About 30% of epilepsy cases are medically intractable and require non-pharmacologic treatment.  While resective surgery is the treatment of choice in these cases, it is not always feasible. In addition, about 20% of patients treated at the most experienced medical centers are not seizure-free after surgery.  Neurostimulation (or neuromodulation) is now an acceptable palliative treatment option for those patients.
From 1997 to 2013, the only neurostimulation modality approved in the United States was vagus nerve stimulation (VNS). The brain-responsive neurostimulator (RNS system) was approved in 2013, and the approval of deep brain stimulation (DBS) is anticipated. There are also other modalities available in other countries, though not on near approval in the United States at this point.
This review will provide an update on VNS since its initial release, as well as discuss the roles of more recent neurostimulation modalities and compare them with respect to each other and to epilepsy surgery.
Vagus Nerve Stimulation (VNS, Cyberonics/Livanova, Houston, TX)
In July 1997, the US Food and Drug Administration (FDA) approved the use of VNS as an adjunctive treatment for refractory partial epilepsy in adults and adolescents aged 12 years and older. The approval was mainly based on the results of the E05 study,  which compared the efficacy and safety of VNS in patients who received high stimulation vs. low stimulation.
The high-stimulation group in the E05 study had a 28% decrease in mean of seizure frequency, vs.15% decrease in the low-stimulation group. Following the acute phase of the study, longer-term results  showed a responder rate (50% or more in seizure reduction) of 36.8% of patients after one year, 43.2% at two years, and in 42.7% at three years. Median seizure reductions compared with baseline were 35% at one year, 44.3% at two years, and 44.1% at three years. It was concluded that the long-term, open-label VNS provided seizure reduction similar to or greater than acute studies.
An 11-year retrospective review of VNS in a consecutive series of 436 adults and children  found that when used in conjunction with a multidisciplinary and multimodality treatment regimen, more than 60% of patients experienced at least a 50% reduction in seizure burden.
Tolerability and safety
In the E05 study, adverse effects reported by more than 10% of the patients during the peri-operative period were pain (29%), cough (14%), voice change (13%), chest pain (12%), and nausea (10%). In the high-stimulation group, voice alteration/hoarseness, cough, throat pain, nonspecific pain, dyspnea, paresthesia, dyspepsia, vomiting, and infection were increased significantly from baseline. Clinical experience has since showed that the most common effects (hoarseness, cough, shortness of breath, paresthesias) appear during stimulation and tend to diminish over time. [3, 5] It was concluded that VNS is safe and well tolerated, with nearly three quarters of the patients choosing to continue therapy. VNS is not associated with the usual adverse effects of antiepileptic drugs (AEDs), such as fatigue, dizziness, depression, insomnia, confusion, cognitive impairment, weight gain, and sexual dysfunction. VNS may have deleterious effects on sleep apnea,  though it is not clear how clinical relevant this is.
Stimulation parameters are many. Output current varies from 0.5 to 3.5 mA, and is typically increased gradually according to tolerance. The default parameters are usually 30-Hz signal frequency, 500-microsecond pulse width, 30 seconds of "on" time and 5 minutes of "off" time (10% duty cycle). The optimal range of device duty-cycles and other parameters remain unclear and largely subject to individual preferences. [7, 8]
The handheld magnet can used on demand to interrupt or reduce the severity of an oncoming seizure. The patient or a companion may activate the generator (triggering an additional stimulation) by swiping the magnet on the generator. Magnet use offers patient and family alike a sense of control or empowerment, and may indeed decrease seizure duration or severity.  The magnet can also be used to turn off automatic stimulation on demand to restore the voice to its normal level if need be.
Quality of life (QOL)
VNS appears to improve quality of life independent of its effect on seizure control. Subjective improvement in Quality of Life (QoL) occurred in 84% of patients, which is likely attributable to additional factors besides seizure control.  VNS is in fact FDA-approved for treatment-resistant depression,  though rarely used due to poor reimbursement.
The PuLsE (Open Prospective Randomized Long-term Effectiveness) trial studied patients with pharmacoresistant focal seizures and divided them into two groups: one received VNS as adjunct to best medical practice (VNS + BMP) and the other received BMP alone. Significant difference was observed in favor of VNS + BMP regarding improvement of seizure frequency and QoL. However, it was noted that more patients in the VNS + BMP group (43%) reported adverse events compared to BMP group (21%) (p = 0.01). adverse events were mostly related to VNS implantation or stimulation. It was concluded that VNS + BMP was superior to BMP alone in improving QoL. 
