Sevoflurane and Parkinson’s Disease: Subthalamic Nucleus Neuronal Activity and Clinical Outcome of Deep Brain Stimulation

Sevoflurane is a widely used volatile anesthetic with a complex mechanism of action including positive allosteric modulation of γ-aminobutyric acid receptors and two-pore-domain K⁺ channels.¹  Although the effects of sevoflurane anesthesia on electroencephalographic (EEG) activity have been described, it is unclear if sevoflurane modulates single neuron activity in human patients.² 

Subthalamic nucleus deep brain stimulation is an effective treatment for patients with Parkinson’s disease with motor fluctuations, but clinical outcomes are critically dependent on accurate placement of the stimulating electrode.³  Microelectrode recordings in awake patients under local anesthesia allow refinement of targeting within the sensorimotor subthalamic nucleus. Further, such recordings have also revealed critical aspects of Parkinson’s disease pathophysiology, including pathologic beta band (13 to 30 Hz) oscillations in neuronal firing and local field potentials that correlate with disease severity, response to medical therapy, and deep brain stimulation efficacy.^4–6  However, patients with severe tremor, painful off-state dystonia, or high levels of anxiety may not tolerate awake stereotactic surgery.⁷ 

For such patients, recordings of subthalamic neurons under propofol or dexmedetomidine general anesthesia may be a feasible alternative.⁸  Indeed, several studies have also reported comparable clinical outcomes for patients with Parkinson’s disease receiving subthalamic nucleus–deep brain stimulation electrode placement surgery using different approaches, such as real-time image guidance and microelectrode recording, under local anesthesia or general anesthesia.^9–12  However, microelectrode recording under general anesthesia presents a unique challenge. Propofol and dexmedetomidine may degrade the quality of microelectrode recordings.^13,14  While the molecular mechanism of sevoflurane and propofol may overlap, EEG analysis revealed a distinct higher theta band (4 to 7 Hz) power under sevoflurane compared to propofol.²  Therefore, it is critical to characterize the effects of different general anesthetics on single neuron spiking properties as well as on population activity as these properties are essential for guidance of deep brain stimulation electrode placement.

The major objective of this study was to compare the electrophysiologic properties of subthalamic neurons in patients with Parkinson’s disease during deep brain stimulation surgery under either sevoflurane general anesthesia or local anesthesia. Based on previous EEG findings, we speculated that sevoflurane would shift the power spectrum to lower frequency oscillations spanning delta, theta, and alpha bands. However, it is unclear whether sevoflurane can also influence the bursting features of subthalamic neurons used for guidance of electrode placement. If not, dorsal and ventral subthalamic borders should be detected under sevoflurane general anesthesia, allowing for electrode placement accuracy comparable to that achieved under local anesthesia. Therefore, in addition to comparing electrophysiologic properties, we also compared electrode placement coordinates, stimulation parameters, and long-term clinical efficacy between patients with Parkinson’s disease receiving subthalamic nucleus electrode placement surgery under sevoflurane general anesthesia or local anesthesia.

The study was approved by the institutional review board of Tzu Chi General Hospital (Hualien, Taiwan), and written informed consent was collected from all patients. From March 2006 to February 2009, 19 patients with Parkinson’s disease with comparable disease severity underwent subthalamic nucleus deep brain stimulation at Tzu Chi General Hospital. These patients were in our study to prospectively compare long-term clinical outcomes between patients with Parkinson’s disease receiving general anesthesia or local anesthesia for deep brain stimulation.¹⁵  Patients were recruited based on our previous experience with analyzing microelectrode recording from patients with Parkinson’s disease.¹⁶  No formal statistical power calculation was conducted to guide sample size. All participants fulfilled surgical criteria of subthalamic nucleus deep brain stimulation. The benefits and risks of local anesthesia and general anesthesia were explained to each patient, and the choice of anesthesia was made on clinical grounds and patient preference. The study was nonrandomized, and each patient was evaluated blindly using the Unified Parkinson’s Disease Rating Scale in four different conditions: on and off medication, and with and without deep brain stimulation. There were no missing data. We evaluated cognitive status with the Cognitive Abilities Screening Instrument and Mini-Mental Status Examination. The Beck Depression Inventory was used to assess patients’ depressive mood. The dosage of dopamine replacement medication was expressed as levodopa equivalent daily dose.¹⁷ 

