Anterior nucleus of the thalamus deep brain stimulation vs temporal lobe responsive neurostimulation for temporal lobe epilepsy

Abstract

Objective:

Based on the promising results of randomized controlled trials, deep brain stimulation (DBS) and responsive neurostimulation (RNS) are used increasingly in the treatment of patients with drug-resistant epilepsy. Drug-resistant temporal lobe epilepsy (TLE) is an indication for either DBS of the anterior nucleus of the thalamus (ANT) or temporal lobe (TL) RNS, but there are no studies that directly compare the seizure benefits and adverse effects associated with these therapies in this patient population. We, therefore, examined all patients who underwent ANT-DBS or TL-RNS for drug-resistant TLE at our center.

Methods:

We performed a retrospective review of patients who were treated with either ANT-DBS or TL-RNS for drug-resistant TLE with at least 12 months of follow-up. Along with the clinical characteristics of each patient’s epilepsy, seizure frequency was recorded throughout each patient’s postoperative clinical course.

Results:

Twenty-six patients underwent ANT-DBS implantation and 32 patients underwent TL-RNS for drug-resistant TLE. The epilepsy characteristics of both groups were similar. Patients who underwent ANT-DBS demonstrated a median seizure reduction of 58% at 12–15 months, compared to a median seizure reduction of 70% at 12–15 months in patients treated with TL-RNS (p > .05). The responder rate (percentage of patients with a 50% decrease or more in seizure frequency) was 54% for ANT-DBS and 56% for TL-RNS (p > .05). The incidence of complications and stimulation-related side effects did not significantly differ between therapies.

Significance:

We demonstrate in our single-center experience that patients with drug-resistant TLE benefit similarly from either ANT-DBS or TL-RNS. Selection of either ANT-DBS or TL-RNS may, therefore, depend more heavily on patient and provider preference, as each has unique capabilities and configurations. Future studies will consider subgroup analyses to determine if specific patients have greater seizure frequency reduction from one form of neuromodulation strategy over another.

1 |. INTRODUCTION

With ongoing development of neuromodulation for epilepsy, patients with drug-resistant epilepsy and their health care providers now have several new options. Although ablative and resective procedures remain a mainstay of surgical treatment for epilepsy, deep brain stimulation (DBS) and responsive neurostimulation (RNS) are particularly important tools in cases of epilepsy that involve eloquent or multifocal regions of cortex. Both devices are often utilized for temporal lobe epilepsy (TLE); however, previous publications provide little evidence with respect to relative outcomes to guide the decision between DBS and RNS.

In randomized controlled trials (RCTs), both DBS of the anterior nucleus of the thalamus (ANT) and RNS were independently shown to have efficacy in reducing seizure frequency.^1–3 In long-term follow-up studies, the benefits of neuromodulation were durable, with gradual increases in median seizure frequency reduction over time. For ANT-DBS trial subjects, median reduction in seizure frequency at 7 years was 75% overall, with a 78% median seizure frequency reduction in the subset of subjects diagnosed with TLE.⁴ Long-term follow-up of RNS trial subjects at 9 years demonstrated a median seizure frequency reduction of 75%, with a 73% median seizure frequency reduction in the subset diagnosed with TLE.⁵

TLE is one of the most common forms of focal epilepsy and has been a longtime focus of surgical management.^6,7 A landmark RCT in 2001 revealed that 64% of patients who underwent temporal lobectomy (as-treated group) for drug-resistant TLE were free of seizures impairing awareness at 12 months postoperatively.⁸ With advances in surgical techniques, minimally invasive ablation of the medial temporal lobe has been adopted increasingly.⁹ One recent meta-analysis of laser ablation for mesial TLE showed that at 2 years postoperatively, 80% of patients achieved either Engel class 1 or 2 outcomes.¹⁰ Nevertheless, for patients with dominant temporal lobe or bitemporal onset epilepsy, destructive procedures may be avoided due to concerns for postoperative cognitive and/or emotional adverse effects.

