Beyond Resection: Neuromodulation and Minimally Invasive Epilepsy Surgery

Abstract

Epilepsy is a common neurological disease impacting both patients and healthcare systems. Approximately one third of patients have drug-resistant epilepsy (DRE) and are candidates for surgical options. However, only a small percentage undergo surgical treatment due to factors such as patient misconception/fear of surgery, healthcare disparities in epilepsy care, complex presurgical evaluation, primary care knowledge gap, and lack of systemic structures to allow effective coordination between referring physician and surgical epilepsy centers.

Resective surgical treatments are superior to medication management for DRE patients in terms of seizure outcomes but may be less palatable to patients. There have been major advancements in minimally invasive surgeries (MIS) and neuromodulation techniques that may allay these concerns. Both epilepsy MIS and neuromodulation have shown promising seizure outcomes while minimizing complications. Minimally invasive methods include Laser Interstitial Thermal Therapy (LITT), RadioFrequency Ablation (RFA), Stereotactic RadioSurgery (SRS). Neuromodulation methods, which are more palliative, include Vagus Nerve Stimulation (VNS), Deep Brain Stimulation (DBS), and Responsive Neurostimulation System (RNS). This review will discuss the role of these techniques in varied epilepsy subtypes, their effectiveness in improving seizure control, and adverse outcomes.

INTRODUCTION

Epilepsy is common neurological disease with a major impact on both patients and health care systems. It affects approximately three percent of adults, with Mesial Temporal Lobe Epilepsy (MTLE) being the most common epilepsy localization worldwide (1). Approximately one third of patients are drug-resistant, requiring the need for surgical options for improved chance of being free from seizures (2). This is supported by high-level evidence from three randomized controlled trials comparing resective therapy to continued medical management in adult and pediatric patients (2–4). Seizure freedom rates of 58%–77% were found in the surgical resection group versus zero to eight percent in the medical management group. A recent large multicenter study from Europe showed 50%–78% of the patients were free of disabling seizures and this rate is influenced by underlying pathology (5). Still, MRI normal epilepsy can have good surgical outcomes, with 38%–60% of the patients achieving good long-term outcomes (Engel class I or Engel class IIA) (6). Expectedly, lack of MRI-detectable lesions results in fewer successful surgery outcomes.

The overarching effectiveness of epilepsy surgery is at odds by the challenge of only a small percentage of surgical candidates receiving epilepsy surgery. This leads to a large surgical treatment gap in both high and low-income countries (7). Physician, patient, and system related factors such as knowledge gap among physicians, patient misperceptions and fear, disparities in epilepsy care, complexity of the pre-surgical evaluation, lack of adequate coordination between epilepsy centers and the referring physician, and inadequate epilepsy surgery research are possible reasons for this gap (8). Even when patients are evaluated, some are not good resective surgery candidates due to reasons such as a lack of discrete seizure focus, multifocal seizure foci, seizure focus involving (or close to) eloquent cortex and complicating medical comorbidities.

Highlights

• Minimally invasive epilepsy surgeries (MIS) include Laser Interstitial Thermal Therapy (LITT), Radio Frequency Ablation (RFA), Stereotactic Radiosurgery.

• Neuromodulatory techniques include Vagus Nerve Stimulation (VNS), Deep Brain Stimulation (DBS), and Responsive Neurostimulation System (RNS).

• MIS and neuromodulatory methods may improve seizure outcomes while minimizing complications.

• These methods can be alternative treatment option for who may not be good candidates to resection.

Fortunately, our armamentarium has expanded with recent advancements in minimally invasive surgeries (MIS) and neuromodulation methods for epilepsy with promising seizure outcomes. Availability of these different treatment modalities increase both physicians’ and patients’ options for decreasing the treatment gap, while hoping for success in the minimization of neurological morbidity and neurocognitive adverse events.

Minimally invasive methods include laser interstitial thermal therapy (LITT), radiofrequency ablation (RFA), stereotactic radiosurgery (SRS) and the newest amongst them, focused ultrasound (FUS). Neuromodulatory methods include vagal nerve stimulation (VNS), deep brain stimulation (DBS), responsive neurostimulation (RNS). This review will discuss the role of these techniques in varied epilepsy subtypes, their effectiveness in improving seizure control, and adverse outcomes.

MINIMALLY INVASIVE METHODS

Laser Interstitial Thermal Therapy

Laser interstitial thermal therapy (LITT) involves low voltage laser energy that is converted to thermal energy resulting in coagulative necrosis and protein denaturation to the region of interest (9). To overcome the inability to monitor tissue destruction in real time, MRI thermography has been used to guide LITT (10) (MRg-LITT) for DRE (11). Since then, MRg-LITT has been successful in various epilepsy etiologies (10,12) and has been an important technique for deeply situated epileptic foci, even when close proximity to eloquent cortex (Figure 1) (13).

