Correlation of antero-dorsal active contact location with weight gain after subthalamic nucleus deep brain stimulation: a case series

Katsuki Eguchi (  k198762@gmail.com ) Department of Neurology, Faculty of Medicine and Graduate School of Medicine, Hokkaido University Shinichi Shirai Department of Neurology, Faculty of Medicine and Graduate School of Medicine, Hokkaido University Masaaki Matsushima Department of Neurology, Faculty of Medicine and Graduate School of Medicine, Hokkaido University Takahiro Kano Department of Neurology, Faculty of Medicine and Graduate School of Medicine, Hokkaido University Kazuyoshi Yamazaki Department of Neurosurgery, Faculty of Medicine and Graduate School of Medicine, Hokkaido University Shuji Hamauchi Department of Neurosurgery, Faculty of Medicine and Graduate School of Medicine, Hokkaido University Toru Sasamori Department of Neurosurgery, Sapporo Azabu Neurosurgical Hospital Kenji Hirata Department of Diagnostic Imaging, Faculty of Medicine and Graduate School of Medicine Mayumi KItagawa Sapporo Teishinkai Hospital Mika Otsuki Faculty of Health Sciences and Graduate School of Health Sciences, Hokkaido University Tohru Shiga Department of Nuclear Medicine, Faculty of Medicine and Graduate School of Medicine, Hokkaido University Kiyohiro Houkin Department of Neurosurgery, Faculty of Medicine and Graduate School of Medicine, Hokkaido University Hidenao Sasaki Department of Neurology, Faculty of Medicine and Graduate School of Medicine, Hokkaido University Ichiro Yabe Department of Neurology, Faculty of Medicine and Graduate School of Medicine, Hokkaido University


Abstract
Background Weight gain is a frequently reported side effect of subthalamic deep brain stimulation; however, the underlying mechanisms remain unclear. The active contact locations in uence the clinical outcomes of subthalamic deep brain stimulation, but it is unclear whether weight gain is directly associated with the active contact locations. We aimed to determine whether weight gain is associated with the subthalamic deep brain stimulation active contact locations.
Methods We enrolled 14 patients with Parkinson's disease who underwent bilateral subthalamic deep brain stimulation between 2013 and 2019. Bodyweight and body mass index were measured before and one year following surgery. The Lead-DBS Matlab toolbox was used to determine the active contact locations based on magnetic resonance imaging and computed tomography. Fluorodeoxyglucosepositron emission tomography data were also acquired before and one year following surgery, and statistical parametric mapping was used to evaluate changes in brain metabolism. The relationship between weight and active contact locations was evaluated with a Spearman rank test with a corrected pvalue < 0.008. We examined which brain regions' metabolism uctuation signi cantly correlated with increased BMI scores and PET data.
Results The body mass index increase was 2.03 kg/m 2 1 year post-surgery. Weight gain was correlated with anterior and dorsal locations of the left-side active contacts, as well as with lateral locations of the right-side active contacts. Furthermore, weight gain was correlated with increased metabolism in the leftside limbic and associative regions, including the middle temporal gyrus, inferior frontal gyrus, and orbital gyrus.
Conclusions Although the mechanisms underlying weight gain following subthalamic deep brain stimulation are possibly multifactorial, our ndings suggest that anterior subthalamic deep brain stimulation alters the activities in the limbic and associative cortical regions, which may then lead to weight gain. Weight gain could be prevented by avoiding stimulation to the anterior part of the subthalamic nucleus.