The preoperative magnetic resonance imaging images were obtained with a 1.5-tesla unit (SIGNA HDXT 1.5T, General Electric, USA). The protocol consisted of T1-weighted axial pre- and postcontrast images (1 × 1 × 0.7 mm voxels) and T2-weighted axial images (0.5 to 1.0 × 0.5 to 1.0 × 2 mm). Stereotactic planning was conducted using a neuronavigation workstation (BrainLAB, Germany). The surgical target was selected to place the tip of the permanent implantable electrode at the ventral border of the posterior subthalamic nucleus. A Leksell G-frame (Elekta Instrument Inc., USA) was used for the stereotactic procedure. The frame was placed under local anesthesia for both groups. Patients then underwent preoperative computed tomography with frame-mounted stereotactic fiducials in place for fusing stereotactic images to the preoperative magnetic resonance imaging.

All general anesthesia group patients received sevoflurane general anesthetics with endotracheal intubation. Anesthesia was initiated by administration of fentanyl (1 to 2 μg/kg), propofol (1 to 2.5 mg/kg), and a muscle relaxant (rocuronium at 0.6 to 1.5 mg/kg or cisatracurium at 0.15 to 0.2 mg/kg). Immediately after the patient’s loss of consciousness, the anesthesiologist stopped propofol infusion and started sevoflurane anesthesia to keep the patient unconscious. Sevoflurane inhalation was maintained for the entire recording period. To prevent inadvertent patients’ responses due to inadequate anesthesia depth, minimum alveolar concentration was maintained around 1.0 to 1.2 before electrophysiologic recordings, gradually decreased over 30 min, and maintained at 0.5 to 0.6 during recording. End-tidal anesthetic concentration was around 0.7 to 0.8 at a minimum alveolar concentration of 0.5 to 0.6.¹⁸ 

Intravenous anesthetics and analgesics demonstrated to dampen subthalamic neuronal firing were avoided for deep brain stimulation surgery.^8,14,19  Patients were closely monitored for heart rate and blood pressure changes during the microelectrode recording procedure.²⁰  If the blood pressure increased beyond 140 mmHg, the anesthesiologists administered intravenous nicardipine or labetalol to control it. Passive movements of the contralateral limb were tested and monitored in both anesthesia groups during the microelectrode recording in the subthalamic nucleus to identify movement-related neuronal firing changes (kinematic responses). We did not encounter any intraoperative awareness. The selection of the final trajectory for electrode implantation depended on the length of the subthalamic neuronal firings and the presence of passive movement-related activity of the subthalamic neurons. Macrostimulation tests were not performed for the patients in the general anesthesia group since there were no parkinsonian symptoms detected during anesthesia.

The microelectrode was 10 to 40 μm in diameter and 200 mm in length, with a less than 50 μm tungsten tip (FHC, Inc., USA) and recording impedance between 0.5 and 1 MΩ. The microelectrode signal was recorded using an intraoperative microelectrode recording system (LeadPoint, Medtronic, USA). The raw signals were amplified (×10) and band-pass filtered (300 Hz to 3 kHz). Recording started 10 mm above the planned target coordinates, and the microelectrode was advanced in steps of 200 to 500 µm with pauses at sites of robust neuronal firing. The discharge from each depth was recorded for at least 10 s.

A typical microelectrode trajectory passes through the thalamus, zona incerta, subthalamic nucleus, and substantia nigra. For both groups, passive movement of the contralateral limbs was tested during microelectrode recording in the subthalamic nucleus to assess for kinematic responses.²⁰  The final trajectory was selected for electrode implantation based on the length of the subthalamic nucleus recording and the presence of kinematic responses. In the local anesthesia group, macrostimulation testing was performed after the deep brain stimulation electrode (Medtronic 3389) was inserted to test for adverse effects and clinical effectiveness. Postoperative brain computed tomography scanning was performed in both groups to check for the presence of intracranial hemorrhage and determine postoperative electrode coordinates. The implantable pulse generator was implanted during a second operation typically 1 week after the initial operation. Detailed clinical evaluation, targeting, surgical procedures, and recording methods are described in our previous publication.²¹ 

Raw spike recordings from the implanted trajectories were analyzed using custom-written scripts in MATLAB R2014b (Mathworks, USA). Raw recordings were visually inspected by a neurophysiologist blind to patient anesthesia group to exclude artifacts (e.g., mechanical artifacts or spike bursts from neurons damaged by the recording microelectrode). Background spiking activity was estimated using the normalized root mean square of each subthalamic nucleus recording.⁶  The root mean square value of each recording was also normalized by the root mean square value of the baseline presubthalamic nucleus recording of each tract (excluding thalamus, zona incerta, and other recordings with prominent spikes) to control for factors such as electrode impedance and electrical noise that vary between subjects.