Figure 1.

a–c. Left mesial temporal laser interstitial thermal therapy postoperative MRI coronal images. Immediate postoperative T2 coronal image (a). Four months postoperative T2 coronal image (b). Four months postoperative postcontrast T1MPRAGE image (c).

Mesial Temporal Lobe Epilepsy

An initial prospective case series evaluated 13 MTLE patients with or without Mesial Temporal Sclerosis (MTS) who underwent MRgLITT to the amygdalohippocampal region from an occipital trajectory (14). Fifty-four percent were freed from disabling seizures (Engel I) and 30.8% were completely seizure-free (Engel IA). The success rate increased when evaluating only MTS with 67% achieving Engel I in the last follow-up. Since this earliest report, there have been multiple case series of 20–58 patients with seizure freedom rates ranging between 44%–67% for MTLE (average 58%) and 44%–74% (average 61.9%) for MTS (15).

A large retrospective multicenter study evaluated the effects of surgical targeting on seizure outcomes in 234 patients with MTLE who underwent MRgLITT (16). At two years, 57.5% had Engel I, and 80.2% had either Engel I or II outcome. Additionally, trajectories that were more anterior, medial, and inferior were more associated with Engel 1 outcomes. Although the presence of MTS on MRI was shown to be associated with favorable outcomes compared to MRI normal TLE, this study (in line with some single center LITT studies) did not find MTS predicted Engel I outcome at six months, even after excluding patients with dual pathologies from the sub-group analysis (16).

In a meta-analysis examining MTLE efficacy of MRg-LITT, (17) the seizure freedom rate in the last follow-up was 58% (95% CI 54%–62%) for MTLE and improved to 66% (95% CI 58%–74%) for those with MTS. The mean follow-up ranged from six to 42.9 months. Regarding complications, visual field deficits were the most common and were typically temporary. This meta-analysis included a study looking at Visual Field Deficits (VFD) in patients and reported a 37.5% rate of new VFD (higher on left) after procedure. In comparison, the average occurrence of VFD after anterior temporal lobectomy has been estimated to be 65% (17).

Another meta-analysis showed treatment efficacy for LITT in MTLE was 0.547 (0.506–0.588; I2 18.7%) (18). The prevalence of Engel IA decreased with time and was estimated to be 0.642 (0.568–0.724), 0.469 (0.388–0.567), and 0.424 (0.339–0.529) at the 12, 24, and 36 months, respectively. About half the population was seizure-free (Engel class IA) for an average of 22 months (18).

Kanner et al. studied long term outcome of 48 patients, in which 60% of patients had Engel I outcome after a mean follow-up of 50 months, the rates decreased from 78% at 18–24 months to 50% at 61–81 months (19). However, whether the ablation volume was associated with seizure freedom was uncertain. Wu et al. demonstrated that more extensive amygdala ablation was associated with Engel I outcome at follow-up while focusing only on the hippocampal body and tail was associated with a decreased chance of Engel I outcomes (16). Additionally, ablations extending posteriorly beyond the coronal plane in line with the lateral mesencephalic sulcus were less likely to be associated with Engel I outcomes. Another study looked at the role of ablation volumes and trajectories in 23 patients who underwent MRgLITT for MTLE (20) and demonstrated that rather than ablation volume itself, sparing of mesial hippocampal head was the only variable significantly associated with persistent seizures after treatment. A lateral position of the laser along the medial-lateral axis of the hippocampal head was associated with worse seizure outcomes. Like some other studies, Brotis et al. were unable to establish the role of total ablation volumes (18).

In terms of complications, Wu et al. identified postoperative hemorrhage in 1.2% (3/2430 patients, one being associated with transient double vision (16). A total of 42 complications were recorded for 35 patients (15.0%), of which eight were transient and 34 were persistent in the follow-up. Similarly, in Kerezoudis et al., visual disturbances were most common (5.1%), followed by worsening of a preexisting affective disorder (4.3%) (17).

While mesial temporal surgery is known to carry the risk of potential neurocognitive decline, data for LITT remains limited. In Jermakowics et al., neuropsychological data was obtained in 20 patients (20). Prior to surgery, nine patients with dominant hemisphere seizures had impairment in at least one measure of verbal memory and/or confrontational naming while seven patients with nondominant hemisphere seizures had impaired scores on visual spatial and/or memory measures. There was a statistically significant decline in delayed verbal memory performance for dominant hemisphere patients only. Verbal memory score changes were not significantly different for seizure-free patients versus patients with persistent disabling seizures (20).

Performance on the WMS-IV Logical Memory subtest decreased following dominant mesial temporal LITT, but the change in the nondominant group did not reach statistical significance (20). Clinically significant changes were observed in both right and left procedures, with the incidence of significant cognitive decline occurring more commonly with dominant mesial temporal LITT. In terms of postsurgical cognitive improvement, one had better visual and verbal memory scores, one improved in visual memory performance, two demonstrated better verbal memory, and three improved in confrontational naming in the nondominant group. In the dominant hemisphere group, one patient had better visual memory performance and one patient showed improvements in both visual memory and naming scores (20).

