Address correspondence and reprint requests to Dr. B. J. Fisch at Department of Neurology, Louisiana State University Medical Center in New Orleans, 1542 Tulane Avenue, New Orleans, LA 70112, U.S.A. E-mail: firstname.lastname@example.org
Summary: Purpose: To assess the effect of vagus nerve stimulation (VNS) on interictal epileptiform activity in the human hippocampus. Clinical studies have established the efficacy of vagus nerve stimulation in patients with epilepsy (VNS Study Group, 1995), although the electrophysiologic effects of VNS on the human hippocampus and mesial temporal lobe structures remain unknown.
Methods: We report a case study in which a patient with an implanted VNS underwent intracranial electrode recording before temporal lobectomy for intractable complex partial seizures. Epileptiform spikes and sharp waves were recorded from a depth electrode placed in the patient's left hippocampus. Spike frequencies and sharp-wave frequencies before and during VNS were compared using both a 5- and a 30-Hz stimulus. Different stimulation rates were tested on different days, and all analyses were performed using a Student's t test.
Results: We found no significant differences in spike frequency between baseline periods and stimulation at 5 and 30 Hz. In contrast, stimulation at 30 Hz produced a significant decrease in the occurrence of epileptiform sharp waves compared with the baseline, whereas stimulation at 5 Hz was associated with a significant increase in the occurrence of epileptiform sharp waves.
Conclusions: VNS produces a measurable electrophysiologic effect on epileptiform activity in the human hippocampus. Although a clinical response to VNS did not occur in our patient before surgery, 30-Hz VNS suppressed interictal epileptiform sharp waves that were similar in appearance to those seen during the patient's actual seizures. In contrast, 5-Hz stimulation appeared to increase the appearance of interictal sharp waves.
Vagus nerve stimulation (VNS) using the Neuro-Cybernetic Prosthesis (NCP; Cyberonics, Inc., Houston, TX, U.S.A.) was approved by the Food and Drug Administration (FDA) in July 1997 for use as an adjunctive therapy for adults or adolescents older than 12 years with medically refractory partial onset seizures. Although its efficacy is well established (1,2), the exact mechanism of action of VNS remains unknown. Animal studies have shown that stimulation of the cervical portion of the vagus nerve can terminate electrographic seizures caused by either topical cortical or systemic strychnine or pentylenetetrazol (PTZ) (3). VNS can also block the development of kindled seizures (4).
McLachlan (5) found that electrical stimulation of the left vagus nerve reduced or abolished penicillin-induced interictal cortical spikes in Wistar rats during and immediately after stimulation. In his study, stimulation consisted of square-wave pulses of 0.01–1.2 mA at 20 or 50 Hz. Woodbury and Woodbury (6) studied the effects of different stimulus parameters and found that stimulation of unmyelinated (C) vagus nerve fibers in male Sprague–Dawley rats could inhibit electrically or chemically induced seizures. They concluded that the optimal stimulus frequency was ∼10–20 Hz. Based on these animal studies, we hypothesized that VNS stimulation could have a direct effect on spontaneous epileptiform activity in the human hippocampus and that the effect might vary according to stimulus frequency. The purpose of this study was, therefore, to determine if such an effect exists and, if so, to examine potential differences in stimulation at different frequencies. In the following report, we describe the effects of VNS on various forms of epileptiform activity (spikes and sharp waves) recorded from the left hippocampus in a patient with left temporal onset seizures who had failed to show improvement during a 10-month trial of VNS.
Human scalp EEG recordings have failed to show any effect of VNS on interictal epileptiform activity or background activity. Hammond et al. (7) studied nine patients with medically intractable seizures participating in a clinical trial of prolonged VNS. Stimulation at various stimulus frequencies and amplitudes had no noticeable effect on EEG background activity regardless of whether the patient was under general anesthesia, awake, or asleep. However, VNS did interrupt ictal EEG activity. Similarly, Salinsky and Burchiel (8) found no VNS effect on background activity. After our initial description of the VNS effects on intracranial epileptiform activity (9), there has been only one other study of intracranial EEG and VNS. Thompson et al. (10) found VNS induced changes in EEG spectra in direct cortical and thalamic recordings. They did not analyze interictal or ictal epileptiform activity or EEG activity recorded from the hippocampus.
