Designing next-generation subscalp devices for seizure monitoring: A systematic review and meta-analysis of established extracranial hardware

Implantable brain recording and stimulation devices apply to a broad spectrum of conditions, such as epilepsy, movement disorders and depression. For long-term monitoring and neuromodulation in epilepsy patients, future extracranial subscalp implants may offer a promising, less-invasive alternative to intracranial neurotechnologies. To inform the design and assess the safety profile of such next-generation devices, we estimated extracranial complication rates of deep brain stimulation (DBS), cranial peripheral nerve stimulation (PNS), responsive neurostimulation (RNS) and existing subscalp EEG devices (sqEEG), as proxy for future implants. Pubmed was searched systematically for DBS, PNS, RNS and sqEEG studies from 2000 to February 2024 (48 publications, 7329 patients). We identified seven categories of extracranial adverse events: infection, non-infectious cutaneous complications, lead migration, lead fracture, hardware malfunction, pain and hemato-seroma. We used cohort sizes, demographics and industry funding as metrics to assess risks of bias. An inverse variance heterogeneity model was used for pooled and subgroup meta-analysis. The pooled incidence of extracranial complications reached 14.0%, with infections (4.6%, CI 95% [3.2 – 6.2]), surgical site pain (3.2%, [0.6 – 6.4]) and lead migration (2.6%, [1.0 – 4.4]) as leading causes. Subgroup analysis showed a particularly high incidence of persisting pain following PNS (12.0%, [6.8 – 17.9]) and sqEEG (23.9%, [12.7 – 37.2]) implantation. High rates of lead migration (12.4%, [6.4 – 19.3]) were also identified in the PNS subgroup. Complication analysis of DBS, PNS, RNS and sqEEG studies provides a significant opportunity to optimize the safety profile of future implantable subscalp devices for chronic EEG monitoring. Developing such promising technologies must address the risks of infection, surgical site pain, lead migration and skin erosion. A thin and robust design, coupled to a lead-anchoring system, shall enhance the durability and utility of next-generation subscalp implants for long-term EEG monitoring and neuromodulation.


Introduction
Implantable devices able to record and/or stimulate the brain are poised to revolutionize neurological and psychiatric care for several chronic brain disorders.As deep brain stimulation (DBS) opened a new field of research and treatment for movement disorders in the late 1980's, implantable EEG monitoring devices may generate critical data and neuromodulation opportunities for epilepsy patients.Such devices may enable long-term exploration of epileptic networks in the patients' home-environment and would address a concrete need for continuous monitoring of epileptic brain activity.Inspired by evolving DBS technologies, epilepsy monitoring and treatment takes already advantage of the emerging 'closed-loop' paradigm in which stimulation is triggered upon sensing and detecting abnormalities in brain signals.This form of adaptive stimulation has shown superiority over traditional 'open-loop' neurostimulation (Leuthardt et al., 2021) with growing evidence in epilepsy (Nair et al., 2020), depression (Scangos et al., 2021) and Parkinson's disease (Meidahl et al., 2017;Torrecillos et al., 2018).Seizure forecasting models may also profit from ultra-long-term EEG data (Duun-Henriksen et al., 2020).The FDA-approved NeuroPace® RNS® System implemented for the first time a closed-loop paradigm for seizure recording and responsive neurostimulation (Jarosiewicz and Morrell, 2021).However, closed-loop systems for brain care currently require the placement of intracranial electrodes and therefore bear a limited, but non-negligible risk of intracranial complications.On the other hand, existing non-invasive recording (e.g.scalp EEG) and brain stimulation methods (e.g.transcranial magnetic stimulation) are poorly suited to longitudinally explore, monitor, and correct abnormal neural dynamics in a long-term closed-loop fashion.If the next generation of implantable devices shows a low-risk profile and reduced invasiveness, they will undoubtedly benefit a broader population of patients.
In this context, several academic-industrial partnerships are developing minimal-invasive devices that require short incisions and minor tissue penetration (Leuthardt et al., 2021) to monitor and/or stimulate the brain.One particularly attractive approach places electrodes below the scalp without skull penetration (i.e.subscalp or subgaleal, Fig. 1) for ultra-long-term ambulatory neuromonitoring (Duun-Henriksen et al., 2020;Haneef et al., 2022) and/or (extracranial) brain stimulation (Kravalis and Schulze-Bonhage, 2020) over months.The scalp comprises five superposed layers, of which the galea is detachable, creating an avascular space.Thus, granted dermal complications are minimized, the large surface between the galea and the periosteum seems exploitable for placing implants that do not penetrate the skull, likely suppressing the risk of intracranial complications.As ideas for minimally-invasive neurotechnologies flourish and large-scale initiatives arise (e.g.DARPA (DARPA, 2019) and BNCI Horizon 2020 (Brunner et al., 2014)), we propose here a needed risk analysis ahead of their deployment to scout for potential complications and confirm the expected benign nature of subscalp implants (Leuthardt et al., 2021).Indeed, in the field of neurotechnologies, minimal-invasiveness does not necessarily imply minimal risk (Leuthardt et al., 2021).
To inform the design, trial and use of future implantable subscalp devices for chronic EEG monitoring, we systematically reviewed past studies on DBS, cranial peripheral nerve stimulation (PNS), RNS and existing single rod subcutaneous EEG (sqEEG) devices.DBS and PNS are broadly used and established technologies that have moved beyond the experimental stage.The literature also offers growing information regarding the safety of the RNS system, and of emerging sqEEG devices that are still at an early commercial stage (Bullard et al., 2020;Weisdorf et al., 2019Weisdorf et al., , 2018)).As these devices require subcutaneous tunnelization of hardware and leave chronically implanted wires and connectors in the subgaleal space, they constitute a good proxy to next-generation subscalp devices.We estimated the pooled incidence of subscalp adverse events and scouted for extracranial complications specific to the design of DBS, PNS, RNS and existing sqEEG implants.As a comparator, we also included intracranial adverse events from studies that reported both.

