Parkinsonian signs in patients with cervical dystonia treated with pallidal deep brain stimulation

Abstract

Pallidal deep brain stimulation is an established treatment in patients with dystonia. However, evidence from case series or uncontrolled studies suggests that it may lead in some patients to specific parkinsonian symptoms such as freezing of gait, micrographia, and bradykinesia. We investigated parkinsonian signs using the Movement Disorder Society Unified Parkinson’s Disease Rating Scale motor score by means of observer-blinded video ratings in a group of 29 patients treated with pallidal stimulation and a non-surgical control group of 22 patients, both with predominant cervical dystonia. Additional assessments included MRI-based models of volume of neural tissue activated to investigate areas of stimulation related to dystonic symptom control and those likely to induce parkinsonian signs as well as an EMG analysis to investigate functional vicinity of stimulation fields to the pyramidal tract. Compared with controls, stimulated patients had significantly higher motor scores (median, 25th–75th percentile: 14.0, 8.0–19.5 versus 3.0, 2.0–8.0; P < 0.0001), as well as bradykinesia (8.0, 6.0–14.0 versus 2.0, 0.0–3.0; P < 0.0001) and axial motor subscores (2.0, 1.0–4.0 versus 0.0, 0.0–1.0; P = 0.0002), while rigidity and tremor subscores were not different between groups. Parkinsonian signs were partially reversible upon switching stimulation off for a median of 90 min in a subset of 19 patients tolerating this condition. Furthermore, the stimulation group reported more features of freezing of gait on a questionnaire basis. Quality of life was better in stimulated patients compared with control patients, but parkinsonian signs were negatively associated with quality of life. In the descriptive imaging analysis maximum efficacy for dystonia improvement projected to the posteroventrolateral internal pallidum with overlapping clusters driving severity of bradykinesia and axial motor symptoms. The severities of parkinsonian signs were not correlated with functional vicinity to the pyramidal tract as assessed by EMG. In conclusion, parkinsonian signs, particularly bradykinesia and axial motor signs, due to pallidal stimulation in dystonic patients are frequent and negatively impact on motor functioning and quality of life. Therefore, patients with pallidal stimulation should be monitored closely for such signs both in clinical routine and future clinical trials. Spread of current outside the internal pallidum is an unlikely explanation for this phenomenon, which seems to be caused by stimulation of neural elements within the stimulation target volume.

Introduction

Pallidal deep brain stimulation (DBS) has been established for over 10 years as an effective and safe treatment for patients with severe generalized or segmental dystonia (Vidailhet et al., 2005; Kupsch et al., 2006). Following case series and smaller non-randomized studies (Hung et al., 2007; Kiss et al., 2007; Walsh et al., 2013), the efficacy of globus pallidus internus (GPi)-DBS has also been demonstrated in cervical dystonia in a more recent large randomized and blinded multicentre trial (Volkmann et al., 2014). Most randomized controlled trials of DBS in dystonia have reported a low frequency of adverse events with no serious side effects on cognitive and neuropsychiatric functions. Negative consequences of current spread to surrounding areas such as the pyramidal tract (including severe dysarthria) are also infrequent (Vidailhet et al., 2013). However, anecdotal evidence suggests that dystonia patients with pallidal DBS can develop specific parkinsonian symptoms such as freezing of gait or micrographia (Tisch et al., 2007; Berman et al., 2009; Zauber et al., 2009; Blahak et al., 2011; Schrader et al., 2011). Further, one systematic study in 10 patients with cervical dystonia observed mild hypokinesia of gait and a relevant decrease in gait variability upon GPi-DBS as compared with no stimulation (Wolf et al., 2016). However, the pathophysiology of these phenomena and their relevance to the overall outcome of pallidal stimulation for dystonia remain poorly understood. None of the studies mentioned above has used a universal instrument for assessment of parkinsonism nor has any study conducted a direct comparison of dystonia patients on pallidal DBS with patients on non-surgical treatment regimens, since the latter patient group can also exhibit mild parkinsonian features (Haggstrom et al., 2017). Moreover, it remains unknown whether the appearance of parkinsonian features has an impact on quality of life.

