Cardiac tamponade during atrial fibrillation (AF) catheter ablation It was accurately predicted using a machine learning model and achieved strong discrimination and clinical utility in a large Chinese cohort.
Cardiac tamponade, a life-threatening accumulation of fluid within the pericardial sac that compresses the heart, remains a rare but fatal complication of AF catheter ablation. Although AF ablation is widely performed to control symptomatic arrhythmias, it remains difficult to identify patients at highest risk for intraoperative complications.
In a retrospective study of 1,481 patients who underwent AF catheter ablation at a tertiary hospital in Nanjing, China, between October 2014 and December 2024, researchers used machine learning techniques to develop a predictive model for cardiac tamponade. After applying minimum absolute shrinkage and selection operator regression to identify candidate variables, eight algorithms were trained and evaluated.
Risk stratification of cardiac tamponade using machine learning
Among the models tested, Extreme Gradient Boosting (XGBoost) had the best overall performance. The model achieved an area under the curve of 0.972 on the training set and 0.908 on internal validation, demonstrating good discrimination. Calibration analysis showed strong agreement between predicted and observed risks, and decision curve analysis suggested the highest net clinical benefit compared to alternative models.
SHapley Additive exPlanations (SHAP) analysis was used to interpret the model output and identify the most influential predictors. Five major determinants of cardiac tamponade were highlighted: operator experience, D-dimer level, total heparin dose, AF type, and left atrial diameter. These variables reflect the treatment technique, coagulation status, arrhythmia characteristics, and structural features of the heart.
Inclusion of operator experience highlighted the procedural component of risk, while elevated D-dimer levels and increased heparin doses pointed to the importance of anticoagulation balance during ablation.
Limitations and implications
Although the findings support the potential of XGBoost-based prediction to improve preoperative risk stratification and guide intraoperative management, there are limitations. This study was conducted at a single center and data were analyzed retrospectively. External validation across multiple institutions is required to confirm generalizability.
If validated, this cardiac tamponade prediction model could keep pace with broader advances in artificial intelligence-driven decision support in cardiology and enable personalized risk assessment prior to AF catheter ablation, improving procedure safety.
reference
Zhou L et al. Explainable machine learning for risk prediction of acute cardiac tamponade during atrial fibrillation ablation. Science representative 2026;Doi:10.1038/s41598-026-40302-2.
