Programming Algorithm for the Management of Speech Impairment in Subthalamic Nucleus Deep Brain Stimulation for Parkinson ’ s Disease

Objectives: Deep brain stimulation (DBS) of the subthalamic nucleus (STN) for Parkinson ’ s disease (PD) has an ambiguous relation to speech. Speech impairment can be a stimulation-induced side effect, and parkinsonian dysarthria can improve with STN-DBS. Owing to the lack of an up-to-date and evidence-based approach, DBS reprogramming for speech impairment is largely blind and greatly relies on the physician ’ s experience. In this study, we aimed to establish an evidence-and experience-based algorithm for managing speech impairment in patients with PD treated with STN-DBS. Materials and Methods: We performed a single-center retrospective study to identify patients with STN-DBS and speech impairment. Onset of speech impairment, lead localization, and assessment of DBS-induced nature of speech impairment were collected. When DBS settings were adjusted for improving speech, the magnitude and duration of effect were collected. We also performed a systematic literature review to identify studies describing the effects of parameter adjustments aimed at improving speech impairment in patients with PD receiving STN-DBS. Results: In the retrospective study, 245 of 631 patients (38.8%) with STN-DBS had signi ﬁ cant speech impairment. The probability of sustained marked improvement upon reprogramming was generally low (27.9%). In the systematic review, 23 of 662 identi ﬁ ed studies were included. Only two randomized controlled trials have been performed, providing evidence for interleaving-interlink stimulation only. Considerable methodologic heterogeneity precluded the conduction of a meta-analysis. Conclusions: Speech impairment in STN-DBS for PD is frequent, but high-quality evidence regarding DBS parameter adjustments is scarce, and the probability of sustained improvement is low. To improve this outcome, we propose an evidence-and experience-based approach to address speech impairment in STN-DBS that can be used in clinical practice.


INTRODUCTION
The ability to speak properly is an important factor for the quality of life of patients with Parkinson's disease (PD). 1,2 Speech impairment (SI) is estimated to occur in the course of the disease in 50% to 90% of patients with PD. 3,4 Deep brain stimulation (DBS) of the subthalamic nucleus (STN) is an established treatment for refractory motor fluctuations or tremor in PD. 5 However, STN-DBS has an ambiguous relation with speech. Although STN-DBS can improve parkinsonian hypokinetic dysarthria, mainly by increasing vocal loudness, it may also impair speech intelligibility by compromising articulation. 6,7 Although the mechanism is not yet elucidated, modulation of corticobulbar fibers but also sensorimotor STN itself are believed to be involved in the occurrence of stimulationinduced dysarthria. 7 Hence, managing SI in patients with PD receiving STN-DBS is challenging. The current generation of implantable pulse generators provides a pleiotropy of parameter adjustments that can be tried to improve SI, which, however, comes at the cost of increasing complexity and is very timeconsuming. Since the publication of a proposed practical programming algorithm to improve SI in STN-DBS in 2016, several programming options (eg, directional stimulation, short pulse width, and interleaving-interlink stimulation) have emerged. 8 Moreover, final recommendations were mainly based on expert opinions. 8 Owing to the lack of an up-to-date and evidence-based approach, DBS reprogramming for SI is largely blind and greatly relies on the physician's experience. In this study, we aim to streamline DBS reprogramming for SI on the basis of the patient's profile, and considering the evidence, probability of effect, and burden of parameter adjustments.

