Clinical aspects

Description

The neurocybernetic prosthesis consists of an electronic generator that delivers stimulation through a flexible bipolar lead that attaches to the vagus nerve.

Nine generator models have been produced, including the no longer distributed models 100 and 101. The available models differ in size, battery capacity, and compatibility with either a single pin or dual pin lead. Models 102 and 102R essentially differ in the lead compatibility. The 102 accepts the single pin of the more recently produced leads and the 102R accepts the dual pin of the older leads. Their appearances are similar with a weight of 25 grams (27 grams for 102R), thickness of 7 mm, and volume of 14 cc (16 cc for 102R). Analogously, models 103 and 104 essentially differ in the lead compatibility with the 103 accepting a single pin lead and the 104 accepting the dual pin. Their thickness is the same as the 102/102R at 7 mm, but they are smaller with volume of 8 cc (10 cc for 104) and weight of 16 grams (18 grams for the 104). Model 105, a newer model, is the same size as model 102, accepts only the single pin lead, provides more additional information about battery depletion and has a larger capacity battery. Model 106 (approved May 2015) and Model 1000 (approved October 2017) are the newer models, and both have an added optional mode called AutoStim mode or automatic stimulation. This option allows for automatically delivered stimulation based on a heart rate change that is predictive of a seizure onset. Limited evidence is available at this time about the relationship between the long-term health outcomes in patients with epilepsy and automatic stimulation, but case series have shown the AutoStim mode may be sensitive to 80% of seizures. Model 106 accepts a single pin and weighs similar to model 105. Model 1000 also accepts a single pin lead, and it weighs 16 grams. Additionally, model 1000 has an accelerometer that provides information regarding whether the patient is prone or supine when in a recumbent position.

Battery longevity depends on stimulation parameters. For the purpose of comparison, longevity has been estimated for parameters that may be used clinically while recognizing that parameters differ for patients depending on clinical needs. With these standard parameters, the 102/102R may last about 8.4 years, the 103/104 for about 6 years, and the 105 for more than 10 years. The expected battery longevities of model 106 and 1000 are 12.8 and 11.5 years, respectively, when the auto-stimulation mode is “OFF.” The battery life reduced to 4 to 7 years depending on the parameters when the automatic stimulation mode is “ON.”

None of the models has a rechargeable battery; therefore, the generator must be replaced when the battery is depleted. The generator is implanted in a subcutaneous pouch at the lateral border of the left pectoral muscle.

The lead attaches to the cervical vagus nerve with a spiral cuff that encircles the nerve. The nerve is accessed through a dissection similar to the approach for a carotid endarterectomy. A special tunneling tool is used to create a path for the lead to reach the generator through the subcutaneous tissue. The implantation requires a 4 cm to 5 cm chest incision and a 2 cm to 2.5 cm neck incision. Replacement of the battery at the end of its life generally requires accessing only the generator because the lead may be disconnected from the generator that is to be removed and then attached to its replacement.

The neurocybernetic prosthesis is programmed through a radio frequency transmitting and receiving wand, which is controlled by a personal digital apparatus with special software.

With this apparatus, the neurocybernetic prosthesis may be tested, and its stimulation parameters may be set. Stimulation occurs by 2 or 3 methods (depending on model). Automatic and chronic stimulation occurs according to a duty cycle of on and off times. Stimulation also may occur on demand when a special magnet is moved across the chest in front of the neurocybernetic prosthesis. In the AutoStim mode, an on-demand stimulation is given when a relative increase in heart rate is detected that surpasses an adjustable threshold. This is an optional feature and works in combination with the normal mode if the AutoStim mode is ON. There are 8 stimulation parameters: (1) stimulation frequency, (2) duty cycle on time, (3) duty cycle off time, (4) duty cycle output current, (5) duty cycle pulse width, (6) magnet-triggered stimulation duration, (7) magnet-triggered output current, and (8) magnet-triggered pulse width. Should the patient desire, the device may be inactivated by placing of the magnet over the generator. Normal cycling resumes on removal of the magnet. The newer model 1000 also has the possibility of programming according to a scheduled difference between day and night.

