Background: Spinal cord stimulation (SCS) has been investigated as a potential therapeutic option for managing refractory symptoms in patients with Parkinson’s disease (PD). Objective: This systematic review and meta-analysis aimed to evaluate the safety and efficacy of SCS in PD. Method: A comprehensive literature search was conducted on PubMed and Web of Science to identify SCS studies reporting Unified Parkinson Disease Rating Scale-III (UPDRS-III) or Visual Analogue Scale (VAS) score changes in PD cohorts with at least 3 patients and a follow-up period of at least 1 month. Treatment effect was measured as the mean change in outcome scores and analyzed using an inverse variance random-effects model. The risk of bias was assessed using the Newcastle-Ottawa Scale and funnel plots. Results: A total of 11 studies comprising 76 patients were included. Nine studies involving 72 patients reported an estimated decrease of 4.43 points (95% confidence interval [CI]: 2.11; 6.75, p < 0.01) in UPDRS-III score, equivalent to a 14% reduction. The axial subscores in 48 patients decreased by 2.35 points (95% CI: 1.26; 3.45, p < 0.01, 20% reduction). The pooled effect size of five studies on back and leg pain VAS scores was calculated as 4.38 (95% CI: 2.67; 6.09, p < 0.001), equivalent to a 59% reduction. Conclusions: Our analysis suggests that SCS may provide significant motor and pain benefits for patients with PD, although the results should be interpreted with caution due to several potential limitations including study heterogeneity, open-label designs, small sample sizes, and the possibility of publication bias. Further research using larger sample sizes and placebo-/sham-controlled designs is needed to confirm effectiveness.

Parkinson’s disease (PD) is a neurodegenerative disorder that can be treated with pharmacological treatments and deep brain stimulation (DBS) [1, 2]. Despite the implementation of various therapeutic interventions, the progression of PD persists, and symptoms such as gait abnormalities, speech impairment, pain, and cognitive dysfunction remain largely refractory to treatment [3‒5]. In recent years, epidural spinal cord stimulation (SCS) has been investigated as a potential therapeutic option for managing refractory symptoms of PD [6‒16].

The concept of using SCS to alleviate the motor symptoms of PD arose from rodent studies, which hypothesized that stimulating afferent somatic pathways could disrupt synchronization of aberrant low-frequency neural oscillations frequently seen in patients with PD [17]. However, the promising preclinical findings were not replicated in an early human trial, which failed to demonstrate improvement in locomotion [18]. Despite the negative results, subsequent studies specifically examining motor outcomes have been published [10, 11, 13‒15]. The effectiveness of SCS in PD remains uncertain due to conflicting results. Some studies have reported significant improvements in motor function and quality of life [10, 13], while others have found minimal or no benefit [14, 16]. To provide a more comprehensive evaluation of the potential benefits and risks of SCS in PD, we conducted a meta-analysis of published studies with the aim to inform clinical decision-making and guide future research in this field.

Protocol and Registration

This study was performed in accordance with the standards established in the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement [19]. The research protocol has been registered and made available in the PROSPERO database (341,626, https://www.crd.york.ac.uk/prospero).

Study Design, Search Strategy, and Selection Criteria

A thorough search for relevant literature was conducted in two electronic databases, PubMed and Web of Science, on April 6, 2022. The complete search parameters can be found in the online supplementary material (for all online suppl. material, see https://doi.org/10.1159/000531089). To identify additional relevant studies, the bibliographies of relevant articles were also reviewed. All references were imported into Covidence, an online systematic review management tool (www.covidence.org). Two authors (Z.S. and A.U.) independently screened the titles and abstracts for relevance, and any conflicts were resolved by a third author (C.S.). The full texts of the relevant articles were then reviewed to determine their eligibility according to the inclusion and exclusion criteria presented in online supplementary Table 1. The data were extracted by two authors (C.S. and O.Y.) independently. The data from the last follow-up visit were used when the same patient cohort was presented in multiple studies, ensuring that no patients were erroneously included in the analysis through double counting.

