Cardiac resynchronization therapy (CRT) has revolutionized the treatment of selected patients with heart failure (HF) ((1),(2),(3),4). The clinical outcome of CRT, however, is influenced by the underlying etiology of HF, with an ischemic etiology being associated with a worse outcome ((5),(6),7). This outcome has been linked to the extent (burden) and location of myocardial scarring ((8),(9),(10),(11),12).

In ischemic cardiomyopathy (ICM), fibrosis usually follows a subendocardial or transmural distribution, in line with the perfusion territories of epicardial coronary arteries. In nonischemic dilated cardiomyopathy (DCM), fibrosis tends to be patchy, subepicardial, or midmyocardial in distribution ((13),(14),15). It has been shown that non-CRT patients with DCM and midwall fibrosis have a higher risk of mortality and unplanned hospitalizations, as well as a higher risk of sudden cardiac death (16). We hypothesized that in patients with DCM, the presence of left ventricular (LV) midwall fibrosis, assessed using late gadolinium enhancement cardiovascular magnetic resonance (LGE-CMR) imaging, predicts the clinical outcome of CRT.

Patients with DCM (n = 97) or ICM (n = 161) who were recruited from a single center (Good Hope Hospital, Birmingham, United Kingdom) and who successfully underwent CRT device implantation and CMR imaging in September 2000 to July 2009 were included in this study. HF was diagnosed on the basis of the clinical features plus echocardiographic evidence of LV systolic dysfunction. ICM was diagnosed if LV systolic dysfunction was associated with a history of myocardial infarction (17) or if there was significant documented coronary heart disease. The pattern of LGE on CMR imaging was also used to differentiate between ICM and DCM (16). Patients with hypertrophic or restrictive cardiomyopathy, primary valvular disease, or myocarditis were excluded. Patients with presumed nonischemic cardiomyopathy with fibrosis in distributions other than midwall (subepicardial, epicardial, or patchy) were excluded. The study conformed with the Declaration of Helsinki. All patients gave written informed consent, and the study was approved by the local ethics committee.

In the United Kingdom, the National Institute of Clinical Excellence guidelines on device therapy were published in 2007 (18), and therefore, most patients with either ICM or DCM before 2007 received CRT pacing (CRT-P). In contrast to other international guidelines, National Institute of Clinical Excellence guidelines recommend CRT-P rather than CRT defibrillation (CRT-D) in patients with DCM (18). With the exception of 2 patients who received CRT-D for secondary prevention, all others with DCM received CRT-P.

CRT device implantation was undertaken using standard techniques under local anesthesia. After implantation, patients were followed up in a dedicated device therapy clinic (Good Hope Hospital, Birmingham, United Kingdom). Clinical response and echocardiographic variables were assessed at 3 months. Patients in sinus rhythm underwent transmitral Doppler-directed optimization of atrioventricular delay using an iterative technique prior to discharge and at every scheduled visit thereafter. Backup atrial pacing was set at 60 beats/min, and the pacing mode was set to DDDR with an interventricular delay of 0 to 4 ms, according to the manufacturer. In the case of patients in permanent atrial fibrillation, right ventricular and LV leads were implanted and a CRT generator was used, with the atrial port plugged and the generator programmed to a ventricular triggered mode. Generators used included the Medtronic InSyncIII models 8040 and 8042, St. Jude Frontier I and II, VitatronCRT 8000, BiotronikStratos LV, and Guidant Contak Renewal TR2.

CMR imaging was performed using a 1.5-T scanner (Signa, GE Healthcare Worldwide, Slough, United Kingdom) and a phased-array cardiac coil. A short-axis LV stack was acquired using a steady state in free precession sequence (repetition time 3.0 to 3.8 ms; excitation time 1.0 ms; image matrix 224 × 224; field of view 36 to 42 cm; flip angle 45°) in sequential 8-mm slices (2-mm interslice gap) from the atrioventricular ring to apex. Acquisition was performed during gated 8-s breath-holds (20 phases). Left ventricular end-diastolic volume (LVEDV) and left ventricular end-systolic volume (LVESV) were quantified using semiautomatic manual planimetry of all short-axis steady state in free precession sequence cine images with MASS analysis software (Medis, Leiden, the Netherlands). The observer was blinded to echocardiographic and clinical data.

