Visualizing real-world outcomes through the lens of a molecular tumor board: Persona-typing pancreatic cancers for clinical decision support.

Abstract

e18100 Background: Despite the molecular diversity of pancreatic adenocarcinoma, standard therapy options remain largely molecularly-untailored except in the context of DNA damage response alterations such as BRCA mutations, NTRK fusions, or microsatellite instability. Methods: To broaden the utility of molecular profiling, we analyzed real-world outcomes and established 12 molecularly-driven clusters, or “persona types” (PTs), across a cohort of 1280 pancreatic cancers with CLIA-grade genomic/proteomic data (commercially available NGS/IHC panels). Patients were consented into Perthera’s precision oncology platform via referrals from advocacy programs and hospital partnerships for case review by a molecular tumor board (MTB). Persona types were generated using unsupervised k-means clustering of molecularly-tailored therapy recommendations formulated by the MTB. Progression-free survival was documented across all lines of therapy. Results: Personas based on multi-omic profiles and expert-driven recommendations were representative of molecular phenotypes reported in previous studies (e.g. DNA damage response deficiencies, BRAF mutations, other non-KRAS-drivers, SWI/SNF alterations, squamous-associated genes, cell cycle regulators, etc.). To streamline the exploration of real-world outcomes, we developed an interactive dashboard that enables users to compare PFS survival curves across Personas for various therapies such as gemcitabine/nab-paclitaxel, FOLFIRINOX, other standard regimens, immunotherapy, PARP inhibitors, RAF/MEK/ERK inhibitors, and more. Persona types associated with increased/decreased sensitivity to several classes of agents were identified (e.g. PD-1/L1 inhibitors in a persona enriched for cell cycle dysregulation, PARP inhibitors in a persona enriched for BRCA mutations). Conclusions: Empowering oncologists with personalized insights into real-world outcomes may promote investigator-initiated trials and augment clinical decision support, particularly when choosing between standard of care regimens or when exploring potential clinical trial options.