To minimize any stimulation induced unpleasant sensations within the neck, treatment begins with a low output current and ramps up over several steps at regularly scheduled outpatient visits. The conventional settings at the initiation of treatment are in Table 1.

Table 1. Conventional Parameters at Initiation of Vagus Nerve Stimulation

Over weeks to months, the output current reaches the 1 mA to 3 mA that is typically used during treatment. Output current beyond 2 mA to 3 mA is rarely used because it consumes the battery faster, may produce more intense adverse effects, and is unlikely to be more effective for seizure control. Other parameters also may be adjusted to maximize the tolerability of stimulation. Decreasing the pulse width to 250 µsec may improve tolerability without apparent loss of seizure control (48). Decreasing the stimulation frequency to 20 Hz is thought by many practitioners to have the same tolerability benefit and lack of drawback (76). We have observed that a frequency of 25 Hz or a pulse width of 250 µsec also is commonly effective and almost always better tolerated. Should the conventional parameters not produce adequate seizure control, adjustment of the duty cycle to increase the percentage of on time may lead to improvement. Various duty cycles have been investigated and compared to the conventional 10% cycle of 30 seconds of stimulation separated by 5 minutes without stimulation. A review of settings used in an extension of one of the neurocybernetic prosthesis' clinical trials failed to show a correlation between stimulation parameters and efficacy, but identified a small subgroup of patients who benefited from a reduction in off time to less than or equal to 1.1 min (15). However, a randomized comparison of 3 stimulation paradigms demonstrated that the conventional settings are equally efficacious to more rapid cycle settings for newly treated patients (14).

Although the neurocybernetic prosthesis is not affected by microwave ovens or other common sources of electromagnetic energy, routine medical care may be affected by the presence of an implanted neurocybernetic prosthesis. MRI with a body coil should not be performed because of the risk of heating injury to the vagus nerve. MRI of the brain may be safely performed if a transmit and receive head coil is used and the scanner’s magnet is less than or equal to 3 tesla. Setting the generator to 0 mA output prior to scanning and testing of the neurocybernetic prosthesis system after scanning is recommended. A survey of 40 centers produced 12 responses and a series of 27 MRI scans of the head on 25 patients with vagus nerve stimulation. All MRIs were with 1.5 tesla scanners and 26 were done with a head coil. Only one patient had a possible adverse event, and this was a mild voice change lasting several minutes (04). Magnetoencephalography is hampered when a generator is in place, regardless of its parameter settings. The electromagnetic noise from the generator produces excessive magnetoencephalography artifact. However, this artifact on the magnetoencephalogram may be minimized with special post-processing techniques. Lastly, diathermy of any sort to any portion of the body is contraindicated for as long as any portion of the neurocybernetic prosthesis is implanted. This is because of the risk for heating of the neurocybernetic prosthesis system and the subsequent potential for permanent nerve, muscle, or vascular damage. This damage may result in loss of vocal cord function or possibly death if there is damage to the carotid artery or jugular vein.

The goal of vagus nerve stimulation therapy is the same as the goal of antiepileptic drug treatment of epilepsy: the elimination of seizures without treatment related adverse effects. However, few patients treated with vagus nerve stimulator have complete seizure control, so the goal in practice is the maximal reduction in seizure frequency with a reduction in antiepileptic medications when possible. Because vagus nerve stimulation, like antiepileptic drugs, has so far been demonstrated to be only antiseizure and not antiepileptogenic, treatment must be continued for as long as the patient is at risk for seizures. That is, vagus nerve stimulation does not alter epilepsy’s natural history.

Indications

The United States Food and Drug Administration has approved the neurocybernetic prosthesis for the adjunctive treatment of partial seizures that are refractory to antiepileptic drugs in patients over 12 years of age. In June 2017, this indication was extended for use in children above the age of 4 years (25).

