responsive neurostimulation in epilepsy therapy: some answers, lingering questions

4
Editorial Responsive neurostimulation in epilepsy therapy: Some answers, lingering questions1. Introduction In November 2013, the Food and Drug Administration (FDA) granted a premarket approval for the RNS System (NeuroPace, Inc.) of closed- loop, responsive stimulation for the treatment of drug-resistant focal epilepsy. The decision followed a favorable vote by the FDA's Neurolog- ical Devices Panel, an advisory board one of the authors (JEC) was a member of, in February 2013 [1]. The RNS System was designed to de- tect focal electrographic seizures in selected brain regions sampled with intracranial electrodes and respond by delivering electrical stimu- lation to those regions aimed at terminating the seizures [2]. The panel's endorsement of the device was primarily based on recently published results [3] and updates [1] from the RNS Pivotal Study, a multicenter, prospective, randomized, double-blind, sham stimulation-controlled trial, which assessed the safety and efcacy of the RNS System as adjunctive therapy of seizures from 2 seizure foci in adults with focal epilepsy that had failed to improve with 2 antiepileptic drug tri- als. During the study, 191 patients with an average of 3 disabling sim- ple partial, complex partial, or secondarily generalized seizures/month over a 12-week baseline period (BP) had the stimulator implanted. These were patients with severe forms of epilepsy: 32% had failed ther- apeutic epilepsy surgery, 34% had been implanted with a vagal nerve stimulator (VNS, which was removed prior to implanted with a), 59% had undergone intracranial monitoring, 55% had two seizure foci, and 50% had mesial temporal lobe epilepsy, of whom 73% had bitemporal epilepsy. The RNS System placement required a craniectomy. The device was connected to one or two of up to four intracranial quadripolar elec- trode leads consisting of subdural strips, no more than two depth elec- trodes, or a combination of depth and subdural electrodes localized to 12 epileptogenic region(s) previously dened by standard presurgical evaluation [3]. The RNS System was equipped with three computation- ally efcient algorithms (half-wave, line length, and area) for early sei- zure detection; two individual detectors could be congured for any two electrode channels [2]. After detection, one to ve stimulation ther- apies, each consisting of two independently programmable bursts, could be delivered sequentially according to several electrode congu- rations dictated by the spread of the seizure onset. All enrolled patients could undergo modication of the seizure detection parameters following device implantation. The subjects were randomized to sham (no stimulation, n = 94) and treatment (n = 97) groups one month after the implant. Following randomization, the subjects assigned to the treatment group began stimulation therapy. Stimulus parameters were adjusted weekly during a four-week optimization period (OP) based on self-reported seizure recurrence, occurrence of side effects to stimulation, and review of the electrocorticogram (ECOG) stored by the RNS System. Because of implant constraints, ECOG was not recorded continuously. The ECOG epochs stored for review included detections of abnormal electrical activity representing presumed ictal discharges, responsive stimulation treatment bursts and their effects on the elec- trographic ictus, and stimulation-related afterdischarges. A 12-week blinded evaluation period (BEP) followed, during which the treatment group continued to receive responsive stimulation, and the sham group did not. All subjects were assessed in monthly clinic visits, and their seizures were logged on a daily basis. All subjects, whether assigned to treatment or sham, were then able to receive stimulation during an 84-week open-label period (OLP) [3]. 2. RNS System clinical safety and efcacy: results and analysis 2.1. Safety The study met its primary safety endpoint, which required that the rate of serious adverse effects (SAEs) not exceed the prespecied SAE rate of historical comparators for the rst four (seizure-related acute implantation of intracranial electrodes and epilepsy resective surgery) and 12 weeks (deep brain stimulation for treatment of movement dis- orders) after the implant [1]. There was no statistically signicant differ- ence in SAEs between the treatment and sham groups during combined OP and BEP. There was also no difference in mood and neuropsycholog- ical status. The most nonserious adverse effects were implant site pain and headache. When all patients enrolled in RNS System trials (Feasibil- ity, Pivotal, and Long-term Treatment studies) were pooled (n = 256), there were a total of seven deaths due to SUDEP, which reects a rate below the corresponding prespecied comparator (9.3/1000 patient- years). One death was due to status epilepticus (SE). Overall, in the pooled population, implant site infection was reported in 6.3% of the pa- tients, premature battery depletion in 4.3%, medical device removal in 4.3%, intracranial hemorrhage not related to seizures in 2.7%, device lead damage in 3.1%, and revision in 2.7% [1]. There were no study with- drawals because of an adverse effect related to a change in seizures. Se- rious adverse effects due to increased seizure frequency or severity or a new seizure type occurred in 16% of the patients. Episodes of SE oc- curred in 3.9% of the subjects, with three episodes considered related to the device. One patient experienced nine SE episodes. The rate of Epilepsy & Behavior 34 (2014) 2528 Submission declaration: The work described has not been published previously and is not under consideration for publication elsewhere. The publication is approved by all au- thors, and it will not be published elsewhere, including electronically in the same form, in English, or in any other language without the written consent of the copyright holder. The views are solely those of the authors and not reecting the views of the University of Texas or the Veterans Administration. The copyright of this manuscript cannot be trans- ferred to the publisher as US laws require this to be in the public domain. http://dx.doi.org/10.1016/j.yebeh.2014.02.014 1525-5050/Published by Elsevier Inc. Contents lists available at ScienceDirect Epilepsy & Behavior journal homepage: www.elsevier.com/locate/yebeh