Effect on Sudden Unexpected Death in Epilepsy (SUDEP)
The risk of premature death is increased in patients with intractable epilepsy. Sudden Unexpected Death in Epilepsy (SUDEP) is the most common cause of death in patients with intractable epilepsy. The effect of VNS on mortality remains unclear. Earlier studies suggested that VNS may reduce the risk of SUDEP after 2 years of treatment,  but other studies did not confirm this.  In a study  of T-wave alternans (a cardiac marker for sudden cardiac death associated with SUDEP) and heart rate variability (an indicator of autonomic function), VNS seemed to have a cardio-protective role.
The (USA) FDA indication for VNS use is "adjunctive therapy in reducing the frequency of seizures in adults and adolescents over 12 years of age with partial onset seizure, which are refractory to anti-epileptic medications." Although the FDA indication for VNS is rather narrow, most epileptologists agree that VNS indications are probably broader. For example, VNS is often the best option for intractable generalized epilepsies,  whether of the Lennox-Gastaut type or idiopathic (genetic) generalized epilepsies. In fact generalized epilepsies of the Lennox-Gastaut type, even though technically “off label”, may be the most common use of VNS. Appropriately, VNS is approved for all types of refractory epilepsies in Europe.
Evolution over time
The models and specifications of the generator have evolved and improved since its first release almost two decades ago. Generally speaking, the device has become smaller, thinner, and its battery life has increased. The latest improvement replaced the blind closed-loop stimulation concept with an open loop therapy. The Aspire device,  in addition to the automatic stimulation, detects any increase in heart rate associated with seizures and delivers an additional stimulation triggered by tachycardia detection. Detection parameters here include heart rate detection and tachycardia threshold. Whether this provides additional efficacy has not been yet determined.
One limitation of VNS is that it limits the use of later MRIs. MRI of the head and body, including 3T, can be performed (with a send and receive coil and the generator off), but the neck area is not considered safe due to possible heating of the lead.
Since its release, VNS has been reviewed periodically by the American Academy of Neurology (AAN) and American Epilepsy Society. Guidelines and practice parameters [18, 19, 20] consistently consider VNS standard of care as a non-pharmacologic treatment. The most recent AAN Quality Measures  rightly include VNS as indicated in patients who are refractory to AEDs and not suitable for epilepsy surgery. The AAN ranks VNS as effective and safe based on a preponderance of class I evidence (level Ia).
Brain Responsive RNS System (NeuroPace, Mountain View, CA)
The brain-responsive RNS system was approved by the FDA in 2013 for medically refractory focal epilepsy. The pivotal multicenter double blind randomized controlled study  observed 191 subjects with medically refractory focal epilepsy. Subjects were randomized into two groups; one received stimulation in response to detections (treatment), and the other received no stimulation (sham).
Groups were assessed over a 12-week blinded period, during which it was found that treatment group had significant reduction in seizures (37.9%) compared to 17.3% in the sham group (p = 0.012) without differences in adverse events. They were assessed again over an 84-week open-label period, during which all subjects received stimulation. This resulted in a sustained reduction in the treatment group and a significant reduction in the sham group. The study concluded that patients who received RNS exhibited a significant reduction in seizure frequency, with an associated improvement in overall QoL (p<0.02), and with no mood and cognitive changes. This study led to FDA’s approval of RNS for adults who average three or more seizures per month, have failed two or more AEDs, and have one or two seizure foci.
A subsequent study on the longer-term effects of RNS  found a median percent reduction in seizures in the open labeled period of 44% at 1 year and 53% at 2 years, representing a significant improvement over time. No significant difference in adverse event rate was noted between the treatment and sham stimulation groups. It was concluded that, similar to VNS, the efficacy of RNS tends to improve over time.
Tolerability and safety
The Pivotal Trial Open Label Period also showed improvements in naming and verbal memory, and in visual memory and executive function. No cognitive adverse effects were seen. 
Quality of life
The Pivotal Trial Open Label Period showed significant long term improvements in quality of life in all domains (Overall QOL, Epilepsy-targeted, Mental Health, and Physical Health) at one and two years, in both neocortical and mesiotemporal scenarios. A modest improvement in mood was also noted. 
Like VNS, RNS is palliative, and therefore is appropriate for patients who are not candidate for, or have failed, resective epilepsy surgery. The most common indications for RNS are true bitemporal epilepsy, and focal epilepsy, where the focus is in an eloquent cortex that cannot be resected.
MRI of the brain is contra-indicated with the RNS system.
