Abstract 6389: An exosome-based liquid biopsy powered by machine learning predicts progression-free and overall survival before commencing first-line EGFR inhibitors for metastatic colorectal cancer: The EXONERATE Study

Abstract Introduction: We developed a liquid biopsy assay based on exosomal and cell-free microRNA (exo- and cf-miRNA, respectively) for patients with chemotherapy-naïve, metastatic, unresectable, RAS and BRAF wild-type colorectal cancer (mCRC), to predict progression-free (PFS) and overall survival (OS) before the commencement of first-line treatment with EGFR inhibitors (EXONERATE). Methods: Patients with mCRC were consecutively enrolled in two open-label, nationwide clinical trials and received first-line cetuximab or panitumumab. Cf- and exo-miRNA expression was quantified using RT-qPCR in plasma samples obtained before exposure to EGFR inhibitors. The prespecified primary endpoint was 12-month PFS, hierarchically tested in left-sided mCRC, right-sided mCRC, and the entire cohort. As secondary endpoints, the EXONERATE status was correlated with PFS and OS hazard ratios (HR). Results: After excluding those with RAS/BRAF mutations and unavailable samples, the complete analysis dataset comprised 120 patients (94 with left-sided and 26 with right-sided mCRC; 87 received cetuximab and 33 panitumumab). As expected, right-sided mCRC patients showed inferior PFS and OS outcomes (HR 1.80 [CI95% = 1.13 - 2.80] and 1.80 [CI95% = 1.00 - 3.27], respectively). Genome-wide small RNA sequencing on 20 patient samples identified 12 cf- and 14 exo-miRNA candidates, then narrowed to 8 and 9 based on best performance. We employed machine learning algorithms to develop two liquid biopsy assays, one based on cf- and the other on exo-miRNAs, predicting 12mPFS with AUC values of 0.84 and 0.81, respectively. These were then combined to establish the EXONERATE assay for left-sided mCRC (AUC = 0.89), successfully validated in right-sided mCRC patients (AUC= 0.84). Among left-sided mCRC patients, 35% were EXONERATE-high, exhibiting shorter median PFS (mPFS = 9.5 vs. 18.5 months, p<.001) and a higher mortality rate (HR 0.21 [CI95% = 0.10 - 0.42]). In the right-sided mCRC cohort, the EXONERATE-high panel robustly identified patients with shorter mPFS (8.6 vs. 23.7 months, p=.0042) and predicted 12mPFS with high accuracy (sensitivity 100%, NPV 81%, PPV 100%). Although in univariate analysis both the EXONERATE panel and sidedness showed associations with PFS and OS, in multivariate analysis only the EXONERATE panel could fully account for the differences in PFS and OS (HR 0.22 and HR 0.19, respectively). Hence, we employed the EXONERATE panel agnostically of tumor sidedness and observed that EXONERATE-high patients demonstrated poorer survival outcomes in mPFS (8.9 vs. 19.6 months, HR 0.27 [CI95% = 0.17 - 0.41]) and mOS (22.6 vs. >60 months, HR 0.18 [CI95% = 0.09 - 0.37]). Conclusion: The EXONERATE panel robustly predicted PFS and OS outcomes in both right- and left-sided mCRC patients before receiving either EGFR inhibitor. Citation Format: Xu Caiming, Alessandro Mannucci, Roy Souvick, Esposito Francis, Oliveres Helena, Alonso-Orduña Vicente, Escudero Pilar, Fernández-Martos Carlos, Salud Antonieta, Gallego Javier, Rodriguez Jose Ramon, Martín-Richard Marta M. Marta, Fernández-Plana Julen, Manzano Hermini, Aparicio Jorge, Feliu Jaime, Maurel Joan, Ajay Goel. An exosome-based liquid biopsy powered by machine learning predicts progression-free and overall survival before commencing first-line EGFR inhibitors for metastatic colorectal cancer: The EXONERATE Study [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 6389.