To estimate the oscillatory entrainment of the population spiking activity, we calculated the power spectral density of the raw recorded subthalamic nucleus population spike firing as previously described.²²  The time series was full-wave rectified and the mean subtracted.^22,23  The power spectral density of the resulting time series was analyzed by the Welch method with a 1-s Hamming window (50% overlap) and zero-padding, yielding a spectral resolution of 1/3 Hz. The power spectral density at each recording depth was normalized by the total power between 0.3 to 100 Hz (excluding ±2 Hz of the 60 Hz line artefacts). We then determined the raw and normalized spectral power from delta (0.1 to 4 Hz), theta (4 to 8 Hz), alpha (8 to 13 Hz), beta (13 to 30 Hz), and gamma (30 to 100 Hz) band ranges in each patient. Each recording’s depth was converted to the relative depth to average across patients. Zero percent was designated as the dorsolateral border of subthalamic nucleus and 100% as ventromedial exit (presubthalamic nucleus represented as less than 0% and postsubthalamic nucleus as greater than 100%). For analysis of topographical distribution of subthalamic nucleus spike features, the subthalamic nucleus was arbitrarily divided into the dorsolateral (0 to 50%) and ventromedial (50 to 100%) components.

Single-neuron spike detection was performed using the voltage threshold method. Putative single neurons were identified offline using principal component analysis and the presence of a central trough in the autocorrelogram (Offline Sorter, USA). Isolation of the spike train was graded by evaluating the fraction of spikes within the refractory period of 1.5 ms out of the total number of spikes in the spike train, and only spike trains with a fraction of less than 1% were processed. Spikes degraded by obvious cardiobalistic or other artefacts were excluded.

The Wilcoxon rank-sum test was used for comparing clinical scores, deep brain stimulation lead coordinates, stimulation parameters, and spike features across relative depths in the two groups. Significance was set at P < 0.05 (two-tailed) for all tests, corrected as needed for multiple comparisons, and two-tailed testing was used. We evaluated all variables for possible outliers and distribution abnormalities. The Kruskal–Wallis test was used to assess groupwise differences in band power among the four combinations of dorsal versus ventral and local anesthesia versus general anesthesia. Pairwise comparisons were assessed using the Wilcoxon rank-sum test with correction for multiple comparisons (five bands and two groups, significance threshold of P < 0.005).

Ten patients (four women/six men, mean age 47 ± 8 yr) underwent subthalamic nucleus deep brain stimulation electrode placement under sevoflurane general anesthesia with endotracheal intubation and nine patients (two women/seven men, mean age 45 ± 11 yr) while awake using local anesthesia. The demographic characteristics of the local anesthesia and general anesthesia patients were comparable. We conducted follow-up visits with patients at 5 yr after subthalamic nucleus deep brain stimulation (general anesthesia group: 62 ± 8 months; local anesthesia group: 60 ± 13 months). The mean duration from disease onset to deep brain stimulation surgery for the general anesthesia group was 11 ± 5 yr (mean age at surgery was 57 ± 6 yr) and for the local anesthesia group was 12 ± 9 yr (mean age at surgery was 57 ± 5 yr). Neither motor disability as assessed by Unified Parkinson’s Disease Rating Scale scores nor responsiveness to dopamine replacement therapy (improvement in percentage) differed significantly between groups (table 1). All patients in both groups used monopolar stimulation. Further, anatomic coordinates of the surgical target and active stimulation contacts did not differ between groups, nor did the number of microelectrode recording tracts used for mapping (table 2).

Implants under general anesthesia and local anesthesia both utilized intraoperative microelectrode recording. Several previous studies have shown that propofol and dexmedetomidine can alter microelectrode recording features, raising doubts about the reliability of electrophysiologic features for target identification across anesthetic states. We analyzed intraoperative recordings with spontaneous activity of subthalamic nucleus units from both groups (19 microelectrode recording tracts with 299 recorded units under general anesthesia and 16 tracts with 224 units under local anesthesia). The mean normalized root mean square value, a metric of multiunit gray-matter spike activity, did not differ significantly between groups (general anesthesia, 2.3 ± 0.8; local anesthesia, 2.2 ± 0.9, P = 0.39; fig. 1A).

