Neural Net Modeling of Checkpoint Inhibitor Related Myocarditis and Steroid Response

Area of Science:

• Oncology
• Cardiology
• Artificial Intelligence

Background:

• Immune checkpoint inhibitors (ICIs) are widely used in cancer treatment.
• Rare cardiotoxicity, including myocarditis, is a serious side effect of ICIs.
• A comprehensive model for ICI-related myocarditis progression and outcomes is lacking.

Purpose of the Study:

• To develop a time-series neural network (NN) model for ICI-related myocarditis.
• To identify predictive biomarkers and determinants of outcomes for ICI-related myocarditis.

Main Methods:

• Utilized electronic medical records from ICI-treated patients with troponin elevation.
• Collected approximately 300 variables per patient, totaling 6000 data elements.
• Developed a supervised machine learning NN model using sequential clinical and laboratory data.

Main Results:

• Identified 23 patients with troponin elevation; 15 had ICI-related myocarditis.
• Troponin was the most predictive biomarker for myocarditis.
• Early and aggressive steroid treatment significantly improved survival in severe (grade 3-4) ICI-related myocarditis.

Conclusions:

• Supervised learning NNs can model rare events like ICI-related myocarditis.
• The model provides insights into disease progression and treatment effectiveness.
• Early detection biomarkers and symptoms are key for timely, life-saving steroid interventions.