A 17-year-old boy began having seizures at age 12 years. Video and EEG scalp monitoring performed at age 15 showed left hemispheric, probably left temporal, seizure onset. Because of medical intractability and multifocal hamartomas, the patient was implanted with a vagus nerve stimulator. During the next 10 months, his seizure frequency was reduced by only 15%. He was therefore referred for surgical evaluation. Stereotactic insertion of one, eight-contact, left hippocampal depth electrode and three, six-contact subdural strip electrodes (left frontal, left anterior temporal, and left posterior temporal; Fig. 1) was performed at the LSU Baptist Hospital Memorial Medical Center Epilepsy Center in New Orleans. Before surgery, the vagus nerve stimulator was turned off. On postoperative day 1 (day 1), the patient's antiepileptic medications [AEDs; felbamate (FBM) and carbamazepine (CBZ)] were discontinued. FBM, 1,200 mg, p.o., twice a day, was reinstituted on the evening of the second postoperative day (day 2) and increased to 1,800 mg, b.i.d., on the third postoperative day (day 3) because of increasing seizure frequency. The serial blood levels of FBM were 17.0 (day 2), 21 (day 3), and 56.0 μg/mL (day 5). CBZ (Tegretol XR) was restarted on day 3 at the dose of 800 mg by mouth twice a day. The serial blood levels of CBZ were 2.5 (day 2), 2.1 (day 3), and 8.4 μg/mL (day 5).
Epilepsy monitoring demonstrated left anterior temporal seizure onset with rapid spread to the left frontal lobe. Five complex partial seizures were recorded on day 2, and two, on the morning of day 3. Electrodes were removed on day 6, and a left temporal lobectomy was performed. The patient had no change in seizure frequency after the surgery (follow-up of >18 months).
The VNS device was reactivated temporarily after sufficient information had been obtained during monitoring to evaluate the patient for possible surgical intervention. The evaluation of VNS and EEG epileptiform activity was performed using visual inspection of data acquired with a Telefactor monitoring system (Astro-Med, Inc., West Warwick, RI, U.S.A.) in the LSU Memorial Medical Center Epilepsy Monitoring Unit. A manual count of the spikes and sharp waves in the left temporal hippocampal depth electrode was performed by two electroencephalographers (P.O. and B.F.) certified by the American Board of Clinical Neurophysiology. The reviewers were not blinded to the condition of stimulation. The intraobserver correlation was 90% for the spikes and 80% for the sharp waves. A spike was defined as a sharply contoured waveform with a duration of 20–70 ms. A sharp wave was defined as a transient of 70–200 ms with an amplitude of ≥200 μV. Spike and sharp wave frequencies during “on” and “off” VNS times were calculated. Simultaneously recorded activity in needle scalp electrodes and the ECG channel was used to detect the electrical artifact produced by the VNS stimulator to monitor when the VNS device was in the “on” mode. Only epochs that appeared to be free of artifact were used.
On postoperative day 3, VNS was studied with the device programmed at the routine clinical settings of current output, 2.75 mA; frequency, 30 Hz; time on, 30 s; time off, 1.1 min; pulse width, 500 μs (Fig. 2). The absolute numbers and frequencies of spikes and sharp waves were analyzed in 18 epochs “on” (with stimulation) alternating with 18 epochs “off” (without stimulation).
On day 4, VNS stimulation was performed using a custom-designed adapter allowing precise artifact registration at a stimulus frequency limit of 5 Hz. An adapter with a resistive voltage divider was placed in series with a capacitive pulse-stretcher to make a longer-duration stimulus artifact. An output current of 2.75 mA; time on, 30 s; time off, 30 s; and pulse width, 500 ms were used (Fig. 3). Twenty-five epochs “on” alternating with 25 epochs “off” were analyzed.
The statistical data were analyzed using Student's t test. Each stimulation session was compared with a preceding “baseline” period that consisted of adjacent 30-s epochs “off.”
On day 3 of monitoring, the first baseline recording (baseline 1) was obtained between 16:00:00 and 16:10:30. This 10-min and 30 s recording (“off” only) was divided into twenty-one 30-s epochs, and the number of spikes and sharp waves occurring during each 30-s epoch was counted. The average spike frequency per epoch was 0.1 (SD, 0.085), and the average sharp wave frequency per epoch was 0.732 (SD, 0.287).
On day 3 of monitoring, intermittent 30-Hz stimulation was performed between 18:30:32 and 18:59:55. Thirty-second-long epochs “on” alternated with 1.1-min epochs “off.” Eighteen epochs “on” and 18 epochs “off” that appeared to be artifact free were saved for analysis. The average spike frequency for “on” periods was 0.194 (SD, 0.057) and for “off” periods was 0.242 (SD, 0.072). The average sharp-wave frequency for the “on” periods was 0.361 (SD, 0.167) and for the “off” periods was 0.478 (SD, 0.135). VNS was deactivated at 19:00 on day 3.
