Improvement of Left Ventricular Function under Cardiac Resynchronization Therapy Goes along with a Reduced Incidence of Ventricular Arrhythmia

Objectives The beneficial effects of cardiac resynchronization therapy (CRT) are thought to result from favorable left ventricular (LV) reverse remodeling, however CRT is only successful in about 70% of patients. Whether response to CRT is associated with a decrease in ventricular arrhythmias (VA) is still discussed controversially. Therefore, we investigated the incidence of VA in CRT responders in comparison with non-responders. Methods In this nonrandomized, two-center, observational study patients with moderate-to-severe heart failure, LV ejection fraction (LVEF) ≤35%, and QRS duration >120 ms undergoing CRT were included. After 6 months patients were classified as CRT responders or non-responders. Incidence of VA was compared between both groups by Kaplan-Meier analysis and Cox regression analysis. ROC analysis was performed to determine the aptitude of LVEF cut-off values to predict VA. Results In total 126 consecutive patients (64±11years; 67%male) were included, 74 were classified as responders and 52 as non-responders. While the mean LVEF at baseline was comparable in both groups (25±7% vs. 24±8%; P = 0.4583) only the responder group showed an improvement of LVEF (36±6% vs. 24±7; p<0.0001) under CRT. In total in 56 patients VA were observed during a mean follow-up of 28±14 months, with CRT responders experiencing fewer VA than non-responders (35% vs. 58%, p<0.0061). Secondary preventive CRT implantation was associated with a higher likelihood of VA. As determined by ROC analysis an increase of LVEF by >7% was found to be a predictor of a significantly lower incidence of VA (AUC = 0.606). Conclusions Improvement of left ventricular function under cardiac resynchronization therapy goes along with a reduced incidence of ventricular arrhythmia.


Introduction
Cardiac resynchronization therapy (CRT) has become an integral component of systolic heart failure (HF) therapy. Up to date there is a bulk of evidence for the beneficial effect of CRT in patients with moderate-to-severe clinical impairment, which includes improved quality of life, fewer hospitalizations and decreased mortality [1]. The antiarrhythmic effects of CRT are well known [2][3][4]. However there are also reports of proarrhythmic effects [5][6][7]. Intriguingly a meta-analysis of large clinical trials, involving patients being implanted with CRT-devices with defibrillator backup (CRT-D), could not show a reduction in device interventions due to ventricular arrhythmias (VA) [8]. Also despite careful patient selection and elaborate efforts in postimplantation management CRT is only successful in about 70% of patients [9]. In patients responding to CRT left-ventricular reverse remodeling leads to a reduction of myocardial stretch and favorable neurohumoral changes [10,11]. Furthermore it can invoke profound changes on the (sub)cellular level [12,13]. Those effects might translate into a stabilizing effect on cardiac electrophysiology [14,15]. Besides reduction in left-ventricular end-systolic volume (LVESV), as the established marker for mechanical remodeling, improved left ventricular systolic function (LVEF) could be a functional and easily obtainable alternative, which might be less afflicted by inter-and intra-observer variability [16][17][18][19]. Importantly, previous studies demonstrated that among CRT responders there is a decrease in VA [20]. However, whether an improvement of LVEF among CRT patients goes along with a decreased incidence of VA has not been prospectively evaluated.
To address the impact of an improved LV function on the occurrence of ventricular arrhythmias we analyzed the incidence of VA in CRT responders in comparison with non-responders.

Patients
126 consecutive heart failure patients who underwent implantation with a combined CRT and cardioverter-defibrillator device (CRT-D) were included. Eligibility for CRT was based on 1) moderate-to-severe HF (NYHA functional class III or IV) despite optimal conventional therapy, 2) LVEF #35% as demonstrated by echocardiographic assessment and 3) left bundle branch pattern on the electrocardiogram (ECG) with a QRS duration .120 ms [21,22].
This study was conducted with approval of the ethics review committees (ERC) appointed by each centre (Heinrich-Heine-University Duesseldorf: ERC of the medical faculty, Building 13.41, Moorenstrasse 5, 40225 Duesseldorf; RWTH Aachen: ERC of the medical faculty, Pauwelstrasse 30, 52074 Aachen). Data collection for scientific purposes was covered under a written consent given by the subjects upon admission to a teaching hospital.

