Electrocardiographic Correlates of Myocardial Remodeling After Cardiac Resynchronization Therapy Authors: Christian Toquica, Amulya Gupta, Tucker Morey, Christopher J. Harvey, Seth H. Sheldon, Madhu Reddy, Amit Noheria Published: 2026 Abbreviations QRS amplitude confidence interval cardiac resynchronization therapy electrocardiogram heart failure heart failure with reduced ejection fraction interquartile range left bundle branch block left ventricular left ventricular end‐diastolic volume left ventricular ejection fraction left ventricular end‐systolic volume New York Heart Association QRS voltage‐time integral change between follow‐up and post‐CRT values Background Cardiac resynchronization therapy (CRT) has been shown to improve survival and quality of life in patients suffering from heart failure (HF) with reduced ejection fraction (HFrEF) and electrical intraventricular dyssynchrony [1, 2]. Identifying ECG markers of electrical dyssynchrony is important to guide selection of patients who are likely to respond to CRT. Previous studies have identified prolonged QRS duration, left bundle branch block (LBBB), and larger 3D (vectorcardiographic) QRS voltage‐time integral (VTIQRS‐3D) or 3D QRS area as predictors of CRT response [3, 4, 5, 6]. Although the role of ECG in selecting patients for CRT has been extensively studied, limited literature exists on how ECG evolves after device implantation. Characterizing these post‐implant changes, especially those linked to salutary left ventricular (LV) remodeling, may provide important pathophysiological insights. In this study, we examined the changes in key ECG metrics from the immediate post‐CRT period to 6‐months follow‐up. Methods A single‐center retrospective analysis of consecutive CRT implants/upgrades between 2010 and 2014 at The University of Kansas Medical Center was performed. For each patient, the first available ECG obtained closest to 7 days after CRT implantation was designated as the “post‐CRT” ECG. The ECG performed closest to 6 months after implantation (within a 3–12‐month window) was designated as the “follow‐up” ECG. Patients were excluded if ECGs from either of these time windows were unavailable. CRT response was defined as a composite endpoint requiring both echocardiographic response [i.e., ≥ 15% reduction in left ventricular end‐systolic volume (LVESV)] and absence of terminal adverse clinical HF outcome at 1 year (i.e., all‐cause mortality, heart transplantation, left ventricular assist device implantation, initiation of home inotrope therapy, or referral to hospice). If data on LVESV was not available, then a change in left ventricular ejection fraction (LVEF) ≥ 10% was substituted for echocardiographic response. The 12‐lead representative beat ECG signals generated by the Philips DXL algorithm were processed using a custom python script. Kors's transformation matrix was applied to convert the 12‐lead ECG signals into orthogonal vectorcardiographic leads (X, Y, and Z). A scalar 3D‐ECG lead signal was then derived as the root‐sum‐square (L2 norm) of the instantaneous voltages from the orthogonal leads. For each lead, the QRS voltage‐time integral (VTIQRS) was calculated as the area under the QRS complex, and the QRS amplitude (AmpQRS) was measured as the difference between the maximum positive and negative deflections of the QRS complex. Changes in ECG variables over time were denoted by the prefix “Δ” (e.g. Δ VTIQRS‐3D), defined as the difference between the variable value on the follow‐up ECG and the post‐CRT ECG. A comparison was made between the