Methods
The present systematic review and meta-analysis was designed in accordance to the PRISMA statement.

Eligibility criteria
Implantable DBS, cranial PNS (i.e.subcutaneous supraorbital and occipital nerve stimulators), RNS and sqEEG (i.e.UNEEG®) devices are included in this study.All share the need for intra-or extracranial leads to record and/or stimulate either brain regions, or peripheral nerves.They all require extracranial subscalp wires and either implantable pulse-generators (IPG, e.g.subclavicular), or recorders (intracalvarial or rarely subclavicular for RNS; subcutaneous for sqEEG).DBS also requires an electrode anchoring to the skull on top of burr-holes.For RNS, the recorder and IPG is placed as a replacement of the craniotomy bone flap.Given the cervical location of vagus nerve stimulation (VNS) systems and the short subscalp course of cochlear implants, these devices were not included in our analysis.As the development of subscalp devices will likely address an adult population in the first place, we did not include children-related literature in the present work.We excluded single case reports, case series of <10 participants (an exception was made for one sqEEG study reporting 9 participants), animal studies and articles in languages other than English.

Information sources and search strategy
Pubmed was last searched in February 2024 for reports published since 2000 with the following key terms: (complication OR adverse OR risk OR bleeding OR hematoma OR hemorrhage OR "tissue damage" OR discomfort OR infection OR "lead fracture" OR "lead migration" OR erosion OR ulceration OR fibrosis OR pain OR "nerve damage" OR irritation OR pressure) AND ("deep brain stimulation" OR "deep-brain stimulation" OR "responsive neurostimulation" OR "occipital nerve stimulation" OR "supraorbital nerve stimulation" OR "trigeminal nerve stimulation" OR "peripheral neurostimulation" OR "responsive Fig. 1.Subgaleal space for subscalp device insertion.The scalp is made of five layers.1) Skin with hair follicles, sebaceous and sweat glands.2) Subcutaneous layer providing blood supply to the scalp with a remarkable arterio-veno-lymphatic matrix.3) Galea aponeurotica, forming a ca.1-2 mm thick membranous non-elastic sheet, which invests the occipitofrontalis muscle and fuses laterally with the muscle temporalis fascia (Seery, 2002;Seline and Siegle, 2005).4) Subgaleal space.The galea is adherent to the overlying layers but detachable from the periosteum, forming a natural dissection plane, i.e. the avascular space of Merkel.5) Periosteum, adherent to the skull.S.L. Barlatey et al. neurostimulation" OR "subcutaneous electroencephalography" OR "subcutaneous EEG" OR intracranial electroencephalography OR electrocorticography) AND (complication OR "adverse event") NOT children

Selection process
Two reviewers (S.L.B., M.O.B) worked independently.After fullread, we excluded studies that reported less than three types of adverse events, as not all outcomes of interest were measured or reported in these cases, with several categories of extracranial complications regrouped as a single entity, preventing a meaningful quantification of complications by type.We excluded reports based on private database searches.Searching Embase, the Cochrane Library and Web of Science did not identify additional reports that may have been unavailable or overseen on Pubmed.