To better understand whether a particular type of neuromodulation is more effective in drug-resistant TLE, we performed a retrospective review to understand the outcome of patients who were treated with either ANT-DBS or temporal lobe (TL) RNS.

2 |. METHODS

2.1 |. Study design

All adult patients who had undergone either ANT-DBS or TL-RNS for TLE at Emory University until December 2021, with at least 12 months of follow-up, were included. Approval for this retrospective review was obtained from the Emory University School of Medicine Institutional Review Board. Of note, Emory University was a clinical site for both the Stimulation of the Anterior Nucleus of the Thalamus for Epilepsy (SANTE)² and the Responsive Neurostimulation Pivotal RCTs.³

All patients who underwent ANT-DBS or TL-RNS were discussed in a multidisciplinary conference, which included neurosurgeons, epileptologists, neuroradiologists, and neuropsychologists. Clinical data discussed at these conferences included semiology, neuroimaging, neuropsychological assessment, and neurophysiology. Localization and lateralization of TLE was determined on a case-by-case basis, considering all data. A designation of bitemporal involvement considered neurophysiological interictal findings, ictal onset and spread patterns, intracranial stimulation results, as well as the above-described data. After agreement on localization, a recommendation of ANT-DBS or TL-RNS was made for all patients included in this study. Patients ultimately elected to proceed with either ANT-DBS or TL-RNS after further discussion regarding the risks, benefits, and alternatives of the procedures.

2.2 |. Surgical procedures

ANT-DBS and TL-RNS procedures were performed by either of two neurosurgeons (R.E.G. or J.T.W.) at Emory University Hospital. Four different standard stereotactic methods were used for electrode implantation, including frameless, frame-based, robot-assisted, and magnetic resonance imaging (MRI) guided. For ANT procedures, targeting included identification of standard landmarks of the anterior commissure (AC), posterior commissure (PC), and midsagittal plane. Frame-based procedures were performed primarily using a CRW Frame (Integra LifeSciences). Robot-assisted procedures were performed using the ROSA system (Zimmer Biomet). Finally, direct real-time MRI-guided procedures were carried out using the ClearPoint system (ClearPoint Neuro) in an interventional MRI scanner (Siemens Espree 1.5T MRI, Siemens).

For ANT-DBS, Medtronic DBS leads (Model #3387 or #3389, Medtronic, Inc.) were used for implantation. In the nine patients who participated in the SANTE trial, initial indirect coordinates were based on the AC/PC (Talairach) coordinate system, with trajectories adjusted subsequently for anatomic considerations (avoidance of vessels and sulci) and further refined based on modeling via the Schaltenbrand-Wahren Atlas in FrameLink software (Medtronic, Inc.), and finally microelectrode recordings. For all other patients implanted with ANT-DBS subsequent to the SANTE trial, targeting was anatomically based by identifying the point at which the mammillothalamic tract entered the ANT on MRI (T2-weighted and FGATIR-weighted sequences),¹¹ and refined on FrameLink software and/or CranialVault software (Neurotargeting).

For TL-RNS, 4-contact NeuroPace leads (Models CL-325-10-K, DL-330-3.5-K, DL-330-10K, DL-344-10-K, NeuroPace) were implanted. Unitemporal hippocampal depth lead implantation typically involved two trajectories in an oblique fashion, which allowed for targeting of the anterior and posterior medial temporal structures separately. For patients with bitemporal lead placement, depth leads were implanted along the long axis of the hippocampus. For all hippocampal leads, the dorsal/ventral and medial/lateral center of the hippocampus was targeted (i.e., dentate gyrus), with the distal contact placed in the amygdala. Depth electrodes were connected to the NeuroPace device in all cases except for two patients with lateral temporal-onset zones, in whom strip electrodes were placed; optimal location was identified anatomically using a neuronavigation system and intraoperative fluoroscopic radiographs and correlated with prior invasive electroencephalography (EEG) electrode contact locations.

2.3 |. Neuromodulation initiation and programming

Neuromodulation programming sessions were supervised by epileptologists trained in neuromodulation (including A.A., K.L.B., B.T.C., I.K., A.R., and D.W.). Battery life and lead impedances were checked during programming visits. Seizure frequency and adverse events were documented during each clinical visit.