CRT implantation
CRT systems were implanted as previously described [21,23,24]. The LV pacing lead was inserted by a transvenous approach through the coronary sinus into either the lateral or posterolateral cardiac vein whenever possible. All patients who received CRT devices had biventricular stimulation of the heart with right ventricular leads.

Follow -up
3.1 Device evaluation. All patients were seen for outpatient clinic visits at 4 weeks post CRT implantation and in 3-month intervals after implantation. A physician contacted patients who failed to present to the prescheduled routine visits by phone. All follow-up visits included an interview, physical examination, an ECG, and a device interrogation. From the device printouts incidence and type of arrhythmias were determined by two trained electrophysiologists. Analysis of the printouts was blinded. Shocks or antitachycardic pacing (ATP) were classified as appropriate when they occurred in response to VT or VF and as inappropriate when triggered by sinus or supraventricular tachycardia, T-wave oversensing, or electrode dysfunction [25,26].
As described before the cycle length of the first ventricular arrhythmia triggering defibrillator (ICD) therapy and the average cut-off rate of the VT detection zone were standardized [27]. VT and VF zones were uniformly programmed with the same cut-offs for primary and secondary prevention patients. A VT-zone with primary application of ATP was programmed for CL from 450 ms to 300 ms, the VF-zone with immediate shock delivery was programmed for CL lower than 300 ms.
3.2 Clinical evaluation. Heart failure symptoms were classified using the New York Heart Association (NYHA) score. Resting 2-dimensional echocardiography was performed at baseline and 6 months follow-up [22]. Biplane LVEF assessment according to Simpson's rule was obtained for each patient in the apical 2-and 4-chamber view [28]. Volumetric parameters were acquired by biplane summation as reliably feasible in the individual [21]. Patients were classified as responders, based on an improvement in NYHA functional class by $1 and/or an improvement by $10% in LVEF 6 month after device implantation [29,30].

Statistical analysis
The primary end point was the occurrence of spontaneous sustained VT or VF leading to appropriate ICD intervention. Continuous data are expressed as mean 6 standard deviation (SD). Categorical data are presented by frequencies and percentages. Differences in baseline characteristics and 6-month follow-up were evaluated using unpaired Student t-test and Fisher's exact test as appropriate. Data within patient groups (to compare the effect of CRT) were compared by the use of paired Student t-tests. Survival curves were calculated and graphically presented using the Kaplan-Meier method for censored failure time data. Cox regression analysis was performed to evaluate associations between risk factors at enrollment and appropriate VT/VF therapy. Variable selection process was performed in two steps. Starting with the univariate analysis of the potential association of baseline findings with the occurrence of sustained VT or VF. This step was used as a model building process with factors showing a p-value of less or equal to 0.2 being used in the corresponding multivariate statistical model. The resulting multivariate Cox regression models were stratified by responders, non-responders and ICM or DCM respectively. In this last step, factors were assessed as significant, if the p-value was less than 0.05. ROC analysis was conducted after prior testing by multivariate analysis. All analyses were performed using SASH statistical software, V9.1.3 (SAS Institute, Cary, NC, USA).

Patients
The baseline characteristics of patients are shown in Table 1. There were no differences in clinical baseline characteristics between patients found to be responders or non-responders except for a slightly higher number of patients with secondary preventive implantation in the responder group. The cause of heart failure was ischemic in 65 patients (52%) and non-ischemic in 61 patients (48%).

CRT response
Mean LVEF increased within 6 month after CRT device implantation from 2567% to 3168% (p,0.001), with a reduction in LVESV from 174673 ml to 138669 ml (p,0.001). There were 74 (59%) patients, among our collective of 126 individuals, fulfilling the criteria for positive CRT response as established above (Table 1). In the responder group the mean LVEF improved from 2567% to 3666% (p,0.001), while in the nonresponder group no significant increase occurred ( Figure 1). Furthermore, in the responder group a decrease in LVESV from 176672 ml to 122668 ml (p,0.001) was noted, while in the nonresponder group no significant decrease in LVESV (baseline: 169673, follow-up: 162669; p = 0.521) occurred. This also applied for the development of NYHA classes. The baseline NYHA classes were matched between responders and nonresponders. At the 6-month interval there was a statistically significant association of CRT-response with development of the NYHA class. In the responder group there was a significant shift of patients from NYHA class IV and III to lower classes ( Table 2).

