Article Text

Download PDFPDF

Original article
Scar size and characteristics assessed by CMR predict ventricular arrhythmias in ischaemic cardiomyopathy: comparison of previously validated models
  1. Stefan de Haan,
  2. Thomas A Meijers,
  3. Paul Knaapen,
  4. Aernout M Beek,
  5. Albert C van Rossum,
  6. Cornelis P Allaart
  1. Department of Cardiology, Institute for Cardiovascular Research (ICaR-VU), VU University Medical Center, Amsterdam, The Netherlands
  1. Correspondence to Stefan de Haan, VU University Medical Center, Department of Cardiology, De Boelelaan 1118, Amsterdam 1081 HV, The Netherlands; s.dehaan{at}vumc.nl

Abstract

Objective Sudden cardiac death is a major cause of mortality in patients with ischaemic cardiomyopathy. Risk stratification remains challenging. Currently, there is growing interest in scar characteristic assessment as a predictor of sudden cardiac death using cardiac magnetic resonance imaging (CMR). Standard analysis methods are lacking. The present study evaluated previously validated methods of scar assessment by CMR with late gadolinium enhancement (LGE) in their ability to predict ventricular tachyarrhythmias.

Methods Patients with ischaemic cardiomyopathy who received an implantable cardioverter defibrillator for primary prevention and in whom a LGE–CMR was performed, were included. Scar core size, peri-infarct zone and total scar size, which is defined as the sum of the core size and peri-infarct zone, were assessed using three previously validated models, and their ability to predict ventricular tachyarrhythmias was evaluated.

Results Fifty-five patients were included (mean age 64.6±10.8 years, 43 men). During a median follow-up of 2.0 years (IQR 1.0–3.0 years) 26% of patients reached the endpoint of ventricular tachyarrhythmia. All scar characteristics (ie, total scar size, scar core size and peri-infarct zone) of the three methods were predictors of the endpoint (p<0.01). Total scar size was comparable, whereas scar core size and peri-infarct zone varied significantly between the tested models. Receiver operating characteristic curves of the different scar characteristics showed comparable areas under the curve varying from 0.721 to 0.812.

Conclusions LGE–CMR-derived scar tissue characteristics are of predictive value for the occurrence of ventricular tachyarrhythmias in patients with ischaemic cardiomyopathy. Additional estimation of scar core size and/or peri-infarct zone does not appear to increase the diagnostic accuracy over total scar size alone.

  • Cardiac imaging
  • cardiac magnetic resonance imaging
  • cardiomyopathy dilated
  • cardiomyopathy hypertrophic
  • clinical cardiology
  • clinical coronary heart disease
  • coronary artery disease (CAD)
  • EBM
  • heart failure
  • hypertrophic cardiomyopathy
  • implantable cardioverter defibrillator
  • late gadolinium enhancement
  • MRI
  • myocardial ischaemia and infarction (IHD)
  • myocardial perfusion
  • NSTEMI
  • nuclear cardiology
  • pacing
  • stable angina
  • STEMI
  • sudden cardiac death
  • ventricular fibrillation
  • ventricular tachycardia
View Full Text

Statistics from Altmetric.com

Patients with an ischaemic cardiomyopathy are at increased risk of sudden cardiac death, especially when left ventricular ejection fraction (LVEF) is below 30%.1 The increased risk is mainly attributable to the occurrence of ventricular tachyarrhythmias.2 3 After introduction of the implantable cardioverter defibrillator (ICD), mortality has been substantially reduced.4 5 However, post-hoc analysis of the MADIT II study showed that only approximately 35% of ischaemic cardiomyopathy patients with an ICD for primary prevention receive appropriate therapy during the first 3 years of follow-up.6 Considering the risk of perioperative complications and adverse events associated with ICD implantation, as well as the high financial impact of ICD implantation and follow-up, better risk stratification protocols are warranted.7–9

