Risk stratification in nonischemic left ventricular dilated cardiomyopathy (LVCM) remains challenging. The identification of irreversibility of LVCM could help establish prognosis, predict response to therapies to improve left ventricular (LV) function, and determine necessity for implantable cardioverter-defibrillator (ICD) implant.

Macroscopic scar suggested by bipolar voltage electroanatomic mapping (EAM) and cardiac magnetic resonance (CMR) imaging can be identified only in a minority of patients with LVCM and, when present, frequently its size does not correlate with the degree of global LV systolic dysfunction ((1),2). An alternative explanation for persistent impairment of LV function in this setting may be diffuse microfibrosis ((3),4).

The purpose of this study was to determine, in patients with LVCM, whether unipolar voltage mapping is useful for identifying myocardium demonstrating irreversible cardiomyopathy in the absence of bipolar electrogram or CMR evidence of macroscopic scar.

We examined patients undergoing catheter ablation of symptomatic ventricular premature depolarization (VPD) or ventricular tachycardia (VT) at the Hospital of the University of Pennsylvania from January 2003 to January 2011. All procedures were performed following the institutional guidelines of the University of Pennsylvania Health System and all patients provided written informed consent. Detailed LV electroanatomical data were obtained from: 1) patients with irreversible LVCM; and 2 reference groups: 2) patients with reversible VPD-mediated LV cardiomyopathy; and 3) patients with structurally normal hearts (SNH).

Twenty-four patients (23 men, mean age: 61 ± 13 years, LV ejection fraction [LVEF]: 28 ± 9%) with idiopathic irreversible LVCM (I-LVCM) who underwent catheter ablation of sustained VT (22 patients) or frequent symptomatic VPD (2 patients). The diagnosis of I-LVCM was established by a persistent global LVEF ≤35% or an improvement <10% after ≥2 years of follow-up under optimal medical therapy, in the absence of potentially reversible cause, prior myocardial infarction, significant stenosis (≥75%) of major epicardial coronary arteries, or significant primary valvular abnormalities. Other causes of dilated cardiomyopathy were excluded. To identify a group with probable extensive microscopic but not extensive macroscopic scar, all patients in the I-LVCM group had to demonstrate diffuse decrease in wall motion but <15% of identifiable endocardial (ENDO) and/or epicardial (EPI) surface manifesting confluent low bipolar voltage abnormalities (<1.5 mV for ENDO and <1.0 mV for EPI) ((5),6). CMR imaging was also used, when possible, to identify sizeable intramural or EPI macroscopic scar. The 24 patients were identified after excluding those from a larger group of 156 patients with LVCM who either had >15% LV ENDO or EPI surface area with confluent bipolar voltage abnormalities or incomplete voltage maps.

Fourteen patients (9 men, mean age: 42 ± 21 years, LVEF: 31 ± 10%) with reversible VPD-induced LV cardiomyopathy (R-LVCM) and LV systolic dysfunction at the time of detailed ENDO EAM served as 1 reference group to define the specificity of the observations. Documented VPD burden was >20% or 20,000/24 h in all cases. Catheter ablation alone was effective in eliminating VPD in 10 of 14 patients. In the remaining 4 patients, VPD elimination was achieved with adjunctive medical therapy. The diagnosis of R-LVCM was established by the normalization of LV systolic function (LVEF ≥50% after treatment and improvement ≥10% from baseline) and absence of coronary or valvular heart disease.

Seventeen patients (13 men, mean age: 41 ± 11 years, LVEF ≥55%) with SNH who underwent detailed ENDO LV mapping and catheter ablation of idiopathic VPD/VT served as the second reference group. Absence of structural heart disease was confirmed with transthoracic echocardiography, stress testing, and/or coronary angiography.

All patients underwent transthoracic echocardiography at the time of the procedure. LVEF, LV end-diastolic diameter, LV end-systolic diameter, interventricular septal and LV posterior wall thickness were evaluated. CMR imaging was performed to further define the presence of macroscopic scar in selected patients (within 3 months of ablation procedure). CMR was performed on a 1.5-T scanner (Avanto, Siemens, Erlangen, Germany) using a standard protocol that included assessment of delayed gadolinium enhancement. To reduce the risk of radiofrequency-related heating of the intracardiac lead in patients with ICDS, a non-balanced echo sequence (TurboFLASH) was used for cine imaging to reduce energy deposition and device artifacts (7).