Retrospective Study
To identify subjects with PD with STN-DBS and SI, the Amsterdam PD DBS database was used. This monocentric database constitutes all patients with PD treated with DBS at Amsterdam UMC location AMC since the year 2000 and contains demographics in addition to multimodal patient information (eg, clinical assessments, medication, and imaging) at all DBS stages (ie, preoperatively, surgically, and postoperatively). The generation and use of this database are according to local regulations and have received a waiver for active consent by the local ethical committee. Database search was performed on December 20, 2021. For this study, inclusion criteria were STN being the DBS target and presence of sufficient information in the electronic health record to assess the presence and management of SI. For these included patients, the data extracted from the Amsterdam UMC location AMC PD DBS database were demographics (age, sex, PD duration), date of DBS surgery, and data regarding SI. The latter were based on an electronic health record search using Dutch words concerning speech (ie, "dysartrie," "spraak," "spreken," and "praten"). The data extracted were presence and date of onset of "significant speech impairment" (defined as "speech impairment perceived as bothersome by the patient and/or caregiver"), execution and outcome of stimulation on-off assessment regarding speech (to assess a stimulation-induced nature of SI), evaluation of electrode location (ie, assessment of lead localization because of occurrence of SI, presence of suboptimal lead placement-ie, too distant from dorsolateral sensorimotor STN as judged by the treating neurosurgeon, and surgical lead repositioning). For each patient with significant SI, execution and outcome of DBS parameter adjustments were recorded. For each of ten possible DBS parameter adjustments (ie, increasing amplitude, decreasing amplitude, alternative stimulation contact, bipolar stimulation, interleaving stimulation, short pulse width [SPW; ie, pulse width <60 μs]), low frequency stimulation (LFS; ie, frequency <100 Hz), interleavinginterlink (IL-IL; ie, two temporally interleaving but spatially overlapping LFS configurations), directional stimulation, and ondemand speech program), the aspects assessed were date of parameter change, magnitude of effect (ie, no effect, some improvement-defined as "improvement perceived as clinically not relevant by the patient and/or caregiver," marked improvement -defined as "improvement perceived as clinically relevant by the patient and/or caregiver," and worsening), duration of effect (ie, transient, or permanent-defined as improvement still present at last follow-up visit), date of loss of effect (if applicable), and reason of loss of effect (if applicable-ie, spontaneous return of SI, abolishment of parameter change due to a side effect, or abolishment of parameter change due to worsening of other PD symptoms).

Systematic Review
An electronic search was performed to identify eligible studies in the database of MEDLINE, via PubMed (from inception to September 14, 2021). Key terms were "speech impairment," "deep brain stimulation," and "Parkinson's disease." The search algorithm for PubMed is reported in Supplementary Data Table S1. No filters or limits were used. No restrictions were imposed on publication date, publication status, or language, to obtain a complete overview of the evidence. The results of the literature search were imported in Endnote X9 (Clarivate, Philadelphia, PA) to manage screening and selection. All duplicates were removed first. Screening of titles and abstracts, to ascertain whether each study fulfilled the research question, was performed by one author (Veronia Lotfalla) independently. Next, full texts were studied to conclude whether inclusion and exclusion criteria were met. The flowchart regarding the literature search was created with Review Manager (RevMan-The Cochrane Collaboration, 2020).
Studies were included if the study population comprised patients with PD who were treated with STN-DBS, and one or more stimulation parameter adjustments aimed to improve speech were examined. Studies whose patient population included SI due to other conditions not associated with stimulation or PD were excluded. No further inclusion criteria regarding the nature, characteristics, or severity of SI were imposed. Eligible study designs were all forms of original research, which comprised all sorts of clinical trials, observational studies that involved prospective and retrospective cohort studies, case-control studies, case series (only if n ≥ 5 patients), and cross-sectional studies.
The data extracted from included studies were study design, number of patients, age, PD duration, DBS parameter intervention(s), speech outcome, motor outcome, and study duration. Speech outcome was subdivided in group effect size, statistically significant outcomes (ie, a p value ≤ 0.05), and proportion of patients with objective or subjective improvement. Motor outcome was assessed with Movement Disorder Society-Sponsored Revision of the Unified Parkinson's Disease Rating Scale (MDS-UPDRS) part II-IV scores.
Risk of bias assessment was performed using the critical appraisal checklist tools provided by the Joanna Briggs Institute (Adelaide, Australia, 2017). Within-subject studies and retrospective SWINNEN ET AL www.neuromodulationjournal.org chart reviews were critically appraised through the checklist for case series because this checklist was deemed most appropriate to assess the risk of bias in these studies.

