Feature Importance And Individual Prediction Of Disease-specific Survival In Biopsy Gleason 8-10 Prostate Cancer:Surgery Vs. Radiotherapy Alone

Background: Prostate cancer is a common malignant tumor of the urogenital system that poses a serious threat to human health.In recent years,probably due to the national guidelines against PSA screening and the increased use of active surveillance in the United States,the proportion of prostate cancer in the early and completely curable stage has been declining,while the incidence of poorly differentiated tumors with Gleason score of 8-10 has been increasing.Objective: According to the National Comprehensive Cancer Network(NCCN)guidelines,prostate cancer patients with a Gleason score of 8-10 are classified as high or very high risk,and the optimal primary treatment regimen remains controversial.In this study,we separately defined independent factors of cancer-specific survival(CSS)after radical prostatectomy(RP)versus radiotherapy(RT)alone,and evaluated relative variable importance.Two preoperative nomograms were developed to help surgeons individually compare the differences in CSS between patients receiving the two treatments before surgery.Methodology: Using the Surveillance,Epidemiology,and End Results(SEER)database,A total of 6,320 patients who received RP and 10,267 who received radiotherapy alone were both separately divided into derivation and internal validation sets using the Surveillance,Epidemiology,and End Results(SEER)database.An independent cohort(n=131)was selected from our center for external validation of the RP model.Cumulative incidence functions for cancer-specific mortality were evaluated,and multivariable competing-risks analyses were used to identify independent predictors.The decision tree CART algorithm was used to analyze the independent factors entered separately,and the random forest model was employed to evaluate the relative importance of predictors in the two derivation cohorts.The independent prognostic factors were integrated into the Cox model and the nomogram was established.The predictive accuracy of the nomogram was evaluated by the consistency index(C-index),receiver operating characteristic curve(ROC)and calibration plot.The risk stratification was performed according to the nomogram and the survival curves were plotted.Results: Age,Gleason pattern,percentage of positive cores and clinical T stage were identified as independent factors related to CSS after RP,while clinical T stage was far more important than other indicators.The Gleason pattern,percentage of positive cores,PSA value and N stage were identified as independent factors related to CSS after radiotherapy alone,while the first two factors were the most important prognostic indicators.Two nomograms were established based on the independent factors of patients undergoing RP and RT alone to predict CSS before surgery,and the models showed good discrimination and calibration in internal validation.Compared with EAU and AJCC stratification criteria,the nomogram performance of patients receiving RT alone was significantly improved by C-index.The risk stratification based on the two nomograms showed a clearer trend of separation of survival curves than the AJCC staging system.Conclusions: The result suggested that the independent preoperative prognostic factors of CSS and their relative importance differed between the patients undergoing RP and radiotherapy alone.There is also a difference in CSS between RP and RT alone for an individual patient.The newly corresponding prognostic stratification can better determine the patient-specific selection of these two treatment options.