Lead Impedance Change as a Prognostic Marker for Cardiac Resynchronization Therapy Response

Background Left ventricular (LV) lead impedance monitoring is a routinely available device-based parameter that may reflect myocardial structural changes and predict cardiac resynchronization therapy (CRT) outcomes, but its prognostic role remains underexplored. When the LV lead is positioned in the coronary sinus, impedance measurements may indirectly reflect structural myocardial changes, offering potential prognostic value beyond resynchronization. Methods We retrospectively analyzed CRT recipients with serial LV impedance and echocardiographic measurements. The primary endpoint was prediction of a ≥ 10% absolute increase in LV ejection fraction (LVEF) at 12 months. Predictive models included univariate logistic regression (LR) using a 150 Ω impedance threshold, multivariable LR, neural network (NN), and eXtreme Gradient Boosting (XGBoost) with and without mean LV size (LV_mean) derived from five echocardiographic assessments and devices measurements. Results A total of 95 patients were included in the final analysis. Univariate LR (≥ 150 Ω) achieved an area under the ROC curve (AUC) of 0.86 and an accuracy of 86.5%. Multivariable LR with age, BMI, and sex yielded an AUC of 0.82. NN performance was lower (AUC = 0.68, accuracy = 65.5%). XGBoost with impedance, age, and BMI achieved an AUC of 0.721 ± 0.075, and adding LV_mean improved the AUC to 0.745 ± 0.064. Impedance change correlated negatively with LV_mean (ρ = − 0.29, p = 0.007). ROC analysis confirmed the highest discriminatory ability for univariate LR, with XGBoost + LV_mean providing a modest but consistent improvement (Fig. 1). To facilitate clinical application, we developed a nomogram based on the logistic regression model, which allows individualized prediction of CRT response (Supplementary Fig. S1). Univariate LR correctly classified 86.5% of patients, as illustrated in the confusion matrix (Supplementary Fig. S2A). Conclusion LV impedance change ≥ 150 Ω is a strong predictor of CRT response. Logistic regression offers high accuracy and clinical interpretability, whereas XGBoost incorporating LV size may improve model robustness and facilitate integration into automated, device-based decision-support systems. These findings support the integration of impedance monitoring into CRT follow-up protocols and highlight its potential for real-time, device-based risk stratification. This approach may facilitate implementation in routine CRT follow-up and device programming.