Stimulating the Facial Nerve to Treat Ischemic Stroke: A Systematic Review

Acute ischemic stroke (AIS) is a common devastating disease that has increased yearly in absolute number of cases since 1990. While mechanical thrombectomy and tissue plasminogen activator (tPA) have proven to be effective treatments, their window-of-efficacy time is very short, leaving many patients with no viable treatment option. Over recent years there has been a growing interest in stimulating the facial nerves or ganglions to treat AIS. Pre-clinical studies have consistently demonstrated an increase in collateral blood flow (CBF) following ganglion stimulation, with positive indications in infarct size and neurological scores. Extensive human trials have focused on trans-oral electrical stimulation of the sphenopalatine ganglion, but have suffered from operational limitations and non-significant clinical findings. Regardless, the potential of ganglion stimulation to treat AIS or elongate the window-of-efficacy for current stroke treatments remains extremely promising. This review aims to summarize results from recent trial publications, highlight current innovations, and discuss future directions for the field. Importantly, this review comes after the release of four important clinical trials that were published in mid 2019.


INTRODUCTION Acute Ischemic Stroke
Stroke is the leading cause of disability and the fifth leading cause of death in the United States (US). Approximately 795,000 people experience a new or recurrent stroke each year (1). Acute ischemic stroke (AIS) occurs when an obstruction within a blood vessel decreases cerebral blood flow, depriving nerve cells of oxygen and leading to severe metabolic failure and neural death (2)(3)(4). Immediately following stroke, a section of the brain referred to as the ischemic core is subject to extreme hypoxia, leading to irreversible brain damage (5). The area surrounding the ischemic core, the ischemic penumbra, is severely hypoperfused and non-functioning, yet can regain functionality if blood flow is restored to the area (5,6). This recovery is highly time-dependent, as the penumbra rapidly evolves into the ischemic core (6,7). The recovery of the penumbra has been demonstrated to have a significant effect on clinical outcomes; Meretoja et al. showed that for every 20-min reduction in time to reperfusion increases the average disability-free life span by 3 months (8).

Current Treatments: Endovascular Thrombectomy and Tissue Plasminogen Activator
Management for AIS relies on rapid treatment times to avoid penumbra evolution. The goal of modern stroke treatment facilities is to reperfuse the ischemic area via endovascular thrombectomy, mechanically removing the blood clot with catheter-based devices (9). The faster the patient achieves reperfusion, the more likely the patient will have excellent neurological outcomes at 90 days (10). The second treatment paradigm for AIS is the use of intravenous (IV) tissue plasminogen activator (tPA), which acts to chemically break down clots (11)(12)(13)(14). IV-tPA is commonly used prior to patient transport for thrombectomy.
Both endovascular thrombectomy and IV-tPA suffer from a limited window-of-efficacy. Currently, stroke guidelines list the acceptable window of treatment for mechanical thrombectomy at 24 h (15,16), and a recommended IV-tPA door-to-treatment time of 60 min (10,17,18). These windows are frequently missed, with fewer than a third of patients in the U.S. treated within the IV-tPA 60 min window (19). Mandatory neuroimaging, presence of a highly-trained neurointerventionalist, and hospital transfer times all reduce event-to-treatment times and result in reductions of successful functional outcomes following recanalization (16,20,21).
Thrombectomy is a well-established treatment (15,16,22), but patients routinely fail to achieve functional independence (mRS ≤ 2) due to extensive event-to-treatment times (15,16). There is a clear need for innovative approaches to extend the window-of-efficacy for endovascular thrombectomy and IV-tPA.

Time Is Brain: Inhibiting the Evolution of the Ischemic Penumbra
Researchers have recently sought to find new approaches to arrest the evolution of the ischemic penumbra and keep this susceptible region from becoming irreversibly damaged. Current approaches aim to either enhance oxygen delivery to the penumbra or reduce tissue oxygen demand (6). In addition to door-to-treatment time limitations, many patients also become ineligible for mechanical thrombectomy due to large ischemic core volumes. Inhibiting the evolution of the penumbra may help buy time by limiting core volume growth from reaching recommended exclusionary levels (23,24). Inhibition of penumbra evolution could also be highly beneficial when combined with IV-tPA, potentially increasing its effectiveness and elongating its window-of-efficacy (6).

