Prognostic implications of late gadolinium enhancement at the right ventricular insertion point in patients with non-ischemic dilated cardiomyopathy: A multicenter retrospective cohort study

Introduction The presence of late gadolinium enhancement (LGE) at the right ventricular insertion point (RVIP) on cardiac magnetic resonance (CMR) is generally believed to be nonspecific, but the clinical implication of this unique LGE pattern in patients with non-ischemic dilated cardiomyopathy (NICM) has not been elucidated. Objectives We investigated the prognostic significance of RVIP-LGE in NICM patients. Methods A total of 360 consecutive NICM patients referred for CMR (102 with no LGE, 50 with RVIP-LGE, 121 with left ventricular [LV]-LGE, and 87 with both an LV and RVIP-LGE) were studied. The primary endpoint was a composite of the all-cause death, hospitalization due to worsening of heart failure, and major arrhythmic events. Results During a mean follow-up of 45.2 ± 36.5 months, 149 (41.4%) patients (22 [21.6%] no LGE vs. 16 [32.0%] RVIP-LGE vs. 62 [51.2%] LV-LGE vs. 49 [56.3%] both LV and RVIP-LGE, P < 0.0001) reached the primary endpoint. A Kaplan Meier curve demonstrated that RVIP-LGE patients had an intermediate trend of an event free survival rate for the composite endpoint (log-rank P < 0.0001). In a multivariable Cox regression model, LV-LGE (P = 0.008) and both LV and RVIP-LGE (P = 0.003) were significantly associated with a worse outcome, whereas RVIP-LGE was not (P = 0.101). In addition, RVIP-LGE patients (n = 32) had a more favorable outcome compared to LV-LGE patients (n = 32) even after matching the extent of the LGE (median 3.4% [interquartile range, 3.1–3.8], 8 [25.0%] RVIP-LGE vs. 20 [62.5%] LV-LGE, P = 0.002). Conclusions LGE confined to the RVIP among NICM patients did not significantly increase the risk of adverse cardiac events, and also showed a better outcome than the same extent of LGE located in the LV. Identification of this unique LGE distribution may help refine the current risk stratification.


Introduction
The presence of late gadolinium enhancement (LGE) at the right ventricular insertion point (RVIP) on cardiac magnetic resonance (CMR) is generally believed to be nonspecific, but the clinical implication of this unique LGE pattern in patients with non-ischemic dilated cardiomyopathy (NICM) has not been elucidated.

Objectives
We investigated the prognostic significance of RVIP-LGE in NICM patients.

Methods
A total of 360 consecutive NICM patients referred for CMR (102 with no LGE, 50 with RVIP-LGE, 121 with left ventricular [LV]-LGE, and 87 with both an LV and RVIP-LGE) were studied. The primary endpoint was a composite of the all-cause death, hospitalization due to worsening of heart failure, and major arrhythmic events.

Results
During a mean follow-up of 45.2 ± 36. 5

Introduction
Non-ischemic dilated cardiomyopathy (NICM) is a primary myocardial disease occurring in approximately one-third of heart failure (HF) patients and is associated with significant morbidity and mortality [1]. Cardiac magnetic resonance (CMR) imaging with late gadolinium enhancement (LGE) has emerged as a promising, noninvasive tool for the detection and quantification of myocardial fibrosis that is frequently found in patients with NICM [2]. Recently, CMR studies in NICM patients have demonstrated that the presence of LGE strongly predicts poor clinical outcomes such as hospitalization for HF, fatal ventricular arrhythmias, and death [3,4].
LGE confined to the anterior and/or posterior right ventricular insertion points (RVIPs) (RVIP-LGE) are commonly seen in patients with hypertrophic cardiomyopathy (HCM), which is associated with a relatively better prognosis [5][6][7]. Previously, this focal LGE region has shown myocardial disarray or interstitial fibrosis rather than replacement fibrosis [7,8]. However, LGE at the RVIP is also found in other medical conditions and its clinical impact still remains uncertain [9][10][11][12]. To the best of our knowledge, there is a paucity of data concerning the role of LGE at the RVIP in the risk stratification of NICM for adverse cardiac events. Hence, the aims of this study were first to investigate the characteristics and prognostic significance of LGE confined to the RVIP, and second to identify the determinants of this unique LGE distribution in patients with NICM.

