2023 World Conference on Lung Cancer (Posters)-EP07.04. Real-World Biomarker Testing and Treatment Patterns Identified Using Machine Learning in Early Non-small Cell Lung Cancer

EP07.04. Real-World Biomarker Testing and Treatment Patterns Identified Using Machine Learning in Early Non-small Cell Lung Cancer - PDF(Abstract)

Back to course

Pdf Summary

This study aimed to evaluate biomarker testing and treatment patterns in patients with early non-small cell lung cancer (eNSCLC) in US clinical practice using machine learning (ML). The researchers analyzed a retrospective observational study based on the Flatiron Health research database, which contained electronic health record (EHR) data. The study included patients with an NSCLC diagnosis between January 2019 and August 2022, with stage I-IIIB disease and clinical activity within 90 days of diagnosis.<br /><br />The results showed that among the 29,314 patients included in the study, the median age was 72 years, 47% were male, and 64% were White. Surgery rates varied across different disease stages, with the highest rates observed in stage II (56%) and the lowest in stage IIIB (10%). Among resected patients, 54% received surgery only, 40% received adjuvant therapy only, 2% received neoadjuvant therapy only, and 4% received both neoadjuvant and adjuvant therapy.<br /><br />Biomarker testing rates and the timeliness of testing improved over time. However, in the latest analysis quarter (Q1 2022), a significant proportion of patients did not receive timely testing for PD-L1 (17%), ALK (31%), or EGFR (19%) markers. Among stage II-IIIB resected patients, 69% were tested for PD-L1, 13% for EGFR, and 2% for ALK markers. Among those who received adjuvant chemotherapy, 22% did not have evidence of a PD-L1 test result before completing chemotherapy.<br /><br />The utilization of atezolizumab, the only adjuvant immunotherapy approved during the study period, increased over time, with an overall rate of 55%. The findings from the ML-extracted dataset were validated using a human-abstracted dataset.<br /><br />In conclusion, this study provides early real-world data on biomarker testing and treatment patterns in eNSCLC patients. While biomarker testing has increased over time, there are still potential gaps in biomarker-informed treatment decisions that require further investigation. The replicated findings from the human-abstracted dataset support the use of the ML-extracted dataset for real-world research in eNSCLC.

Asset Subtitle

Jay Lee

Meta Tag

Speaker Jay Lee

Topic Early-Stage NSCLC: Progress in Pathology

Keywords

biomarker testing

treatment patterns

non-small cell lung cancer

machine learning

US clinical practice

surgery rates

adjuvant therapy

biomarker testing rates

PD-L1

atezolizumab