Cardiovascular magnetic resonance-based predictors of complete left ventricular systolic functional recovery after rhythm restoration in patients with atrial tachyarrhythmia

Abstract

Aims

To establish a cardiovascular magnetic resonance (CMR)-based prediction model for complete systolic left ventricular ejection fraction (LVEF) recovery for the distinction of ‘arrhythmia-induced’ from ‘arrhythmia-mediated’ cardiomyopathy in patients with atrial tachyarrhythmias.

Methods and results

Two hundred and fifty-three tachyarrhythmia patients referred for catheter ablation were enrolled and underwent CMR baseline imaging; patients with a reduced LVEF <50% at baseline and CMR imaging at 3-month follow-up after successful rhythm restoration constituted the final study population (n = 134). CMR at baseline consisted of standard functional cine imaging, determination of extracellular volume, and late gadolinium enhancement (LGE) imaging; follow-up CMR comprised standard functional cine imaging. Left ventricular end-diastolic volume index (LVEDVI) measurements were categorized in ‘opposite’, ‘normal’, and ‘enlarged’. At follow-up, 80% (107/134) presented with complete LVEF recovery, while in 20% (27/134) persistent LVEF impairment was observed. LVEDVI and LGE were independent predictors of complete LVEF recovery with LGE adding significant incremental value on logistic regression modelling. Model-derived probabilities for complete LVEF recovery in LVEDVI categories of opposite, normal, and enlarged for LGE negativity and positivity were 94%, 85%, and 29% and 77%, 55%, and 8%, respectively.

Conclusion

CMR-derived assessment of LVEDVI category and LGE allowed for identification of arrhythmia-induced cardiomyopathy with acceptable discriminative performance. Probabilities for complete LVEF recovery for the combination of opposite LVEDVI/LGE negativity and enlarged LVEDVI/LGE positivity were 94% and 8%, respectively. The CMR-based prediction model of complete LVEF recovery can be used to perform upfront stratification in atrial tachyarrhythmia-related LVEF impairment.

Open in new tabDownload slide

Introduction

The pathophysiologic concept of arrhythmia-related cardiomyopathy associates the presence of atrial or ventricular tachyarrhythmias to the concomitant left ventricular (LV) functional impairment and resultant systolic heart failure symptoms. The hallmark of this condition is partial or complete reversibility of LV function once arrhythmia control has been achieved. Hence, by definition, proof that a ‘culprit arrhythmia’ constituted the sole reason for or a contributing factor to worsening systolic LV function can only be obtained in retrospect by evidence of improved LV functional status on follow-up examinations after rhythm restoration.^1,2 At the time of presentation, however, clinicians face the problem to discern whether an arrhythmia is the initiator or the consequence of cardiomyopathy in a patient with tachyarrhythmia and heart failure (‘chicken-egg dilemma’) and usually a high degree of clinical suspicion is needed to pinpoint subtle diagnostic clues.³ Cardiovascular magnetic resonance (CMR) imaging is the recognized standard of reference for assessment of ventricular volumes and function and provides myocardial tissue characterization during short single session examinations (<30 min).

Thus, the primary aim of the present study was to establish a CMR-based prediction model for complete systolic LV functional recovery following rhythm restoration in order to identify ‘arrhythmia-induced’ cardiomyopathy.

Methods

Patient population

The study was conducted in accordance with the local institutional review board and the standards of the University of Leipzig ethics committee; written informed consent was obtained from all patients prior to the CMR examination. Consecutive patients with sustained atrial tachycardia (heart rate ≥ 100 beats per minute) referred for rhythm control using catheter ablation were prospectively included. Patients were not considered for study inclusion in the presence of general contraindications to CMR imaging (e.g. pacemaker, intra-orbital metallic debris and other CMR-incompatible metallic implants), pregnancy, known adverse reaction to gadolinium-based contrast agents, or chronic renal failure (estimated glomerular filtration rate ≤ 30 mL/min). Patient characteristics were recorded at the time of study inclusion.

