Long-term effects of subthalamic nucleus deep brain stimulation on speech in Parkinson’s disease

Bilateral subthalamic nucleus deep brain stimulation (STN-DBS) is an effective treatment in advanced Parkinson’s Disease (PD). However, the effects of STN-DBS on speech are still debated, particularly in the long-term follow-up. The objective of this study was to evaluate the long-term effects of bilateral STN-DBS on speech in a cohort of advanced PD patients treated with bilateral STN-DBS. Each patient was assessed before surgery through a neurological evaluation and a perceptual-acoustic analysis of speech and re-assessed in the long-term in different stimulation and drug conditions. The primary outcome was the percentage change of speech intelligibility obtained by comparing the postoperative on-stimulation/off-medication condition with the preoperative off-medication condition. Twenty-five PD patients treated with bilateral STN-DBS with a 5-year follow-up were included. In the long-term, speech intelligibility stayed at the same level as preoperative values when compared with preoperative values. STN-DBS induced a significant acute improvement of speech intelligibility (p < 0.005) in the postoperative assessment when compared to the on-stimulation/off-medication and off-stimulation/off-medication conditions. These results highlight that STN-DBS may handle speech intelligibility even in the long-term.

Primary outcomes.In the long-term and at the group level, speech intelligibility did not significantly worsen with respect to preoperative values when comparing the postoperative on-stimulation/off-medication condition with the preoperative off-medication condition (z = − 0.371, p = 0.710).The shutdown of the stimulation led to an acute significant worsening of speech intelligibility in the postoperative assessment (z = − 3.500, p < 0.001) when comparing the on-stimulation/off-medication and off-stimulation/off-medication conditions.Furthermore, the assessment of disease progression effects on speech intelligibility obtained when comparing the postoperative off-stimulation/off-medication condition with the preoperative off-medication condition showed a significant worsening with respect to preoperative values (z = − 2.92, p < 0.005).Analyzing the longterm postoperative changes of speech intelligibility at the individual level for each patient, sixteen patients were classified as "stable" while the remaining nine patients integrated a "worsened" subgroup (Fig. 1).Compared to Table 1.Stimulation parameters at postoperative evaluation.off-stimulation/off-medication condition when compared with the on-stimulation/off-medication condition (p < 0.005) and with the preoperative on-medication condition (p < 0.005).In addition, the mean intensity of sustained phonation was significantly reduced in the off-stimulation/off-medication condition if compared with the preoperative off-medication condition (p < 0.005) highlighting the negative effects of disease progression on this acoustic parameter.
A direct correlation between the variation of speech intelligibility and tremor (p = 0.026) and PIGD subscores (p = 0.040) by comparing the off-stimulation/off-medication with the on-stimulation/off-medication one was also found, highlighting that the rebound of tremor and axial symptoms can worsen speech in the acute testing situation.On the contrary, no correlations were found between the variation of speech intelligibility and the changes in the different motor scores and subscores when comparing the on-stimulation/off-medication and preoperative off-medication condition.Furthermore, a negative correlation was also found in the on-stimulation/ off-medication condition between speech intelligibility and akinesia subscore (p = 0.025) meaning that patients with a more severely decreased movements amplitude and speed have less intelligibility.The negative correlation between speech intelligibility and akinesia subscore was also confirmed in the on-stimulation/on-medication condition (p = 0.013).