Increased Collateral Blood Circulation Through Ganglion Stimulation
Facial nerve-induced vasodilation of the cerebral arteries is an emerging therapeutic target that seeks to increase collateral blood flow in ischemic brain tissue, improve oxygen availability, and improve patient functional outcomes after stroke. Collateral circulation refers to alternative, pre-existing vascular pathways that deliver blood to target tissue when the primary vessel is occluded (25). Imaging of the brain and vessels has shown that collateral blood flow can preserve brain tissue for hours after major arteries to the brain are blocked (26). Figure 1 illustrates how facial nerve-induced increased collateral blood flow has the ability to ameliorate clot-induced tissue death by limiting ischemic penumbra evolution.
The sphenopalatine ganglion (SPG), also known as the pterygopalatine ganglion, is the largest and most superior ganglion of the sympathetic and parasympathetic nervous system and contains the largest collection of neurons in the calvarium outside of the brain (27). Humans have two SPGs, on each side of the midface, located within the viscerocranium in a space called the pterygopalatine fossa. This fossa has direct connections to the middle cranial fossa, nasal cavity, orbit, infratemporal fossa and oral cavity (28,29). The SPG contains parasympathetic fibers that synapse in the ganglion and innervate the internal carotid artery (ICA) through the deep petrosal nerve (30). Electrical stimulation of the SPG activates the fibers, releasing several neurotransmitters such as acetylcholine, vasoactive intestinal polypeptide, peptide histidine isoleucine, and nitrous oxide that play a role in inducing vasodilation of blood vessels in the anterior circulation (26,31).
The geniculate ganglion is a small collection of somatosensory and gustatory ganglion cells (32) located within the temporal bone along the axis of the ear canal. Similar to the SPG, the geniculate ganglion has parasympathetic connections to cerebral arteries and has also been demonstrated to increase cerebral blood flow (33,34). Its curved shape allows for greater susceptibility to be activated location makes it easier to access through non-invasive routes (33,35).

Search Strategy and Selection Criteria
The literature search was conducted in three electronic databases MEDLINE, EMBASE, and SCOPUS, to obtain pre-clinical and clinical studies dating from 1986 to 2019. Search terms included "isch(a)emic, " "brain, " "cerebr * , " "core, " and "penumbra" with intervention approach terms such as "parasympathetic, " "ganglion, " "facial nerve, " "stimulation, " "neuromodulate * , " "electrical acupuncture, " and "magnet * ." Studies selected for final review included those assessing the effects of stimulation in the ischemic stroke indication, as well as the effects on blood flow in the brain from ganglion stimulation. The inclusion criteria for pre-clinical studies had at least one of the following outcome measures: (1) healthy and ischemic stroke-induced animal models, (2) invasive or noninvasive ganglia stimulation, and/or (3) studies providing at least one outcome measurement of cerebral or collateral blood flow, brain infarct size, neuronal survival, and neurological outcomes. Respectively, the inclusion criteria for clinical studies had at least one of the following outcome measures: (1) healthy human volunteers or patients with ischemic stroke, (2) invasive or non-invasive ganglia stimulation, (3) sham stimulation control group, (4) safety and tolerability of intervention, and/or (5) studies providing at least one outcome measurement of cerebral or collateral blood flow, brain infarct size, neuronal survival, functional and neurological outcomes including but not limited to mRS at 90 days post-stroke. (C) SPG stimulation induces vasodilation of blood vessels in the anterior circulation, increasing blood flow to the affected area limiting the evolution of the penumbra and reducing infarct volume.
Pre-clinical and clinical studies were excluded if they evaluated facial nerve stimulation for other indications such as hemorrhagic stroke, chronic headache, migraine, rhinitis, or stroke rehabilitation. Reviews, meta-analyses, observational studies, and commentaries were also excluded.