Therefore, we sought to assess parkinsonism and its relevance for daily motor functioning and quality of life in patients with predominant cervical dystonia with validated universal tools, comparing patients treated with chronic pallidal DBS with matched patients under conservative treatment regimes. Moreover, we performed a descriptive imaging analysis on volumes of neural tissue activated (VNTAs) in and around the GPi related to (i) reduction of dystonia; and (ii) induction of parkinsonian signs, to elucidate potential mechanisms by which pallidal DBS produces these effects. Lastly, we assessed whether current spread to the pyramidal tract due to pallidal stimulation as assessed by EMG is related to the advent of parkinsonism in a subgroup of DBS treated patients.

Materials and methods

Participants and DBS implantation

Patients with bilateral GPi-DBS for cervical dystonia or segmental dystonia with predominant involvement of the neck were consecutively recruited from our movement disorders outpatient clinic. All patients had been carefully selected to undergo DBS lead implantation between 2004 and 2014. Clinically relevant parkinsonian signs had been excluded before surgery as part of our routine clinical preoperative assessment performed by movement disorder experts (T.F. and P.L.), as were other associated clinical signs such as additional movement disturbances (e.g. ataxia and chorea). Operations were performed using an MRI-guided and MRI-verified approach under general anaesthesia as previously published (Tisch et al., 2007). In brief, implantation of bilateral quadripolar DBS electrodes (MDT-3389 Medtronic or STJ-6164-6149 St Jude Medical) was performed with direct targeting of the posteroventral GPi, as visualized on proton density sequence stereotactic MRI at 1.5 T without microelectrode recording (Hirabayashi et al., 2002; Holl et al., 2010; Nakajima et al., 2011). Intraoperative dynamic impedance recording was used to delineate grey and white matter boundaries (Zrinzo and Hariz, 2009). Confirmation of electrode placement in the posteroventral GPi was obtained in all patients with immediate post-implantation stereotactic MRI. Stimulation parameters were postoperatively individualized in each patient for the best clinical effect and the least side effects. Stimulation parameters had been kept constant at least 6 months prior to the present assessment. A group of patients with predominant cervical dystonia on conservative treatment regimens (without DBS) were also consecutively recruited from our movement disorders outpatient clinic and included as controls. Patients were not systematically screened for inherited dystonia, but two patients included in our study (both in the GPi-DBS group) had a known gene mutation—one a TOR1A mutation (DYT 1) and one a THAP1 (DYT 6) mutation. All patients provided written informed consent for the study according to the Declaration of Helsinki. The study was approved by the local ethics committee.

Experimental design

In this case-control and multimodal study clinical and neurophysiological assessments of patients were performed within 1 day (D.G. and F.B.). Clinical assessments were videotaped and videos were rated by a clinician blinded to treatment group and stimulation condition (P.M.). Imaging data were retrieved from intraoperative MRI scans and processed as detailed below (H.A.).

Clinical assessments

For this study, DBS treated patients and control patients underwent clinical assessment in 2015 consisting of standardized neurological examination with validated rating scales complemented by validated questionnaires. The motor section of the Movement Disorder Society Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) (Goetz et al., 2008) as well as the Toronto Western Spasmodic Torticollis Rating Scale (TWSTRS) torticollis severity scale (Comella et al., 1997) were used. Furthermore, the TWSTRS disability scale, TWSTRS pain scale, MDS-UPDRS motor experiences of daily living questionnaire, Freezing of Gait (FOG) questionnaire (Giladi et al., 2000), and the descriptive system of the EQ-5D-3L (European quality of life, five dimensions, three level version; EuroQol Group, 1990) were applied to assess disability and quality of life. After initial assessment, GPi-DBS was switched off and, in patients who tolerated this condition without excessive recurrence of dystonia, stimulation was kept off for a median of 90 (25th–75th percentile: 60–100) min before reassessing TWSTRS torticollis severity and MDS-UPDRS motor sores.

All clinical assessments were videotaped and later assessed by a movement disorder neurologist (P.M.) experienced in the rating of the two scales and blinded to group assignment and stimulation condition. In the GPi-DBS group TWSTRS torticollis severity scores were compared with scores that had been recorded preoperatively to determine the benefit of chronic pallidal stimulation in reducing severity of cervical dystonia. MDS-UPDRS motor ratings did not include the item for neck rigidity in any of the patients, nor any other items in body parts with significant dystonia in individual patients. Rigidity ratings were taken from the actual clinical assessments. MDS-UPDRS motor scores were subgrouped into a rigidity subscore (item 3), a bradykinesia subscore (items 4–8), an axial motor subscore (items 9–14), and a tremor subscore (items 15–18).