For scar imaging using LGE, short-axis slices identical to the LV stack were acquired using a segmented inversion-recovery technique 10 min after the intravenous administration of gadolinium-diethylenetriaminepentaacetic acid (0.1 mmol/kg). Inversion times were adjusted to null normal myocardium (260 to 400 ms). Scars were classified into subendocardial, midwall, epicardial, transmural, or patchy, according to McCrohon et al. (15). Scars in a subendocardial or transmural distribution following coronary artery territories were regarded as ischemic in etiology, whereas midwall scars and absence of scar were regarded as indicative of a nonischemic etiology. Patients were dichotomized according to presence or absence of midwall hyperenhancement (MWHE), assessed visually. Examples of scars typical of ICM and DCM with MWHE are shown in (Figure 54_gr1). Scar volume was calculated by multiplying the manually planimetered area of LGE in each slice by the slice thickness. Scar burden was expressed as a percentage of LV myocardial volume in the diastolic phase.

Intraventricular dyssynchrony was assessed using the CMR–tissue synchronization index (TSI), as previously described (19). Briefly, segmental radial wall motion data were quantified for up to 20 phases (time points) in each RR interval and fitted to an empirical sine wave function y = a + b × sin (2πt/RR + c). The mean segmental radial wall motion (a), the segmental radial wall motion amplitude (b), and the segmental phase shift of the maximum radial wall motion (c) were extracted from the fit. The CMR-TSI, a global measure of radial dyssynchrony, was expressed as the standard deviation of all segmental phase shifts of the radial wall motion extracted from the fit. This measure was previously shown to predict mortality and morbidity after CRT ((19),20). In a previous study (19), intraobserver and interobserver variabilities for CMR-TSI were 3.01% and 8.84%, respectively.

Two-dimensional echocardiography was performed using systems 5 and 7 with EchoPAC (General Electric Healthcare Worldwide). LVEDV and LVESV were assessed by planimetry of apical 4-chamber views and the Simpson equation. LV reverse remodeling was defined as a ≥15% reduction in LVESV (21) at 3 months.

A clinical response to CRT was quantified in terms of the composite clinical score, according to which response was defined as survival with freedom from HF hospitalizations for 1 year after implantation as well as improvement by ≥1 New York Heart Association (NYHA) functional class or by ≥25% increase in 6-min walking distance. The primary endpoint was cardiovascular death (including transplantation). Secondary endpoints included death from any cause, the composite endpoints of cardiovascular death/unplanned hospitalization for worsening HF, death from any cause/unplanned hospitalization for major adverse cardiovascular events (MACE), and sudden cardiac death/hospitalization for major arrhythmic events. Hospitalizations for worsening HF, myocardial infarction, unstable angina, arrhythmia, stroke, or pulmonary embolism were included in this endpoint. The first event was included in the analysis. Sudden cardiac death was defined as a “natural, unexpected death due to cardiac causes, heralded by an abrupt loss of consciousness within 1 hour of the onset of acute symptoms” (22). Death from pump failure was defined as “death after a period of clinical deterioration in signs and symptoms of heart failure despite medical treatment” (23). Mortality data was collected prospectively through medical records and, when appropriate, from interviews with patients' caregivers. Clinical outcome data were collected every 3 months by an investigator who was blinded to clinical and imaging data. Events were adjudicated by the investigators (P.W.F. and F.L.) every 3 months.

Continuous variables are expressed as mean ± SD. Normality was tested using the Shapiro-Wilk test. Comparisons between normally distributed continuous variables were made using analysis of variance with the Scheffe F procedure for multiple comparisons. Variables that did not follow a normal distribution, such as NT pro–brain natriuretic peptide, were log-transformed for statistical analyses. Categorical variables were analyzed using chi-square tests and the Scheffe post hoc test. Changes in variables from baseline to follow-up were analyzed using paired t tests and Wilcoxon matched-pairs analyses. The ability of MWHE to predict the various endpoints was assessed using Kaplan-Meier survival curves and the log-rank (Mantel-Cox) test, as well as Cox proportional hazards analyses. For large estimates of the coefficient, as in our case of sudden cardiac deaths in the DCM +MWHE group, the standard error is typically inflated, resulting in a lower Wald statistic, falsely considering the variable not relevant in the model. To overcome this, we used the likelihood ratio test, which is considered superior for testing the Cox regression model. The profile likelihood was used to estimate the lower 95% confidence interval (CI) bound. Variables reaching p < 0.10 on univariate analyses were considered for entry in multivariable models. Statistical analyses were performed using Statview (Cary, North Carolina) and SPSS version 15.0 (Chicago, Illinois). A 2-tailed p < 0.05 was considered statistically significant.

