**13. Prostate genomics**

The six-tier NCCN risk group classification system provides a highly-validated basic framework for standard treatment recommendations [8]. However, a variety of advanced risk stratification tools have been developed and are in various stages of validation that independently improve stratification. The NCCN recommends ordering these tests for borderline cases as an extra data point that may potentially change management; patients with low-risk, favorable intermediate-risk, unfavorable intermediate-risk, and high-risk tumors with a life expectancy ≥10 years may be candidates for Decipher®, Oncotype Dx Prostate®, or Prolaris® [63]. Indeed, research has demonstrated that the current risk stratification systems are frequently poor prognosticators for clinically-meaningful endpoints; for distant metastases rates at 10-years, the concordance index (c-index) for the NCCN classification system was 0.73 (95% CI, 0.60–0.86) and for CAPRA was 0.74 (95% CI, 0.65–0.84) [64]. Per the NCCN, Decipher® currently has the highest level of evidence for validation among the major genomic classifiers (GCs), having been validated in the context of multiple clinical trials with consistent results (see **Table 4**) [63].

Numerous studies have been performed, and are ongoing, to validate GCs. One notable example includes validation of the 22-gene Decipher® GC from the biobank from the phase III randomized trial NRG Oncology/RTOG 0126 [64]. This trial compared men with intermediate-risk prostate cancer randomized to 70.2 Gy versus 79.2 Gy, without androgen deprivation therapy. RNA was extracted from the highest grade tumor foci, and for 215 patients (of the 1532 patients in the study), the material passed quality control. GC data were generated and compared to the patients' respective clinical outcomes on the study, for a retrospective analysis of the prospective trial. The GC proved independently prognostic for disease progression (p = 0.03),


*Level 1: Validation in the context of multiple clinical trials with consistent results.*

*Level 2: Validation in multiple prospective registry/observational cohorts with consistent results.*

*Level 3: Validation in multiple independent retrospective studies with consistent results.*

*Level 4: Validation in a single retrospective study, or multiple independent retrospective studies with inconsistent results.*

#### **Table 4.**

*Levels of evidence for major prostate genomic classifiers [8].*

biochemical failure (p < 0.001), distant metastasis (p = 0.01), and prostate cancerspecific mortality (p < 0.001). The authors deemed that the GC can be used to help personalize treatment for intermediate-risk prostate adenocarcinoma.

In 2021, prostate cancer researchers published "A Systematic Review of the Evidence for the Decipher Genomic Classifier in Prostate Cancer" in *European Urology* [65]. This systematic review incorporated 42 studies and 30,407 patients with localized, post-prostatectomy, non-metastatic castration-resistant, or metastatic hormonesensitive prostate adenocarcinoma. The patients were part of retrospective studies (n = 12,141), prospective registries (n = 17,053), and prospective and post-hoc randomized trial analyses (n = 1213). For 32 studies, the GC proved independently prognostic for all study endpoints (adverse pathology, biochemical failure, metastasisfree survival, cancer-specific survival, and OS) on multi-variate analysis, and improved discrimination over the standard of care in 24 studies. As well, the GC changed management for the AS (NNT = 9) and post-prostatectomy (NNT = 1.5–4) settings. Its utility was deemed strongest for decision-making with intermediate-risk prostate adenocarcinoma and post-prostatectomy. Indeed, despite the ongoing debates about adjuvant and salvage radiotherapy in the setting of adverse pathologic risk factors without biochemical failure, the NCCN Prostate Guidelines now (Version 1.2023) recommends that Decipher® "should be considered if not previously performed to inform adjuvant treatment if adverse features are found post-RP;" [63] as well, as discussed previously, the NCCN recommends strongly considering post-prostatectomy radiotherapy and ADT when the Decipher® GC score is high (>0.6) [63].

A clinical-genomic model has been developed that incorporates the NCCN risk groups with the Decipher® GC, thereby creating a clinical-genomic point system. This model has an improved c-index of 0.84 (95% CI, 0.61–0.93), versus 0.73 (95% CI, 0.60– 0.86) for the NCCN six-tiered classification system alone (see **Figure 3**) [64].

Regarding patients on AS, several studies have demonstrated the utility of GCs in determining which patients would have biopsy reclassification on serial biopsies, and therefore stop AS in favor of definitive treatment. A study at the University of California, San Francisco studied men with clinically low-risk prostate cancer prospectively enrolled on AS between 2000 and 2016 [66]. In this study, biopsy reclassification was defined as Gleason grade group ≥2 on subsequent biopsy. On multivariate analysis, biopsy re-classification at 3–5 years was strongly associated with a high genomic score (HR = 2.81); it was also strongly associated with a PSA density ≥ 0.15 (HR = 3.37), rapid PSA kinetics (HR = 2.19), and percentage biopsy cores positive (HR = 1.27). Of note, a PI-RADS 4–5 score on MRI was not associated with

#### **Figure 3.**

*Clinical-genomic point system using a genomic classifier [63, 64].*

biopsy re-classification. In a multi-institutional study led by the University of Michigan, 855 men underwent Decipher® testing of their prostate biopsies between February 2015–October 2019, of whom 264 (31%) proceeded with AS [67]. For the men who chose AS, after adjusting for NCCN risk group and all risk factors, a high-risk Decipher® score was independently associated with a shorter time to treatment failure (HR = 2.51, p < 0.001). Of note, for the men who proceeded with radical therapy, a high Decipher® score was independently associated with a shorter time to treatment failure on multi-variate analysis.

Numerous prospective trials are currently ongoing to evaluate the effectiveness of these markers, and to better assess their utility for enhancing risk stratification, including the following studies, as demonstrated in **Table 5** [68–70].

Additionally, while the usage of artificial intelligence is still an NCCN category IIB recommendation at this time for aiding with risk stratification, new studies suggest it will have an increasing role in cancer therapy precision [8, 71].