Detailed maps of the ENDO LV surface were obtained during sinus rhythm or ventricular pacing using the CARTO EAM system (Biosense Webster Inc., Diamond Bar, California) and a 4-mm standard tip catheter (NaviStar, Biosense Webster Inc.) or 3.5-mm open irrigated-tip catheter (Navistar Thermocool, Biosense Webster Inc.). Bipolar (bandpass filtered at 30 to 500 Hz) and unipolar (bandpass filtered at 1 to 240 Hz) electrograms were recorded and displayed at 200 mm/s sweep speed. Wilson central terminal was used as an indifferent electrode to record unipolar electrograms. A detailed assessment of individual electrogram characteristics was made off-line before being displayed as 3-dimensional voltage maps to: 1) ensure a correct measurement of peak-to-peak voltage amplitude; 2) avoid the repolarization signal on unipolar electrograms; and 3) exclude noise and pacing artifact. To ensure adequate sampling density and a complete representation of voltage distribution, a fill threshold of ≤20 mm was maintained. Valvular sites were identified and intracavitary points with poor contact were edited and excluded.

EPI mapping was performed at the decision of the primary operator and based on clinical criteria such as characteristics of VT on surface 12-lead electrocardiogram <!---->and/or failure of prior ENDO ablation. EPI access was obtained using percutaneous techniques described by Sosa et al. (8). A 4-mm standard tip or 3.5-mm open irrigated-tip catheter was also used. The fill threshold was maintained at ≤20 mm and all aspects of the LV EPI were sampled.

Abnormal LV bipolar signal amplitude was identified according to previously established criteria for ENDO (≤1.5 mV) and EPI (≤ 1.0 mV) ((5),6). To avoid overestimation of EPI scar due to areas of “normal” fat distribution, confluent areas of low bipolar voltage (<1.0 mV) also had to demonstrate >20% abnormal electrograms including: 1) wide, >80 ms duration; 2) split, 2 or more distinct components with >20 ms isoelectric segment between peaks of individual components; or 3) late, distinct electrograms with onset after the end of the QRS complex (6).

The area of interest for the study was the LV ENDO surface on the electroanatomical map with no evidence of macroscopic scar according to bipolar voltage EAM (ENDO or opposite EPI) and/or CMR criteria. Confluent areas (>2 cm²) of bipolar electrogram abnormality were identified on the LV ENDO and EPI electroanatomical maps. When detected, the area of bipolar abnormality on the epicardium was projected on the endocardium using the CARTO software mesh feature (Figure 77_gr1). If CMR was available, areas of delayed gadolinium enhancement consistent with macroscopic scar were also identified and located on the LV ENDO electroanatomical map using the standardized myocardial 17-segment model (9) (Figure 77_gr2). The aggregate of probable macroscopic scar areas on the LV ENDO map thus defined was excluded, leaving the remaining larger area of myocardium as the area of interest for further signal analysis (Figure 77_gr3).

The area of probable macroscopic fibrosis on the LV ENDO map according to the defined criteria was measured and its proportion to the total LV ENDO area calculated. As the purpose of the study was the evaluation of dysfunctional but otherwise “normal” myocardium according to bipolar voltage EAM and CMR imaging criteria, patients manifesting areas of probable macroscopic scar >15% in the LV ENDO map were not included.

Unipolar voltage amplitudes of LV ENDO electrograms within the area of interest as defined were analyzed. Abnormal LV ENDO unipolar signal amplitude was defined as ≤8.27 mV, based on previously published criteria (10). To further maximize specificity of unipolar recordings for identifying abnormal myocardium, points within 1 cm around the mitral annulus and within 1 cm around areas of probable macroscopic scar (according to EAM and/or CMR criteria) were excluded (Figure 77_gr3). Differences in unipolar electrogram amplitude distributions among the 3 groups were evaluated. A receiver-operating characteristic (ROC) analysis was performed to evaluate the ability of the 8.27-mV cutoff value to identify amplitude differences between the groups.

