If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Brain stimulation has become a major field of innovation in epilepsy therapy during the last five years.
Both direct stimulation of the brain and stimulation of brain nerves have class I evidence for efficacy in focal epilepsy.
Diverse mechanisms of action are involved in the antiepileptic action, depending on stimulation parameters and site.
Ictal closed-loop stimulation of the epileptic focus provides the first on-demand treatment available for epileptic seizures.
Brain stimulation is increasingly used in epilepsy patients with insufficient therapeutic response to pharmacological treatment. Whereas vagus nerve stimulation with implanted devices has been used in large and heterogeneous patient groups, new devices also enable targeted brain stimulation at the site of seizure generation (responsive neurostimulation) or at network hubs (thalamic stimulation). Both responsive neurostimulation systems targeting the epileptic focus and the latest vagus nerve stimulators are intended to stimulate during the ictal phase to disrupt clinical seizure manifestation of reduce seizure severity. Furthermore, transcutaneous stimulation approaches are now available, although their efficacy remains uncertain. This review explains the concepts underlying brain stimulation, provides an overview of efficacy and tolerability data and discusses the rational use of the growing spectrum of neuromodulatory strategies available.
Unlike systemic pharmacological treatments affecting all brain areas expressing the individual drug ligands, neuromodulation can be applied to a defined target region and its associated network circuitry. This enables clinicians specifically to design stimulation approaches for different focus regions and help to reduce unwanted effects. However, this approach may be of limited efficacy if epileptogenic regions are extended or if there is rapid spread of epileptic activity. Beyond the stimulation sites, the effects of electrical stimulation critically rely on the stimulus parameters chosen from a wide und multidimensional parameter space. At present, the mechanisms contributing to the antiepileptic efficacy of neuromodulatory approaches are not completely understood. High frequency stimulation may cause local inactivation of target brain tissue by preferential activation of GABA-ergic inhibitory neurons and alter extracellular potassium concentrations. It may furthermore desynchronize neural activities and lower the recruitability of neurons to epileptic rhythms. Low frequency stimulation may reduce excitability by induction of long-term depression, and DC stimulation may diminish action potential generation by hyperpolarization of neuronal membrane potentials (e.g. [
]). Activation of brainstem nuclei with widely divergent projections may have extended net inhibitory effects.
Presently, the limited efficacy of the neurostimulation approaches available means that they can only be considered as palliative treatments, used to reduce patients’ seizure burden and improve quality of life. This review explores the efficacy and tolerability of currently available and certified neuromodulation techniques using peripheral nerve stimulation and direct brain stimulation (Table 1) and discusses possible criteria for the selection of individual approaches.
Table 1Approved invasive neuromodulatory approaches for epilepsy.
Vagus nerve stimulation
Responsive focus stimulation
1997 (FDA)/(EU) (since 2015: also heart-rate triggered)
Left vagus nerve (neck)
Anterior nuclei of the thalamus (bilaterally)
Epileptic focus (cortex)
Subcutaneous, left pectoral/subclavicular
Within the skull
Open-loop/closed-loop based on detection of tachycardia
Closed-loop based on detection of ictal EEG patterns
Intensity: 0,25–3 mA Frequency: 20–30 Hz Pulse width: 250–500 μs Duty cycle: 30 s on/5 min off (standard); 7 s on/30 s off (“rapid cycling”)
Intensity: 5 V Frequency: 145 Hz Pulse width: 95 μs Duty cycle: 1 min on/5 min off
12,7% infections 4.5% incranial bleeding 18.2% paresthesia at implantation site 10.9% local pain
7.8% infections 4.7% intracranial bleeding 9.9% transient pain at implantation site
Side effects of stimulation
Hoarseness (intensity-dependent up to 66%), cough (up to 45%)
14.8% depression 13.0% memory impairment
Efficacy in blinded studies (stimulation — control/sham)
Seizure frequency Δ — 12.7%/−18.4% responder Δ 7%/18%
Seizure frequency Δ — 25.9% responder rate Δ 3%
Seizure frequency Δ — 20.6% responder rate Δ 2%
Basis characteristics of invasive stimulation approaches. Adverse events and efficacy is stated based on data from randomized controlled trial phases; this may underestimate long-term efficacy and overestimate experience-dependent complication rates of the implantation procedure. Stimulus parameters can usually be chosen over wider areas and are given as most frequently applied. Δ: difference between stimulation group and active control (VNS) resp. sham stimulation (thalamic stimulation, responsive focus stimulation).
