Responsive brain stimulation in epilepsy

Abstract

Stimulation devices are considered in patients with drug-resistant epilepsy and who are not surgical candidates. Responsive neurostimulation (RNS) is a cortically based stimulator activated by electrocorticography (ECoG) patterns. Stimulation is applied directly to the seizure focus. The vagal nerve stimulator AspireSR 106 is also a responsive device which, in addition to basal stimulation, is activated by tachycardia. Deep brain stimulation of the anterior nucleus of the thalamus is used in Europe for intractable epilepsy and yields similar response rates to RNS using duty cycle stimulation. Chronic subthreshold cortical stimulation is an experimental form of constant, low-level stimulation applied to a seizure focus. These modalities are discussed and compared in this review.

Introduction

Epilepsy is a common neurologic disease, affecting 1.2% of the US population.¹ In the majority of patients, epilepsy can be controlled with medication, but approximately one third have drug-resistant disease, defined as continued seizures in spite of two appropriate trials of medication.² Resective surgery can be considered if a seizure focus is localized. Surgery results in seizure freedom for over 60% of patients, with better outcomes in certain types of epilepsy.³ If there are multiple seizure foci, the focus cannot be localized, or when the seizure onset zone resides in the eloquent cortex then resection may not be an option. In these cases patients could be candidates for treatment with stimulation.

Currently available responsive stimulation devices include the responsive neurostimulator (RNS, NeuroPace Inc., Mountain View, CA, USA) and the vagal nerve stimulator (VNS, LivaNova Inc., Houston, TX, USA), responsive to tachycardia. A duty-cycle device that delivers intermittent stimulation is the deep brain stimulator (DBS, Medtronic Inc., Minneapolis, MN, USA) which is not available for epilepsy in the US but has the European CE mark. Chronic subthreshold cortical stimulation (CSCS) is an experimental form of stimulation not widely available and currently under investigation, which delivers continuous, low-level stimulation.

Responsive neurostimulation was approved in the US for drug-resistant focal epilepsy in 2013. This system consists of an implanted stimulator connected to one or two subdural strips or depth leads, each containing four electrodes (Figure 1). These are placed at seizure foci and deliver a stimulation in response to detected electrocorticography (ECoG) patterns. The pattern recognition is programmed by the physician and is tailored to the patient’s ictal ECoG patterns.⁴

Figure 1.

Responsive neurostimulator (RNS): the top left image is the stimulator attached to subdural and depth electrodes; the top right diagram shows placement of the stimulator in the skull; the bottom three images are radiographic images of lead placement and wires.

Vagal nerve stimulation has been approved by the US Food and Drug Administration since 1997 as an adjunctive treatment for refractory focal epilepsy. VNS pulses stimulation to the vagal nerve on the left. The frequency and timing of stimulation can be adjusted. In 2015 a new model was released which generates additional electrical stimulation in response to tachycardia.⁵

The responsive neurostimulation system

Responsive neurostimulation is based on the observation that electrical stimulation can abort epileptic afterdischarges during intracranial functional mapping.⁶ Functional mapping has been performed for decades to identify eloquent areas of the cortex. It consists of delivering stimulation to intracranial electrodes to localize brain function. During functional mapping the stimulation can produce afterdischarges which may evolve into clinical seizures.⁷ Delivering subsequent stimulation aborts the afterdischarges and prevents seizures from occurring (Figure 2). RNS also delivers electrical stimulation based on identified patterns but offers a shorter pulse of lower current activity compared with functional mapping.

Figure 2.

Afterdischarges seen following cortical stimulation during mapping. Repeat stimulation is given which aborts the afterdischarges.

Between one and two leads with four contacts each are placed subdurally or as depth electrodes at the seizure focus. The leads are connected to a stimulator which resides in the skull. The contacts can be programmed as anodes or cathodes, and the stimulator itself can serve as a cathode. Patients are supplied with a programmer in the form of a laptop which is connected to a wand. Following a seizure, patients use the wand to download ECoG data onto an online database for physicians to review. With this stored information the epilepsy providers personalize the stimulation to detect seizure ECoG patterns. The patterns detected by the implanted electrodes can then be programmed to trigger up to five bursts of electrical stimulation (0.5–12 mA) for 100–200 ms each (Figure 3). This stimulation can occur thousands of times per day.