We next examined the oscillatory features of the spike trains. Sevoflurane general anesthesia shifted subthalamic nucleus spike oscillations toward lower frequencies. Under general anesthesia, the spectral power across delta, theta, and alpha bands was each significantly greater than in the local anesthesia group (general vs. local anesthesia, mean ± SD; delta: 13 ± 7.3% vs. 7.8 ± 4.8%; theta: 8.4 ± 4.1% vs. 3.9 ± 1.6%; alpha: 8.1 ± 4.1% vs. 4.8 ± 1.5%; all P < 0.001). In contrast, higher frequency beta and gamma band oscillations were more prominent in the local anesthesia compared to general anesthesia group (general vs. local anesthesia, mean ± SD; beta: 12.7 ± 5% vs. 18 ± 6.1%; P < 0.001; gamma: 53.6 ± 11.3% vs. 62.5 ± 9.7%; P < 0.001). In the local anesthesia group, we observed the characteristic gradient of decreasing beta-band power along the recording trajectory from dorsolateral (sensorimotor) to ventromedial (associative-limbic) subthalamic nucleus (mean ± SD; dorsal vs. ventral percent of beta band power: 20.5 ± 6.6% vs. 15.4 ± 4.3%; P < 0.001). This topographic pattern of beta oscillations was not found in the recordings under general anesthesia (fig. 1B, fig. 2). Alternatively, neuronal spiking characteristics were similar between anesthetic states, with no significant group differences in mean spike firing rate (general vs. local, 32 ± 4 Hz vs. 41 ± 5 Hz, P = 0.17), peak burst firing rate (56 ± 3 Hz vs. 60 ± 5 Hz, P = 0.5), and burst index²⁴  (4.5 ± 0.3 vs. 4.6 ± 1.2, P = 0.87). The coefficient of variation of the interspike interval was relatively higher in the general anesthesia group, but the difference did not reach significance (1.6 ± 0.1 vs. 1.4 ± 0.1, P = 0.056, fig. 3).

At 5 yr after subthalamic nucleus deep brain stimulation surgery, general anesthesia and local anesthesia groups also showed comparable improvements in Unified Parkinson’s Disease Rating Scale scores relative to postoperative medication off/deep brain stimulation off and preoperative medication off conditions (Supplemental Digital Content, tables 1 https://links.lww.com/ALN/C246, and 2 https://links.lww.com/ALN/C247). The fractional improvement in off-levodopa Unified Parkinson’s Disease Rating Scale part III (motor rating) with deep brain stimulation on did not differ between general and local anesthesia groups (45 ± 17% vs. 38 ± 16%, P = 0.37). The reductions in levodopa equivalent daily dose (P = 0.41) and motor complications (Unified Parkinson’s Disease Rating Scale part IV, P = 0.57) were also comparable, with no significant differences between groups in subthalamic nucleus deep brain stimulation efficacy according to fractional changes. The disease progressed at a similar rate in both groups (Supplemental Digital Content, table 3, https://links.lww.com/ALN/C248), as evidenced by score changes from preoperative medication off to the postoperative medication off/deep brain stimulation off conditions. Similarly, neuropsychological evaluation results, including Mini-Mental Status Examination, Cognitive Abilities Screening Instrument, and Beck Depression Inventory scores, were comparable between groups at 5 yr (Supplemental Digital Content, table 4, https://links.lww.com/ALN/C249). Furthermore, adverse events frequencies were comparable across groups. One patient implanted under general anesthesia developed an intracerebral hematoma and presented with mild hemiparesis. Another patient implanted under general anesthesia developed infection at the electrodes and recovered fully after explantation and antibiotic treatment. He underwent successful reimplantation. Other stimulation-related adverse events included hypophonia (general anesthesia, two; local anesthesia, one), dysarthria (general anesthesia, two), dyskinesia (local anesthesia, two), and weight gain (general anesthesia, three; local anesthesia, one).

Stimulation parameters were comparable between groups at 5 yr follow-up. The mean stimulation parameters in the general anesthesia group were (left vs. right subthalamic nucleus) 3.8 ± 0.4 V and 3.5 ± 0.9 V, 60 ± 0 and 60 ± 0 microseconds, and frequency 123 ± 16 Hz and 123 ± 16 Hz. In the local anesthesia group, parameters were voltages 3.2 ± 0.6 V and 3.5 ± 0.3 V, 64 ± 11 and 64 ± 11 microseconds, and 128 ± 5 and 128 ± 5 Hz. There were no significant differences between groups (stimulation amplitude: P = 0.15, pulse width: P = 0.35, frequency: P = 0.26).

