Outcome in Dilated Cardiomyopathy Related to the Extent, Location, and Pattern of Late Gadolinium Enhancement

Objectives This study sought to investigate the association between the extent, location, and pattern of late gadolinium enhancement (LGE) and outcome in a large dilated cardiomyopathy (DCM) cohort. Background The relationship between LGE and prognosis in DCM is incompletely understood. Methods The authors examined the association between LGE and all-cause mortality and a sudden cardiac death (SCD) composite based on the extent, location, and pattern of LGE in DCM. Results Of 874 patients (588 men, median age 52 years) followed for a median of 4.9 years, 300 (34.3%) had nonischemic LGE. Estimated adjusted hazard ratios for patients with an LGE extent of 0 to 2.55%, 2.55% to 5.10%, and >5.10%, respectively, were 1.59 (95% confidence interval [CI]: 0.99 to 2.55), 1.56 (95% CI: 0.96 to 2.54), and 2.31 (95% CI: 1.50 to 3.55) for all-cause mortality, and 2.79 (95% CI: 1.42 to 5.49), 3.86 (95% CI: 2.09 to 7.13), and 4.87 (95% CI: 2.78 to 8.53) for the SCD endpoint. There was a marked nonlinear relationship between LGE extent and outcome such that even small amounts of LGE predicted a substantial increase in risk. The presence of septal LGE was associated with increased mortality, but SCD was most associated with the combined presence of septal and free-wall LGE. Predictive models using LGE presence and location were superior to models based on LGE extent or pattern. Conclusions In DCM, the presence of septal LGE is associated with a large increase in the risk of death and SCD events, even when the extent is small. SCD risk is greatest with concomitant septal and free-wall LGE. The incremental value of LGE extent beyond small amounts and LGE pattern is limited.

D espite advances in therapy, outcomes in dilated cardiomyopathy (DCM) remain poor (1). DCM is a heterogeneous disease affecting a diverse group of patients and response to therapy is varied (2). Precise phenotyping, enabling targeted and personalized management to improve outcomes and avoid unnecessary interventions remains a therapeutic goal (3).
Late gadolinium enhancement (LGE)cardiovascular magnetic resonance (CMR) detects nonischemic LGE in approximately 30% of patients, which correlates with replacement fibrosis on histology (1,4). LGE provides incremental value in addition to left ventricular ejection fraction (LVEF) for predicting all-cause mortality and sudden cardiac death (SCD) events; therefore, it has the potential to guide therapy such as during the selection of patients for implantable cardioverter-defibrillators (ICDs) (1,4). Nonischemic LGE most often occurs in a linear pattern in the mid-wall of the septum; however, subepicardial patterns and LGE occurring in the free-wall of the left ventricle (LV) are also recognized. The nature of the dose-response relationship between LGE and outcome is poorly understood. Data examining the association between the location and pattern of LGE and specific clinical outcomes are also lacking. Identifying an amount, location, or pattern of LGE that provides the optimal mode of risk stratification will help guide the use of this technique in clinical practice.