For ANT-DBS, electrode contacts were selected based on localization within anatomic reconstructions using CranialVault (Neurotargeting). In general, the initial programming visit occurred at 2 and 4 weeks after implantation of the internal pulse generator (IPG). A monopolar stimulation protocol was typically used as the initial setting with 0.5–1 V, pulse width 90 μs, and frequency 145 Hz, as reported previously in the SANTE trial.² Cyclic stimulation was usually used, starting with 1 min of stimulation (“on”) followed by 5 min without stimulation (“off”). Follow-up visits occurred every 1–3 months, at which stimulation voltage was increased by 0.5–1 V at each visit depending on tolerability.

For TL-RNS, the device began collecting electrocorticography (ECoG) data at the time of implantation. The initial programming visit typically occurred 2–4 weeks postoperatively, during which ECoG tracings were reviewed and correlated with the patient’s documentation of seizures, which allowed adjustments to be made to detection settings. Initial stimulation settings were typically 0.5–1 mA, pulse width 160 μs, frequency 200 Hz, and 100 ms burst duration, using a bipolar configuration. Detection parameters and stimulation settings were adjusted at subsequent programming visits.

2.4 |. Clinical data

Retrospective chart review was performed to collect patient demographics and clinical details. This included operative reports, MRI images and reports, seizure frequency, follow-up dates, stimulation parameters, and adverse surgical or stimulation events. As part of their clinical care, patients were encouraged to use seizure diaries in the form of written, mobile app, or online calendars, though some patients opted to instead report intervisit average seizure frequency at clinical visits. At each visit, patients were asked about any adverse side effects; this documentation was gathered and compiled throughout each patient’s postoperative course.

Baseline seizure frequency was established using the following method. At the first postoperative neuromodulation visit, patients were asked to report their seizure frequency prior to implantation of their device. To ensure accuracy, this report was checked with clinical reports in the medical record within 6 months pre-operatively. Data used for comparison included clinical visits, epilepsy monitoring unit notes, and epilepsy multidisciplinary conference notes.

Because the benefits of ANT-DBS and RNS have been reported to increase over time,^4,5 we gathered clinical data throughout each patient’s clinical course. At clinical visits, patients would report their seizure frequency during the intervisit interval. Seizure frequency at each clinic visit was calculated using one of two methods. For patients who provided a total number of seizures in the inter-visit interval, the seizure frequency for that inter-visit interval was calculated as the total number of seizures divided by the number of days in the inter-visit interval, which resulted in a seizure frequency per day. For patients who provided an average seizure frequency since their last clinic visit, this value was calculated as seizures per day and extended to apply to each day of the intervisit interval. Clinical notes that did not provide numeric documentation, except for notes that stated that the seizure frequency had not changed from the prior visit, were omitted from the analysis. For overlapping intervisit intervals, which could occur due to the presence of multiple providers, the seizure frequency for each overlapping day was averaged. Due to heterogenicity of patient reports, total seizure frequency data, including focal aware type, were collected.

Seizure frequency was calculated for the following postoperative periods as an average over 3 months: 0–3, 3–6, 6–9, 9–12, 12–15, 15–18, 18–21, and 21–24 months. In addition, seizure frequency for the last 3 months of follow-up was calculated. Two approaches were used to determine seizure frequency during these periods. The first involved calculating the median using only reported per-day data points (as calculated above). However, this approach resulted in some patients having incomplete data during the specified time interval, as it was possible to not have contiguous clinic visits during that period. Therefore, a second approach was also used, using the last observation carried forward, which would cover these noncontiguous periods. This approach has been used previously in other studies on neuromodulation.¹² Ultimately, this approach allowed for a seizure frequency per day to be obtained for the entire time interval, up until each patient’s last follow-up, which was then averaged over the 3 months for each interval.