MRI acquisition and processing

DBS contacts’ VNTAs modelling was performed on MRI scans obtained intraoperatively on a 1.5 T Siemens Avanto interventional MRI scanner (28 scans available). SureTune® (Medtronic Inc.), a DBS therapy planning platform, was used to model VNTAs around individual contacts. The platform applies homogeneous finite element simulations of the distribution of the electric potential together with coupled axon cable models (Aström et al., 2009), where the latter were composed of 21 nodes, with a diameter of 2.5 µm. Intraoperative MRI scans were uploaded and post-implantation MPRAGE (magnetization-prepared rapid gradient echo) used to fit the DBS lead model within the MRI artefact produced by the leads. Individual VNTAs were then generated according to the respective patient’s chronic DBS settings in terms of stimulation amplitude and pulse width as previously described (Aström et al., 2009). Binary image files of VNTAs along with corresponding transformation matrices were exported and processed in MATLAB (Mathworks Inc.) using in-house software to generate NIfTI (Neuroimaging Informatics Technology Initiative) files. MPRAGE MRI scans were then registered to the Montreal Neurological Institute (MNI) ICBM (International Consortium of Brain Mapping) 152 non-linear 6th Generation Symmetric Average Brain Stereotaxic Registration Model (Grabner et al., 2006) using non-linear registration. The resulting transformation warps were in turn used to transform all VNTAs to MNI space. Transformed VNTAs were thresholded at 95% using FSLmaths (Functional MRI of the Brain Software Library; FSL 5.0) to remove the interpolation effect. Geometrical centres of VNTAs correspond to the active contact location and are shown in Supplementary Fig. 1 for illustrative purposes. A group average VNTA was generated from the MNI warped individual VNTAs by using the FSLmaths function with -Tmean flag, which is equivalent to a sum of all voxels in all VNTAs. To generate efficacy and side-effect average clusters, patients were independently ranked according to per cent dystonia reduction upon chronic DBS as measured by the TWSRTS torticollis severity scale, as well as appearance of bradykinesia (hemi-body) and axial motor symptoms as measured by the respective MDS-UPDRS derived subscores (consistent differences across groups and stimulation conditions were observed for bradykinesia and axial motor symptoms, but not for rigidity or tremor; see ‘Results’ section). Individual VNTAs were weighted by multiplying each VNTA by the rank for each of these three effect categories separately. Group average clusters for the three clinical effects were then generated from the weighted VNTAs. The resulting clusters where thresholded at 95% for illustrative reasons using FSLeyes. Additionally, the MNI coordinates of the clusters’ centre of gravity were calculated using the FSLcluster function yielding coordinates for the stimulation side of maximum effect for the three categories (dystonia reduction, severity of bradykinesia and axial motor symptoms).

EMG recordings

To estimate the functional proximity to the pyramidal (corticobulbar and corticospinal) tract, EMG activity of the orbicularis oris muscle and the first dorsal interosseous muscle upon contralateral pallidal stimulation was recorded in patients, who tolerated being off high frequency stimulation for the necessary time period. Recordings were performed using 9 mm Ag-AgCl surface cup electrodes with the same equipment in a standardized fashion as recently reported in a cohort of patients with Parkinson’s disease with subthalamic stimulation (Mahlknecht et al., 2017). In brief, signal amplification gain was set at 2000, recording frequencies at 10 kHz, and a band pass frequency filter at 20–1000 Hz. The signal was digitized and saved for offline analysis blinded to the patients’ clinical examination using Signal V4.08 (CED). For the orbicularis oris muscle, the active electrode was placed 1 cm lateral to the mouth corner, and the reference 2 cm lateral. For the first dorsal interosseous, the muscle belly and the tendon of the same muscle were used, respectively.

First, motor evoked potentials (MEPs) were assessed while patients were sitting in a comfortable armchair and instructed to relax, but not speak or sleep. Monopolar stimuli were elicited by the impulse generator at contacts used for chronic stimulation on each electrode. In patients with interleaved stimulation the contact with the higher stimulation amplitude and in patients with bipolar stimulation the contact representing the cathode were used. Stimulation at a low frequency of 3 Hz (allowing enough time for MEP recordings before subsequent stimuli) and at pulse width used for chronic stimulation was increased in 0.5 mA steps up to 8.0 mA or until bothersome side effects appeared. Thirty sweeps of EMG triggered by the stimulation artefact were averaged per condition to detect the resting motor threshold (RMT) of pyramidal tract activation. RMTs were determined as the lowest stimulus intensity inducing MEPs clearly recognizable above background activity upon visual inspection (in most instances this was the case when MEPs reached >10 µV in amplitude).