Data collection process and data items
Patient's characteristics were collected including age, sex and indication for device implantation.Average follow-up duration and number of subjects lost to follow-up were also compiled.We compiled devicerelated mortality, the number of necessary re-operations (urgent explantations, device replacements or revisions), as well as extracranial and intracranial adverse events related to the implantation procedure or the presence of hardware at any timepoint post-operatively, unequivocally categorized as described in Table 1 (Engel et al., 2018).Additionally for subgroup analyses, each study was categorized as belonging to the PNS, DBS, RNS or sqEEG categories.DBS and PNS studies were stratified by periods of publicationfrom year 2000 to 2010 and from 2011 to 2023.Funding sources were collected when available.

Study risk of bias assessment
We compared individual sizes of cohorts for PNS, DBS, RNS and sqEEG reports, as well as baseline demographic data and Industry funding.Lost-to-follow-ups and availability of all outcomes were assessed as a quality metric.

Effect measures
From each included publication, we extracted the number of included patients and numbers of each described complication (Table 2).The effect measure in the present meta-analysis is the incidence of a given post-surgical complication, calculated for each included study as the ratio between the number of occurrences and the number of included patients that received an implant.One included patient could have had more than one complication.Studies with missing data on an outcome of interest were ignored in the meta-analytic estimation of the effect size, i.e. we did not perform any missing data imputation.

Synthesis methods and certainty assessment
Estimates of complication rates across studies were obtained as a weighted average.For all estimates, statistical analysis was performed using MetaXL with an inverse variance heterogeneity model, as previously described by Doi et al. (Doi et al., 2015).The incidence rate and 95% confidence interval of each complication was calculated including extracted data from all identified studies, and selectively for each category: DBS, PNS, RNS, sqEEG, 2000's studies, 2010's studies.We compared statistical findings between device categories, as well as 2000's vs 2010's studies for significant discrepancies.

Results
We screened 3141 articles providing abstract and full text and identified 104 publications meeting our inclusion criteria.We excluded 57 publications for discerning less than three different complication entities, reporting less than ten participants (an exception was made for one sqEEG publication reporting nine participants) or mixing distinct adverse event categories into one entity.Out of the remaining 47 publications reviewed in detail (listed in Table 3), 36 were retrospective studies (5 multicentric), 2 were meta-analyses, 9 were prospective trials including 4 multicentric studies.29 studies concerned DBS, 10 PNS, 7 RNS and 2 sqEEG (Table 2, Fig. 2a).One comparative study (Yang et al., 2022) compiled data of distinct DBS and RNS cohorts, analyzed here twice independently.
An assessment of potential bias revealed that PNS and sqEEG studies tend to be smaller (median cohort size of respectively 40 and 23 participants) and prospective, compared to DBS and RNS studies that were retrospective and larger (median cohort size of respectively 130 and 126).Industry funding was more frequent for PNS, RNS and sqEEG.As a quality metric, lost-to-follow-ups were reported in 8/10 PNS, 5/7 RNS, 1/2 sqEEG, and 12/29 DBS studies.All outcomes were available for 10/ 10 PNS, 5/7 RNS, 2/2 sqEEG, and 13/29 DBS reports respectively.Studies were based on cohorts in northern and southern Europe, Canada, and the USA.DBS data also came from Japan and China.Fig. 3 In total, 7329 subjects were included, 5457 with DBS, 610 with PNS, 1215 with RNS and 47 with sqEEG.DBS was implanted for Parkinson's disease, dystonia, tremor (essential or secondary), pain disorders, and obsessive-compulsive disorder.PNS was implanted for chronic migraine, cluster headache, occipital neuralgia or cervicogenic headache, as well as other primary or secondary headaches.RNS and sqEEG were implanted respectively for palliative treatment and monitoring of medication-refractory epilepsy.Given these different patient populations, expected notable differences (Table 2) included a preponderance of females in PNS studies (74%, vs 40%, 50% and 57% for DBS, RNS and sqEEG respectively) and an older age of DBS participants (61, vs 44, 34 and 40 years for PNS, RNS and sqEEG respectively).The mean follow-up was only 2 months for sqEEG devices, compared to several years for the other systems.All publications reported on full-implantable devices, none included cases with an externalized IPG.Heterogeneity across studies was very high (Q=424.1,I 2 =88.9%) for the main outcome of interesti.e.extracranial complicationsmotivating the use of an inverse variance heterogeneity model for pooled estimates.Overall, statistical meta-analysis of the reviewed literature results in an average incidence of extracranial adverse events of 14.0% (CI 9.3 -19.0%) following DBS, PNS, RNS or sqEEG implantation, including mostly mild to moderate adverse events (Table 2), as detailed below.
To gain detailed insight into the extracranial complications that might be specific to non-skull-penetrating subscalp devices, we performed a stratified analysis by device type.The overall incidence of extra-cranial adverse events was significantly higher for PNS (42.1%, CI 21.1 -64.0%) and sqEEG (36.3%, CI 23.2 -50.6%) in comparison to DBS (11.8%,CI 7.1 -17.0%) and RNS (10.8%,CI 5.5 -16.8%).This discrepancy mainly relies on a dramatically higher incidence of reported pain at implantation site following PNS and sqEEG, as well as frequent lead migration following PNS, but does not concern adverse events such as infections, as described below in order of higher to lower incidence (Fig. 2b).