Study population
Between May 2003 and February 2016, 430 consecutive patients with newly diagnosed NICM referred for CMR at two tertiary hospitals in South Korea were retrospectively identified. All patients underwent CMR just after the diagnosis of NICM that was confirmed by clinical assessment, echocardiography and coronary angiography. NICM was defined according to the World Health Organization/International Society and Federation of Cardiology criteria [13]. Patients with symptoms or signs of heart failure, a reduced left ventricular ejection fraction (LVEF < 50%) without regional wall motion abnormalities, an increased LV end-diastolic diameter (LVEDD > 55mm), no prior history of a myocardial infarction or revascularization, by the Ministry of Science, ICT & Future Planning (NRF-2015R1C1A1A02037085 to BJ), and a grant from the Korean Healthcare Technology R&D Project funded by the Ministry of Health & Welfare (H16C0058, H15C1200 to BJ). The funders had no role in study designm data collection and analysis, decision to publish, or preparation of the manuscript. and absence of significant coronary artery disease on coronary angiography (obstruction > 50% of 2 or more epicardial vessels or the left main or proximal left anterior descending coronary artery) [14] were included in this study. We excluded patients with acute myocarditis, severe valvular heart disease, hypertrophic cardiomyopathy, infiltrative cardiomyopathy or other specific cardiomyopathies and congenital heart disease. Forty patients who had contraindications to CMR and gadolinium-based contrast agents or that withdrew their consent for participation in the study were excluded.
The study protocol conformed to the principles outlined in the Declaration of Helsinki and was approved by the local ethics committee (EUMC 2017-09-019-004). All data were fully anonymized before access and the requirement for informed consent was waived due to the retrospective nature of this study. The names of ethics committees are as follows: Ewha Womans Mokdong Hospital and Severance Cardiovascular Hospital.

Electrocardiogram and echocardiography
On the day of the CMR study, standard resting 12-lead electrocardiograms (ECGs) were obtained at a paper speed of 25 mm/s with a calibration of 10 mm/mV (GE Healthcare, Marquette, MAC 5500, Waukesha, WI). All ECGs were analyzed by 1 experienced reader who was blinded to the clinical data of the patient. We documented the rhythm and measured the heart rate (ventricular rate), PR interval, and QRS duration according to the established guidelines. The corrected QT (QTc) interval was automatically calculated using Bazett's formula (QTc Baz = QT= ffi ffi ffi ffi ffi ffiffi RR p ) and a prolonged QTc was defined as a QTc interval of >440ms [15]. Transthoracic Doppler echocardiography was performed within one week of the CMR and all measurements were obtained as by the recommended standard guidelines [16].

