Diagnostic performance of left ventricular mechanical dyssynchrony indices using cardiovascular magnetic resonance feature tracking

1 Background

Cardiac resynchronization therapy (CRT), induces reverse remodeling, improves symptoms, and reduces mortality in carefully selected heart failure (HF) patients with left ventricular (LV) dyssynchrony [1]. However, a nonresponse rate of 30-40% is still observed [2], indicating a need for further refinement of current criteria for selecting patients to undergo CRT. Left ventricular dyssynchrony is a discoordination of intraventricular regional mechanical LV activation. Mechanical dyssynchrony has been proposed as one of the principal therapeutic targets of CRT [3]. As such, mechanical dyssynchrony makes for a potential predictor of CRT response.

Cardiovascular magnetic resonance (CMR) feature tracking has been used to characterize mechanical contraction features in patients with left bundle branch block (LBBB) compared to healthy controls [4]. The COMPANION, REVERSE, and MADIT-CRT trials demonstrated that patients with LBBB derived greater benefit from CRT compared to those with other conduction patterns [5–7]. Furthermore, positive CRT response and “super-response” to CRT have to an even greater extent been found to be associated with strict LBBB, as defined by Strauss et al. [8], when compared to LBBB defined by conventional criteria [9, 10]. This is likely a result of greater specificity for the strict LBBB criteria for identifying significant LV dyssynchrony, the substrate that is corrected by CRT.

Abnormalities of regional myocardial mechanics can be caused by conduction abnormalities, especially LBBB, or by focal myocardial fibrosis, also referred to as scar [11]. Myocardial scarring has been found to be negatively associated with CRT response [12–15]. Indices of mechanical dyssynchrony are often derived from measurements of cardiac deformation such as strain. Strain measurements have been found to be greatly affected by the presence of myocardial scar [16–18] or myocardial edema [17, 18]. Importantly, the failure of most indices of mechanical dyssynchrony can be attributed to a lack of specificity [19]. Consequently, the goal of dyssynchrony analysis should be detection of a mechanical dyssynchrony pattern amenable to CRT, and not the detection of mechanical dyssynchrony per se [19]. Information about mechanical dyssynchrony characteristics in patients with a high likelihood of CRT response, while at the same time controlling for concomitant pathology, e.g myocardial scar, is of great interest in identifying predictors of CRT response. However, information about this is scarce.

Therefore, the aim of this study was to investigate mechanical dyssynchrony characteristics of patients with strict LBBB and normal QRS duration, a left ventricular ejection fraction ≤35% and no LV myocardial scar assessed by CMR late gadolinium enhancement (LGE).

2 Methods

This is an observational case-control study where patients were retrospectively identified by cross-referencing the CMR and electrocardiography (ECG) databases from three centers (Duke University Medical Center, NC, USA; Pittsburgh University Medical Center, PA, USA; and Karolinska University Hospital, Stockholm, Sweden). The study was approved by the local human subject research ethics committee at each site, and all subjects either provided written informed consent or were included following a retrospective waiver of informed consent provided by the local ethics committee.

3 Results

3.2 Dyssynchrony measurements

Both CURE and SSI showed group differences between LBBB and controls (Figure 1). Consistent with a greater amount of mechanical dyssynchrony in LBBB, CURE was lower in LBBB compared to controls (0.63 [0.54-0.75] vs 0.79 [0.69-0.86], p<0.001), and SSI was higher in LBBB compared to controls (9.4 [7.4-12.7] vs 2.2 [1.2-3.6], p<0.001). Compared to CURE, SSI had a greater area under the ROC curve for detecting mechanical dyssynchrony associated with strict LBBB (0.98 [95% CI: 0.95-1.00] vs 0.77 [95% CI: 0.66-0.86], p< 0.001, Figure 2), and this corresponded to a higher sensitivity and specificity for SSI compared to CURE (Figure 3). The odds ratio (OR) for identifying LBBB for SSI was 3.40 [95% CI: 1.84-6.28] per 1 percentage unit increase in SSI value, and for CURE was 2.16 [95% CI: 1.44-3.24] per 0.10 decrease in CURE value. In evaluating the need for covariate-adjusted and/or covariate-specific ROC curves, linear regression models were used to test the association between dyssynchrony indices and covariates: age, LVEDVI, LVMI, and sex, allowing for interactions between LVMI and LVEDVI, with sex, respectively. No evidence was found in support of an association between dyssynchrony indices and covariates in controls, or between CURE and covariates in LBBB. Furthermore, neither CURE nor SSI associated with QRS duration in either LBBB or controls (p ≥0.4 for all). There was an association between SSI and age, keeping other covariates fixed, in LBBB. This suggests that the discriminatory ability of SSI might vary with respect to age. For purposes of clarity, the unadjusted ROC curve is presented.

