The Prostate Cancer Risk Stratification Project: Database Construction and Risk Stratification Outcome Analysis

Abstract

This investigation reports on the biochemical and clinical outcomes of a newly created pan-Canadian Prostate Cancer Risk Stratification (ProCaRS) database developed by the Genitourinary Radiation Oncologists of Canada (GUROC). GUROC ProCaRS template-compliant data on 7974 patients who underwent radiotherapy were received from 7 unique databases. Descriptive analysis, Cox proportional hazards, and Kaplan-Meier analyses were performed using American Society for Radiation Oncology (ASTRO) biochemical failure-free survival (BFFS), prostate cancer-specific survival, and overall survival. Multivariable modeling for the primary ASTRO BFFS end point showed that age, prostate-specific antigen, T stage, and Gleason score and components such as hormonal therapy, and radiation treatment (brachytherapy with better outcome than external-beam) were predictive of outcome. Kaplan-Meier analysis of the existing GUROC and new NCCN classification system both showed good separation of all clinical outcome curves. The construction of a pan-Canadian database has informed important prostate cancer radiotherapy outcomes and risk stratification.

Brachytherapy Improves Biochemical Failure–Free Survival in Low- and Intermediate-Risk Prostate Cancer Compared With Conventionally Fractionated External Beam Radiation Therapy: A Propensity Score Matched Analysis

“…For example, information on both percent positive core biopsy and Gleason 7 subpattern (3 þ 4 vs 4 þ 3) was generally incompletely entered and could not be controlled for in our analysis (48,49). The variation in RT management of prostate cancer across different treatment centers could potentially impact outcomes (63); however, the method of data collection from the various cancer institutions made it impossible to include a "treatment center" variable in the PS models (48,49). Finally, any conclusions regarding OS should be made cautiously given the limited median followup, ranging from 2.7 to 7.3 years depending on the cohort being assessed.…”

“…The analysis performed was limited by the treatment and outcome data available in the ProCaRS database. For example, no toxicity data were available, which would have allowed for both mortality and morbidity assessment (48,49). Matching was restricted to patients with complete data for all variables included in the PS models (logistic regression models) and based on specific sets of exclusion criteria used to create homogenous comparison groups.…”

“…Details of the ProCaRS database ethics approval, construction, quality assurance process, and results of recursive partitioning riskstratification analysis and Cox multivariable regression outcome analysis have been previously described (48,49).…”

“…The GUROC 3 risk-category system was used to stratify patients as low-or intermediate-risk (2). Treatment comparisons in the high-risk cohort were abandoned owing to low numbers of patients receiving BT (nZ21) compared with EBRT (nZ268) in the ProCaRS database (48,49). Patients with T2b T stage with missing data on TNM staging year were removed before PS matching to ensure all T2b patients were properly identified with unilateral or palpable bilateral disease (50)(51)(52) …”

External validation of the ProCaRS nomograms and comparison of existing risk-stratification tools for localized prostate cancer

“…Clinical outcomes and patient characteristics for the ProCaRS database have been described in detail previously. 11,19 Median followup was 3.67 years for the CHUM database compared to 6.57 years in the ProCaRS database (p<0.001).…”