The extent of confluent areas of unipolar voltage abnormality (AUA) was also measured within the area of interest using the software and measurement algorithm included in the CARTO system. The percent AUA of the total ENDO LV surface area was determined and compared among the 3 study groups. Two independent observers analyzed the AUA and the interobserver variability of these measurements was evaluated.

Continuous data are expressed as mean ± SD or median (interquartile range) when appropriate. All continuous data were tested using the 1-sample Kolmogorov-Smirnov test against a normal distribution. One-way analysis of variance was used for comparisons of normally distributed continuous variables among the 3 groups. When appropriate, further post-hoc subgroup analysis was performed using the Tukey correction. Comparisons of non-normally distributed continuous variables among the 3 groups were performed using the Kruskal-Wallis test. Differences in unipolar electrogram amplitude distributions were also evaluated using the Kolmogorov-Smirnov test. Discrimination power of unipolar electrogram amplitude and percent AUA was evaluated with ROC analysis. Multiple linear regression was used to examine the relation between percent AUA and irreversibility, adjusting for age, sex, presence of antiarrhythmic drugs at the time of procedure, and LV size. The intraclass correlation coefficient was used to measure the degree of agreement between 2 independent observers, assuming a 2-way random model with absolute agreement. For comparison of noncontinuous variables, the chi-square test was used. A p value ≤0.05 was considered statistically significant.

Baseline characteristics of the 55 patients in the study are listed in (Table 1). Patients in the I-LVCM group were older (61 ± 13 years) than those in the R-LVCM (42 ± 21 years) and SNH groups (41 ± 11 years) (p < 0.01). Female prevalence was also lower in the I-LVCM group (4%) than in the R-LVCM (36%) and SNH (23%) groups (p = 0.04). Presence of antiarrhythmic drugs at the time of procedure was significantly higher in the I-LVCM group (p < 0.01). Mean New York Heart Association functional class was similar in the groups of patients with LV systolic dysfunction: 1.6 ± 0.8 in the R-LVCM group and 2.1 ± 0.9 in the I-LVCM group (p = 0.12).

Main echocardiographic findings are listed in (Table 1). Groups with irreversible and reversible LVCM exhibited no difference in the degree of LV systolic dysfunction (LVEF: 28 ± 9 vs. 31 ± 10%; p = 0.33) at the time of the procedure. LV size was larger in the I-LVCM group, reflected by greater LV end-diastolic diameter (66 ± 8 mm vs. 54 ± 7 mm; p <0.01) and LV end-systolic diameter (56 ± 10 mm vs. 44 ± 7 mm; p < 0.01), than in the R-LVCM group. There were no significant differences in LV wall thickness between the groups (mean interventricular septal and posterior wall size: 10.6 ± 1.9 mm and 9.8 ± 1.7 mm in the I-LVCM group; 9.9 ± 1.8 mm and 9.4 ± 1.7 mm in the R-LVCM group; and 9.8 ± 1.5 and 9.7 ± 1.7 mm in the SNH group; p = 0.49 and 0.79, respectively). Eight patients (33%) in the I-LVCM group and 2 (14%) in the R-LVCM group exhibited some degree of LV hypertrophy (all mild except 1 moderate in the I-LVCM group). All patients in the SNH group exhibited normal LV size and function and normal wall thickness.

A moderate negative correlation between LV chamber size and LV function was found among patients with LV systolic dysfunction, I-LVCM and R-LVCM groups (r = −0.58, p < 0.001). A moderate significant correlation between LV chamber size and irreversibility of LV function was also found among patients in these groups (r = 0.59, p < 0.001).

All patients underwent detailed LV ENDO EAM (189 ± 99 points sampled). Twenty-two patients (40%) also underwent detailed EPI EAM (409 ± 184 points): 17 (71%) in the I-LVCM group; 2 (14%) in the R-LVCM group; and 3 (18%) in the SNH group. CMR was available in 19 patients (34%): 7 (29%) in the I-LVCM group; 9 (64%) in the R-LVCM group; and 3 (18%) patients in the SNH group. Overall, 19 patients (79%) in the I-LVCM group and 9 patients (64%) in the R-LVCM underwent EPI EAM and/or CMR in addition to ENDO EAM.