Cardiovascular magnetic resonance imaging

All CMR studies were carried out on a 1.5 Tesla MR scanner (Philips Ingenia, Best, The Netherlands) equipped with a 28-element array coil with full in-coil signal digitalization and optical transmission. CMR protocols adhered to well-established, standardized procedures.⁴ Baseline CMR exams consisted of cine imaging, a contrast‐enhanced three-dimensional CMR angiographic scan of the left atrium and the pulmonary veins, late gadolinium enhancement (LGE) imaging and T1 mapping using a modified Look-Locker inversion recovery (MOLLI) sequence prior to and 15 min after contrast administration. CMR angiography was solely acquired for clinical purposes: volume-rendering reconstruction of the three-dimensional angiographic dataset was performed and a surface mesh model of the left atrium and the pulmonary veins was generated with the exported mesh data used for electroanatomical mapping and anatomical guidance during subsequent ablation procedures. Follow-up CMR examinations comprised of contrast-agent free cine imaging adhering to standard protocols for ventricular and atrial volumetric assessment.

Cardiac volumes and function

Cine images were acquired with a steady-state free precession (SSFP) sequence over 30–40 cardiac phases with a reconstructed spatial resolution of 1.3 × 1.3 × 8.0 mm³ using multiple gap-less short-axis views covering the entire left ventricle, multiple gap-less four-chamber views covering the entire left atrium and a standard three- and two-chamber view. Cardiac chamber parameters were measured by level III certified CMR readers and indexed for body surface area.⁵ In order to ease interpretation and facilitate routine clinical applicability, LV end-diastolic volume index (LVEDVI) measurements were categorized in three groups in accordance with recommended gender-dependent reference values as follows: ‘opposite’ (i.e. values that outside the normal range but in the opposite direction of typical pathology; female < 56 mL/m², male < 57 mL/m²); ‘normal’ (i.e. values within the normal range; female 56–96 mL/m², male 57–105 mL/m²); ‘enlarged’ (i.e. values outside and above the reference range; female ≥ 97 mL/m², male ≥ 106 mL/m²).^6,7 In compliance with current European Society of Cardiology/American Heart Association (ESC/AHA) guidelines, a systolic LV ejection fraction (LVEF) ≥50% was considered to be normal and on follow-up indicated complete systolic LV functional/LVEF recovery.⁸

Late gadolinium enhancement

Three-dimensional LGE imaging was performed in all cardiac standard geometries about 10–15 min following intravenous gadolinium-based contrast administration (Dotarem^®, Guerbet, France; total dosage, 0.2 mmol/kg) with a spatial resolution of 1.6 × 1.6 × 5.0 mm³. LGE images were obtained using inversion-recovery gradient echo sequences with individually adjusted inversion times set to null myocardial tissue signal (inversion time, 230–320 ms). The presence and extent of myocardial LGE, considered synonymous with replacement fibrosis or scar tissue, was assessed by level III certified CMR experts.

Extracellular volume measurements

An ECG-gated modified Look-Locker inversion (MOLLI) sequence with SSFP image readout was acquired at a representative mid LV short-axis geometry pre- and 15 min post-contrast. Extracellular volume (ECV) was determined for the six LV segments of the mid short axis (IntelliSpace Portal 9.0, Philips, Best, The Netherlands) using the haematocrit level from a same day blood sample. The corresponding short-axis LGE image was reviewed in a side-by-side fashion: LGE positive myocardial segments were excluded from subsequent ECV calculations and LGE negative segments were averaged yielding a global fibrosis/scar-free ECV value per patient.⁹ ECV measurements were dichotomized into normal/abnormal according to previously published gender-dependent reference values with abnormal exceeding the upper limit of the reported 95% confidence interval (mean ± 2 SD for male 26.1 ± 2.3%, and for female 28.7 ± 2.6%).¹⁰

Follow-up

For all patients with baseline CMR and initially successful restoration of sinus rhythm, a follow-up visit after 3 months was scheduled. Follow-up data included questioning clinical symptoms, a 12-lead ECG for rhythm documentation, and in case of sinus rhythm a follow-up CMR examination. Patients with a documented relapse of their rhythm disorder underwent a repeat rhythm control attempt and were invited for another follow-up visit 3 months later.