Our study demonstrates that sevoflurane general anesthesia has no clinically significant impact on the spike bursting characteristics of subthalamic neurons recorded during deep brain stimulation electrode placement surgery for Parkinson’s disease compared to local anesthesia. While sevoflurane did induce a shift in the power band spectrum of subthalamic neuron spiking toward subbeta bands (delta, theta, and alpha) compared to local anesthesia, these changes did not influence stimulation electrode placement accuracy, deep brain stimulation parameters, or the clinical efficacy of deep brain stimulation for reducing motor symptoms at 5 yr postsurgery.

For microelectrode recording to be useful for electrode targeting under general anesthesia, one must be able to reliably identify the characteristic electrophysiologic features and transitions into and out of the target structure (such as the subthalamic nucleus for deep brain stimulation treatment of Parkinson’s disease). We report that the fundamental subthalamic nucleus firing properties and gray–white matter transitions can be readily identified under sevoflurane anesthesia as precisely as under local anesthesia. While subthalamic nucleus mapping under inhalational anesthetics has been reported previously, there was no detailed electrophysiologic characterization to ensure preservation of features required for accurate electrode placement.^9,16,20  Further, no single anesthetic regimen has emerged in the literature as optimal for preserving the critical electrophysiologic features required for electrode target identification. For instance, the α-2-adrenergic receptor agonist dexmedetomidine has been reported to decrease bursting activity and variably alter subthalamic nucleus spike rates even at low doses.^13,25,26  Alternatively, propofol has been reported to qualitatively preserve subthalamic nucleus discharge patterns,²⁷  but several studies have reported that higher dosage reduces the discriminability of spike features in the subthalamic nucleus and globus pallidus.^{14,19,24,28–30}  Remifentanyl and ketamine anesthesia may also preserve the electrophysiologic properties of the subthalamic nucleus, but concerns about increased intracranial pressure and postanesthetic neuropsychiatric effects have limited use of ketamine.³¹  Thus, sevoflurane may be an alternative choice for subthalamic nucleus deep brain stimulation patients intolerant of local anesthesia.

Although the fundamental spiking properties of the subthalamic neurons were similar under sevoflurane and local anesthesia, sevoflurane induced a shift in the oscillatory entrainment toward subbeta bands. Under general anesthesia, subthalamic nucleus spiking demonstrated increased power in the delta, theta, and alpha (less than 12 Hz) range, and decreased power in the beta (13 to 30 Hz) range. Despite distinct molecular targets, sevoflurane was shown to induce unconsciousness in a manner similar to propofol and ketamine by interfering with coherent oscillations between cortical layers of frontal and parietal lobes.^31,32  These effects on the group oscillations of subthalamic neurons further confirm that sevoflurane has a similar influence on basal ganglia–related oscillatory dynamics and anesthesia as propofol. In contrast, Velly et al. suggested that sevoflurane and propofol produce unconsciousness and analgesia through distinct effects on cortical and subcortical structures.³³  A recent study using desflurane during microelectrode recording of deep brain stimulation surgery showed enhanced power over theta band range (4 to 8 Hz) oscillation only, which further indicates that different volatile anesthetics work through different mechanisms on drug-induced unconsciousness and analgesia.^32,34 

Under general anesthesia, we did not observe increased beta entrainment of the dorsolateral subthalamic nucleus compared to the ventromedial subthalamic nucleus. The inability to detect this electrophysiologic signature of sensorimotor subthalamic nucleus could potentially interfere with electrode placement accuracy for some applications. Beta band oscillations and beta-gamma phase-amplitude coupling have been proposed as biomarkers of Parkinson’s disease neuropathology and have been demonstrated to correlate with severity of Parkinson’s disease motor disability, response to levodopa, and response to deep brain stimulation.^35–39  In addition, the spatial extent of beta oscillations at the implanted deep brain stimulation site correlated with its efficacy.⁶  Detection of such oscillations intraoperatively may be useful to refine optimal targeting and guide deep brain stimulation programming. Also, adaptive deep brain stimulation strategies triggered by beta-band oscillations may offer additional advantages including reduced battery consumption and side effects.⁴⁰  In this case, implantation at a site with an optimal control signal would be important. Thus, while we found no differences in clinical efficacy at the 5-yr follow-up, we cannot exclude potential disadvantages of sevoflurane anesthesia for some deep brain stimulation applications.