Figure 3.

Detection of preprogrammed ECoG (electrocorticography) pattern followed by a train of five electrical stimulations which aborts the ECoG pattern.

Responsive neurostimulation is considered in patients for whom resective surgery is not possible but whose seizure focus or foci can be identified. Ideal patients include those with multiple foci such as bitemporal epilepsy, or in patients with a seizure focus in the eloquent cortex, such as motor or speech regions. Intracranial electroencephalography is often used to identify the seizure focus and confirm that the patient is not a surgical candidate. The RNS is implanted and programmed after an initial detection period.

The initial clinical trial for the RNS was a multicenter randomized controlled trial with 191 adult patients.⁸ After a 4-week baseline stabilization period there was a double-blind period involving sham or actual stimulation followed by an open-label period. During the 12-week blinded period seizure frequency was significantly reduced in the treatment group (down 37.9% compared with 17.3% in the sham group). The initial responder rate (defined as 50% or greater reduction in seizure number) was 29% in the treatment group and 27% in the sham group, but by month 3 the sham group response rate was 16% and the treatment group response rate was 27%. One year post implant during the open-label phase the responder rate was 43%.

Several noncontrolled longer-term studies with the same patient cohort have reported favorable long-term outcomes with continued reduction in seizures. Geller and colleagues reported on 111 subjects with mesial temporal epilepsy and RNS.⁹ These patients were followed for an average of 6 years, during which time disabling seizures were reduced by 66.5% and responder rate (>50% seizure reduction) continued to improve, achieving 64.6% at the 6-year mark. Forty-five percent of patients had a seizure-free interval of 3 months or more, and 29% had an interval greater than 6 months. A study of 126 patients with neocortical epilepsy showed a mean seizure reduction of 51–70%,depending on epilepsy type.¹⁰ Twenty-six percent of this group went 6 months or more without a seizure. These favorable long-term results suggest that the mechanism of this stimulation method may not be the immediate abortion of electrographic seizures by stimulation but rather a long-term neuromodulatory effect. They also suggest that RNS outcomes improve over time.

Adverse events with RNS included intracranial hemorrhage (4.7%), infection (5.2%), and lead damage (2.6%).^8–10 These percentages are similar to other implantable neurostimulators.¹¹ Eleven patients died during the 6-year study, including two suicides, one from status epilepticus, one from lymphoma and seven from suspected or definite sudden unexpected death in epilepsy (SUDEP).^9,10 This death rate is below the expected SUDEP rate in a highly refractory epilepsy population.

Several studies have looked at quality of life (QOL) and mood in patients using RNS.^4,12,13 The study of Meador and colleagues from 2015 used the quality of life in epilepsy inventory 89 (QOLIE-89) as well as depression scales. At 2 years, 44% of patients reported an improvement in QOL and 16% reported a decline. Patients with neocortical epilepsy fared better than those with mesial temporal epilepsy. There was also a modest improvement in depression scores at 2 years.

Cognitive dysfunction is a common comorbidity of epilepsy. Up to 70–80% of patients with epilepsy have cognitive complaints.¹⁴ The mechanism for this dysfunction is likely multifactorial, stemming from the epilepsy itself, seizure medications, underlying brain networking, and comorbid mood issues.¹⁵ Neurocognitive function has been studied in the setting of RNS use.¹⁶ At 2 years there were improvements in naming (23.5% of patients scored higher on a Boston Naming Test), no change in verbal learning, and no significant difference in overall memory based on the Rey Auditory Verbal Learning Test. Patients with neocortical temporal epilepsy had greater naming improvement compared with mesial temporal epilepsy.

Responsive vagal nerve stimulation

Vagal nerve stimulation has been widely available since 1997 as a treatment for drug-resistant focal epilepsy. Off label this device has also been implanted in patients with forms of generalized epilepsy. Previous models of VNS offered programmable cyclic stimulation to the vagus nerve in the neck. The proposed mechanism of action is downregulation of excitatory pathways, especially deep brain structures, through vagally mediated external stimulation.^17,18 The initial data on VNS suggested that 44% of patients experienced a 50% or greater reduction in number of seizures at 2 years of treatment, with a median decline in seizures of 46%.¹⁹ There are also recent data suggesting that VNS use is associated with lowered risk of SUDEP.²⁰

