Usefulness of Left Atrial Emptying Fraction to Predict Ventricular Arrhythmias in Patients With Implantable Cardioverter De ﬁ brillators

,

Impaired left atrial emptying fraction (LAEF) is an important predictor of mortality in patients with heart failure.As it may reflect increased LV wall stress, it might predict ventricular arrhythmia (VA) specifically.This study evaluated the predictive value of LAEF assessed with cardiovascular magnetic resonance (CMR) imaging with respect to appropriate device therapy (ADT) for VA and compared its role with CMR assessed scar size and other risk factors.In total, 229 patients (68% male, 63 -10 years, 61% ischemic cardiomyopathy) with LV ejection fraction £35% who underwent CMR and implantable cardioverter defibrillator (ICD) implantation for primary prevention in 2005 to 2012 were included.CMR was used to quantify LV volumes and function.LV scar size was quantified when late gadolinium enhancement was available (n [ 166).Maximum and minimum left atrial volumes and LAEF were calculated using the biplane area-length method.The occurrence of ADT and mortality was assessed during a median follow-up of 3.9 years.Sixty-two patients (27%) received ADT.Univariable Cox analysis showed that male gender, creatinine level, minimum left atrial volume, LAEF, and total scar size were significant predictors of ADT.In multivariable Cox analysis, LAEF (hazard ratio 0.75 per 10%, p <0.01), and scar size (hazard ratio 1.03 per g, p [ 0.03) remained the only independent predictors of ADT.Patients with both LAEF > median and scar size < median were at low risk (13% ADT at 5 years), whereas those with LAEF < median and scar size > median experienced 40% ADT at 5 years (logrank p [ 0.01).In conclusion, LAEF independently predicts ADT in patients with primary prevention ICDs.Combined assessment of LAEF and scar size identifies a group with low risk of ADT.Therefore, LAEF assessment could assist in risk stratification for VA to select patients with the highest benefit from ICD implantation.Ó 2017 The Authors.Published by Elsevier Inc.This is an open access article under the CC BY license (http://creativecommons. org/licenses/by/4.0/).(Am J Cardiol 2017;120:243e250) The implantable cardioverter defibrillator (ICD) implantation is the first choice therapy in patients with heart failure and left ventricular ejection fraction (LVEF) 35% for the primary prevention of sudden cardiac death (SCD). 1 However, follow-up studies report an ICD discharge rate of only 9% to 35% after 30 to 36 months in patients treated with ICDs for primary prevention. 2,3Enhanced risk prediction of ventricular arrhythmia (VA) beyond LVEF assessment, therefore, is needed.Previous studies have demonstrated that enlarged left atrial (LA) volumes (LAVs) and impaired LA emptying fraction (LAEF) predict the risk of heart failure progression and mortality.4e7 As the prognosis in patients with heart failure is partially driven by SCD, markers of LA dysfunction may also relate to VA in particular.Impaired LA function might reflect increased LV filling pressure and wall stress which may contribute to the formation of fatal arrhythmias due to modulation of action potential duration, calcium handling, and conduction.8e11 Data on the value of markers for LA dysfunction in predicting VA, however, are lacking.Using cardiovascular magnetic resonance (CMR) imaging, LA volumes and function can be accurately evaluated.In addition, LV scar tissue, which is known to be associated with the occurrence of VA, can be quantified.12e14 This study aimed to evaluate the predictive value of LA function with respect to appropriate device therapy (ADT) for VA and to compare its role with scar size and other risk factors.