Evolution of safety over time
We compared incidence rates of each complication between two subgroups of DBS and PNS studies according to their publication decade 2000-2010 vs 2011-2023.The overall complication rate did not significantly change over time with an incidence of 16.4% (CI 9.2-24.4%)over 2000-2010 and 13.7% (CI 6.4 -22.0%) over 2011-2023.Of all adverse events, none showed significant decrease or increase over time.Devices show however a trend of increasing reliability over timestatistically non-significantwith hardware malfunction rates of 4.5% (CI 0.9 -9.1%) and 1.3% (CI 0.2 -2.8%) over the periods 2000-2010 and 2011-2023 respectively.Severe to life-threatening intracranial adverse events were relatively rare overall, and entirely absent among 657 patients implanted with PNS and sqEEG devices (Table 2).Following DBS or RNS implantation, 453 patients out of 6672 suffered an intracranial adverse event, regrouping all intracranial complications listed in Table 1 (incidence 6.2%, CI 4.3 -8.3%).Under the binomial distribution, the probability of finding zero complication among 657 patients with PNS or sqEEG, had the incidence been the same as in DBS and RNS is p=0.Thus, the incidence rate of intracranial complications in PNS and sqEEG is with statistical certainty under that of DBS and RNS, but the current total number of patients cannot exclude a probability below 0.5%.

