Presentation (CTI)

OBJECTIVES: TGCT is a rare, locally aggressive soft-tissue tumor comprising distinct subtypes: localized/nodular (L-TGCT), diffuse (D-TGCT), and rare malignant (M-TGCT). Real-world evidence from administrative claims data can help characterize these populations, but challenges like diagnostic ambiguity, misclassification, and non-specific coding hinder accurate identification and timely treatment. As novel systemic therapies emerge, robust methodologies for identifying TGCT patients in real-world datasets are essential to inform clinical practice and research.
METHODS: We conducted an analysis of administrative claims databases in the United States (Marketscan) and Japan (JMDC) to develop algorithms for identifying TGCT cases. Clinical input from orthopedic surgeons ensured effective subtype differentiation. TGCT case definition was operationalized using specific ICD-10 codes: L-TGCT (D48.1 + D21.X), and D-TGCT and M-TGCT (M12.2 with associated combinations). Inclusion criteria required ≥1 inpatient/emergency department or outpatient claim, and/or received TGCT-targeted systemic therapy (e.g., pexidartinib, vimseltinib). We assessed algorithm performance based on scalability, reproducibility, and subtype particularity.
RESULTS: The algorithm leveraged ICD-10 diagnosis and procedure codes, WHO-Anatomical Therapeutic Chemical codes, specialist visits, and treatment patterns. Challenges included ICD subtype coding heterogeneity, diagnostic ambiguity, documentation variability, and absence of histopathological confirmation. Preliminary analyses supported the validity and clinical plausibility of the identification strategy, indicating that claims-based algorithms can enhance TGCT case ascertainment.
CONCLUSIONS: Terminological inconsistencies and coding limitations complicate TGCT identification in secondary datasets. Standardizing diagnostic criteria and coding practices across geographies is crucial for improving global TGCT surveillance, particularly as targeted systemic therapies gain prominence. Descriptive analyses of the algorithm indicate that the identification strategy has face validity and clinical plausibility; integration with complementary data sources, such as pathology, imaging, and natural language processing of electronic health records will be essential for accuracy. This scalable framework offers potential to support rare tumor surveillance and inform research, drug development, and health policy.