Upload: jose-e

Post on 23-Dec-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Responsive neurostimulation in epilepsy therapy: Some answers, lingering questions

Epilepsy & Behavior 34 (2014) 25–28

Contents lists available at ScienceDirect

Epilepsy & Behavior

j ourna l homepage: www.e lsev ie r .com/ locate /yebeh

Editorial

Responsive neurostimulation in epilepsy therapy: Some answers,lingering questions☆

1. Introduction

In November 2013, the Food andDrugAdministration (FDA) granteda premarket approval for the RNS System (NeuroPace, Inc.) of closed-loop, responsive stimulation for the treatment of drug-resistant focalepilepsy. The decision followed a favorable vote by the FDA's Neurolog-ical Devices Panel, an advisory board one of the authors (JEC) was amember of, in February 2013 [1]. The RNS System was designed to de-tect focal electrographic seizures in selected brain regions sampledwith intracranial electrodes and respond by delivering electrical stimu-lation to those regions aimed at terminating the seizures [2]. The panel'sendorsement of the device was primarily based on recently publishedresults [3] and updates [1] from the RNS Pivotal Study, a multicenter,prospective, randomized, double-blind, sham stimulation-controlledtrial, which assessed the safety and efficacy of the RNS System asadjunctive therapy of seizures from ≤2 seizure foci in adults withfocal epilepsy that had failed to improve with≥2 antiepileptic drug tri-als. During the study, 191 patients with an average of≥3 disabling sim-ple partial, complex partial, or secondarily generalized seizures/monthover a 12-week baseline period (BP) had the stimulator implanted.These were patients with severe forms of epilepsy: 32% had failed ther-apeutic epilepsy surgery, 34% had been implanted with a vagal nervestimulator (VNS, which was removed prior to implanted with a), 59%had undergone intracranial monitoring, 55% had two seizure foci, and50% had mesial temporal lobe epilepsy, of whom 73% had bitemporalepilepsy. The RNS Systemplacement required a craniectomy. The devicewas connected to one or two of up to four intracranial quadripolar elec-trode leads consisting of subdural strips, no more than two depth elec-trodes, or a combination of depth and subdural electrodes localized to1–2 epileptogenic region(s) previously defined by standard presurgicalevaluation [3]. The RNS System was equipped with three computation-ally efficient algorithms (half-wave, line length, and area) for early sei-zure detection; two individual detectors could be configured for anytwo electrode channels [2]. After detection, one to five stimulation ther-apies, each consisting of two independently programmable bursts,could be delivered sequentially according to several electrode configu-rations dictated by the spread of the seizure onset. All enrolled patientscould undergo modification of the seizure detection parameters

