ABSTRACT
INTRODUCTION: Early biochemical recurrence after prostate cancer surgery is associated with higher risk of aggressive disease and cancer specific death. Many new tests are being developed that will predict the presence of indicators of aggressive disease like early biochemical recurrence. Since recurrence occurs in less than 10% of patients treated for prostate cancer, validation of such tests will require expensive testing on large patient groups. Moreover, clinical application of the validated test requires that each new patient be tested. In this report we introduce a two-stage classifier system that minimizes the number of patients that must be tested in both the validation and clinical application of any new test for recurrence. MATERIALS AND METHODS: Expressed prostatic secretion specimens were prospectively collected from 450 patients prior to robot-assisted radical prostatectomy for prostate cancer. Patients were followed for 2.5 years for evidence of biochemical recurrence. Standard clinical parameters, the levels proteolytic activity of prostate specific antigen (PSA) and the levels of PCA3 RNA, PSA RNA and TMPRSS2:ERG fusion RNA were determined in each prospective patient specimen for subsequent correlation with biochemical recurrence. RESULTS: While levels of PCA3 and PSA proteolytic activity (PPA) in prostatic secretions provided an effective pre-surgical predictor of early biochemical recurrence in prostate cancer, application of the two-stage classifier shows that only 60% of the patients need these tests. CONCLUSION: Two-stage classifiers can provide a parsimonious approach to both the validation and clinical application of biomarker-based tests. Adoption of the two-stage neutral zone classifier can reduce unnecessary testing in prostate cancer treatment.