], and a new paradigm for implantable VNS aimed to automate ictal stimulation have been introduced. Furthermore, transcutaneous trigeminal nerve stimulation has undergone initial clinical trials and may offer a novel, non-invasive alternative using a similar mode of action as VNS treatment.
1.2 Vagus nerve stimulation
Stimulation of the vagus nerve activates brain stem nuclei including the N. tractus solitarii; secondary activation of the N. coeruleus and its noradrenergic projections are critical for its antiepileptic efficacy [
] Fig. 1); there are, however issues of blinding in all studies of peripheral nerve stimulation as patients are aware of the stimulation. Nevertheless, these earlier results were confirmed in a prospective trial by Amar et al. [
] found an odd’s ratio of 1.73 [1.13–2.64] in favor of a positive treatment response with 20–30 Hz VNS vs. active controls based on five prospective studies.
In regulatory trials, VNS implantation was associated with an up to 11% risk of infection (E05) and an about 1% risk of vocal cord paralysis. A recent retrospective study reported infections in 2.6%, postoperative hematoma in 1.9%, vocal cord paralysis in 1.4%, pain in 1.4%, cable break in 0.2%, and other local surgical complications in 0.6% of cases. Over time, lead fracture occurred in 3.0%, lead disconnection from the stimulator in 0.2%, and spontaneous VNS activation in 0.2% of cases in a large surgical series [
Depending on the stimulation parameters, hoarseness of the voice and dyspnea during stimulation periods are the most common stimulation-related side effects, coughing and local pain occur less frequently. VNS does not have cognitive side effects [
The variation of duty cycles from the standard setting of 30 s stimulation on-time and 5 min of pause to shorter intervals (“rapid cycling”) has been suggested for early non-responders by the company Cyberonics (producers of the first implantable VNS device); however, there is no evidence from larger patient cohorts that this improves efficacy [
1.2.1 Long-term neuromodulation versus acute effects
VNS has repeatedly been claimed to exert neuromodulatory effects with prolonged treatment rather than as an immediate consequence of stimulation. The facts that efficacy may increase over a period of months and that long-term improvement may be seen in patient cohorts over several years have been put forward as evidence for such claims [
]. Uncontrolled data suggesting longer term improvement need careful interpretation. However, such effects would be of considerable interest because they suggest that VNS may not only work via immediate stimulation effects but also through the remodulation of networks toward a less epilepsy-prone state. Fig. 2 shows reported seizure reduction in three published patient cohorts which suggest the VNS has additional longer term effects.
Of course, a delayed onset of an anti-seizure effect may be at least in part be due to the gradual up-titration of stimulation intensities which may occur over periods of several weeks to months. Improvement over periods of years, on the other hand, may be due to factors unrelated to VNS therapy such as changes in antiepileptic medication, resective surgical epilepsy treatment or enrichment of VNS patient cohorts with responsive subgroups (e.g. [
Complete seizure control is rarely achieved by VNS; in a recent survey based on the Cyberonics patient registry of 5554 patients, 49.0% of patients are classed as “responders” with a reduction of seizure frequency by at least 50% and 5.1% as seizure-free patients after 4 months of treatment; 63.0% are considered responders and 8.2% are seizure-after 2–4 years of treatment. A literature review suggested responder rates of 40.0% by 4 months with 2.6% of patients becoming seizure-free, and responder rates of 60.1% (8.0% seizure-free) at last follow-up [
]. In the registry data, earlier seizure onset was associated with better seizure control. These data were, however, not corrected for the lower number of patients and visits with longer follow-up periods and for changes in the accompanying drug regimen.