The common clinical practice includes trials of at least 2 or 3 antiepileptic drugs and consideration of epilepsy surgery prior to consideration of vagus nerve stimulation. Because therapy with vagus nerve stimulation requires an operation, and most individuals with epilepsy experience seizure control with antiepileptic drugs alone, antiepileptic drugs are always tried first. However, some individuals may wish to consider vagus nerve stimulation early to avoid the undesired effects of antiepileptic drug treatment. When they are not effective in preventing seizures and the patient has partial (with or without generalization) seizures, a search for a resectable focus should be undertaken because certain syndromes, such as mesial temporal lobe epilepsies, are likely to be treatable with resective surgery.

In 1999, the American Academy of Neurology endorsed vagus nerve stimulation therapy for medication refractory epilepsy in patients older than 12 years with partial seizures, but it recommended an epilepsy surgery evaluation prior to vagus nerve stimulation therapy for consideration of surgery and to rule out a nonepileptic condition as the cause of the seizures (23). The American Academy of Neurology updated this evidence-based guideline in 2013 based on subsequently published research results (56; 76; 27; 50; 87). The update expands the utilization of vagus nerve stimulation and states that it may be considered in children and for seizures associated with Lennox-Gastaut syndrome. Furthermore, the update concludes that vagus nerve stimulation was found to have improving seizure control efficacy over time and a mood improving effect in adults.

Clinical practice often extends beyond official guidelines, and this is true for vagus nerve stimulation. In addition to Lennox-Gastaut syndrome, other generalized epilepsy syndromes also have been found to respond well to vagus nerve stimulation. Thus, use of the procedure extends beyond just partial seizures. Moreover, conventional clinical practice now employs vagus nerve stimulation prior to corpus callosotomy in the treatment of epileptic drop attacks.

Vagus nerve stimulation’s use as a therapy for treatment-resistant depression has been approved by the regulatory agencies of the United States, European Union, and Canada. This use of vagus nerve stimulation for depression is supported by an observed improvement in standardized mood scores among patients with epilepsy following initiation of treatment with vagus nerve stimulation (32). The mood improvement did not differ between responder and nonresponders to vagus nerve stimulation treatment for seizures, so appears to be independent to the antiepileptic effect. Improvement in mood scales is a finding that has been duplicated in other uncontrolled studies involving patients with epilepsy or patients with severe, treatment-resistant, non-psychotic depression (70). A meta-analysis of the 6 trials that have compared vagus nerve stimulation to treatment as usual for treatment-resistant depression has found a significantly increased likelihood of response (odds ratio 3.19) and remission (odds ratio 4.99) when vagus nerve stimulation was included in the treatment (05). However, the only randomized controlled trial did not identify a significant difference, and a meta-analysis that included a meta-regression found that a positive response to vagus nerve stimulation is dependent by about 84% on severity of depression with more severe depression predicting a better response (51). This suggests the possibility of regression to the mean influencing the outcome and the value of another randomized controlled trial. The randomized controlled trial lasted 10 weeks, which may not have allowed sufficient time to observe a response to treatment (66).

Other clinical indications for vagus nerve stimulation that have undergone clinical investigation include Alzheimer disease, chronic pain, migraine, and cardiac arrhythmias and left ventricular dysfunction associated with heart failure (71; 54; 07; 90; 13). The antinociceptive effect of vagus nerve stimulation appears to be a chronic change within central pain inhibition centers as the effect was independent to the on/off cycle of the stimulator and specific to pain that is amplified by central processing. The response to experimentally induced pain before and after initiation of vagus nerve stimulation therapy was found to significantly differ (39).

Contraindications

Left cervical vagotomy is the only standard contraindication to vagus nerve stimulation, but any situation in which the left vagus nerve may be especially vulnerable should lead to a detailed consideration of the risk for permanent vagus injury. In situations considered to have high risk of left vagus nerve injury, 3 patients have been reported to receive right vagus nerve stimulation with the neurocybernetic prosthesis and no complications arose (55).

