Felix Preisser1, Marco Bandini2, Sebastiano Nazzani3, Elio Mazzone2, Michele Marchioni4, Zhe Tian5, Felix K H Chun6, Fred Saad5, Alberto Briganti7, Alexander Haese8, Francesco Montorsi7, Hartwig Huland8, Markus Graefen8, Derya Tilki9, Pierre I Karakiewicz5. 1. Martini-Klinik Prostate Cancer Center, University Hospital Hamburg-Eppendorf, Hamburg, Germany; Cancer Prognostics and Health Outcomes Unit, Division of Urology, University of Montreal Health Center, Montreal, Quebec, Canada; Department of Urology, University Hospital Frankfurt am Main, Frankfurt am Main, Germany. Electronic address: felixpreisser@gmx.de. 2. Cancer Prognostics and Health Outcomes Unit, Division of Urology, University of Montreal Health Center, Montreal, Quebec, Canada; Department of Urology and Division of Experimental Oncology, URI, Urological Research Institute, IRCCS San Raffaele Scientific Institute, Milan, Italy. 3. Cancer Prognostics and Health Outcomes Unit, Division of Urology, University of Montreal Health Center, Montreal, Quebec, Canada; Academic Department of Urology, IRCCS Policlinico San Donato, University of Milan, Milan, Italy. 4. Cancer Prognostics and Health Outcomes Unit, Division of Urology, University of Montreal Health Center, Montreal, Quebec, Canada; Department of Urology, SS Annunziata Hospital, "G.D'Annunzio" University of Chieti, Chieti, Italy. 5. Cancer Prognostics and Health Outcomes Unit, Division of Urology, University of Montreal Health Center, Montreal, Quebec, Canada. 6. Department of Urology, University Hospital Frankfurt am Main, Frankfurt am Main, Germany. 7. Department of Urology and Division of Experimental Oncology, URI, Urological Research Institute, IRCCS San Raffaele Scientific Institute, Milan, Italy. 8. Martini-Klinik Prostate Cancer Center, University Hospital Hamburg-Eppendorf, Hamburg, Germany. 9. Martini-Klinik Prostate Cancer Center, University Hospital Hamburg-Eppendorf, Hamburg, Germany; Department of Urology, University Hospital Hamburg-Eppendorf, Hamburg, Germany.
Abstract
BACKGROUND: Prostate cancer (PCa) staging is crucial in clinical decision making and treatment assignment. OBJECTIVE: To develop a predictive tool that is capable of predicting the probability of metastases at initial PCa diagnosis. DESIGN, SETTING, AND PARTICIPANTS: Within the Surveillance, Epidemiology, and End Results database (2010-2014), we identified patients with newly diagnosed PCa and available clinical tumor stage, prostatic-specific antigen value (PSA), and Gleason grade group (GGG), and with or without metastases. OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: We relied on PSA, clinical tumor stages, and GGG to discriminate between M1 and M0 patients. Patients were randomly divided according to the registry of origin between development (n=102469) and validation (n=98755) cohorts. Logistic regression modeling coefficients were used to devise a lookup table to discriminate between M0 and M1 stages. Receiver operating characteristic-derived area under the curve was tested for model accuracy, within the validation cohort. A total of 2000 bootstrap resamples were applied to 95% confidence intervals (CIs). Decision curve analysis (DCA) and calibration plots were used to test the performance of the lookup table. RESULTS AND LIMITATIONS: Of 201224 patients, 3.5% harbored metastatic PCa (mPCa). PSA >40ng/ml, GGG5, and GGG4, in that order, represented the strongest predictors of mPCa. Overall, PSA, clinical tumor stage, and GGG were 94.3% (95% CI: 94.2-94.3%) accurate in predicting the probability of mPCa, in the external validation cohort. Up to 39.4% probability of mPCa, the model demonstrated accurate predictions in the calibration plot. In DCA, a net benefit was recorded up to a threshold probability of approximately 54%. CONCLUSIONS: The proposed lookup table for the prediction of the probability of mPCa may represent a useful clinical tool based on its high accuracy, excellent calibration, and robust nature of predictions. PATIENT SUMMARY: Our study provides a highly accurate lookup table for the prediction of the probability of metastatic prostate cancer patients. This clinical tool can be useful in staging decisions.
BACKGROUND:Prostate cancer (PCa) staging is crucial in clinical decision making and treatment assignment. OBJECTIVE: To develop a predictive tool that is capable of predicting the probability of metastases at initial PCa diagnosis. DESIGN, SETTING, AND PARTICIPANTS: Within the Surveillance, Epidemiology, and End Results database (2010-2014), we identified patients with newly diagnosed PCa and available clinical tumor stage, prostatic-specific antigen value (PSA), and Gleason grade group (GGG), and with or without metastases. OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: We relied on PSA, clinical tumor stages, and GGG to discriminate between M1 and M0 patients. Patients were randomly divided according to the registry of origin between development (n=102469) and validation (n=98755) cohorts. Logistic regression modeling coefficients were used to devise a lookup table to discriminate between M0 and M1 stages. Receiver operating characteristic-derived area under the curve was tested for model accuracy, within the validation cohort. A total of 2000 bootstrap resamples were applied to 95% confidence intervals (CIs). Decision curve analysis (DCA) and calibration plots were used to test the performance of the lookup table. RESULTS AND LIMITATIONS: Of 201224 patients, 3.5% harbored metastatic PCa (mPCa). PSA >40ng/ml, GGG5, and GGG4, in that order, represented the strongest predictors of mPCa. Overall, PSA, clinical tumor stage, and GGG were 94.3% (95% CI: 94.2-94.3%) accurate in predicting the probability of mPCa, in the external validation cohort. Up to 39.4% probability of mPCa, the model demonstrated accurate predictions in the calibration plot. In DCA, a net benefit was recorded up to a threshold probability of approximately 54%. CONCLUSIONS: The proposed lookup table for the prediction of the probability of mPCa may represent a useful clinical tool based on its high accuracy, excellent calibration, and robust nature of predictions. PATIENT SUMMARY: Our study provides a highly accurate lookup table for the prediction of the probability of metastatic prostate cancerpatients. This clinical tool can be useful in staging decisions.