Multiple Urology Practice Cohort Validation of the Prostate Cancer Prevention Trial (PCPT) Calculator for Predicting Prostate Cancer Detection
Stephen J. Eyre, MD1, Donna P. Ankerst, PhD2, John T. Wei, MD3, Prakash V. Nair, MS3, Meredith Regan, Sc.D5,7, Geri Bueti, BS1, Jeff Tang, BS1, Mark Rubin, MD6, Michael Kearney, MD1, Ian M. Thompson, MD7, Martin G. Sanda, MD1,7
1Division of Urology, Beth Israel Deaconess Medical Center, Boston, MA, 2Department of Mathematics, Technical University; Munich, Germany, 3Departments of Urology, Epidemiology and Biostatistics, UTHSCSA, San Antonio, TX, 4Department of Urology, University of Michigan, Ann Arbor, MI, 5Depertment of Biostatistics and Camputational Biology, Dana Farber Cancer Institute, Boston, MA, 6Department of Pathology and Laboratory Medicine, Weill-Cornell Medical College, New York, NY, 7Harvard Medical School, Boston, MA
The PCPT Calculator of risk for detecting prostate cancer on biopsy was developed in a restricted clinical trial cohort wherein age, race, and PSA distribution (eg most having PSA 60; 10 ng/mL, nearly half over 70 years old) do not fully represent men referred for biopsy in standard urology practice. We sought to determine if the calculator is generalizable by evaluating its performance in a prospective, multi-site cohort of urology practices.
Patients undergoing prostate biopsy at 5 U.S. clinical sites were prospectively enrolled in this NCI/Early Detection Research Network (EDRN) cohort. Eligibility for this analysis was limited to patients undergoing their first prostate biopsy. PCPT Risk Calculator validity was evaluated by examining area underneath the receiver operating characteristic curve (AUC), sensitivity, specificity, and by calibration comparing observed versus predicted risk of detecting prostate cancer on biopsy.
Cancer incidence on biopsy was greater (43.4% versus 21.9%, p=0.001) in the EDRN validation cohort (N=645) compared to the PCPT group (N=5519). The EDRN cohort was younger, more racially diverse (87% vs 96% white), had more abnormal digital rectal exams, and higher average PSA than did the PCPT cohort (all p60;0.001). Cancer severity was worse in EDRN than in PCPT (Gleason 7 or higher 60% vs 21%, p60;0.001). The AUC of the PCPT Risk calculator was improved over PSA alone (Fig 1; AUC=.691 vs .655, p=0.009). Calibration confirmed calculator validity (Fig 2).
Differences in cancer prevalence, severity, patient age, PSA, and race between our EDRN validation cohort and the PCPT cohort underscore the importance of ascertaining calculator performance in the multi-center urology practice setting. Our findings validate utility of the calculator in a cohort that is more representative of American urology practice, and extend its applicability to the routine urology practice setting.