Vagus nerve stimulation may result in the worsening of preexisting obstructive sleep apnea, but the effect may be at least partially lessened by stimulation parameter adjustments (53). Some patients experience swallowing difficulty during stimulation, but this is usually minor. The possibility of worsening obstructive sleep apnea or preexisting swallowing dysfunction should be considered against potential benefits prior to implantation.

Outcomes

Five clinical trials of the neurocybernetic prosthesis were conducted prior to its approval by the FDA. Three trials were longitudinal, and 2 were randomized, blinded, and with an active control (74; 75; 31). The active control trials used 2 different parameter settings, high and low. The high parameters are essentially those that are conventionally used in current clinical practice, and the low parameters were selected because of a presumed lesser effect on seizure control. In total, 451 subjects received stimulation in the 5 studies with 313 of these in the randomized controlled trials. The responder rate (percentage of subjects who experienced greater than or equal to 50% reduction in seizure frequency compared to baseline) for the 3 longitudinal trials ranged from 29% to 50%. The responder rates for the 2 randomized controlled trials were 23% and 30% for those receiving high stimulation parameters and 14% and 16% for those receiving low parameters.

Open extensions of the initial trials have provided additional efficacy data. The average responder rate after one year of high stimulation was analyzed for the first randomized controlled trial with an intent to treat approach (68). All 114 subjects were included in the analysis, although 14 discontinued treatment prior to 1 year of stimulation. Subjects who received low stimulation in the trial were changed to high stimulation on the 3-month trial’s completion and followed for 1 year. After 1 year of high stimulation, the average responder rate was 32%. A similar analysis was performed for the other randomized controlled trial, and its responder rate at 1 year was 35% (15). This follow-up of the subjects from second randomized controlled trial also found that 20% of subjects had less than 75% reduction in seizures. Responder rates over even longer periods of time were calculated using all 5 initial trials with data from 440 of the 451 subjects (57). The responder rate at one year was 37%. At 2 years and 3 years, it was 43%.

More recent, prospective, uncontrolled series have demonstrated similar results. In a group of 118 patients who were followed for at least 6 months, 50% were responders (78). In another group of 138 patients, 47% were responders (59). Similar results also were found in this series for the patients with generalized epilepsies. Of the 13 with symptomatic generalized epilepsy, 46% were responders. Of the 14 with idiopathic generalized epilepsy, 57% were responders. Such registries also have demonstrated a sustained efficacy with seizure reduction increasing from 45% to 58% from 3 to 12 months of vagus nerve stimulation in a group of 269 patients whose antiepileptic medications were unchanged (46). A prospective study of 43 patients whose anti-seizure medications were not changed identified 63% were responders at 18 months of vagus nerve stimulation (28). A prospective registry of 4483 patients, 1104 of whom reached 24 months of treatment, identified a 44% response rate at 3 months and progressive increases in the response rate to 56% by 24 months (20). Although data were gathered prospectively, selection bias remains as a potential confound. An analysis of the same prospective registry of 5554 patients showed 49% and 63% response rate at 0 to 4 months and 24 to 48 months of treatment, respectively. These rates of seizure freedom observed over time were supported with a systematic literature review of 2869 patients treated with vagal nerve stimulation across 78 studies (21). A retrospective study of 65 consecutive patients who had been treated with vagus nerve stimulation for more than 10 years identified a progressively increasing response (18). With a mean vagus nerve stimulation duration of 10.4 years, the mean decrease in seizure frequency at last follow-up was 76%. The decrease at 1 year was 36%, at 4 years was 58%, and at 8 years was 66%. An analysis from the same center that included 436 consecutive patients with a mean vagus nerve stimulation duration of 4.9 years identified a mean seizure frequency reduction of 56% (19). A retrospective study of 74 adults followed between 10 to 17 years agues vagus nerve stimulation should not be discontinued due to lack of efficacy until 2 years is complete due to the large increase in efficacy in the second year (responder rate 38% year 1 and 51% year 2, respectively) (10).

