and Prostate Specific Antigen for Predicting Recurrence After Radical Prostatectomy: Results from the Search Data Base

Abstract

Prostate specific antigen (PSA) density based on the surgical weight of the radical prostatectomy specimen has previously been shown to be an independent predictor of biochemical recurrence after radical prostatectomy. We determined whether preoperative PSA density calculated using transrectal ultrasound prostate volume was a better predictor of advanced pathological findings or biochemical recurrence after radical prostatectomy relative to PSA. We examined 552 men from the newly established Shared Equal Access Regional Cancer Hospital data base of men treated with radical prostatectomy at equal access medical centers to determine whether preoperative PSA density was a significant predictor of an adverse pathological condition or PSA recurrence after radical prostatectomy. Models using PSA density were compared with models using PSA to determine whether PSA density improved risk stratification relative to PSA. PSA density was examined as a continuous and a categorical variable using cutoffs to separate patients into groups at different risks for PSA failure. PSA density and PSA were significant predictors of adverse pathological findings on univariate analysis. Using PSA density in the multivariate model resulted in slightly better but statistically insignificant improvement in prediction of positive surgical margins (p = 0.134) and extracapsular extension (p = 0.771) relative to using PSA in the model. Neither PSA nor PSA density were significant independent predictors of seminal vesicle invasion. Area under the ROC curves for predicting biochemical recurrence for PSA and PSA density were not significantly different (0.589 and 0.58, respectively, p = 0.691). On separate multivariate analyses PSA density and PSA were significant independent predictors of biochemical failure. The multivariate model using PSA density provided only slight improvement in risk assessment relative to the model using PSA (index C = 0.589 and 0.581, respectively). To determine whether using PSA density as a categorical variable would result in improved prognostication we evaluated PSA density to determine the cutoff points that would provide the greatest risk stratification. PSA density cutoffs of less than 0.4, 0.4 to 1 and greater than 1 ng./ml./cc separated patients into 3 distinct groups at increasing risk for biochemical failure (p <0.001). While these cutoffs provided better risk stratification than when PSA density was examined as a continuous variable (index C = 0.684 versus 0.58), they provided only marginal improvement relative to the standard PSA cutoffs of less than 10, 10 to 20 and greater than 20 ng./ml. (index C = 0.676). The use of preoperative PSA density relative to PSA provided only slight improvement for predicting adverse pathological findings and biochemical recurrence after radical prostatectomy. The minimal and statistically insignificant improvement in preoperative risk assessment provided by PSA density does not justify the time and effort necessary to calculate this value.