WCLC 2025 - Posters & ePosters-EP.08.13 Machine Learning Prediction of Overall Survival in Stage III Unresectable NSCLC: Data Analysis From a Single Center Prospective Cohort

EP.08.13 Machine Learning Prediction of Overall Survival in Stage III Unresectable NSCLC: Data Analysis From a Single Center Prospective Cohort

Back to course

Pdf Summary

This study presents a retrospective analysis of a prospective cohort of 45 patients with locally advanced (stage III) unresectable non-small-cell lung cancer (NSCLC), not driven by oncogene addiction, treated with concurrent chemoradiotherapy (cCRT) followed by Durvalumab immunotherapy (IT). Given the current standard of care combining cCRT and Durvalumab, the research aimed to evaluate overall survival outcomes and investigate the immunologic impact of radiotherapy (RT), particularly concerning radiation dose to mediastinal nodal stations.<br /><br />All mediastinal nodal stations—both involved and uninvolved—were meticulously contoured for each patient, and radiation doses delivered to these areas were recorded. Of note, radiotherapy to uninvolved mediastinal nodes can induce lymphopenia, potentially compromising immune responses vital for effective immunotherapy. This study sought to understand how RT dose distribution may influence treatment efficacy and immunomodulation.<br /><br />Using a Random Forest (RF) machine learning model with death as the endpoint, the cohort was split into a training group (32 patients) and a validation group (13 patients). This model achieved a high predictive accuracy with an area under the curve (AUC) of 0.8667, indicating robust ability to predict overall survival (OS). Key RT-related factors, particularly doses to mediastinal lymph nodes and heart, emerged as significant predictors of OS. Data also suggested the mediastinal RT dose could influence outcomes, emphasizing the importance of optimized radiotherapy planning.<br /><br />The authors highlight the necessity for further, deeper investigation into how RT dose distribution affects immunomodulation and overall treatment success, reinforcing that enhancing RT targeting may improve patient responses to combined chemoradiotherapy and immunotherapy. This work aligns with current literature underscoring the balance between effective tumor control and minimizing adverse immune effects during stage III NSCLC treatment.

Asset Subtitle

Sara Ramella

Meta Tag

Speaker Sara Ramella

Topic Local-Regional Non-small Cell Lung Cancer

Keywords

non-small-cell lung cancer

stage III NSCLC

concurrent chemoradiotherapy

Durvalumab immunotherapy

mediastinal nodal stations

radiation dose

lymphopenia

Random Forest model

overall survival prediction

immunomodulation