Discussion
In this systematic review and meta-analysis of 7329 subjects across 48 studies, we found a 14.0% overall rate of extracranial complications associated with the implantation of DBS, PNS, RNS or sqEEG devices.The number of included patients was sufficient to obtain detailed poolestimates of extracranial complications and inform the design of future subscalp devices.We identified scalp infection (4.6%) and persistent surgical site pain (3.2%) as potentially serious complications.Other categories of adverse events may cause device dysfunctions, limit clinical usefulness, reduce implants' durability and trigger premature removal: lead migration, lead fracture, hardware malfunction, skin erosion, and subcutaneous hematoma or seroma, all under 3%.Surgical site infection remains an inherent risk to any invasive intervention, specifically with foreign material implantation.Once the surgical wound closed, skin itself acts as the main barrier against surgical site infection, except for rare hematogenous events.Noninfectious cutaneous complications occurred with a relatively low incidence, but an interplay between skin erosion and infection is expected, especially at thick connector sites.Surgical factors that appear to reduce the risk of wound dehiscence include incisions transverse to the course of cables and connectors, as well as flap incisions that circumvent hardware (Park et al., 2011).Consequently, subscalp monitoring and stimulation devices require a thin design to avoid skin atrophy, as well as a dedicated implantation system through short incisions to promote prompt wound healing.As proposed for shunt-catheters, antimicrobial-impregnation and coating may contribute to reduce the risk of hardware infections (Konstantelias et al., 2015).Subgroups DBS, PNS, RNS and sqEEG were analyzed independently to identify potential advantages or disadvantages in the design of their respective extracranial subscalp components.Sustained device-related pain or discomfort was significantly higher in the PNS (12.0%) and sqEEG (23.9%) subgroups.For PNS, this may be due to predispositions in a population primarily treated for refractory pain syndromes, reflecting a potential selection bias leading to an over-estimation of pain incidence.Mechanical stress of PNS leads over sensory nerves might also play a role.Anatomically, subscalp devices for long-term brain recording are not expected to apply pressure on major peripheral sensory structures -such as the occipital and supraorbital nerves.Of note, persisting pain data was well reported across the analyzed PNS (8/10 studies) and sqEEG (2/2) publications, suggesting a potential reporting bias and under-estimation in DBS (9/29) and RNS (1/7) studies that focused on more severe intra-and extra-cranial adverse events.Even though the mean follow-up period for sqEEG devices in only 47 participants was low (2 months) and post-implantation pain may resolve further on, a rate of 23.9% persisting pain/discomfort following sqEEG implantation is concerning and requires attention.
From the analyzed data, minimizing lead migration appears as a major challenge for PNS electrodes (12.4% lead migration) that lack lead fixation systems.Supraorbital and occipital nerve stimulation electrodes are inserted in anatomical areas of intrinsic mobility, increasing the risk of lead migration, which may be lower for electrodes laying on the convexity.None of the two sqEEG studies presented data about lead migration.For epilepsy monitoring, the localization of brain signals plays a crucial role.Non-identified displacement of such device may fail to record epileptic activity, or wrongly attribute epileptic potentials to normal tissue.Three possible strategies may address the risk of lead migration: 1) accept electrode displacement as a property of the device, mitigated by means to track lead positions over time, 2) design an electrode fixation system that does not develop mechanical constraints susceptible to increase rates of lead fracture or 3) develop an insertion technique that promotes natural anchorage and minimizes the risk of displacement.Burr-hole-coverssuch as Stimloc, SureTek or Neuropaceappear to secure safely DBS and RNS electrodes and may inspire similar systems for PNS, sqEEG and future subscalp monitoring devices.
To our knowledge, ethical thresholds defining acceptable complication rates are not established in the current neuro-ethics literature (Cheung, 2009;Clausen et al., 2017;Leuthardt et al., 2021).Any procedure requiring foreign material implantation may lead to adverse events, with an expected benefit that needs to overweigh the likelihood of severe complications.In the absence of intracranial complications for PNS and sqEEG devices, we found evidence for safety (i.e.lack of unacceptable risk) of foreign material implantation beneath the scalp, but also an opportunity to reduce rates of extra-cranial complications.Additionally, neurological information gain (Leuthardt et al., 2021)  expected from subscalp EEG recordings is very high, as the signals are similar to those obtained with scalp EEG (Weisdorf et al., 2018), but available anytime and anywhere.Taken together, this data suggests the "high-need, limited-risk" profile of subscalp monitoring and/or stimulating devices (Fig. 4).
Our meta-analysis has limitations.Case reports, small case series and publications in languages other than English were excluded and may have provided relevant data.Overall heterogeneity was high, and the different study designs, sizes, and populations with different underlying disorders limit comparability between extracranial complications from DBS, PNS, RNS and sqEEG.DBS and RNS studies present larger cohorts than PNS and sqEEG reports, with more frequent industry-funding for PNS, RNS and sqEEG publications.These two factors may introduce a selection bias in our estimation of complication rates.Also, the heterogeneity of occurrences per complication category between device types prevents generalization of a single safety profile to all kinds of subscalp devices.Complication data was not available for all four devices, with information lacking about lead migration for RNS and sqEEG, and about lead damage for sqEEG.Nevertheless, most subscalp complications are in our opinion independent enough from these factors to draw a comparison between the designs of extracranial components of these implants.Future larger cohorts may confirm trends such as higher rates of infection for PNS and RNS compared to DBS.Reflecting the lack of data, our risk assessment focused on adverse events following a first subscalp implantation but did not report on sometimes necessary (durability risk) Fig. 3. Meta-analysis of extracranial adverse events.Forrest plot of the total incidence of adverse events for each included study.Blue: DBS studies, green: PNS studies, red: RNS studies, grey: sqEEG studies.Point size refers to the total number of patients included in each publication.Diamond-shaped pooled incidence across studies, CI 95%.Fig. 4. Risk-need-information profile.According to the classification proposed by Leuthardt et al. subscalp devices are in high need, show a limited-risk, and yield valuable information to assess physiological and pathological states of the brain.The information is not at the maximum, because as compared to intracranial devices, the spatial resolution is lower (but the spatial coverage is higher).or desired material explantation, which may come with different risks.Neither the performance of recordings, nor the efficacy and safety of brain stimulation were assessed, as we focused on the surgical and hardware aspects of subscalp devices.As future receptors/transmitters of subscalp devices will likely be cranially implantable, we did not analyze subclavicular complications associated to pulse-generators (IPG).The lower rate of skin erosion following RNS implantation (1.4%) compared to PNS (6.4%) and DBS (2.3%) suggests that imbedding an IPG/receptor within the skull and suppressing the tunnelization of hardware to the subclavicular region may render future implants safer.Other minimally-invasive approaches such as intravascular (Oxley et al., 2016) and intracalvarial (Gribetz, 2019) device placement were not covered here and may show different risk profiles, including low, but non-zero catastrophic risks (Leuthardt et al., 2021).
In our opinion, upcoming subscalp devices show potential for chronic EEG monitoring.As they do not penetrate the skull and require a similar implantation technique as PNS and current sqEEG hardware (i.e.subgaleal tunnelization of electrodes and wires connected to a receptor), we expect these subscalp devices to show a similar safety profile as PNS and existing single rod sqEEG.Due to specific risk factorse.g. head trauma with skin erosion and/or infectious complicationsin patients with epilepsy, dedicated studies about the safety of subscalp devices for seizure monitoring are required.Nevertheless, addressing high rates of lead migration, hardware-associated pain, infections, and cutaneous complications needs crucial attention.In contrast, intracranial events or device-related death are very unlikely.From the identified adverse events, we propose the following must-haves for these systems: 1) a thin design reducing the risk of pressure-induced skin erosion and associated infection, 2) an anchoring system that prevents subgaleal lead migration, 3) the use of stretchable materials robust to anatomical mechanical constraints, preventing lead fracture.An exploration of potential reasons for persisting pain or discomfort following the implantation of existing PNS and sqEEG devices is needed.Future comparative studies will then be required to assess comprehensively the real-world performance and long-term risk-benefit ratio of implantable subscalp EEG devices compared to non-invasive monitoring technologies.

