Comparison between endocardial and epicardial cardiac resynchronization in an experimental model of non-ischaemic cardiomyopathy

Abstract

Aims

Pacing from the left ventricular (LV) endocardium might increase the likelihood of response to cardiac resynchronization therapy. However, experimental and clinical data supporting this assumption are limited and controversial. The aim of this study was to compare the acute response of biventricular pacing from the LV epicardium and endocardium in a swine non-ischaemic cardiomyopathy (NICM) model of dyssynchrony.

Methods and results

A NICM was induced in six swine by 3 weeks of rapid ventricular pacing. Biventricular stimulation was performed from 16 paired locations in the LV (8 epicardial and 8 endocardial) with two different atrioventricular (80 and 110 ms) intervals and three interventricular (0, +30, −30 ms) delays. The acute response of the aortic blood flow, LV and right ventricular (RV) pressures, LVdP/dt_max and LVdP/dt_min and QRS complex width and QT duration induced by biventricular stimulation were analysed. The haemodynamic and electrical beneficial responses to either LV endocardial or epicardial biventricular pacing were similar (ΔLVdP/dt_max: +7.8 ± 2.2% ENDO vs. +7.3 ± 1.5% EPI, and ΔQRS width: −16.8 ± 1.3% ENDO vs. −17.1 ± 1.9% EPI; P = ns). Pacing from LV basal regions either from the epicardium or endocardium produced better haemodynamic responses as compared with mid or apical LV regions (P < 0.05). The LV regions producing the maximum QRS complex shortening did not correspond to those inducing the best haemodynamic responses (EPI: r² = 0.013, P = ns; ENDO: r² = 0.002, P = ns).

Conclusion

Endocardial LV pacing induced similar haemodynamic changes than pacing from the epicardium. The response to endocardial LV pacing is region dependent as observed in epicardial pacing.

Introduction

Cardiac resynchronization therapy (CRT) is indicated in patients with chronic heart failure, who present ventricular dyssynchrony and QRS complex widening.¹ Left ventricular (LV) pacing is generally performed at the epicardium via the coronary venous system. However, the accessibility and distribution of the coronary veins may limit the success of the technique. Left ventricular endocardial pacing might allow a rapid engagement of the Purkinje network resulting in a faster activation of the LV and a more efficient resynchronization.² However, endocardial LV electrode implantation can lead to specific complications.³

The potential superiority of LV endocardial pacing has not been demonstrated. Most of the scientific evidence that shows advantages of endocardial LV pacing comes from canine models.^4–7 Strikingly, the only study that used a sheep animal model did not find any difference between endo and epi LV pacing.⁸ Evidence in humans has been collected in a recently published review and metaanalysis.⁹ This study showed that the response rates to endo pacing were heterogeneous but not superior to epi LV pacing. The authors also acknowledge important limitations in the analysed studies.

To get closer to the clinical scenario of CRT in patients with advanced cardiac disease, we used a swine model of pacing-induced non-ischaemic cardiomyopathy (NICM) that has never been used before in this setting. Dilated cardiomyopathic human hearts are characterized by increased epicardial eccentricity and decreased wall thickness, and this has been related to a reduced gradient of conduction velocities between the endocardium and epicardium.² The swine model is characterized by an homogeneous transmural conduction velocity¹⁰ and, therefore, could mimic more accurately the condition of these patients. We aimed to compare the haemodynamic and electrical response of endo and epi LV pacing in CRT in this animal model of NICM.

Methods

Study population

Six female domestic swine (Landrace-Large White cross, body weight 37 ± 1 kg) were submitted to two interventions.

First intervention: induction of non-ischaemic cardiomyopathy

Animals were premedicated with a combination of midazolam and ketamine, anesthetized with propofol, intubated and mechanically ventilated under general anaesthesia with a mixture of oxygen and sevofluorane. Remifentanil was administered during the procedure for analgesia. A conventional 12-lead ECG and a 2-D echocardiography were recorded. A bipolar pacing electrode (1888TC/58 cm, St. Jude Medical Inc., Sylmar, CA, USA) was introduced via the jugular vein to the right ventricular (RV) apex for continuous ventricular stimulation. The electrode was connected to a custom-modified pacemaker generator (St. Jude Medical Inc., Sylmar, CA, USA) that was implanted subcutaneously at the pig’s neck. One week after the index procedure, the pacemaker was turned on and rapid RV pacing at 200 beats per minute was maintained for 3 weeks.