One study examined varied cognitive outcomes for 26 patients 6 to 12 months after MTLE MRgLITT (21). At the time of neurocognitive testing, 21 (81%) patients were free from disabling seizures, five (19%) experienced a reduction but continued to experience recurrent events. In line with previous research, presurgical evaluations revealed dominant MTLE LITT patients demonstrated deficits in verbal learning and memory, whereas nondominant hemisphere patients scored lower on visually mediated tests (21).

When looking at cognitive domains outside of memory, improvement in at least one cognitive measure occurred in 90% of subjects with similar frequency in either temporal lobe. Most improvements occurred in non-memory domains including processing speed, mental set shifting, and fine motor dexterity. Although a select number of dominant hemisphere ablation patients experienced a decline in language and memory performance (one in verbal learning, two in delayed recall, two in contextual verbal learning, three in delayed visual recall, two in confrontation naming), overall outcomes were favorable when compared to estimated rates of impairment following ATL (21,22).

Hypothalamic Hamartomas

Hypothalamic hamartomas (HH) are a non-neoplastic malformation located in the ventral hypothalamus, commonly associated with gelastic seizures. Endoscopic and open resections result in approximately half of patients becoming seizure-free. Still, the deep surgical location imparts an increased surgical risk. One study evaluated a heterogenous group of 71 patients who underwent MRg-LITT and found that 93% achieved seizure freedom from gelastic seizures, however only 12% were seizure-free without ASD on last follow-up (23). Complications occurred in only 1.5% of patients with once case having severe short term memory impairment and another with worsened diabetes insipidus, compared to the complication rate of 15% with open surgery (23).

Focal Cortical Dysplasia

Focal cortical dysplasia (FCD) is a type of malformation of cortical development characterized by abnormal cortical lamination, neuronal migration, and differentiation (24). Seizure freedom rates after surgical resection range from 50–70% and favorable prognostic factors include temporal location, complete resection, and an MRI detectable lesion (9,25). In one systematic review, MRgLITT report approximately 50% seizure freedom, though rates were slightly lower with 1-year follow-up. These series treated FCD in deep locations and had infrequent or transient complications (9). In a pediatric case series that included 12 patients with FCD, Engel I outcome was noted in 28.6% (2/7) of FCD after LITT (26). The study by Gupta et al. included five patients with FCD, all of whom achieved Engel I or II outcomes (27).

Cavernomas

Cavernomas are irregularly shaped enlarged capillaries that are frequently related to seizures. A study evaluated 17 patients with Cavernoma who underwent MRgLITT and found 82% of patients were seizure-free in the 12-month follow-up (14). There were no hemorrhagic complications. Two patients had temporary neurological deficits; one with motor hand deficit, and another with visual field deficit (14).

Extra Temporal Lobe Epilepsy

The data on the use of MRgLITT for extra temporal lobe epilepsy is limited. One meta-analysis investigated the efficacy of MRgLITT in drug resistant epilepsy including extratemporal lobe epilepsy showed pooled prevalence of Engel I outcome of 50% (95% CI 0.41% to 0.59%). Lesional cases had better Engel I outcome (65/128, 50.78%) compared to non-lesional cases (12/27, 44.44%) of extra temporal lobe epilepsy. The complication rates for extra temporal epilpesy were not reported however they noted motor deficits, including transient perioperative deficits, more commonly in insular cortex ablation (28).

In one multicenter prospective study, 60 patients with varied epilepsy subtypes underwent LITT (29). Engel I outcome was achieved in 64.3% at one-year follow-up with MTLE comprising 56.7% of this cohort. Other significant etiologies included FCD, HH, cavernoma, heterotopias, and tuberous sclerosis. Five adverse events were reported, one was categorized as serious. Two patients had headaches, one patient had a small subdural hematoma that resolved on its own, one patient had mild aphasia that improved in the 1-year follow-up. The patient with a serious adverse event needed a second ablation with a three-trajectory procedure which led to intraparenchymal hemorrhage requiring decompressive craniectomy (29).

In one study evaluating extra temporal lobe epilepsies which was also included in the meta analysis by Barot et al, 44% achieved Engel class I. Subgroup analysis revealed better outcomes for patients with lesional Extra Temporal Lobe Epilepsy than for those who were non-lesional, multifocal, or who had failed prior interventions. Favorable seizure-onset patterns on sEEG predicted higher likelihoods of success. Six patients had transient sensorimotor deficits and four patients had asymptomatic hemorrhages along the fiber tract. Two patients had major adverse events, one with brain abscess needing stereotactic drainage and one with persistent hypothalamic obesity. Three deaths that were unrelated to procedures occurred in this cohort, two were seizure-associated and one was a suicide (27).