Conclusion
In the present review, we collected and synthesized existing data about the risk of complications related to subscalp material, to inform the development and deployment of future recording technologies in this anatomical space.Infections, persisting post-implantation pain, non-infectious skin complications and lead migration appear as the main adverse events to minimize.As expected, strictly extracranial devices did not cause intracranial complications.Inspired by current existing hardware, an opportunity exists to improve the safety of future implantable subscalp devices for long-term epilepsy monitoring.Although the risk of minor complications discussed above is currently high, it may be mitigated with improved designs, materials, coatings, and insertion techniques.Their success will rely on their performance in providing exploitable electrophysiological data.

Fig. 2 .
Fig. 2. Incidence of extracranial adverse events.(a) Flow chart of articles' identification, selection and inclusion.(b) Detailed incidences of adverse events for DBS, PNS, RNS and sqEEG, excluding intracranial complications.Note a significantly higher rate of lead migration and cutaneous complications (skin erosion) for PNS, as well as persistent pain for both PNS and sqEEG.

Table 1
Categories of adverse events.
high electrode impedance, electrode malfunction, premature battery depletion, abnormal current spread along extension wires, as well as short or open circuits; Pain Persisting pain or discomfort S.L. Barlatey et al.

Table 2
Statistics.Characteristics of included reports, demographic data, and incidence rates of adverse events by order of severity and incidence.

Table 3
Reports included to the meta-analysis.