Smaller studies of children demonstrate vagus nerve stimulation efficacy for seizure control and improved quality of life (81). A prospective study of 13 children with localization-related epilepsy and who were between 5 and 18 years showed a 3 month 50% responder rate of 77% (88). A multicenter study on children with Lennox-Gastaut syndrome suggests that children with this epilepsy syndrome may be particularly responsive to vagus nerve stimulation (27). This retrospective analysis demonstrated overall seizure reductions of 58% at both 3 and 6 months. Particularly striking was the reductions in drop attacks by 47% at 3 months, and 88% by 6 months. Drop attacks are the most disabling type of seizures in these children, often result in significant physical injury, and are also particularly refractory to antiepileptic drugs.

The indication for vagus nerve stimulation was expanded to 4 years of age and older by the Food and Drug Administration in June 2017 (25). This has been based on extrapolation of data from 4 premarket studies, 1 post market study, and a database of clinical use. Of the 847 patients included in the approval submission, 117 patients (13.8%) were between 4 to 11 years of age. At 12 months, 582 patients were evaluated for efficacy outcome. The median reduction in seizure frequency was -24.7% and -40.4%, respectively, for the 4 to 11 years age group (n = 54) and over 12 years age group (n = 528). Overall response rate was 35% (19 out of 54) for 4 to 11 year age compared to 42% (224 out of 528) for over 12 years (25). A metaanalysis compared corpus callosotomy versus vagus nerve stimulation implantation in children with atonic seizures (84). It found that although corpus callosotomy was more effective (effect size 0.73, 95% CI 0.69-0.77), there was a higher rate of complications (6.6% vs. 3.8%), and a 14% risk of symptomatic disconnection syndrome.

The improvement in seizure control that vagus nerve stimulation may provide occurs without the typical adverse effects of antiepileptic drugs such as drowsiness, dizziness, ataxia, and memory impairment. Moreover, some patients may have an overall decrease in typical antiepileptic drug side effects by supplanting part or all of their epilepsy treatment with vagus nerve stimulation. This was demonstrated in a prospective, case-matched study of 21 patients after 13 months of vagus nerve stimulation. Fifteen of these patients had a decrease in antiepileptic drug treatment without deterioration in seizure control (73). This decrease was through reduction in the number of antiepileptic drugs (5 patients), antiepileptic drug dose(s) (6 patients), or both (4 patients). In a prospective, uncontrolled series of patients with mental retardation and developmental disabilities, the average number of antiepileptic drugs per patient decreased from 3.3 to 2.3 (37). Independent to the decrease in antiepileptic drugs and their side effects, vagus nerve stimulation may improve memory function by enhancing consolidation with subsequently improved retention (29). In a Spanish retrospective cohort of patients, there was a significant impact on quality of life with an average 8 point reduction (8.5 +/- 7.2) (52). In a retrospective cohort study, vagus nerve stimulation over the long-term showed a significant reduction in age-adjusted risk of sudden unexpected death in epilepsy (SUDEP) (67). The SUDEP risk at follow up between 1 to 2 and 3 to 10 years were 2.4 out of 1000 and 1.68 out of 1000 patients years, respectively. The findings should be interpreted carefully given the limitations of the study.

A study suggested that the standard deviation of successive RR intervals significantly decreases in the 10 minutes before a seizure in patients who were likely to be nonresponders to vagus nerve stimulation (8 of 11), noting that this measure is highly individualized (35). Due to the small size of this cohort and a lack of replication, the results should be considered exploratory.

