- Top of page
Summary: Purpose: Approved neural-stimulation therapies for epilepsy use prolonged intermittent stimulation paradigms with no ability to respond automatically to seizures.
Methods: A responsive neurostimulator that can automatically analyze electrocortical potentials, detect electrographic seizures, and rapidly deliver targeted electrical stimuli to suppress them was evaluated in an open multicenter trial in 50 patients, 40 of whom received responsive cortical stimulation via subdural electrodes implanted for epilepsy surgery evaluations.
Results: Four patients, ages 15 to 28 years, monitored at three institutions, with clinical and electrographic response to neurostimulation, are described. Electrographic seizures were altered and suppressed in these patients during trials of neurostimulation lasting ≤68 h, with no major side effects. In one patient, stimulation appeared also to improve the baseline EEG.
Conclusions: Responsive cortical neurostimulation may be a safe and effective treatment for partial epilepsy. This information was derived from a small group of patients in an observation study. A double-blind, controlled Food and Drug Administration (FDA)-approved study of a permanently implanted responsive neurostimulation system to treat medically refractory partial seizures is under way.
The vagus nerve stimulator (VNS) is the only approved treatment modality for epilepsy using electrical stimulation (1). Patients with auras can activate the stimulator by using a magnet, but they must have the magnet nearby and recognize their aura. Reports of patient activation of the VNS suggest benefits in selected individuals; no controlled trials of patient activation have been performed. VNS is postulated to act via polysynaptic brainstem and cerebral pathways to modulate background synchronization (2). Other human studies using electrical stimulation of the cerebellum, thalamus, and subthalamic nucleus either are inconclusive or are still being evaluated (e.g., anterior thalamus) (3–9). All of these stimulus modalities are nonresponsive.
Electrical cortical stimulation has terminated induced electrographic epileptiform activity as well as spontaneous epileptiform activity in animal and human studies. The earliest report of cortical stimulation being applied in response to spontaneously occurring epileptiform activity in humans is by Penfield and Jasper in 1954 (10). Subsequently, in 1983, interictal spikes from epileptic foci in cats were suppressed by stimulation applied to the caudate nucleus, whereas random stimulation had no effect (11). In another study of slices of rat hippocampi, a computer-controlled system applied current to the stratum pyramidal on detection of spontaneous “abnormal” electrical activity (12). Other investigators reported that elicited and spontaneous epileptogenic discharges arising in human cortex can be terminated by responsive stimulation (13,14). More recently, a computer-controlled stimulator successfully terminated spontaneously arising epileptiform activity in subjects monitored with intracranial electrodes (15).
A multicenter study was recently completed to evaluate the safety of an external responsive neurostimulator (eRNS; NeuroPace, Inc., Mountain View, CA, U.S.A.). Ultimately, the intent is to investigate whether an implantable device with similar capabilities can reduce or abort spontaneous seizures. This device uses seizure-detection algorithms that can be tuned to patient-specific epileptiform activity to deliver electrical stimulation through implanted subdural electrodes at the epileptic focus to reduce or abort seizure activity (5–9). The potential benefits of such a device include the ability to deliver therapy when epileptiform activity occurs, the avoidance of side effects from anticonvulsants (AEDs), and the provision of an option for patients with medically intractable epilepsy in whom epilepsy surgery either has failed or was deemed too risky (14–16). In this study, patients undergoing implantation of invasive electrodes for seizure localization as part of an assessment for epilepsy surgery were connected to the eRNS device. By December 2003, 50 patients were enrolled in the study, and 40 of them received responsive neurostimulation (17,18). This study was primarily done to establish safety and preliminary information for future trials. Implantation of an internal responsive neurostimulator (RNS) system is currently under way as a clinical trial in 2004. We report the electrographic and clinical response to responsive neurostimulation in four patients, aged 15 to 28 years, undergoing subdural electrode monitoring. These cases represent several of the cases with interesting clinical or electrographic responses or both that occurred in this trial.
- Top of page
All four patients were admitted for intracranial monitoring to localize seizure onset for potential resection and had the device programmed to stimulation mode after the clinical evaluation was completed. All patients were studied over a 9-month period from November 2002 until August 2003. This study was approved by the Food and Drug Administration (FDA) and the Institutional Review Board (IRB) of each respective institution including the Western Institutional Review Board. All patients provided informed consent.