The baseline characteristics of patients with ICM and DCM are shown in (Table 1). Among the entire cohort, 20 of 258 patients (7.8%) with DCM had MWHE. Of the 97 patients with DCM, 20 (26%) had MWHE, and the rest of the DCM group had no myocardial scarring at all. The DCM +MWHE and DCM −MWHE groups were well matched for age, device type, comorbidities, presence of permanent atrial fibrillation, and QRS duration. The +MWHE group had a worse NYHA class (p = 0.0271) and LV ejection fraction (LVEF) (p = 0.0007), a higher LVEDV and LVESV (both p < 0.0001), lower systolic (p = 0.0048) and diastolic (p = 0.008) blood pressures, higher NT pro–brain natriuretic peptide levels (p = 0.0064), higher plasma creatinine levels (p = 0.0070), but similar estimated glomerular filtration rates than the DCM −MWHE group. These groups were also matched for treatment with angiotensin-converting enzyme inhibitors/angiotensin receptor antagonists, beta-blockers, and aldosterone antagonists, but the +MWHE group was more likely to require loop diuretics (p = 0.0257). In comparison with the DCM groups, the ICM group had a higher proportion of men and was more likely to receive CRT-D (as required by contemporaneous guidelines) but was well matched to the DCM +MWHE for age, medication, and LV volumes and LVEF. The CMR-TSI was higher in the ICM group than in the DCM +MWHE group (p < 0.0001).

After a maximum follow-up period of 3,166 days (8.7 years; median follow-up time 1,038 days [2.84 years]), total mortality was 10 of 20 (50%) in DCM +MWHE and 5 of 77 (6.5%) in DCM −MWHE. Cardiovascular mortality was 9 of 20 (45%) and 2 of 77 (2.6%) in the DCM +MWHE and DCM −MWHE groups, respectively. In the ICM group, total mortality was 53 of 161 (31.8%) and cardiovascular mortality was 49 of 161 (30.4%).

Among patients with DCM, +MWHE predicted cardiovascular mortality (hazard ratio [HR]: 18.1; p < 0.0001), the composite of total mortality or hospitalization for MACE (HR: 7.57; p < 0.0001), and the composite endpoint of cardiovascular mortality or HF hospitalizations (HR: 9.90; p = 0.0004), independent of NYHA class, QRS duration, presence of atrial fibrillation, LV volumes, LVEF, and CMR-TSI (Table 2).

Kaplan-Meier survival curves for the DCM and ICM groups are shown in (Figure 54_gr2). In multivariable analyses comprising the DCM and ICM subgroups, both DCM +MWHE (HR: 18.5 in model 1; HR: 18.6 in model 2; both p = 0.0002) and ICM (HR: 21.0; p < 0.0001) emerged as strong predictors of cardiovascular mortality, independent of NYHA class, treatment with beta-blockers, QRS duration, presence of atrial fibrillation, LV volumes, LVEF, and CMR-TSI (Table 3).

Scar burden in the DCM group was 2.12% ± 4.96% (range 0 to 26.9%). In Cox proportional hazards analyses, scar burden did not emerge as a predictor of total mortality (HR: 0.99; 95% CI: 0.87 to 1.14; p = 0.94) or cardiovascular mortality (HR: 1.02; 95% CI: 0.89 to 1.16; p = 0.82).

Of the 60 cardiovascular deaths, 46 were due to pump failure (DCM +MWHE 6 of 20 [30%]; DCM −MWHE 2 of 77 [2.6%]; ICM 38 of 161 [23.6%]). In univariate analyses comprising the DCM and ICM subgroups, both ICM (HR: 10.5; 95% CI: 2.52 to 43.5; p = 0.0012) and DCM +MWHE (HR: 14.1; 95% CI: 2.85 to 70.0; p = 0.0012) emerged as predictors of death from pump failure.

Of the 60 cardiovascular deaths, 14 were sudden cardiac deaths (DCM +MWHE 3 of 20 [15%]; DCM −MWHE 0 of 77 [0%]; ICM 11 of 161 [6.8%]). In univariate and multivariate analyses comparing with DCM −MWHE, both DCM +MWHE and ICM significantly improved the Cox regression model using the likelihood ratio test (p = 0.0029), and both DCM +MWHE (HR lower 95% CI: 2.65) and ICM (HR lower 95% CI: 4.54) emerged as predictors of sudden cardiac death.

