Vagal stimulation in heart failure

The vagus nerve (VN) is a mixed nerve, composed of approximately 80% afferent and 20% efferent axonal projections. Efferent fibers originate in the dorsal motor nucleus and in the nucleus ambiguous located in the brainstem, and project to the heart, lungs, larynx, pharynx, stomach, spleen, pancreas, liver, intestines, and ovaries. At the cardiac level, it is now well established that efferent vagal fibers do not directly innervate cardiomyocyte as previously believed, but relay on neurons located in epicardial fat pads, which constitute a complex neuronal network better known as the “intrinsic cardiac nervous system” (ICNS; [6]). The human ICNS contains at least 10 major groupings of ganglionated plexuses, each one providing a preferential, albeit not exclusive, and largely overlapping distribution to different cardiac regions. The right and the left VN do not share exactly the same distribution on the plexuses of the ICNS; as a result, the right VN has a greater influence on the sinus node activity, whereas the left VN has a predominant control over the atrioventricular node function.

From a structural point of view, the VN contains a mixture of different types of nerve fibers, which are organized into bundles (fascicles). Unfortunately, despite decades of extensive research, it is still a matter of debate whether fascicles into the VN are arranged according to the type of fiber (afferent or efferent) they contain or according to their peripheral end-organ distribution (namely, somatotopic arrangement). The available evidence, however, favors the latter possibility [7]. Neuronal fibers within the VN are classified according to their diameter and their conduction velocity, with Aα fibers being the largest and fastest, unmyelinated C fibers the smallest and slowest, and Aβ, Aγ, Aδ, and B fibers intermediate. Cardiac vagal control in mammalians is mediated by B- (efferent) and C‑type fibers (afferent and efferent). Notably, VNS recruits neuronal fibers based on proximity to the stimulation electrode, the local electric field strength, and inversely related to size—with A‑type fibers being recruited first and C‑type fibers last.

Vagal nerve stimulation was first studied as an antiarrhythmic intervention. The first report dates back to 1859 [8], when Einbrodt incidentally observed that dogs receiving VNS were less likely to die from direct delivery of electrical current to the ventricles. More than 100 years later, between 1973 and 1978, several studies on anesthetized animals [9‐12] confirmed that VNS reduces the risk of ventricular fibrillation (VF) during acute myocardial ischemia. In the study by Myers, the protective effect on VF was not abolished by preventing heart rate (HR) decrease and was only mildly reduced by VN decentralization, therefore suggesting a direct efferent effect of VNS at the ventricular level [12]. A definite mechanistic demonstration came in 1991, when a conscious canine model of sudden death was used to prove that cholinergic antagonism favors VF, while right VNS applied a few seconds after coronary occlusion protects against VF; almost 50% of this protective effect was related to the significant HR lowering achieved during right VNS [13, 14].

During the same period, we also provided the first evidence of the protective and only partially HR-dependent effect of VNS on reperfusion arrhythmias [15]. The mechanism underlying this effect was only recently unraveled, together with the understanding that reperfusion arrhythmias share common pathways with ischemia/reperfusion injury. In both settings, VNS exerts its protective effects [16‐18] by complex intracellular pathways including activation, through Gi-coupled muscarinic receptors, of phosphatidylinositol 3‑kinase and Akt. These are the same pathways implicated in the protective effect of ischemic preconditioning [19]. Finally, VNS leads to the upregulation of the anti-apoptotic protein BCL‑2 and to the suppression of caspase‑3 in cardiomyocytes, thus protecting from cell death [20]. Notably, an increase in cardiac efferent vagal tone is also likely to mediate the beneficial effect of remote preconditioning [21, 22].

Studies of VNS on inflammation also produced interesting findings. In 2002, Tracy [23] provided conclusive evidence supporting the so-called cholinergic anti-inflammatory pathway, a neural mechanism that inhibits pro-inflammatory cytokine release via signals that require α7 cholinergic receptor expression on macrophages and other immunocompetent cells. In 2011 Calvillo et al. [24] demonstrated the cardiac effects of this pathway, underlying the crucial role played by nicotinic receptors in this HR-independent anti-inflammatory and anti-apoptotic effect of VNS, which produced a marked infarct size reduction following ischemia/reperfusion in rats.