Second intervention: comparison of left ventricular endocardial vs. left ventricular epicardial biventricular pacing

Experimental preparation

Animals were premedicated, anesthetized, and maintained as described in the first intervention. The femoral vein and artery were cannulated and two catheters (Millar Instruments, Inc., Houston, TX, USA) were placed into the RV and the LV for continuous monitoring of the LV and RV pressures. A mid thoracotomy was performed and an ultrasonic flow probe (Transonic Systems Inc., Ithaca, NY, USA) was deployed around the aortic root to measure the aortic blood flow (ABF). A plunge electrode was inserted into the right atrium for subsequent atrial pacing. RV pacing was performed from the previously implanted RV electrode. Endocardial LV pacing was performed by means of a standard tetrapolar steerable electrophysiology catheter (Blazer II, Boston Scientific Corp., Natick, MA, USA) introduced through a retro-aortic approach. A plunge electrode was used to pace the LV epicardium. The atrial, RV, and LV pacing electrodes were connected to a CRT device (Promote Accel™ CD3215-36, St. Jude Medical Inc., Sylmar, CA, USA) (Figure 1). The ECG and the LV and RV intra-cavitary pressure probes were connected to a Nihon Kohden amplifier (RMP-6004, Nihon Kohden Corporation, Tokyo, Japan). The ABF probe was connected to a Transonic Perivascular Flowmeter (T206, Transonic Systems Inc., Ithaca, NY, USA). Both amplifiers were connected to a PowerLab 8/35 (ADInstruments, CO, USA), for continuous data storage and signals were digitized at a rate of 1 kHz for subsequent off-line analysis.

Study variables

Left ventricular ejection fraction (LVEF), end-systolic (ESLVD), and end-diastolic (EDLVD) LV diameters were measured from the echocardiography performed at the beginning of both interventions to verify the induction of a successful NICM model. Additionally, the heart rhythm, RR interval, QRS width, and QT interval in the ECG were also evaluated. We analysed the differential effect of endocardial and epicardial pacing on the following variables at each pacing configuration: LV peak pressure (LVP), LV dP/dt_max, LV dP/dt_min, RV peak pressure (RVP), mean ABF, as well as QRS complex width and QT interval. Custom Matlab R2013a scripts (Mathworks, Inc., Natick, MA, USA) and LabChart 7 Pro software (ADInstruments, CO, USA) were used for signal analysis. The values obtained for each biventricular pacing configuration were compared with those obtained during previous dyssynchronous RV DDD pacing at the same AV delay. These differences were expressed as percentage of change using the formula: 100 × [(value of variable X/dyssynchronous value of variable X) − 1], obtaining the new study variables: ΔLVP, ΔLVdP/dt_max, ΔLVdP/dt_min, ΔRVP, ΔABF, ΔQRS and ΔQT. The AV and VV intervals producing the best local haemodynamic or electrical response at each site per animal was then selected for further analysis. The best haemodynamic response was defined as the maximum improvement in LVdP/dt_max (positive ΔLVdP/dt_max) and the best electrical response was defined as the maximum shortening of the QRS (negative ΔQRS width). Moreover, the sites inducing a haemodynamic improvement (ΔLVdP/dt_max > 0%) and a super-response (ΔLVdP/dt_max >10%) were also assessed.

Protocol

Prior to the beginning of the study protocol, RV pacing was discontinued for at least 30 min. A 2-D echocardiography and 12-lead ECG were recorded verifying a successful induction of the NICM. During the study, the atrium was paced at a constant rate 10% over the intrinsic sinus rhythm. The biventricular pacing protocol (RV + LV pacing) was performed at the fixed location of the RV, and at eight pre-established regions of the LV, selected in a random order. The following LV regions were evaluated: basal-anterior, basal-lateral, basal-posterior, mid-anterior, mid-lateral, mid-posterior, apical-anterior, apical-lateral. We performed endocardial and epicardial pacing in each LV region site, thus resulting in a total of 16 LV pacing sites per animal. The endocardial catheter was positioned opposite to the epicardial plunge electrode with the help of orthogonal fluoroscopic projections to ensure regional concordance between epicardial and endocardial pacing sites. Biventricular pacing was performed at two different atrioventricular (AV) delays (80 and 110 ms) and at three different interventricular (VV) intervals (VV = 0: simultaneous pacing of both ventricles; VV = +30 ms: LV paced after the RV, and VV = −30 ms: LV paced before the RV). All measurements were performed at each site after thirty seconds of pacing at each configuration. Although the underlying tachycardia-induced cardiomyopathy surely promoted certain degree of dyssynchrony, continuous RV DDD pacing was used to maximize the dyssynchronic contraction of the heart. Therefore, RV DDD pacing was performed at each location before the beginning of the biventricular pacing protocol and used as a reference for comparison. At the end of the second intervention, animals were euthanized and their hearts removed to obtain myocardial samples for histopathologic studies. The specimens were fixed in 10% buffered formalin, embedded in paraffin and sliced (4-µm thickness). After this, the samples were stained with Masson’s trichrome and analysed by an anatomopathologist.