Myocardial scar has been demonstrated to serve as substrate for ventricular arrhythmias.3 Cardiac magnetic resonance imaging (CMR) with late gadolinium enhancement (LGE) has been shown to be able to assess myocardial scar accurately.10 11 Therefore, several studies have evaluated the risk stratification potential of myocardial scar assessment by LGE–CMR. These reports showed that the extent of left ventricular scarring as well as the extent of the peri-infarct zone were independent predictors of ventricular tachyarrhythmia inducibility in ischaemic cardiomyopathy.12 13 More recently, these scar characteristics have been linked to appropriate ICD therapy in this patient population.14 However, uniformity in the analysis of the LGE–CMR parameters is lacking, and the aforementioned studies have applied different criteria.12–14

The current study set out to compare and validate these previously described methodological approaches in their ability to predict ventricular tachyarrhythmias in cardiomyopathy patients with an ICD for primary prevention.

Methods

Study population

Study procedures were in accordance with the Declaration of Helsinki. National and institutional guidelines did not require institutional review board approval because the study was retrospectively performed and patient data were anonymised and solely patients from the VU University Medical Center (Amsterdam, The Netherlands) were included. The study population consisted of 55 patients with ischaemic cardiomyopathy, defined as a LVEF less than 35% and a history of myocardial infarction or percutaneous or surgical revascularisation, who were referred for LGE–CMR on clinical indication and subsequently scheduled for ICD implantation for primary prevention of sudden cardiac death between January 2006 and May 2009.15 The clinical indication for CMR with LGE in most patients was the assessment of myocardial viability and in a smaller proportion of patients for diagnosis of ischaemic cardiomyopathy. A combined device for additional cardiac resynchronisation therapy was implanted when appropriate according to current European Society of Cardiology guidelines.16

CMR image acquisition

CMR studies were performed on a 1.5-Tesla whole-body scanner (Magnetom Sonata/Avanto, Siemens, Erlangen, Germany), using a six-channel phased-array body coil. After survey scans, a retrospectively gated, balanced steady-state free precession gradient-echo sequence was used for cine imaging. Image parameters included slice thickness of 5 mm, slice gap 5 mm, temporal resolution less than 50 ms, repetition time 3.2 ms, echo time 1.54 ms, flip angle 60° and a typical image resolution of 1.3×1.6 mm. The cardiac cycle consisted of 20 phases. After obtaining four, three and two-chamber view cines, stacks of 10–12 short axis slices were acquired to cover the left ventricle fully. Cine images were acquired during breath-hold in mild expiration.

Contrast images were acquired 10–15 min after the administration of 0.2 mmol/kg gadolinium-DTPA in the same views used in the cine images, using a two-dimensional segmented inversion–recovery prepared gradient echo sequence (TE 4.4 ms, TR 9.8 ms, inversion time 250–300 ms, typical voxel size 1.3×1.6×5 mm3).

CMR image analysis

Images were analysed off-line, using the software package MASS and were performed blinded from ICD results. First, short axis cine images were analysed. Epicardial and endocardial borders of the left ventricle were outlined manually in both the end-diastolic and end-systolic phase in all short axis images. End-diastolic volume, end-systolic volume, ejection fraction and end-diastolic mass were computed using these analyses.

Subsequently, LGE images were analysed using MASS to calculate infarct core size and peri-infarct zone according to three previously validated models (table 1 and figure 1).13 14 17 The first method (method 1) by Roes et al,14 which is derived from the full width half max method, defines scar core as myocardium with signal intensity of 50% or greater of the maximum signal intensity of the hyperenhanced area. The peri-infarct zone is defined as myocardium with signal intensity between 35% and 50% of maximum signal intensity. The definition of the scar core in the second method (method 2), by Schmidt and colleagues,13 is identical to the first method. The peri-infarct zone was defined as myocardium with signal intensity higher than peak signal intensity of a remote reference area, and lower than the 50% threshold of the scar core zone. The third method (method 3) is the scar analysis based on the method based on Yan et al,17 which calculates the average signal intensity and SD of a remote non-enhanced myocardial segment. Scar core is subsequently defined as myocardium with signal intensity higher than 3 SD above the mean signal intensity of the normal remote myocardium, whereas the peri-infarct zone was defined as signal intensity between 2 and 3 SD greater than remote. In all three methods total scar was defined as the sum of scar core and peri-infarct zone. Scar was expressed as grams of myocardium.