Finally, the results of the first experimental study showing the benefit of chronic VNS for the treatment of HF were published in 2004 [25]. At 14 days after a large anterior myocardial infarction (MI) leading to HFrEF, rats were randomized to sham stimulation or active VNS (10 s on, 50 s off), at 20 Hz, with 0.2-ms pulses. The intensity of VNS was adjusted to reduce HR by 20 to 30 bpm from a starting value of 360 bpm. After 6 weeks of VNS, a duration mostly determined by the generator’s life, the treated animals showed a significantly better left ventricular (LV) function, a lower normalized biventricular weight, lower norepinephrine and B‑type natriuretic peptide (BNP) levels, and a strikingly better survival (86% vs. 50%) at 140 days compared to sham-operated animals, despite a similar infarct size. Notably, the lack of difference in the infarct size was expected due to the relatively long time elapsed between coronary artery ligation and the beginning of VNS.

Subsequently, between 2005 and 2013, the group of Hani Sabbah conducted several studies assessing the effects of right VNS on a canine model of chronic HF induced by coronary microembolizations [26, 27]. In the first two studies [26], the authors used the CardioFit system (see below) to deliver VNS. This system includes an intracardiac sensing lead aimed to synchronize VNS to the cardiac cycle and to continuously modulate VNS intensity based on the extent of acute HR reduction achieved during stimulation, which was set at 10% of the baseline HR. In the first study, 3 months of VNS, compared with sham operation, improved LV hemodynamics, decreased tumor necrosis factor (TNF)-α and interleukin (IL)-6 levels, tended to normalize nitric oxide synthase (NOS) expression, and dramatically increased the expression of Cx43, a gap-junction-forming protein whose reduction is associated with conduction blocks and increased arrhythmic susceptibility. Notably, the histological comparison between the right and the left VN performed at the end of the study revealed normal right VN fascicular and cellular structure. In the second study, the beneficial effects of a 3-month combination therapy of VNS and metoprolol on LV hemodynamics were proven to be additional to those achieved with metoprolol alone. Finally, in the third study [27], the group used a different VNS system (Boston Scientific Corporation) specifically set with a low stimulation intensity, unable to affect HR. Nonetheless, the beneficial effects of VNS on LV function were still evident, together with an improvement in NT-proBNP, pro-ANP, TNF‑α, IL‑6, BCL‑2, caspase‑3, Cx43, and the three isoforms of NOS.

In 2009, Zhang et al. [28] studied the effects of chronic VNS in a different model of HF, induced by high-rate ventricular pacing. Dogs in the active group received 8 weeks of high-rate ventricular pacing with concomitant VNS (20 Hz, pulse width 0.5 ms, duty cycle 14 s on and 12 s off), while the sham group only received pacing. The intensity of VNS was individually set before the start of pacing to reduce sinus rate by approximately 20%. Again, although HR was kept constant by pacing, VNS elicited significant benefits in LV hemodynamics, C‑reactive protein, norepinephrine and angiotensin II levels, HR variability, and baroreflex sensitivity compared to controls.

Notably, despite several demonstrations of the beneficial effects of VNS in post-MI animal models, data on timing, spatial distribution, and functional consequences of parasympathetic remodeling occurring after MI have long been lacking. Very recently, Vaseghi et al. [29] demonstrated that, as opposed to norepinephrine levels, cardiac acetylcholine levels are preserved 6–8 weeks after MI in border zones and in viable myocardium of infarcted hearts, supporting the anatomical integrity of the cardiac parasympathetic neuronal network. Yet, in vivo neuronal recordings from postganglionic parasympathetic neurons unraveled both abnormal firing frequency at rest and abnormal responses to stimuli, proving a profound disruption of parasympathetic cardiac control after MI and reinforcing the strong rationale for therapeutic interventions aimed at restoring a proper cardiac vagal output, such as VNS [30].

Chronic electrical unilateral (left-sided) VNS has been used clinically for decades for the management of drug-refractory epilepsy [31] and, more recently, also for resistant depression [32] with more than 100,000 patients treated all over the world and no major safety concerns.

Table 1 lists the five clinical studies of cervical VNS in HFrEF (four published, one ongoing), showing their main inclusion/exclusion criteria, the objectives, and the stimulation protocols. Table 2 presents the patient characteristics and 6‑month results of the four published studies.