☆ Submission declaration: Thework described has not been published previously and isnot under consideration for publication elsewhere. The publication is approved by all au-thors, and it will not be published elsewhere, including electronically in the same form, inEnglish, or in any other language without the written consent of the copyright holder.The views are solely those of the authors and not reflecting the views of the University ofTexas or the Veterans Administration. The copyright of this manuscript cannot be trans-ferred to the publisher as US laws require this to be in the public domain.

http://dx.doi.org/10.1016/j.yebeh.2014.02.0141525-5050/Published by Elsevier Inc.

following device implantation. The subjects were randomized to sham(no stimulation, n = 94) and treatment (n = 97) groups one monthafter the implant. Following randomization, the subjects assigned tothe treatment group began stimulation therapy. Stimulus parameterswere adjusted weekly during a four-week optimization period (OP)based on self-reported seizure recurrence, occurrence of side effects tostimulation, and review of the electrocorticogram (ECOG) stored bythe RNS System. Because of implant constraints, ECOGwas not recordedcontinuously. The ECOG epochs stored for review included detectionsof abnormal electrical activity representing presumed ictal discharges,responsive stimulation treatment bursts and their effects on the elec-trographic ictus, and stimulation-related afterdischarges. A 12-weekblinded evaluation period (BEP) followed, during which the treatmentgroup continued to receive responsive stimulation, and the shamgroup did not. All subjects were assessed in monthly clinic visits, andtheir seizureswere loggedon adaily basis. All subjects,whether assignedto treatment or sham, were then able to receive stimulation during an84-week open-label period (OLP) [3].

2. RNS System clinical safety and efficacy: results and analysis

2.1. Safety

The study met its primary safety endpoint, which required that therate of serious adverse effects (SAEs) not exceed the prespecified SAErate of historical comparators for the first four (seizure-related acuteimplantation of intracranial electrodes and epilepsy resective surgery)and 12 weeks (deep brain stimulation for treatment of movement dis-orders) after the implant [1]. Therewas no statistically significant differ-ence in SAEs between the treatment and shamgroups during combinedOP and BEP. Therewas also no difference in mood and neuropsycholog-ical status. The most nonserious adverse effects were implant site painand headache.When all patients enrolled in RNS System trials (Feasibil-ity, Pivotal, and Long-term Treatment studies) were pooled (n = 256),there were a total of seven deaths due to SUDEP, which reflects a ratebelow the corresponding prespecified comparator (9.3/1000 patient-years). One death was due to status epilepticus (SE). Overall, in thepooled population, implant site infectionwas reported in 6.3% of the pa-tients, premature battery depletion in 4.3%, medical device removal in4.3%, intracranial hemorrhage not related to seizures in 2.7%, devicelead damage in 3.1%, and revision in 2.7% [1]. Therewere no studywith-drawals because of an adverse effect related to a change in seizures. Se-rious adverse effects due to increased seizure frequency or severity or anew seizure type occurred in 16% of the patients. Episodes of SE oc-curred in 3.9% of the subjects, with three episodes considered relatedto the device. One patient experienced nine SE episodes. The rate of

Page 2: Responsive neurostimulation in epilepsy therapy: Some answers, lingering questions

26 Editorial

adverse effects due to changes in seizures and that of SE episodes didnot exceed those reported in antiepileptic drug trials of refractoryfocal epilepsy or as would be expected in patients with severe focal-onset seizures [1].

2.2. Efficacy

Similar to reports from the anterior thalamic stimulation trial [4], animportant “surgical” or implant-related effect was apparent, in thattherewas approximately a 30%median percent reduction in seizure fre-quency following device implantation in both treatment and shamgroups and prior to the initiation of stimulation in the treatmentgroup [1]. The implant-related effect in the sham group largely resolvedby the fifth month after implantation, mostly due to very high seizurefrequencies in two subjects at the end of the BEP relative to baseline.In contrast, subjects in the treatment group maintained a postimplantseizure improvement throughout the BEP [1,3]. The positive implant-related clinical effect seemed to extend to quality-of-life changes: boththe treatment and the sham group experienced statistically significantand clinically meaningful improvements in the overall and subscalescores derivedwith QOLIE-89 inventory at the end of the BEP comparedwith the BP, although the difference between groupswas not significant[1]. Improvements in quality of life continued to increase during theOLPwhen all subjects could receive stimulation. This suggests that the stim-ulation itself improved quality of life in most patients but that the BEPmay have been too short to detect significant stimulation-related effectsabove those offered by the implant or sham.