Adverse effects

The most common adverse effects are directly related to the stimulation of the vagus nerve, and predominantly occur only during the on portion of the duty cycle (64). However, complications from surgery can also occur. A retrospective series including 105 patients and 118 vagus nerve stimulation operations over 10 years in a German hospital reported 6 patients having complications from the operation (72). Complications included wound infection, poor wound healing, and nerve injury. In the randomized controlled trials, the effects of nerve stimulation have included voice alteration, cervical paresthesia, dyspnea, coughing, hoarseness, and throat pain and occurred in 20% to 60% of subjects. We have found that cough and hoarseness can be substantially reduced by lowering the stimulus frequency to 25 Hz and lowering the pulse width to 250 µsec. Less frequently, dyspepsia, exacerbation of sleep apnea or of a swallowing difficulty occurs (61). Preliminary evidence of weight loss due to vagus nerve stimulation was found in a retrospective review of 32 patients at one center. A loss of 5% or more of body weight from prevagus nerve stimulation baseline was found in 25% of patients, and no patient had consistent weight gain (08). However, a series of 21 patients at another center found no significant change in weight over a 2-year period (42). One convincing case of vagus nerve stimulation-induced diarrhea has been reported (69). The more serious adverse effects are rare and related to surgery. They include left vocal cord paralysis, left facial nerve paralysis, left hemidiaphragm paralysis, infection, and hypesthesia. Vagus nerve stimulation does not produce any significant change to the cardiac function as assessed by electrocardiography, including prospective 24-hour recordings (65). However, bradycardia and asystole have occurred during the intraoperative testing of the neurocybernetic prosthesis for several patients (01). This was without any permanent consequence in all cases, but continuous electrocardiographic monitoring is warranted in during initial stimulation. One patient with post-implantation stimulation-related bradyarrhythmia has been reported (02). The bradyarrhythmias first arose more than 2 years after implantation and have been manifested by syncopal episodes. The basis for this problem has not been determined. There are additional cases of syncope, which have been reported (11; 62; 09; 38). In 1 case, syncope was the presenting symptom for third degree heart block caused by vagus nerve stimulation (38). There is a case report of stridor developing after anterior cervical discectomy and fusion through a right-sided approach in a patient with vagus nerve stimulation, which disappeared when the vagus nerve stimulation was turned off (36). Presumably this happened due to right vocal cord paralysis from the surgery and intermittent vocal cord paresis form the vagus nerve stimulation.

There is some evidence of safety of combination therapy with vagus nerve stimulation and either responsive neurostimulation or deep brain stimulation, with no ill effects in 11 patients receiving dual therapy (26).

Prognosis of vagus nerve stimulation is discussed in the Outcome section.

Special considerations

Vagus nerve stimulation has no known teratogenic effect and does not appear to alter fertility or pregnancy.

Clinical vignette

A 36-year-old man developed epilepsy at 15 years of age after approximately 4 days of status epilepticus and coma without a clear precipitant. After recovery from this, he developed complex partial seizures that typically follow an aura of déjà vu or fear and are manifested by staring and lip smacking. Secondarily generalized seizures occur rarely. Because the seizures continued at an average frequency of 5 to12 times monthly despite trials of 6 antiepileptic medications, an evaluation for epilepsy surgery was performed. This demonstrated independent right and left temporal lobe ictal onsets, bilateral hippocampal atrophy with signal abnormality on MRI, left temporal hypometabolism on PET, and severe impairment of both verbal and nonverbal learning. The memory deficit initially was noted on the patient’s recovery from status epilepticus and coma. The evaluation indicated that the patient was not an epilepsy surgery candidate, so he was offered vagus nerve stimulation. The procedure was considered, in part, because of a desire to maximize memory function through minimization of both seizures and antiepileptic drugs, if possible.

The patient’s best seizure control was during the months prior to vagus nerve stimulation implantation, when seizures averaged 2 to 8 monthly on a combination of 2 antiepileptic drugs that were new to him. The seizures became less intense during the ramping up of the neurocybernetic prosthesis output current. Four seizures occurred in the month following implantation, and the patient became seizure free 4 months following implantation. This lasted until 1 year ago (6 years from implantation) when auras returned at a frequency that increased to the present frequency of 2 to 3 monthly. Because of these breakthrough seizures and concern for a recurrence of complex partial seizures, the stimulation duty cycle was increased from 10% to 29%, but the aura frequency has not changed. Overall, the patient is highly satisfied with the treatment and is not experiencing any stimulation side effects. He continues on the same antiepileptic medications, but is experiencing less memory impairment and drowsiness. One year after implantation, he obtained a driver’s license.