The most recent VNS model has a feature which makes it responsive to tachycardia. The rationale behind this change is the observation that both focal and generalized seizures are associated with tachycardia in up to 82% of patients.²¹ Autonomic activation is thought to play a role in this process. Heart rate changes can precede electrographic and clinical changes during seizure.²²

Fisher and colleagues implanted VNS with tachycardia detection in 20 patients with known ictal tachycardia.²³ This was defined as a heart rate above 100 beats per minute during a seizure and at least a 55% increase or 35 beats per minute increase from baseline. In addition to the basal cyclic stimulation, all VNS models have the option of patient-activated stimulation in the form of a magnet which is swiped over the stimulator. The Aspire SR 106 (VNS, LivaNova Inc., Houston, TX, USA) also senses and stimulates in response to tachycardia. The tachycardia detection can be turned on or off and stimulation is set at 0.125 mA below the magnet current. A change in heart rate ranges from 20% to 70% increase over baseline triggers stimulation (this is modifiable). There is a sensitivity setting ranging from 1 to 5 (most to least sensitive). Fifty percent of patients were considered responders (defined as greater than 50% reduction in seizures) at 1 year. Similar responder rates have been found with previous VNS models.^24,25

Common side effects from VNS include hoarseness, dysphagia and cough.²⁶ More serious side effects of lead damage (3%) and infection (3–7%) have also been reported.²⁷

Responsive versus continuous deep brain stimulation

Responsive neurostimulation is novel in its mode of stimulation delivery. Other implantable devices, aside from the VNS with tachycardia detection, provide stimulation intermittently or continuously. Deep brain stimulation is one such system. Since 2010, DBS has been used in Europe and other regions for treatment of refractory focal epilepsy. Deep brain stimulation is approved in the US for Parkinson’s disease and essential tremor but not for epilepsy. The best investigated target for DBS in focal epilepsy is the anterior nucleus of the thalamus, although the centromedian nucleus of the thalamus, cerebellum and hippocampus have also been studied in uncontrolled clinical trials with varying success.²⁸ DBS delivers duty cycle stimulation which is intermittent electrical simulation, similar to VNS, to the designated electrode. The amplitude, pulse width and frequency of the stimulation are programmable (Figure 4).

Figure 4.

X-Ray showing placement of the leads of a deep brain stimulator in the anterior nucleus of the thalamus.

The landmark trial for DBS was the SANTE study in 2010, which involved 110 patients randomized to stimulation and control groups. In the control group the stimulator was implanted but remained off and these patients could then participate in the unblinded arm. After a 3-month blinded phase there was a 9-month unblinded phase. During the blinded phase focal seizures improved by 36.3% as opposed to 12.1% in the control group, with the greatest improvement in patients with temporal onset. During the open-label phase the mean seizure frequency decreased by 41% at 13 months and 56% at 25 months.²⁹ Longer-term outcomes were reported by Salanova and colleagues in 2015. At 5 years there was a median seizure reduction of 69% and a responder rate (defined by greater than 50% reduction in seizures) of 68%.³⁰

The most common adverse event was paresthesia (18.2%). Eighteen patients in the initial trial withdrew after implantation because of adverse events. Other more serious events included hemorrhage and infection.²⁹

Neurocognitive outcome was also studied. During the blinded phase of the SANTE study 14.8% of patients in the active group reported depression, compared with 1.8% of controls. Memory issues were also reported in 13%, compared with 1.8% of controls.³¹ These effects diminished with time and at 7 years of treatment there were no significant cognitive or mood effects associated with DBS. Oh’s group found that there were improvements in verbal fluency after a year of DBS.³²

There have been no studies comparing RNS and DBS. Based on the two major long-term trials the responder rate at 5–6 years is similar in RNS and DBS (64.6% and 69% respectively). RNS seemed to have a favorable neurocognitive profile with respect to memory and depression incidence. Does the type and location of stimulation matter? DBS delivers higher overall current density and has prolonged stimulation duration. DBS is based on the principle of affecting seizure networks while RNS is based on modulating the seizure onset zone. While RNS was designed to abort seizures with direct stimulation the continued improvement in patients over months to years also suggests that there is a beneficial neuromodulatory effect, likely due to the many stimulations which occur daily. The underlying mechanisms of both of those stimulation paradigms are unclear. Further study is warranted to render judgement on which mode of stimulation is more advantageous. With a longer clinical experience using these modalities certain subgroups or trends could emerge. For example, DBS is more effective for temporal compared with frontal epilepsies, while the RNS seems to be equally effective in neocortical and mesial temporal lobe epilepsy.^30,10 Will the trend of improved response rate over years continue with prolonged therapy or will it level out? Is the neuromodulatory effect due to re-networking of epileptic pathways and if so can this be accomplished with responsive or intermittent stimulation?