Background
Subthalamic nucleus deep brain stimulation (STN-DBS) is an established and effective treatment strategy for advanced Parkinson's disease (PD) [1]. However, STN-DBS is associated with several adverse effects, of which weight gain (WG) is one of the most common, although the underlying mechanisms have not been fully elucidated. Previous reports have suggested that WG is associated with lowered resting energy needs [2,3], fewer motor complications (especially dyskinesia) [4][5][6], changes in eating behaviors [7,8], and hormonal factors [9,10]. However, these factors do not completely explain WG, suggesting that WG following STN-DBS is a multifactorial process [11].
The locations of electrodes and active contacts affect the clinical outcomes and adverse effects of STN-DBS [12,13]. Based on animal studies [14,15] and studies examining the human brain using diffusion tensor imaging [16][17][18], the STN is divided into three functional subregions: the sensorimotor, associative, and limbic regions. Previous studies have suggested that superior motor improvement is achieved by stimulating the sensorimotor area, which is located in the dorsolateral part of the STN and is linked to the primary motor and supplementary motor cortices [17,[19][20][21]. Nevertheless, an increased risk of neuropsychiatric side effects may be associated with stimulation of the limbic area, which exists in the anteromedial part of the STN [17,22]. However, few studies have evaluated whether WG is directly correlated with the positions of the active contacts during STN-DBS. Therefore, this study prospectively evaluated patients who underwent STN-DBS for PD and aimed to determine whether WG was correlated with the positions of the active contacts. Moreover, we evaluated whether WG was associated with altered glucose metabolism in speci c brain regions to estimate which functional subregion(s) in uenced WG.

Patients
We aimed to determine whether WG is associated with the subthalamic deep brain stimulation active contact locations. This prospective study recruited PD patients who underwent bilateral STN-DBS surgery at Hokkaido University Hospital between 2013 and 2019. We recruited 16 patients with PD who underwent bilateral STN-DBS surgery, although we excluded one patient who withdrew from participating and another who required electrode removal that was related to an infection. Thus, this study analyzed data from 14 patients, including 12 women. The median age at DBS surgery was 62.5 (55.5-68) years, and the median disease duration at DBS surgery was 14.3 (12-20.5) years. The eligibility for DBS surgery was determined based on the guidelines of the International Parkinson and Movement Disorders Society [23].
This study was conducted following the 1964 Declaration of Helsinki and its later amendments and was approved by the ethics panel of the institutional review board of Hokkaido University Hospital. All patients provided written informed consent before their inclusion in the study. Clinical assessment Clinical assessments were conducted at the preoperative baseline and one year following surgery.
Bodyweight and height were measured and used to calculate the body mass index (BMI, kg/m 2 ). Motor symptoms were assessed using the United Parkinson's Disease Rating Scale (UPDRS) part III in the MedOff state at baseline and in the MedOff and DBS-on state at one year following surgery. Dyskinesia was also evaluated using items 32 and 33 of the UPDRS part IV. Neuropsychological evaluations were based on the Mini Mental State Examination, Frontal Assessment Battery (FAB), Apathy Scale [24], and Patient Health Questionnaire-9 (PHQ-9) [25]. The levodopa-equivalent daily dose (LEDD) [26] was calculated at both baseline and one year following DBS surgery. Brain metabolism was evaluated using uorodeoxyglucose-positron emission tomography (FDG-PET) at baseline and one year following surgery.
The DBS stimulation parameters were recorded after one year and included the stimulation voltage, pulse width, frequency, and active contact location. Assessing the active contact positions The active contact positions were evaluated using Lead-DBS, which is a validated Matlab toolbox [27,28] that was implemented using Matlab 2019b (MathWorks, Natick, MA, USA). Preoperative results were obtained via T2-weighted magnetic resonance imaging (MRI, slice thickness: 1 mm; echo time; 222 ms; repetition time: 2,000 ms), and postoperative results were obtained via computed tomography (CT). The preoperative MRI and postoperative CT images were co-registered using advanced normalization tools [29] and then nonlinearly normalized into Montreal Neurological Institute (MNI) standard space (MNI_ICBM_2009b_NLIN_Asym). The DBS electrodes were automatically reconstructed using the TRAC/CORE algorithm [28] and manually re ned to evaluate their coordinates in MNI space. If more than one contact was used for the stimulation, the mean coordinates of all active contacts were recorded. The positional relationships between the active contacts and the STN were assessed using the DISTAL atlas [30], which is a composite atlas that is based on histology, structural connectivity, and manual segmentations of a multimodal brain template that is normalized in MNI space.