Radiofrequency Ablation

RFA involves the establishment of a current flow between two electrodes to produce a thermoablative lesion through a burr-hole (Figure 2). A systematic review evaluated stereotactic RFA in HH in which 142 patients (76.8%) experienced some benefit and 69.2% achieved complete seizure freedom. Shirozu et al. made up the largest HH group (150 patients) of which 135 (90%) experienced freedom from gelastic seizures, 90 (60%) were free of non-gelastic seizures, and 110 (73.3%) had overall seizure freedom following RFA (30). The most common temporary complications reported were hyponatremia (52 patients, 28.1%), hyperphagia (48 patients, 25.9%), hyperthermia (47 patients, 25.4%), short-term memory loss (17 patients, 9.2%), disturbance of consciousness (4 patients, 2.25%), Horner’s syndrome (103 patients, 55.7%), weight gain (75 patients, 40.5%), and hemorrhage (9 patients, 4.9%) (30).

Figure 2.

Postoperative coronal T2 MRI of radiofrequency ablation of left cingulate gyrus cortical tuber of a patient with Tuberous Sclerosis Complex and drug resistant epilepsy.

A study evaluated 61 patients with MTLE who underwent MRI guided RFA with mean follow-up of 5.3 years, and found that 70.5% had Engel I, 9.8% had Engel II, 14.8% had Engel III, and 4.9% had Engel IV outcomes (31). The most serious complication was a small intracerebral hematoma in the probe trajectory and a small intraventricular hemorrhage causing obstructive hydrocephalus requiring temporary ventricular drain placement without further complications. Transitory anomia occurred in one patient while another had upper quadrantanopia. Two patients had small, clinically silent hematomas and one patient had a small, clinically silent, subdural hematoma that self-resolved. Two patients had meningitis requiring antibiotic treatment. Headaches were noted in 44 patients that were otherwise without any neurological complications. Neuropsychological testing was completed in 31 patients and showed no change in global memory in 86%, improved global memory in ten percent and worsened in four percent (31).

A recent meta-analysis compared RFA and MRg-LITT to open resection in patients with MTLE with or without MTS (32). An Engel I outcome was achieved using RFA in 44% (54/123, range=0%–67.2%), MRgLITT in 57% (315/554, range=33.3%–67.4%), selective amygdalohippocampectomy (sAHE) in 66% (887/1326, range=21.4%–93.3%), and ATL in 69% (1032/1504, range=40%–92.9%) of patients. The subgroup analysis between MRgLITT and RFA revealed no significant differences. There was a significant difference between RFA and ATL and sAHE, more pronounced when the follow-up was longer than 60 months. In contrast to other studies (16), Engel I outcomes for MTS was higher compared to MRI normal MTLE in this MRgLITT meta-analysis (32).

The use of sEEG is a well-tolerated procedure that can improve seizure outcomes through targeted RFA of the epileptogenic zone (33–36). The success of RFA can be used to guide surgical resection in patients with inadequate seizure control. In the study by Bourdillon et al., the positive predictive value of being responder (patients with >50% improvement) 2 months after sEEG-guided RFA and to be Engel’s class I or II after subsequent surgery was 93% while the negative predictive value was low (40%). Being non-responder appeared to be an unreliable predictor for outcome of surgery (37,38). One study evaluated 89 patients with varied epilepsy etiologies who underwent sEEG guided RFA of the epileptogenic zone. Twenty-five patients (28.1%) exhibited a persistent significant improvement in their seizures at 12 months. More favorable results were observed in patients with nodular heterotopia, those with a lesion found on MRI, and those with MTS, although the latter two did not reach statistical significance (35,39). Only one patient had a severe, unexpected neuropsychological impairment, one patient had an expected permanent motor deficit, and one pediatric patient had an expected transient motor deficit (35).

Another study evaluated 162 patients with varied epilepsy etiologies who underwent sEEG guided RFA over a ten-year period and found that 25% (41) were seizure-free in two months and 67% (108) were responders in two months, while 58% of responders maintained their status long-term. The seizure outcome was significantly better when the RFA involved the occipital region (40). Six patients experienced adverse events; two patients had permanent motor deficits which were expected. Transient neurological deficits included abnormal sensation, hemiparesis, brachyfacial palsy and aphasia (40,41).

One meta-analysis evaluated six studies on sEEG guided RFA and found one-year seizure-free rate varied greatly across studies (4%–71%). The pooled seizure-free rate was 23% (95% CI 8%–50%) which was not affected by the number or volume of RFA lesions. The pooled responder rate at one year (patients with 50% or greater reduction in seizure frequency) was 58% (95% CI 36%–77%) (37). In terms of tolerability, the pooled rate of permanent neurologic deficit was 2.5% (95% CI 1.2%–5.3%) reported in five patients across three studies (37).

