Examination of Decipher prostate genomic classifier in patients with de novo metastatic disease from a large scale real-world clinical and transcriptomic data linkage.

Abstract

224 Background: Prognostic gene expression testing of primary tumor tissue has become widely adopted for localized prostate cancer risk stratification. Recent retrospective analyses of clinical trials have examined such testing in metastatic hormone-sensitive disease, but little has been reported outside of this context. We aim to evaluate the prognostic value of the Decipher prostate genomic classifier (GC) in patients with de novo metastatic prostate cancer (mPC) in real-world clinical practice (RWD). Methods: Clinical and transcriptomic data from clinical use of the GC between 2016-2024 were linked with RWD aggregated from insurance claims, pharmacy records, and electronic health record data. Patients (pts) were anonymously linked between datasets by deterministic methods through a de-identification engine using encrypted tokens. A hierarchical claims-based algorithm was used to identify de novo distant metastasis in the patient’s record. De novo metastasis was defined using claims and diagnosis codes recorded within 90 days of initial prostate cancer diagnosis, excluding cases with other primary malignancies diagnosed within 90 days and omitting codes for unspecified or pelvic lymph node–only metastases. The distribution of GC scores was compared between patients with de novo metastases and (1) all patients with localized prostate cancer and (2) a matched subset of localized patients with comparable baseline clinical and pathologic features. Here we focus on comparison to the latter group. Results: 135,044 pts with Decipher prostate GC from biopsy tests were successfully linked to RWD. De novo mPC was identified in 509 patients and compared to a matched set of 10,689 pts with localized disease. Among pts with de novo mPC, the median age at Decipher testing was 71 years (IQR 65, 77), median percentage of positive cores was 75% (IQR 50-100%), median PSA was 17 ng/mL (IQR 6.8,73) and 75% had NCCN high or very high-risk disease at diagnosis. Compared to the matched set for localized pts, 29% of mPC pts had PSA > 50 vs. 6% for localized patients. Median Decipher score for mPC pts was 0.94 (IQR 0.71, 0.99) compared to 0.75 (IQR 0.5, 0.9) in the matched localized patient cohorts. Compared to localized prostate cancer pts, mPC pts exhibited a higher proportion of Luminal B subtype (65% vs 54%), and a higher prevalence of PTEN inactivity (25% vs 15%). Conclusions: Using the largest linkage of transcriptomic and clinical data to date, we developed algorithms to identify de-novo mPC from a cohort of patients tested with a GC. These pts tended to have higher PSA, higher rates of PTEN inactivity and luminal B subtype tumors, higher NCCN risk groups at time of diagnosis and had substantially elevated GC scores. The use of the GC test may enhance understandings of de novo metastatic disease biology, patterns of care, and treatment effectiveness.