Statistical Analysis
All statistical analyses were performed in IBM SPSS Statistics for Windows, version 28.0.1.0 (IBM Corp, Armonk, NY). Variables were expressed as mean ± SD or median with interquartile range (25th-75th percentile) for parametric and nonparametric data, respectively. The Shapiro-Wilk test was used to test for normality. Chi-square test was used to assess the difference of the DBS adjustment effect between SI onset before and after DBS. This was only performed for "marked permanent improvement" and "worsening of speech" because these were deemed clinically relevant, and was only performed if the DBS parameter adjustments were used in ≥30 patients. Statistical significance was set at p ≤ 0.05.

Retrospective Study
Of 631 patients with PD treated with STN-DBS, 245 (38.8%) had significant SI somewhere along the postoperative course. Patient characteristics are described in Supplementary Data Table S2. In most patients (n = 197, 80.4%), onset of SI was after STN-DBS (median 1.26 years).
One or more parameter changes aiming at improving SI have been performed in 140 of 245 subjects, resulting in a marked permanent improvement in 39 subjects (27.9%). Altogether, 227 parameter changes for SI have been performed. These are summarized in Table 1, except for parameter changes performed in no more than five patients, that is, directional stimulation (n = 5), interleaving stimulation (n = 3), SPW (n = 2), and on-demand speech program (n = 1). For the latter parameter changes, only on-demand speech program induced a permanent marked improvement of SI. Amplitude alterations were the most frequent parameter changes. Permanent marked improvement of SI was most likely with amplitude reduction (22/83-26.5%), bipolar stimulation (3/13-23.1%), and change in stimulation contact (7/ 36-19.4%). LFS induced a permanent marked improvement in only two of 20 subjects (10.0%). Worsening of SI was most frequent with amplitude increase (14/64-21.9%) and change in stimulation contact (7/36-19.4%).
When considering the onset of SI, increasing the amplitude was more likely to worsen SI in subjects with onset of SI after (13/ 43-30.2%) than in those with onset of SI before (1/21-4.8%) DBS (p = 0.021; Table 1). Permanent marked improvement was more likely when changing the stimulation contact if onset of SI was after (7/30-23.4%) than if it was before (0/6-0.0%) DBS (p < 0.001). Similarly, changing the stimulation contact more frequently worsened speech if onset of SI was before (3/6-50%) than if it was after (4/30-13.3%) DBS (p < 0.001).
Assessing the relation between timing of parameter change and effectiveness, increasing the amplitude was ineffective when performed more than five years after STN-DBS (Fig. 1). Potential effectiveness of decreasing the amplitude, however, was maintained, even at more than ten years after STN-DBS (Fig. 1). LFS, alternating stimulation contact and bipolar stimulation, also seemed to remain potentially effective, although the small number of adjustments warrants caution with interpreting these results.
A formal clinical assessment of a stimulation-induced origin of SI (ie, assessment in DBS on vs DBS off) has been performed in 31 subjects. In 22 subjects (71.0%), SI was not stimulation-induced. In four subjects (12.9%), SI was entirely stimulation induced. In these four patients, speech was markedly improved by decreasing amplitude. Of note, in one patient, this was implemented as an ondemand speech program, and in another, there was some return of SI four years later. In five subjects (16.1%), the origin of SI was mixed (ie, partially stimulation induced and partially constituting a PD symptom). In only one of these patients could a permanent marked improvement of SI be obtained (by changing the stimulation contact); other parameter changes had no effect.
In 11 subjects, lead localization was determined (partly) because of the presence of SI and revealed suboptimal lead placement in six subjects. In five of these six subjects, but none of the five subjects without suboptimal lead placement, lack of DBS effectiveness (ie, insufficient effect on motor fluctuations) accompanied the SI. Surgical repositioning was performed in three of these subjects; however, marked permanent improvement of SI was obtained in none of the patients, despite an effective repositioning (ie, beneficial effect on motor fluctuations). In the other three subjects, only a permanent unilateral stimulation discontinuation in one subject resulted in a permanent marked improvement of speech.