The first proof-of-concept study of electrical VNS in HFrEF was performed as a single-center experience in Italy, with eight participants (all males: [33]). Based on the favorable preliminary results at 6 months, the study was subsequently extended to a multicenter single-arm open-label phase-II European study including 32 patients and using the same device (CardioFit 5000; [34]). The study, usually referred to as European Multicenter CardioFit Study, aimed to assess the safety and tolerability of VNS (primary endpoints) as well as to collect detailed efficacy data (secondary endpoints). The device had already been favorably tested in the first animal studies by Sabbah’s group and consisted of an implantable neurostimulator and two leads: one stimulating the right cervical VN, the other connected to a standard bipolar sensing electrode to sense each QRS complex in the right ventricle. The vagal electrode had been specifically designed to achieve a preferential, albeit not exclusive, stimulation of efferent fibers, through simultaneous cathodic stimulation and asymmetrical anodal block combined to a multi-contact cuff design (to maximize, at relatively high-current amplitudes, B‑type fiber recruitment while minimizing A‑type fiber recruitment). The implanted system provided a pulse synchronous (1–3 pulses per cardiac cycle) VNS at 1–3 Hz, with 10 s on and 30 s off, and was programmed to temporarily stop in the case of HR lowering below a predefined safety value, which was initially set at 55 bpm, thus forming a so-called closed-loop system. In five to six visits, VNS intensity was up-titrated to reach a mean level of 4.1 ± 1.2 mA; hoarseness and jaw pain were the main limiting symptoms to a further increase. Overall, the mean HR reduction achieved during the on-phase was modest (1.5 bpm as measured by a 24‑h Holter recording) but consistent across patients, with a small minority of patients showing as much as a 10-bpm acute decrease. At 6 months, ECG resting HR decreased from 82 ± 13 to 76 ± 13 bpm. Notably, HRV tended to increase during the study and the change in pNN50 was statistically significant, despite no changes in the mean HR on Holter ECG. The latter finding was thought to reflect mainly a greater level of physical activity due to the improved HF status. Patients were in a very advanced baseline condition and on optimized and stable medical therapy including beta-blockers, angiotensin-converting enzyme inhibitors/angiotensin receptor blockers, and loop diuretics. Still, after 6 months of VNS, quality of life score (QoL) and functional capacity based on NYHA class and 6‑minute walk test (6MWT) significantly improved (see Table 2), together with a significant improvement in LV end-systolic volume (LVESV) and left ventricular ejection fraction (LVEF, from 22 ± 7% to 29 ± 8%) as assessed by blinded ultrasound analysis. The results were maintained in those who reached the 1‑ and 2‑year follow-up, with no major safety concerns. Notably, 59% of patients had an implantable cardioverter defibrillator (ICD) at baseline, but none had cardiac resynchronization therapy (CRT).

The second study assessing the effects of VNS in HFrEF was the Autonomic Regulation Therapy for the Improvement of Left Ventricular Function and Heart Failure Symptoms (ANTHEM-HF) study [35]. The ANTHEM-HF was a multicenter, open-label, phase II clinical trial enrolling patients at 10 Indian sites. Overall, 60 patients (NYHA class II–III, LVEF ≤ 40%, left ventricular end diastolic diameter [LVEDD], between 50 and 80 mm, QRS < 150 ms, none with ICD or CRT) were randomized to receive either left (n = 31) or right (n = 29) VNS system implantation (Demipulse Model 103 pulse generator and PerenniaFLEX Model 304 lead, Cyberonics, Houston, TX, USA). The stimulation system had already been approved for clinical use for drug-refractory epilepsy and tested in the high-rate pacing-induced canine model of HF. Vagal nerve stimulation was performed through an open-loop system lacking ECG synchronization capability and delivering a 14‑s on and 66‑s off stimulation, at 10 Hz; the vagal electrode was not designed for asymmetrical stimulation. After a 10-week up-titration period, a mean output current of 2.0 ± 0.6 mA was reached, with a good safety profile. The pooled right and left efficacy analysis at 6 months showed a significant (+4.5% in absolute values) increase in LVEF and a nonsignificant decrease of LVESV (co-primary endpoints). Further, QoL and NYHA also significantly improved. Right VNS was associated with a trend toward a greater benefit compared to left-sided stimulation. Improvements in cardiac function and HF symptoms seen after 6 months of VNS were maintained at 12 months [36], and until 42 months [37], with no significant difference between right- and left-sided VNS. A recent subanalysis including a 24- and a 36-month follow-up, confirmed the long-lasting, beneficial effect of VNS on autonomic tone (HRV), baroreceptor sensitivity (HR turbulence), and cardiac electrical stability as assessed by T‑wave alternans, R‑wave and T‑wave heterogeneity [38]. Additionally, a significant reduction of not-sustained ventricular tachycardia episodes at 12, 24, and 36 months was also reported [38].