CMR acquisition and analysis
All patients underwent CMR studies on a 1.5 Tesla whole-body scanners (InteraAchieva, Phillips Medical System, Best, The Netherlands or Philips Healthcare, Andover, MA, USA) with a phase array cardiac coli and the same CMR protocol was adopted in both centers. Electrocardiogram (ECG)-gated cine images were acquired in the horizontal, vertical, and short-axis views using a breath-hold cine steady-state, free precession sequence (field of view, 360 ⅹ360 mm; matrix, 256 ⅹ 256 mm; slice thickness, 8 mm; TR [repetition time]/TE [echo time], 3.4/ 1.7 ms; flip angle, 50˚; 25 frames for the cardiac cycle). A stack of contiguous short-axis slices from the base to apex were obtained without an interslice gap. Late enhancement images were acquired 10 minutes after infusing intravenous gadolinium-DTPA (0.2mmol/kg; gadoteratedimeglumine; Dotarem, Geurbet) using a T1-weighted 2-dimensional gradient echo inversionrecovery sequence (field of view, 360 ⅹ360 mm; matrix, 521 ⅹ 512 mm; slice thickness, 8 mm; TR/TE, 5.3/1.6 ms; flip angle, 15˚; no interslice gap) in the same views used for the cine images. An ECG-synchronized image acquisition in the mid-diastole phase was performed in order to minimize any cardiac motion induced artifact. The inversion time (T1) was individually adapted to nullfy the signal of normal myocardium using a dedicated T1 determining sequence [17].
All CMR images were analyzed off-line using a dedicated software program (CMR42, Circle Cardiovascular Imaging, Calgary, Alberta, Canada) by the consensus of 2 experienced radiologists blinded to the patient clinical data and outcome. The left and right ventricular volumes and ejection fraction (EF) were measured from the cine images using a semi-automatic segmentation in the software and all volume measurements were normalized to the body surface area. The LGE was visually assessed for each segment using a modified American Heart Association (AHA) 16-segment left ventricular (LV) model [18] and considered to be present if there was a region of discernible high signal intensity upon visual inspection. The presence of LGE at the RVIP was defined as a focal, hyperintensity area confined to the junction of the RV wall into the anterior and/or posterior interventricular septum in at least two consecutive short-axis images. Factors affecting the image quality were checked, and the potential pitfalls and artifacts (e.g. blood pool, epi-and pericardial fat, partial volume effects, vessels, or ghosting artifacts) mimicking myocardial scar in the delayed enhancement CMR imaging were also excluded [19]. If LGE was detected, patterns were classified either as subendocardial, midwall, subepicardial, transmural or patchy [20]. The myocardial volume was obtained using a semiautomatic segmentation of the endocardial and epicardial borders in each short axis image. The LGE volume was quantified from a short-axis stack of images using a full width with a half maximum (FWHM) technique and the region of interest was drawn in the area of the maximum signal intensity of a visible LGE for the FWHM threshold. Both the myocardial and LGE volumes were estimated from the sum of each area for each slice, multiplied by the slice thickness. The extent of the LGE was expressed as a percentage of the LGE, which was derived by dividing the LGE volume by the myocardial volume, and the quotient multiplied by 100. The presence and extent of the LGE were evaluated twice by 2 independent expert readers who were blinded to all the patient details. The inter-observer agreement between the two readers in regard to the presence or absence of LGE was substantial (kappa value = 0.827, P < 0.005). The intra-observer variability of the LGE quantification for the coefficient of variation (CV) and intra-class correlation coefficient (ICC) were 12.5% and 0.99 (95% confidence interval [CI] 0.97-0.99), respectively. Also, there was good inter-observer agreement for RVIP-LGE (kappa value = 0.785, P < 0.005), and the intra-observer variability of RVIP-LGE quantification for the CV and ICC was 12.1% and 0.97 (95% CI 0.96-0.98), respectively.

Clinical follow-up and outcome events
Patients underwent regular follow-up every 3 to 6 months via clinic visits or telephone interviews. The primary endpoint was a composite of the all-cause death and hospitalization due to worsening of heart failure (CHF) and major arrhythmic events. Hospitalization for HF was defined as the first readmission with worsening signs or symptoms of HF requiring treatment with i.v. therapies (i.e. diuretics, vasodilators, and inotropes), mechanical or surgical interventions, or initiation of ultrafiltration, hemofiltration or dialysis. Major arrhythmic events included sustained ventricular tachycardia (VT), ventricular fibrillation (VF), appropriate implantable cardioverter defibrillator (ICD) discharges, and sudden cardiac death (SCD). Sustained VT was defined as three or more consecutive ventricular extrasystoles of �100 beats/ min lasting for �30 sec on the ECG or 24-h Holter monitoring or that requiring intervention for termination. VF was determined by QRS complexes with a markedly different morphology, axis, and amplitude, no identifiable P waves, and irregular ventricular rates of �300 beats/min. Appropriate ICD discharges were defined as a device shock or anti-tachycardia overdrive pacing for sustained VT based upon a programmed rate cutoff of the ICD (>180 beats/min) or VF. The cause and date of the death were identified using the information from the National Population Registry of Korea National Statistical Office as well as the medical records at the time of the death. SCD was defined as an unexpected death with or without documented ventricular arrhythmias occurring within 1h of the cardiac symptoms, or nocturnal death with no antecedent history of progressive cardiac deterioration.