Percent change in seizure frequency was calculated by subtracting the postimplantation seizure frequency from the preimplantation seizure frequency and dividing by the preimplantation seizure frequency. Responder rate was calculated as the percentage of patients who had a reduction in seizure frequency of 50% or greater. In cases where subsequent procedures were performed, seizure frequency continued to be collected using an as-treated approach.

2.5 |. Statistical analysis

Due to the expected non-normalized distribution of the data, median values are reported. For statistical comparison, the nonparametric Wilcoxon rank-sum test was performed to determine whether the difference in effect of each treatment group was significant. For categorical variables, Fisher exact test was used, based on a consideration of sample size. We note that with a comparison of 26 vs 32 patients, we are powered to detect only a large difference in responder rate. For example, we have just over 80% power to detect a difference of a 69% responder rate in one group vs a 30% responder rate in the other group using the Fisher exact test (R package “Exact,” R version 4.1.1, www.r-project.org).

3 |. RESULTS

3.1 |. Patient characteristics

A total of 58 patients with drug-resistant TLE were included in this study: 26 treated with ANT-DBS and 32 treated with TL-RNS (Table 1). Overall demographic and epilepsy background characteristics were not significantly different between groups (p > .05). Median age at time of implantation of ANT-DBS patients was 38.5 years and median age of TL-RNS patients was 32 years. Fourteen of the ANT-DBS and 19 of the TL-RNS patients were women. The median time of last follow-up was 751.5 days for ANT-DBS patients and 1684.5 days for TL-RNS patients (Wilcoxon rank-sum test, p = .049).

TABLE 1.

Subject characteristics

	ANT-DBS	TL-RNS
Characteristic	N = 26	N = 32	p-value
Age at implantation (median)	38.5	32	.53
Gender (women)	14	19	.79
Median number of years with epilepsy	12	13	.87
Median baseline number of seizures (per day)	0.4	0.25	.6
Median number of epilepsy medications	2	3	.06
Median number of epilepsy medications trialed	5	4	.73
MRI findings
Normal	7	11	.58
Mesial temporal sclerosis	4	5
Postoperative changes	6	8
Heterotopia/dysplasia	1	1
Encephalocele/dural defects	3	1
Prior surgical procedures at time of implant
VNS	7 (3 removed)	5	.34
RNS	2 (1 removed, 1 nonfunctional)	0
DBS	0	1 (removed)
Resection or ablation	6	8	1
Lateralization
Unilateral temporal	11	16	.79
Bilateral temporal	13	16
Unable to determine	2	0
Location of seizure onset
Mesial	20a	30b	.27
Lateral	5a	3b
Unable to determine	5	0
Intracranial monitoring performed	16	26	.14
Median seizure frequency after implantation (per day)
Last 3 months of follow up	0.06	0.03	.41
12–15 month period	0.13	0.09	.59
Follow-up time (median)
Last follow up (days)	751.5	1684.5	.049
Median number of epilepsy medications at follow-up
Last follow up	2.5	3	.13
Median change in seizure frequency after implantation
Last 3 months of follow up	78%	84%	.41
12–15 month period	58%	70%	.77
Therapeutic responders (≥50% reduction in seizure frequency)
Last 3 months of follow up	18	24	.77
12–15 Month period	14	18	1
Additional neuromodulation postoperatively
DBS	–	5
RNS	1	–

Abbreviations: ANT, anterior nucleus of the thalamus; DBS, deep brain stimulation; N, number; RNS, responsive neurostimulation; VNS, vagus nerve stimulation.

^a3 ANT patients had both lateral and mesial onsets.

^b1 RNS patient had both lateral and mesial onset.