Table 1

LGE analysis techniques

Figure 1

(A) Short axis cardiac magnetic resonance with late gadolinium enhancement (LGE) image from a patient with previous anteroseptal myocardial infarction. Epicardial and endocardial contours are outlined. The extent of LGE according to method 1 (B), method 2 (C) and method 3 (D) as described in table 1. Red indicates scar core and yellow peri-infarct zone. Please note the variation in scar core and peri-infarct zone between the different methods.

ICD devices

Patients in this study received a CRT-D device (Concerto or InSync Sentry, Medtronic Inc Minneapolis, USA; Promote, St Jude Medical St Paul, USA; Kronos or Lumax, Biotronik Berlin, Germany), dual chamber ICD (Virtuoso, Medtronic Inc; Current, St Jude Medical; Lumos, Biotronik) or single chamber ICD (Entrust or Virtuoso, Medtronic Inc; Lumos, Biotronik). Although device settings for therapies were different between patients, typically all ICD and CRT-D devices were programmed to monitor ventricular arrhythmias with a cycle length of 400 ms or less.

Follow-up

Follow-up was performed by device interrogation every 3–6 months and chart review. The median follow-up duration was 2.0 years (IQR 1.0–3.0). The endpoint of the study was the occurrence of a ventricular tachyarrhythmia with a cycle length of 400 ms or less with a duration of more than 30 s, or the delivery of appropriate ICD therapy. All device interrogations were performed blinded from CMR results and all events were reviewed by an experienced cardiologist.

Statistical analysis

Continuous variables are presented as mean±SD, and categorical data are summarised as frequencies and percentages. For comparison of two datasets, unpaired Student's t test or Fisher's exact test were performed when appropriate. Levene's test for equality of variances was used to verify if the application of the unpaired Student's t test was appropriate. Comparison of multiple datasets was performed using analysis of variance, and specific differences were identified by a Student's t test with the Bonferroni inequality adjustment. Receiver operating characteristic (ROC) curves were created for all scar characteristics and areas under the curve (AUC) were calculated. AUC were compared using the method of DeLong et al.18

Intra-observer and interobserver agreement for LGE measurements was calculated using the intraclass correlation coefficient (ICC) for absolute agreement.

All tests were two-sided, and a p value less than 0.05 was considered statistically significant.

Results

Study population

Baseline characteristics of the study population are listed in table 2. Heart failure symptoms were compatible with NYHA class I (n=14), II (n=14), III (n=26), or IV (n=1). Of the baseline parameters, only age was significantly higher in patients who reached the primary endpoint.

Table 2

Clinical baseline characteristics

Follow-up

During follow-up, 14 patients (26%) reached the endpoint of a ventricular tachyarrhythmia. In 13 patients at least one single episode of sustained ventricular tachycardia was documented, one patient experienced ventricular fibrillation. The mortality rate during the study was 15% (n=8) due to progressive heart failure. None of the patients died of ventricular tachyarrhythmia.

CMR scar parameters

Scar characteristics are shown in table 3. Total scar size was comparable for each of the methods, scar core size was significantly increased using method 3. The peri-infarct zone displayed variability between methods with the largest size for method 2 and the smallest for method 3. Total scar size, scar core size and peri-infarct zone size of all three methods were significantly larger in the patient group that reached the endpoint compared with patients who did not reach the endpoint of the study (table 3 and figure 2). The ICC for the different scar measurements for intra-observer and interobserver agreement are shown in table 4. As indicated in figure 2, there was considerable overlap between the various scar characteristics of patients who did and did not reach the endpoint of ventricular arrhythmia, regardless of the methodology used. Nonetheless, a lower threshold could be identified in each of the methodological approaches beyond which no ventricular arrhythmias could be detected. The proportions of patients below these thresholds varied from 18% of patients for total scar size assessed by method 2 to 35% of patients for peri-infarct zone assessed by method 3.