Methods
This study included consecutive patients with ischemic and dilated cardiomyopathy (ICMP and DCMP) and LVEF 35% who received an ICD for primary prevention of SCD according to the American College of Cardiology/American Heart Association/European Society of Cardiology 2006 guidelines from January 2005 to December 2012 in the VU University Medical Center. 1 The patient population has been described in a previous study. 15For the present study, patients were included when CMR was performed within 6 months before ICD implantation.CMR was typically performed for LVEF assessment, detection of LV thrombus, guiding of LV lead placement by scar assessment in resynchronization therapy, or as part of other study protocols.Exclusion criteria were chronic atrial fibrillation, atrial flutter, or when no follow-up data were available.Primary prevention was defined as no history of sustained ventricular tachycardia (VT) or ventricular fibrillation (VF; >48 hours after acute myocardial infarction).Patient characteristics before device implantation were collected from medical records.The Medical Ethics Review Committee of the VU University Medical Center approved the data collection and management of this study.
CMR studies were performed on a 1.5-T whole body scanner (Magnetom Sonata/Avanto; Siemens, Erlangen, Germany) using a dedicated phased array body coil as described previously. 15Cine imaging was performed using a retrospectively electrocardiogram-gated, steady-state free precession sequence during breath holds in mild expiration.Standard long-axis slices were acquired from the 4-, 3-, and 2-chamber views.Subsequently, consecutive short-axis slices were acquired, fully covering the LV.Late gadolinium enhanced (LGE) images were acquired approximately 10 to 15 minutes after the administration of 0.2 mmol kg À1 gadolinium diethylenetriamine penta-acetate in the similar orientations as used in the cine images, using a 2-dimensional (2D) segmented inversion-recovery gradient echo sequence.Images were analyzed using the dedicated software package MASS (Mass v.5.1 2010-EXP beta; Medis, Leiden, the Netherlands).Endocardial LA contours and length were drawn manually in both the apical 2-and 4-chamber views in the frames just before mitral valve opening to obtain maximum LA volume (LAVmax) and immediately after mitral valve closure for minimum LA volume (LAVmin; Figure 1).LA length (L) was defined as the distance from the center of the mitral annulus to the posterior border of the LA area, perpendicular to the mitral annular plane.The LA appendage and pulmonary veins were excluded from the LA measurements.LA volumes were calculated using the biplane area-length formula: LAV ¼ 8ðA1ÞðA2Þ 3pðLÞ , according to the guidelines of the American Society of Echocardiography and European Association of Cardiovascular Imaging, where A1 and A2 represent the planimetered LA areas in the 2-and 4-chamber views, whereas for L, the shortest length of both views was used. 16LAEF was calculated as LAVmaxÀLAVmin LAVmax Â 100.LV end-systolic and enddiastolic volumes (LVEDV and LVESV) were measured using a standard method, and LVEF was calculated.Finally, LGE short-axis images were outlined manually at the endocardial and epicardial borders.Myocardial scar size was quantified automatically using the full width at half-maximum method. 17car size was expressed as total grams and percentages of LV mass.All CMR images were analyzed by experienced observers who were blinded to clinical follow-up data.Clinical follow-up with device interrogation was routinely performed with regular intervals of 6 months.The ICDs were typically programmed according to the PRE-PARE study with detection rates of > w180 beats/min (VT zone) and > w250 beats/min (VF zone), extended detection intervals of 30/40 ventricular beats or 8/10 ventricular beats plus a 5 to 8 seconds delay, depending on the device manufacturer, and appropriate utilization of antitachycardia pacing (ATP). 18,19Event transmissions of patients connected with home monitoring were reviewed instantly when they occurred.All recorded events and ADT were reviewed by specialized cardiac technicians or by electrophysiologists. ADT was defined as an episode of ATP and/or defibrillation shock to terminate VT or VF.The primary end point was defined as the occurrence of first ADT.The secondary end point was defined as either the occurrence of ADT or all-cause mortality.
Continuous data were expressed as means AE SD and were compared using the Student t test or Mann-Whitney U test when appropriate.Dichotomous and categorical data were expressed as frequencies and percentages and were compared using the chi-square test or, when appropriate, the Fisher's exact test.Univariable and multivariable Cox proportional hazard regression analyses were performed to identify predictors of the primary and secondary end points.Multivariable Cox proportional hazards models were performed using backward elimination with inclusion of variables with a p value below 0.10 in univariable Cox analysis.
Scar size was evaluated in a separate multivariable model (model 2) for the subgroup with LGE available (n ¼ 166).Both LAVmax index and LVESV index were not entered in multivariable analyses due to significant collinearity with LAVmin index and LVEDV index, respectively (correlation coefficients >0.90).Patients were categorized into low-and high-risk groups according to the median of LAEF (38.7%).In addition, patients with ICMP were categorized into lowand high-risk groups using the median of scar size (15.5 g), whereas DCMP patients were separately divided according to the presence or absence of scar size (median 0 g).Finally, a combined risk factor score was generated consisting of LAEF and scar size: 0 risk factors (both LAEF > median and scar size < median), 1 risk factor (LAEF < median or scar size > median), and 2 risk factors (both LAEF < median and scar size > median).Time to the primary and secondary end points were compared between risk groups using Kaplan-Meier curves and the log-rank test.A p value of 0.05 or less was considered statistically significant.All statistical analyses were performed using SPSS software package, version 20.0 (IBM SPSS Statistics, Chicago, Illinois).

