Dataset Information

Population-Based Comparison of Different Risk Stratification Systems Among Prostate Cancer Patients.

ABSTRACT:

Background

It is not known which risk stratification system has the best discrimination ability for predicting prostate cancer death.

Methods

We identified patients with non-metastatic primary prostate adenocarcinoma diagnosis between 2004 and 2015 using the Surveillance, Epidemiology, and End Results database. Patients were categorized in different risk groups using the three frequently used risk stratification systems of the National Comprehensive Cancer Network guideline (NCCN-g), American Urological Association guideline (AUA-g), and European Association of Urology guideline (EAU-g), respectively. Associations between risk classification and prostate cancer-specific mortality (PCSM) were determined using Kaplan-Meier analyses and multivariable regression with Cox proportional hazards model. Area under the receiver operating characteristics curve (AUC) analyses were used to test the discrimination ability of the three risk grouping systems.

Results

We analyzed 310,062 patients with a median follow-up of 61 months. A total of 36,368 deaths occurred, including 6,033 prostate cancer deaths. For all the three risk stratification systems, the risk groups were significantly associated with PCSM. The AUC of the model relying on NCCN-g, AUA-g, and EAU-g risk stratification systems for PCSM at specifically 8 years were 0.818, 0.793, and 0.689 in the entire population; 0.819, 0.795, and 0.691 in Whites; 0.802, 0.777, and 0.681 in Blacks; 0.862, 0.818, and 0.714 in Asians; 0.845, 0.806, and 0.728 in Chinese patients. Regardless of the age, marital status, socioeconomic status, and treatment modality, AUC of the model relying on NCCN-g and AUA-g for PCSM was greater than that relying on EAU-g; AUC of the model relying on NCCN-g system was greater than that of the AUA-g system.

Conclusions

The NCCN-g and AUA-g risk stratification systems perform better in discriminating PCSM compared to the EAU-g system. The discrimination ability of the NCCN-g system was better than that of the AUA-g system. It is recommended to use NCCN-g to evaluate risk groups for prostate cancer patients and then provide more appropriate corresponding treatment recommendations.

SUBMITTER: Xie M

PROVIDER: S-EPMC8076565 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:INTRODUCTION:The purpose of this study was to perform a direct comparison of several existing risk-stratification tools for localized prostate cancer in terms of their ability to predict for biochemical failure-free survival (BFFS). Two large databases were used and an external validation of two recently developed nomograms on an independent cohort was also performed in this analysis. METHODS:Patients who were treated with external beam radiotherapy (EBRT) and/or brachytherapy for localized prostate cancer were selected from the multi-institutional Genitourinary Radiation Oncologists of Canada (GUROC) Prostate Cancer Risk Stratification (ProCaRS) database (n=7974) and the Centre Hospitalier de l'Université de Montréal (CHUM) validation database (n=2266). The primary outcome was BFFS using the Phoenix definition. Concordance index (C-index) reported from Cox proportional hazards regression using 10-fold cross validation and decision curve analysis (DCA) were used to predict BFFS. RESULTS:C-index identified Cancer of the Prostate Risk Assessment (CAPRA) score and ProCaRS as superior to the historical GUROC and National Comprehensive Cancer Network (NCCN) risk-stratification systems. CAPRA modeled as five and three categories were superior to GUROC and NCCN only for the CHUM database. C-indices for CAPRA score, ProCaRS, GUROC, and NCCN were 0.72, 0.72, 0.71, and 0.72, respectively, for the ProCaRS database, and 0.66, 0.63, 0.57, and 0.60, respectively, for the CHUM database. However, many of these comparisons did not demonstrate a clinically meaningful difference. DCA identified minimal differences across the different risk-stratification systems, with no system emerging with optimal net benefit. External validation of the ProCaRS nomograms yielded favourable calibrations of R2=0.778 (low-dose rate [LDR]-brachytherapy) and R2=0.868 (EBRT). CONCLUSIONS:This study externally validated two ProCaRS nomograms for BFFS that may help clinicians in treatment selection and outcome prediction. A direct comparison between existing risk-stratification tools demonstrated minimal clinically significant differences in discriminative ability between the systems, favouring the CAPRA and ProCaRS systems. The incorporation of novel prognostic variables, such as genomic markers, is needed.