Five patients had unplanned hospitalizations for major arrhythmic events (DCM +MWHE 1 of 20 [5.0%] for atrial fibrillation; DCM −MWHE 1 of 77 [1.3%]; ICM 1 atrial fibrillation, 1 ventricular tachycardia, and 1 ventricular fibrillation [total 3 of 161 (5.0%)]). In univariate analyses comprising the DCM and ICM subgroups, DCM +MWHE emerged as a predictor of sudden cardiac death or major arrhythmic events (HR: 16.7; 95% CI: 1.87 to 149.7; p = 0.0118), whereas ICM reached only borderline significance (HR: 7.03; 95% CI: 0.92 to 53.4; p = 0.06). In univariate analyses comprising the DCM subgroups, DCM +MWHE emerged as a predictor of sudden cardiac death or major arrhythmic events (HR: 16.1; 95% CI: 1.81 to 144.8; p = 0.0128).

Whereas no significant changes in NYHA class were observed in the DCM +MWHE group (−1.25; p = 0.17) or the DCM −MWHE group (−1.20; p = 0.07), a significant reduction in NYHA class was observed in the ICM group (−1.10; p = 0.0429). The 6-min walking distance, however, increased in all groups (DCM +MWHE +101.6 ± 75.8 m, p = 0.0007; DCM-MWHE +54.9 ± 74.2 m, p < 0.0001; ICM +63.1 ± 98.17, p < 0.0001). Quality of life scores (a reduction denoting an improvement in quality of life) decreased in all groups (DCM +MWHE −13.7 ± 40.6, p = 0.0202; DCM −MWHE −25.7 ± 29.6, p < 0.0001; ICM −18.5 ± 21.7, p < 0.0001). Responder rates, in terms of the clinical composite score, were similar across the groups (DCM +MWHE 65.0%; DCM −MWHE 80.5%; ICM 68.2%; p = 0.19).

As shown in (Figure 54_gr3), LV reverse remodeling, defined as a ≥15% reduction in LVESV, was observed in the DCM −MWHE (p = 0.0007) and ICM groups (p = 0.0428) but not in the DCM +MWHE group. Similarly, significant reductions in the LVEDV were observed in the DCM −MWHE group (p = 0.0019) and the ICM group (p = 0.0238) but not in the DCM +MWHE group. The LVEF increased in the DCM −MWHE group (p = 0.0395) and the ICM group (p = 0.05) but not in the DCM +MWHE group.

We have shown that in patients with DCM, midwall fibrosis (detected by MWHE on LGE-CMR imaging) predicts mortality and morbidity after CRT. Compared with patients without midwall fibrosis, patients with midwall fibrosis were 18 times more likely to die from cardiovascular causes after adjustment for NYHA class, beta-blocker use, QRS duration, atrial fibrillation, LVEF, and dyssynchrony. Midwall fibrosis was also predictive of the combined endpoint of total mortality or hospitalizations for MACE and the combined endpoint of cardiovascular mortality or HF hospitalizations. Patients with midwall fibrosis were less likely to exhibit LV reverse remodeling, assessed by echocardiography.

A novel finding from this study is that the outcome after CRT of patients with midwall fibrosis was similar to that of patients with ICM. Conversely, the outcome of patients with DCM and without midwall fibrosis was dramatically better. These findings have emerged in the context of major CRT trials showing that patients with nonischemic DCM have a better clinical outcome (5), as well as a better LV reverse remodeling response ((6),(7),24) to CRT. Importantly, however, no major trial has used CMR in the characterization of the etiology of HF nor have they used CMR in the differentiation between DCM with and without midwall fibrosis. It therefore remains unknown whether the superresponders to CRT described in such studies are patients with DCM without midwall fibrosis. Our findings have major implications for prognostic stratification in clinical practice as well for the design of CRT outcome trials. It would appear that patients with DCM without midwall fibrosis have a particularly low clinical event rate.

Since its description in autopsy studies (25) and in an LGE-CMR study (15), midwall fibrosis has been recognized as a prognostic marker for patients with DCM. A prospective cohort study of 65 patients with nonischemic DCM and LVEF ≤35% undergoing implantable cardioverter defibrillator therapy showed that LGE on CMR in any distribution was associated with an 8.2-fold increase in the risk of the composite endpoint of hospitalization for HF, appropriate implantable cardioverter defibrillator shocks, and cardiac death (26). Assomull et al. (16) showed that MWHE on LGE-CMR predicted the risk of death or hospitalizations in non-CRT patients with DCM. Our findings are largely consistent with these studies and extend the application of this technique to the risk stratification of patients undergoing CRT.