Results
In total, 240 patients were evaluated.Eleven patients showed insufficient CMR image quality due to severe motion artifacts or irregularity in heart rhythm and were The cumulative 5-year incidence of ADT was 27% for the total study population and did not differ between patients with ICMP or DCMP (log-rank p ¼ 0.55).Parameters that were significantly associated with the occurrence of ADT in univariable Cox analyses included male gender, creatinine level, LAVmin index, LAEF, and scar size (Table 3).The multivariable Cox model for the total study population (model 1, excluding scar size) revealed that male gender and LAEF were independent predictors of ADT.When adding scar size to the multivariable analysis (model 2), LAEF and scar size remained independent predictors of ADT (Table 3).The secondary end point of ADT or mortality occurred in 97 of 229 patients (42%), with a cumulative incidence of 40% at 5 years.Univariable Cox regression analyses revealed that male gender, nonsustained VT, creatinine level, LAVmin index, LAEF, LVEDV index, LVESV index, and LVEF were significantly related to the occurrence of ADT or mortality (Table 4).In multivariable analysis (model 1), LAEF, nonsustained VT, creatinine level, and LVEDV index independently predicted ADT or mortality.When adding scar size to the multivariable model (model 2), LAEF and LVEDV index remained the only independent predictors of ADT or mortality, whereas for scar size, only a trend was observed.
Patients with LAEF < median were at significantly higher risk of experiencing ADT and ADT or mortality.Patients with scar size > median were at significantly higher risk of experiencing ADT and showed a trend toward a higher risk of ADT or mortality (Figure 2).A combined risk score using LAEF and scar size identified a low-risk group (13% ADT and 23% ADT or mortality at 5 years), an intermediate-risk group (24% ADT and 37% ADT or mortality at 5 years), and a high-risk group (40% ADT and 54% ADT or mortality at 5 years; Figure 3).