The study protocol was approved by the Animal Care and Use Committee of our Institution, in accordance with the regulation for the treatment of animals established by the Guide for the Care and Use of Laboratory Animals (Eighth edition, National Research Council, Washington DC, The National Academies Press 2010).

Statistical analysis

Results are expressed as mean ± standard error of the mean (SEM). The statistical differences in the study variables were assessed by means of t-tests, χ² tests, Pearson’s correlation and analysis of variance (ANOVA) with Bonferroni correction for post-hoc comparison, as appropriate. A P-value <0.05 was considered statistically significant. All analyses were performed using SPSS v. 22.0 (IBM SPSS Inc., Chicago, IL, USA).

Results

Non-ischaemic cardiomyopathy model

Animals underwent 24 ± 2 days of RV pacing. After discontinuing RV pacing all animals were in sinus rhythm and showed LV dilatation and heart dysfunction with marked QRS widening associated to a certain degree of non-specific intraventricular conduction defect, as shown in Table 1. An increase of the voltage and duration of the p waves was also observed. The successful induction of the model was also confirmed post-mortem based on the histopathological examination of LV sections. Haemodynamic parameters of NICM animals before the pacing protocol were: LVP = 71 ± 4 mmHg; LV end diastolic pressure = 12 ± 3 mmHg; LVdP/dt_max=567 ± 148 mmHg/s; LVdP/dt_min = −687 ± 52 mmHg/s; and ABF = 2 ± 1 L/min.

Endocardial vs. epicardial biventricular pacing

A total of 576 endocardial and epicardial pacing sites were analysed in the 6 studied animals. Baseline conditions (RV DDD pacing) remained stable during the entire procedure in all the studied variables (data not shown). Both endocardial and epicardial LV biventricular pacing were associated with an improvement in the haemodynamic and electrical parameters when compared with dyssynchronous RV DDD pacing (ΔLVdP/dt_max: +7.8 ± 2.2% and +7.3 ± 1.5% for endo and epi pacing respectively, P < 0.001; QRS duration: −16.8 ± 1.3% and −17.1 ± 1.9% for endo and epi pacing respectively, P < 0.001). Figure 2 shows the haemodynamic and electrical response of biventricular pacing from the epicardium and the endocardium of the LV. Regarding haemodynamic response, similar aortic blood flow and cardiac LV contractility (LVdP/dt_max and LVdP/dt_min) responses after biventricular pacing were observed in endocardial and epicardial pacing. Likewise, QRS shortening after biventricular pacing was similar. Moreover, the percentage of sites inducing a haemodynamic improvement (ΔLVdP/dt_max > 0%) was not different between endocardial and epicardial pacing (56% of endocardial and 59% of epicardial pacing sites). Furthermore, a super-response (ΔLVdP/dt_max > 10%) was present in 17% of endocardial and at 15% of epicardial pacing sites (P = ns).

Additionally, we analysed LV sites with the best and less favourable response to stimulation (Table 2). The best haemodynamic response was obtained at the epicardium in three animals and at the endocardium in the other three. The less favourable response was epicardial in two cases and endocardial in the remaining four. Based on the electrical response, the pacing site inducing the narrowest QRS complex was found to be epicardial in four animals and endocardial in the other two. No correlation between the haemodynamic response during LV pacing at a specific site and the degree of QRS shortening was found (EPI: r² = 0.013, P = ns; ENDO: r² = 0.002, P = ns).