METHODS
Consecutive patients with DCM referred to our unit between 2000 and 2011 were screened for a registry.
All participants provided written informed consent and the study was approved by the National Research Ethics Service. The diagnosis of DCM was confirmed using the World Health Organization/International Society and Federation of Cardiology definition, based on reduced LVEF and elevated LV end-diastolic volume indexed to body surface area (BSA) (LVEDVi), compared to published age-and sex-specific reference values (5). Exclusion criteria ( Figure 1) included ischemic heart disease, defined as a stenosis of >50% in a major coronary artery or evidence of inducible ischemia on functional testing; evidence of acute myocarditis, or ongoing inflammatory myocardial disease; hypertrophic cardiomyopathy; arrhythmogenic right ventricular cardiomyopathy; significant valve disease; and infiltrative disease. In keeping with guidelines (6,7), an ischemic etiology was considered in all patients and ruled out as follows: All those with infarct patterns of LGE were excluded.
All of the remaining patients (n ¼ 130) were free of angina and considered to have a low risk of ischemic heart disease by their attending cardiologists; the majority (n ¼ 82) were 40 years of age or younger. In the absence of a class 1 indication, coronary angiography was not performed (6,7). None of these patients underwent coronary revascularization or suffered an acute coronary syndrome during follow-up. The final cohort included 682 from previous studies, all of whom underwent extended follow-up for this study (1,4).
CMR was performed on a 1.5-T system (Sonata/ Avanto, Siemens, Erlangen, Germany) using a standardized protocol (4). Late gadolinium imaging was performed 10 min after intravenous injection of 0.1 mmol/kg of gadopentetate dimeglumine or gadobutrol (Bayer AG, Berlin, Germany) using an inversionrecovery gradient echo sequence. Images were acquired in standard long-axis planes and consecutive short axis slices (8-  Late Gadolinium Enhancement and Outcome in DCM A U G U S T 2 0 1 9 : 1 6 4 5 -5 5 CMR data. The primary outcome of interest was allcause mortality. The cause of death was confirmed from a combination of medical records, death certification, and postmortem results using American College of Cardiology/American Heart Association guidance (8). The secondary endpoint was an SCD composite including SCD and aborted SCD. SCD was defined as "unexpected death either within 1 h of the onset of cardiac symptoms in the absence of progressive cardiac deterioration; during sleep; or within 24 h of last being seen alive" (9). Aborted SCD was defined as "an appropriate ICD shock for ventricular arrhythmia, successful resuscitation following ven- LGE was present only in the septum in 142 (16.2%) cases, only in the LV free-wall in 42 (4.8%), and in both locations in 116 (13.3%) ( Figure 2). LGE was     Table 3).
The estimated HRs were similar after additionally adjusting for RVEF, NYHA class, LVEDVi, LV mass index, and LAVi as part of a sensitivity analysis (Supplemental Tables 3 and 4, Supplemental Figure 3).
This was superior to those based on extent or pattern of LGE and the LGE cutoff with the largest C-statistic for the prediction of the primary endpoint. Adding the presence of any LGE and the presence of septal LGE to the baseline multivariable models without LGE improved the C-statistic for the prediction of allcause mortality (Supplemental Table 3).
Following adjustment for LVEF, age, and sex, LGE was associated with SCD and aborted SCD (HR: 3.96; 95% CI: 2.41 to 6.52; p < 0.001) (Supplemental Table 5). The estimated HRs were similar following adjustment for additional covariates as part of a sensitivity analysis (Supplemental Tables 5 and 6, Supplemental Figure 6).   Overall, the model with the smallest AIC that best predicted the SCD endpoint was based on the presence and location of LGE within the septum, the freewall, or in both locations ( Table 3). This was superior to models based on extent and pattern of LGE. Adding the presence of any LGE and the presence of LGE by location to the baseline multivariable models without

Extent
LGE improved the C-statistic for the prediction of the SCD composite (Supplemental Table 5).

DISCUSSION
This is the largest study to date to examine the association between the extent, location, and pattern of LGE and outcome in a large, well-phenotyped DCM cohort. We show the superiority of models based on the presence and location of LGE for the prediction of all-cause mortality and SCD events over those based on LGE extent and pattern ( Figure 5). Our data also establish a nonlinear association between LGE extent and outcome, with a large increase in risk with small degrees of LGE and less marked increases with greater extents thereafter ( Figure 5). The increase in risk with small amounts of LGE was most marked for SCD events (Figure 3).

Previous studies have shown that nonischemic LGE
is associated with an increased risk of death and arrhythmic events (1,11). It has been proposed that LGE-CMR may be able to improve the selection of patients who benefit from ICD implantation (12).
However, up until now there has been a paucity of data examining the relationship between LGE extent, location, pattern, and specific outcomes.
Our data suggest that measures based on LGE location are better than those based on extent for risk prediction. We show that patients with septal LGE were at highest risk of death whereas those with freewall LGE were at similar risk to those without LGE.