Follow-up of responders vs. non-responders
The mean duration of follow-up was 28614 months (range 9 to 48 month). There were 10 (8%) deaths in our whole collective. Of those 3 occurred in the responder group and were all due to noncardiac causes, the remaining 7 deaths occurred in the nonresponder group and were all attributed to worsening heart failure. There were 8 hospitalizations due to worsening heart failure among 7 Patients in the responder group (11%) and 43 such hospitalizations among 22 patients in the non-responder group (42%). There were no heart transplantations in the responder group whereas there were 2 transplantations among the nonresponders.

Incidence and therapy of ventricular arrhythmias
Patients in the responder group received less appropriate ICD therapy than patients in the non-responder group (35% vs. 58%, log-rank X 2 = 7.5, p,0.0061, Figure 2A). At the end of follow-up, the time between implant and first appropriate ICD therapy was comparable in both groups (19613 months vs. 15614 months, p = 0.09). Furthermore, the cycle length of the first ventricular arrhythmia triggering ICD therapy was the same in both groups (responders: 301648 ms; non-responders: 290654 ms, p = 0.43). The likelihood of arrhythmia events decreased with increasing LVEF during follow-up (40% vs. 58%, log-rank X 2 = 5.7, p,0.0168). By employing ROC analysis an EF increase by .7% during CRT was determined as a significant cut-off value ( Figure 3) to distinguish individuals with a lower risk of VA occurrence ( Figure 2B). The group of patients with secondary preventive ICD-implantation had greater quota of individuals receiving appropriate ICD therapies (n = 20, 71%) than the group of patients with primary preventive indication (n = 36, 37%, p = 0.0002). There was no significant difference in the incidence of appropriate ICD therapies between in patients with ischemic or non-ischemic heart failure.
We also conducted a subgroup analysis comparing nonresponders with first VA event in the initial 6 months of CRT (early events) against with first event beyond this period (late events). In this analysis patients with early events had a lesser increase in LVEF during the first 6 months (11.7627.9% vs. 5.9624.1%), although the number of patients in those groups was small (n = 28 vs. 24) and the effect was statistically not significant (p = 0.435).

Predictors of ICD therapy
Cox Regression analysis of clinical baseline parameters indicates that patients on therapy with an ACE inhibitor or an angiotensin receptor antagonist, as well as females a significantly have a reduced risk of VA (HR = 0.24-0.47, p,0.05). Also response to CRT therapy by 6 months after implantation is associated with an reduced risk for VA (HR = 0.53, p,0.05). Secondary preventive indication for device implantation is associated with a significantly higher risk of such events (HR = 3.13, p,0.05) ( Table 3).

Inappropriate ICD therapy
Twelve patients (9.5%) experienced inappropriate ICD shocks. The trigger for inappropriate therapy was atrial arrhythmia in 3 patients, sinus tachycardia in 1 patient, T-wave oversensing in 7 patients, and sensing of diaphragm potentials in 1 patient. Of these 12 patients 7 belonged to the responder group and 5 to the nonresponder group, representing 9.5 respectively 9.6% of each group (p = 1.0).