Statistical analysis

All analyses were done using SPSS (version 21, IBM Corporation, Armonk, NY, USA). Continuous variables were given as mean ± standard deviation for normally distributed data; frequencies and percentages were used to describe categorical data. Differences between continuous and categorical variables were assessed using Student’s t-test, analysis of variance, Chi-square test, and Wilcoxon signed ranks test as appropriate. All tests were two-tailed and a P-value of <0.05 was considered significant.

The treatment effect of rhythm restoration on LVEF was determined with Cohen’s d (effect size calculated as the difference of the means divided by the standard deviation) using the following scale: d = 0.2 indicates small; d = 0.5, medium; d = 0.8, large; d = 1.2, very large; d > 2.0, huge treatment effect.

Binary logistic regression analysis was used to model the relationship between LVEDVI and LGE measurements and to assess whether LGE added incremental value to predict complete LVEF recovery. When the difference between ‘−2 log likelihood’ for the model (which has a χ² distribution with one degree of freedom) is statistically significant, the parameter added significant incremental value.

Receiver-operating characteristic analysis was performed for model prediction of complete LVEF recovery and the areas under the curves (AUCs) were compared.¹¹ Global measures of goodness of fit were calculated for the model: Pseudo-R² (Nagelkerke), the ‘c-statistic’ and a test of model calibration (Hosmer–Lemeshow). In order to gauge the discriminative performance of the prediction models for dichotomous outcomes, the AUC is equivalent to the concordance probability (‘c-statistic’) and was graded as acceptable (0.70–0.79), excellent (0.80–0.89), and outstanding (≥0.90) discrimination.¹²

Results

Patient population

A total of 253 study participants were enrolled during July 2016 to September 2019; all patients with reduced LV systolic function (i.e. LVEF <50%) at baseline formed the final study population (n = 134) and were further analysed for predictors of complete LVEF recovery (Figure 1). Rhythm at baseline was atrial fibrillation, typical/atypical atrial flutter, and ectopic atrial tachycardia in 67% (89/134), 31% (42/134), and 2% (3/134), respectively, with a mean heart rate of 121 ± 18 b.p.m. (range 100–166 b.p.m.). The majority of patients exhibited normal ECV values with a mean ECV of 26.5 ± 2.9% (Table 1).

At follow-up, 107 of these 134 patients (80%) presented with complete LVEF recovery, while in 27 patients (20%) persistent LVEF impairment was observed.

Treatment effect of rhythm restoration

CMR parameters at baseline and follow-up in patients with and without complete LVEF recovery after rhythm restoration are summarized in Table 2. Figure 2 illustrates the comparison of baseline vs. follow-up LVEF per LVEDVI category: baseline LVEF was significantly different between groups (P < 0.001) but the gain in absolute LVEF percentage points did not significantly differ (opposite 19 ± 7%, normal 18 ± 8%, and enlarged 23 ± 9%, respectively; P = 0.092).

As a measure of treatment effect size Cohen’s d with corresponding 95% confidence interval was determined for all patients and in LVEDVI categories of opposite, normal, and enlarged and amounted to 2.27 (1.96, 2.58) and 2.55 (2.03, 3.07), 2.05 (1.64, 2.45) and 4.58 (2.99, 6.17), respectively. Thus, an overall huge treatment effect regarding LVEF improvement after successful rhythm restoration was observed (see also supplementary material).

Left ventricular LGE was present in 16% (22/134) of all patients. There was no significant difference in LVEDVI between LGE-negative (67 ± 24 mL/m²) and LGE-positive patients (71 ± 21 mL/m²; P = 0.455). LGE-negative patients had a higher LVEF at baseline (37 ± 9% vs. 32 ± 9%, P = 0.028), while the gain in LVEF improvement was similar (19 ± 7% and 18 ± 10%; P = 0.511). Overall, LGE-negative patients were more likely to experience full LVEF recovery after successful rhythm restoration (88%, 94/112) than LGE-positive patients (59%, 13/22; P = 0.008).

