Machine learning models for identifying predictors of clinical outcomes with first-line immune checkpoint inhibitor...

Area of Science:

• Oncology
• Machine Learning
• Real-World Evidence

Background:

• Immune checkpoint inhibitors (ICIs) are standard first-line therapy for advanced non-small cell lung cancer (aNSCLC) lacking driver mutations.
• Observed outcome variations in clinical trials suggest a need for better prediction models in real-world practice.

Purpose of the Study:

• To develop and validate machine learning (ML)-based survival models for predicting outcomes in aNSCLC patients receiving first-line ICI therapy.
• To identify key predictors of overall survival (OS) and progression-free survival (PFS) in this patient population.

Main Methods:

• Applied ML survival models to a real-world cohort of aNSCLC patients from a US electronic health record database.
• Evaluated model performance using concordance index (c-index) and employed explainability techniques to identify significant predictors.
• Analyzed Kaplan-Meier survival curves to compare outcomes between identified risk groups.

Main Results:

• The ML model achieved c-indices of 0.672 for OS and 0.612 for PFS.
• Significant differences in OS and PFS were observed between low- and high-risk groups (log-rank test p < 0.0001).
• Top predictors for both OS and PFS included Eastern Cooperative Oncology Group performance status, PD-L1 expression, and serum albumin.

Conclusions:

• ML-based survival models can effectively predict outcomes in real-world aNSCLC patients treated with first-line ICIs.
• Identified predictors align with clinical expectations and can help understand treatment response heterogeneity.
• Further validation on independent datasets is necessary to confirm these findings.