FDG-PET acquisition and preprocessing
Patients were instructed to fast overnight before the PET scans. PET scans were performed in the MedOn at baseline and in the MedOn and DBS-off state post-operation. DBS stimulation was turned off 30-60 minutes before PET scans. 18 F-FDG (4.5 MBq/kg) was administered intravenously, and serum glucose was measured to exclude the patients showing fasting hyperglycemia (> 150 mg/dL). The images were acquired 60 min following FDG administration for 10-min emission scanning. Pet data were obtained using either a GEMINI TF64 (Philips, Amsterdam, Netherlands) PET-CT scanner or a Biograph 64 (Siemens, Munich, Germany) TruePoint PET-CT scanner for different patients. Time-of-ight technology was used with GEMINI TF64 but not with Biograph 64. Images were reconstructed using ordered subset expectation-maximization. The PET images were preprocessed using statistical parametric mapping (SPM12; Wellcome Department of Cognitive Neurology, London, UK) running on Matlab 2019b. The PET images were initially subjected to a ne and nonlinear spatial normalization into the MNI brain space, although we used an FDG-PET-speci c template that was described by Della Rosa et al [31] instead of the default 15 H 2 O-SPM template. The normalized images were then smoothed using an 8-mm isotropic Gaussian lter to compensate for individual anatomical variability. Finally, we created a percent signal change map (PSC map) using the following formula: where V 1y and V b represent voxel values at one year following surgery and baseline, respectively.

Statistical analysis
Scores are reported as the median ± interquartile range (IQR), and nonparametric analyses were used based on the small sample size. Preoperative and postoperative values for BMI, UPDRS part III and IV scores, neuropsychological data, and LEDD were compared using the Wilcoxon signed-rank test, and p < 0.001 was considered as statistically signi cant using Bonferroni correction. Stimulation voltage and active contact coordinates for the left and right sides were compared using Wilcoxon's rank-sum test. The correlation between WG after STN-DBS and the active contact positions (based on X-, Y-, and Z-axis coordinates on both sides) was evaluated using the Spearman rank test, and p < 0.008 was considered as statistically signi cant using Bonferroni correction. To detect potential confounders for WG other than the active contact positions, analyses were performed to determine the correlation between WG and preoperative BMI, improvement in the UPDRS part III and IV (items 32 + 33) scores, LEDD reduction, and stimulation voltage using the Spearman rank test (p < 0.05 was considered as signi cant). Statistical analyses were conducted using JMP Pro software (version 14; SAS Inc., Cary, North Carolina, USA).
The PET data were analyzed using SPM12 (Wellcome Department of Cognitive Neurology, London, UK). To identify which brain regions' metabolism change correlated signi cantly with increased BMI scores, a general linear model was tested at each voxel with the BMI score as a covariate using a PSC map. LEDD was included in the SPM analysis as a covariate, A voxel-level threshold of P < 0.05 (family-wise error corrected for multiple comparisons) was used to assess the SPM t values. If statistical signi cance was not reached, we performed the same analysis with a voxel-level threshold of P < 0.005 (uncorrected for multiple comparisons), considering the small sample size and the study's exploratory nature. Only clusters containing > 100 voxels were reported.

Results
Clinical outcomes Table 1 shows the clinical values at preoperative baseline and one year following DBS surgery. A signi cant motor improvement was observed at 1-year post-surgery, based on the decreased UPDRS part III scores in the MedOff state. A non-signi cant trend toward decreased values for items 32 and 33 of the UPDRS part IV was also observed. Bodyweight and BMI values increased signi cantly following STN-DBS. The neuropsychological assessments revealed a small but statistically signi cant increase in the FAB score but no signi cant changes in the other scales. There was a non-signi cant trend toward decreasing LEDD after DBS surgery.