The extracted variables included bibliographic data, the year the surgeries were performed, the study design, the number of patients, the follow-up period, demographic variables (age and gender), clinical presentation (disease duration, preoperative DBS status, indication for SCS surgery), preoperative and postoperative clinical outcome scores (levodopa equivalent daily dose, Unified Parkinson Disease Rating Scale [UPDRS], Visual Analogue Scale or Numeric Rating Scale [VAS/NRS], Hoehn and Yahr Scale, freezing of gait questionnaire [FOG-Q], timed up and go [TUG] and stand-walk-sit [SWS] tests, the Parkinson’s Disease Questionnaire – 39 and 8 [PDQ-39 and PDQ-8], quantitative gait analysis), medication and DBS status during testing, lead type and location, stimulation type and parameters, system manufacturer, and complications.

The primary outcome of the meta-analysis was the mean difference in UPDRS-III scores relative to baseline at the last follow-up. The secondary outcomes were the mean difference between baseline and last follow-up for (1) VAS/NRS; (2) UPDRS-II; (3) UPDRS-III axial subscores (sum of the items: arising from chair, posture, gait, postural stability, body bradykinesia), OFF-medication status, paddle-lead cohorts, and no-low back pain cohorts; (4) FOG-Q; (5) TUG and SWS; (6) PDQ-39 and PDQ-8; (7) gait analysis parameters; and (8) the complication profile.

Risk of Bias Assessment

The risk of bias was assessed using the Newcastle-Ottawa Scale for single-arm observational studies, with adjustments made by removing the questions related to the nonexposed cohort and comparability [20]. The publication bias was evaluated using funnel plots and Egger’s regression test to assess asymmetry. Heterogeneity was assessed using the χ2 test and the I2 measure of inconsistency. When the I2 value was greater than 50%, outlier and influence diagnostic procedures were performed on the random-effect model. Outliers were identified by comparing the pooled effect confidence interval (CI) to the individual study CIs. Influential studies were identified using the leave-one-out method, which involves recalculating the results of the meta-analysis by leaving out one study at a time.

Data Synthesis and Statistical Analysis

Descriptive statistics were reported as the mean ± standard deviation or median and range for continuous variables and the number of cases with valid percentage for categorical variables. When the original study only reported the median and range, the mean and standard deviation were estimated using appropriate methods [21]. When mean scores were not reported, the mean and standard deviation were calculated using individual patient data extracted from the manuscripts or using the figures provided in the studies. The treatment effect was quantified as the mean change in outcome scores and analyzed using an inverse variance random-effect model based on the mean difference reported in each study. Statistical analyses were performed using R Studio (2022.02.3 + 492, “Prairie Trillium”; https://www.rstudio.com) with “meta” and “metafor” packages. p values of < 0.05 were considered statistically significant.

Study Selection and Characteristics of the Included Studies and Patients

We found a total of 761 and 1,176 studies in the PubMed and Web of Science databases, respectively. After removing duplicate entries, 1,492 studies were evaluated for relevance. After initial abstract screening, 49 full-text studies were assessed for eligibility, and 11 were included in the meta-analysis (online suppl. Fig. 1). Characteristics of the eligible studies are given in Tables 1 and 2. The study included a total of 72 patients with PD and 4 patients with atypical parkinsonism. The mean age of the patients ranged from 60 to 74 years, and the mean duration of the disease ranged from 9 to 21 years. The average length of follow-up in the studies was between 2 and 36 months. Of the patients, 19 (25%) received a cervical lead, 5 (7%) received an upper thoracic lead, and 52 (68%) received a lower thoracic lead. Burst stimulation was tested in three studies, and stimulation intensity was calibrated to range from 60 to 70% of the paresthesia threshold to elicit stimulation without paresthesia [7, 9, 11]. Weighted mean baseline scores were 31.6 ± 10 for UPDRS-III, 11.6 ± 1.7 for UPDRS-III axial subscores, and 7.4 ± 1.7 for VAS/NRS while Med-ON/SCS-ON. The Abbott system (n = 4) was the most commonly used, followed by Medtronic (n = 3) and Boston Scientific (n = 2). A minimum of 23 (30%) of the patients underwent DBS treatment prior to receiving SCS treatment. No complications related to surgical procedure, hardware, or stimulation were reported (nstudy = 7, npatient = 46) [9‒15] except a transient postoperative delirium in 1 patient [14].

Table 1.