Accordingly, a model based on the presence of septal
LGE best predicted all-cause mortality. Whereas septal LGE was also associated with increased SCD events, the greatest risk was seen with concomitant septal and free-wall LGE. A model accounting for the greater risk associated with concomitant LGE in the septum and free-wall was most effective for SCD.
Additionally, sub-epicardial or multiple patterns of LGE were associated with a high-risk of SCD events.
These data add important new information on how to best to use LGE-CMR in risk stratification, an area of unmet need (12,13).
Similar to our results, septal LGE has been associated with worse prognosis in myocarditis (14). The variation in risk based on location may be explained by differences in etiological substrate, scar microstructure, and geographical effects. Idiopathic DCM is most commonly associated with septal mid-wall LGE whereas a previous episode of myocarditis, the cause of a third of DCM, is often associated with free-wall LGE (%)

Late Gadolinium Enhancement and Outcome in DCM
A U G U S T 2 0 1 9 : 1 6 4 5 -5 5 LGE (15). Different insults may create fibrosis with different microstructures and varying levels of risk.
Septal LGE also has greater interaction with the right ventricle and the conduction system. We also show a nonlinear relationship between LGE extent and outcome, such that small degrees of fibrosis are associated with a large increase in risk, particularly with regards to SCD events. This may be explained by the multifactorial disease process.
Replacement fibrosis is 1 of several processes contributing to ventricular arrhythmogenesis (3). It is likely that the synergistic presence of multiple features leads to ventricular arrhythmia rather than 1 factor in a linear dose-dependent manner. In addition, it appears that risk is influenced by fibrosis microstructure and heterogeneity, not simply mass.
Areas of scar with the greatest heterogeneity will cause the largest variation in conduction velocities and the greatest chance of creating re-entrant arrhythmia. Computational scar modeling offers the potential to provide important insights (19).
Localized LGE at the ventricular insertion areas is common, even in healthy volunteers. What this represents and its significance is uncertain. Examining this was beyond the scope of this study; therefore, localized LGE at the ventricular insertion areas was not included. Quantifying the "gray-zone" surrounding an area of replacement fibrosis was proposed in the context of myocardial infarction (20).
There is a lack of histologic correlation examining this     Values are n or n (%) unless otherwise indicated. p values are quoted for each model overall and for the individual components. *The model with the smallest Akaike information criterion and the most optimal for prediction of all-cause mortality.
AIC ¼ Akaike information criterion; C statistic ¼ Harrell's C-statistic; CI ¼ confidence intervals; HR ¼ hazard ratio; Pts ¼ number of patients in each sub-group; other abbreviations as in Table 1. Values are n or n (%) unless otherwise indicated. p values are quoted for each model overall and for the individual components. *The model with the smallest AIC and the most optimal for prediction of SCD.

FIGURE 5 Late Gadolinium Enhancement and Outcome in DCM
LGE extent Our study of dilated cardiomyopathy patients shows a nonlinear relationship between late gadolinium enhancement (LGE) extent and all-cause mortality and sudden cardiac death (SCD) events with a large increase in risk with small degrees of LGE. We show the superiority of models based on the location of LGE for the prediction of these end-points. DCM ¼ dilated cardiomyopathy; other abbreviations as in Figure 3. smaller number of patients in sub-groups such as those with focal or sub-epicardial LGE does, however, limit the interpretation of this specific data.
We recognize that not all arrhythmias resulting in appropriate shocks may have resulted in SCD if untreated. However, we have selected the most robust definition available, excluding antitachycardia pacing (8). We acknowledge that the use of different contrast agents has the potential to impact LGE quantification.
However, there was no difference in the quantity, pattern, or location of LGE for patients scanned with gadobutrol compared to gadopentetate dimeglumine.
In addition, the associations between LGE and outcome remain similar when patients are divided based on contrast agent and there was no difference in the estimated effect of LGE on outcome between groups (Supplemental Table 7). The impact of the use of different contrast agents on the results of the study, therefore, appears to be minimal.
Parametric mapping was not available at the outset of the current study and was therefore not included in the analysis. This technique has the advantage of identifying diffuse myocardial changes which LGE imaging may not detect. Previous work has shown associations between native T1 values and mortality and heart failure outcomes in DCM (21). Given the possible role of diffuse fibrosis in arrhythmia generation and heart failure, parametric mapping offers hope in the identification of those at risk of adverse outcomes. We eagerly await further data examining the incremental value of parametric mapping. Our data suggest the need to examine the incremental value of this technique in addition to the presence of septal LGE.

CONCLUSIONS
We show a large increase in all-cause mortality and