Stimulation parameters and active contact positions
All patients received bilaterally implanted quadripolar (from contact number '0' for the most ventral contact to '3' for the most dorsal one) DBS electrodes (3387; Medtronic, Minneapolis, MN, USA). Each patient's stimulation parameters at one year following DBS surgery are shown in Table 2. The stimulation intensities were similar between the right and left sides. Figure 1 shows the electrode locations for all patients regarding the STN as de ned using the DISTAL atlas and MNI space. The median coordinates of the right-side active contacts were 12.5 (11.7-13.5) mm on the X-axis, − 14.5 (-13.7 -− 14.5) mm on the Y-axis, and − 6.4 (-5.6 -− 7.6) mm on the Z-axis. The mean left-side coordinates were − 13.6 (-11.5 -− 14. 2) mm on the X-axis, − 13.6 (-11.8 -− 15.6) mm on the Y-axis, and − 6.8 (-5.1 -− 6.9) mm on the Zaxis. The X-, Y-, and Z-axis coordinates were not signi cantly different between the right and left sides, although the left-side Y-axis coordinates had a larger IQR than the right-side Y-axis coordinates.  Correlations with increased BMI On the left side, WG was signi cantly correlated with the Y-axis coordinates (r = 0.77, P = 0.001) and the Zaxis coordinates (r = 0.77, P < 0.001), which suggests that WG was associated with contacts located more anteriorly and dorsally within the STN (Fig. 2). WG was not signi cantly correlated with the active contact coordinates on the left-side X-axis, right-side X-axis, Y-axis, and Z-axis. Preoperative BMI, improvements in the UPDRS part III and IV scores, LEDD reduction, and stimulation voltage also did not correlate with WG.
PET image analysis We analyzed PET data from 13 patients since one patient's PET data were missing. A GEMINI TF64 PET-CT scanner was used for 12 patients, while a Biograph 64 TruePoint PET-CT scanner was used for the remaining patient. None of the voxels were signi cant at a voxel-level threshold of P < 0.05 (family-wise error corrected for multiple comparisons). However, at a threshold of P < 0.005 (uncorrected for multiple comparisons), we identi ed several clusters with positive correlations between WG and increased metabolism in the left hemisphere (Table 3 and Fig. 3). The correlations were observed in the left middle temporal gyrus, inferior frontal gyrus, lateral orbital gyrus, anterior orbital gyrus, and planum polare. We did not observe any negative correlations between WG and brain metabolism in the various areas. The uncorrected thresholds were signi cant (P < 0.005) at the voxel level. Only clusters with > 100 voxels are reported.