Chronic subthreshold cortical stimulation (CSCS) is an experimental form of neurostimulation currently under clinical investigation. This provides constant, low-level stimulation to subdural electrodes on or near the seizure onset zone.³³ This is based in part on earlier observations that continuous stimulation to the mesial temporal structures has beneficial effect in some patients with drug-resistant epilepsy.³⁴ A continuous, low-level stimulation is delivered (0.2–0.4 mA; in comparison RNS delivers 0.5–12 mA stimulation). Data from 13 patients revealed a subjective improvement in life satisfaction and reduction in epilepsy severity in 76.9%.³⁵ There was also a reduction in interictal epileptiform discharges in some patients. Further rigorous study is needed on this novel modality of neurostimulation.

Comparing stimulation modalities

In the third of patients who are not adequately controlled on seizure medications, resective surgery is the single best option for complete cure. In patients for whom this is not possible a stimulation device may be considered.

Responsive neurostimulation may be the device of choice in patients with multiple foci or in patients where a focus has been identified but cannot be resected. For example, the focus resides in the eloquent cortex. In our experience RNS is a great option in these patients. If a focus cannot be identified or there are more than two potential foci, VNS is the best option. There may be other considerations, including patient preference, infection risk, and ease of use. For RNS to be successful patients or caregivers must be able to regularly record and download data. VNS does not require this type of input. Both systems benefit from adjustments in settings and the best results are not achieved immediately.

If DBS becomes available in the US it may be an option in a similar patient population to those who are RNS candidates.

Additional benefits

One advantage of RNS is the ability for a patient to identify and download seizure data using a wand and laptop (Figure 5). This information can be used to track long-term trends and patterns. A recent paper by Spencer and colleagues described emerging cyclic trends. Interictal activity was more robust during sleep. Neocortical regions had an increase in seizure detections nocturnally. In contrast, limbic regions had a diurnal peak in detections.³⁶ There was a suggestion of longer-term trends in frequency. In the future medication regimens or stimulation intensity could be individualized based on these types of trends. With time, RNS may also teach us more about the underlying mechanisms of epilepsy.

Figure 5.

Patient holds a wand to the stimulator and downloads electrocorticography (ECoG) data.

Another potential use for RNS is in the case of bitemporal epilepsy, which represents a significant number of patients with temporal epilepsy.³⁷ Implanting electrodes into both temporal lobes in cases of suspected bitemporal epilepsy could provide additional long-term intracranial ECoG. In a study of 82 patients implanted with bitemporal RNS electrodes, 71 had bitemporal seizures. The average time to record both seizures was 41.6 days, suggesting that the standard 1-week epilepsy monitoring unit stay may not be enough to diagnose these patients.³⁸ Whether the RNS is useful as a long-term ambulatory monitoring device to localize seizure onset needs to be proven.

Conclusion

In patients with drug-resistant epilepsy the best option is resective surgery because of the high likelihood of seizure freedom in the appropriate patient. In patients for whom this is not possible there is an increasing number of stimulator options. Some devices such as VNS and DBS provide continuous cyclic stimulation. Responsive stimulators provide on-demand stimulation in response to intracranial ECoG patterns. Longer-term data are still needed on these devices. The most recent studies on DBS and RNS give 5–8 years of clinical experience and suggest that there are continued long-term improvements with these devices. Both modes of stimulation have a neuromodulatory effect. These devices have proven safe and well- tolerated over the long term. More widespread use will also help physicians discover which patients benefit the most from these interventions. Longer clinical use with the tachycardia responsive VNS will be needed to see if this feature provides additional benefit compared with traditional VNS. Questions also remain regarding the optimal frequency and amplitude of stimulation.

In summary, responsive stimulation provides an option to patients whose seizures cannot be controlled medically. Further long-term clinical experience will enhance our understanding of this technology.