baseline characteristics and ECG parameters between responders and non‐responders at 6 months follow‐up. We assessed the Pearson correlation coefficients (r) of echocardiographic changes with changes in ECG features. A p value < 0.05 was considered the threshold of statistical significance. Results A total of 426 patients (age 68.5 ± 11.1 years, 31.5% female, 57.3% NYHA class III) were included. Change in LVESV ≤ 15 was used to evaluate echocardiographic response in a majority (97.2%) of patients, and rest were evaluated using change in LVEF ≥ 10% (due to missing values of LVESV). Overall, 202 (47.4%) patients were CRT responders and the rest were non‐responders. Responders were less likely to have ischemic cardiomyopathy/history of myocardial infarction (39.6% vs. 50.4%, p = 0.0318) as compared to non‐responders. The baseline clinical characteristics of responders and non‐responders are detailed in Table 1. TABLE 1Baseline characteristics.Variable Overall (N = 426) Responder (n = 202, 47.4%) a Non‐responder (n = 224, 52.6%) p Age68.51 ± 11.0968.75 ± 11.2168.30 ± 11.050.6728Female (%)134 (31.5%)65 (32.2%)69 (30.8%)0.8410 Race Black White Other 59 (13.8%) 348 (81.7%) 19 (4.5%) 23 (11.4%) 167 (82.7%) 12 (5.9%) 36 (16.1%) 181 (80.8%) 7 (3.1%) 0.1639 Ischemic cardiomyopathy/ history of myocardial infarction 193 (45.3%)80 (39.6%)113 (50.4%)0.0318 NYHA class b I II III IV 3 (0.7%) 99 (23.2%) 244 (57.3%) 32 (7.5%) 2 (1.0%) 56 (27.7%) 112 (55.4%) 8 (4.0%) 1 (0.4%) 43 (19.2%) 132 (58.9%) 24 (10.7%) 0.0318History of atrial fibrillation203 (47.7%)96 (47.5%)107 (47.8%)1.0000History of ventricular tachycardia/fibrillation76 (17.8%)32 (15.8%)44 (19.6%)0.3700Chronic kidney disease265 (62.2%)109 (54.0%)156 (69.6%)0.0012Diabetes mellitus212 (49.8%)90 (44.6%)122 (54.5%)0.0517Hypertension392 (92.0%)188 (93.1%)204 (91.1%)0.5614 aLVESV≤15% was used to evaluate echocardiographic response in a majority (97.2%) of patients, and rest were evaluated using LVEF≥10% due to missing values. bWherever available.John Wiley & Sons, Ltd. Baseline characteristics. Overall (N = 426) Responder (n = 202, 47.4%) a Non‐responder (n = 224, 52.6%) Race Black White Other 59 (13.8%) 348 (81.7%) 19 (4.5%) 23 (11.4%) 167 (82.7%) 12 (5.9%) 36 (16.1%) 181 (80.8%) 7 (3.1%) Ischemic cardiomyopathy/ history of myocardial infarction NYHA class b I II III IV 3 (0.7%) 99 (23.2%) 244 (57.3%) 32 (7.5%) 2 (1.0%) 56 (27.7%) 112 (55.4%) 8 (4.0%) 1 (0.4%) 43 (19.2%) 132 (58.9%) 24 (10.7%) aLVESV≤15% was used to evaluate echocardiographic response in a majority (97.2%) of patients, and rest were evaluated using LVEF≥10% due to missing values. bWherever available. At a median follow‐up of 6.7 [IQR 4.4, 9.8] months, the mean LVEF increased from 26.4 ± 8.0 % to 37.0 ± 12.9 % (p < 0.0001) and LVESV decreased from 109.9 ± 51.4 mL to 94.3 ± 53.0 mL (p < 0.0001). QRS changes from post‐implant to the follow‐up ECG (at median 5.8 [IQR 4.6, 7.3] months post‐implant) included slight increase in the QRS duration (136.1 ± 21.8 ms to 138.0 ± 22.0 ms, p = 0.0412), and reductions in VTIQRS‐3D (59.2 ± 28.6 µVs to 56.3 ± 26.2 µVs, p = 0.0266), VTIQRS‐X (38.7 µVs ± 27.5 µVs, p = 0.0056), AmpQRS‐3D (0.94 ± 0.42 mV vs. 0.90 ± 0.37 mV, p = 0.0182), and AmpQRS‐X (0.79 ± 0.44 mV vs. 0.71 ± 0.41 mV, p = 0.0002) [Table 2]. TABLE 2Baseline and follow‐up values for clinical, echocardiographic, and ECG variables.Variable (N = 426)Post‐CRT6‐month follow‐up p Clinical variables Creatinine (mg/dL)1.44 ± 1.171.51 ± 1.210.0700BNP (pg/mL)942.68 ± 1138.55796.83 ± 1136.050.0906NYHA class2.81 ± 0.582.60 ± 0.69< 0.0001 