Furthermore, active motor thresholds (AMT) were assessed during sustained muscle contraction of ∼25% of maximum voluntary force production (provided to the participants as visual feedback with a line on the EMG screen, which they were asked to match), of the first dorsal interosseous muscle by squeezing a roll of tape and of the orbicularis oris muscle by forming a smiling mouth. Stimuli were elicited via the clinically used contacts on each side as for the RMT. AMT was determined as the lowest stimulus intensity inducing MEPs clearly recognizable above background activity upon visual inspection.

Statistical analysis

As data were largely non-normally distributed, as shown by the Shapiro-Wilk test, we used non-parametric tests for comparative statistics. Continuous variables are uniformly given by medians and 25th–75th percentiles. EQ-5D-3L states were transformed into index values based on the UK EQ-5D index tariffs (Sullivan et al., 2011) anchored at 1 (full health) and 0 (dead). For the descriptive analysis we used two-sided Mann-Whitney U-tests to compare age, disease duration, and scale scores between GPi-DBS versus control patients and paired two-sided Wilcoxon tests to compare scale scores under off versus on stimulation condition. Binary variables were compared using the chi-squared test. Spearman rank test was used for all correlation analyses. The association of parkinsonian signs with quality of life was assessed using the Spearman rank test adjusted for age, sex, disease duration, cervical dystonia severity, and treatment group. Additionally, a linear regression analysis adjusted for the same confounders was performed. For the latter analysis continuous variables were log₂-transformed to achieve an approximately normal distribution; results from the regression analysis are given as per cent change on the EQ-5D index value per doubling of value in the MDS-UDPRS motor scores. As dystonia is known to potentially cause (postural and/or action) tremor, the same analyses were repeated leaving tremor out from the MDS-UPDRS motor score. Finally, Euclidean distances from active contact locations to the centres of gravity for the three clinical effects (dystonia reduction, severity of bradykinesia and axial motor symptoms) were calculated for each side. They reflect direct geometric distances in MNI space and are independent of the direction between points. Distances from active contact locations to the centres of gravity for the three clinical effects were then correlated with the amount of the corresponding clinical effects (per cent dystonia reduction with chronic stimulation, contralateral MDS-UPDRS bradykinesia subscore and MDS-UPDRS axial motor symptom subscore) using the Spearman rank test. Correlation analysis for dystonia improvement and axial motor symptoms was undertaken using the mean of the distances of both sides. SPSS 22.0 (IBM Corp., Armonk, NY) was used for all statistical analyses. The significance level was set at two-sided P-value of <0.05.

Results

Patient characteristics

Twenty-nine dystonia patients with chronic GPi-DBS and 22 dystonia patients without DBS were recruited and included in the present study. At the time of the present evaluation, all patients received monopolar stimulation on one or two adjacent active contacts, except for three patients who had been switched to an interleaved stimulation mode on two adjacent contacts and two patients who had been switched to a bipolar stimulation mode on two adjacent contacts. Stimulation was set at a frequency of 130 Hz in all patients, except for the three on interleaved stimulation mode (125 Hz). Median pulse widths were 60 (60–90) μs on both sides and median stimulation amplitudes 3.7 (3.5–4.6) V on the left side and 3.7 (3.3–4.6) V on the right side.

At the time of assessment at a median of 5.0 (2.0–7.0) years after DBS implantation, stimulated patients had significantly lower TWSTRS torticollis severity scores compared with their preoperative status (11.0, 5.5–14.5 versus 18.0, 11.3–20.0; Z = −3.4, P = 0.0008).

Patient controls did not significantly differ from GPi-DBS patients in terms of age, sex, disease duration, or dystonia severity (GPi-DBS patients on stimulation). Characteristics of the two patient groups are presented in Table 1.