• Download figure
• Open in new tab

Figure 1.

Values for the circumferential uniformity ratio estimate (CURE) and the systolic stretch index (SSI) for left bundle branch block (LBBB) and controls. The box and whisker plots show the median (horizontal line), interquartile range (box), and data points within 1.5× interquartiles ranges of the first and third quartile, respectively (whiskers). Note, there is more pronounced mechanical dyssynchrony (lower CURE, higher SSI) in LBBB compared to control. CURE is more homogenously distributed between groups compared to SSI.

• Download figure
• Open in new tab

Figure 2.

Receiver operating characteristics (ROC) for univariable logistic regression models to differentiate between left bundle branch block (LBBB) and controls using the circumferential uniformity ratio estimate (CURE) and the systolic stretch index (SSI), respectively. Better discriminatory ability for LBBB is seen for SSI compared to CURE. AUC = area under the ROC curve, CI = confidence interval.

• Download figure
• Open in new tab

Figure 3.

Bar plot showing sensitivity and specificity for detecting left bundle branch block using the systolic stretch index (SSI) and the circumferential uniformity ratio estimate (CURE), respectively. Error bars denote 95% confidence intervals.

4 Discussion

The main finding of the study is that the ability to discriminate between LBBB and normal QRS duration among patients with severely reduced LV ejection fraction and no scar was fair for CURE and excellent for SSI. This highlights that, when developing and evaluating indices aimed at accurately identifying mechanical dyssynchrony amenable to CRT, it is important to evaluate the performance of a proposed index both in patients with LBBB, and in comparison to control subjects with normal QRS duration.

The lack of response and even harmful effects of CRT when implanted in patients with narrow QRS complexes indicates that CRT requires an electrical substrate [32, 33]. Findings that LBBB patients derive greater benefit from CRT compared to those without support that LBBB is this electrical substrate [5–7]. Response to CRT can partly be explained by correction of the discoordinated contraction of myocardial wall segments, a consequence of abnormal electrical activation. However, not all patients with LBBB by conventional electrocardiographic criteria have complete LBBB [34, 35], and mechanical dyssynchrony is not uniquely associated with electrical dyssynchrony [11, 36]. For example, focal LV myocardial scarring is also known to cause abnormalities in regional myocardial mechanics [11]. Approximately one third of patients with LBBB by conventional criteria do not have strict LBBB [37]. Consequently, Strauss, et al, proposed more strict criteria for LBBB [8]. Increased rate of CRT response has been found when using this strict definition of LBBB [9, 10, 38], and strict LBBB is associated with greater mechanical dyssynchrony than non-strict LBBB [39]. Mechanical dyssynchrony per se might therefore not be of primary interest in predicting CRT response but instead focus should be on identification and quantification of the mechanical dyssynchrony pattern associated with an abnormal electrical activation as seen in complete LBBB. Importantly, the current study evaluates the discriminatory ability of two recently proposed mechanical dyssynchrony indices for the mechanical dyssynchrony pattern associated with strict LBBB compared to patients with equally reduced ejection fraction but normal QRS, while controlling for confounding factors such as scar.