Discussion
The key findings of this study are: (1.) the incidence of VA is reduced in patients responding to CRT during long-term followup. (2.) An increase of LVEF during the first 6 months after commencement of therapy goes along with a decreased incidence VA.
Although there is some evidence favoring antiarrhythmic effects of CRT [4,31,32], it's impact on arrhythmia susceptibility is not fully understood with multiple studies providing differing outcomes. Some studies reported a decrease of the number of VA after CRT [33][34][35]; however others reported the opposite [7,36]. One meta-analysis of large randomized CRT trials found no significant effect of CRT on sustained VA occurrence compared with ICD therapy only [8]. Additionally, a recent meta-analysis including data from the extension phase of Cardiac Resynchronization-Heart Failure (CARE-HF) provided evidence against a benefit of CRT alone on risk of sudden cardiac death (SCD) [37].
In a recent study Lin et al. [38] could not determine any benefit of CRT on the occurrence of VA in a collective of patients undergoing an upgrade from ICD to CRT-D. This might be due to the relatively small sample size and more importantly to the preselection of the patients. All patients included in this study were previously implanted with an ICD, demonstrating a markedly higher risk for VA. Our patient collective in contrast included patients on an ''all comers'' basis and thus was more heterogenous.
The time course of VA incidence in our study appears to be in line with reports from other investigators, delineating a period   (29) There was as a significant shift of patients from NYHA class IV and III to lower classes in patients responding to CRT. Abbreviations: LVEF = left ventricular ejection fraction; CRT = cardiac resynchronization therapy; NYHA class = New York Heart Association functional heart failure classification; * = p,0.0001 for comparison against baseline. doi:10.1371/journal.pone.0048926.t002 from 3 to 12 months after implantation in which electrical reverse remodeling seems to occur [39,40]. This seems to apply to responders as well as to non-responders, implying that even in non-responders there seems to be some consequence of CRT -just not to the extent as in responders. We emphasize that the population of our uncontrolled nonrandomized observational study is smaller than that included in recent hallmark trials looking for benefit from CRT responding. However, our study design pursued a different aim and was conducted to investigate the incidence of VA in association to improved cardiac function in daily clinical practice. Contradictory to evidence from previous studies [41], in the presented COX regression analysis of baseline patient characteristics treatment with beta-blockers had no significant effect on VA incidence (HR = 0.657, p = 0.431), This might be in part due to the overpowering impact of other factors in this statistical model.   Table 3. Cox Regression analysis of the correlation of selected patient characteristics with the incidence of ventricular arrhythmias (VA) during cardiac resynchronization therapy (CRT). In addition, relatively low beta-blocker dosing, due to adverse effects [31], might be of importance but remains speculative. Also we chose a definition of reverse remodeling that diverged in some points from the definition used in other current studies. This might impede the comparability of our results.
For the assessment of functional myocardial remodeling under CRT we monitored the development of LVEF during patient follow-up visits. Although LVESV has been shown to be a good indicator of reverse remodeling in previous publications, LVEF was chosen as it is well established and can be measured with comparably low intra-and inter-observer variation during routine echocardiography [42]. Nevertheless this might limit comparability of the presented data.
Defining a positive response to CRT is still crucial. Defining a response as .10% increase of LVEF and/or an improvement of NYHA functional class by $1 may limit the number of patients fulfilling these criteria. Also there exist more elaborate classifications of HF, which might aid in the better discrimination of HF stages (e.g. Metra et al. [43]).
However this cut-off is well established in the literature and the observed responder rate of about 70% in our patient population with an LVEF improvement by 7% is in line with various reports [44,45]. In another recent publication Ypenburg et al. nicely demonstrated that a response to CRT, defined as an improvement in NYHA functional class, was associated with a lower risk of receiving ICD therapy [37,46].
Herein, we provide evidence during a longer follow-up that improved systolic LV function is associated with a reduced incidence of VA.
Importantly, a possible explanation for the relatively high number of patients with adequate ICD interventions might be early detection and first-line treatment of VA with ATP that may be in fact nsVT. Therefore, ICD therapy would not be a reliable surrogate for SCD, because some VA would likely have terminated spontaneously in the absence of the ICD [7]. However, to determine the risk of sudden cardiac death was beyond the scope of the present study.

Conclusion
The present findings suggest that improvement of LVEF under CRT in patients with moderate-to-severe heart failure and left bundle branch block, apart from hemodynamic benefits, goes along with a decreased incidence of ventricular arrhythmias.

Supporting Information
Supplement S1 Table containing patient characteristics at time of implantation (physiological parameters, prevalence and discrimination of AF types, mitral regurgitation, conduction disorders, indication for ICD therapy and history of ventricular tachycardia or fibrillation before implantation), for responders and non-responders (p-values are given for comparison of both.