Discussion
Our results add to further observations regarding the long-term effects of bilateral STN-DBS on speech in PD patients 7,12 .Globally, the studied cohort showed a long-term maintenance of speech intelligibility after surgery, highlighting some possible beneficial long-term effects of STN-DBS when comparing off stimulation and on stimulation conditions.This is consistent with previous studies that showed no worsening of speech following STN-DBS when assessed either with the UPDRS speech item 15 or dedicated perceptual assessments 16 , which was reduced with the introduction of L-DOPA 17 , in line with our findings.Nevertheless, when using a proper perceptual assessment of speech intelligibility, it is commonly accepted that a significant worsening of speech intelligibility should be expected one year after surgery, when comparing postoperative on-stimulation/offmedication condition with the preoperative off-medication condition 9,18 .This has also been reported in the long-term by Aviles-Olmos et al. 12 , who reported a significant worsening of speech intelligibility at five (− 43.7%) and eight years (− 21.4%) in the off-medication conditions if compared with preoperative values.The different results obtained in our cohort may be due to the different evaluation of speech intelligibility (i.e., sentence task of the Assessment of Intelligibility for Dysarthric Speech and single word intelligibility).However, by looking at the individual cases of our cohort in detail, two distinct subgroups could be identified based on the long-term changes of speech intelligibility, with different clinical characteristics.This is in line with similar approaches that tried to disentangle different possible speech outcomes following STN-DBS 18,19 .Patients in the "worsened" subgroup showed a worse preoperative and postoperative motor disease severity together with a worse cognitive function particularly regarding executive-frontal domain.These results are in line with a recent 2-year follow-up  study that underlined the potential predictor role of preoperative cognitive function for speech deterioration after STN-DBS 19 .The trend toward the significance of PIGD subscores between the two "worsened" and "stable" subgroups suggests a possible link between speech deterioration and axial features, confirming the correlation between hypokinetic dysarthria and PD axial symptoms 8 .A similar tendency towards significance was also found for the over-distribution of GBA1-PD patients in the "worsened" group.GBA1-PD patients complain about higher axial and cognitive burden when compared with wild-type PD patients 20 leading to the assumption that GBA1-PD patients may be at higher risk of developing post-operatively not only gait and cognitive worsening but also speech deterioration.Interestingly, disease duration, sex, duration of follow-up and the total electrical energy delivered by STN-DBS 21 did not significantly differ between the two groups.This result is in line with a previous study that did not find any association between speech intelligibility deterioration and disease duration 7 even if another study by Tripoliti et al. 9 found that disease duration was a predictive factor for speech outcome.
In addition, concerning preoperative speech variables we did not find differences between the two groups, while postoperatively, both speech intelligibility and mean intensity of spontaneous speech were significantly lower in the worsened group if compared with the stable one.With regards to the comparison between the different speech variables, a statistically significant reduction of the intensity of sustained phonation emerged in the offstimulation/off-medication condition if compared with the on-stimulation/off-medication condition highlighting the possible positive effect of STN-DBS on voice intensity, as already confirmed in previous studies 7,18 .The improvement of speech intensity due to STN-DBS has been linked to its effects on hypokinesia and rigidity of language-related organs 18,22 which represent one of the major pathophysiological bases of hypokinetic dysarthria, characterized by hypophonia, monotony, hypoarticulation of consonants and inappropriate silences 23 .
A further result that emerged in our sample was the significant reduction of the maximum phonation time (MPT) in the off-stimulation/off-medication condition when compared to the on-stimulation/off-medication condition.This result shows how stimulation, even 5 years after surgery, can positively influence MPT that is significantly reduced following the device's shutdown.This finding is in line with previous studies that reported a beneficial effect of stimulation on MPT in the short-term after surgery 24 .Even in this case, the reduction of MPT can be traced back to the underlying laryngeal dysfunction and the reduced respiratory volume mainly related to the rigidity and hypo-bradykinesia of the laryngeal and diaphragmatic muscles 23 .
This study also allows for evaluation of the possible effects of dopaminergic therapy alone on speech parameters by comparing the preoperative off-medication and on-medication conditions.The results obtained did not show statistically significant differences between these two conditions, suggesting that dopaminergic therapy alone did not significantly improve or maintain speech postoperatively.These findings were consistent with previous studies that showed little or no effects of levodopa alone on acoustic speech parameters 8,[25][26][27] .In particular, no significant changes following levodopa intake were found in phonatory, articulatory or prosodic parameters, as previously reported 27 confirming the possible involvement of non-dopaminergic pathways in the pathophysiology of hypokinetic dysarthria in PD 8 .
We also found a direct correlation between the variation of speech intelligibility and tremor and PIGD subscores highlighting that the rebound of tremor and axial symptoms may worsen speech in the acute testing situation when stimulation is turned off.In addition, speech intelligibility also correlated with the UPDRS akinesia subscore in two out of the three postoperative conditions tested, highlighting a possible link between the decrease in limb movement speed and amplitude and speech.This is in line with other studies that have reported a correlation between speech variables and limb bradykinesia 28,29 .Our study has several limitations, including a lack of definition for the position of the electrodes, the small sample size, the lack of a control group and the lack of the assessment of speech intelligibility also at the sentence level.Nevertheless, our findings highlight the possibility of positive effects of STN-DBS on speech intelligibility after surgery in the long-term.Interestingly, it also gives a better understanding of PD characteristics associated with long-term speech worsening after STN-DBS.This information may allow clinicians to improve candidates' selection for DBS and refine prognostic accuracy (e.g., GBA1 genetic status influence on speech) and that, if necessary, early speech interventions should be used after surgery in PD patients treated with STN-DBS at a higher risk of speech deterioration.