Discussion
This study showed a mean BMI increase of 2 kg/m 2 at one year following DBS surgery, which agrees with previous ndings [32]. Furthermore, the mean active contact coordinates in our study were similar to the preferred coordinates for motor improvement from previous studies [33]. Moreover, WG after STN-DBS was signi cantly correlated with some active contact coordinates and with increased glucose metabolism in the left frontal and temporal lobes. We could not detect a signi cant correlation between WG and dyskinesia reduction, while previous ndings remain con icting [4][5][6]33].
The STN plays an important role in reward processing [34], and several studies have indicated that the STN is involved in controlling appetite and eating behavior. For example, a study on non-human primates illustrated that STN activity increased during food reward anticipation and delivery [35]. In humans, stroke or tumors affecting the STN causes hyperphagia and increases appetite [36,37], while abnormal eating behaviors have been reported following STN-DBS [7,[38][39][40][41][42]. The anteromedial part of the STN is thought to be involved in reward and emotion processing, and dysfunction in this area can induce stereotyped and violent behaviors in non-human primates [43]. Moreover, during STN-DBS, stimulation of the anteromedial part of the STN led to an increased risk of abnormal behavior [22,44,45]. Considering these ndings, our results regarding the correlation between WG and anterior active contact locations suggest that WG is associated with the stimulation of the anteromedial part of the STN.
We observed correlations between WG and increased metabolism in the limbic and associative areas but not the sensorimotor areas; however, these were not statistically signi cant in multiple comparisons. A previous FDG-PET study showed that WG after STN-DBS was correlated with increased metabolism in the limbic and associative regions, including the orbitofrontal cortex, lateral and medial parts of the temporal lobe, anterior cingulate cortex, and retrosplenial cortex [33]. Other PET and functional MRI studies have also suggested that a broad network of limbic and paralimbic network structures mediates the desire for food [46][47][48][49][50][51][52][53]. This network is thought to integrate sensory information with the cognitive desire for food and induces behaviors that aim to obtain food [54,55]. Regions with increased brain metabolism in our study were also associated with the processing of desire for food, which suggests that stimulating the anterior part of the STN changed the activities in the limbic and associative areas, which modi ed foodrelated behavior and ultimately WG. Nevertheless, a larger prospective study with correction for multiple comparisons is warranted to con rm this hypothesis.
We also observed that WG was correlated with the left-side active contact Z-axis coordinates, which agrees with a previous report indicating that active contacts located in the zona incerta (dorsally out of the STN) were correlated with increased appetite after STN-DBS [56]. The zona incerta contains neurons expressing melanin-concentrating hormone, which is involved in the regulation of feeding [57]. Thus, our nding of a correlation between WG and the Z-axis coordinates might be explained by the stimulation of zona incerta and neurons that express melanin-concentrating hormone. However, dorsally located contacts are usually also located anteriorly, since the DBS electrodes are usually inserted from the anterodorsal aspect to the posteroventral aspect. Thus, the Y-and Z-axis coordinates acted as confounders for each other. To determine which direction of current spread to the anterior part of STN or Zi is more important for WG, we need to conduct further studies.
We observed that WG was correlated with active contact locations and increased brain metabolism only on the left side. Several studies have also indicated that unilateral STN-DBS causes WG [58,59], although the laterality of this relationship remains unclear. In this study, left-side active contact Y-axis coordinates had a larger IQR value than the right-side coordinates; the greater variability in the left-side coordinates may explain the laterality of our FDG-PET ndings. However, further studies are warranted to determine whether stimulating the right-side anterior part of the STN could in uence brain metabolism, subsequently causing WG.
This study has several limitations. First, we did not assess eating habits or daily food intake. Thus, future studies must con rm whether WG is caused by stimulation of the limbic area, which induced changes in eating behaviors, using preoperative and postoperative data on eating behaviors and food intake. Second, we did not assess hormonal factors or swallowing function, which could have confounded our analyses. Third, although we measured the active contact coordinates, we did not assess the volume of tissue activation (VTA) or connectivity between the stimulation site and brain regions. Recent studies have indicated that the clinical outcomes of DBS could be predicted based on the connectivity pro le of the VTA and cortical areas [60,61]. Therefore, the association between WG and stimulation of the limbic area should be con rmed by analyzing the connectivity between the VTA and cortical area. Fourth, we obtained PET data only in an "off DBS" state after the surgery. Therefore, it is di cult to attribute the changes in brain metabolism after DBS to the plasticity of the neural circuit or the washout process of therapeutic DBS.

Conclusion
In conclusion, we found that WG was correlated with active contact coordinates on several axes. Since WG is likely a multifactorial process, it is di cult to determine which axes might have the greatest effects. However, based on our PET ndings, WG might be associated with stimulating more anteriorly within the STN and subsequent changes in eating behaviors. Further investigations are warranted to con rm this hypothesis by accurately assessing eating behaviors and food intake. This study was conducted following the 1964 Declaration of Helsinki and its later amendments and was approved by the institutional review board of Hokkaido University Hospital. All patients provided written informed consent prior to their inclusion in the study.

Not applicable
Availability of data and materials The dataset(s) used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.