2023 World Conference on Lung Cancer (Posters)-P2.03. Predicting Risk of Distant Brain Failure with Radiosurgery and Systemic Therapy in a Diverse NSCLC Brain Metastasis Population

P2.03. Predicting Risk of Distant Brain Failure with Radiosurgery and Systemic Therapy in a Diverse NSCLC Brain Metastasis Population - PDF(Slides)

Back to course

Pdf Summary

A study conducted at Miami Cancer Institute aimed to develop a machine-learning algorithm to predict the risk of distant intracranial failure (DIF) in a diverse population of patients with non-small cell lung cancer (NSCLC) brain metastasis. The study analyzed a total of 844 brain metastases in 226 patients, with a median age of 64 years and 55% being female. The expression of programmed cell death ligand 1 (PDL1) was assessed, with 27% having high expression, 36% moderate expression, and 25% negative expression. After a median follow-up of 14.5 months, the 1-year cumulative incidence of DIF was 52.1%. A machine-learning model was developed, generating 100 trees and including variables such as Karnofsky Performance Score (KPS), age, use of immunotherapy, gender, and use of chemotherapy. The model had an accuracy of 76% in predicting the risk of DIF.<br /><br />The methodology included the inclusion of consecutive NSCLC brain metastasis patients treated with stereotactic radiosurgery (SRS) between 2017 and 2021. Patient characteristics, treatment details, and outcomes were extracted from electronic medical records. Random survival forest analysis was used to predict the risk of DIF, taking into account various factors such as time to DIF, age, gender, KPS, number and size of lesions, extracranial disease burden, molecular status, and systemic therapy type. The data was divided into training and test sets, and the ML model was evaluated based on its ability to predict the risk of DIF.<br /><br />The study concluded that the developed ML model had an accuracy of 76% in classifying patients and estimating the risk of DIF. The model could be used in predicting intracranial tumor control and treatment selection for high-risk patients, including considering alternative systemic therapy or whole-brain radiotherapy. The findings of this study provide insights into the prediction and management of DIF in patients with NSCLC brain metastasis.

Asset Subtitle

Rupesh Kotecha

Meta Tag

Speaker Rupesh Kotecha

Topic Metastatic NSCLC: Local Therapies

Keywords

Miami Cancer Institute

machine-learning algorithm

distant intracranial failure

non-small cell lung cancer

NSCLC brain metastasis

programmed cell death ligand 1

PDL1 expression

Karnofsky Performance Score

immunotherapy

chemotherapy