Statistical analysis
All statistical analyses were conducted using SPSS version 18.0 software (SPSS Inc., Chicago, Illinois, USA). Statistical significance was accepted for two-tailed values of P < 0.05.
Continuous variables were expressed as the mean ± SD or median with a corresponding interquartile range (IQR) and categorical variables as the n (%). Comparisons of variables across groups were performed using a one-way ANOVA combined with a Bonferroni post hoc analysis or Kruskal-Wallis H test and Chi-square (χ 2 ) or Fisher's exact test, as appropriate. Hazard ratios of the primary endpoint and its components were calculated with univariate Cox proportional hazard models with a computed 95% confidence interval (CI). The Kaplan-Meier method with log-rank tests was used to estimate the event free survival curves for each category of NICM. To control for any potential confounders, multivariate Cox proportional hazard models were built using covariates identified as significant in the univariate analysis or known to affect the primary endpoint. A post-hoc subgroup analysis was performed for patients with an LVEDVI less than the median of 160ml/m 2 . A multivariate binary logistic regression analysis was performed to assess the determinants of an LGE confined to the RVIP.

Patient characteristics
A total of 360 patients (36.9% female, mean age 54 ± 15 years) were enrolled and categorized into 4 groups according to the presence or absence and location of the LGE in the heart: no LGE (n = 102, 28.3%), RVIP-LGE (n = 50, 13.9%), LV-LGE (n = 121, 33.6%), and both LV and RVIP-LGE (n = 87, 24.2%) (Fig 1). The group baseline characteristics are summarized in Table 1. RVIP-LGE patients were younger and had higher systolic and diastolic blood pressure (BP) levels, but the gender distribution, BMI, and co-morbidity profiles did not significantly differ among the groups. Loop diuretics were more frequently used in the RVIP-LGE patients than those without LGE. Left bundle branch block (LBBB) was more common in the no LGE and RVIP-LGE patients than in the LV-LGE groups. RVIP-LGE patients had a more prolonged QTc interval, higher NT-proBNP level than the other groups, and greater E/E' ratio than the no LGE group.
The mean LVEF in the total study population was 25 ± 9%. RVIP-LGE patients had a similar LVEF as the LV-LGE patients, but had lower LV end-diastolic and end-systolic volume indices compared to other groups. Although the RVIP-LGE patients tended to have a higher RVEF, there were no statistically significant differences in the RV end-diastolic and end-systolic volume indices between the groups. LGE at only the posterior RVIP was more common in the RVIP-LGE patients, whereas LGE at both RVIPs was more frequently observed in the patients with LGE in both the LV and RVIP. LGE at only the anterior RVIP was not found in either group. Patchy pattern of LGE was predominantly found more often in the RVIP-LGE patients than other groups and the extent of the LGE was greatest in the patients with LGE at both the LV and RVIP.

Clinical outcome
During a mean follow-up duration of 45.2 ± 36.5 months, 149 (41.4%) patients reached the primary endpoint. Thereof 63 (17.5%) patients died, 110 (30.7%) were hospitalized due to worsening of HF and 49 (13.6%) suffered major arrhythmic events. Overall, the composite endpoint most commonly occurred in patients with an LGE in both the LV and RVIP, followed by the LV-LGE patients, RVIP-LGE patients, and lowest in the patients without LGE. Similar results were observed for the all-cause death, but the incidence of worsening HF did not differ between the LV-LGE patients with and without RVIP-LGE. Major arrhythmic events least frequently occurred in the RVIP-LGE patients and none of them underwent implantable cardioverter defibrillation (ICD) implantations for primary or secondary prevention (Fig 2). The Kaplan-Meier survival curves significantly differed among the groups. For the major arrhythmic events, the RVIP-LGE patients had a higher event free survival rate than the LV-LGE patients with or without RVIP-LGE, which was similar to that of the patients without LGE (Fig 3D, log-rank P < 0.0001).
A multivariate Cox regression analysis demonstrated that the presence of LGE at only the RVIP did not significantly increase the risk of adverse outcomes compared with the absence of an LGE (Table 2). Additionally, we compared the clinical outcomes between the patients with LGE isolated to the RVIP (n = 32) and those with LV-LGE only (n = 32) after matching the extent of the LGE to examine if the favorable outcome of the RVIP-LGE was related to the   (Table 3), RVIP-LGE patients had a significantly lower incidence of the composite endpoint and worsening HF than the LV-LGE patients (Table 4). Kaplan-Meier curves for the composite endpoint and worsening HF also demonstrated significant differences between them (Fig 4).