An area of probable macroscopic scar, based on either ENDO ± EPI bipolar voltage EAM and/or CMR criteria, was present in 22 patients (92%) with I-LVCM and occupied a mean ENDO area of 20.4 ± 13.1 cm², which represented 8.6 ± 4.7% (range 0% to 14.8%) of total LV ENDO surface area. In the R-LVCM group, an area of probable macroscopic scar was present in 6 patients (43%), with a mean area of 4.1 ± 6.9 cm², which represented 2.7 ± 3.9% (range 0% to 10.5%) of LV surface area. The proportional area of macroscopic scar was significantly larger in the I-LVCM group than in the R-LVCM group (p < 0.001). However, when present, the total area of macroscopic scar was <15% of the LV ENDO surface area in all cases, which is consistent with described study entry criteria. None of the 17 patients in SNH group exhibited any area macroscopic scar.

A total of 7,514 LV ENDO recordings were obtained from the area of interest: 4,148 recordings in the I-LVCM group; 1,342 in the R-LVCM group; and 2,024 in the SNH group. Main findings of LV EAM are summarized in (Table 2). The unipolar signal amplitude was lower in the I-LVCM group (median: 7.6 mV [interquartile range (IQR): 5.5 to 9.7]) than in the R-LVCM (median: 13.2 mV [IQR: 10.4 to 16.2 mV]) and SNH groups (median: 16.3 mV [IQR: 13.6 to 19.8 mV]) (p < 0.001). The shift in unipolar voltage distribution observed in the I-LVCM group was also reflected by an increase in distribution skewness, from 0.38 ± 0.05 and 0.30 ± 0.07 in the SNH and R-LVCM groups to 0.72 ± 0.04 in the I-LVCM group (p < 0.001) (Figure 77_gr4).

Low unipolar amplitude, <8.27 mV, was identified in 2,437 (58.8%) signals in the I-LVCM group, 132 (9.8%) in the R-LVCM group, and 24 (1.2%) in the SNH group. The ROC analysis for unipolar electrogram amplitude in patients with LV systolic dysfunction (R-LVCM and I-LVCM groups) showed an area under the curve of 86.1% and demonstrated that the 8.27-mV amplitude cutoff had a 58.9% sensitivity and 90.2% specificity for detecting an AUA related to irreversibility of LV dysfunction (p < 0.001).

AUA >5 cm² in size were present in 24 of 24 patients (100%) in the I-LVCM group, 7 of 14 (50%) in the R-LVCM group, and 1 of 17 (6%) in the SNH group (Table 2). Median unipolar signal amplitude within AUA (<8.27 mV) was smaller in the I-LVCM group (5.84 ± 1.61 mV) than in the R-LVCM (7.17 ± 1.00 mV) and SNH (7.47 ± 1.07 mV) groups (p < 0.001).

The percent AUA represented a larger proportion of the total LV ENDO surface area in the I-LVCM group (median: 64.7% [IQR: 47.5% to 75.9%]) than in the R-LVCM (median: 5.2% [IQR: 0.0% to 19.1%]) and SNH (median: 0.1% [IQR: 0.0% to 0.9%]) groups (p < 0.001). There was no significant difference in percent AUA between the R-LVCM and SNH groups (p = 0.26) (Figs. (Figure 77_gr5) and Figure 77_gr6). The ROC analysis for the percent AUA among patients with LV systolic dysfunction (I-LVCM and R-LVCM groups) showed an area under the curve of 99.7% and indicated that a 32% cutoff value for the percent AUA is 96% sensitive and 100% specific for identifying irreversibility of LV dysfunction (p < 0.001) (Figure 77_gr7). The intraclass correlation coefficient for AUA measurement was 1.00 (range 0.99 to 1.00), suggesting low interobserver variability.