Scientific basis

The mechanism by which vagus nerve stimulation reduces seizure frequency apparently relies on at least some of the widespread neural connections between the vagus nerve and regions of the brainstem, midbrain, diencephalon, and forebrain (79). Approximately 80% of the vagus nerve fibers are afferent and primarily project to the nucleus of the solitary tract. Stimulation of the vagus nerve or the nucleus of the solitary tract may either directly synchronize or desynchronize EEG (40). The specific effect depends on the frequency of stimulation whether stimulation of the nerve is sufficiently intense to activate both myelinated A and B fibers and unmyelinated C fibers. The issue of fiber type activation is bypassed by direct stimulation of the nucleus of the solitary tract, which produces desynchronization at frequencies over 30 Hz and synchronization at frequencies between 1 to 17 Hz. However, the relevance of this effect to understanding the mechanism of vagus nerve stimulation is limited because few studies of the antiepileptic effect of nucleus of the solitary tract stimulation have been performed. One reported result is that the nucleus of the solitary tract stimulation interferes with feline amygdaloid kindling in cats (49).

The effect of vagus nerve stimulation on the EEG is complicated because of the different fiber types within the nerve. Early animal experiments indicated that C fiber activation produced desynchronization and vagus nerve stimulation effectiveness was proportional to the degree of C fiber recruitment. This was contradicted by rat experiments that demonstrated that the vagus nerve stimulation effect is independent of C fiber function (83; 44), and vagus nerve stimulation-induced slow hyperpolarization of cortical pyramidal cells diminishes with activation of C fibers (86). Electrophysiologic recordings of the human vagus nerve during vagus nerve stimulation implantation have provided complimentary evidence that the human vagus nerve stimulation effect is through A and B fibers only (41). Overall, the data based on fiber type recruitment are discordant and do not readily lend itself to an explanation of the efficacy of vagus nerve stimulation.

The nucleus of the solitary tract and the locus coeruleus are brainstem structures that each may have a role in the antiepileptic effect of vagus nerve stimulation. Increases in GABA transmission or decreases in glutamate transmission in the mediocaudal nucleus of the solitary tract reduce the severity of limbic motor seizures in rats (80). Thus, the balance of excitation and inhibition within the nucleus of the solitary tract may affect seizure susceptibility. The locus coeruleus is indirectly connected to the nucleus of the solitary tract and has been found to have an increased discharge rate in rats during vagus nerve stimulation (30). Although the changes within the locus coeruleus during vagus nerve stimulation are not known, chemical ablation of the locus coeruleus attenuates the seizure suppressing effect of vagus nerve stimulation (43). This is consistent with the recognized antiepileptic effect of norepinephrine and may be a more downstream neurochemical mechanism (24). The hippocampus is one potential downstream location, and vagus nerve stimulation efficacy in a rodent model of limbic epilepsy has been found to be dependent on vagus nerve stimulation’s increase to hippocampal norepinephrine concentration (63). This furthers the support for this anti-seizure mechanism for vagus nerve stimulation.

Evoked potential recordings have shown a major pathway from the nucleus of the solitary tract to the intralaminar and ventroposterior thalamus via the parabrachial nucleus (16). As measured by [O-15]water PET, vagus nerve stimulation produces regional cerebral blood flow changes within the thalamus bilaterally that correlate to changes in seizure frequency (34). This increase in thalamic activity is the most consistent finding among the other rCBF and functional MRI studies of chronic vagus nerve stimulation. (58; 33). A perfusion SPECT study found significant correlations between right amygdalar and hippocampal perfusion changes and seizure control (77).

