A prospective validation of the genomic classifier to define high-metastasis risk in a subset of African American men with early localized prostate cancer: VanDAAM study.

Abstract

5005 Background: Risk stratification of prostate cancer (PC) using routine clinical variables remains suboptimal as they do not account for underlying tumor biology. The genomic classifier provides information on underlying biology and independently predicts an individual patient’s risk of metastasis. Although the performance of the genomic classifier has been tested across different cohorts primarily comprised of White men, its validation as an optimal genomic risk classifier for African American men (AAM) is thus far lacking in a prospective trial. We report the initial results on the prospective validation of the genomic classifier in a matched cohort of AAM and non-AAM (NAAM). Methods: This was a multisite, prospective validation trial of the genomic classifier i.e. Decipher score in AAM. Participants were recruited on a 1:1 enrollment ratio of AAM to NAAM diagnosed with low-intermediate risk PC. Patient on active surveillance were ineligible. NAAM were matched to AAM on PSA, age, biopsy Gleason score, clinical stage, and percent positive biopsy cores. Diagnostic biopsy specimens were processed at a CLIA certified laboratory and Decipher score was assessed using whole transcriptome profiling platform. Total target accrual was 250 men treated for low-intermediate PC over three years. Statistical analyses include categorical comparison of race dependent risk group migration between NCCN risk group and genomic classifier. Relative risk of metastasis was estimated using negative binomial model. Results: Final analytical cohort included 207 evaluable cases (AAM = 102 and NAAM = 107) with comprehensive genomic information. Risk of metastasis was determined based on pretreatment biopsy Decipher score, and patients were classified as low, favorable-, and unfavorable intermediate risk. Despite achieving a robustly matched clinical cohort, we observed significant genomic heterogeneity between AAM and NAAM across NCCN risk groups. In a comparative analysis, 49% of low-favorable intermediate risk AAM harbored high genomic risk tumors as compared to only 10% NAAM, p = 0.02. Similarly, using the modified clinico-genomic risk classifier (cGC), comprised of both Decipher score and clinical variables, AAM experienced an extreme deviation of risk status (difference [δ] between cGC and NCCN ≥ 2) as compared to NAAM (26.8% vs 8.1%, p = 0.03). In a binomial model, low-favorable NCCN risk AAM were 3.9 times more likely to be reclassified as high genomic risk for distant metastasis compared to NAAM (RR = 3.99, 95% CI, 1.15 – 13.86, p = 0.02). Conclusions: Clinical NCCN risk classification is an inadequate surrogate of tumor biology and offers suboptimal risk stratification for AAM with PC. Integration of patient specific genomic classifier into standard of care will improve accuracy in disease risk classification and treatment recommendations for AAM. Clinical trial information: NCT02723734.