QRS duration cannot accurately characterize the spectrum of conduction abnormalities, and so it seems unlikely that any singular mechanical dyssynchrony index will be able to capture the full spectrum of variation inherent to dyssynchronous ventricular contraction. While both CURE and SSI capture differences in the pattern of discoordination on a group level, we found SSI superior to CURE with regards to the ability to differentiate between LBBB and controls. Our results suggest that the incremental mechanical dyssynchrony component associated with LBBB is better characterized by quantification of the absolute extent of stretch during the opposing directions of movement in the septum and lateral wall.

SSI was developed by Lumens et al. [26] in an attempt to characterize the electromechancial substrate that may respond to CRT. They used a computational model to simulate electromechanical and non-electrical substrates of mechanical dyssynchrony and identified strain characteristics specific for the different substrates of mechanical dyssynchrony. In a study of patients enrolled in the Adaptive CRT trial, it was found that SSI by echocardiography was independently associated with CRT outcome, adjusting for QRS morphology, QRS duration, sex, heart failure etiology, and treatment with angiotensin-converting enzyme inhibitors/angiotensin II receptor blockers [25]. However, whether SSI has added prognostic value over strict LBBB morphology is still unknown. It remains to be explored whether SSI can be used to identify non-LBBB patients that may be suitable for CRT, and such studies are justified.

CURE was first evaluated in a canine model of heart failure and LBBB conduction delay [22, 24]. It was found that biventricular pacing leads to greater synchrony (increased CURE), improved global function, and that circumferential dyssynchrony indices had greater dynamic range when compared to longitudinal indices [24]. Importantly, CURE was found sensitive to regionally clustered dyssynchrony [24]. Regionally clustered dyssynchrony might show equal variance as dispersed dyssynchrony when compared to variance-based dyssynchrony measures, although with very different effects on cardiac mechanics [24]. While CURE can be considered a more general measure of dyssynchrony, it has been found to be predictive of CRT response in clinical cohorts [13, 23, 40]. An advantage of CURE over commonly used time-to-peak based indices is that CURE utilizes information of the full cardiac cycle. Additionally, considering that CURE is derived from the relative positions of included segments and less so on their absolute value of strain, CURE is theoretically less sensitive to inter-vendor variations of strain measurements. However, despite these theoretical advantages of using CURE, the current study shows that CURE had a modest performance in identifying LBBB specific mechanical dyssynchrony.

The association between mechanical dyssynchrony, quantified as the systolic dyssynchrony index, and myocardial scar has been studied in patients with systolic heart failure [41]. They conclude that 25% of patients with narrow QRS (<130 ms) presented with mechanical dyssynchrony, despite no difference in scar burden compared to narrow QRS patients without mechanical dyssynchrony. Those findings suggest that mechanical dyssynchrony in such patients might be secondary to myocardial scar rather than electrical dyssynchrony. There is no general agreement upon the definition of mechanical dyssynchrony, and the difference in vendor software for strain measurements limits straightforward comparisons between studies. CURE has been shown to identify a greater magnitude of dyssynchrony (lower CURE values) in patients with non-ischemic cardiomyopathy compared to healthy controls (0.79±0.14 vs 0.97±0.02) [42]. In a different study, CURE in healthy controls volunteers has been shown to be 0.87±0.07 [43]. The current study shows that patients with severely reduced LVEF and normal QRS duration have some degree of mechanical dyssynchrony even in the absence of scar (median CURE 0.79). This would suggest that other factors beyond scar and LBBB contribute to mechanical dyssynchrony detected by CURE. Such factors may include variations in pre-load and/or afterload, and regional wall motion abnormalities due to chronic ischemia or other non-ischemic cardiomyopathies that impair contractile function without causing myocardial scar.

While CURE and SSI both displayed group differences between LBBB and controls, the current study found no evidence in support of a relationship between either CURE or SSI, and QRS duration within LBBB and control groups, respectively. CURE and QRS duration have previously been found to be modestly correlated (r = −0.58; p<0.001) in a cohort (n = 43) of cardiomyopathy patients with similar reductions in ejection fraction and prolongation of QRS duration, though QRS morphology was not reported [23]. However, when only evaluating the correlation between CURE and QRS duration in those patients referred for CRT (n = 20) those authors found that the evidence did not support any correlation (r = −0.40; p = 0.08) [23]. The apparent lack of correlation between dyssynchrony and QRS duration is of interest considering that current guidelines are still unclear regarding the group of patients with intermediate QRS width (QRS 120-149ms). These exploratory results add to the notion that there is a complex relationship between electrical and mechanical dyssynchrony.