Intracranial and scalp EEG recordings have demonstrated vagus nerve stimulation associated changes that may provide some mechanistic insights. In one patient, the frequency of epileptiform spikes and sharps were recorded from the hippocampus with depth electrodes with vagus nerve stimulation at 5 Hz and 30 Hz (60). The change in vagus nerve stimulation frequency did not alter the frequency of spikes but did affect the frequency of sharps. Sharp discharge frequency decreased from baseline with 30 Hz stimulation and increased from baseline with 5 Hz stimulation. This is consistent with the observation that the antiepileptic vagus nerve stimulation effect occurs with stimulation above approximately 20 Hz. If the physiological difference between epileptiform spikes and sharps is their degree of synchronization, then perhaps vagus nerve stimulation is limited to affect activity only up to a certain degree of synchronization.

A transcranial magnetic stimulation study also has demonstrated vagus nerve stimulation associated changes with possible mechanistic implications. In a series of 5 patients, stimulation produced a significant increase in intracortical inhibition without a change in resting motor threshold (17). This suggests a GABAa mechanism. Within this series, the patient without a change in intracortical inhibition was the only one without an improvement in seizure control.

Other investigations have identified vagus nerve stimulation-induced synchronization or a combination of synchronization and desynchronization. Intracranial EEG recordings from 3 patients demonstrated increased EEG power and decreased complexity, indicating greater synchronization (82). Serial EEGs performed on 21 patients during vagus nerve stimulation therapy showed an early increase in clustering of the interictal epileptiform discharges for the 5 patients with highly frequent interictal epileptiform discharges prior to vagus nerve stimulation and a later reduction in interictal epileptiform discharge frequency and duration for all 21 patients (40). The investigator interprets the initial clustering of interictal epileptiform discharges as evidence for initially increased synchronization with progressively greater desynchronization over months. This novel observation of a combination response may lead to new insights into the physiologic basis of vagus nerve stimulation. In a study that did not follow patients over time, the stimulation period was correlated with a decrease in interictal epileptiform discharge frequency (45). Scalp EEG was used to assess interictal functional connectivity in 19 patient with chronic vagus nerve stimulation. The degree of resting state synchronization was lower in responders compared to nonresponders in both ON and OFF periods. Moreover, this degree of synchronization was lower in the ON compared to the OFF period (06). A study evaluated 5 drug resistant epilepsy patients with vagus nerve stimulation undergoing stereotactic EEG to assess synchronization and functional connectivity. Synchronization was seen during ON period in 4 patients. The 1 patient with a decreased synchronization in the ON period was the clinical responder. There was no change in functional connectivity with increased stimulation (03). These studies suggest the role of vagus nerve stimulation in the synchronization or desynchronization depending the degree of functional connectivity and clinical response. More controlled studies are required to study this mechanism in a systematic manner.

In part because the antiepileptic effect of vagus nerve stimulation increases slowly over months to years, vagus nerve stimulation has been investigated for a possible antiepileptogenic effect that is distinct from its acute effects. Kindling experiments during vagus nerve stimulation support an antiepileptogenic effect with a marked delay in the kindling process (22; 49). Indeed, none of the animals subjected to vagus nerve stimulation reached a stage 6 seizure within the framework of the study. At present, there is no evidence for an antiepileptogenic effect in humans. Based on a statistical analysis of “seizure loads” in patients who were involved in the clinical trials leading to FDA approval of the device, Dasheiff and colleagues concluded that vagus nerve stimulation does not “unkindle” seizures (12). However, the demonstration of kindling in humans has been tenuous, and “unkindling” has not been defined in animals (such as by demonstration of progressively increasing afterdischarge thresholds and decreasing afterdischarge durations). Moreover, it is generally understood that the changes produced by kindling are permanent. Although the analysis by Dasheiff and colleagues is interesting, it is not evidence to suggest that some plastic changes are not involved in contributing to the efficacy of vagus nerve stimulation.