402 Background: Active surveillance (AS) has been accepted as the standard management for lower-risk prostate cancer (PCa) by major clinical guidelines. Genomic tools such as Decipher (Veracyte) have improved risk stratification and clinical decision-making in PCa, though their optimal role in AS is still under investigation. Decipher GRID, a proprietary platform including validated prognostic signatures, hallmark pathway scores, and other expression signatures, represents a potential resource for refining risk assessment in AS. Methods: The Urologic Outcomes Database (UODB) at the University of California San Francisco (UCSF) was queried for men on AS with ≥2 biopsies with Decipher testing performed on biopsy tissue. Outcomes were any upgrade (any increase in Gleason Grade Group, GG), major upgrade (upgrade to ≥GG3), and development of unfavorable histology (expansile, or large, cribriform or intraductal carcinoma). An average genomic risk (AGR) score was calculated as previously described. Multivariable Cox proportional hazards regression using stepwise selection models were performed to determine associations between 18 validated prognostic gene signatures and 37 hallmark cancer pathways with risk of outcomes, adjusting for UCSF Cancer of the Prostate Risk Assessment (CAPRA) score. Results: 486 men were included. Median (IQR) follow-up for the cohort was 70 (42-104) months. On diagnostic biopsy, 378 (78%) and 108 (22%) patients had GG1 and GG2, respectively. CAPRA risk at diagnosis was low (0-2), intermediate (3-5), and high risk (6-10) in 359 (74%), 126 (26%) and 1 patients, respectively. Median (IQR) AGR was 0.25 (0.20-0.31). After adjusting for CAPRA, multivariable models demonstrated that genomic signatures Long 2014 and Yu 2007 provided independent prognostic value for both upgrade and major upgrade outcomes and demonstrated that Lapointe 2004 provided independent prognostic value for unfavorable histology outcome. After adjusting for CAPRA, hallmark signatures of PI3K/AKT/mTOR signaling and reactive oxygen species pathways also provided independent prognostic value for all outcomes. Conclusions: Our analysis uncovered multiple prognostic and hallmark gene expression signatures that prognosticated adverse AS outcomes independent of CAPRA, a well-validated risk stratification tool. These findings support further validation of these genomic signatures to enhance risk stratification and guide AS management in PCa.
Shee et al. (Sun,) studied this question.