Stereotactic Radiosurgery

SRS involves the use of focal ionizing radiation to destroy epileptogenic foci via a linear proton beam accelerator or photon/linear accelerators. Early applications in epilepsy were in patients with cerebral tumors, vascular malformations, and HH with varying seizure outcomes. Previous case series and reports have shown mixed results in MTLE (42). In 2004, Regis et al published the first prospective study on 20 MTLE patients who underwent SRS at 24 Gy radiation. Seizure freedom at 2 years was observed in 65% of patients, without neuropsychological worsening and nine patients had VFD (43).

The following prospective study included 30 MTLE patients with MTS who underwent SRS at either 20 or 24 Gy radiation. Seventy-seven percent of patients had seizure freedom in the 24 Gy group versus 59% in the 20 Gy group (44). In terms of adverse events, new onset headaches were observed in 10/47 patients in the low dose group versus 11/85 patients in the high dose group. Seven percent of patients had VFD in the low dose group vs 8% in the high dose group. Only three of 12 patients with dominant hemisphere and one of 14 with nondominant hemisphere SRS had verbal memory impairment with no subject experiencing decline on more than one measure on neurocognitive assessment. One patient in the high dose group had a serious adverse event of increased intracranial hypertension that resolved with temporal lobectomy. Both studies found initial increased aura frequency, before reduction or cessation of seizures which occurred after 12 months (44).

These two studies culminated in the only randomized trial comparing SRS to ATL. The Radiosurgery or Open surgery for Epilepsy (ROSE) trial was a 6-year multinational and multicenter study (45) in which 64% were seizure free overall with 52% (16/31) in SRS group and 78% (21/27) in ATL at 36 months. Of note, only 6% of SRS group was seizure free at 3-month follow-up versus 81% in ATL group suggesting immediate treatment effect compared with a more gradual increase in SRS treatment effect. Overall, this study failed to show non-inferiority of SRS as compared to ATL. Twelve (39%) patients in SRS group and three (11%) patients in ATL group experienced treatment related adverse events. In SRS group, adverse events were mostly seen between 11–27 months, corresponding to expected transient cerebral edema. Like previous studies, 65% of the SRS patients were on steroids. Visual field deficits occurred in 34% (10/29) of SRS patients and 42% (11/26) of ATL patients in early postoperative period while verbal memory measures were similar between the groups.

One meta-analysis compared treatment efficacy of SRS to MRgLITT and RFA, yielding a pooled seizure-free rate estimate per person-year of 0.38 (95% CI 0.14–1.00) for RFA, 0.59 (95% CI 0.53–0.65) for MRg-LITT, and 0.50 (95% CI 0.34–0.73) for SRS after a treatment delay of 24 months (46). Two of 133 (1.5%) patients died, one case of SUDEP and another of cerebellar hemorrhage.

A separate meta-analysis evaluating efficacy and complications of SRS and MRgLITT showed that mean incidence of overall seizure freedom at 12 to 36 months was comparable between the 2 procedures (MRgLITT 50%, 95% CI 44%–56% vs SRS 42%, 95% CI 27%–59%). Complication rate was 20% (95% CI 14%–26%) for MRg-LITT and 32% (95% CI 149% -46%) for SRS with VFD being the most common complication in both groups (n=12 in MRgLITT and n=21 in SRS). In the SRS group, cerebral edema (n=11), psychotic and cognitive symptoms (n=7) and cranial nerve deficits (n=2) were among the other complications (47).

One meta-analysis analyzing SRS in arteriovenous malformations showed that seizure control was achieved in 73.1% (95% CI 66.9%–78.9%) while seizure freedom was achieved in 55.7% (95% CI 44.5%–66.6%). The reported mean time to seizure control was 25.8±2.0 months (48) while another demonstrated overall seizure freedom rates of 0.49 (95% CI 0.38–0.59) with no death or permanent complication. Temporary morbidity was noted in 17.1% of patients (49).

Focused Ultrasound

FUS, a small target is heated by ultrasound with the aim of producing coagulative necrosis while minimizing harm to surrounding tissue (50). Focused US has been shown to be effective for tumors, and more recently for the treatment of essential tremor, Parkinson’s, and neuropathic pain (50). While showing efficacy in murine models, long term human data is limited as clinical trials for DRE are ongoing (51). One published study looked at sEEG guided FUS in six patients. Two patients were seizure free at three days and one patient had an increase in subclinical seizures. One patient developed transient naming and memory impairment that resolved in three weeks. Post-treatment MRIs demonstrated that the procedure was well tolerated without any radiological changes related to FUS (51).

