Stratification of patients by tumor type using molecular profiling in real-world data.

Abstract

e19262 Background: While Next-Generation Sequencing (NGS) tests become increasingly more common for diagnosis, molecular characterization, and treatment, a significant amount of molecular data derives from single-gene or analyte tests. Single gene test information is stored in disparate sources including electronic medical record (EMR) and data access for clinical use remains a challenge. A solution that harmonizes biomarker data beyond standard NGS-centric data and linked to rich clinical data is required for the complete patient picture. Methods: Health Catalyst’s extended real-world database, Touchstone includes a molecular data mart that integrates data from provider and life sciences proprietary NGS panels, Laboratory Information Systems, and other repositories. A portion of the data is derived from single-gene tests documented in the EMR. Biomarker data from EMRs was extracted from six health systems via a proprietary pipeline for extracting biomarker data. The algorithm relies on a curated ontology for molecular terms and publicly available terminologies for human genetics. Minor transformations extract pertinent variant information where available to harmonize with NGS-level data. Results: Over 44 thousand molecular labs from over 24 thousand patients were identified with this method. The oncology classes for which molecular data was identified in the greatest number of patients include skin, hematological, breast, digestive, and lung cancers (Table). PRTN3, EGFR, BRAF, JAK2, ERBB2, and KRAS are among the most commonly tested genes. Conclusions: Integrated real-world clinical and biomarker data from single gene tests can inform clinical decision-making and support clinical trial recruitment across a broader set of patient population. [Table: see text]