THE JOURNAL OF UROLOGY姞
1982 HEAD TO HEAD COMPARISON OF THE ERSPC RISK CALCULATOR AND ANN PROSTATACLASS IN 4685 PATIENTS FOR PROSTATE CANCER RISK ON BIOPSY Carsten Stephan, Berlin, Germany; Monique J Roobol*, Rotterdam, Netherlands; Henning Cammann, Klaus Jung, Kurt Miller, Berlin, Germany INTRODUCTION AND OBJECTIVES: We compared two prostate-specific antigen (PSA) and percent free PSA (%fPSA) based multivariate models for prediction of prostate cancer (PCa) risk on biopsy. METHODS: Data from 4685 biopsied men (6-10 core) with tPSA, %fPSA, age, prostate volume and digital rectal examination (DRE) status were applied to the ERSPC based risk calculator and the artificial neural network (ANN) “ProstataClass”. The screening population consisted of 3887 men (884 PCa) and the referred population included 798 men (468 PCa). All PSA and free PSA measurements were performed with the Access assays (Beckman Coulter). Data within the PSA range 1-10 ng/ml (n ⫽ 3332 ⫹ 590 patients) were additionally evaluated. The risk prediction values of both multivariate models were analysed by ROC curve analysis and the comparison of observed versus predicted PCa cases. RESULTS: Age (66 vs. 64 y), prostate volume (41.5 vs. 38 ml) and positive DRE status (34 vs. 39%) differed only slightly while PSA (4.4 vs. 6.6 ng/ml) and %fPSA (20.6 vs. 12.8%) showed larger differences between the screening and referred cohort. Areas under the ROC curves (AUC) for PSA, %fPSA, the ANN “ProstataClass” and the risk calculator in the ERSPC cohort were 0.699, 0.725, 0.729 and 0.785 (P⫽0.63 between %fPSA and ANN, all other comparisons with P⬍0.05, respectively). The AUCs for PSA, %fPSA, ANN and the risk calculator in the referred cohort were 0.688, 0.797, 0.885 and 0.831 (P always ⬍0.05). The performance of the risk calculator in the 798 referred patients showed an approximately 2-fold underestimation of the actual PCa risk with an intra class correlation coefficient (ICC) of 0.66. The “ProstataClass” ANN overestimated the real risk in the screening cohort about 1.5-3-fold (ICC: 0.54). CONCLUSIONS: The difference in the performance of both models is most likely due to the cohort composition (referred versus screened) and due to PSA and %fPSA differences. However, the risk calculator had a significant larger AUC than %fPSA in the referred cohort. The use of models in external cohorts with different compositions should be performed with caution. Source of Funding: None
1983 INITIAL DIAGNOSIS OF HIGH-GRADE PROSTATIC INTRAEPITHELIAL NEOPLASIA IS A RISK FACTOR FOR SUBSEQUENT DEVELOPMENT OF PROSTATE CANCER Michael Lee*, Ayman Moussa, Changhong Yu, Michael Kattan, J. Stephen Jones, Cleveland, OH INTRODUCTION AND OBJECTIVES: The management of patients diagnosed with high-grade prostatic intraepithelial neoplasia (HGPIN) remains controversial. Our aim was to evaluate the effect of HGPIN on initial biopsy toward future presence of prostate cancer. METHODS: From December 1997 to February 2008, 328 men had a second prostate biopsy after being initially diagnosed with HGPIN. Men with prostate cancer (CaP) or atypia were excluded. 225 men without HGPIN, CaP or atypia underwent a second prostate biopsy based on clinical suspicion. A total of 663 patients were included in the analysis. Data were prospectively obtained after IRB approval. A Cox proportional hazards model was used to estimate the effect of HGPIN on the subsequent diagnosis of CaP, after adjustment for PSA, age, presence of inflammation, abnormal DRE and number of cores taken at biopsy. Kaplan-Meier plots were generated to estimate the rates of CaP.
Vol. 183, No. 4, Supplement, Wednesday, June 2, 2010
RESULTS: HGPIN alone on initial prostate biopsy has a significant effect on the subsequent diagnosis of CaP. The hazard ratio is 1.89 (95% CI: 1.39, 2.55) for PIN versus no PIN with a p-value 0.00004, increasing with time as shown in the Kaplan-Meier curves. The estimated cancer rates by year 1, 3 and 5 years are 9%, 30% and 49% for patients with PIN and 5%, 17% and 30% for patients without PIN. CONCLUSIONS: HGPIN remains a risk factor for future development of CaP despite adjusting for other clinical indicators such as PSA and abnormal DRE.
Source of Funding: None
1984 PROSTATE CANCER RISK IN THE PSA RANGE 1-4NG/ML IN THE PROSPECTIVE SCREENING STUDY Maciej Kwiatkowski*, Daniel Seiler, Martin Baumgartner, Stephan Bauer, George Koritsiadis, Andreas Huber, Michael Kurrer, Aarau, Switzerland; Kurt Lehmann, Baden, Switzerland; Franz Recker, Aarau, Switzerland INTRODUCTION AND OBJECTIVES: The aim of this study was to investigate the cumulative risk of having prostate cancer (PCa) in low PSA range 1-4 ng/ml in the screened population. In addition the utility of free-to-total PSA ratio (F/T Ratio) was investigated in that setting. The data were gathered prospectively within an ongoing PCa screening trial setting. METHODS: 4923 men, participants of the swiss site (Aarau) of the ERSPC screening study, underwent total and free PSA test in the first screening round between 1998 and 2003. 1935 (39.3%) had PSA between 1 and 2.99ng/ml and 238 (4.8%) between 3 and 3.99ng/ml. Free PSA was measured in 1900 and 236 respectively. The TRUS guided prostate biopsy was offered for all men with PSA above 3.0 ng/ml and, as a side study protocol (implemented in the 1st screening round only) all men in the PSA range 1-3 ng/ml were offered a biopsy when F/T Ratio was below 20%. PCa incidence data were available until September 2008. The cumulative incidence rates (CIR) of PCa were calculated. RESULTS: Out of 1900 men in the 1-3ng/ml PSA range 430 (22.6%) had F/T Ratio below and 1470 (77.5%) above 20%. 84.4% (363/430) agreed to undergo prostate biopsy resulting in 46 PCa cases (12.7% detection rate). During further follow-up the CIR increased up to 22.8% for this group (98/430) whereas for the F/T Ratio above 20% the CIR of 10% (147/1470) was observed. Interval cancers outside the screening study protocol constituted 12.2% (12/430) and 13.6% (20/