NEUROMODULATION DEVICES

Vagus Nerve Stimulators (VNS)

The modern VNS first underwent animal testing in 1985, with first human implantations in 1988 (52,53). Since then, over 100,000 patients with drug resistant epilepsy (DRE) have had VNS implantation (54). The VNS device is an implanted pulse generator with a lead approximated to the left vagus nerve. As VNS devices have evolved over time, additional capabilities have also been added. Ictal tachycardia is known to be present in many patients with epilepsy (55). Automatic stimulation functionality allows a VNS to automatically deliver therapy following detection of sudden tachycardia (56), further reducing seizure frequency (57). The device can have separate settings for day or night which can be useful for some groups of patients. Scheduled programming allows a settings titration regimen on a pre-determined schedule without need for additional office visit. The device also detects bradycardia or a prone position after automatic stimulation or magnet stimulation, although this feature is currently only used for reporting purposes (54).

Recently, transcutaneous VNS has also been developed, whereby stimulation is applied transcutaneously to the vagus nerve from its cervical path or its auricular branch through the ear concha (58). While benefiting from a non-surgical implantation, the efficacy and role currently remain under investigation

The exact mechanism by which VNS improves seizure control is not fully determined; however, it appears to have both immediate and long-term reduction in seizure frequency over time (59). One proposed mechanism is that afferent vagus nerve stimulation results in nucleus tractus solitarius activation. This, in turn, projects diffusely to many different regions of the brain. Specific projections to the locus coeruleus and raphe nuclei result in norepinephrine and serotonin production, which may result in anti-epileptic effects (60).

The efficacy and safety of VNS were first evaluated in initial clinical studies in the 1990s. Two pivotal studies, E03 and E05, were parallel randomized controlled trials that evaluated responses to VNS at high and low levels of stimulation. As a result of these trials, VNS was initially approved by FDA in 1997 as adjunctive treatment of focal DRE for patients older than 12 years old. Approval was extended to 4–12 year olds in 2017 after the E06 and JPAS trials (61,62). In E03, patients with focal DRE were implanted and randomized to high- or low-level stimulation. Seizure frequency decreased by 24.5% while 31% of patients had at least 50% reduction in seizure frequency (>50% responder rate). No patients achieved seizure freedom during the trial (63).

The E05 study used a similar parallel study design comparing high and low level stimulation in patients with focal DRE. Similar degrees of improvement were reported, with approximately 28% reduction in seizure frequency in the high-stimulation group, compared with 15% reduction in the low stimulation group. In an open label extension, XE5, patients were followed for an additional 12 months. After this period, seizure frequency reduction improved to 45%. The >50% responder rate was 35%, and 20% experienced at least a 75% reduction in seizure frequency. No patients achieved seizure freedom (59).

Initial adverse effects were documented in the E03 and E05 trials. The most common adverse event was voice hoarseness or tremulousness during stimulation in a current-dependent fashion. While noted by most patients, it was only reported as an adverse event in 37.2% of patients in the high stimulation group and 13.3% in the low stimulation group (p<0.01) (63). Other side effects related to surgical intervention, such as superficial infection or nerve injury, occur at relatively low rates, <5% (58).

Deep Brain Stimulation

The deep brain stimulation (DBS) is more recently used in epilepsy while known to be effective in other disease processes such as tremor. Currently, there are a variety of DBS device systems available. Modern DBS utilizes an open-loop system allowing for parameter adjustments (64). Currently, no controlled trials compare DBS stimulation patterns for epilepsy treatment, (65) so DBS parameters may be determined by the clinician.

Like VNS, the exact mechanism DBS seizure reduction is not fully understood. One hypothesis One hypothesis is that epileptogenic networks involve corticothalamic connections. Regular stimulation of thlamus may inhibit nearby neurons via sodium channel inactivation and activate more distant neurons and nearby fiber tracts. In doing so, DBS would provide a desynchronization of epileptic networks (66).

There have been multiple potential targets that have been evaluated for DBS for epilepsy. The most examined target is the anterior nucleus of the thalamus (ANT) (Figure 3). The ANT is part of the Papez circuit as well as having a wide range of influence throughout the cortex. Based on studies done by Cooper (67) and subsequent others, a multicenter randomized placebo-controlled trial involving DBS of the bilateral ANT in Epilepsy (SANTE) was completed in 2010 (68). This study led to FDA approval for DBS of the bilateral ANT for adjunctive treatment of focal DRE in adults (69). This study assessed 110 focal DRE patients with at least 6 seizures per month. After a three-month baseline, DBS was implanted with patients randomized to active or sham stimulation. After three months, active DBS patients experienced a mean seizure reduction of 40.4%, compared to 14.5% in the sham group. >50% responder rates, Liverpool Seizure Severity Scale (LSSS) scores, and Quality of Life in Epilepsy (QoLIE-31) scores did not significantly change during this time frame. These patients were studied at 13 and 25 months with improved seizure frequency reduction: 41% and 56%, respectively. There were also improvements in >50% responder rates, LSSS, and QoLIE-31with increased follow up (68). A 10-year follow-up cohort study disclosed that, at seven years post-implantation, median seizure frequency reduction from baseline was 75%, with a >50% responder rate of 74%. 18% of patients were seizure free for at least 6 consecutive months, including nine (8%) subjects who were seizure-free for more than 2 years. Improvements were demonstrated on QoLIE-31 and LSSS and the percentage of patients reported to be satisfied or greatly satisfied was 84% (70).

