Baseline substrate and response after cardiac resynchronization therapy in non-left bundle branch block heart failure

Background: Response to cardiac resynchronization therapy (CRT) is heterogeneous in patients with non-left bundle branch block (non-LBBB) heart failure. Whether pre-implant substrate or procedural characteristics provide the more stable framework for predicting 1-year echocardiographic response remains uncertain. Methods: We retrospectively analyzed 120 non-LBBB patients undergoing CRT. The primary logistic model included left ventricular end-diastolic diameter (LVEDD), left ventricular ejection fraction (LVEF), left atrial diameter, log-transformed NT-proBNP, baseline QRS duration, fragmented QRS burden across V1?V6 leads, and pulmonary artery pressure. Missing predictor data were handled using multiple imputation with 20 datasets. Model performance was assessed using bootstrap internal validation and recalibration. A prespecified procedural extension added pacing strategy, posterolateral biventricular left ventricular lead location, left ventricular pacing threshold, and right ventricular lead position. Exploratory phenotyping and sensitivity analyses were performed. Results: Echocardiographic response occurred in 51 patients (42.5%). LVEDD (OR, 0.899 [95% CI, 0.826?0.978]; P=0.013) and LVEF (OR, 1.068 [95% CI, 1.000?1.140]; P=0.050) were the most informative predictors. The primary model showed apparent AUC 0.811 and Brier score 0.173, with optimism-corrected AUC 0.766 and calibration slope 0.765. Procedural extension showed no retained incremental value after validation. Exploratory phenotyping identified three response patterns with moderate stability. Conclusions: In non-LBBB CRT, baseline structural, biomarker, and electrocardiographic substrate provided the most stable framework for predicting 1-year echocardiographic response. Procedural variables added limited retained value, suggesting that pacing strategy should be interpreted alongside baseline substrate.