Risk of Bias Analysis
The final emerging pre-clinical studies were subjected to the Systematic Review Center for Laboratory Animal Experimentation (SYRCLE) risk of bias tool (36). The SYRCLE's tool for assessing risk of bias contains 10 domains which fall under six types of bias; selection bias, performance bias, detection bias, attrition bias, reporting bias, and other. Judgement of bias was indicated as low by a positive ("yes"), high by a negative ("no"), and imprecise ("unclear") when details reported were insufficient. Physiotherapy Evidence Database (PEDro) scale was selected as the risk of bias assessment tool for clinical studies (37). The PEDro scale consists of 10 items; they are random allocation, concealment of allocation, baseline equivalence, blind subjects, blind therapists, blind assessors, intention to treat analysis, adequate follow-up, between-group statistical analysis, and measurement of data variability and point estimates. Studies with a PEDro score of more than 6 were considered good quality.

RESULTS
The initial search yielded 1,106 studies: 263 duplicates were removed, 843 studies were screened by title and abstract, and 717 studies were deemed irrelevant articles. A total of 126 studies were assessed for eligibility and were subjected to the predefined inclusion and exclusion criteria via full-text review, resulting in the removal of 100 studies. The final emerging 26 studies combined in this review assessed 22 pre-clinical studies and 5 human studies. One publication contained both a human and a pre-clinical trial. Figure 2 outlines the flow of study selection.

Quality of Studies
The methodological quality assessments of included pre-clinical and clinical studies are provided in Tables 1, 2, respectively. Animal studies assessed by the SYRCLE found a prevalence of 41.3% for items classified as "unclear, " and 5.9% for items classified as "no" ( Table 1). Human trials assessed by the PEDro scale had an average score of 6.0 with scores ranging from 4 to 7 ( Table 2).
Key parameters of stimulation protocols included intensity, frequency, pulse, and duration (Tables 3, 4). Several studies sought to find a dose-response relationship or a maximum tolerated frequency and intensity, causing variability between reported stimulation protocols. Electrical stimulations reported intensity in Amps or Volts, ranging from 0.5 to 5mA and 2 to 20V. CBF significantly increased along with blood pressure in a stimulation frequency-dependent manner (41,46). Even in studies where this dependency was not clear, lower powers and durations were markedly less effective in increasing CBF (56).
Intervention groups undergoing stimulation routinely demonstrated increased CBF. Three studies did not report an increase in CBF, with one not recording CBF (42), one reporting no change (64), and another reporting a decrease following vagus nerve stimulation (39). Goadsby et al. (49) reported the largest shift from baseline following stimulation, reporting a 150% increase following 200 s of stimulation. CBF increased throughout the brain without an obvious preference for the hemisphere ipsilateral to stimulation (54,56). The ability for stimulation to increase CBF flow was found to be lower in ischemic models compared to CBF increase in health models (39,40,53). Animals that underwent invasive stimulation were found to have improved neurological scores in five of the seventeen studies (38,39,42,53,59). Only one study tracking neurological scores found a decrease following ganglion stimulation, however, they also reported a reduction in infarct volume compared to controls (40). In all studies that reported infarct volume stimulated models performed better (38-40, 42, 53, 55, 59). Bar-Shir et al. (42) found both improved neurological scores and infarct volumes in stimulated models at 8 days compared to control, but no difference in a 28 day follow-up.