Comparison between groups in terms of parkinsonian signs, disability, pain and quality of life

GPi-DBS patients had higher MDS-UPDRS motor scores as well as bradykinesia and axial motor subscores compared with controls (Fig. 1, see Supplementary Table 1 for numerical data). Tremor subscores were higher in the control group, but this difference was not statistically significant. Tremor was mainly a postural and/or action tremor and only two patients exhibited a concomitant rest tremor (one in the GPi-DBS group and one in the control group). Rigidity subscores were not different between groups. On a questionnaire basis, GPi-DBS patients had less torticollis-related disability but more parkinsonism related disability, although this was not statistically significant (Table 1). FOG scores were higher in the stimulation group. EQ-5D-3L index values were higher in DBS patients compared with control patients indicating better quality of life.

A correlation analysis and a regression analysis, both adjusted for potential confounders, showed that higher MDS-UPDRS motor scores were associated with decreased quality of life (Table 2).

Off stimulation condition

As a sensitivity analysis, clinical assessments were repeated in an off stimulation condition in the GPi-DBS group to examine parkinsonism domain severity changes upon switching off stimulation. In the 19 participants tolerating this condition, TWSTRS torticollis severity scores compared with those not tolerating being off stimulation were lower, although this was not statistically significant (14.0, 7.8–20.0 versus 18.5, 17.3–21.5; Z = −3.4, P = 0.0879). All other baseline characteristics outlined in Table 1 were not different between these two groups (all P > 0.3). Switching off stimulation led to a worsening of dystonic symptoms (Fig. 2, see Supplementary Table 2 for numerical data). The percentage TWSTRS torticollis severity increase when stimulation was switched off correlated with the percentage TWSTRS torticollis severity reduction on stimulation versus the preoperative status (r = 0.599, P = 0.0099). Switching stimulation off lowered MDS-UPDRS scores as well as bradykinesia and axial motor subscores (Fig. 2 and Supplementary Table 2).

Potential factors driving the occurrence of parkinsonian features

Neither TWSTRS torticollis severity on stimulation nor the percentage reduction of torticollis severity upon stimulation (pre- versus postoperative status) correlated with the MDS-UPDRS motor scores or any of its subscales (all Spearman rank correlations P > 0.4). Similarly, upon switching stimulation off, the percentage increase in torticollis severity did not correlate with the decrease in MDS-UPDRS motor scores or any of the subscales (all Spearman rank correlations P > 0.5). The duration of stimulation and stimulation amplitude did not correlate with the MDS-UPDRS motor or any of the subscores (all Spearman rank correlations P > 0.2).

Imaging analysis

The descriptive imaging analysis showed that the ‘sweet spot’ for dystonia reduction upon stimulation was localized in the posteroventrolateral GPi (Fig. 3). On visual inspection the clusters associated with bradykinesia and axial motor symptoms largely overlapped with the ‘sweet spot’, but extended slightly more inferiorly and anteriorly. The coordinates of centres of gravity of clusters are illustrated in Table 3 and were similar for the three clinical effects. The distances of active contact locations to the centres of gravity of the three clinical effects ranged between 0.3 and 6.3 mm and medians were between 2.1 and 2.6 mm. Distances to the centres of gravity correlated inversely with the amount of the corresponding three clinical effects (per cent dystonia reduction with chronic stimulation, contralateral MDS-UPDRS bradykinesia subscore and MDS-UPDRS axial motor symptom subscore), but this was statistically significant only for dystonia reduction (Supplementary Fig. 2).

Motor evoked potentials

A total of 28 DBS electrode contacts in 14 patients, who tolerated being off stimulation for the necessary time period, were assessed while at rest and upon activation using stimulation at a low frequency (3 Hz) at the clinically used contacts. RMT and AMT in the contralateral orbicularis oris muscle were elicited in 21 contacts and 27 contacts and in the contralateral first dorsalis interosseus muscle in 15 contacts and 26 contacts, respectively, using stimulation strengths up to 8 mA. In those contacts where RMT or AMT could not be elicited, putative thresholds of 9 mA were assumed for correlation analysis to avoid missing values. Median thresholds are presented in Supplementary Table 3 and were lower in the orbicularis oris muscle compared with the first dorsalis interosseus and upon activation compared with the resting condition. None of the thresholds were correlated with the MDS-UPDRS motor scores or with single parkinsonian signs (all P > 0.15, n = 14 data points). Also, lateralized scores for parkinsonian features were not correlated with any of the thresholds (all P > 0.3, n = 28 data points).