The stimulation treatment effect was estimated in the primary effi-cacy analysis by the difference between the reduction in mean seizurefrequency noted in the treatment (37.9%, 95% confidence interval =27.7–46.7%) and that noted in the sham (17.3%, 95% confidenceinterval = 2.3–29.9%) group during the BEP relative to the BP [1]. Thetreatment effect was statistically significant (p b 0.012) when severalmodifications to the generalized estimating equation model prespeci-fied by the sponsor and agreed by the FDAwere applied post hoc to im-prove datafitting. Thesemodificationswere as follows:monthly insteadof daily seizure counts, negative binomial instead of Poisson distributionof the seizure counts, and inclusion of several clinical covariates used inrandomization [1]. The treatment effectwas not significantly dependenton seizure onset (mesial temporal vs. other), number of seizure foci(unifocal vs. bifocal), and prior epilepsy surgery vs. no surgery [1].Most benefit from stimulation was gained by subjects with a high num-ber of seizures at baseline (≥84 seizures/month). The greatest reduc-tion was noted in disabling simple partial motor seizures, with lessdecreases noted in complex partial and secondarily generalized seizures[1].

Additional efficacy measures during the BEP relative to the BP dem-onstrated a positive trend in treatment vs. sham subjects: responderrate (percent of subjects with ≥50% reduction in seizures, 29% vs.27%), change in mean seizure frequency (−11.4 vs. −5.3 seizures/month from baseline 33.8 vs. 35.2), median percent change in seizures(−28% vs. −19%), change in days with seizures (−19% vs. −18%from baseline 10.7 vs. 10.6 days/month), and mean change in a seizureseverity score [1]. None of the differences between groups for these sec-ondary outcomes was statistically significant. With the exception of theresponder rate, however, the study protocol did not specifywhether thetrial was adequately powered to detect significant differences.

Overall, 76% and 70% of the subjects in the treatment and shamgroups, respectively, reported a decrease in the number of seizures inthe BEP relative to the BP. A total of 2.1% of the treatment subjects andnone of the sham subjectswere seizure-free during the BEP. Conversely,20% and 19% reported no change or≤50% increase in seizure frequency,and 4% and 11% reported a ≥50% increase in seizures frequency in thetreatment and sham groups, respectively [1].

During the OLP, all subjects could receive stimulation and wereallowed changes in antiepileptic medications but were not disclosed

their prior assignments to either the treatment or the sham group. Sim-ilar to other neurostimulation modalities (VNS, anterior thalamic), theefficacy of RNS appeared to improve over time above that noted duringthe BEP. The proportion of responders (≥50% reduction in seizures)was43.6% at one year and 54.6% at two years. The median percentage sei-zure reduction was 44% at one year and 53% at two years. During theOLP, the subjects previously assigned to the sham group experiencedseizure reduction that approached the extent seen in those initiallytreated with stimulation [1]. The device's retention rate was high: 92%of implanted subjects completed the entire two-year study. Moreover,high patient satisfactionwith the devicewas reflected in that 93%decid-ed to replace their stimulator at the end of battery service (≤3 years).