Table 3

MRI scar variables

Figure 2

Scatter plots of the amount of scar according to method and scar characteristic for patients with (ventricular tachyarrhythmias; VT) and without ventricular arrhythmias (no VT). The mean is indicated by a horizontal line.

Table 4

Intra and interobserver variability

Predictors of ventricular tachyarrhythmias

ROC curves of the different scar characteristics all show comparable AUC (figure 3). AUC ranged from 0.72 for the scar of methods 1 and 2 to 0.81 for the peri-infarct zone of method 2. Comparison of the AUC of the different methods showed no significant differences. LVEF had no predictive value for ventricular tachyarrhythmias in the current study population.

Figure 3

Receiver operating characteristic curves of the prediction of ventricular tachyarrhythmias of the different scar characteristics by method. AUC, area under the curve.

Discussion

The current study was conducted to compare three previously described methods for the analysis of CMR with LGE and evaluate their predictive values for ventricular tachyarrhythmias. There was no significant difference in the extent of total scar between methods; however, variability in scar core and peri-infarct zone could be detected. All scar characteristics were shown to be significantly higher in patients in whom ventricular tachyarrhythmias occurred after ICD implantation. Nonetheless, ROC curves showed comparable AUC for all scar characteristics, deeming each parameter of comparable predictive value. Therefore, of the tested methodology no scar characteristic could be labelled superior in the prediction of ventricular arrhythmias in patients with ischaemic cardiomyopathy and ICD implantation for primary prevention.

The majority of patients with cardiomyopathy at risk of sudden death are of ischaemic aetiology.3 In these patients, scar tissue serves as an important substrate for ventricular tachyarrhythmias, based on the re-entry phenomenon.19 Therefore, assessment of scar extent might serve as a useful tool for risk stratification. CMR with LGE is an excellent technique to assess scar size accurately in humans in vivo.10 11 Bello et al12 initially demonstrated that absolute scar size assessed by LGE–CMR was correlated with the inducibility of ventricular tachycardias during electrophysiological studies. Those results could successfully be reproduced in the studies of Schmidt et al,13 Roes et al14 and Yan et al,17 despite the fact that LGE analysis in estimating scar was not uniform.13 14 The current study reveals that these methodological differences do not lead to appreciable variation in total scar size estimation and that all methods equally predict the occurrence of ventricular tachyarrhythmias. Furthermore, ICC for the different scar measurements for intra-observer and interobserver agreement are high and comparable between methods, except for the peri-infarct zone of the method by Yan et al.17 These ICC are comparable to results of previous studies.13 20

In addition, the authors distinguished the scar core from the peri-infarct zone around it. This peri-infarct zone is thought to be composed of both fibrosis and preserved myocytes characterised by inherent conduction abnormalities, and presumably identifies myocardium susceptible to ventricular arrhythmias.13 As a result of its mixed composition, the peri-infarct zone displays an intermediate signal intensity on LGE images that is higher than normal myocardium but lower than the infarct core. The region with intermediate signal intensity can be quantified using the signal intensity of remote non-enhancing and/or that of a hyperenhanced infarcted region, although a standard quantification method has not yet been developed. An important issue in the assessment of the peri-infarct zone is the partial volume effect. It hampers the assessment of the peri-infarct zone by overestimating it. A study by Schelbert et al21 showed that the peri-infarct zone might be overestimated more than twofold. They showed in an animal model with high resolution CMR that the peri-infarct zone increases when resolution diminishes due to the partial volume effect. Currently, there are no methods to compensate for this effect, as it is inherent to the spatial resolution. However, in each of the aforementioned studies, regardless of the quantification method, infarct heterogeneity appeared to be of stronger diagnostic value for the prediction of ventricular arrhythmias than total scar size. When these methodological approaches to quantify the border zone were applied in the current study population peri-infarct zone size varied considerably. These results imply that, in contrast to total scar size estimation, these methodological approaches may not be interchangeable. More importantly, the peri-infarct zone, irrespective of the methodology used, was not superior to total scar size in identifying patients at risk of ventricular tachyarrhythmias. Similar observations were made regarding scar core size, which also displayed heterogeneity between analysis methods and did not enhance diagnostic accuracy over total scar size. These data suggest that quantification of total scar size is sufficient for risk stratification purposes and the scar core and borderzone quantification can be disregarded.