Determinants of an LGE confined to the RVIP
In the multivariate logistic regression analysis, the age, a female gender, diastolic BP, presence of LBBB, and a prolonged QTc interval (>440msec) were independently associated with LGE at only the RVIP (Table 5).

Discussion
The main findings of this study among the patients with NICM were as follows. First, the presence of LGE at only the RVIP did not significantly increase the risk of adverse cardiac events. Second, a young age, female gender, elevated diastolic BP, presence of LBBB, and prolonged QTc interval were independent predictors of LGE confined to the RVIP. To the best of our
Although the presence of LGE has been described as a poor prognostic indicator in patients with NICM [3,4], only a few studies have mentioned the clinical significance of LGE at the RVIP [21,22]. Neilan T et al. found in a study of 162 NICM patients that a focal LGE at the RVIP was not associated with an increased risk of adverse outcomes [21]. On the contrary, data from Gaztanaga J et al. showed a worse outcome in patients with RVIP-LGE than in those with various patterns of LV-LGE [22]. However, the aforementioned studies did not focus particularly on LGE confined to the RVIP and accordingly, had a relatively small sample size and limited number of events in the patients with RVIP-LGE. In the present study, we demonstrated a more favorable outcome in patients with LGE at only the RVIP than in those with LV-LGE, but at the same time, it was less favorable than that in patients without LGE.
Myocardial fibrosis is one of the most common histopathologic findings of the failing heart, which is characterized by a progressive accumulation of collagen components leading to an increased myocardial stiffness, cardiac remodeling, and worsening ventricular systolic function [23]. CMR imaging with LGE (CMR-LGE) allows a sensitive and reproducible assessment of replacement myocardial fibrosis, but does not detect interstitial fibrosis, which is another subtype of myocardial fibrosis in dilated cardiomyopathy (DCM) [24]. In our study, despite a small amount of myocardial scar, the RVIP-LGE group had an elevated LV filling pressure comparable to that of the LV-LGE groups and the use of loop diuretics was more common in the RVIP-LGE patients than in those without LGE. Furthermore, the RVIP-LGE group had a significantly longer QTc interval than the other groups, and a prolonged QTc interval was an independent predictor of LGE confined to the RVIP. Recently, Malaty AN et al. demonstrated a higher LV filling pressure in DCM patients without LGE than in those with LGE, postulating a causal role of interstitial fibrosis in increased LV stiffness [25]. The QTc interval has been also reported as an electrocardiographic parameter associated with LV diastolic dysfunction [26]. Indeed, in a study of 241 HF patients, a prolonged QTc interval was a strong predictor of death due to worsening HF [15]. Considering that the LV diastolic, rather than systolic function is closely related to exercise tolerance [27], LGE confined to the RVIP could be an important surrogate marker reflecting increased LV stiffness beyond the extent of the LGE. Interestingly, in a post-hoc subgroup analysis of patients with a less dilated LV cavity (LVEDVI � the median value of 160ml/m 2 ), we found that the RVIP-LGE patients had a greater risk of experiencing the composite endpoint than the patients without LGE (adjusted HR, 4.20; 95% CI, 1.21-14.52, P = 0.024), and had a similar risk as the LV-LGE patients with and without RVIP-LGE (S1 Table). The Kaplan-Meier analysis also indicated that the RVIP-LGE patients were at a significantly increased risk for adverse outcomes as compared to the patients without LGE during the follow-up (S1 Fig, log-rank P = 0.003). These results suggested that the identification of this unique LGE distribution may be useful in the prognostic stratification from the early stages of NICM [28]. On the other hand, in patients with LV myocardial scar on the CMR-LGE, the additional clinical impact of an RVIP-LGE was not significant (Fig 3A, P = 0.571). Although the myocardial scar observed by CMR-LGE is thought to be a potential substrate for re-entrant circuits [29], we found no significant difference in the occurrence of major arrhythmic events between the patients with RVIP-LGE and those without LGE. Previous studies documented that the characteristics of the LGE, including the pattern and extent, are Table 3