Overall, the results of the RNS Pivotal Study support the applicationin the clinical realm of a very attractive concept that seizures can be ter-minated by concurrently applied electricity. This concept is grounded inobservation and solid experimental evidence (reviewed in [2,5]). TheRNS System is reasonably safe when compared with other neuro-invasive procedures. The device has a high retention rate by the patientsin the study, suggesting in part that the neurosurgical risk may be bal-anced overall by the absence of drug-related adverse effects, whichcould complicate the alternative of an exclusive use of pharmacologictherapy and comparatively higher antiepileptic drug doses. Efficacy re-sults are encouraging if mixed. A direct comparison between the RNSPivotal and VNS trials is difficult given different trial designs. Neverthe-less, it is reassuring that the RNS System showed a trend towards great-er seizure frequency decrease than VNS E03 and E05 trials during theblinded periods (37.9% vs. 24.5% and 27.9%, respectively) and compara-ble responder rates during the blind (29% vs. 31% and 23.4%, respective-ly) and at one year (43.6% vs. 35% in the E05 trial). However, RNSSystem use did not lead to a significant difference in responder ratesbetween the treatment vs. sham (29% vs. 27%) groups during theblinded period. This suggests that combinedwith effectiveness at reduc-ing seizures skewed towards patients with high baseline seizure counts,the blinded period in the pivotal trial may have been too short to allowfor significant results in patients with rarer seizures. Alternatively, thesefindingsmay be partly explained by highly variable device effectivenessin individual subjects associated with different baseline seizure fre-quencies reflecting various degrees of disease severity. Furthermore, itis important that long-term improvement bemeasured against the nat-ural history of the disease [6,7]. Based on the trial, a set of candidates tobenefit most from responsive neurostimulation is identified: patientswith severe refractory focal epilepsy, especially those with very fre-quent disabling seizures; those who are not candidates for epilepsy sur-gery; and those refusing surgery or in whom surgery did not have asatisfactory result.

3. RNS Pivotal trial and beyond: concerning rational electrotherapy

The RNS System trial results reviewed above invite several questionsabout how the RNS System can be used clinically on a case-by-case basisin future practice. First, did the RNS System work the way it wasintended to? In other words, to what extent was the reduction inseizures attributed to the activation of responsive stimulation duringthe trial due to immediate, seizure-aborting or mitigating effects ofstimulation or to longer-term presumed neuromodulatory influenceson seizure initiation or propagation? These questions are important be-cause the rational use of the device is based on the idea of immediatecessation of seizures by stimulation; as such, adjusting the detectionand stimulation parameters during the trial was based mainly on ashort-term, weekly-to-monthly sequential feedback of seizure recur-rence. That responsive neurostimulation is able to abort seizures record-ed electrographically has been demonstrated in small pilot studies(reviewed in [2]), but no adequate pivotal trial data to support it werepooled and released. Using the device without large-scale trial data toverify its operational premise may invite untoward outcomes in the fu-ture. Such an outcome may be merely a lack of efficacy at the cost of

Page 3: Responsive neurostimulation in epilepsy therapy: Some answers, lingering questions

27Editorial

cranial surgery, device maintenance, and operation. For example, a pa-tient may experience no initial overall clinical seizure reduction afterthe device was implanted and the first adjustments of detection andstimulation parameters have beenmade, despite examples of individualseizures recorded on ECOG that were indeed aborted after stimulation.In such case, should serial changes in parameters be performed to im-prove clinical seizure control in a reasonable amount of time, with theinherent trial-and-error risks, or could a neuromodulatory effect bebet on to improve seizure outcome? What parameter adjustments, ifany, and in which domain, detection or stimulation or both, have mostconsistently resulted in positive outcomes after an initial limited treat-ment response and may most likely be of benefit in this case?

Crucial to answering such questions is understanding the link be-tween the clinical and electrographic activity changes associated withthe device. A step to this end is establishing the relationship betweenthe rate of clinical seizures logged during the trial and a treatmentmarker, such as the rate of ictal discharges recorded on ECOG. It isnatural yet not proven to think that while suchmarker does not captureall stimulation treatment effects, including seizure duration, intensity,and spread, it is indeed linked to the prespecified outcome, i.e., clinicalseizure rate changes. The RNS Pivotal trial, however, did not provideelectrographic metrics such as determining the device's seizure detec-tion sensitivity and specificity nor did it provide electrographic surro-gate outcomes, such as the effect of stimulation on the rate of seizurescaptured by ECOG. Such determinations are not trivial. Calculating thesensitivity and specificity of the detection tools would require the re-cording of all clinical events or subclinical seizures, including the onesnot detected by the device. Assuming that lead implantation was accu-rate enough to detect all electrographic ictal discharges in each subject,such recordings would require long-term continuous ECOG epochsstored by the RNS Systemavailable for offline review, a feat not includedin the trial but recently reported in a groundbreaking first-in-man study[5]. In addition, calculating treatment metrics is complicated bythe need of both early seizure detection and early stimulation for thedevice to be effective in most cases [2]. Such a short window of oppor-tunity (several seconds) confounds detection specificity calculations,since stimulation therapy may be applied to self-limited, nonevolvinginterictal activity appearing quasi-identical to a would-be seizure dis-charge terminated by stimulation.

