Predictive Model to Guide Brain Magnetic Resonance Imaging Surveillance in Patients With Metastatic Lung Cancer:...

Area of Science:

• Oncology
• Radiology
• Machine Learning

Background:

• Brain metastasis is a frequent complication in lung cancer, leading to significant patient morbidity.
• Early detection of brain metastasis is crucial for improving outcomes, yet prediction models for targeted surveillance are lacking.

Purpose of the Study:

• To develop a machine learning-based clinicogenomic prediction model for estimating patient-level brain metastasis risk in lung cancer.
• Identify high-risk individuals for enhanced brain magnetic resonance imaging (MRI) surveillance.

Main Methods:

• A penalized regression competing risk model was developed using data from 330 lung cancer patients.
• Patient data included diagnosis, treatment, and genomic alterations, with follow-up through June 2021.

Main Results:

• The model achieved high discrimination (AUC 0.75) using features like histology, stage, age, primary site, and RB1/ALK alterations.
• High-risk patients (1-year risk >14.2%) had a significantly higher incidence of brain metastasis (30.8% vs 6.1%).
• Delayed detection in high-risk patients was associated with larger brain metastases.

Conclusions:

• The developed model effectively identifies lung cancer patients at high risk for brain metastasis.
• Intensive brain MRI surveillance for high-risk patients may reduce treatment morbidity via early detection.