Echocardiographic variables a LVEF (%)26.37 ± 7.9736.98 ± 12.88< 0.0001LVESV (mL)109.88 ± 51.3994.27 ± 52.95< 0.0001LVEDV (mL)161.37 ± 56.81147.06 ± 56.71< 0.0001 BiV‐paced ECG variables b , c QRS duration136.07 ± 21.76137.95 ± 22.010.0412VTIQRS‐3D 59.15 ± 28.6456.32 ± 26.240.0266VTIQRS‐X 38.67 ± 27.4835.21 ± 25.040.0056VTIQRS‐Y 24.72 ± 14.7623.86 ± 13.620.2454VTIQRS‐Z 25.82 ± 16.3726.30 ± 16.240.5444AmpQRS‐3D 0.94 ± 0.420.90 ± 0.370.0182AmpQRS‐X 0.79 ± 0.440.71 ± 0.410.0002AmpQRS‐Y 0.51 ± 0.260.50 ± 0.250.5090AmpQRS‐Z 0.55 ± 0.280.56 ± 0.290.3862AmpQRS‐V1 0.79 ± 0.510.81 ± 0.470.4160AmpR‐V1/ AmpS‐V1 3.66 ± 10.296.34 ± 25.810.0863AmpR‐I/ AmpS‐I 3.95 ± 17.062.51 ± 6.880.0845QRS frontal axis (°)127.00 [−14.75, 208.50]125.50 [0.00, 223.00]0.1833Abbreviations: BiV, biventricular; BNP, brain natriuretic peptide; LVEF, left ventricular ejection fraction, LVESV, left ventricular end systolic volume; LVEDV, left ventricular end diastolic volume, VTI; voltage‐time integral. a Median time to follow up echocardiogram was 6.7 [4.4, 9.8] months. b Baseline ECG is the first ECG post‐CRT. c Median time to follow up ECG was 5.8 [4.6, 7.3] months.John Wiley & Sons, Ltd. Baseline and follow‐up values for clinical, echocardiographic, and ECG variables. Abbreviations: BiV, biventricular; BNP, brain natriuretic peptide; LVEF, left ventricular ejection fraction, LVESV, left ventricular end systolic volume; LVEDV, left ventricular end diastolic volume, VTI; voltage‐time integral. Median time to follow up echocardiogram was 6.7 [4.4, 9.8] months. Baseline ECG is the first ECG post‐CRT. Median time to follow up ECG was 5.8 [4.6, 7.3] months. Stratification by responder status (Tables S1 and S2) showed that QRS duration did not increase in responders (p = 0.6727) but showed a statistically significant increase in non‐responders (136.3 ± 21.9 ms to 140.3 ± 23.2 ms, p = 0.0018). Responders showed a decrease in VTIQRS‐3D (62.2 ± 29.1 µVs to 53.6 ± 22.7 µVs, p < 0.0001), VTIQRS‐X (41.0 ± 28.7 µVs to 32.6 ± 21.1, p < 0.0001), VTIQRS‐Y (25.0 ± 15.6 µVs to 22.8 ± 13.0 µVs, p = 0.0017), AmpQRS‐3D (0.98 ± 0.43 to 0.89 ± 0.35 mV, p = 0.0004), and AmpQRS‐X (0.83 ± 0.45 mV to 0.69 ± 0.37, p < 0.0001). Non‐responders did not show statistically significant changes in any of these metrics. Table 3 shows correlation coefficients for change in ECG parameters (difference between follow‐up value and immediate post‐CRT value) with change in LVEF and LVESV. Positive correlation with ΔLVESV was shown by ΔVTIQRS‐3D [Pearson's r 0.21 (95% CI 0.11, 0.30), p < 0.0001], ΔVTIQRS‐X [r 0.20 (0.11, 0.29), p < 0.0001], ΔAmpQRS‐3D [r 0.17 (0.08, 0.26), p = 0.0004], ΔAmpQRS‐X [r 0.18 (0.09, 0.27), p = 0.0002]. TABLE 3Correlation between change in echocardiographic and ECG parameters.Pearson's correlation coefficients (r)ECG parameterΔLVEF p ΔLVESV p ΔQRS duration−0.12 (−0.21, −0.02)0.01470.08 (−0.01, 0.18)0.0844ΔVTIQRS‐3D −0.17 (−0.26, −0.08)0.00040.21 (0.11, 0.30)< 0.0001ΔVTIQRS‐X −0.15 (−0.24, −0.05)0.00230.20 (0.11, 0.29)< 0.0001ΔVTIQRS‐Y −0.14 (−0.24, −0.05)0.00300.07 (−0.02, 0.17)0.1337ΔVTIQRS‐Z −0.04 (−0.13, 0.06)0.40810.04 (−0.05, 0.14)0.3774ΔAmpQRS‐3D −0.15 (−0.25, −0.06)0.00110.17 (0.08, 0.26)0.0004ΔAmpQRS‐X −0.11 (−0.20, −0.02)0.02290.18 (0.09, 0.27)0.0002ΔAmpQRS‐Y −0.11 (−0.20, −0.01)0.02420.07 (−0.03, 0.16)0.1631ΔAmpQRS‐Z −0.02 (−0.12, 0.07)0.63770.09 (−0.01, 0.18)0.0746ΔAmpQRS‐V1 −0.04 (−0.13, 0.06)0.41290.07 (−0.02, 0.17)0.1442ΔAmpR‐V1/ AmpS‐V1 0.03 (−0.08, 0.13)0.6203−0.07 (−0.17, 0.04)0.2123ΔAmpR‐I/ AmpS‐I −0.03 (−0.13, 0.06)0.49560.01 (−0.09, 0.11)0.8252ΔQRS frontal