Discussion
The present study evaluated the role of markers of LA dysfunction in predicting ADT for VA in patients with ICMP and DCMP.In addition, the predictive value of LA dysfunction was compared with known risk factors for VA such as LV function and scar size.It was demonstrated that both impaired LAEF and scar size were independent predictors of ADT for VA.Furthermore, combined assessment of LAEF and scar size further stratified the patient cohort into high-, intermediate-, and low-risk groups for receiving ADT.Comparable results were obtained for predicting ADT or mortality.
The hemodynamics of the LV and LA are continuously intertwined, and LA function is related to LV volumes, filling pressure, and wall stress. 10For this reason, a renewed interest has emerged in the assessment of LA size and function for risk stratification in patients with heart failure.Several LA parameters can be calculated including LAVmax, LAVmin, and LAEF, which may each provide different information.Although the LAVmax mainly reflects increased atrial pressure and volume, the LAVmin is more closely related to the diastolic LV filling pressure (atrial afterload). 20Furthermore, total LAEF can be divided in active LAEF and passive LAEF.In a study by Posina et al, 21 total LAEF was found to be the strongest predictor of increased LV end-diastolic pressure, whereas active LAEF and passive LAEF were of less significance.The present study therefore evaluated total LAEF.Although 2D echocardiography is commonly used for this purpose, the present study used CMR which has a higher reproducibility and accuracy. 22,23Median LA volumes and LAEF as observed in the present study were comparable with the study by Pellicori et al 6 who found a CMR assessed median LAEF of 42% in patients with heart failure, whereas in the general population, the median LAEF was 52% to 55%. 5 The association of LA volumes or function and mortality or heart failure progression has been well established by multiple studies. 4,6,7As the prognosis in patients with heart failure is partially driven by SCD, it is of interest to evaluate whether the predictive value of LA dysfunction holds true for VA specifically.The present study is the first to show that impaired LAEF is an independent predictor of ADT for VA in patients with implanted ICDs.In a small retrospective study by Kaplan et al, 24 LAVmax assessed by 2D echocardiography was related to the occurrence of VA within the preceding year in 32 patients with ICDs implanted.In addition, Koilpillai et al 25 found a correlation between LA width and nonsustained VA frequency on Holter monitoring.Conversely, the present study found that LAEF and, to a lesser extent, LAVmin were related to VA, whereas LAVmax was not.The superior value of LAEF over LAVmax in predicting mortality was also demonstrated in more recent studies using CMR. 5,6Pellicori et al 6  Follow-up (years) Event-free survival (ADT or mortality) Figure 2. Kaplan-Meier curve analyses of the event-free survival of ADT (A and C) and ADT or mortality (B and D), compared between patients with LAEF < median and LAEF > median (A and B) and between scar < median and scar > median (C and D).Patients with LAEF < median showed higher incidence of ADT and ADT or mortality (HR 1.98, 95% CI 1.18 to 3.33, p ¼ 0.009, and HR 1.92, 95% CI 1.28 to 2.90, p ¼ 0.002, respectively, both compared with LAEF > median).Patients with scar size > median showed a higher incidence of ADT compared between scar size < median (HR 1.94, 95% CI 1.04 to 3.62, p ¼ 0.04), whereas a trend was observed toward a higher incidence of ADT or mortality (HR 1.55, 95% CI 0.94 to 2.56, p ¼ 0.08).multivariable analyses.Impaired LAEF was also found to be an independent predictor of mortality in the general population in a large study performed by Gupta et al. 5 The present study confirms that impaired LAEF is an important marker of poor prognosis and suggests that this might be partly explained by the occurrence of life-threatening arrhythmias.
The exact pathophysiological link between impaired LAEF and VA remains uncertain.Impaired LAEF may reflect high diastolic LV filling pressure and increased wall stress, which are associated with proarrhythmogenic effects on a cellular level including modulation of refractoriness, calcium handling, and conduction resulting in increased susceptibility of VA. 8,9,11 This hypothesis is underlined by studies that have evaluated the relation of VA and level of B-type natriuretic peptide (BNP), which is released in response to increased LV wall stress and pressure overload.They consistently demonstrate that increased BNP is an independent predictor of VA or SCD in patients with impaired LVEF. 26,27Although BNP level and LAEF are intertwined and may reflect comparable pathophysiological substrates, Pellicori et al 6 showed that LAEF was a more powerful predictor of mortality in patients with heart failure as compared with the level of BNP.The present study, however, did not evaluate the predictive value of BNP as this parameter was not available in a large part of the study population.As BNP is a marker which is easily obtained at low costs, it would be of interest to evaluate whether the value of LAEF is superior to BNP in predicting VA.
The pathophysiology of SCD is complex and multifactorial; therefore, it is likely that a combination of risk markers is needed to identify patients at highest risk of SCD.In addition to LAEF, this study identified total scar burden as an independent predictor of ADT for VA.These results are consistent with previous studies that also demonstrated the relation of scar size and VA in patients with ICMP and DCMP.12e14 Combined assessment of LAEF and scar size using CMR resulted in an improved risk stratification of receiving ADT, suggesting that the combination of a large anatomical substrate with increased wall stress may in particular increase vulnerability to VA.This notion is supported by an experimental study performed in chronically infarcted canine hearts by Calkins et al, 11 which revealed that increased ventricular loading, in particular at sites of extensive fibrosis, resulted in exaggerated shortening of refractoriness and increased inducibility of VA.More recently, the value of LGE assessed scar size combined with a different marker of increased LV wall stress, NT-proBNP, in risk stratification for VA has been evaluated by Mordi et al. 26 They found that patients in the lowest risk group showed 3% ADT or mortality per year during follow-up, whereas the incidence in high-risk patients was 10% per year.
Several limitations of the present study should be acknowledged.First, this study should be regarded as hypothesis-generating due to its retrospective character.Consequently, the acquired CMR images in the present study were not intended for LA evaluation specifically, and 3D LA volume assessment was not possible.However, the biplane area-length method is an accurate and reproducible method which is easily and rapidly obtained from the standard long-axis views. 28It has been used in most large studies that used CMR evaluation of the LA for risk stratification, which allowed comparisons with the results of the present study.4e6 Whether 2D echocardiographic measurements of LAEF could also identify patients at high risk of VA remains unclear and should be evaluated in future studies.Second, the present study did not evaluate intraobserver and interobserver variations.Nonetheless, excellent reproducibility of CMR measurements of LA volumes has been reported previously. 18,19Third, the occurrence of ADT for VA is a surrogate end point and does not equal the incidence of SCD.Finally, the included patient population with both LGE and LAEF assessment available was relatively small; therefore, adequate subgroup analyses could not be performed.Further prospective studies are needed to confirm the results of this study and define potential optimal cut-off values of LAEF for predicting VA.Follow-up (years) Event-free survival (ADT or mortality) Figure 3. Kaplan-Meier curve analyses of the time to ADT (A) and ADT or mortality (B), compared between patients at low risk (0 risk factors), intermediate risk (1 risk factor), and high risk (2 risk factors) according to combined LAEF and scar size assessment.The corresponding HRs for ADT in the high-and intermediate-risk groups were 3.89 (95% CI 1.51 to 9.97, p ¼ 0.005) and 2.18 (95% CI 0.87 to 5.46, p ¼ 0.10), respectively, compared with the low-risk group.For the occurrence of ADT or mortality, the HRs were 3.26 (95% CI 1.54 to 6.90, p ¼ 0.002) for the high-risk group and 2.05 (95% CI 1.00 to 4.20, p ¼ 0.049) for the intermediate-risk group compared with the low-risk group.

