The EMPaCT Classifier: A Validated Tool to Predict Postoperative Prostate Cancer-related Death Using Competing-risk Analysis

TY - JOUR

T1 - The EMPaCT Classifier: A Validated Tool to Predict Postoperative Prostate Cancer-related Death Using Competing-risk Analysis

AU - Tosco, Lorenzo

AU - Laenen, Annouschka

AU - Briganti, Alberto

AU - Gontero, Paolo

AU - Karnes, R Jeffrey

AU - Bastian, Patrick J

AU - Chlosta, Piotr

AU - Claessens, Frank

AU - Chun, Felix K

AU - Everaerts, Wouter

AU - Gratzke, Christian

AU - Albersen, Maarten

AU - Graefen, Markus

AU - Kneitz, Burkhard

AU - Marchioro, Giansilvio

AU - Salas, Rafael Sanchez

AU - Tombal, Bertrand

AU - Van den Broeck, Thomas

AU - Van Der Poel, Henk

AU - Walz, Jochen

AU - De Meerleer, Gert

AU - Bossi, Alberto

AU - Haustermans, Karin

AU - Van Poppel, Hendrik

AU - Spahn, Martin

AU - Joniau, Steven

AU - European Multicenter Prostate Cancer Clinical and Translational Research group (EMPaCT)

N1 - Copyright © 2017 European Association of Urology. Published by Elsevier B.V. All rights reserved.

PY - 2018/4

Y1 - 2018/4

N2 - BACKGROUND: Accurate prediction of survival after radical prostatectomy (RP) is important for making decisions regarding multimodal therapies. There is a lack of tools to predict prostate cancer-related death (PCRD) in patients with high-risk features.OBJECTIVE: To develop and validate a prognostic model that predicts PCRD combining pathologic features and using competing-risks analysis.DESIGN, SETTING, AND PARTICIPANTS: This was a retrospective multi-institutional observational cohort study of 5876 patients affected by high-risk prostate cancer. Patients were treated using RP and pelvic lymph node dissection (PLND) in a multimodal setting, with median follow-up of 49 mo.OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: For PCRD prediction, a multivariate model with correction for competing risks was constructed to evaluate pathologic high-risk features (pT3b-4, Gleason score ≥8, and pN1) as predictors of mortality. All possible associations of the predictors were combined, and then subgroups with similar risk of PCRD were collapsed to obtain a simplified model encoding subgroups with significantly differing risk. Eightfold cross-validation of the model was performed.RESULTS AND LIMITATIONS: After applying exclusion criteria, 2823 subjects were identified. pT3b-4, Gleason score ≥8, and pN1 were all independent predictors of PCRD. The simplified model included the following prognostic groups: good prognosis, pN0 with 0-1 additional predictors; intermediate prognosis, pN1 with 0-1 additional predictors; poor prognosis, any pN with two additional predictors. The cross-validation yielded excellent median model accuracy of 88%. The retrospective design and the short follow-up could limit our findings.CONCLUSIONS: We developed and validated a novel and easy-to-use prognostic instrument to predict PCRD after RP+PLND. This model may allow clinicians to correctly counsel patients regarding the intensity of follow-up and to tailor adjuvant treatments.PATIENT SUMMARY: Prediction of mortality after primary surgery for prostate cancer is important for subsequent treatment plans. We present an accurate postoperative model to predict cancer mortality after radical prostatectomy for high-risk prostate cancer.

AB - BACKGROUND: Accurate prediction of survival after radical prostatectomy (RP) is important for making decisions regarding multimodal therapies. There is a lack of tools to predict prostate cancer-related death (PCRD) in patients with high-risk features.OBJECTIVE: To develop and validate a prognostic model that predicts PCRD combining pathologic features and using competing-risks analysis.DESIGN, SETTING, AND PARTICIPANTS: This was a retrospective multi-institutional observational cohort study of 5876 patients affected by high-risk prostate cancer. Patients were treated using RP and pelvic lymph node dissection (PLND) in a multimodal setting, with median follow-up of 49 mo.OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: For PCRD prediction, a multivariate model with correction for competing risks was constructed to evaluate pathologic high-risk features (pT3b-4, Gleason score ≥8, and pN1) as predictors of mortality. All possible associations of the predictors were combined, and then subgroups with similar risk of PCRD were collapsed to obtain a simplified model encoding subgroups with significantly differing risk. Eightfold cross-validation of the model was performed.RESULTS AND LIMITATIONS: After applying exclusion criteria, 2823 subjects were identified. pT3b-4, Gleason score ≥8, and pN1 were all independent predictors of PCRD. The simplified model included the following prognostic groups: good prognosis, pN0 with 0-1 additional predictors; intermediate prognosis, pN1 with 0-1 additional predictors; poor prognosis, any pN with two additional predictors. The cross-validation yielded excellent median model accuracy of 88%. The retrospective design and the short follow-up could limit our findings.CONCLUSIONS: We developed and validated a novel and easy-to-use prognostic instrument to predict PCRD after RP+PLND. This model may allow clinicians to correctly counsel patients regarding the intensity of follow-up and to tailor adjuvant treatments.PATIENT SUMMARY: Prediction of mortality after primary surgery for prostate cancer is important for subsequent treatment plans. We present an accurate postoperative model to predict cancer mortality after radical prostatectomy for high-risk prostate cancer.

KW - Journal Article

U2 - 10.1016/j.euf.2016.12.008

DO - 10.1016/j.euf.2016.12.008

M3 - SCORING: Journal article

C2 - 28753838

VL - 4

SP - 369

EP - 375

JO - EUR UROL FOCUS

JF - EUR UROL FOCUS

SN - 2405-4569

IS - 3

ER -