Assessing the persistence of effect of parameter changes, 61 of 227 parameter changes (26.9%) caused a transient (either some or marked) improvement. This transient nature mostly concerned a spontaneous return of SI, that is, a loss of effect without any additional parameter change (Table 2). Oftentimes, however, the transient nature was due to the parameter change being reverted either because of stimulation-induced side effects (increased amplitude-2/14) or because of worsening of PD symptoms (decreased amplitude-8/23, LFS-3/6, and bipolar stimulation-4/6).

Systematic Review
The search strategy yielded 662 records, 23 of which were included. The flowchart regarding identification, screening, and inclusion of records is depicted in Supplementary Data Figure S1. Detailed study characteristics are summarized in Table 3. In general, the number of records investigating a particular intervention was limited, with only frequency modulation and alternating amplitude being assessed in more than five records. Only two records concerned randomized controlled trials (RCTs), assessing SPW and IL-IL, respectively. 27,30 Both had a double-blind cross-over design, limited study duration (four and two weeks), and small study population (n = 16 and 25). The RCT assessing SPW showed no effect on speech, whereas that assessing IL-IL showed an improvement of SI on the group level (effect size not mentioned). Fourteen records concerned within-subjects studies. SI was defined as being stimulation-induced in only four records. 24,27,28,31 In six of the 23 studies, patients already experienced speech impairment preoperatively-as judged by the preoperative MDS-UPDRS-II/III speech items being >0. [9][10][11]13,14,22 In the remaining studies, the onset of speech impairment has not been assessed/described.
Group effect sizes were numerically reported in only three records but mostly comprised a qualitative description. 10,11,15 Considerable heterogeneity in intervention modality and outcome measures precluded the conduct of a meta-analysis and the determination of concatenated effect sizes. The definition of SI was mostly not specified regarding origin (eg, stimulation-induced, PD symptom, mixed), severity, and characteristics, causing significant heterogeneity among study populations. Furthermore, the definition of "low" and "high" in units for the different parameters SPEECH IMPAIRMENT IN STN-DBS PD PATIENTS www.neuromodulationjournal.org diverged notably between studies. Outcome measures for assessing speech also varied considerably, and clinically meaningful measures were often lacking. Frequency modulation was studied most often (n = 14), with a positive effect of LFS on SI reported in nine records. Changing the stimulation amplitude was assessed in six records, of which four reported statistically significant speech improvement with decreasing amplitude. SPW showed improvement in three of four records. Evidence regarding stimulation of an alternative contact, IL-IL, directional stimulation, and interleaving stimulation was scarce. At the individual level, responder rates were reported in only a few records (n = 8); altogether, these were 46 of 117 for LFS (39.3%), 14 of 20 for SPW (70%), 6 of 17 for IL-IL (35.3%), and 2 of 13 for directional stimulation (15.4%).  Results are not displayed for directional stimulation, SPW, interleaving stimulation, and speech program on demand because the number of patients did not exceed n = 5. Proportions are indicated between brackets and are relative to the number of patients who underwent the respective programming option stratified regarding onset. Permanent improvement is defined as "improvement still present at last follow-up visit." Not performed in our retrospective cohort: interleaving-interlink. p Values concern chi-square test comparing onset before DBS with onset after DBS. † Chi-square test was used to assess the difference of the DBS adjustment effect between SI onset before and after DBS. Most records (n = 14) assessed speech outcomes only in the acute setting (defined as less than one day after applying the intervention). Only five records performed longitudinal speech assessments, with the second assessment ranging from two weeks to more than one year after initial assessment. 13,14,17,22,[27][28][29] Speech improvement was transient in one record and sustained in two. 17,22,28 Risk of bias was evaluated to be low for only two records and was unclear or high for all other records. 9,27 Effect of DBS parameter change on general PD symptom control has been reported in approximately half of the records (n = 13).