Nevertheless, we believe that no effort should be spared towards anevaluation of ECOG measures of detection and stimulation. In parallelwith continued research on the biophysics of electrical stimulation atmultiple neural scales [8], acute clinical trials set in the EMU [9] or in-volving continuous long-term portable ECOG recordings are necessaryto assess rigorously the performance of the RNS System. These trialscould be designed to include the statistical framework adopted byprior studies evaluating individual detection features adopted by theRNS System and stimulation treatments [9]. Applying the full (com-bined) detectionmodule of the device to a large-scale, publicly availableECOG database [10] may define better its ability to recognize a widerange of ictal patterns characterizing focal epilepsy. Working towardsthe release of an extracranial version of the device made available toEMUs nationwide for training and trial during phase II monitoringmay complement such studies in providing a large body of data of itsshort-term use and may guide lead implantation for subsequent long-term operation according to current specifications.

On a shorter time span, the disclosure of gathered ECOG data by thesponsor during the pivotal trial would also allow partial yet informativeanswers as to how the device performs. Potentially useful measures ofdetection accuracy include the difference and correlation between thenumber of daily definite ictal patterns (subclinical and clinical seizuresrecorded on ECOG and scored by expert reviews) that were detectedbut not terminated by stimulation (failed and suboptimal stimulations)and the number of daily clinical events recorded in the seizure log; foran individual subject, a consistently null or positive difference andhigh correlation over time would indicate a robust and clinically useful

level of detection. Stimulation efficacy could be estimated by comparingthe number of ictal patterns detected during a period when the stimu-lation was off with the number of electrographic seizures detected butnot aborted by stimulation during periods when stimulation was en-abled. A comparatively high value of the latter does not necessarilyimply ineffective stimulation, as it may be due to an increase in subclin-ical seizures at the expense of clinically overt ones or occur withincidental increases in ictogenicity in some patients during the on-stimulation periods; a relatively low number of seizures stimulatedbut not aborted, however, would confirm a positive electrographic re-sponse to stimulation. Charting the number of detections–stimulationsof epileptiform activity across time could help measure fluctuationsin ictogenic potential and, by doing so, would help uncover potentiallong-term neuromodulatory influences in addition to immediateseizure-aborting effects associated with stimulation. As such, a consis-tently downward trend in the number of epileptiformpatterns detectedover timewould support a gradual antiepileptogenic effect on epilepticnetworks that may explain in part the progressive improvement in sei-zure counts reported with long-term RNS use, similarly observed withother neurostimulation modalities (VNS, anterior thalamic).

Although these and similarmeasures applied to ECOGdatamay gen-erate a limited number of easy-to-interpret data patterns, they can offerimportant clues about how the RNS System works.

Second, have a few stimulus selection and optimization strategiescrystalized to a degree necessary for clinical application in individualpatients? Can such strategies be devised? Allowing stimulus locationto vary by targeting various supratentorial structures with intracranialelectrodes tuned to individual epileptic networks, the RNS System hasafforded unprecedented degrees of freedom in selecting stimulation at-tributes, even when compared with other stimulation devices, such asVNS or anterior thalamic. In addition, the RNS Pivotal trial has proposeda wide range of parameter values open for optimization for a number ofsequential stimulations, independently programmable bursts, and elec-trode configurations, to include current frequency (1 to 333 Hz), ampli-tude (0.5 to 12 mA), pulse width per phase (40 to 1000 μs), and burstduration (10 to 5000 ms). During the trial, except for 1-Hz pulse fre-quency and 1000-μs pulse duration, the full range of stimulus optionswas used, but how these options were selected is unclear. Although aset of initial stimulation settings was recommended, an optimizationstrategy for stimulation parameters was not prespecified. In contrast,such road map guiding stimulation changes was included in the proto-col of a previous short-term, EMU-based, closed-loop stimulation trial[9] or that of the VNS E03 trial, which included stepwise current incre-ments to tolerable levels [11]. This wide range of stimulus selectionadds to the high variability of electrographic phenotypes that RNS isintended to detect and respond to [8]. Optimization across that manyparameter options based on pure empiricism is not practical.