Figure 3.

Coronal CT scout image of deep brain stimulator leads targeting bilateral anterior nucleus of the thalamus.

Adverse device-related events were reported from the SANTE trial, the most common of which were parasthesias in 18.2% of patients, implant site pain in 10.9% and implant site infection in 9.1%. Within the blinded phase, eight patients in the active stimulation group reported depressive symptoms compared to one in the control group; seven of these patients had a history of depression, and four had resolution of symptoms within three months. Memory impairment was reported in 13.0% of active stimulation patients; the severity was deemed to be non-serious and resolved in all patients (68). In the 10-year follow up, the most frequent adverse events included implant site infection in 12.7% and leads not at target in 8.2%. 37.3% of subjects at seven years reported symptoms of depression, with two-thirds of these patients having a history of depression. Eight deaths occurred, none of which were determined to be due to the implant, device, or therapy. One case of SUDEP occurred in the unblinded phase, and one possible and one definite case of SUDEP occurred in the long-term follow-up which is lower than the rate expected in patients with DRE (70).

The centromedian nucleus of the thalamus (CM) is another target for the treatment of DRE. Only one controlled double-blind crossover trial exists, with seven patients receiving repetitive intermittent stimulation for two hours daily. One group was treated with active stimulation or placebo for 3 months, followed by a three-month washout period, followed by three months cross-over treatment. Tonic-clonic seizure frequency decreased by 30% with active stimulation, compared with 8% with placebo; this was not statistically significant. The long-term extension phase of the study, however, did demonstrate that 3 of the 6 patients reported at least a 50% decrease in seizure frequency (71). Additional studies have suggested that CM stimulation may be more efficacious in generalized seizures compared to partial onset seizures with impaired awareness (72). A separate non-randomized single-blind controlled trial demonstrated all six patients with generalized epilepsy displayed >50% reduction in seizure frequency in the initial blinded phase, and 5 of 6 in the long-term phase. This contrasted to 1 of 5 frontal lobe epilepsy patients and 3 of 5 patients in the long-term phase (73). Given the low patient population, assessments of safety are rather limited. In the former study, one patient experienced asymptomatic post-operative hemorrhage. A cognitive battery was completed and there was no significant difference between baseline, placebo, and active treatment states (71). In the latter study, one patient required device explantation after 6 months due to infection, one patient had transient agraphia for 4 days (73).

The hippocampus is a third site that has been evaluated with DBS. A prospective double blind randomized controlled study assessed hippocampal DBS in drug resistant temporal lobe epilepsy over a 6-month blinded phase. Of the eight patients in the active treatment group, seven had a >50% reduction in seizure frequency, and four were seizure free within the time frame (74). Six of these patients had MRI evidence of MTS; interestingly, an earlier study did suggest that patients with normal MRIs had greater seizure reduction than those with evidence of hippocampal sclerosis (>95% vs 50–70%) and responded to DBS more quickly (75). Reported adverse events in these studies were skin erosions in two (12.5%) and three patients (33%), respectively, although sample sizes were small (74,75). In one study, there was no decrease in memory testing scores after stimulation (75). Other targets that have been explored also include the caudate nucleus, subthalamic nucleus, hypothalamus, caudal zona incerta, nucleus accumbens, fornix, (65) and medial pulvinar (76).

Responsive Neurostimulation

The brain-responsive neurostimulation (RNS) (Figure 4) differs from the above discussed methods of neuromodulation. RNS is a closed loop system delivering stimulus based on device recognition of prespecified electrocorticography (ECoG) patterns defined by physicians (eg. a seizure pattern). This contrasts to the open loop paradigm for VNS (excluding AutoStim) and DBS that deliver a prespecified metronomic stimulation regardless of ictal or non-ictal state (77).

Figure 4.

a, b. Coronal CT scout images of Responsive Neurostimulation System. Bilateral depth electrodes targeting bilateral hippocampus in a patient with bilateral mesial temporal epilepsy (a). Bilateral depth electrodes targeting bilateral supplementary motor area (b).