Only one of these 13 records, assessing decreasing amplitude, reported decreased PD symptom control. 23

DISCUSSION
The high prevalence of bothersome SI (almost 40%) in our retrospective study supports the relevance and importance of SI management in STN-DBS for PD. The relatively small proportion of subjects experiencing a persistent significant improvement of SI

High
Unclear effect size comprises the following reasons: inconsistent results, assessing different combinations of stimulation parameters, and not explicitly describing or mentioning the effect (size) of the intervention. Within-subjects study design refers to a study design in which every subject receives all stimulation conditions (eg, every study subject is exposed to all examined frequencies). cHFS, (conventional) high frequency stimulation; DA, decreasing amplitude; DIR, directional stimulation; IA, increasing amplitude; ILS, interleaving stimulation; Improvement nos , improvement not otherwise specified (improvement reported by the authors, not substantiated by numeric effect sizes); IQR, interquartile range; IR, inconsistent results; NA, not applicable; PW, pulse width; SID, stimulation-induced dysarthria. *Unless specified otherwise (eg, IQR, range). **Proportion of patients with objective or subjective improvement. ***Median. † Age/PD duration at implant. † † Overlap in patient cohort. † † † Pearson's r.   [2] speech characteristics regarding phonation, articulation, and prosody are needed for baseline assessment and to guide speech therapy, preferably performed by a speech therapist; [3] primarily based on speech characterization, limited stimulation amplitude titration, history, patient files, and monopolar review; if unclear, speech assessment with DBS enabled and 15 to 30 minutes after stimulation discontinuation might be useful; titration of stimulation (eg, in 0.5 mA increments) to identify the therapeutic window regarding dysarthria might be useful, especially in cases with PD-related dysarthria with DBS off and stimulation-induced dysarthria with DBS on; [4] for example, insufficient motor effectiveness, narrow therapeutic window on monopolar review; [5] new MRI (if IPG allows) or fusion of (perioperative or new) CT with DBS parameter adjustments (<30%) indicates its refractoriness to DBS adjustments but also the need for a better and more systematic approach to managing SI in STN-DBS for PD. This responder rate is lower than that deducted from literature (ie, 42.1%), probably owing to issues related to trial designs (eg, too short a follow-up to assess transient nature of effect, and exclusively in-hospital assessments). Responder rate was highest for decreasing amplitude, bipolar stimulation, and changing the stimulation contact; hence, these may be the parameter adjustments with the highest probability of effect. Conversely, probability of SI worsening was highest with increasing amplitude and changing the stimulation contact. Effective DBS adjustments often needed to be reversed owing to worsening of other PD symptoms; hence, counseling patients regarding possible PD symptom deterioration upon parameter change is recommended. Patient characteristics may guide the selection of parameter adjustments to a limited extent. Timing of SI onset and timing of DBS adjustment seem to be related with adjustment effectiveness. This observation probably comes down to the important distinction between SI constituting a PD symptom or a stimulation-induced side effect. Although performed only rarely in our cohort, a stimulation on-off assessment evaluating speech may be useful in this regard. Although our retrospective study provides only limited data, SI seems unlikely to be caused by suboptimal lead placement if a DBS lead otherwise effectively controls PD symptoms. Hence, SI alone should probably not suffice to consider lead repositioning.
The systematic review identified a large body of evidence supporting the notion that decreasing amplitude and LFS can improve SI, but this has not been assessed in an RCT. SPW has been shown to improve SI in a few studies, but this was not corroborated in the single RCT assessing SPW. IL-IL is the only parameter adjustment whose beneficial effect on SI has been indicated in an RCT, but confirmation is lacking. Regarding other parameter adjustments, studies are scarce. In general, most studies lack a powerful design, exhibit a very short study duration, and were appraised as containing an unclear or high risk of bias. Moreover, interpretation of study results is compromised by considerable methodologic heterogeneity at the levels of study population, intervention, and outcome measures. The effect of parameter changes on general PD symptom control has been reported in just more than half of studies.