Patient and study characteristics

First author, journal, yearCountry of originEnrollment periodStudy designCT.gov identifiern (patient)n (male)Age, mean±SD [range], yearsDisease duration, mean±SD [range], yearsn (patients with DBS)Primary indication for surgeryDBS status during testingMed. status during testingF/U, mean [range], monthsHoehn Yahr stage, mean±SD or median [range]Levodopa equivalent daily dose, mean±SD, mg
preoppostoppreoppostop
Agari, Neurol Med Chir [6], 2012 Japan NA RO NA 15 71.1 [63–79] 17.2 [7–31] LBP ON ON 12 [12–12] 3 [3–4] NA NA NA 
Nishioka, Neuromodulation [12], 2015 Japan NA PO NA 74.3 [67–80] 9.3 [5–13] LBP NA ON? 12 [12–12] 4 [4–5] 4 [4–4] 882±77 1,113±241 
Pinto de Souza, Mov Disord, 2017 & de Lima-Pardini, eLife, 2018 [8,13] Brazil NA OLT 02388204 64.2±5.9 21.2±10.1 ON OFF 6 [6–6] 3±0.7 7.8±4.1 813±470 763±423 
Samotus, Mov Disord, 2018 &Brain Stimul, 2020 [15,16] Canada 2016 OLT 03079310 5/4* 5/4 71±10 14±4 0? NA ON 36 [36–36] NA NA 1,330±441 1,215±420 
Mazzone, Brain Sci [11], 2019 Italy 2009–2017 RO NA 6/12 6/10 71.1±7.3/65.6±11.1 17.1±6.1/11.1±5.3 0?/3? LBP/G NA ON 12 [12–12] 3.6±0.6 2.6±0.9 1,334±472/835±310 1,083±263/730±263 
HubschParkinsonism Relat Disord [10], 2019 France NA OLT 02381951 68.8±3.9 14.8±7 NA ON and OFF 2 [2–2] ** NA NA NA NA 
Chakravarthy, Biolectron Med [7], 2020 USA, Japan NA PO NA 7/8 NA 74±5.2 17±8.7 0/8 LBP NA/ON ON 22 [4–33] 4.3±0.5/3.8±0.5 4.4±0.5/3.9±0.7 NA NA 
Furusawa, Parkinsonism Relat. Disord [9], 2020 Japan 2018 RO NA 74 [66–81] 12.4 [5–31] 0? LBP NA ON? 6 [6–6] 3 [2–4] NA NA NA 
Prasad, Mov Disord [14], 2020 Canada NA OLT NA 59.6±15.8 15.1±2.3 NA ON and OFF 12 [12–12] 4.1±1.1 4.2±1.3 NA NA 
First author, journal, yearCountry of originEnrollment periodStudy designCT.gov identifiern (patient)n (male)Age, mean±SD [range], yearsDisease duration, mean±SD [range], yearsn (patients with DBS)Primary indication for surgeryDBS status during testingMed. status during testingF/U, mean [range], monthsHoehn Yahr stage, mean±SD or median [range]Levodopa equivalent daily dose, mean±SD, mg
preoppostoppreoppostop
Agari, Neurol Med Chir [6], 2012 Japan NA RO NA 15 71.1 [63–79] 17.2 [7–31] LBP ON ON 12 [12–12] 3 [3–4] NA NA NA 
Nishioka, Neuromodulation [12], 2015 Japan NA PO NA 74.3 [67–80] 9.3 [5–13] LBP NA ON? 12 [12–12] 4 [4–5] 4 [4–4] 882±77 1,113±241 
Pinto de Souza, Mov Disord, 2017 & de Lima-Pardini, eLife, 2018 [8,13] Brazil NA OLT 02388204 64.2±5.9 21.2±10.1 ON OFF 6 [6–6] 3±0.7 7.8±4.1 813±470 763±423 
Samotus, Mov Disord, 2018 &Brain Stimul, 2020 [15,16] Canada 2016 OLT 03079310 5/4* 5/4 71±10 14±4 0? NA ON 36 [36–36] NA NA 1,330±441 1,215±420 
Mazzone, Brain Sci [11], 2019 Italy 2009–2017 RO NA 6/12 6/10 71.1±7.3/65.6±11.1 17.1±6.1/11.1±5.3 0?/3? LBP/G NA ON 12 [12–12] 3.6±0.6 2.6±0.9 1,334±472/835±310 1,083±263/730±263 
HubschParkinsonism Relat Disord [10], 2019 France NA OLT 02381951 68.8±3.9 14.8±7 NA ON and OFF 2 [2–2] ** NA NA NA NA 
Chakravarthy, Biolectron Med [7], 2020 USA, Japan NA PO NA 7/8 NA 74±5.2 17±8.7 0/8 LBP NA/ON ON 22 [4–33] 4.3±0.5/3.8±0.5 4.4±0.5/3.9±0.7 NA NA 
Furusawa, Parkinsonism Relat. Disord [9], 2020 Japan 2018 RO NA 74 [66–81] 12.4 [5–31] 0? LBP NA ON? 6 [6–6] 3 [2–4] NA NA NA 
Prasad, Mov Disord [14], 2020 Canada NA OLT NA 59.6±15.8 15.1±2.3 NA ON and OFF 12 [12–12] 4.1±1.1 4.2±1.3 NA NA 