On day 4 the second baseline recording (baseline 2) was performed between 15:30:00 and 15:42:15. Similar to the first baseline recording, it was divided into 21 epochs (all “off”) of 30 s duration, and epochs with artifacts were excluded. The spike and sharp-wave frequencies were calculated for each 30-s epoch and then averaged. The average spike frequency was 0.111 (SD, 0.085). The average sharp-wave frequency was 0.571 (SD, 0.208).
Stimulation on day 4 was performed at 5 Hz between 18:37:18 and 19:06:11. Thirty-second “on” epochs alternated with 30-s epochs “off.” Only one epoch in the middle of the studied period had to be excluded because of the presence of artifacts. Twenty-five epochs “on” and 25 epochs “off” were analyzed. The average spike frequency for “on” periods was 0.157 (SD, 0.109), and for “off” periods, 0.137 (SD, 0.121). The average sharp-wave frequency for “on” periods was 0.836 (SD, 0.186), and for “off” periods was 0.719 (SD, 0.187).
The comparison of average spike frequencies is depicted in Table 1. There was no meaningful difference between spike frequency during 30-Hz stimulation for “on” and “off” epochs or between the baseline 1 period and either the “on” (p = -0.8795) or “off” epochs (p = 0.421). Similarly, stimulation at 5 Hz did not produce significant spike-frequency differences for “on” versus “off” epochs or for baseline period 2 versus either the “on” (p = 0.3953) or “off” epochs (p = 0.1178). VNS was deactivated at 19:10 on day 4, to be restarted on day 7 (1 day after left temporal lobectomy).
Table 1. Spike frequency comparison between the baseline recording and alternating VNS-on and VNS-off epochs
Comparison to baseline (t test p values)
The comparison of average sharp-wave frequencies is shown in Table 2. Stimulation at 30 Hz produced a significant decrease in sharp waves compared with the baseline period 1 for the “on” epochs (p = 5.363E-05) and for the “off” epochs (p = 0.001). Comparison of the “on” versus “off” periods also showed a significant reduction of sharp waves during the “on” periods (p = 0.05; Fig. 2). Stimulation at 5 Hz was associated with an increase in sharp-wave frequency, which was greater for “on” (p = -5.406E-05) than “off” epochs (p = 0.0165; Fig. 3), as compared with “baseline 2.”
Table 2. Sharp-wave frequency comparisons between the baseline recording and alternating VNS-on and VNS-off epochs
Comparison to baseline (t test p values)
VNS on epochs
VNS off epochs
Negative (−) p value means that average test value is greater than average baseline value.
Similar evaluations using the recording from needle scalp electrodes at electrode positions F7, T7, and P7 were not performed. As expected, the majority of interictal epileptiform events recorded from intracranial electrodes did not produce apparent changes in simultaneous recordings using the F7, T7, and P7 electrodes referenced to Cz.
Our findings suggest that vagus nerve stimulation has a direct electrophysiologic effect on the human hippocampus, and that VNS stimulation at 30 Hz may reduce the occurrence of hippocampal epileptiform sharp waves more effectively than stimulation at 5 Hz. Our findings are consistent with those of animal studies in which faster rates of stimulation appear to be more effective in blocking seizures (3,6) and in producing EEG desynchronization (11).
The selection of our patient was necessarily biased against an anticonvulsant VNS effect. Depth electrode implantation would not have been undertaken had VNS been effective in stopping seizures. The failure of our patient to respond to VNS gave us the opportunity to observe that stimulation of the vagus nerve does not necessarily produce an all-or-none effect on epileptic phenomena in the human hippocampus. That is, in the absence of a clinical VNS effect, there may still be a suppressant effect on epileptiform activity.
