Introduction:
Genomic classifiers are endorsed by national guidelines and are commonly used to inform prognosis in patients with localized prostate cancer (PCa). Despite their availability, integration of genomic classifiers within optimized clinical prognostic stage group systems are lacking. Therefore, using two prospective population-based tumor registries, we aimed to understand the distribution of genomic risk within the previously validated, American Joint Committee on Cancer (AJCC)-compliant, staging collaboration for cancer of the prostate (STAR-CAP). We hypothesized that genomic heterogeneity would have implications on clinical risk estimates and may inform treatment options.
Methods:
Genomic risk distribution was assessed using the Decipher genomic classifier (Veracyte, San Diego, CA) in two separate multi-institutional, prospectively collected population-based tumor registries: (1) the Decipher Genomics Resource for Intelligent Discovery (GRID) and (2) the community-based Michigan Urological Surgery Improvement Collaborative (MUSIC) (n=50,981 GRID + 1584 MUSIC). Our primary endpoint was PCa-specific mortality (PCSM) with a secondary endpoint of distant metastasis (DM). Marginal risk estimates provided by STAR-CAP stage were combined with hazard ratios of Decipher on PCSM and DM along with the distribution of Decipher scores by STAR-CAP stage to calculate integrated estimates. Risk estimates were calculated to have a weighted average consistent with the marginal STAR-CAP estimates while simultaneously adhering to the hazard ratio constraints.
Results:
Median age was 68 years and 66 years and median PSA was 6.2 ng/mL and 10.5 ng/mL in the GRID and MUSIC populations, respectively. The GRID population had 31.4%, 18.5%, and 50.1%, with high-, intermediate-, and low-Decipher risk, compared to 35.7%, 16.2%, and 48.1% in MUSIC. Decipher-based genomic risk varied across all STAR-CAP stages (IA-IIIC) in both the Decipher GRID (Figure 1a) and MUSIC registries (Figure 1b). Among those with low STAR-CAP stage (IA-IC), there was a proportion of men with high genomic risk in both GRID (18.7%) and MUSIC (17.7%) populations. Estimates of PCSM (0.1% to 49.2%) and DM (0.3%-73.4%) varied widely depending on clinical-genomic risk. Use of a Decipher-informed STAR-CAP system led to a significant degree of stage reclassification, including 25.6% or 45.5% upstaging or downstaging at least one stage and 7.2% or 11.8% upstaging or downstaging at least two stages, respectively.
Conclusion:
In a large scale, population-based, prospective assessment, we found wide variation in genomic risk distribution within and across clinical stage groups, suggesting integration of genomic and clinicopathologic risk groups may lead to more nuanced prediction of relevant clinical outcomes in PCa. These findings may help individualize risk assessment, yielding more precise prognosis estimates for patients.
Funding: Urology Care Foundation/AUA Research Scholar Award, Prostate Cancer Foundation
Image(s) (click to enlarge):
UNDERSTANDING POPULATION-WIDE GENOMIC RISK DISTRIBUTION AND INTEGRATING CLINICAL-GENOMIC RISK FOR PROGNOSITICATION IN PATIENTS WITH CLINICALLY LOCALIZED PROSTATE CANCER
Category
Prostate Cancer > Potentially Localized
Description
Poster #200
Friday, December 1
9:00 a.m. - 10:00 a.m.
Presented By: Udit Singhal
Authors:
Udit Singhal
Ralph Jiang
Matthew Schipper
James Proudfoot
Elizabeth Chase
Krithika Suresh
Elai Davicioni
William C. Jackson
Daniel E. Spratt
Robert T. Dess
Todd M. Morgan