Without rational stimulus optimization strategies, the multitude ofsources of operational uncertaintymay render RNS use in commonprac-tice ineffective and, in selected cases, potentially deleterious given thecosts of device implantation. The challenges to stimulation optimizationdesign are many. A selected stimulus parameter set must be tried on aminimumnumber of events for proper statistical validation of its effects,which may not be realized during relatively short optimization andblinded evaluation periods [8]. Such longer periods may be necessaryfor improved optimization, but they are difficult to implement. Withoutelectrographic metrics and surrogate treatment markers, such as thosediscussed in the prior paragraph, defining the immediate seizure-aborting effects of stimulation and the timeline of an appropriate opti-mization schedule is not possible. Optimal stimulation parameters mayvary considerably among patients and targeted networks. In addition,defining a dose–response relationship between stimulus attributevalues and seizure occurrence, presently not required by the FDA formedical device approval, or defining reasonable assumptions of it is nec-essary for rational optimization strategies. Such a dose–response curvemay not be smooth or monotonous and is also likely to vary among

Page 4: Responsive neurostimulation in epilepsy therapy: Some answers, lingering questions

28 Editorial

subjects. While challenges of an effective, rational stimulation selectionand optimization strategy briefly exposed here seem daunting, severalshort- and medium-term measures may be of benefit. First, we encour-age the release by the RNS pivotal trial team of outcome data stratifiedaccording to whether preimplant intracranial extraoperative monitor-ing was performed to characterize the epileptic foci and determine thelocation of future stimulation delivery. Based on these data, recommen-dations as towhether suchmonitoring restricted toNAEC level-4 epilep-sy centers is needed before device implantation can be made. Second,the publication of subject-specific outcome data together with individu-al optimized stimulus parameter sets (i.e., inmany cases, the latest set tobe selected for a subject and active at the last follow-up during the OLPor during the Long-termTreatment study)may offer a narrower “useful”parameter range. While not necessarily optimal in the strictest sense,these parameter sets may serve as reasonable initial optimizationchoices in the future common use of the device. In this way, optimizingthe RNS System use is simplified, avoiding excessive trial and error and“effectiveness minima” resulting from myriad theoretical parametercombinations. Third, we propose the creation of a publicly available, lon-gitudinal RNS registry. Such a registry would contain data from the RNSpivotal and other clinical trials and also from less rigorous clinicalreports and single cases. Retrospective observations would includestimulation parameter sets sequentially uploaded during optimization,along with outcome data on seizure control and potential side effects.A potential model registry would be the NINDS supported InternationalEpilepsy Electrophysiology Portal (www.ieeg.org). Built along similarproposals [12], the registry would assist both researchers and clinicalepileptologists in summing up the role of the RNS System in refractoryfocal epilepsy and improve its use for the benefit of patients with mostsevere forms of epilepsy. The academic community must collaborate tounderstand the optimal parameters of RNS rather than repeating thesame history with VNS, where we still are not certain about optimalstimulation parameters or preimplantation predictors for greater effica-cy despite almost 17 years of clinical experience.

4. FDA regulatory framework: proposed changes

The measures proposed here subscribe to the larger call for tighten-ing of the FDA regulatory standards for therapeutic medical devices re-quiring a premarket approval application to complement a departurefrom the “black box” view of devices prevalent nowadays. We agreethat for neurostimulation devices, as for therapeutic drugs, detaileddata on mechanisms of action are often not needed for good use andmay be expensive to accrue. However, whenmechanistic presumptionsdirectly affect device operation, we believe that these should be ade-quately evaluated. In addition, a more stringent application to charac-terizing the effects of device parameter choices is needed. At present,the FDA regulatory process examines the evidence of safety and efficacyof the stimulation devices as a whole, without limiting the range of pa-rameters available to that for which such evidence is adequate. As longas some stimulation parameters are used in a regulatory trial, even ifonly once and briefly, they could be included in the marketed versionof the device. Therapeutic devices should be released with limitedrange of parameters for clinical use reflecting those parameters forwhich safety and efficacy data clearly exist (i.e., used by the majorityof patients within pivotal trials with positive efficacy and safety data).In conclusion, future FDA approval standards should require data on