RVIP-LGE (n = 32) LV-LGE (n = 32) P Value
Age   [3,4]. In a recently published study, the extent of the myocardial scar was an independent predictor of arrhythmic events [21]. However, in our study, most patients with RVIP-LGE demonstrated a patchy pattern of the LGE, and a midwall pattern was observed in only 2 (4.0%) patients. Moreover, the extent of the LGE was significantly lower in the patients with RVIP-LGE than those with LV-LGE. These CMR characteristics could explain why patients with LGE confined to the RVIP did not have a significant arrhythmic risk compared to the patients without LGE in this study.
In the present study, we found an independent association between RVIP-LGE and LBBB, which was a consistent finding of the previous reports that described the paradoxical motion of the interventricular septum as a major determinant of LGE at the VIP [9]. Also, LGE at the RVIP was related to a young age, female gender, elevated diastolic BP, and prolonged QTc interval. However, our data showed no significant relationship between RVIP-LGE and SPAP, despite the involvement of the RVIP has been reported as a typical pattern of LGE in pulmonary hypertension. This seems to be attributed to our population characteristics with a relatively lower proportion and lesser degree of pulmonary hypertension (no [60.7%], mild [28.0%], moderate [11.3%], severe [0%]), compared with previously published studies of patients with pulmonary hypertension [10].

Limitations
Our study had several inherent limitations given its retrospective design with a potential selection and information bias. However, we acquired and analyzed the complete data on the clinical outcomes of all patients. Second, although the hemodynamic data was not obtained through cardiac catheterization, we used Doppler-derived measurements that have been extensively validated to obtain a good correlation with the invasive hemodynamic indexes in the patients with DCM [30]. Third, in a subgroup analysis, we arbitrarily used the median value of the LVEDVI (� 160ml/m 2 or >160ml/m 2 ) as a cutoff point to stratify the degree of LV remodeling. Although it is well known that an increase in the LV volume reflects an increased LV wall stress and LV mass, leading to myocardial remodeling [28], there is no consensus on the optimal threshold of the LVEDV for quantification of LV remodeling in patients with NICM. Fourth, in this study, some patients were found to have an atypical LGE pattern for NICM (13 subendocardial and 29 transmural types), and these findings could suggest a previous history of spontaneous recanalization in myocardial infarction. However, we excluded the patients with evidence of significant coronary artery disease on angiogram before the CMR imaging. In addition, the incidence of an atypical LGE pattern in our study was similar to that in the previously published studies that have shown various LGE patterns in NICM patients with different etiologies [31][32][33]. Fifth, in the present study, LGE volume was

Conclusions
First, the presence of LGE confined to the RVIP among NICM patients did not significantly increase the risk of adverse cardiac events.
LGE limited to the RVIP was an indicator of a small LGE extent, with limited prognostic significance only in a selected subpopulation, particularly in patients with a less dilated LV. Furthermore, RVIP-LGE had a more favorable outcome than LV-LGE even with the same extent of the LGE, which strongly suggested the benign nature of LGE located in the RVIP. Second, a young age, female gender, elevated diastolic BP, presence of LBBB, and prolonged QTc interval were independent predictors of RVIP-LGE. Those variables may be associated with a small LGE extent, as well as the RVIP location. However, considering that the clinical impact of the RVIP-LGE could not be determined solely by the location or extent of the LGE, these variables may also sufficiently predict the RVIP-LGE. This unique LGE distribution may have an important prognostic implication in refining the current risk stratification for adverse cardiac events in NICM patients.