Optimal clinical outcomes from subthalamic nucleus deep brain stimulation, including maximal motor benefit and minimal stimulation-related sided effects, depend on accurate placement of electrodes. Despite 30 yr of clinical experience, the optimal anatomic target and the best surgical approach to reach it are still debated.⁴¹  Awake implantation facilitates microelectrode mapping and intraoperative assessment of the clinical response to stimulation. However, awake surgery is less comfortable for patients, and may be contraindicated for patients with painful dystonia, large-amplitude tremors, or severe anxiety. As advances in magnetic resonance imaging have improved direct targeting and real-time intraoperative imaging has become available, the relative benefits of awake surgery and microelectrode recording have been questioned.

In our study, surgical procedures were standardized between patients, save for anesthetic modality. Both anesthesia groups achieved approximately 50% improvement in off-levodopa Unified Parkinson’s Disease Rating Scale part III score with stimulation at 5 yr postsurgery, comparable to long-term outcomes reported in other studies of deep brain stimulation under local anesthesia.^42,43  Thus, our results are consistent with the noninferiority of microelectrode recording-guided deep brain stimulation lead implantation under general anesthesia. Similarly, Fluchere et al. reported comparable 1- and 5-yr subthalamic nucleus deep brain stimulation outcomes for a large cohort of patients with Parkinson’s disease implanted under general anesthesia with microelectrode recording.⁹  The current results are also in line with several smaller studies and a meta-analysis reporting clinical outcomes at short- to midterm follow-up of deep brain stimulation under general anesthesia for Parkinson’s disease.^10,11,27  Complication rates and hospital length of stay have also been reported to be comparable between anesthetic modalities.⁴⁴  However, these studies did not describe electrophysiologic changes to subthalamic neurons compared to local anesthesia. Given the importance of neuronal firing properties for defining the optimal stimulation target, the current findings provide a plausible explanation for the comparable clinical success of subthalamic nucleus deep brain stimulation under general anesthesia.

Our study has several limitations. First, patients were not randomized to the sevoflurane anesthesia and local anesthesia groups. All were included according to strict surgical criteria. Second, there were differences in some motor domains at baseline between groups, such as levodopa response, which may have influenced electrophysiologic features or clinical outcome.⁴  For instance, it has been suggested that responsiveness to dopaminergic treatment correlates with deep brain stimulation efficacy.^45,46  Third, we did not assess differences in electrophysiologic properties according to the depth of sevoflurane anesthesia, and all surgical and recording procedures were standardized. Fourth, this relatively small cohort limits the statistical power to discern small differences in electrophysiologic responses and long-term clinical outcomes between anesthetic groups. Finally, as we mentioned, we were unable to perform trial stimulation tests to evaluate the initial effectiveness and possible side effects in the general anesthesia group. However, due to precise localization of subthalamic nucleus targets and similar postoperative active contacts coordinates, patients in both groups achieved comparable benefits and exhibited similar side effect profiles at the long-term follow-up.

In conclusion, we report that the basic firing properties of subthalamic neurons are preserved using sevoflurane anesthesia for deep brain stimulation electrode placement surgery. Alternatively, the characteristic beta oscillations in the dorsolateral (sensorimotor) subthalamic nucleus are obscured by sevoflurane-induced entrainment of delta, theta, and alpha oscillations. Nonetheless, this does not impair electrode placement accuracy or affect long-term clinical outcome. This study supports the application of electrophysiology-based approaches for deep brain stimulation implantation under inhalational anesthesia. Whether this combined approach is widely applicable will depend on future deep brain stimulation platforms employing real-time electrophysiologic data to guide therapy.

The authors thank Todd Michael Herrington, M.D., Ph.D., Department of Neurology, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts, for helpful discussion and English grammar editing.

This work was funded by the Buddhist Tzu Chi Medical Foundation (TCMF-MP 108-01-03), Ministry of Science and Technology, Taiwan (107-2314-B-303-002), Tzu-Chi University, Hualien, Taiwan (TCIRP 104001-05), and Hualien Tzu-Chi General Hospital, Hualien, Taiwan (TCRD106-32).

The authors declare no competing interests.