Regional differences in LV pacing

The left ventricular basal pacing locations induced the highest ΔLVdP/dt_max improvement as compared with the mid and apical regions, both in endocardium and epicardium (Table 3, Figure 3). Likewise, a higher percentage of pacing sites associated to a haemodynamic benefit was noted to the basal region of the LV (EPI: 72% vs. 52% vs. 49% for basal, mid and apical pacing sites respectively, P < 0.01; ENDO: 81% vs. 39% vs. 42% for basal, mid, and apical pacing sites respectively, P < 0.001). No statistical differences were observed when anterior, lateral, and posterior sites for each region were compared, neither endocardial nor epicardial (Figure 3).

No regional differences were found with respect to the QRS shortening (Table 3).

Cardiac resynchronization therapy device configuration

The CRT device configurations that induced the best haemodynamic response were similar between endocardial and epicardial LV pacing sites (Table 4). No specific atrioventricular delay (AV) was found to induce better responses to biventricular pacing.

Discussion

Main findings

The main finding of the study is that endocardial biventricular pacing has similar acute haemodynamic effects than pacing at the epicardium in a NICM swine model. LV endocardial and epicardial biventricular pacing induced the best haemodynamic response at the basal regions, as compared with mid or apical positions. A lack of correlation between haemodynamic and ECG responses to CRT was also found.

Endocardial vs. epicardial biventricular pacing

Our results are in concordance with those reported by Dzemali et al.⁸ in a sheep model of NICM. These authors observed that both endocardial and epicardial pacing of the lateral wall of the LV led to an improvement in LV function, with no differences between endocardium and epicardium in any of the haemodynamic and echocardiographic parameters analysed. In the present study we used a swine animal model of dilated NICM and further evaluate all LV epicardial–endocardial representative regions, recording multiple haemodynamic (aortic blood flow, intraventricular pressures) and electrocardiographic parameters. Furthermore, the use of multiple device programming configurations in our study design allowed us the full evaluation of all LV pacing combinations, resulting in 96 endocardial/epicardial paced paired-sites per animal.

In contrast, various studies have demonstrated the advantages of LV endocardial pacing over epicardial pacing in canine models of isolated acute left bundle branch block (LBBB) and heart failure plus acute LBBB.^4–7 These studies showed that endocardial biventricular pacing produced a more homogeneous and rapid ventricular electrical depolarization and repolarization, associated to an additional marked improvement in acute systolic LV pump function.

The large multicentre prospective human study ALternate Site Cardiac ReSYNChronization (ALSYNC) included patients who had failed or were unsuitable for conventional CRT. This study demonstrated the clinical feasibility of endocardial pacing in CRT and, although it was not a comparative study, that the rate of responders to endocardial pacing was similar to the rate of responders to classical CRT.¹¹ A recently published review and meta-analysis showed that the response rates to endo pacing were heterogeneous but not superior to epicardial LV pacing. The authors also pointed out that there is a large inconsistence between the results of small and large studies, probably due to publication bias.⁹

It can be hypothesized that the discrepancies observed between the different animal and human studies could be related to differences in the species-specific intrinsic characteristics of the Purkinje system and its functional remodelling under myocardial disease. The distribution of the Purkinje system in dogs is similar to that of healthy humans, being circumscribed to the endocardium,^10,12 resulting in faster conduction velocities in the endocardium than in the epicardium.² On the contrary, sheep and swine have a transmural Purkinje system^12,13 which accelerates the transmural conduction and explains similar conduction velocities in the endocardium and the epicardium. These differences between dogs, sheep and swine would explain the discordant results of the animal studies evaluating endocardial–epicardial pacing in CRT. Studies in humans with heart failure (HF) and advanced myocardial disease have demonstrated damage of the distal Purkinje network and conduction remodeling.¹⁴ Dilated cardiomyopathic human hearts are characterized by increased epicardial eccentricity and decreased wall thickness, shortening the transmural path length. This has been related to a reduced gradient of conduction velocities between the endocardium and epicardium.² In this scenario, the pig model of NIDCM used in this study could mimic more accurately what happens in the diseased human heart.