From a clinical point of view, a cut-off value is desired to link scar size to the identification of patients who are most likely to benefit from ICD implantation. Unfortunately, as depicted in figure 2, there is considerable overlap in scar size between arrhythmic and non-arrhythmic patients. The latter holds true for each of the investigated methodologies and each of the scar parameters (ie, total scar size, scar core size and peri-infarct zone). Nonetheless, a lower threshold could be identified in each of the methodological approaches beyond which no ventricular arrhythmias could be detected. On average this could hypothetically exclude approximately one quarter of patients eligible for ICD implantation for primary prevention in ischaemic cardiomyopathy. However, the study consisted of a rather small population and only 14 patients experienced ventricular tachyarrhythmia. Therefore, it would be premature to conclude that these suggested thresholds will be valid. Obviously, more studies in larger numbers of patients using uniform LGE–MRI analysis are warranted to explore this hypothesis further.

A number of limitations of the current study need to be addressed. First, the endpoint in the current study was the occurrence of ventricular fibrillation or ventricular tachycardia of at least 30 s with a cycle length of 400 ms or less. The incidence of ventricular tachycardia, however, far exceeds the incidence of sudden cardiac death. Deeming implantation of a ICD as appropriate based on the occurrence of ventricular tachycardia alone is therefore questionable. As a result of the limited number of patients in the present analysis, however, the use of this surrogate endpoint was required. Second, the present study included only ischaemic cardiomyopathy patients in whom a ICD had been implanted for primary prevention of sudden cardiac death. These data can therefore not be extrapolated to patients with a preserved LVEF, non-ischaemic cardiomyopathy, or patients with secondary prevention. Differences in patient population characteristics may also account for the discrepancies in the diagnostic accuracy of LGE–CMR compared with the aforementioned studies. Furthermore, the present analysis also included patients in whom ICD therapy was combined with cardiac resynchronisation. As biventricular pacing may also diminish the susceptibility to ventricular tachyarrhythmias, this may have introduced a bias,22 23 although no significant difference was seen in the prevalence of cardiac resynchronisation therapy between the patients with and without ventricular arrhythmias. Furthermore, patients with a recent myocardial infarction were excluded from the current study, as these patients are not eligible for ICD therapy according to the current guidelines. However, subclinical ischaemia/necrosis that may have caused oedema could not be excluded and, therefore, the extent of LGE might be overestimated due to the presence of oedema. Another point is the medication during follow-up in the study population. Alterations in medication may have taken place and might potentially have some effect on the occurrence of ventricular arrhythmias. Furthermore, there was a difference in age between patients who experienced ventricular tachyarrhythmia and patients who did not. Although age is not known as an important risk factor for ventricular arrhythmias, one could not exclude that differences in age had some effect on the results. In addition, the study population was rather small. The current study was not able to demonstrate any significant differences in AUC; however, in a larger patient population some difference might be detected. Finally, the partial volume effect hampers the assessment of the peri-infarct zone and might influence the analysis methods differently.

In conclusion, LGE–CMR-obtained scar tissue characteristics were shown to be predictors of ventricular tachyarrhythmias in ischaemic cardiomyopathy patients. The quantity of total scar size estimation is relatively independent of the methodologies investigated in the current study. Furthermore, analysis of scar core zone and peri-infarct zone does not seem to enhance the predictive value over the quantification of total scar size alone in the current study. Finally, considerable overlap in scar size between patients with and without documented ventricular arrhythmias exits. However, in each of the methodological approaches a lower threshold could be identified beyond which no ventricular arrhythmias could be detected.

References

View Abstract

Footnotes

  • Funding This research was performed within the framework of CTMM, the Center for Translational Molecular Medicine (http://www.ctmm.nl), project COHFAR (grant 01C-203), and supported by The Netherlands Heart Foundation.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.