Discussion

Clinical findings

In this study we evaluated parkinsonian symptoms in patients with predominant cervical dystonia treated with bilateral pallidal DBS and compared findings with those of a group on conservative treatment regimens. Patients with GPi DBS, while significantly improved compared to their preoperative status, had similar dystonia severity as the control group of patients on non-surgical treatments. However, GPi-DBS treated patients had high median MDS-UPDRS motor scores of 14 points compared with 3 points in control patients. The difference in median scores to control patients of 11 points exceeds the ‘minimally clinically important difference’ of ∼4 points as determined in patients with Parkinson’s disease (Horváth et al., 2015). Parkinsonian features that drove this difference were bradykinesia and axial motor symptoms, similar to descriptions in earlier case series and uncontrolled studies in selected patients (Tisch et al., 2007; Berman et al., 2009; Zauber et al., 2009; Blahak et al., 2011; Schrader et al., 2011). Our findings therefore confirm these observations of parkinsonian signs emerging due to GPi-DBS in dystonia patients that by far exceeds the amount of mild parkinsonian features thought to accompany idiopathic and inherited isolated dystonias (Haggstrom et al., 2017). Various forms of tremor, mainly with an action and/or postural component, can occur more commonly in dystonia patients, potentially resembling Parkinson’s disease tremor and being a frequent cause of ‘SWEDDs’ (subjects without evidence of dopaminergic deficit) (Erro et al., 2016; Gigante et al., 2016). In line with these observations, in this cohort there was less tremor in DBS patients compared with control patients (although not statistically significant), which may be due to the therapeutic effect of pallidal DBS on dystonic tremor (Fasano et al., 2014; Volkmann et al., 2014). Rigidity was not different between groups.

After switching stimulation off for a median of 90 min, a significant reoccurrence of dystonic symptoms (worsening of 29%) and significant reduction of overall parkinsonian symptoms (improvement of 18%) was observed. Reduction in parkinsonism was seen in the same subdomains bradykinesia (improvement of 19%) and axial motor symptoms (improvement of 39%), whereas changes were not seen for rigidity or for tremor. Interestingly, in two of our DBS-treated patients, dopamine transporter scans had been performed as part of their clinical workup because parkinsonian symptoms were severe enough to warrant exclusion of nigrostriatal dopaminergic deficit. Although both patients were markedly bradykinetic, one with a severe hypokinetic gait disorder (MDS-UPDRS motor score of 42 points) and the other a rest tremor (MDS-UPDRS motor score of 23 points), imaging results in both patients were normal. In general, rest tremor was uncommon (only observed in two patients overall together with postural and/or action tremor). This further argues against nigrostriatal dopaminergic dysfunction accounting for parkinsonian signs seen in GPi-DBS treated patients with dystonia. Also, the observed median of 14 motor score points on the MDS-UPDRS is lower compared with motor scores in patients with Parkinson’s disease even in early disease stages up to 3 years of disease duration, where mean scores range around 20–28 points (Holden et al., 2018). Although we did not formally dichotomize groups as there is no validated cut-off for the MDS-UPDRS motor score indicating the start of parkinsonism, descriptively, nearly one-quarter of DBS-treated patients (7 of 29) were in this range or above compared with none in the control group.

Patients with DBS tended to experience less torticollis-related disability compared with patients on non-surgical treatments, but experienced more parkinsonism-related disability, which also exceeded the minimally clinically important difference (Horváth et al., 2017). Although we have not applied objective measures of gait such as kinematic assessment, questionnaire-assessed features of freezing of gait were significantly greater in the stimulation group compared with the control group and indicated decreased mobility. Quality of life was significantly better in patients with DBS, as expected from results from large randomized controlled trials (Kupsch et al., 2006; Mueller et al., 2008; Volkmann et al., 2014). However, there was a variability in quality of life across patients in the two groups and part of this variability can be accounted for by parkinsonian symptoms that have a significant negative impact on quality of life as shown in a correlation and regression analysis adjusted for potential confounders. This was also true if tremor was left out from analysis. Although these observations should not impact on the decision to treat dystonia patients with pallidal DBS in the first place, it highlights the importance of a thorough screening of those treated with chronic pallidal stimulation for features of parkinsonism.