Invasive Ganglion Stimulation: Clinical Studies
Four clinical studies using invasive electrical stimulation to treat AIS were identified ( Table 5) (60)(61)(62)(63). All four targeted  the SPG using an implanted device (ISS, BrainsGate, Israel). The procedure involved surgically placing a platinum-iridium stimulator trans-orally through the greater palatine canal in the extracranial pterygopalatine fossa, adjacent to the SPG. A transmitter coil was then placed on the patient's cheek to induce an electronic circuit. Khurana et al. (60) reported the outcomes of the implant for augmentation of cerebral blood flow trial-1 (ImpACT-1). The prospective, single-arm, feasibility trial was conducted between 2006 and 2009. The trial recruited 98 subjects and reported no major safety concerns. At least one SAE was reported in 23 patients, with three classified as "related" or "possibly related" to the intervention, mostly due to the re-implantation and/or implant misplacement. Pain during stimulation (15/92) was the most frequently reported AE. Secondary efficacy analysis demonstrated a very slight reduction in mRS score (−0.76) between stimulation and historical controls. There was also a significant improvement in functional independence (0-2 mRS) at 90 days, with 48% of stimulated patients (40/84) reaching functional independence compared to 29% (48/165) in the historical control group (60).
Bornstein et al. (61) reported the outcomes of the implant for augmentation of cerebral blood flow trial-24A (ImpACT-24A), the follow-up efficacy trial of the ImpACT-1. The trial used a two-arm, randomized, double-blind design with a sham-control. Within the sham procedures the trans-oral device was implanted but no electrical current was activated. The trial was conducted from 2009 to 2011. The trial recruited patients with evidence of stroke in the anterior circulation, were able to undergo treatment within 24 h, and were ineligible for IV-tPA or endovascular thrombectomy. Bornstein et al. reported enrollment of 327 patients, however six exited prior to implantation and 18 had incomplete implantations. Patients were randomized 2:1, with 202 undergoing stimulation and 101 undergoing sham. Due to issues during implantation a new optic navigation system was introduced midway through the trial, however, only 75.7% of patients in the active group received stimulation. Due to the inconsistency of accurate device placement the trial was ended at the first interim analysis. A modified intent to treat (mITT) analysis looking at patients who underwent at least one full successful active or sham treatment found no benefits in either the primary outcome or any of the secondary endpoints. Post-hoc analysis demonstrated significant improvements within a subsection of the population that had confirmed cortical involvement (CCI), defined as patients with NIHSS score ≥10 and signs of hypodensity or tissue swelling in at least one cortical region on initial imaging (61).
The results of the follow-up trial, ImpACT24B, was conducted between 2011 and 2018 and reported by Bornstein et al. (62). Trial set-up was nearly identical to ImpACT24A, however, the CCI subsection was added as a primary outcome measure of interest. The trial protocol was changed several times, including the introduction of a new guidance system, neurostimulator, implantation technique, and electrical transmitter-control unit. Of the 1,078 patients recruited, 481 underwent stimulation and 519 underwent a sham operation. Again, no benefit in the mITT population was seen in any analysis, however the CCI subgroup showed statistically significant improvements in every endpoint. An inverted U-shape relationship between stimulation intensity and effectiveness in the CCI subgroup, with lower-mid-level stimulation indicated as most effective (62).
The ImpACT24M study, reported by Saver et al. (63), was a single-arm study assessing the effect of two changes to the SPG stimulation procedure required for implanting and optimizing the device. Primary outcomes were difference in 7-day NIH Stroke Scale (NIHSS), and proportion of patients with improvement in stroke symptoms, compared to historical controls (65). The 49 patients that received stimulation demonstrated a median NIHSS improvement of 75% compared to 50% in the historical control. Stimulation was also shown to significantly increase CBF and hand motor function within the cohort, but was not compared to the historical control (63).
There is one ongoing trial (NCT04014621) to determine if 6 h of SPG stimulation using the same procedure in the ImpACT studies will "freeze" the ischemic penumbra in patients with acute ischemic stroke and reduce brain tissue death. The primary outcome measure is volume of core expansion determined by CT scan post SPG stimulation. This trial is expected to complete in spring 2021.

Non-invasive Ganglion Stimulation: Pre-clinical Studies
Six of the 26 publications identified assessed non-invasive interventions to stimulate facial nerves or ganglions ( Table 4) (43-45, 51, 52, 56). Three studies used a non-invasive magnetic intervention, only one of which used an ischemic stroke model (43,44,56). Four studies used a non-invasive electrical intervention, all of which were tested in healthy models (43,45,51,52). Borsody et al. (43) reported outcomes on both a magnetic and electrical non-invasive intervention, with each intervention targeting separate ganglions.
Magnetic stimulation interventions were primarily deployed through tesla coils located on either side of the subject's head. Magnetic intensities were reported in Tesla, and ranged from 0.5 to 1.9 T in the three studies (43,44,56). Non-invasive electrical stimulation was induced over 30-90 min, while the magnetic stimulation was induced over several 5 min increments with a 30 min recovery period.
All non-invasive stimulation studies found an increase in CBF compared to controls (43-45, 51, 52, 56), however, Borsody et al. (43) found a reduction in CBF following non-invasive electrical stimulation through the middle ear. They also reported the largest change in CBF following magnetic stimulation, with a 120% increase compared to baseline using a 10 Hz frequency at 1.5 T. Comparatively, Sanchez et al. (56) found the increase in CBF occurred throughout the brain without obvious preference for the hemisphere ipsilateral to stimulation, however this finding was not reported in any other study. No studies recorded neurological scores, and only two reported on infarct size (44,45), with non-invasive stimulation significantly reducing infarct size compared to control in both studies.