The country of origin refers to the country where the first and last authors are affiliated.

Mazzone et al. and Chakravarthy et al. presented data from two separate cohorts in a single article, with the information for each cohort being demarcated by a “/” symbol.

DBS and medication status during testing refer to whether the patient was on medication or active DBS during the assessment of the primary outcome.

CT.gov identifier, clinicaltrials.gov identifier number; DBS, deep brain stimulation; F/U, follow-up; G, gait dysfunction; LBP, low back pain; med, medication; n, number of; NA, not available or non-applicable; OLT, open-label trial; PO/RO, prospective/retrospective observational cohort studies; preop/postop, preoperative and postoperative; SD, standard deviation.

*Samotus et al. included 5 patients in their first study and 4 patients in their last study.

**Satisfaction scales and stimulation parameters at postoperative month 29 were reported in the study.

?The authors of this meta-analysis estimated this information based on the study design, as it was not explicitly stated in the manuscript.

Table 2.

Hardware and stimulation characteristics

First author, journal, yearLead typeLevel of lead tipStimulation typeFrequency, HzAmplitude, VPulse width, µsSystem manufacturer
minmaxminmaxminmax
Agari et al. [6] 2012 2 linear leads T7-12 Cont-T 20 210 330 NA 
Nishioka et al. [12] 2015 Paddle lead T8-L1 Cont-T 65 0.6 5.8 360 450 Medtronic 
Pinto de Souza, [13] 2017 Paddle lead T2-4 Cont-T 300* 300 Mean±SD4.6±1.9 (standing)2.0±0.5 (supine) 90 90 Medtronic 
Samotus, Mov Disord et al. [15] 2018 2 linear T8-10 Cont-T 30 130 2.4 mA 11.2 mA 300 400 Boston Scientific 
Mazzone et al. [11] 2019 1 linear lead C1-3 G1: Cont-TG2: Cont-B G1: 130G2: 40–250 G1: 185G2: 40–500 G1: 1.3VG2: 0.2 mA G1: 4VG2: 0.9 mA G1: 60G2: 1,000 G1: 210G2: 1,000 Medtronic/Abbott 
Hubsch et al. [10] 2019 1 linear lead T10-11 Cont-T 100 200 At the threshold of paresthesia 150 300 Abbott 
Chakravarthy et al. [7] 2020 1 or 2 linear leads C2-5 (n = 1)T2-4 (n = 1)T6-12 (n = 13) Cycl-B or cont-B or cont-T Tonic: 10Burst: 40–500 Tonic: 40Burst: 40–500 Tonic: 2.6 mA Burst: 0.15 mA Tonic: 4.5 mA Burst: 1.45 mA Tonic: 350Burst: 1,000 Tonic: 350Burst: 1,000 Abbott 
Furusawa et al. [9] 2020 1 linear lead T8-9 Cont-B 40–500 40–500 At 60% of paresthesia level 1,000 1,000 Abbott 
Prasad et al. [14] 2020 2 linear leads T10 Cont-T 50 130 2.7 mA 15.9 mA 50 450 Boston Scientific 
First author, journal, yearLead typeLevel of lead tipStimulation typeFrequency, HzAmplitude, VPulse width, µsSystem manufacturer
minmaxminmaxminmax
Agari et al. [6] 2012 2 linear leads T7-12 Cont-T 20 210 330 NA 
Nishioka et al. [12] 2015 Paddle lead T8-L1 Cont-T 65 0.6 5.8 360 450 Medtronic 
Pinto de Souza, [13] 2017 Paddle lead T2-4 Cont-T 300* 300 Mean±SD4.6±1.9 (standing)2.0±0.5 (supine) 90 90 Medtronic 
Samotus, Mov Disord et al. [15] 2018 2 linear T8-10 Cont-T 30 130 2.4 mA 11.2 mA 300 400 Boston Scientific 
Mazzone et al. [11] 2019 1 linear lead C1-3 G1: Cont-TG2: Cont-B G1: 130G2: 40–250 G1: 185G2: 40–500 G1: 1.3VG2: 0.2 mA G1: 4VG2: 0.9 mA G1: 60G2: 1,000 G1: 210G2: 1,000 Medtronic/Abbott 
Hubsch et al. [10] 2019 1 linear lead T10-11 Cont-T 100 200 At the threshold of paresthesia 150 300 Abbott 
Chakravarthy et al. [7] 2020 1 or 2 linear leads C2-5 (n = 1)T2-4 (n = 1)T6-12 (n = 13) Cycl-B or cont-B or cont-T Tonic: 10Burst: 40–500 Tonic: 40Burst: 40–500 Tonic: 2.6 mA Burst: 0.15 mA Tonic: 4.5 mA Burst: 1.45 mA Tonic: 350Burst: 1,000 Tonic: 350Burst: 1,000 Abbott 
Furusawa et al. [9] 2020 1 linear lead T8-9 Cont-B 40–500 40–500 At 60% of paresthesia level 1,000 1,000 Abbott 
Prasad et al. [14] 2020 2 linear leads T10 Cont-T 50 130 2.7 mA 15.9 mA 50 450 Boston Scientific 

