3262 - Utilization of a 46 Gene Assay Genomic Test with Clinicopathologic Factors in Unfavorable Intermediate to Very High Risk Prostate Cancer To Predict Upstaging Via PSMA PET/CT Scan

Purpose/Objective(s):

In this project, we analyzed the predictive role of a 46 gene assay with clinicopathologic features in unfavorable intermediate to very high risk prostate cancer for upstaging via PSMA PET/CT scan

Materials/Methods:

We performed a single institution retrospective review of all unfavorable intermediate to very high risk prostate cancer patients who received an initial workup including a PSMA PET/CT and Prolaris genomic testing. The Prolaris genomic test is a 46 gene assay used to assess risk 10-year rates of distant metastasis with active surveillance, single-modal therapy (radiation, surgery), or multi-modal therapy (radiation and ADT). A multivariate logistic regression model was made to predict metastatic disease on initial staging PSMA (PSMA+), using clinicopathologic features of grade group, AJCC T-stage, PSA at the time of diagnosis, prostate gland size, and number of biopsy cores positive for malignancy. Youden's Index was used to establish a probability threshold for high and low risk for PSMA+ with goodness-of-fit assessed using calibration plots. Final predictions were then compared to the Prolaris test for PSMA+ in terms of sensitivity, specificity, accuracy, and AUC. The threshold for statistical significance for all tests was set to p < 0.05.

Results:

We identified 88 patients eligible for analysis with a median age of 69.5 years (IQR = 63 - 74 years). Most patients had unfavorable intermediate risk disease (70.5%; N = 62/88) with stage IIC (58.0%; N = 51/88) disease. Patients had an initial median PSA of 9.2 ng/ml (IQR = 6.2 - 14.0 ng/ml) with screen-detected (78.4%; N = 69/88) and grade group 3 disease (43.2%; N = 38/88). Median Prolaris Scores were 3.6 (IQR = 3.2 - 4.1) with 10-year risk of distant metastasis estimated at a median of 5.6% (IQR = 3.4 - 8.9%) and 3.4% (IQR = 2.3 - 5.2%) with single- and multi-modal therapy, respectively. On multivariate analysis, PSA was the only clinicopathologic variable significantly associated with PSMA+ (OR = 1.09, 95% CI = 1.01 - 1.18, p = 0.024). Patients predicted as high-risk by this model had an incidence of 37.9% PSMA+ versus 1.7% for low-risk patients (p < 0.001). Sensitivity (76.3%), AUC (0.808), and overall accuracy (78.4%) of the model greatly outperformed the Prolaris test (50.0%, 0.603, and 50.0% respectively), while specificity was identical at 91.7%.

Conclusion:

A model consisting of T-stage, prostate gland size, number of biopsy cores positive for malignancy, grade group, and initial PSA greatly outperformed the Prolaris genomic assay for detecting distant metastatic disease on staging PSMA scans. With validation, this model could be used as an adjunct to modern prostate staging paradigms to help inform decision-making on utilization of PSMA scans, potentially leading to lower healthcare expenditures and more efficient workups in resource-poor environments.