Non-invasive Ganglion Stimulation: Clinical Studies
Sanchez et al. (56) was the only clinical study that assessed the ability of non-invasive ganglion stimulation ( Table 5). The geniculate ganglion was targeted and stimulated at 1, 1.3, 1.6, and 1.9 T in 280 µs pulses for 3 min in health volunteers. All tests were conducted using a single device (VitalFlow, NeuroSpring, California). CBF was found to significantly increase following stimulation, with higher intensity correlating to larger CBF flux. No adverse effects were reported within a 24-h follow-up.

DISCUSSION
There is a substantial amount of pre-clinical and clinical data to suggest that stimulation of the parasympathetic fibers of the facial nerve system is a promising option for rapid AIS treatment. Data from early trials with both invasive and noninvasive approaches have proven safe with very minimal reports of treatment-associated adverse events. Stimulation of the SPG or geniculate ganglion clearly results in increased CBF, and the technological means required to rapidly induce ganglion stimulation are certainly feasible. While the SPG has been the most well-studied ganglion in clinical trials to date, preclinical evidence suggests a variety of targets for the potential therapeutic development.
Despite the tremendous upside it poses for treating AIS, ganglion stimulation has yet to demonstrate clinically meaningful outcomes. To date the clinical trials for SPG stimulation via a trans-oral device have suffered from various methodological shortcomings that obscure the reasons behind the lack of positive results. A recurring issue with incorrect device placement resulted in the study protocols for both the ImpACT-24A and ImpACT-24B trials to be repeatedly modified during the trial course. The mid-trial change of the neurostimulator, implantation technique and electrical transmitter-control unit in the ImpACT-24B trial is especially disconcerting and was not addressed in the outcome analysis. Additionally, while the trials claim to be double or triple blinded, it is unclear how successful the blinding can be for the subject and physician when a key element of the intervention's deployment is mild facial discomfort. It is also not clear why the ImpACT-1 and ImpACT-24A were published almost 10 years after completion, and simultaneously with ImpACT-24B and ImpACT-24M.
Current strategies for employing ganglion stimulation to treat AIS suffer from three core limitations: (1) current trial design, (2), time-to-deployment, and (3) accessibility to trained physicians.
The ImpACT trials have focused on independently improving AIS-related disability, but their negative results suggest that tPA and mechanical thrombectomy will remain the gold standard for AIS treatment. It is therefore optimal for facial nerve stimulation to augment and improve the applicability of these existing treatments. There is a clear need for interventions that elongate  . CBF-derived lesion volumes in stimulated animals were ∼10% smaller (non-significant, P > 0.05) than in non-stimulated controls Neurological score: The 24-h neurological scores (mean ± SD min, max, range) were improved in the SPG group (2.5 ± 0.8, 2, 4.2) relative to controls (3.7 ± 0.8, 3, 5, 2). CBF: In the non-ischemic brain, SPG stimulation significantly elevated CBF predominantly within areas supplied by the anterior cerebral artery (by 0.64 mL/g/min relative to baseline. In the ischemic brain, CBF only marginally increased within the penumbra and core (by up to 0.08 and 0.15 mL/g/min relative to pre-stimulation, respectively) Frontiers in Neurology | www.frontiersin.org Dorsal facial area CBF CBF: Electrical stimulation of the DFA appeared to increase the regional blood flow of both cerebral hemispheres (intracranial tissues) and to increase predominantly the regional blood flow of extracranial tissues on the side ipsilateral to stimulation Infarct size CBF Infarct size: In SPG-stimulated rats (n = 6), both the size of the stained cortical region and the intensity of the dye were smaller compared to those in RB-treated non-stimulated animals (n = 6). Rats stimulated either at 15 min or 24 h after photothrombosis showed a significant reduction in loss of brain tissue compared to non-treated controls (37.1 ± 6.0; 20.6 ± 4.8 and 17.9 ± 5.1% reduction in cortical volume compared to the contralateral hemisphere for RB, n = 9; RB-SPG-15 min, n = 8 and RB-SPG-24 h, n = 6, respectively. CBF: Stimulation at 1 mA (200 ms) did not induce a significant increase in vascular diameter or rCBF (n = 7), however stimulation at 2 mA resulted in a significant increase in both vascular diameter and rCBF in 12 of 14 (86%) rats. Prolonging the pulse duration to 500 ms was associated with an additional significant increase in diameter and rCBF ( Figure 1C). No significant changes in vascular diameter or rCBF were observed in non-stimulated animals during 3 h of recordings (n   *Improved mRS score at 3 months beyond expectation. **(NIHSS score ≤1 or improved ≥9) at 3 months. ***(mRS 0-2) at 3 months.
Frontiers in Neurology | www.frontiersin.org the current window-of-efficacy (19), with extended door-totreatment times leaving only 3-22% of AIS patients eligible for mechanical thrombectomy (66), and <5% eligible for tPA (67). Primary outcomes within ganglion stimulation trials should be focused on increasing the efficacy of mechanical thrombectomy for AIS patients that miss the current window-of-efficacy. The upcoming trial (NCT04014621) looking at penumbra "freezing" suggests future ImpACT studies involving invasive approaches may soon follow this suggestion. However, if the goal of future trials is to stop the evolution of the penumbra prior to recanalization, it is key that ganglion stimulation protocols add minimal delay to the already extensive door-to-treatment time. The design of the current clinical stimulation devices and associated procedures are limited in their ability to accomplish this. The ImpACT-24B trial saw a 1.2-h difference in time from "last known well" on NIHSS to the first stimulation session between sham and control groups; this is likely due to the procedure associated with finding the ideal stimulation intensity, a step inherently missing from the sham procedure. Even without this optimization step, the implantation of the invasive device takes a reported average of 20 min, not including time for setting a sterile environment. This substantially eats away at thrombectomy's window-of-efficacy or the even shorter 60-min window for tPA induction (33,68). Transfer time from intake hospitals to stroke centers is an ideal time for rapid ganglion stimulation, however, requirements associated with invasive implantation limit their applicability in an ambulatory care setting. Even the non-invasive methods used by Sanchez et al. (56) require TMS coils, which are often large and require substantial energy supply to induce magnetic fields.
The requirement of highly trained physicians, specialized imaging, and sterile fields further limit the currently studied innovations. Current invasive techniques for ganglion stimulation can be employed only by specialized physicians, demonstrated by 58% of implantations in the ImpACT-24B trial being performed by surgeons and anesthesiologists, with the remaining performed by neurologists (62). Implanting the device also requires an advanced optical guidance system that would likely only be available to stroke centers, limiting its ability to be implemented in peripheral hospitals prior to transfer.
The research to date on ganglion stimulation is extremely promising, but further innovation is required to find a workable integration of ganglion stimulation into current clinical procedures. An inherent focus must be on limiting the evolution of the penumbra and minimizing the size of the ischemic core, with the primary goal of elongating the window-of-efficacy for mechanical thrombectomy or tPA. Solutions should look to benefit these current treatment options instead of supplanting them, and find a way to be used primarily at peripheral nonstroke centers.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.

AUTHOR CONTRIBUTIONS
AC is the guarantor. TB conceptualized the study. TB and JR developed the search strategy and drafted the manuscript. JR extracted the data. All authors contributed to the development of the selection criteria, risk of bias assessment strategy, and data extraction criteria and also read, provided feedback, and approved the final manuscript.