Figure 1 .
Figure 1.Example of CMR evaluation of the left atrium using the biplane area-length method in the apical 4-chamber view (A and B) and 2-chamber view (C and D).Measurements were performed in the frame just before mitral valve opening to obtain maximum LA volumes (B and D) and in the frame immediately after mitral valve closure for minimum LA volume assessment (A and C).

Table 1
Baseline characteristicsVariable N(%), median (interquartile range), or mean AE standard deviation Diuretics include mineralocorticoid receptor antagonists.ADT ¼ appropriate device therapy.*Dataavailableforn¼178.†Compared using the chi-square test for trend.zDataavailableforn¼ 218.xCompared using the Mann-Whitney U test.Arrhythmias and Conduction Disturbances/LAEF Predicts Ventricular Arrhythmias excluded from the analysis.Therefore, 229 patients were included in the present study.Clinical baseline characteristics are presented in Table1.Total median follow-up time was 3.9 years (interquartile range 2.5 to 5.7 years).The primary end point ADT was reached in 62 patients (27%).

Table 2 .
Median time between CMR and ICD implantation was 56 days (interquartile range 8 to 92).No differences were observed in (indexed) LAVmin or LAVmax.However, patients who received ADT showed significant lower LAEF (p ¼ 0.003).In addition, a trend was observed toward a larger LVEDV in patients who received ADT (p ¼ 0.06), although LVEF did not differ.LGE data were available in 166 patients.Main reasons for missing LGE data were impaired renal function, LGE assessment not requested, or LGE assessment during a previous CMR assessment > 6 months before ICD implantation.Total scar size, quantified in grams, was significantly larger in patients who experienced ADT (n ¼ 41).Supplementary Table1lists a comparison of CMR characteristics between patients with ICMP and DCMP.Mean LAEF did not differ between both CMP groups.Patients with DCMP showed significant larger (indexed) LV volumes and lower LVEF compared with patients with ICMP (all p <0.001).LGE assessment was available in 104 of 140 patients (74%) with ICMP and 62 of 89 patients with DCMP (70%; p ¼ 0.45).LGE was more often observed and of greater extent in patients with ICMP.

Table 2
Cardiovascular magnetic resonance imaging characteristics according to appropriate device therapy Variable N(%), median (interquartile range), or mean AE standard deviation † ADT ¼ appropriate device therapy.* Data available for n ¼ 166.† Compared using the Mann-Whitney U test.Model 1: excluding scar size (total study population).Model 2: including scar size (subgroup with late gadolinium enhancement assessment available).CI ¼ confidence interval; HR ¼ hazard ratio.* Data available for n ¼ 178.† Data available for n ¼ 166.Model 1: excluding scar size (total study population).Model 2: including scar size (subgroup with late gadolinium enhancement assessment available).* Data available for n ¼ 178.† Data available for n ¼ 166.Arrhythmias and Conduction Disturbances/LAEF Predicts Ventricular Arrhythmias showed that although both LA volumes and LAEF were associated with heart failure hospitalization and mortality among 664 patients with heart failure, only LAEF remained predictive in