We consider the significance of our observations to be preliminary, not only because they are restricted to one individual, but because of limitations inherent in this kind of analysis. Both the recent implantation of intracranial electrodes and electrographic and clinical seizures preceding the comparison of VNS and epileptiform activity may have influenced the occurrence of epileptiform activity during baseline or stimulation periods (12). Given these circumstances, the relationship between baseline and “off” times were remarkably stable: spikes and sharp waves occurred at a statistically similar rate during both the baseline and “off” periods between trains of stimulation. A statistically greater suppression of sharp waves also occurred during trains of stimulation compared with the intervening “off” periods. It is generally recognized that a minority of individuals may show dramatically changing interictal epileptiform patterns during the first 3 days after electrode implantation. We attempted to minimize immediate postoperative effects by waiting until postoperative days 3 and 4 to begin baseline and stimulation observations. The occurrence of epileptiform changes can also vary according to sleep/wake cycles or circadian/ultradian factors. We attempted to limit circadian effects by performing all observations between 15:00 and 19:00 h on consecutive days. Even so, a comparison of the stimulation periods with the baselines could have been compromised by a potential order effect as well as unobserved biologic or environmental effects. The investigators were not blinded to the presence or absence of stimulation during analysis. However, both investigators were blinded to each other's observations, and their observations produced the same statistical conclusions. Finally, a slight difference in the duration of “off” times complicates the comparison of the effects of 5- vs. 30-Hz stimulation.
Although Hammond et al. (7) found one patient whose seizures appeared to be blocked by VNS, no other investigators have found an effect of VNS on human interictal epileptiform activity (9). This may be because no other investigators have used direct intracranial recording, which is far more sensitive than scalp recording. In our patient, this expected difference was confirmed by simultaneous scalp and intracranial recording. Our own experience, and that of others, indicates that detectable volume conduction from a hippocampal source in a routine scalp recording without spike averaging is extremely unlikely.
The morphology of sharp waves suppressed by VNS in our patient was highly similar to that seen during the patient's actual electrographic seizures. Interictal spikes that were unaffected by VNS in our patient did not appear as part of the ictal pattern. Seizure morphology remained highly consistent between attacks (Fig. 4). Despite the similarity between the interictal sharp waves and the components of the electrographic seizure discharge, VNS stimulation had little clinical effect on our patient: there was a negligible reduction in seizures during the 10-month VNS treatment period before intracranial recording.
Recent studies have suggested that some individuals who do not initially respond to VNS may gradually improve after 1–2 years of continued use. Therefore, we have continued to use VNS in our patient. Obviously, any assessment of the effect of VNS will be confounded by the effects of surgery, including a possible “running down” phenomenon. Our patient did not undergo positron emission tomography (PET) examination before VNS, but Henry et al. (13,14) have proposed that patients with greater VNS thalamic PET activation have a greater chance of seizure reduction with VNS.
The differential effect of VNS on epileptiform spikes versus sharp waves in our patient suggests that VNS did not exert a uniform effect throughout the epileptic irritative zone. The biophysical characteristics of volume conduction and time course observed in epileptiform spikes and sharp waves indicate that they arise from different groupings of neuronal cell populations. We believe that VNS may have had a greater effect on sharp waves because they are less highly synchronized events, and it is known that higher frequency VNS has a desynchronizing effect on the EEG. Similarly, Chase et al. (15,16) and Magnes et al. (11) have shown that slower rates of stimulation (1–17 Hz) can actually increase synchrony of certain EEG activities. This is consistent with our observation that sharp waves appeared to be enhanced at a slower (5-Hz) stimulus rate. In addition to the effects of VNS on ictal and interictal epileptiform activity, a variety of other changes occur in the EEG in animals (7,8,14–16). Zanchetti et al. (16) discovered that VNS blocked spindle-like activity and reduced the amplitude of the EEG background in cats. Woodbury and Woodbury (6) found that VNS could produce EEG desynchronization, whereas Chase et al. (15,16) and Rutecki (17) found a complex relation between VNS and EEG with either synchronization or desynchronization depending on the population of stimulated nerve fibers.
Animal studies suggest that a suppression of epileptiform activity during stimulation “on” periods might also persist into the “off” periods (3). However, in our patient the “on” and “off” conditions for the 30-Hz stimulation were significantly different for sharp waves. Although these periods differed, there was also a trend toward a decrease in sharp waves during the “off” periods compared with the baseline 1 recording. This trend may have been due to a persisting effect, or it could simply reflect an overall trend toward decreased interictal activity during the period of testing.
Thus far >7,000 patients have been implanted with vagus nerve stimulators. Currently, there is no clear way to predict the response to VNS. Depending on the therapeutic approach at different epilepsy centers, some patients who would require invasive recording have been encouraged to first try VNS instead of epilepsy surgery. In those cases in which VNS fails to suppress the majority of seizures, invasive presurgical evaluations are more likely to be undertaken. In those cases, we strongly encourage other investigators to extend our observations and those of others (10) on the functional anatomy of VNS and its effect on human epilepsy.