Pathophysiological aspects

The mechanism by which pallidal DBS induces parkinsonism is largely unknown. It could be speculated that parkinsonism occurring upon pallidal stimulation is a result of current spread outside the GPi and subsequent activation of adjacent fibre tracts such as the pyramidal tract. However, our EMG assessment of resting and activation motor thresholds of the corticospinal and corticobulbar tract in association with pallidal DBS did not correlate with the severity of overall parkinsonism or any parkinsonian features. This underlines the observation of an earlier case study that did not find any correlation between bradykinesia and the structural proximity of GPi DBS electrodes to the internal capsule as assessed on neuroimaging (Berman et al., 2009).

Another speculation concerns the stimulation of different functional zones within the GPi as observed in GPi-stimulated patients with Parkinson’s disease, where activation of lower contacts may lead to pronounced improvement in levodopa-induced dyskinesias but development of akinesia (Krack et al., 1998). In contrast, stimulation of more dorsal contacts may lead to moderate improvements of akinesia and may even induce dyskinesias in some patients. In our descriptive imaging analysis, the ‘sweet spot’ for dystonia reduction with chronic stimulation was located only few millimetres above the ventral border of posteroventrolateral GPi, and in proximity of the medial medullary lamina. This ideal target for stimulation coincides with previous reports on the ‘sweet spot’ in three imaging studies (Starr et al., 2006; Cheung et al., 2014; Schönecker et al., 2015). In addition to confirming this ‘sweet spot’, the ‘hot spot’ of stimulation potentially responsible for the advent of parkinsonian bradykinesia and axial motor symptoms was assessed. Clusters for bradykinesia and axial motor symptoms largely overlapped with the ‘sweet spot’, but extended slightly more inferiorly and anteriorly. However, coordinates of maximum effects for the three clinical categories were very similar to one another. This finding allows for two possible simplified interpretations: (i) stimulation of the neural elements within the GPi that are responsible for alleviation of dystonia are also responsible for the induction of parkinsonian symptoms, both most likely attributable to altered outflow activity of pallido-thalamo-cortical pathways; and (ii) different functional neural elements within the GPi are responsible for the two potential effects. The observation that not all GPi DBS dystonia patients had high MDS-UPDRS motor scores would speak to this second hypothesis. However, the true explanation for the advent of parkinsonian signs upon pallidal stimulation is likely to be more complex. Evidence from neurophysiological studies suggests that pallidal DBS reverts pathologically increased cortical excitability and enhanced long-term potentiation-like synaptic plasticity in cortical circuits in patients with dystonia (Kuhn et al., 2003; Quartarone et al., 2003; Ruge et al., 2011). This effect may be mediated via suppression of pathologically enhanced pallidal low frequency (theta) activity, which may no longer be propagated along basal ganglia nuclei to the cerebral cortex upon chronic stimulation (Barow et al., 2014). Excessive reduction in cortical plasticity upon pallidal stimulation may be related to the induction of parkinsonian signs, which would explain their delayed emergence. The fact that in our cohort short-term withdrawal of stimulation led only to a partial, but significant, alleviation of bradykinesia and axial motor features underlines this hypothesis. Interestingly, a recent neurophysiological study found pallidal peak theta activity exactly at the ‘sweet spot’ for dystonia reduction (Neumann et al., 2017). The coordinates of the ‘hot spots’ for bradykinesia and axial motor symptoms found in our study were similar, which would further add weight to the hypothesis that excessive suppression of theta activity is a driving factor for the advent of parkinsonian signs.

Decreasing amplitude of stimulation, moving the stimulation field towards higher contacts, or lower frequency of stimulation have anecdotally been shown to potentially help decrease parkinsonian symptoms (Tisch et al., 2007; Berman et al., 2009; Schrader et al., 2011; Ba et al., 2016). This may, however, come at the cost of (partially) losing dystonic symptom control. Another recent study found that high-frequency stimulation led to deterioration in a finger tapping task as compared with no and low frequency stimulation, thus suggesting a frequency-specific modulation of hand motor function in pallidal DBS (Huebl et al., 2015). In our sample, the advent of parkinsonian features was independent from clinical variables such as dystonia severity and amplitude of stimulation. Future interventional studies should systematically assess the potential of changing stimulation parameters such as using low frequency stimulation of 60–100 Hz, manipulating pulse width, or using directional stimulation with novel, commercially available DBS leads in preventing or alleviating parkinsonism. Interestingly, subthalamic DBS has also been successfully used in dystonia patients (Ostrem et al., 2011, 2017; Cao et al., 2013). These studies, however, were rather small and did not include an assessment of parkinsonism. Whether STN-DBS in dystonia patients can provoke dyskinesia also requires careful study.