Because of the demonstrated differences in the groups with respect to age, sex, presence of antiarrhythmic drugs, and LV size (LV end-diastolic diameter), a further analysis with multiple linear regression was performed to evaluate their effect on the unipolar voltage findings in patients with low LVEF (R-LVCM and I-LVCM groups). This analysis found that only irreversibility of LVCM was an independent predictor of a higher percent AUA (p < 0.001). Sex, presence of antiarrhythmic drugs, and LV size were not independent predictors on multivariate analysis (p = 0.89, 0.38, and 0.95, respectively). Interestingly, this analysis found a trend of older age predicting larger percent AUA (p = 0.09).

An additional subgroup analysis based on age was performed to further evaluate its effect on percent AUA among patients with low LVEF (R-LVCM and I-LVCM groups). In patients below 65 years of age, the median percent AUA was 51.0% (IQR: 41.1% to 74.3%) in the I-LVCM group and 2.7% (IQR: 0.0% to −8.9%) in the R-LVCM group. In patients above 65 years of age, the median percent AUA was 69.6% (IQR: 55.7% to 76.8%) in the I-LVCM group and 19.5% (IQR: 19.2% to 24.6%) in the R-LVCM group. According to these findings, the analysis based on age adjusting the previous 32% cutoff of percent AUA to 22% for younger patients (<65 years) and to 35% for older patients (>65 years) demonstrated a 100% sensitivity and 100% specificity to detect irreversibility of LV dysfunction among patients of both age-based subgroups, below and above 65 years (p < 0.001 and p = 0.01, respectively) (Figure 77_gr7).

The ROC analysis for unipolar electrogram amplitude estimated a 53.7% sensitivity and 94.9% specificity for the 8.27-mV cutoff when performed in the subgroup of younger patients (<65 years). Interestingly, this age-based ROC analysis also suggested that a higher cutoff for unipolar voltage of 9.65 mV would increase the sensitivity to 69.9% while maintaining the same high specificity of 90.0% to identify AUA. Using the 9.65-mV cutoff in the subgroup of patients <65 years in age, the median percent AUA was 74.8% (IQR: 58.2% to 81.4%) in the I-LVCM group, 5.3% (IQR: 1.0% to 26.4%) in the R-LVCM group, and 0.8% (IQR: 0% to 4.7%) in the SNH group. Using the higher cutoff in patients <65 years of age, the percent AUA was still <10% in 8 of 11 patients (73%) in the R-LVCM group, whereas >54% in 12 of 13 patients (92%) in the I-LVCM group. The higher unipolar voltage cutoff in the population <65 years of age provided a greater potential for discrimination between the I-LVCM and R-LVCM groups, with a larger difference between the ranges of percent AUA, while maintaining no difference between the R-LVCM and SNH group (p = 0.18) (Figure 77_gr8).

This study demonstrates the ability of ENDO unipolar voltage EAM, <8.27 mV, to identify permanent LV dysfunction in the absence of macroscopic scar defined with standard bipolar EAM and CMR criteria. Our data reveal a progressive shift in the median amplitude of unipolar LV ENDO electrograms from patients without structural heart disease, 16.3 mV, to those with reversible, 13.2 mV, and irreversible, 7.6 mV, LVCM. Patients with I-LVCM demonstrated confluent low unipolar voltage areas comprising a median of 64.7% (IQR: 47.5% to 75.9%) of the total LV ENDO surface. In contrast, patients with R-LVCM and comparable LV systolic dysfunction at the time of EAM and patients with SNH exhibited either an absence of or smaller AUA on the LV ENDO maps: median: 5.2% (IQR: 0.0% to 19.1%) in the R-LVCM group; and 0.1% (IQR: 0.0% to 0.9%) in the SNH group. A cutoff AUA of >32% of total ENDO surface area accurately identified irreversibility of LVCM in patients with systolic dysfunction.

