Complication probability models for radiation-induced heart valvular dysfunction: do heart-lung interactions play a...

Area of Science:

• Radiation Oncology
• Medical Physics
• Cardiovascular Research

Background:

• Thoracic irradiation, particularly for Hodgkin lymphoma, can lead to cardiac toxicity, including radiation-induced valvular disease (RVD).
• Accurate prediction of RVD risk is crucial for optimizing radiotherapy planning and patient outcomes.
• Normal tissue complication probability (NTCP) models are used to estimate the likelihood of such complications.

Purpose of the Study:

• To compare the effectiveness of different NTCP models in predicting RVD after thoracic irradiation.
• To evaluate the predictive performance of models based solely on heart dose-volume histograms (DVHs) versus those incorporating lung parameters.

Main Methods:

• Analysis of 90 Hodgkin lymphoma survivors treated with 3D conformal radiotherapy (median dose 32 Gy).
• Extraction of heart and lung DVHs to fit Lyman-Kutcher-Burman (LKB) and Relative Seriality (RS) NTCP models.
• Model performance assessed using Area Under the Curve (AUC) from receiver operating characteristic analysis, with bootstrap refitting for robustness.

Main Results:

• NTCP models using only heart DVHs yielded AUCs of 0.67 (LKB) and 0.66 (RS).
• Models using only lung DVHs showed similar performance (AUCs 0.68 for LKB, 0.66 for RS).
• A multivariate model incorporating maximum heart dose, heart volume, and lung volume achieved the highest AUC of 0.82, indicating superior predictive capability.

Conclusions:

• NTCP models relying solely on heart dose-volume distributions are insufficient for predicting radiation-induced valvular disease.
• Improved RVD risk prediction necessitates incorporating both heart and lung volume parameters, highlighting the significance of heart-lung interactions.