5 Image availability

The current study shows that patients with severely reduced ejection fraction and normal QRS duration have a baseline level of mechanical dyssynchrony that is not attributable to myocardial scarring or prolonged depolarization of the myocardium. Consequently, when developing an index of mechanical dyssynchrony, specificity for its intended use should be of interest. To our knowledge, no study to date has included patients free of myocardial scar, with a normal QRS duration, and with severely reduced ejection fraction when comparing or developing indices of mechanical dyssynchrony. In order to facilitate future research where baseline mechanical dyssynchrony is accounted for, the images from the current study are made available online. See the section ‘Availability of data and materials’ below.

6 Limitations

Identification of the time point for aortic valve opening, and aortic valve closure was performed by visual assessment of CMR cine images, and the accuracy of this assessment is limited by the temporal resolution of CMR images. However, any variations in accuracy would equally affect the analysis of both patient groups, and hence should not have a major effect on the overall results. The software used for strain analysis reports segmental strain measurements according to the AHA 17-segment model. Hence, CURE calculation was limited to Fourier transformation applied to six individual myocardial segments in the midventricular short-axis slice. The impact of the spatial resolution of measurement in CURE quantification has not previously been reported. However, it cannot be excluded that quantification of CURE using higher spatial resolution could potentially influence the results.

7 Conclusion

SSI was superior to CURE regarding the ability to discriminate between strict LBBB and normal QRS duration among patients with severely reduced ejection fraction and no scar. The amount of dyssynchrony in patients with no scar, and severely reduced ejection fraction, needs to be taken into account when developing and evaluating indices aimed at accurately identifying mechanical dyssynchrony amenable to CRT. We suggest that control subjects with normal QRS duration, severely reduced ejection fraction, and the absence of myocardial scar, be included when developing and evaluating mechanical dyssynchrony indices aimed at predicting CRT response.

Funding

Not applicable.

Availability of data and materials

The datasets generated and/or analyzed during the current study, as well as code needed to reproduce all aspects of the current study, are available in the Figshare repository, doi:10.6084/m9.figshare.15155596. The most recent version of the analysis code is available in the Github repository, https://github.com/dloewenstein/dillacs-study.

Ethics approval and consent to participate

The study was approved by the local human subject research ethics committee at each site, and all subjects either provided written informed consent or were included following a retrospective waiver of informed consent provided by the local ethics committee.

Competing interests

EH is the founder of the company Medviso AB which develops medical image analysis software. RN has received research grants from Philips Volcano and Biotronik. RJK is a consultant for Abiomed. BDA has received research grants from Boston Scientific and Abbott, and consultation fees from Abbott, Medtronic, Biotronik, and Biosense Webster. MU is principal investigator on a research and development agreement regarding cardiovascular magnetic resonance between Siemens and Karolinska University Hospital. The remaining authors have nothing to disclose that is relevant to the contents of this paper.

Consent for publication

Not applicable.

Authors’ contributions

DEL: conception and design of the work, data acquisition and analysis, interpretation of data, drafting the work. BW: data acquisition and analysis, interpretation of data, substantively revised the work. EH: data analysis, interpretation of data, substantively revised the work. JA: data acquisition, substantively revised the work. IK: design of the work, data acquisition, interpretation of data, substantively revised the work. RN: data acquisition, interpretation of data, substantively revised the work. EBS: data acquisition, substantively revised the work. PS: data acquisition, substantively revised the work. AS: data acquisition, substantively revised the work. DGS: design of the work, substantively revised the work. RJK: data acquisition, substantively revised the work. BDA: conception and design of the work, interpretation of data, drafting the work. MU: conception and design of the work, interpretation of data, drafting the work. All authors read and approved the final manuscript.

Authors’ information

Not applicable.