OBJECTIVES: A challenge in performing retrospective observational studies using treatment data is deriving clinically appropriate treatment sequences, especially when working across cancer types. The objective of this study is to assess the performance of a disease-agnostic treatment sequencing heuristic for deriving LOT in a biomarker-defined, multi-tumor cohort.

METHODS: This study used the nationwide de-identified Flatiron Health-Foundation Medicine Enhanced Pan-tumor Clinico-genomic Database (data cut-off 09-30-2023). Data originated from approximately 280 US cancer clinics. A cohort of patients with advanced/metastatic disease across 17 rare cancer types, known TMB status, and ≥2 LOTs was selected. The disease-agnostic heuristic for deriving LOT was applied to patient-level systemic, non-oral antineoplastic treatment data recorded in the electronic health record and chart-abstracted oral therapy. Lines were evaluated by cancer type and sequence indexed to advanced/metastatic diagnosis date. Treatment sequencing, visualization of sequencing and qualitative assessment of LOT compared to National Comprehensive Cancer Network (NCCN) preferred/recommended regimens was performed with patients stratified based on TMB-high (TMB-H) [≥10 mutations/megabase (mut/Mb)] versus non-TMB-H status (<10 mut/Mb). Patterns of immune checkpoint inhibitor (ICI) use across cancer types were also evaluated by characterizing ICI prevalence, sequencing, and mono-/combo-ICI use among the TMB-H and non-TMB-H sub-cohorts.

RESULTS: Among 4,314 patients, clinically appropriate LOTs and sequencing of LOTs were represented in a consistent manner across cancer types and aligned with NCCN preferred/recommended regimens. For patients with TMB-H tumors, ICI-therapy was well represented across most diseases with sequencing consistent with tumor-specific and tumor-agnostic approved indications for ICI-therapy.

CONCLUSIONS: A disease-agnostic treatment sequencing heuristic for deriving LOT is feasible and can define clinically appropriate LOTs across a multi-tumor patient cohort. This was confirmed by assessing exposure to ICI-therapy stratified by TMB status. This approach may be applied to other real-world disease-agnostic datasets to facilitate treatment-related analyses.