A photograph and block diagram of the eRNS System is shown in Figure 1. The eRNS is a battery-operated desktop device with four sense amplifiers that differentially amplify the inputs from up to eight contacts. The amplifiers have a bandwidth from 3 to 90 Hz. The signals are digitized at 500 Hz. The digital data are processed in real time by using algorithms consisting of up to three detection tools operating on one or two channels. The three detector tools are designated as half-wave, line length, and area (19). The half-wave tool can be viewed either as a pattern-recognition algorithm or as a computationally efficient method of detecting a specific onset frequency. Line length and area tools detect increases in complexity or energy from background trends. Tool outputs can be logically manipulated to tune the detection criteria to a patient's specific epileptiform activity.
Figure 1. A: External responsive neurostimulator (eRNS). B: Block diagram of system.
Results are periodically examined by the microprocessor to determine whether detection has occurred. The microprocessor then controls therapy and the storage of electrocorticograms (ECoGs) and diagnostics. The device has a 32-min ECoG buffer. When detection occurs, the device can deliver responsive therapy consisting of biphasic, charge-balanced, electrical pulses. The programmable ranges for stimulation parameters are as follows: frequency, 1 to 200 Hz; current, 0.5 to 12.0 mA; pulse width, 40 to 1,000 µs; stimulation pathway (any combination of the eight connected contacts), burst duration (≤5 s), and number of bursts, 1–5. After a pulse-train therapy has been delivered, a redetection algorithm determines whether the epileptiform activity is still present. If the activity is still present, up to four additional bursts of therapy can be delivered per episode. Various safety features including a “prescribed therapy limit” keep the amount of stimulation delivered within safe boundaries. The closed-loop stimulation in this investigation is performed only after the presurgical evaluation is complete and the patient is waiting for removal of intracranial electrodes. The period of such closed-loop stimulation is therefore limited.
A 15-year-old girl was admitted for localization and resection of her epileptogenic focus. Seizures began at age 5 years. Seizures were occurring weekly despite trials of carbamazepine (CBZ), valproic acid (VPA), and levetiracetam (LEV). Clinical manifestations of her seizures included aphasia, head deviation to the right, shortness of breath, sense of dissociation, and right-hand clenching lasting 60 s. Examination revealed a dense right homonymous hemianopsia but no hemiparesis. Magnetic resonance imaging (MRI) revealed a large left posterior cerebral artery distribution infarct, and multiple EEGs showed left occipital and temporal slowing.
Subdural electrode arrays were implanted, including an 8 × 8 grid over the left frontotemporal region, covering the sylvian fissure, a 4 × 5 grid over the region of infarction (parietooccipital), and two 1 × 6 strips covering the mesial temporal lobe. Ictal recordings revealed epileptiform activity that appeared to be emanating independently at different times from three separate foci: inferior frontal, angular gyrus, and the inferior temporal gyrus. Unfortunately, the former two were functionally determined to be near both the anterior and posterior speech areas, based on the results of cortical mapping.
During her 7-day monitoring period, eight seizures occurred, mostly at the end of her monitoring period. No subclinical (electrographic only) seizures occurred. For 41 h total, the eRNS was connected, and after several seizures, was tuned to detect accurately a typical electrographic seizure by using the inferior frontal region as the primary detection area. Before engaging responsive stimulation, manually delivered impulses were used to assess the patient's tolerance. Initial stimulation, when increased to 8 mA, produced some tongue tingling that resolved when the current was decreased to 7 mA. The final programmed stimulation parameters were as follows: current, 7 mA; 95-ms burst duration; 280-μs pulse width; pulse count, 20; and pulse frequency, 200 Hz. The anode was four contacts on the anterior row of the posterior 4 × 5 grid; the cathode was four electrodes in the anterior region of the frontotemporal 8 × 8 grid (Fig. 2). Both regions were active electrographically and were thought to be functional for language, as described previously.
During responsive stimulation, 24 therapies were delivered for a cumulative total of 2.4 s of stimulation with no clinical seizures occurring. With several stimulations, clear evidence was seen of electrographic resolution of seizure activity (Fig. 3). After 7.2 h, the device reached a prescribed therapy limit (four stimulated events before it required resetting) and did not deliver further stimulation. Fifty minutes later, the patient had a typical clinical seizure. The subdural grid was removed 8 h later, and an inferior temporal lobectomy was performed to remove the electrographically active region not involved in speech. Now 6 months after surgery, the subject remains seizure free.