Univariable analysis

In univariable analysis, gender, LVEDVI and LGE were significantly associated with complete LVEF recovery (Table 1) and were used for subsequent binary logistic regression analysis and model generation. Distribution of arrhythmia type did not significantly differ between patients with complete LVEF recovery [atrial fibrillation, 65% (70/107); atrial flutter, 32% (34/107); ectopic atrial tachycardia, 3% (3/107)] and those with persistent LVEF impairment [atrial fibrillation, 70% (19/27); atrial flutter, 30% (8/27); P = 0.648].

Binary logistic regression and receiver-operating characteristic analysis

In logistic regression analysis, LVEDVI category and LGE proved independent significant predictors of complete LVEF recovery, while gender did not significantly contribute (P = 0.298) and was removed. Thus, models for prediction of complete LVEF recovery were generated incorporating LVEDVI category (Table 3, model A) and LVEDVI category and LGE (Table 3, model B). In model A, a change in category (i.e. from enlarged to normal or normal to opposite) increased the odds for complete LVEF recovery about five-fold. In model B, LGE added incremental value and LGE negativity increased the odds for complete LVEF recovery about four-fold. The models had good overall fit, as demonstrated by their likelihood ratios and the decrease in likelihood ratios from model A to model B indicating that LGE added significant independent incremental value to LVEDVI category (Table  3, P = 0.011 for the comparison of model A vs. B).

Receiver-operating characteristic curves of logistic regression models for prediction of complete LVEF recovery are displayed in Figure 3; discrimination performance measures of model A and B resulted in acceptable discrimination (AUC of model A and B, 0.72 and 0.77, respectively).

Model-derived probabilities for complete LVEF recovery for all possible combinations of LVEDVI category and LGE were calculated using reference (‘dummy’) coding and are given in Table 4.

Discussion

The main findings of the present study can be summarized as follows: (i) CMR-based parameters (LVEDVI category, presence/absence of LGE) can be used for prediction of complete systolic LV functional recovery in tachyarrhythmia patients, (ii) binary logistic regression modelling demonstrated acceptable discriminative performance with LGE adding significant incremental diagnostic value; model-derived probabilities for the combinations of LVEDVI category and LGE may be used as a guide in clinical decision making and demonstrated that (iii) probability of complete LVEF recovery was highest for the combination of opposite LVEDVI category and LGE negativity (94%) and lowest for enlarged LVEDVI and LGE positivity (8%), and (iv) identified patients with normal LVEDVI and positive LGE as indistinct regarding complete LVEF recovery (i.e. about 50% probability). Finally, a huge treatment effect of rhythm restoration could be observed overall and in all LVEDVI categories.

Still, there are no established diagnostic criteria for detection of ‘tachycardiomyopathy’. Historically and conceptually, the distinction between ‘arrhythmia-induced’ and ‘arrhythmia-mediated’ cardiomyopathy has mainly been based on a high degree of clinical suspicion e.g. taken the time course of symptoms or prior baseline routine assessments of one-dimensional LV diameters/LVEF into account which are frequently missing or are unavailable at the time of initial symptomatic presentation.^1–3 Unfortunately, objective criteria for profound clinical decision making regarding identification of tachyarrhythmia as ‘culprit arrhythmia’ and attributing it to constitute the sole reason for deterioration of systolic LV function (generally termed ‘arrhythmia-induced’ cardiomyopathy) were missing. Thus, the final diagnosis of ‘arrhythmia-induced’ cardiomyopathy could only be established in retrospect demonstrating complete systolic LV functional recovery on follow-up examinations after successful rhythm restoration.³ Our study data demonstrated that the clinically important distinction between ‘arrhythmia-induced’ and ‘arrhythmia-mediated’ cardiomyopathy can be substantiated at the time of initial clinical presentation by readily available CMR-based parameters in a rather simple binary logistic regression model. The assessment of categorized LVEDVI measurements (i.e. opposite, normal and enlarged) together with visual assessment of the presence or absence of LGE proved to be of high predictive value for complete systolic LV functional recovery following rhythm restoration. The study data suggest that the combination of LVEDVI/LGE may be used as an upfront classifier for early detection of ‘arrhythmia-induced’ cardiomyopathy: this is corroborated by the high sensitivity (90%) and the model-derived probabilities in the opposite LVEDVI group (LGE negative: 94%, LGE positive: 77%). Notably, in the normal LVEDVI group, LGE negative subjects predominantly constitute ‘arrhythmia-induced’ cardiomyopathy (probability of complete LVEF recovery, 85%), while LGE positive subjects remain indistinct (probability of about 50%). On the other hand, ‘arrhythmia-mediated’ cardiomyopathy may be early recognized by an enlarged LVEDVI category. However, it must be stressed that the treatment effect defined as improvement of systolic LV function was similar in all LVEDVI categories (i.e. an about 20% gain in absolute LVEF percentage points per LVEDVI category): this is of particular importance in the enlarged LVEDVI subgroup since it may lift these patients out of the impending indication for an implantable cardioverter-defibrillator in case of a severely reduced LVEF < 35% at initial presentation.⁸ In summary, patients with ‘arrhythmia-mediated’ cardiomyopathy exhibited a larger LV with a significantly lower baseline LVEF and a higher likelihood of positive LGE when compared to patients with ‘arrhythmia-induced’ cardiomyopathy. Future research shall be directed to further improve CMR-profiling and characterization in this very important predominantly ‘arrhythmia-mediated’ subgroup.

