Abstract
Purpose :
In uveal melanoma (UM), tumor-based molecular prognostic testing using the prospectively validated 15-gene expression profile (15-GEP) test is widely used to stratify patients into low- (Class 1) and high-risk (Class 2) groups. There is growing interest in augmenting the 15-GEP with additional UM-associated prognostic factors including tumor size and PRAME status, however, whether the integration of these factors enhances prognostic accuracy is unclear. This study describes the development and validation of an integrated model that incorporates 15-GEP, PRAME, and tumor diameter to provide an optimized survival score for metastatic risk assessment.
Methods :
A total of 1763 UM patients were prospectively enrolled across 26 ocular oncology centers in the Collaborative Ocular Oncology Group study 2 (COOG2), including 1586 patients ≥ 18 years of age with posterior tumors and available 15-GEP/PRAME test results that were randomly assigned to a development cohort (n=793) and a validation cohort (n=793). A variety of Cox regression models were generated from combinations of 15-GEP, PRAME, and tumor diameter as continuous or categorical variables, and models were compared for performance using Net Reclassification Improvement (NRI), Integrated Discrimination Improvement (IDI), and receiver operating characteristic (ROC) analysis.
Results :
The top performing model used 15-GEP discriminant score, PRAME expression levels, and tumor diameter as continuous variables in an integrated model (i15-GEP). The i15-GEP, which yields a continuous survival score ranging from 0-100%, performed substantially better than the same model without 15-GEP, with a NRI of 0.91 (95% CI 0.76-1.07, P<0.001) and IDI of 0.12 (95% CI 0.09-0.15, p<0.001), demonstrating the strong contribution of 15-GEP to the model. The i15-GEP model also outperformed categorical 15-GEP class alone, with a NRI of 0.42 (95% CI 0.24-0.60, p<0.001) and IDI of 0.08 (95% CI 0.04-0.12, p<0.001). Further, the i15-GEP demonstrated high performance in the validation cohort, with an AUC of 0.84 (95% CI 0.81-0.88) vs. 0.79 (CI 0.76-0.83) for the baseline model (15-GEP).
Conclusions :
Incorporating tumor diameter and PRAME with 15-GEP provides better discriminatory power to predict metastatic risk than any factor alone or in combination, leading to improved accuracy for clinical pathway decision-making.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.