Can prostate specific antigen derivatives and pathological parameters predict significant change in expectant management criteria for prostate cancer?

PURPOSE: A prior report established that pretreatment criteria based on clinical and biopsy pathology parameters can predict men who harbor small volume prostate cancer who might be followed expectantly. However, some of these men will exhibit disease progression with time and will need definitive therapy. To detect those in whom disease may progress, repeat prostate biopsies are performed at yearly intervals. Therefore, we determined whether biomarkers could be used to determine those in whom disease is likely to progress and thus those who require definitive therapy.
MATERIALS AND METHODS: Initial and repeat biopsy information along with transrectal ultrasound measurements of gland volume, total prostate specific antigen (PSA), %free PSA (%fPSA) and total PSA velocity were evaluated in 78 men, 45 from the prior study, in whom disease was being managed expectantly. Univariate and multivariate logistic regression analyses determined variables that predicted a favorable tumor burden based on biopsy pathology status at each subsequent repeat biopsy. A Cox proportional hazards model was produced using 67 of 78 evaluable cases having adequate temporal data to predict hazard ratios for conversion from favorable to unfavorable tumor burden status.
RESULTS: At time zero for 78 patients %fPSA, total PSA, and gland volume univariately and multivariately differentiated unfavorable and favorable tumor burden groups (p <0.05). The receiver operator characteristic area under the curve (ROC-AUC) was 83%. At the first followup biopsy 17 of 67 (25.4%) men converted to unfavorable tumor burden status. The %fPSA, PSA velocity and gland volume univariately distinguished these 2 groups (p <0.05) with 82% ROC-AUC. At second repeat biopsy 6 of 36 (16.7%) men converted to unfavorable tumor burden status and the ROC-AUC was 76%. Of the 14 men who had a third repeat biopsy all demonstrated favorable tumor burden status. A Cox proportional hazards model stratified the 67 of 78 men into high (48) and low risk (19) groups based on %fPSA at a 20% cutoff (p <0.01). Classification and regression tree analysis using logistic regression multivariately selected variables predicted favorable tumor burden status with an accuracy that ranged from 75% to 84% during our study.
CONCLUSIONS: PSA velocity, %fPSA and gland volume information improves the prediction of men undergoing expectant management who are more likely to have small volume disease based on a 12-core biopsy interpretation within the time of our observations. %fPSA proved to be a valuable marker to stratify the 2 risk groups. Therefore, based on these factors it may be possible to consider deferment of repeat prostate biopsy until adverse results are detected. This rational approach to the management of prostate cancer in older men with small volume cancer seems to be a reasonable strategy.