Finally, all CMR parameters needed for stratification of patients with atrial tachyarrhythmias can be slipstreamed into established comprehensive standard CMR protocols at no additional cost and without prolonged examination duration (total examination duration ≤ 30 min).¹³ Such a comprehensive approach can be employed to derive all clinically relevant information regarding the presence of structural heart disease (myocardial tissue characterization and detection of haemodynamically relevant stenoses in obstructive coronary artery disease during pharmacological stress CMR), depiction of pulmonary venous anatomy (three-dimensional pulmonary venous angiography for anatomical guidance during mesh-model based electrophysiological interventions), and for preprocedural stratification of tachyarrhythmia patients (‘arrhythmia-induced’ vs. ‘arrhythmia-mediated’).

Study limitations

This prospective study was conducted at a highly specialized, tertiary care centre: while the study cohort closely reflected the clinical spectrum typically encountered in the given setting, the data as such may only be applicable to a similar clinical scenario. In addition, subgroup analysis resulted in a disparity of sample size between groups since the majority of patients returned to normal LVEF (≥50%) after successful rhythm restoration. Hence, the present study results shall be corroborated in future large-scale multicentre studies. Patients who had failed attempts at rhythm control did not undergo CMR imaging at follow-up and consequently, a respective control group of patients with persistent atrial arrhythmia was not considered in the current study.

In general, and from a methodological standpoint, a derivation cohort is used to develop the model and subsequently, the validity of the model shall be tested in a validation cohort. As mentioned above, the prediction model of the current prospective study was derived from consecutive patients referred to a single tertiary care centre for catheter ablation of atrial tachyarrhythmia. Basically, techniques of model validation include ‘internal’ validation (i.e. splitting of the dataset of the own study which is known to result in over-optimistic estimates of the model’s accuracy) or ‘external’ validation (i.e. assessing the applicability of the model to data collected at another centre or by other researchers) with the latter approach addressing the additional issue of generalizability of the model either.¹⁴ However, external model validation was beyond the scope of the current study and shall be addressed in future research efforts.

Conclusion

The assessment of CMR-defined parameters of LVEDVI and LGE allowed for identification of ‘arrhythmia-induced’ cardiomyopathy with acceptable discriminative performance; probabilities for complete LVEF recovery for the combination of opposite LVEDVI/LGE negativity and enlarged LVEDVI/LGE positivity were 94% and 8%, respectively. Hence, the CMR-based prediction model of complete LVEF recovery can be used to perform upfront stratification of patients presenting with atrial tachyarrhythmia-related LVEF impairment.

Supplementary material

Supplementary material is available at Europace online.

Conflict of interest: none declared.

Data availability

All data are incorporated into the article and its online supplementary material.

References