All patients agreed to have TLE based on clinician judgment and/or epilepsy surgery conference review of diagnostic data. Overall distribution in terms of localization and lateralization was not significantly different between the two groups. Twenty patients in the ANT-DBS group and 30 patients in the TL-RNS group were found to have mesial temporal lobe involvement, and 13 patients in the ANT-DBS group and 16 patients in the TL-RNS group had bitemporal involvement. Seven patients in the ANT-DBS group and 11 patients in the TL-RNS group had normal MRI, and 16 patients in the ANT-DBS group and 26 patients in the TL-RNS group underwent intracranial monitoring. There was no statistically significant difference between the two groups in any of these characteristics. Medical treatment with antiseizure medication was similar in both groups preoperatively, with a median of 2 medications in the ANT-DBS group and 3 medications in the TL-RNS group (Wilcoxon rank-sum test, p > .05). Patients in the ANT-DBS group had trialed a median of five antiseizure medications, whereas patients in the TL-RNS group had trialed a median of four antiseizure medications (Wilcoxon rank-sum test, p > .05).

Rates of prior surgical treatment in the ANT-DBS and TL-RNS groups were not significantly different. Six patients in the ANT-DBS group and eight patients in the TL-RNS group had undergone prior resective or ablative surgery. Preoperatively, seven patients in the ANT-DBS group and five patients in the TL-RNS group underwent concurrent neuromodulation with vagus nerve stimulation (VNS). In the ANT-DBS group, the settings of the VNS devices either remained constant (three patients) or were decreased (one patient). Similarly, in the TL-RNS group, the settings of VNS devices either remained the same (two patients) or were decreased (three patients).

Finally, a small number of patients from both groups underwent additional neuromodulation after the initial ANT-DBS or TL-RNS procedure. One patient in the ANT-DBS group was implanted with TL-RNS, which occurred after 36 months. Five patients were implanted with ANT-DBS after being treated with TL-RNS therapy; these occurred after 15, 17, 33, 56, and 63 months. Finally, one patient in the TL-RNS group had an ablation performed at 28 months.

3.2 |. Clinical outcomes

When considering clinical outcome at the last 3 months of follow-up (median last follow-up time of 751.5 days for ANT-DBS patients and 1684.5 days for TL-RNS patients), median seizure frequency reduction was 78% in the ANT-DBS group and 84% in the TL-RNS group (Figure 1, Wilcoxon rank-sum test, p > .05). To address whether the disparate time to last follow-up was a confound, since prior long-term studies have indicated that effect of ANT-DBS or RNS could increase in a time-dependent manner, we also compared clinical outcome at 3-month intervals, up to 24 months after implantation of the devices. During this specific time-frame, patients with ANT-DBS did not have a significantly different seizure frequency reduction compared to patients who were treated with TL-RNS, using either actual clinical data (Figure 1) or last-observed carried-forward approach (Figure 2). For instance, in the 12–15 month follow-up period, patients with ANT-DBS had a median seizure frequency reduction of 58%, compared to patients with TL-RNS who had a median seizure frequency reduction of 70% (Wilcoxon rank-sum test, p > .05). Although antiseizure medication adjustments could be made throughout the follow-up period, there was no significant difference in the number of antiseizure medications at last follow-up (Wilcoxon rank-sum test, p > .05).

FIGURE 1.

Treatment responses at 3-month intervals, up to 24 months. Differences between ANT-DBS and RNS, at each time point, were not significant (Wilcoxon rank-sum test, p > .05). Box plot: Upper boundary indicates 75th percentile; lower boundary indicates 25th percentile. Red line indicates median. Whiskers show 1.5 times the interquartile range. Comparison at each time interval, using Wilcoxon rank-sum test, demonstrated no significant difference (p > .05). ANT-DBS, anterior nucleus of the thalamus deep brain stimulation; RNS, responsive neurostimulation.

FIGURE 2.

Last observation carried forward approach with treatment responses at 3-month intervals, up to 24 months. Differences between ANT-DBS and RNS, at each time point, were not significant (Wilcoxon rank-sum test, p > .05). Box plot: Upper boundary indicates 75th percentile; lower boundary indicates 25th percentile. Red line indicates median. Whiskers show 1.5 times the interquartile range. Comparison at each time interval, using Wilcoxon rank-sum test, demonstrated no significant difference (p > .05). ANT-DBS, anterior nucleus of the thalamus deep brain stimulation; RNS, responsive neurostimulation.