Figure 3. Patient 1: Nonstimulated electrographic seizures (left) and stimulated activity (right), as recorded from external responsive neurostimulator (eRNS).
Download figure to PowerPoint
A 24-year-old man was admitted for epilepsy surgery after five AEDs failed. Seizures began at age 3 years, 1 year after a concussion with posttraumatic amnesia. Complex partial seizures were occurring monthly and were described as an unusual sensation with tachycardia, followed by staring, fumbling, and head turning to the left, with auras occurring more frequently. MRI revealed bilateral hippocampal hyperintensity, and ictal single-photon emission computed tomography (SPECT) showed increased activity in the right temporal and parietal regions.
Eight-contact depth electrodes were placed in the left and right hippocampi and temporal lobes via occipital insertion. Over a 5-day period, all medications were tapered and discontinued, and 14 clinical and two subclinical seizures occurred. Seizures consistently occurred during the period from 4 a.m. to 6 a.m. each morning. All seizure onsets were localized to the left temporal depth electrode contacts 3 and 4, with contact 1 most proximal.
Stimulation was set at 95-ms burst duration, 200-μs pulse width, and 200-Hz pulse frequency. At 10 mA, the patient felt a cheek sensation that resolved when the stimulation was reduced to 6 mA. Electrode contacts initially used for stimulation were left 3 and 4 (anode) and 5 and 6 (cathode). The next morning, the patient complained of seeing a flash, so the stimulus current was lowered to 5.5 mA, and the stimulus contacts were changed to a more proximal final location (2 and 3 as the anode, 4 and 5 as the cathode) with no further visual phenomena.
Responsive stimulation was delivered 960 times over a period of 66 h for a total of 97.7 s of total stimulation. No clinical seizures occurred during this time period nor were side effects noted (Fig. 4). Apparent electrographic activity was stimulated and altered (Fig. 5). Before discontinuing neurostimulation, the patient was loaded intravenously with AEDs and then taken to the operating room for depth electrode removal.
A 23-year-old woman was admitted for subdural grid implantation to localize seizure onset, after having tried four AEDs with no success. Seizures began at age 19 years with no clear etiology. Seizures were occurring 10 times per day, most at night, and were described as a head turn to the left followed by a fencing posture with right arm extension and confusion. MRI revealed a small cyst in the left putamen, which was not thought to be significant.
Subdural electrode arrays were implanted including an 8 × 8 grid over the left frontal lobe and sylvian fissure, a 4 × 5 grid over the parietooccipital region, a 2 × 8 grid placed interhemispherically to the left of the falx cerebri, and two 1 × 8 strips covering the orbitofrontal and mesial temporal regions. Ictal recordings demonstrated an epileptic focus at the posterior four contacts of the interhemispheric grid.
Initial stimulation parameters were 3-mA current intensity, 95-ms burst duration, 240-μs pulse width, and 200-Hz pulse frequency. The three superior posterior contacts on the interhemispheric grid were used as the anode, with the three contacts immediately below or inferior as the cathode. By using this stimulation configuration, seizures were detected and stimulated, but none was aborted. The next day, a second configuration was designed: the posterior six electrodes as an anode, with two contacts in the center of the left frontal 8 × 8 grid as the cathode. The stimulus current was increased to 6 mA. Some evidence was noted of electrographic efficacy in this configuration, but the patient subsequently experienced a seizure that tangled the external wiring. Therapy was disabled until the next day. The final stimulation parameters used a pulse width of 280 μs. During the next 6 h, 20 responsive stimulations and no seizures occurred. An aborted seizure as recorded from the eRNS is shown (Fig. 6). One hour after discontinuing the stimulation in preparation for surgery, the patient had two clinical seizures. A limited resection was performed without any clear seizure improvement.
A 28-year-old man was admitted for seizure-onset localization. Blunt head trauma at age 18 years caused a left subdural hematoma requiring evacuation and subsequent seizures. Although he was seizure free with AEDs for 6 years, seizures later became intractable. Complex partial seizures causing a behavioral arrest and unresponsiveness occurred 2 to 3 times per day, and generalized tonic–clonic events occurred 10 times per year. MRI revealed a region of encephalomalacia in the left temporooccipital region. Positron emission tomography (PET) and interictal EEG suggested a left temporofrontal focus. Wada testing was concerning for left-sided language and memory.