Relation between haemodynamic and electrical changes

Our data show a lack of correlation between haemodynamic and ECG response to CRT. This absence of haemodynamic–ECG correlation occurred both in epicardial as well as to endocardial pacing. To the extent of our knowledge, studies correlating pacing locations with acute haemodynamic response and ECG changes are scarce in the literature and their results are limited. Spotnitz et al.¹⁵ tested only two epicardial LV regions at different VV intervals in an intraoperative study performed in a limited number of patients. These authors found a negative correlation (improved haemodynamics with shorter QRS) only in five of seven patients, although the best haemodynamic response did not correspond to the shortest QRS. Interestingly, the remaining two patients had a positive (paradoxical) haemodynamic/ECG correlation. Derval et al.¹⁶ conducted a larger acute human study examining eleven LV sites during exclusively DDD LV pacing (biventricular pacing was not performed). They found that QRS shortening correlated with acute haemodynamic response and that LV pacing location was a primary determinant of paced QRS duration. However, QRS shortening did not predict the maximum haemodynamic response. Porta-Sanchez et al. have recently shown that the benefit of LV epicardial pacing in human cardiomyopathic hearts could be achieved without abbreviating the QRS duration.¹⁷

Regional differences in left ventricular pacing

The response of epicardial and endocardial LV pacing was regional dependent and the best response was obtained at the basal regions. Our observation is in agreement with previous experimental studies.^6,8 The study by Dzemali et al.⁸ found that stimulation of the left ventricle at the basal lateral wall in NICM sheep resulted in a significant increase of the cardiac output (CO). Similarly, Bordachar et al.⁶ concluded that the optimal biventricular pacing site (defined as the max % increase of dP/dt_max) in dogs with HF and LBBB was basal for both endocardial and epicardial pacing. Our study extends their results by evaluating all endocardial and epicardial LV representative regions, and corroborates that the region dependency present in epicardial pacing is also observed in endocardial pacing. In addition, these results fully agree with long term studies in humans that report worse CRT outcomes associated to pacing from apical regions.^18,19

Device programming configuration

We found a specific device programming configuration (AV and VV interval) for each site and animal that induced the best haemodynamic response to biventricular pacing. These results are comparable with clinical observations stressing the need for an individual optimization of the device pacing parameters.²⁰

Clinical implications

Left ventricular pacing in CRT is generally performed epicardially via the coronary venous system. This study suggests non-inferiority of endocardial versus epicardial LV pacing, and therefore endorses the clinical strategy of endocardial pacing when conventional epicardial pacing is not feasible or for non-responders.

Limitations

Some limitations of our study should be noted. The LV dilatation and dys-synchrony model did not generate a definitive LBBB pattern but a certain degree of non-specific intraventricular conduction defect. Thus, to maximize dys-synchrony, the experiments were performed during continuous RV DDD pacing. We did not collect data on regional electrical delays due to technical issues related to the experimental setting. However, our pacing results were regional dependent and the best response was obtained at the basal regions, suggesting that the electrical properties of the model are comparable to those of human diseased hearts. We only analysed the acute haemodynamic and ECG effects of biventricular endocardial and epicardial pacing. Whether acute results correlate to the long-term response to CRT needs further investigation. Left ventricular dP/dt_max was used as a measurement of myocardial contractility as in most experimental studies. It is also the standard contractility parameter used in the clinical setting. Additional parameters of contractility like those based on pressure volume loops were not assessed since LV volumes were not measured. The right ventricular pacing electrode in this study was positioned in the apex. Our findings regarding segmental LV lead position may differ from other right ventricular pacing sites (i.e. mid/high septum or outflow tract). However, this should not significantly alter the comparison between LV endocardial and epicardial pacing that was the main aim of the study. Additionally, the use of an electrophysiological endocardial catheter and a plunge epicardial electrode, and the methodology for assuring the correspondence of the endocardial and epicardial sites may arise some concerns. The plunge electrode was minimally inserted in the epicardium to obtain pure epicardial regional capture. The radiological approach to position the endocardial catheter facing the plunge electrode and the inter-electrode separation of the catheter seems accurate enough as to allow the capture of the same segment of the heart, allowing direct comparison of paired pacing segments.

Conclusions

The study shows that endocardial pacing has similar haemodynamic and electrocardiographic benefit than pacing from the epicardium in a dilated NICM swine model. The response to endocardial LV pacing is region dependent as observed in epicardial pacing. In addition, no correlation between haemodynamic and ECG response to CRT was found.

Funding

This work was supported by a grant from the Spanish Ministerio de Economía y Competitividad, Instituto de Salud Carlos III [FIS-PI10/01149] and Fondo Europeo de Desarollo Regional (FEDER).

Conflict of interest: none declared.

References

Author notes