Responder rates between the ANT-DBS and TL-RNS groups were similarly not significantly different at any of the 3-month intervals up to 24 months or during the last 3 months of follow-up (Fisher exact test, p > .05). For instance, during the 12–15 month follow-up period, the responder rate in the ANT-DBS group was 54% (14 patients), and the responder rate in the TL-RNS group was 56% (18 patients). In considering the last 3 months of follow-up, the responder rate was 69% (18 patients) in the ANT-DBS group and 75% (24 patients) in the TL-RNS group. Seizure freedom during the last 3 months of follow-up was also not statistically significant, with four patients in the ANT-DBS group (15%) and seven patients in the TL-RNS group (22%) reporting seizure freedom (Fisher exact test, p > .05).

These results were based on an as-treated approach. As described earlier, one patient in the ANT-DBS group was implanted with TL-RNS afterward, and five patients in the TL-RNS group were treated with ANT-DBS afterward. Furthermore, one patient in the TL-RNS group had an ablation performed postoperatively. Shortening the follow-up period in these select patients, such that the data collected reflected only treatment with one intracranial neuromodulatory therapy, did not result in significant differences in seizure frequency reduction between the ANT-DBS and TL-RNS groups.

Surgical adverse events were infrequent and included wound complications (two in the ANT-DBS group and two in the TL-RNS group) and intracranial hemorrhage related to lead placement (two in the ANT-DBS group and one in the TL-RNS group) (Table 2). Stimulation-related adverse effects were similarly uncommon and included headache and sensorimotor changes that resolved with adjustment of settings. One patient in the ANT-DBS group, who was part of the SANTE trial, committed suicide, which was considered secondary to psychiatric comorbidities and not directly due to DBS therapy.

TABLE 2.

Adverse events

	ANT-DBS	TL-RNS
Adverse event	N = 26	N = 32	p-value
Surgical (no. of patients)	5	5	.74
Wound infection	1	0
Wound dehiscence	1	2
Hardware malfunction	0	0
Hardware revision	1	2
Intracranial hemorrhage	2	1
Stimulation (number of patients)	9	18	.12
Headache	3	6
Mood change	1	1
Paresthesias	0	4
Drowsiness	3	0
Motor Twitching/jerking	0	5
Visual phenomenon	0	2
Memory changes	1	0

Abbreviations: ANT, anterior nucleus of the thalamus; DBS, deep brain stimulation; N, number; RNS, responsive neurostimulation.

4 |. DISCUSSION

In this study, there was no clear difference in seizure frequency reduction for patients with TLE who were treated with either ANT-DBS or TL-RNS. At 3-month intervals up to 24 months, and at the last 3 months of follow-up, there was no statistically significant difference in seizure frequency reduction. At 12–15 months after device implantation, we found a median seizure frequency reduction of 58% in patients treated with ANT-DBS and 70% in patients treated with TL-RNS. We did not find a statistically significant difference between the ANT-DBS and TL-RNS groups regarding the proportions of patients with unilateral vs bilateral or lateral vs mesial involvement. Neither did the 12-to 15-month responder rates (≥50% reduction in seizure frequency) differ significantly between ANT-DBS and TL-RNS. However, a limitation of this study is that our sample size was powered to detect only a large difference in responder rate.

Our median seizure frequency reduction rates appear to be higher than reported previously in the ANT-DBS and RNS trials. Although responder rates specifically for TLE were not reported, median seizure frequency reduction of ANT-DBS for temporal lobe seizures has been reported as 44.2% at the month 3–4 time point,² 44% at 1 year,¹³ 76% at 5 years,¹³ and 78% at 7 years.⁴ For RNS in mesial TLE, the median percent change in seizure frequency has been reported as 55% at the end of the open-label period at 26 months¹ and 66.5% (disabling seizures; interquartile range [IQR] 31.8%–93.7%) at 6 years. For all temporal lobe seizures (including mesial and lateral), a median percent change in seizure frequency of 73% (IQR 47%–93%) was reported at 9 years.⁵ The difference in median seizure frequency reduction may be due to differences in the measurement of seizure frequency, in patient populations, or ongoing seizure medication adjustments. Other factors that may have contributed to better than expected outcomes include our use of an open-label retrospective observational design rather than a randomized sham-controlled trial design, our center’s gravitation toward frequent utilization of diagnostic intracranial monitoring prior to neuromodulation, and our center’s high-volume experience with surgical placement and stimulation programming relative to the multicenter pivotal trials.¹⁴ It remains reassuring that the real-world seizure frequency reduction appears as good or better than the seizure frequency reductions reported in earlier clinical trials. Furthermore, our results are consistent with prior real-world experiences with neuromodulation.¹⁴ Of interest long-term studies of ANT-DBS and RNS have suggested slow gradual increases in median change in seizure reduction over time.^4,5 In our ANT-DBS and RNS patients, the overall differences in clinical response were not significant at multiple time intervals.