Deep Brain Stimulation (DBS)
Deep brain stimulation has been FDA-approved as a treatment for Parkinson’s disease since 1997. DBS was studied in adults with medically refractory localization-related epilepsy by stimulating bilateral anterior nuclei of the thalamus in a multi-center, double-blind, randomized trial.  Adults were divided into two halves, one received stimulation and the other did not over a three-month blinded phase, then all received open label stimulation. Baseline median seizure frequency was 19.5 per month. In the last month of the blinded phase the stimulated group had a 29% greater reduction in seizures compared with the control group. By the end of the second year, there was a 56% median reduction in seizure frequency; 54% of patients had a seizure reduction of at least 50%. It was concluded that bilateral stimulation of the anterior nuclei of the thalamus reduces seizures for adults with medically refractory epilepsy, however, it has not yet been FDA approved in the United States.
Other Neurostimulation Modalities
Transcutaneous Vagus Nerve Stimulation (t-VNS)
Non-invasive t-VNS is a device that stimulates the afferent auricular branch of the vagus nerve located medial to the tragus of the left ear. It may be a useful, safe and well tolerated alternative treatment option. In a pilot-study by Stefan et al,  t-VNS was applied to 10 patients with pharmaco-resistant epilepsy where they received stimulation 3 times/day for 9 months. Three patients aborted the study. Five out of the remaining seven patients had overall reduction of seizure frequency after 9 months. T-VNS is not approved by the FDA due to minimal research.
External Trigeminal Nerve Stimulation
External trigeminal nerve stimulator (eTNS) is another non-invasive neuromodulation therapy that is not approved in the US. A double-blind randomized controlled study of 50 subjects  concluded that eTNS is associated with significant within-group improvement in the responder rate, which increased to 40.5% at the conclusion of the double blind period (p =0.01). It is also associated with a significant improvement in mood. Given promising results of phase 2 trial providing class II evidence that eTNS may be safe and effective, it was approved in Europe and Canada. It is still under investigation with larger multicenter phase 3 clinical trials.
Research in cerebellar stimulation for cortically or hippocampal induced epilepsy has been conducted since 1941 on animal models and in small clinical trials, resulting in mixed results. In 2005, bilateral cerebellar stimulation for intractable motor seizures was re-evaluated  in a small randomized double-blind study of 5 patients. Four-contact plate electrodes were placed on the cerebellar superomedial surface through two suboccipital burr holes. In the first month post-placement, none of the patient received stimulation, then patients were randomized into two groups: three received stimulation and two didn’t. After 4 months, all five patients received stimulation. No change was found in the mean tonic clonic seizure rate in the two patients who did not receive stimulation in the initial 3-month double blind phase whereas the three patients who received stimulation had a 33% seizure reduction (p = 0.023). A 41% seizure reduction was achieved in all five patients after 6 months of stimulation. At 24 months, tonic seizure reduction in all four patients who had them was 43%. The statistical analysis showed a significant reduction in tonic-clonic seizures (p < 0.001), and tonic seizures (p < 0.05). Complications included migration in three patients and wound infection in one patient.
Clinical Use of Neurostimulation
Being a surgical candidate is not a binary (Y/N) proposition. It can be viewed as a continuum. At one extreme, Idiopathic (genetic) generalized epilepsies and symptomatic generalized epilepsies of the Lennox-Gastaut type are never treated with focal resections. At the other extreme, because the likelihood of seizure freedom is high, clear-cut unilateral mesiotemporal epilepsy or straightforward focal lesional epilepsy should be treated with focal resections. Several scenarios fall somewhere in between, such as bitemporal epilepsy (but it depends how bitemporal is defined), extratemporal non-lesional epilepsy, and lastly extratemporal non-lesion poorly localized epilepsy.
In comparing VNS vs. RNS, the following factors are relevant: RNS is more invasive and requires seizure localization. VNS is less invasive, and thus may be used (and often is) in truly multifocal or generalized epilepsies of the Lennox-Gastaut type, or in focal epilepsy that is completely unlocalizable. At this point, the efficacy of VNS and RNS appears roughly comparable to many, [16, 32] so that it would seem logical to try VNS first. For others, the 3-year seizure reduction of 44% for VNS  and 60% for RNS  would seem to favor RNS. One advantage of RNS (over VNS) is the capability to be used as a recording device, and especially for bitemporal epilepsy the ability to record for weeks to months or years, while providing treatment, may allow lateralization and an eventual resection. [33, 34, 35] Both VNS and RNS can be used after failed epilepsy surgery. [36, 37]
Based on the above, every patient’s situation is different, and this is the reason for the multi-disciplinary management conference held at level-4 (surgical) epilepsy centers. There is certainly room for centers' preferences and differences in opinion, but within reason, and comprehensive centers should be … comprehensive and use all options available.  Lastly, patient preferences is of course also a factor.