axis (°)0.03 (−0.07, 0.12)0.5710−0.10 (−0.20, −0.01)0.0333 Note: “Δ” signifies difference between follow‐up and baseline. Positive numbers mean decrease in the metric at follow‐up.Abbreviations: LVEF, left ventricular ejection fraction; LVESV, left ventricular end systolic volume; VTI, voltage‐time integral.John Wiley & Sons, Ltd. Correlation between change in echocardiographic and ECG parameters. Note: “Δ” signifies difference between follow‐up and baseline. Positive numbers mean decrease in the metric at follow‐up. Abbreviations: LVEF, left ventricular ejection fraction; LVESV, left ventricular end systolic volume; VTI, voltage‐time integral. Sensitivity analysis after restricting follow‐up ECGs to 4.5–9.0 months (instead of 3–12 months post‐ablation) did not substantially change the correlation coefficients or statistical significance. Discussion In this evaluation of ECG changes associated with myocardial remodeling after CRT implantation, our key findings were as follows. First, CRT responders demonstrated stable QRS duration at follow‐up, accompanied by significant reductions in QRS VTI and amplitudes in the 3D and X leads. Second, CRT non‐responders exhibited a slight increase in QRS duration, while QRS VTI and amplitude metrics remained largely unchanged. Finally, changes in QRS VTI and amplitudes in the 3D and X (right‐to‐left or transverse) leads showed a moderate positive correlation with changes in LVESV and a negative correlation with LVEF. It is important to note that, unlike most prior studies, we compared follow‐up ECGs (3–12 months post‐implantation) with the immediate post‐CRT paced ECGs rather than with pre‐CRT baseline ECGs. The pattern of stable or decreased biventricular‐paced QRS duration in responders and increased paced QRS duration in CRT non‐responders over short‐term follow‐up has been observed across several previous studies [7, 8]. The mechanism of this may be related to shorter conduction time related to decrease in size of the LV as well as electrical remodeling leading to improved cell‐to‐cell conduction [9]. Previous work by De Pooter et al. has shown that decrease in 3D QRS area in the Z‐axis from baseline pre‐CRT ECG with left bundle branch block is associated with improved acute hemodynamic response [10], and other publications have demonstrated that this reduction in 3D QRS area is associated with subsequent echocardiographic and clinical outcomes [11, 12]. However, the evolution of biventricular‐paced vectorcardiographic ECG changes during post‐CRT period related to myocardial remodeling itself has previously not been comprehensively evaluated. The finding of decreases in QRS‐VTI and amplitudes in the 3D and X axes in responders, and the corresponding lack of such decreases in non‐responders, demonstrating the impact of the reversal of adverse myocardial eccentric remodeling on the biventricular paced electrical properties is a novel result. Limitations: A retrospective single‐center study design, automated ECG extraction, and the absence of strict temporal ECG evaluation protocols are important limitations of our study. Funding The authors have nothing to report. Ethics Statement This study was conducted under ethical approval by the Institutional Review Board of the University of Kansas. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Supporting Table 1: Baseline and follow‐up values for clinical, echocardiographic, and ECG variables in CRT responders. Supporting Table 2: Baseline and follow‐up values for clinical, echocardiographic, and ECG variables in CRT non‐responders.