Using an 8.27-mV reference value, we have recently shown that unipolar ENDO LV EAM can help identify EPI bipolar low voltage consistent with macroscopic scarring in patients with LVCM and normal ENDO bipolar voltage (10). It appears that the presence of an AUA not correlating with ENDO or EPI scar and the finding of a significantly lower mean unipolar signal amplitude in the LV of patients with LVCM compared with those without heart disease suggest that the larger “antenna” of unipolar electrograms could also detect a more diffuse underlying myocardial process with persistent LV dysfunction. Moreover, an inverse relation between amplitude of unipolar endocardial recordings and myocardial fibrosis burden has been already described in an experimental model of nonischemic cardiomyopathy (11). In the absence of macroscopic fibrosis, diffuse microscopic fibrosis may explain the permanent impairment of LV function in the setting of LVCM. Based on findings described in previous anatomic studies, we assert that a marked unipolar abnormality (<8.27 mV), in the absence of macroscopic scar, is consistent with a larger degree of myocardial damage and diffuse microscopic fibrosis ((3),4). Interestingly, the use of a higher cutoff at 9.65 mV would further optimize unipolar signal amplitude discrimination in younger patients (<65 years), providing greater sensitivity and still being specific for detecting abnormalities associated with irreversibility in LV dysfunction and probable fibrosis. The reason why age may have an effect on unipolar electrograms remains unknown and could be due to some degree of age-related fibrosis in older patients.

The ability of unipolar electrograms to identify irreversibility of LV systolic dysfunction in this setting has important clinical implications and may potentially assist clinicians in assessing sudden cardiac death risk and the need for ICD implant. In our study, 4 of 6 patients with R-LVCM had an ICD implanted based on a primary prevention indication because of severe LV systolic dysfunction in the absence of sustained ventricular arrhythmias. The identification of the potential for recovery of LV function in these 4 patients based on the presence of limited unipolar electrogram abnormalities would have been helpful for better risk stratification and avoiding unneeded ICD implantation. Moreover, a better quantification of LV myocardial fibrosis burden provided by LV ENDO unipolar signals, and not possible with other current techniques, might potentially help in predicting response to medical, ablative, or device therapy and prognosis in the setting of LVCM and warrant additional study.

The study population underwent EAM during catheter ablation of ventricular arrhythmias. Moreover, the patients included in the I-LVCM group included those patients with sustained VT and the R-LVCM group included patients with premature VPD–induced cardiomyopathy. Thus, although hopeful, these findings might not be extrapolated to the entire population of patients with LVCM.

An effort was made to detect and exclude from the analysis all areas of probable macroscopic scar using either EAM ENDO and, when available, EPI mapping or CMR. However, 5 patients (21%) in the I-LVCM group and 5 patients (36%) in the R-LVCM group had neither EPI EAM nor CMR for identification of areas of probable macroscopic scar. Importantly, the fact that these patients did not require an epicardial approach for success of catheter ablation of VT makes the presence of significant epicardial macroscopic substrate less likely. Nevertheless, prospective validation of the current findings in a cohort without a history of VT with a lower probability of any bipolar voltage abnormality or macroscopic scar is certainly warranted.

Furthermore, our study was an observational analysis. However, electroanatomical data collection in patients with ventricular arrhythmias was systematically and prospectively performed in a detailed fashion whenever attempting to characterize the arrhythmia substrate as an institutional strategy, making the data robust and highly representative. In addition, artifact from the pulse generator or lead remains a limitation for cardiac imaging in patients with devices to define scar, but usual location (anterior or apical) does not generally alter image characterization at the lateral or the septal aspect of the LV, which are common locations for macroscopic scar in the setting of LVCM (12).

Finally, this study does not provide histopathological correlation in areas of unipolar voltage abnormality. Previous studies have described the diffuse microscopic process of myocardial fibrosis responsible for irreversibility of LV dysfunction in the setting of LVCM. Although our observations should be confirmed by ongoing studies in explanted hearts, the correlation of unipolar voltage abnormality and irreversibility of LVCM suggests a probable relation to myocardial fibrosis.

LV unipolar voltage abnormality (>32% of LV ENDO surface area) in the absence of macroscopic scar defined by bipolar voltage abnormality or CMR criteria suggests irreversibility of LV myocardial dysfunction consistent with a more diffuse microscopic fibrotic process. Detailed unipolar voltage may serve as valuable prognostic tool in patients with LVCM for identifying irreversible LV dysfunction and might facilitate clinical decision making.