On the basis of the systematic review and retrospective study, we propose a systematic approach to managing SI in STN-DBS for PD (Fig. 2).
There are several limitations to the retrospective study. Parameter adjustments for SI have not been performed at all in 105 of 245 subjects with SI. This illustrates that for a considerable proportion of patients, SI probably is considered not severe enough to justify parameter adjustments but also may have introduced a certain level of selection bias. Owing to the retrospective chart-based approach, assessment of bothersome SI presence was qualitative and not standardized. Characterization of SI (regarding articulation, phonation, and prosody) has not been performed. Similarly, effect size assessments (eg, "marked improvement") of DBS adjustments were subjective, qualitative, and not standardized. In addition, duration of effect was not assessed in a standardized and systematic manner; hence, "permanent" improvement is dependent on the follow-up duration. Several parameter adjustments have been performed very rarely (eg, SPW and directional stimulation) or even not at all (ie, IL-IL), precluding assessment of their performance. In line with this, the current data mostly reflect the use of quadripolar electrodes and provide only limited insights into the (directional) use of octopolar electrodes. When performing LFS, it is generally recommended to keep the total electrical energy delivered constant to maintain PD symptom control. 8 However, since this was mostly not implemented, the performance of LFS may have been underestimated in our retrospective study. The systematic review also contained several potential limitations. Only statistically significant results were considered. However, statistical significance and clinical relevance are not interchangeable. Regarding the effect on motor performance, the outcome measures chosen were MDS-UPDRS part II-III-IV scores because these are universal outcome measures used to assess motor performance in PD. Some studies, however, used different assessment methods, which now have escaped our systematic review.
To increase the scientific evidence supporting DBS adjustments for SI in STN-DBS for PD, future clinical trials should preferably have an RCT design with long-term (at least three months) follow-up. The onset, nature (ie, stimulation-induced, PD symptom, or mixed), and characteristics (ie, disturbance of phonation, articulation, or prosody) of SI should be actively determined, clearly defined, and duly reported. Next to standardized and patientrelevant speech outcome measures, assessment of general PD symptoms also should be performed. Because of the pleiotropy of DBS adjustments for SI, head-to-head RCT comparisons of all possible parameter changes are unfeasible. Therefore, large prospective cohorts constitute an alternative approach to assess performance of DBS adjustments for SI. Prospective validation of the programming algorithm proposed here is required because it is currently unclear whether implementation of this algorithm might lead to a higher probability of sustained clinically significant SI improvement. Collaboration with speech pathologists in both the determination of SI nature/origin and standardized assessments to evaluate the outcome of parameter adjustments can be helpful in further improving the performance of the algorithm. Voice recognition and analysis software might assist further. Technologic advances like adaptive DBS and variable frequency stimulation might overcome current programming hurdles and might be a better solution for SI. 32,33 A better understanding of the pathological underpinnings of and anatomical structures/networks related to SI will, however, be pivotal in moving the field forward. Current anatomical lead localization and volume of tissue activated estimation may already provide some guidance when changing the stimulation contact or using directional stimulation. However, more anatomical and functional refinements like 7T magnetic resonance imaging and functional connectivity analyses may allow pinpointing of optimal patient and symptom-specific anatomical stimulation locations. These advancements will be crucial in propelling SI management in STN-DBS for PD.
images with preoperative MRI; [6] too distant from dorsolateral sensorimotor STN; [7] surgical risk and uncertainty regarding SI improvement need to be considered; [8] probability of sustained marked improvement of SI, as determined in our retrospective study; [9] can be based on monopolar review; [10] at this stage, refraining from further DBS adjustments can be considered given the lack of evidence and high burden for patient and physician; [11] one or all can be assessed in isolation or combined with other programming options; [12] if IPG and/or leads allow; [13] only in clinical use in Japan; [14] can be considered earlier, depending on speech characteristics. MRI, magnetic resonance imaging; CT, computed tomography; IPG, implantable pulse generator.

SPEECH IMPAIRMENT IN STN-DBS PD PATIENTS
www.neuromodulationjournal.org