Limitations and strengths

There are some limitations that should be taken into account when interpreting the findings of our study. There was no preoperative and longitudinal formal assessment of parkinsonism in our dystonia patients by the use of standardized scales such as the MDS-UPDRS, and the temporal course of the advent of parkinsonian signs due to GPi-DBS in dystonic patients remains unknown. Nevertheless, clinically obvious parkinsonism had been ruled out before DBS implantation and the use of a comparable control group with blinded assessment confirms that troublesome parkinsonian signs occur in many dystonia patients under pallidal stimulation after some time (months to years) of stimulation. Most, but not all dystonia patients tolerated switching stimulation off in order to examine whether parkinsonian signs were alleviated. However, those evaluated for this sub-analysis were not different in any of the characteristics compared with those not being switched off. Voxel-wise statistical analysis of imaging data was not performed due to the large number of voxels in the pallidal area of interest relative to the limited number of clinical observations. Instead, a correlation analysis between non-directional distances between active contacts to the centres of gravities of the three clinical effects and the size of the clinical effects was carried out. Here we found inverse correlations between these two parameters as expected, but this was statistically significant only for dystonia reduction. Nevertheless, multicentre imaging studies in larger number of patients are needed to confirm findings of our descriptive imaging analysis. As this was a hypothesis-driven study, we did not formally adjust analyses for multiple comparisons. However, main outcomes of our study would largely survive such an adjustment as indicated in Supplementary Tables 1 and 2. Strengths of our study include the standardized and multimodal, blinded assessments in a large group of patients. Of note, after a median of 5 years of pallidal DBS torticollis severity was still improved by 32% as compared with the preoperative status, similar to improvements seen after 6 months of open label stimulation in randomized controlled trials (Volkmann et al., 2014). These findings underline the representativeness of our cohort of dystonia patients treated with long-term pallidal DBS.

Conclusion

In summary, parkinsonian signs, particularly bradykinesia and axial motor signs due to pallidal stimulation in dystonic patients are frequent and negatively impact on motor functioning and quality of life. Although these observations should not influence the decision to treat dystonia patients with pallidal DBS, our findings have implications for the follow-up of dystonia patients and potentially also for other movement disorders that are treated with pallidal DBS. Patients under chronic pallidal stimulation should be monitored closely for features of parkinsonism, both in clinical routine and future clinical trials. Spread of current outside the GPi is an unlikely explanation for this phenomenon, which seems to be caused by stimulation of neural elements within the stimulation target volume. Future interventional studies should systematically look into the effect of switching active contacts, manipulating frequency and/or pulse width of stimulation with regard to the prevention or alleviation of parkinsonian signs in dystonia patients under pallidal DBS.

Funding

This study was funded by a grant from the Brain Research Trust (BRT). The Unit of Functional Neurosurgery, UCL Institute of Neurology, Queen Square, London is also supported by the Parkinson’s Appeal and the Sainsbury Monument Trust. This work was done both at UCL and UCL Hospitals NHS Trust and was funded in part by the Department of Health National Institute for Health Research Biomedical Research Centres funding scheme. P.M. was supported by a research grant from the Austrian Society of Neurology.

Competing interests

P.M. and D.G. received an educational grant from Medtronic. M.H. received honoraria from Medtronic and Boston Scientific. L.Z. received honoraria from Medtronic and Boston Scientific. T.F. received honoraria from Medtronic, St Jude Medical, Profile Pharma, Bial, Abbvie Pharmaceuticals, UCB Pharmaceuticals, and Oxford Biomedica. P.L. received honoraria from Medtronic, St Jude Medical, and Boston Scientific.

Abbreviations

Abbreviations
 
• AMT

  active motor threshold

 
• DBS

  deep brain stimulation

 
• EQ-5D-3L

  European quality of life, five dimensions, three level version

 
• GPi

  globus pallidus internus

 
• MDS-UPDRS

  Movement Disorder Society Unified Parkinson’s Disease Rating Scale

 
• MEP

  motor evoked potential

 
• RMT

  resting motor threshold

 
• TWSTRS

  Toronto Western Spasmodic Torticollis Rating Scale

 
• VNTA

  volume of neural tissue activated

References

Author notes