Bayesian Learning to Reduce Cardiac Risk of Patients With Locally Advanced Non-Small Cell Lung Cancer Based on...

Area of Science:

• Radiation oncology
• Cardiovascular medicine
• Biostatistics

Background:

• Radiation therapy for non-small cell lung cancer (NSCLC) poses a risk of heart damage.
• Individual patient susceptibility to cardiac adverse events (CAEs) varies.
• Standardized heart radiation dose constraints may not be optimal for all patients.

Purpose of the Study:

• To develop and evaluate a framework for personalized adaptive radiation treatment (PART) to minimize CAEs in locally advanced NSCLC patients.
• To implement a Bayesian risk prediction model for guiding heart dose constraints.
• To assess the feasibility and effectiveness of PART in a prospective clinical trial.

Main Methods:

• A prospective study utilizing Bayesian continuous learning and adaptation was conducted.
• A Bayesian personalized risk prediction model was developed to guide heart dose constraints.
• High-sensitivity cardiac troponin T (hs-cTnT) elevation served as a surrogate biomarker for CAEs.

Main Results:

• 100 patients with locally advanced NSCLC were enrolled; 50 received standard treatment and 50 received PART.
• The PART cohort showed a lower incidence of hs-cTnT elevation (20.5%) compared to the standard cohort (31.9%).
• Within the PART cohort, patients adhering to dose constraints had significantly lower hs-cTnT elevation (9.7%) than those exceeding them (46.2%).

Conclusions:

• Clinical implementation of the PART model is feasible for guiding treatment decisions in NSCLC.
• The PART-derived mean heart dose constraints are clinically relevant and effective.
• PART is associated with a reduced incidence of hs-cTnT elevation, indicating potential for decreased cardiac damage.