Evaluating Real-World Data Fitness-for-Use in Early Stage NSCLC Within a US Based Community Health System Network

OBJECTIVES: The early-stage cancer treatment landscape is changing rapidly as precision oncology finds applications beyond advanced disease. The Syapse Learning Health Network (LHN), consisting of community health systems within the U.S., presents a unique opportunity to understand disease management in the real-world setting. Characteristics of an early-stage non-small cell lung cancer (eNSCLC) cohort from Syapse LHN were evaluated for fitness-for-use in generating real world evidence.

METHODS: A previously developed data assessment framework^1,2, with focus on early-stage cancers, was applied. The eNSCLC cohort with initial diagnosis of stage I-III since 2016 was assessed for its fitness in three use cases: characterizing the patient cohort, understanding biomarker testing trends, and analyzing primary treatment patterns. Assessments were performed on data availability, conformity, completeness, and plausibility for each use case.

RESULTS: High completeness of variables such as initial diagnosis, clinical and pathological staging, histology, and other history of primary cancers in the data enables selection of an eNSCLC cohort (>7,000 patients). The baseline demographic characteristics of the selected cohort had low missingness (< 5%), and comprehensive information for PD-L1 and EGFR testing. More than 90% of patients had treatment information, including treatment received/not received and associated dates. Approximately 70% of patients resided in the Midwest U.S., and approximately 70% of patients had information related to Charlson Comorbidity Index at diagnosis. The cohort had a limited number of patients within the most recent 12 months due to data latency from cancer registries and processing.

CONCLUSIONS: This community health system based eNSCLC cohort demonstrated fitness-for-use for three common use cases, with potential to provide insights into broader clinical practice beyond clinical trials or academic practices. Inherent biases relating to data recency and geographic representation require careful consideration in assessing fitness-for-use particularly for studies requiring the most recent data or focusing on specific populations.