Introduction:

Despite efforts to reduce prostate cancer (PCa) overdiagnosis, over 27% of newly diagnosed cases are classified as low-risk or less by National Comprehensive Cancer Network (NCCN) guidelines. As such, a paradigm shift has endorsed active surveillance (AS) to manage low-risk and selected intermediate-risk localized PCa, however the risk of progression to treatment at five-years ranges between 20% to 50%. Therefore, clinical tools including multiparametric magnetic resonance imaging (mpMRI) and novel genomic classifiers (GC) such as Decipher® GC are recommended to risk stratify newly diagnosed PCa. Given guidelines remark on limited longitudinal evidence validating tissue biomarker testing with mpMRI findings independently improve csPCa screening among AS patients, our primary aim was to examine imaging (mpMRI) and genomic (Decipher GC) risk stratification at PCa diagnosis with AS clinical progression to treatment. Our secondary aim was to determine if combining preliminary imaging with genomic testing would provide any additional predictive utility.

Methods:

Consecutive AS patients (n=338) identified from biopsies submit for Decipher® GC testing (Veracyte, Inc, San Francisco, CA) between December 2016 to December 2023 were retrospectively reviewed (n=973). Prostate-specific antigen (PSA), index biopsy Gleason Grade Group (GGG), NCCN risk group, mpMRI Prostate Imaging-Reporting and Data System (PI-RADS) classification, and Decipher® GC score during AS initiation were collected with treatment timing, if patients progressed. Initial PI-RADS 1-3 or PI-RADS 4-5 classification and Decipher GC low-risk (<0.45), intermediate-risk (0.45-0.6), or high-risk (>0.6) stratified patients. The association between initial risk profiles to AS outcomes were assessed by a chi-square statistic. Kendall’s tau-b (τb) correlation coefficient evaluated the relationship between Decipher GC (≥0.45) and mpMRI (PI-RADS 4-5). Cox proportional hazard regression models predicting progression were created on Decipher GC or mpMRI risk groupings independently, and when stratifying by the non-modeled variable to determine whether combining assessments would enhance prognostic precision.

Results:

The mean PSA was 7.4ng/mL for the 338 AS patients, 34% progressed during median follow-up of 26 (IQR:16-40) months. Clinical characteristics and risk classifications stratified by outcome are presented in Table 1.

Figure 1A illustrates independent Decipher GC and mpMRI Cox hazard models. Decipher GC ≥0.45 generated a hazard ratio (HR) of 2.04 ([1.39-2.99], p<0.001) while PI-RADS 4-5 yielded an HR of 1.85 ([1.21-2.81], p=0.004) for progression. As genomic and imaging high-risk features were weakly correlated (τb = 0.133[0.06-0.21], p=0.021), Figure 1B models progression for Decipher GC groups after PI-RADS stratification and Figure 1C models progression for PI-RADS groups after Decipher GC stratification. Decipher GC ≥0.45 remained a predictor for all patients, including the subset with PI-RADS 4-5 (HR:1.58[0.99-2.50], p=0.05); however, PI-RADS 4-5 was not predictive for men with Decipher GC ≥0.45 (HR:1.27 [0.59-2.71]).

Conclusion:

Based on our analysis, elevated risk evaluations during AS inititiation by Decipher GC (≥0.45) and mpMRI (PI-RADS 4-5) were weakly correlated (τb = 0.13), which suggests that imaging and genomic features capture distinct, complementary prognostic information and should ideally be combined given that both initial Decipher GC ≥0.45 and PI-RADS 4-5 were associated with progression to treatment. Our results validate a Decipher GC ≥0.45 as an independent predictor for AS progression in patients across PI-RADS classification, while mpMRI PI-RADS 4-5 was not when Decipher GC ≥0.45. As such, an individualized approach to surveillance is augmented by Decipher GC testing, even for men with concerning mpMRI (e.g., PI-RADS 4-5).

Funding: This work was supported in part by National Institutes of Health (R01-CA258690 and R01-CA282022 to E.H.K. and J.E.I.).