Additional utilization of RNS has suggested additional possible benefits. First, it can potentially predict seizures up to a few days before a seizure occurs. One study assessed an initial cohort of 18 patients and a validation cohort of 157. The individual patient’s training dataset, which included various temporal features were input into point process generalized linear models to create a probability over time that a seizure may occur. The model was able to reliably predict a seizure within the next calendar day for 83% in the former cohort, and 66% in the latter. This was able to be extended to three days in 11% and 39% of patients respectively (78). A second potential benefit to RNS is improved epilepsy localization. One study assessed 82 RNS patients with bilateral mesial temporal lead implantations. Implantation was bilateral in 71 patients due to presumed bilateral temporal lobe epileptogenic zones. However, follow-up (mean 5 years) revealed only unilateral onset seizures in 16% of the patients. Furthermore, 27.4% of patients did not have bilateral seizures until after four weeks, a time point commonly outside of inpatient intracranial EEG. Of the 11 patients with suspected unilateral seizures, seven (64%) had bilateral seizure onset, captured a median 35 days after the contralateral seizure (79). Therefore, long term ECoG may provide additional clinical information for patients that can inform future surgical intervention (80).

A full understanding of the mechanism by which the RNS reduces seizures is not fully known, but similar to all neuromodulation, likely has both immediate and delayed effects. The idea for RNS arose from the observation that provoked epileptiform afterdischarges during cortical mapping could be aborted by administration of brief pulses of electrical stimulation (81). As such, it was hypothesized that epileptiform discharges at a seizure focus may be able to be extinguished, potentially reducing the likelihood of seizure propagation. Additionally, given long term improvements in seizure control, there are still to be understood mechanisms that affect the underlying epileptic network. Studies on movement disorder neuromodulation implicate local modulation of neurotrophic factors and network-distributed changes in gene expression (77).

The RNS was approved by the FDA in 2013 as an adjunctive therapy in reducing the frequency of seizures in patients older than 18 years old with focal DRE, with frequent and disabling partial onset seizures, and two or fewer epileptogenic seizure foci (82). The pivotal study that led to approval was a double-blind randomized controlled trial assessing safety and efficacy of the RNS in patients with drug-resistant epilepsy. 191 patients were implanted with the RNS and ECoG was monitored. They were randomized to treatment or sham groups. After a two-month optimization phase, the patient entered a 12-week blinded evaluation period. Mean seizure frequency reduction was 37.9% in the treatment group, compared to 17.3% in the sham group. Additionally, 29% of patients in the treatment group, and 27% of patients in the sham group, had >50% reduction in seizures. In a 2-year open label-follow up, the >50% responder rate had increased to 46%. QoLIE-89 had improved in both treatment and sham groups, and improvements in all patients at 1 and 2 years in measures of verbal functioning, visuospatial ability, and memory (83). A long-term follow-up study displayed a median seizure reduction of 58% after 3 years and 75% after 9 years in a group of these patients. The >50% responder rate was 73% after 9 years, with a >90% reduction in 35% of patients. QoLIE-89 displayed continued improvement over 9 years (84).

The most common adverse events (AE) were implant site pain and headache at 15.7% and 10.5% at 1 year, respectively; with most AEs occurring within the 4-week post-op period and later incidence declining. Intracranial hemorrhage occurred in 9 patients (4.7%), 6 of which occurred post-operatively; the other 3 were due to seizure-related head trauma. Implant or incision site infection occurred in 5.2% of subjects (10 patients), requiring explantation in 4 patients. All infections involved soft tissue only (83). In the long-term follow-up study, implantation side infection was reported in 12.1% of all participants, none involving meningitis or encephalitis. Status epilepticus occurred in 8.2% of patients, 89.7% of which were not device related. Depression was reported in 64% of patients, of whom 71% had a history of depression. The depression was deemed not device related in 82%. Non-device related memory impairment was reported in 12.5%, with69% having memory impairment before enrollment. The rate of probable or definite SUDEP was 2.8 per 1000 patient-stimulation years, which is lower than that of epilepsy surgical candidates (9.3 per 1000 patient-years) (84).

CONCLUSION

Our understanding and management of epilepsy over the past few decades have greatly increased. This has led to the development and implementation of newer and ever-evolving methods to control seizures in patients with DRE, as well as increased access to new methods. Minimally invasive surgical techniques can ablate an epileptic focus, and in doing so, can be curative in a number of cases. The minimal invasiveness appears to reduce the relative frequency and severity of post-surgical adverse events, albeit with mildly lower efficacy rates (32). In cases where surgery is not an option, the development of neuromodulatory devices can benefit patients with seizure frequency reduction and relatively minimal AE rates.

While neuromodulation has demonstrated various levels of safety and efficacy, they also uncover additional questions regarding the nature of epilepsy and its management, as well as opening additional therapeutic avenues. Evaluation of minimally invasive surgical methods requires additional research into their relative clinical utility with respect to invasive surgical methods. While initial outcome data for neuromodulation is positive, continued long term follow up studies as well as research examining the underlying mechanisms of action are crucial for next steps. Additional devices, such as transcutaneous vagus nerve stimulation, or DBS applied to regions outside ANT require more data. Overall, however, these discussed advancements provide additional options for treatment of DRE that improve seizure control and quality of life.

Author’s Note: Daniah Shamim and Obiefuna Nwabueze are equal contributors to this work and designated as co-first authors.