The Neuronal Cardiac Therapy for Heart Failure (NECTAR-HF) study was a phase II, multicenter, sham-controlled study including 96 patients with NYHA class II–III, LVEF < 35%, a QRS < 130 ms and LVEDD > 55 mm [39]. All patients underwent implantation and were then randomized 2:1 to active VNS or sham treatment for the first 6 months; subsequently, VNS was turned on in all patients. The majority had an ICD, only few had CRT. The primary safety endpoint was 18-month all-cause mortality. The VNS system, sponsored by Boston Scientific (MN, USA), consisted of a rechargeable device already approved for chronic pain therapy (Precision) and an investigational helical bipolar vagal electrode not pursuing preferential stimulation and relatively similar to the one used in ANTHEM. After a 4-week up-titration period, a mean stimulation intensity amplitude of 1.4 ± 0.8 mA was reached at a stimulation frequency of 20 Hz. At 6‑month follow-up, left ventricular end systolic diameter (LVESD, primary efficacy endpoint) was unchanged, as was LVEDD, LVESV, LVEF, peak VO₂ at cardiopulmonary exercise test, and NT-proBNP (all additional secondary endpoints). Yet, there was a statistical improvement in QoL score and in NYHA class in treated patients. The 18-month results (with all patients on active VNS) showed the persistence of a favorable adverse event profile with acceptable survival rate (95%). However, LVESD and LVEF did not change in the crossover group, confirming the 6‑month findings. Finally, QoL scores did not change either, while NYHA class improved in all patients over the 18-month study period. Additionally, a new technique to detect subtle HR changes, namely, tridimensional heat maps, was applied to 24‑h Holter ECG obtained at 6 and at 12 months: only 12% of the studies showed VNS-evoked HR responses, as opposed to zero studies in the sham arm [40]. This rare occurrence of even minimal HR effects during vagal stimulation is at variance with the findings with the same device in the experimental studies and may suggest lower amplitudes of stimulation and lower fiber recruitment as a reason for the different effects between the preclinical and the clinical settings. In the NECTAR-HF trial, no apparent differences were found in conventional measures of frequency and time domain HR variability when comparing treated patients with and without evidence of VNS-evoked HR changes in the heat map analysis.

The last and largest clinical trial of VNS in HFrEF completed so far is the Increase of Vagal Tone in Heart Failure (INOVATE-HF), a phase III international, multicenter, randomized trial aimed at assessing the efficacy and safety of chronic right-sided VNS using the CardioFit system [41]. Beginning in 2011, 707 patients with NYHA class III, LVEF ≤ 40%, and LVEDD between 50 and 80 mm were enrolled at 85 centers in the United States, Europe, and Israel and randomized in a 3:2 ratio to either active treatment (implanted) or continuation of medical therapy (not implanted). Except for a slightly lower LVEF in the VNS group, the remaining baseline characteristics were well balanced. Most patients had cardiac devices, including one-third with CRT. The primary efficacy endpoint was the composite of all-cause mortality or unplanned HF hospitalization equivalent, using a time-to-first-event analysis. Freedom from procedure and system-related complication events at 90 days and number of patients with all-cause death or complications at 12 months were the two co-primary safety endpoints. The study was stopped for futility in December 2015 after the second planned interim analysis. The mean follow-up period was 16 months (range: 0.1–52) and the mean stimulation current was 3.9 ± 1.0 mA, with 73% of patients achieving the goal of ≥ 3.5 mA. The LVESV index, a pre-specified secondary endpoint, was unchanged, whereas QoL, NYHA class, and 6 MWT distance were improved by VNS (p < 0.05), although the unblinded nature of the study should be kept in mind. Subgroup analysis failed to detect any significant interaction between the primary endpoint and age, 6MWT at baseline, HF etiology, diabetes, and CRT.

The strong pathophysiological rationale, combined with the favorable results of preclinical studies and of the first two, not controlled, clinical studies set great expectations for VNS in HFrEF, which were met by disappointment due to the neutral results of NECTAR-HF and of INOVATE-HF, leading to a general loss of confidence in the technique. It should be considered, however, that the trials were undertaken in the absence of adequate knowledge on the dose–response relationship of VNS, on the subgroups of patients most likely to benefit from VNS, and on the most suitable study design.