Patients who have received a VNS implant and who are capable of doing this are usually given a magnet in order to activate the stimulator when they or their relatives notice that a seizure is starting. Some patients believe that this can alleviate or disrupt seizures [
]. Only a minority of patients can, however, activate the stimulator themselves during a seizure. Since 2014 a system has been available intended to trigger ictal stimulation automatically. Whereas the standard VNS treatment regime uses intermittent stimulation for 30 s every 5 min (pauses are required to preserve nerve integrity), the new implantable VNS device is capable of sensing heart rate changes through constant electrocardiographic (ECG) monitoring and of triggering VNS activation if the heart rate crosses a pre-defined threshold. This threshold is chosen relative to a moving average of 10 s to allow the device to adapt to physiological heart rate variability. The functionality of heart rate detection has been proven in a prospective trial [
], however, so far there are no studies proving that ictal stimulation improves the efficacy of VNS treatment.
1.2.3 Transcutaneous vagus nerve stimulation
Transcutaneous vagus nerve stimulation (tVNS) is an approach intended to have similar effects as implanted VNS generators. This approach uses the auricular skin branch of the vagus nerve for non-invasive stimulation with a smartphone-like programmable generator. Only 4 h of transcutaneous stimulation per day have been applied so far with the aim of achieving a longer-lasting neuromodulatory effect. tVNS was recently studied in a prospective but underpowered multicenter trial comparing 25 Hz and 1 Hz stimulation (active control; [
]). Whereas there was a significant decrease of seizure frequency in the 25 Hz stimulation group during month 4 compared to baseline, there was no statistically significant difference between the two treatment arms (Fig. 1). Most patients tolerated tVNS well; local skin irritation and headache may occur if high stimulus amplitudes are chosen. Both efficacy of tVNS and the potential predictive value of tVNS for an implantable VNS response need further investigation. A higher efficacy using bilateral tVNS was reported by a Chinese group [
]. Acute trigeminal nerve stimulation (TNS) has been effective in acute seizure models. This observation was the basis of two clinical studies of its efficacy in human epilepsy. In a pilot study using overnight bilateral stimulation of V1 or V2 [
], 7 out of 13 patients continued treatment for 12 months. Seizure frequency was reduced by 66% after 3 months, and by 59 after 12 months. A subsequent randomized, controlled study in 50 patients comparing 120 Hz stimulation with active control failed to achieve statistically significant differences in terms of seizure frequency reduction, responder rate, or time to the fourth seizure. Within the high frequency stimulation group, a responder rate of 40.5% was achieved [
]. Within one year after the primary treatment phase, responder rates had dropped to 30.6% of all patients starting the study, with a difference of 11.8% between high frequency stimulation and active control [
]. Side effects were mostly limited to local skin irritation and headaches. Larger studies are needed to prove the clinical efficacy of TNS.
1.3 Intracranial stimulation
Intracranial stimulation can pursue different strategies and be aimed at a wide range of anatomical targets. Both, the epileptogenic regions and relevant network hubs are possible targets for neuromodulation. A variety of targets have been studied in pilot applications with equivocal results, including cerebellum, subthalamic nucleus, anterior and centromedian thalamic nuclei, caudate nucleus and hippocampus. Based on larger multicenter trials, in Europe, stimulation of the anterior nucleus of the thalamus (ANT-stimulation) is available for the treatment of pharmacoresistant focal epilepsy, whereas in the US a closed loop stimulation system applied to the region of the epileptic focus has received FDA approval.
1.3.1 Stimulation of the anterior nuclei of the thalamus
The initial evidence for the efficacy of thalamic stimulation emerged from a US multicenter trial (Stimulation of the Anterior Nucleus of the Thalamus for Epilepsy, “SANTE”) involving 110 patients with pharmacoresistant focal epilepsy and the application of high frequency stimulation (>100 Hz) bilaterally using a transventricular approach to the anterior nuclei of the thalamus (ANT). Following an insertional effect generating about 20% reduction in seizure frequency, seizure control gradually improved in the group receiving high frequency (145 Hz) stimulation (40.4% reduction in seizure frequency vs. a decrease by 14.5% in the unstimulated control group) [
]. As the difference between stimulated and control group was not significant over the whole 3-month period analyzed, the FDA considered the primary outcome a “not met”, and thalamic stimulation was only made available in other countries, e.g. Europe, Australia and South America. Side effects of stimulation included mood deterioration (14.8% of patients vs. 1.8% in the control group) and memory impairment (13.0% vs. 1.8% of patients). The number of more severe seizure types declined particularly strongly, possibly pointing toward a major effect of ANT stimulation on the propagation of ictal epileptic discharges within the brain.