A grid (single left temporooccipital, 8 × 8), eight strip electrodes (1 × 8 and 1 × 10 covering the frontal, temporal, and occipital poles, and the interhemispheric regions), and two depth electrodes (eight and 12 contacts into the hippocampus and eight contacts into the region of encephalomalacia) were placed on the left (Fig. 7). Approximately 15 to 20 subclinical seizures per day were detected, with stereotyped gamma activity was seen over the anterior superior temporal grid and the center of the bordering superior 1 × 10 strip (LSPF). Five clinical seizures occurred before activation of responsive neurostimulation.
Figure 7. Patient 4: subdural electrode placement (asterisk, = depth electrodes, straight black line with += cathode, −= anode).
Download figure to PowerPoint
Stimulation parameters were set at 10-mA stimulus current, 95-ms burst duration, 240-μs pulse width, and 200-Hz stimulus frequency. The anode was contacts four to seven in the LSPF strip, and the cathode was four contacts in the second row of the temporal grid (Fig. 7). RNS was available for a total of 43 h over a 3-day period, and was available during the days but not every night. No clinical seizures occurred during this time period, and 223 therapies were delivered (total stimulation time, 21.2 s). Interestingly, seizures did not occur during 8-h nighttime periods when stimulation was not available. The EEG appeared to improve from baseline once stimulation began (Fig. 8). The patient was taken to the operating room, and the electrodes were removed. A resection of the encephalomalacia and a subpial transection in the region where the eRNS appeared to suppress seizures (believed to be eloquent cortex) were performed. To date, seizures remain unchanged.
Figure 8. Patient 4: EEG at baseline (A) and after 2 days of neurostimulation (B), as recorded from the external responsive neurostimulator (eRNS).
Download figure to PowerPoint
- Top of page
These four case reports provide some evidence that RNS alters and may terminate seizures. During the periods, ranging from 6 to 68 h, when the eRNS was programmed to deliver stimulation, electrographic seizures were detected, and some were apparently terminated. Stimulation seemed to reduce the numbers of seizures over both temporal and extratemporal locations, by using both strip and depth electrodes. In patient 3, an apparent clinical effect was not seen until stimulation parameters were modified, perhaps indicating that stimulation parameters require individual adjustment. Interestingly, in the fourth patient, the subdural EEG recordings appeared to normalize after several days of stimulation, suggesting a possible neuromodulatory effect of this repetitive direct brain stimulation. In addition, patient 1 seemed to have had electrographic change even though the current may not have been directly over the epileptogenic cortex.
Two patients had brief, transient side effects. In the first patient, tongue tingling stopped when the stimulation current was reduced. The second patient had both facial tingling and visual flashes that stopped after reducing the stimulation current and selecting different stimulation contacts. When stimulation was perceived, it was not painful or intolerable. The other two patients had no perception of stimulation.
This study was not designed to test efficacy of a permanently implanted responsive cortical neurostimulator in ambulatory patients with epilepsy. The study was nonblinded, and the primary outcome measure was safety; results from the entire study in regards to safety will be published separately. The study was conducted in the setting of an inpatient admission for intracranial EEG monitoring. During the course of the study, AEDs were not held constant, and patients received other medications likely to alter the seizure threshold, such as analgesics, steroids, and antibiotics. Responsive stimulation could be enabled only at the completion of the epilepsy surgery evaluation and before electrode deplantation, which allowed only a short period of testing. In this interval of time, it is not possible to exclude the possibility that improvement in seizures and/or electrographic discharges with stimulation was serendipitous or a response to some other therapeutic manipulation. However, direct observation of the electrocorticographic response to stimulation suggests discharges were favorably altered in some patients.
In summary, eRNS appeared to be safe and well tolerated in these four patients. Analysis of these individual cases suggests that some patients may have suppression of their seizures, and in some situations, repetitive stimulation may suppress further events and normalize the background activity. As with other types of neurostimulation, responsive stimulation has the benefits of lack of drug-related CNS side effects. Potentially, if efficacy is demonstrated in more extensive trials, the RNS system would be applicable in a variety of patients with focal epilepsy, including failures of previous seizure surgery, and in those patients with seizures originating from eloquent areas.