Our study is the first of its kind to directly compare therapeutic benefits between ANT-DBS and TL-RNS for drug-resistant TLE. To some extent, this may be due to limitations in access to both options for neuromodulation. For example, although ANT-DBS has been available outside the United States since 2008, it only became available in the United States in 2018. Conversely, RNS has been available in the United States since 2014 but is still not available outside the United States. Nevertheless, even now that both have been available in the United States since 2018, there are several form and functional features that differentiate the two therapies and which may influence the adoption of each by both patients and physicians.¹⁵

For RNS, the current version of the RNS device can target one or two epileptic foci, using either depth or strip electrodes. Because the RNS system depends on programming to identify seizures in the recorded ECoG, patients must upload data regularly so that clinicians can fine-tune detection and stimulation settings. Unique benefits to RNS additionally include potentially objective enumeration of electrographic seizure-like events, clarifying physiologic responses to medications¹⁶ and chronic seizure lateralization in presumed bilateral cases.¹⁷ In comparison, ANT-DBS has been an open-loop stimulation therapy, although there has been recent integration of technologies that allow for sensing of local field potentials (LFPs) around the DBS leads.¹⁸ An additional difference is the need for a craniectomy with RNS, as well as the need in many cases for prior invasive intracranial neurophysiological monitoring to localize the seizure-onset zones. Ultimately, the decision between using ANT-DBS or RNS may depend on patient preference and willingness or ability to comply with maintenance. Although this work specifically addresses patients with TLE, the general preference at our center is additionally influenced by clinical suspicion of involvement of a wider epileptic network. In these cases, DBS may potentially influence a wider network via the thalamocortical circuits.

Although our study was not statistically powered to perform analyses of subgroups of TLE (unilateral vs bilateral, medial vs lateral, etc.), other factors that may be worthwhile to examine in the future include epilepsy-specific factors, such as duration of epilepsy, electrophysiological findings, and imaging findings such as mesial temporal sclerosis. For instance, although analysis of patients with mesial temporal lobe epilepsy in the prospective RNS trials found no difference in clinical response based on the presence of mesial temporal sclerosis,¹² there fewer data indicating whether the presence of mesial temporal sclerosis influences the effectiveness of ANT-DBS.¹⁹ Further analyses on statistically powered, larger groups of patients will be needed to better understand whether specific patient factors should play a role in the selection between DBS or RNS in TLE.

Another form of DBS targeting the hippocampus in patients with temporal lobe epilepsy has been reported, with varying results in four small RCTs. In 2006, Tellez-Zenteno et al.²⁰ reported the results of a double-blind cross-over RCT of four patients for 3 month periods, with a 15% median seizure reduction in the stimulation group, which was not significant in comparison to the control group. Subsequent trials have had conflicting results. Velasco et al.²¹ reported, after a 1-month blinded period, the 18-month open-label results of a 93% median seizure frequency reduction with hippocampal DBS. McLachlan et al.²² showed, in two patients, a mean seizure frequency reduction of 33% in their double-blind, cross-over randomized trial over 3 month periods. Finally, Cukiert et al.²³ demonstrated the efficacy of hippocampal DBS in a randomized controlled double-blind trial of 16 patients, which showed a statistically significant reduction in focal aware and focal impaired awareness seizures at the end of the 6-month blinded phase.