] showed a gradual increase in efficacy in the patients remaining in follow-up, with 59 out of 110 patients initially treated experiencing a median seizure reduction of 69% (Fig. 3) and 16% achieving seizure-freedom for at least 6 months. There were also statistically significant improvements in seizure severity and quality of life. However, this analysis did not control for changes in the antiepileptic drug regimen. Presently outcome data of thalamic stimulation performed in Europe is being collected for further analyses in an industry-sponsored registry (“MORE”), comprising >150 patients as of July 2016.
1.3.2 Responsive focus stimulation
A different intracranial stimulation approach involves responsive stimulation of the epileptogenic focus (RNS). Here, an advanced implantable device capable of using three algorithms is intended to detect ictal electrographic patterns and to trigger timely interference with ongoing ictal activity in a closed-loop fashion. Unlike with VNS and deep brain stimulation of the anterior nucleus of the thalamus, the generator is implanted into the skull. Recordings and stimulation are performed via two four-contact strip or depth electrodes placed in the suspected or intracranially established seizure onset zone. Stimulation is performed using brief bursts (100 ms) of 200 Hz stimulation when ictal EEG patterns are detected online. The device is capable of storing a subset of peristimulation EEG data for later offline analysis.
In a randomized, double-blind prospective trial in 191 patients, superiority of stimulation versus a sham control was proven [
]. After an insertional effect resulting in a >25% seizure reduction, the overall reduction in seizure frequency was 37.9% in the stimulation group vs. 17.3% in the sham control. During the final month of blinded stimulation, the reduction in seizure frequency compared to baseline was 41.5% in the stimulation group and 9.4% in sham controls. Perioperative morbidity was rare with the exception of headaches and implant site pain reported during the first four weeks after implantation by a total of 21.4% of patients. Other problems included implant site infection (5%), and muscle twitching (2%), dizziness and paresthesia, whereas there was no evidence of mood changes or cognitive deterioration. In a study long-term efficacy study of open-label treatment the median reduction in seizure frequency was 44% at 1 year, 53% at 2 years and ranged between 48 and 62 percent in the subsequent 3 years, with a similar course as reported with ANT-stimulation (Fig. 3; [
]). Quality of life improved during these longer time periods, and neuropsychological tests showed cognitive improvements in learning with mesiotemporal stimulation, and in naming with neocortical stimulation after 2 years of treatment [
There is a rapidly growing spectrum of neuromodulatory approaches to the treatment of pharmacoresistant focal epilepsy, using different stimulation sites, stimulus parameters and intervention timings. So far, there is class I evidence for the therapeutic efficacy of vagus nerve stimulation and responsive focus stimulation, whereas stimulation of the anterior nuclei of the thalamus marginally failed to achieve the predefined primary positive outcome threshold. Transcutaneous approaches are intriguing possible early treatment options, but so far have not undergone sufficiently powered clinical trials to provide evidence of their efficacy. Recently, new approaches like fornix stimulation [
] and transcranial modulation of cortical foci using a subgalear electrode with Laplacean design have been proposed for treatment of pharmacoresistant focal epilepsy (http://www.precisis.de/en/epilepsy/easee-system-development-project.html). Sufficiently powered prospective clinical trials will need to assess their clinical potential. Aside from such additional stimulation targets, the wide range of possible stimulus parameters pose a difficult challenge in the field of neuromodulation.
], current neuromodulation techniques cannot compare with resective epilepsy surgery in terms of seizure control and thus have to be considered as complementary rather than alternative treatment options to excisional surgery in patients with pharmacoresistant focal epilepsy. The lack of systemic side effects, pharmacokinetic interactions, compliance problems, teratogenicity, and the potential for immediate intervention during an ongoing seizure represent significant advantages of stimulation techniques over the addition of another antiepileptic drug. In trials of VNS therapy, retention rates were considerably higher than those seen in treatments with antiepileptic drugs. One prospective study has indicated that the combination of VNS with pharmacotherapy has greater positive effects on patients’ quality of life than optimized pharmacotherapy [
]. Neurostimulation may also offer advantages in terms of cognitive tolerability as shown for VNS and RNS.