We found that the degree of dyssynchrony was similar in the +MWHE and −MWHE groups. Yet, the clinical outcome was worse in the +MWHE group. This suggests that the detrimental effects of midwall fibrosis on clinical outcome is mediated through mechanisms that are independent of dyssynchrony. Several aspects may be relevant in this respect. Fibrosis effectively replaces viable myofibrils, thus reducing the amount of functional myocardium. It interacts mechanically with the complex fiber architecture of the myocardium, which may lead to global mechanical effects remote from the affected area, such as LV stiffness, reduced compliance, and reduced contractile reserve. The observation that patients with midwall fibrosis did not exhibit LV reverse remodeling or an improvement in LVEF is consistent with this notion. Myocardial scars are not readily excitable ((27),28) and therefore reduce the volume of excitable myocardium available to ventricular depolarization.

It has been shown that myocardial fibrosis can form a substrate for ventricular arrhythmias ((29),30). Accordingly, midwall fibrosis has been linked to a higher risk of sudden cardiac death in patients with DCM. We too have found that midwall fibrosis predicts sudden cardiac death, as well as the composite endpoint of sudden cardiac death or major arrhythmic events. In fact, the association between midwall fibrosis (HR: 16.7; p = 0.0118) and this endpoint was stronger than for ICM (HR: 7.03; p = 0.0597). The occurrence of only 3 sudden cardiac deaths in the whole DCM group in this study, however, precluded reliable statistical analysis of midwall fibrosis in relation to sudden cardiac death. On the other hand, the LVEF was very low (16%) in comparison with that of CRT-D trials (20% to 22%) (2), and deaths from pump failure occurred relatively early in the follow-up period. Therefore, the possibility arises that patients succumbed to pump failure before the occurrence of lethal ventricular arrhythmias.

Despite worse outcomes, the symptomatic response in patients with midwall fibrosis was similar to that in patients without midwall fibrosis. This discordance between outcomes and symptomatic response after CRT is well recognized. Yu et al. (21), for example, found no relationship between LV reverse remodeling and changes in NYHA class, 6-min walking distance, or quality of life scores after CRT. Ypenburg et al. (31) also showed similar improvement in NYHA class, quality of life scores, and 6-min walking distance in patients exhibiting ≥15% reduction in LVESV and in those exhibiting a reduction in LVESV of <14% after CRT. Foley et al. (32) found similar symptomatic response rates in survivors and nonsurvivors 1 year after CRT device implantation. There is, in addition, evidence of a discordance between outcomes and symptomatic response according to etiology. In the REVERSE (Resynchronization Reverses Remodeling in Systolic Left Ventricular Dysfunction) study, in which patients in NYHA class I or II and LVEF ≤40% were randomized to CRT or no CRT for 12 months, HF etiology was not predictive of the composite clinical response (24). This is in keeping with our finding of similar symptomatic response rates in patients with and without midwall fibrosis.

CMR has already gained credence as an ideal investigation for patients with HF because it provides unparalleled quality of information on cardiac function and disease etiology. In addition, LGE-CMR is also unique in its ability to allow quantification and localization of myocardial scarring in patients with ICM, in whom it has proven to be valuable in prognostic stratification ((8),(10),11). Our findings extend the utility of CMR to the prognostic stratification of patients with DCM undergoing CRT.

This study was observational and did not include a control group on maximum tolerated pharmacological therapy only. We therefore cannot ascertain whether patients with MWHE have a worse outcome than patients not undergoing CRT, and we cannot therefore assume that patients with MWHE do not benefit from CRT. In addition, the number of patients in the +MWHE group was small. Therefore, the lack of an effect of CRT on LV reverse remodeling may be attributable to statistical underpowering. Moreover, we have not quantified the severity of mitral regurgitation, which could also contribute to differences in outcomes between the groups. The lack of systematic collection of arrhythmic events at device interrogation is also a limitation. The relatively small number of events in some multivariable models render these liable to overfitting, and further validation is desirable. The strength of the association between the DCM +MWHE group and the various endpoints is, however, unlikely to be affected by further validation. In contrast to other studies in patients with DCM (16) or coronary heart disease (33), we have not found a graded relationship between scar burden and mortality in patients with DCM. This, however, is likely to be due to statistical underpowering.

We conclude that midwall fibrosis, detected by MWHE on LGE-CMR, is a powerful predictor of mortality and morbidity in patients with DCM undergoing CRT, independent of QRS duration, NYHA class, LVEF, atrial fibrillation, LV volumes, and mechanical dyssynchrony. Pump failure as well as sudden cardiac death and arrhythmic events mediate this association. These findings provide further evidence for CMR in the prognostic stratification of patients undergoing CRT.