Article Impact Level: HIGH Data Quality: STRONG Summary of Journal of the American College of Cardiology, 81(10), 949–961. https://doi.org/10.1016/j.jacc.2022.12.024 Dr. Curtis Ginder et al.
Points
- The study aimed to determine if daily remote monitoring data from ICD and CRT-D devices can predict clinically significant ventricular fibrillation (VF) or ventricular tachycardia (VT).
- The investigators conducted a post hoc analysis of a randomized, controlled trial of 2,718 patients to develop separate multivariable logistic regression and neural network models to predict appropriate device therapies.
- Logistic regression identified shock lead impedance and ventricular ectopy as significantly associated with increased risk of appropriate device therapy (sensitivity 39%, specificity 91%, area under the curve [AUC]: 0.72).
- Neural network modeling yielded significantly better predictive performance with a sensitivity of 54%, specificity of 96%, and AUC of 0.90. It also identified patterns of change in atrial lead impedance, mean heart rate, and patient activity as predictors of appropriate therapies.
- The study findings indicate that remote monitoring data collected by ICD and CRT-D devices can predict malignant ventricular arrhythmias, which has significant implications for managing patients with heart failure and ICD or CRT-D devices.
Summary
In the post hoc analysis, 151 patients received appropriate device therapies, consisting of 141 shocks and ten anti-tachycardia pacing (ATP) treatments. The patients included in the study had a mean age of 64 ± 11 years, with 26% women and 64% ICD. Logistic regression analysis identified shock lead impedance and ventricular ectopy as significantly associated with increased risk of appropriate device therapy (sensitivity 39%, specificity 91%, area under the curve [AUC]: 0.72). In comparison, neural network modeling yielded significantly better predictive performance (p < 0.01 for comparison) with a sensitivity of 54%, specificity of 96%, and AUC of 0.90.
Furthermore, the neural network model identified patterns of change in atrial lead impedance, mean heart rate, and patient activity as predictors of appropriate therapies. This complements and expands upon variables of interest in the conventional logistic regression analysis. The authors suggested that additional studies and risk models combining conventional and neural network approaches are indicated to assess complex and nonlinear interactions between remote monitoring data and potentially enhance dynamic risk stratification for VT and VF.
Overall, the study findings indicate that daily remote monitoring data collected by ICD and CRT-D devices can be utilized to predict malignant ventricular arrhythmias. This has significant implications for managing patients with heart failure and ICD or CRT-D devices, as it may help predict appropriate device therapies before they occur.
Link to the article: https://www.jacc.org/doi/10.1016/j.jacc.2022.12.024
References Ginder, C., Li, J., Halperin, J. L., Akar, J. G., Martin, D. T., Chattopadhyay, I., & Upadhyay, G. A. (2023). Predicting malignant ventricular arrhythmias using real-time remote monitoring. Journal of the American College of Cardiology, 81(10), 949–961. https://doi.org/10.1016/j.jacc.2022.12.024