Closed loop approaches make use of a specific advantage of electrical brain stimulation, its immediate effects on brain activity without interference of resorption, metabolization, distribution and passage of a pharmacological agent across the blood–brain-barrier, slowing down presently available antiepileptic drug application routes. Obviously, the recording of a seizure pattern from the seizure onset zone (as used in responsive focus stimulation) should provide the best basis for early intervention during the course of a seizure. This approach is, however, only possible in those patients in whom the seizure onset zone is known and who are yet not suitable for epilepsy surgery. Furthermore, the seizure detection algorithms implemented so far provide fast interventions at the expense of low specificity, with responsive focus stimulation being activated several hundred times per day. Improvements in seizure detection algorithms [
] may open up larger time windows for timely interventions and requires new studies on stimulation paradigms which are capable of reversing pro-ictal alterations in brain dynamics.
The use of ictal tachycardia as a surrogate trigger for ictal stimulation is again limited to the subgroup of patients with early cardiac effects; this group is, however, quite large, especially amongst patients with temporal lobe epilepsy [
]. As in the case of EEG-based approaches, the specificity of detection of seizure-related changes needs to be improved, and data on the efficacy of ictal vagus nerve stimulation have not been provided yet.
So far, there are no prospective data comparing the efficacy of available neuromodulatory approaches, and there is a lack of valid predictors of individual treatment response. Rational patient selection is thus difficult and mostly limited to arguments based on plausibility. Possible reasons to select VNS could include an unclear site of seizure generation, comorbid depression, or a susceptibility to cognitive side effects of antiepileptic drug treatment; patients with ictal tachycardia may become candidates for closed-loop VNS if the additional efficacy of ictal stimulation can be established. ANT stimulation may be particularly effective in patients whose seizure generation zone is in the limbic and mesiofrontal target areas of ANT neurons, and in patients in whom the ANT is a hub relevant for seizure propagation. Responsive focus stimulation may be of particular interest if the seizure onset zone is known but resective strategies are not an option, for instance because seizures start in eloquent cortex such as language areas or functional hippocampi. Both thalamic and responsive focus stimulation have proven effective in patients who had previously failed to respond to VNS.
Many questions remain open. Overall, stimulation efficacy needs to be improved. Improvements should be possible through the optimization of patient selection, electrode positioning, choice of stimulation parameters and through new approaches to focus and network modulation. However, progress is hampered by problems with patient enrollment in prospective clinical studies, and the limited number of stimulation paradigms which can be studied systematically after selection from the wide parameter space available. Case registers of patients receiving treatment with stimulation therapies could improve our knowledge about optimal case selection, treatment or stimulation parameter choice. However, it would be critical for such registers to be independent of companies with a financial interest in devices or in stimulation treatments. Specific subanalyses e.g. for optimal electrode positioning in thalamic stimulation [
] can produce hypotheses which can then be studied in larger cohorts. Notably, possible pharmacodynamic interactions between brain stimulation and antiepileptic pharmacotherapy have not been addressed in the analyses of clinical studies although there is some experimental evidence for interactions (e.g. [
]), which may be relevant in terms of both, clinical efficacy and a better understanding of the mechanisms involved in neuromodulation.
Transcutaneous and transcranial stimulation approaches may lower the threshold for an early use of neuromodulatory therapies, provided that they prove sufficiently efficacious. Despite all progress neuromodulation, the technologies used in present-day implants are — with the exception of the RNS device — a long way behind what would be considered state of the art in the 21st century; this is true for the way EEG signals are picked up, the miniaturization and processing power of the microprocessors used, storage capacities of devices, options for external readout of data and flexibility in terms of temporospatial stimulation patterns. It has to be hoped that the rapid progress in engineering will translate into more versatile and powerful devices, and that increasing safety demands of regulatory bodies will not prove beyond the financial capabilities of innovative startup companies intending to bring modern technology to our patients [
A. Schulze-Bonhage has received honoraria for lectures from pharmaceutical companies (Bial, Desitin, EISAI, Novartis, UCB), honoraria for advice from Bial, EISAI and Precisis, research support for participation in clinical trials and registries from Medtronic, Bial, Precisis and UCB, and research support from funding institutions ( BMBF , DFG , EU and NIH ).
This review was supported by the Excellence Cluster BrainLinks-BrainTools (DFG, Grant # 3 EXC 1086 ).
Long term multicenter experience with vagus nerve stimulation for intractable partial seizures.