The use of RNS for targeting thalamic nuclei, including the ANT, has been reported in a small number of cases.^24,25 Elder et al.²⁴ reported a case series of three patients with multifocal epilepsy affecting the frontal and temporal lobes who were treated with unilateral ANT-RNS. In these patients, the second RNS electrode array was a cortical strip placed over the region associated with seizure onset, which was similar to the strategy used in a case report by Herlopian et al.^24,25 After at least 33 months of follow-up, Elder et al. reported a 50%–60% reduction in disabling seizures.

As observed in both RCTs that established the efficacy of ANT-DBS and RNS, a subset of patients did not demonstrate improvement in median seizure frequency with neuromodulation. Ongoing questions remain about the management of these neuromodulation-resistant patients. At our center, patients who do not exhibit significant improvement have been re-investigated with invasive neurophysiology (stereo-EEG) to further characterize the seizure-onset zone. If a single focus is identified in a non-eloquent region, ablative procedures are offered. Indeed, in this work, one patient in the RNS group underwent this evaluation and demonstrated improved seizure reduction after ablation of an identified seizure focus. However, in patients with bilateral or multifocal seizure-onset zones that are not amenable to destructive procedures, treatment strategies include either switching between ANT-DBS and RNS or the combined use of both ANT-DBS and RNS. The optimal strategy has not yet been well identified, but our initial experience will be reported in a subsequent publication.

Limitations to our study include the number of patients, which reduces the statistical power in terms of comparisons of subgroups. In addition, the use of a single center in our data set may limit the generalizability to other epilepsy centers. The selection of patients who undergo neurostimulation at our center may be unique, such as a tendency to trial neuromodulation in patients with dominant hemisphere, lesional TLE. Medical management practices, such as the number of antiseizure medications used and trialed at our center, may be different from those at other centers. Finally, the stimulation protocols employed for the ANT-DBS and TL-RNS patients may be different from the parameters used at other centers, which could affect outcomes.

Other confounders include that the two cohorts were not completely concurrent, leaving a possible role for covert biases related to changes in practice patterns. In addition, was the lack of being able to control for other concomitant neuromodulation therapies, as small numbers of patients in our series had additional devices in place. We also did not specifically examine objective side-effect profiles, such as cognitive, psychiatric, and quality-of-life measures. Finally, the lack of definitive seizure diaries (i.e., the reporting of an average intervisit seizure frequency by patients) may have led to nonstandardized seizure frequency collection, which can also limit accuracy in our results, but is consistent with epilepsy management in real clinical practice.

Future work will attempt to identify whether subgroups of patients with temporal lobe epilepsy patients have the potential to benefit from specific therapies. Ultimately, an RCT of DBS vs RNS in a uniform patient population would be needed to better understand whether one form of neuromodulation is more effective. Such a study may be difficult to perform, as patients may resist randomization due to individual preferences for one device form factor or features over another.

5 |. CONCLUSION

We demonstrate that ANT-DBS and TL-RNS are similarly effective in the treatment of patients with drug-resistant TLE who are not candidates for destructive neurosurgical procedures. The choice between each distinct neuromodulation device or strategy to use in individual cases will ultimately require in-depth discussions among patients and their providers to identify how differences in device capability and form factor may impact quality of life.

Key Points.

• In a first-of-its-kind single-center retrospective review, anterior nucleus of the thalamus deep brain stimulation (ANT-DBS) and temporal lobe responsive neurostimulation (TL-RNS) for temporal lobe epilepsy (TLE) had similar median seizure reduction at 3-month intervals up to 24 months after device implantation, as well as similar rates of adverse events.

• ANT-DBS and TL-RNS for TLE both additionally demonstrated similar responder rates (percentage of patients with a decrease of 50% or more in seizure frequency).

• Given the similarity of clinical response, patient and/or physician preferences for form factor and unique features of each mode of neuromodulation may be more important in the selection between ANT-DBS and TL-RNS.