surrogate neurophysiological markers and electrographic treatmentmetrics and characterization of responses in treatment groups stratifiedbased on narrower stimulation parameter ranges. The advent of the RNSSystem as a clinical tool in the treatment of severe focal epilepsy is oneof the most exciting new development since the VNS was approved in1997. Undertaking steps to improve what we know about how itworks will facilitate its due implementation in clinical practice.

Disclosure

Dr. Cavazos owns 4% of shares of Brain Sentinel (AKA: LGCH, Inc.), astart-up company developing a convulsion-alerting device based onEMG recordings.

References

[1] http://www.fda.gov/AdvisoryCommittees/CommitteesMeetingMaterials/MedicalDevices/MedicalDevicesAdvisoryCommittee/NeurologicalDevicesPanel/ucm340251.htm. [Accessed on April 9, 2013].

[2] Sun FT, Morrell MJ, Wharen Jr RE. Responsive cortical stimulation for the treatmentof epilepsy. Neurotherapeutics 2008;5(1):68–74.

[3] Morrell MJ, RNS System in Epilepsy Study Group. Responsive cortical stimulation forthe treatment of medically intractable partial epilepsy. Neurology 2011;77(13):1295–304.

[4] Fisher R, SalanovaV,Witt T,Worth R, Henry T, Gross R, et al. SANTE Study Group. Elec-trical stimulation of the anterior nucleus of thalamus for treatment of refractory epi-lepsy. Epilepsia 2010;51(5):899–908.

[5] Cook MJ, O'Brien TJ, Berkovic SF, Murphy M, Morokoff A, Fabinyi G, et al. Predictionof seizure likelihood with a long-term, implanted seizure advisory system in pa-tients with drug-resistant epilepsy: a first-in-man study. Lancet Neurol2013;12(6):563–71.

[6] Callaghan BC, Anand K, Hesdorffer D, Hauser WA, French JA, et al. Likelihood of sei-zure remission in an adult population with refractory epilepsy. Ann Neurol2007;62(4):382–9.

[7] Choi H, Heiman GA, Munger Clary H, Etienne M, Resor SR, Hauser WA, et al. Seizureremission in adults with long-standing intractable epilepsy: an extended follow-up.Epilepsy Res 2011;93(2–3):115–9.

[8] Sunderam S, Gluckman B, Reato D, Bikson M. Toward rational design of electricalstimulation strategies for epilepsy control. Epilepsy Behav 2010;17(1):6–22.

[9] Osorio I, Frei MG, Sunderam S, Giftakis J, Bhavaraju NC, Schaffner SF, et al. Automatedseizure abatement in humans using electrical stimulation. Ann Neurol 2005;57(2):258–68.

[10] Klatt J, Feldwisch-Drentrup H, Ihle M, Navarro V, Neufang M, Teixeira C, et al. TheEPILEPSIAE database: an extensive electroencephalography database of epilepsy pa-tients. Epilepsia 2012;53(9):1669–76.

[11] The Vagus Nerve Stimulation Study Group. A randomized controlled trial of chronicvagus nerve stimulation for treatment of medically intractable seizures. Neurology1995;45(2):224–30.

[12] Synofzik M, Fins JJ, Schlaepfer TE. A neuromodulation experience registry for deepbrain stimulation studies in psychiatric research: rationale and recommendationsfor implementation. Brain Stimul 2012;5(4):653–5.

Octavian V. LieJose E. Cavazos⁎

Department of Neurology,University of Texas Health Science Center at San Antonio,

San Antonio, TX, USASan Antonio VA Epilepsy Center of Excellence, San Antonio, TX, USA

⁎Corresponding author at: Department of Neurology,University of Texas Health Science Center at San Antonio,

San Antonio, TX, USA.E-mail address: [email protected] (J.E. Cavazos).

27 November 2013Available online xxxx