Methods
Participants.Patients treated with bilateral STN-DBS from 2012 to 2017 at the Neurological Unit of the OCB University Hospital were included.All patients fulfilled the diagnosis of PD according to the UK Brain Bank criteria 30 and suffered from disabling motor complications.Data from non-native Italian speakers were excluded.This study was approved by the ethics committee of the Area Vasta Emilia Nord, Italy (protocol Clinical assessment.The clinical evaluation was performed in accordance with the CAPSIT-PD protocol 31 . Levodopa responsiveness was evaluated through an acute levodopa challenge.Hoehn and Yahr scale (H&Y) and the Unified Parkinson's Disease Rating Scale (UPDRS) 32 were applied to quantify disease severity both in the "off-medication" condition (obtained after a 12-h antiparkinsonian medication withdrawal) and "on-medication" condition (obtained after 60 min and the administration of a 30% higher dose of the usual levodopa morning intake) 31 .Different subscores were extrapolated from the UPDRS including tremor, postural instability/gait disorders (PIGD), akinesia and UPDRS item 18 (speech) subscores.PD motor phenotype (tremor dominant [TD], indeterminate and PIGD) 33 was also extrapolated and patients were screened for the presence of mutations in the glucocerebrosidase-1 (GBA-1), leucine-rich repeat kinase-2, α-synuclein and parkin genes 34,35 .The total amount of dopaminergic medications was calculated as levodopa equivalent daily dose (LEDD) milligrams 36 .
All patients underwent 3-Tesla brain-MRI to evaluate the presence/absence of white matter hyperintensities of vascular origin.All subjects were re-evaluated with a median five-years follow-up (range 3-7 years) after surgery.
Neurological evaluation (superimposable with the preoperative one) and the speech assessment were carried out on the same day and in the following conditions: on-stimulation/off-medication (washout of at least 12-h of dopaminergic medications); off-stimulation/off-medication (stimulation was temporarily turned off for at least 1-h); on-stimulation/on-medication (stimulation was turned on and dopaminergic therapy was administered [early morning LEDD plus 30%]).Each patient underwent a complete neuropsychological assessment before surgery and at long-term evaluation, including phonemic fluency, spatial perception (localization of numbers), Raven's progressive matrices, Stroop test and Trail making test part B.
Speech evaluation.Speech evaluation was performed both preoperatively during the acute levodopa challenge and at postoperative assessment.Each evaluation was carried out in a silent environment with conversation voice intensity and was recorded with a digital voice recorder (model SONY ICDPX240) kept 20 cm from the patient's lips 25 .The perceptual-acoustic analysis was performed using Praat software 37 and blinded to the patient's condition.The following tasks were included: word intelligibility (calculated as the percentage of words correctly transcribed by the examiner among a set of 25 recorded words) 25 ; oral diadochokinesis task in which participants produced the syllables /pa/, /ta/, /ka/ and the pseudoword /pataka/, as fast as they could with habitual pitch and loudness (irregular rhythm [presence of absence], uncontrolled acceleration [presence of absence]); sustained production of the phoneme /a/ for as long as possible, performed three times (duration [sec], intensity [dB]); counting from 1 to 20 (speech rate [syllables/second]).Single words intelligibility was selected due to its advantage of eliminating a number of other variables that can affect intelligibility, such as sentence level syntactic and prosodic variables.Furthermore, the use of single words to assess intelligibility is a much less difficult task for dysarthric participants than sentence level productions.As such, if an intelligibility impairment is noted at the single word level, intelligibility deficits are more than likely at higher/more complicated levels of speech productions, such as the sentence level 38 .A calibration tone (80 SPL dB, 1 kHz) was included at the beginning of each recording to serve as a reference in the determination of speaking amplitude.
According to recent guidelines 39 , these parameters have been selected because they represent acoustic characteristics previously reported as altered in hypokinetic dysarthria 23,27 .The presence and severity of hypokinetic dysarthria were perceptually determined by two speech language pathologists, both Italian native speakers.Speakers' severity of dysarthria was categorized on a coarse scale ranging from none, mild, moderate to severe (1: severe, 2: moderate; 3: mild: 4: none) 40,41 .

Statistical analysis.
Descriptive statistics were used for describing demographic and clinical data.The primary objective of the study was to evaluate the long-term effects of bilateral STN-DBS on speech intelligibility in advanced PD patients.As primary outcomes, we selected the change of speech intelligibility between postoperative on-stimulation/off-medication condition and preoperative off-medication condition; postoperative on-stimulation/off-medication and off-stimulation/off-medication condition.Positive changes represented improvement of speech intelligibility while negative changes represented speech worsening.The presence of significant differences in speech intelligibility in the different conditions tested was calculated using the Friedman test with subsequent post-hoc Wilcoxon signed rank test with Bonferroni correction for multiple comparisons because of the use of multiple tests (statistical adjusted significance was set at p-value < 0.005).Based on the presence/absence of long-term postoperative worsening of speech intelligibility, patients were divided into two groups ("stable/imp" [absent or positive variation] and "worsened" [negative variation]) that were compared to find significant differences in demographic, clinical and speech variables.With regards to continuous and ordinal variables, the Mann-Whitney test was used, while for categorical variables the chi-square independence test was applied (statistical significance was set at p-value < 0.05).Secondary outcomes included: the changes of the other speech variables in the different conditions tested; the correlation between the variation of speech intelligibility and the variation in the different motor scores and subscores in the on-stimulation/off-medication condition compared with the off-stimulation/off-medication condition and in the on-stimulation/off-medication compared with preoperative off-medication condition; the correlation between the different motor scores and subscores and speech intelligibility in the postoperative conditions tested.Correlation analyses were performed by using the Spearman Correlation analysis.Statistical analysis was performed using the IBM SPSS Statistics for Windows version 20.0 (IBM, Armonk, NY, USA).

Figure 1 .
Figure 1.Changes of speech intelligibility in the stable and worsened groups.

Table 3 .
Differences in clinical and speech variables between "worsened" and "stable" subgroups.PD Parkinson disease, LEDD Levodopa Equivalent Daily Dose, PIGD dominant postural instability and gait disorder, SD standard deviation, UPDRS Unified Parkinson's Disease Rating Scale.: 2019/0,056,629), and written informed consent was obtained from participants.The study was performed in accordance with the Declaration of Helsinki. number