The stimulation parameters reported were those in place at the final follow-up visit.

The stimulation types were (1) continuous tonic (cont-T), (2) continuous burst (cont-B), and (3) cycling burst (cycl-B).

The leads were at (1) thoracal (T), (2) cervical (C), and (3) lumbar (L) levels.

The frequency of burst stimulation was reported using the “number-number” format, with the first number representing the burst frequency and the second number representing the intraburst frequency.

µs, microseconds; G1/2, groups 1 and 2; Hz, hertz; mA, milliampere; max, maximum; min, minimum; NA, non-applicable or not available; SD, standard deviation; V, volts.

*60 Hz tested as a separate experiment.

Clinical Outcomes

Effects on UPDRS

Nine studies [6, 7, 9‒14, 16] involving 72 patients reported an estimated decrease of 4.43 points (95% CI: 2.11; 6.75, p < 0.01, random-effects model, equivalent to a 14% [6.6–21.3%] reduction) in UPDRS-III score at the latest follow-up compared to baseline (medication-ON) in the stimulation-ON/medication-ON state (Fig. 1). No significant heterogeneity was detected among studies (Q = 8.3, degrees of freedom [df] = 10, p = 0.59, I2 = 0% [95% CI: 0–60%]). We performed several subgroup analyses. First subgroup analysis revealed an estimated effect size of 4.52 points (95% CI: 1.35; 7.68, p < 0.05) for tonic stimulation and 4.2 points (95% CI: −2.53; 10.92, p = 0.14) for burst stimulation (Fig. 1). Another subgroup analysis demonstrated an estimated effect size of 12.63 points (95% CI: 7.63; 17.64, p < 0.05) for paddle-lead cohorts and 3.65 points (95% CI: 1.86; 5.44, p = 0.01) for linear-lead (1 or 2 leads) cohorts (online suppl. Fig. 2). Last subgroup analysis, which exclusively focused on cohorts treated primarily for motor symptoms and not for low back pain, showed an estimated effect size of 5.55 points (95% CI: −0.2; 11.31, p = 0.06) (online suppl. Fig. 3a). In contrast, the estimated effect size was 4.15 points (95% CI: 0.73; 7.58, p < 0.05) for the cohorts treated primarily for low back pain. Axial subscores of UPDRS-III reported by 5 studies for 48 patients decreased by 2.35 points (95% CI: 1.26; 3.45, p < 0.01, equivalent to a 20% [10.8–29.7%] reduction) (online suppl. Fig. 3b). Three studies [10, 13, 14] reported medication-OFF status results with an estimated mean difference of 7.12 points (95% CI: −9.64; 23.87, p = 0.20, random-effects model) (online suppl. Fig. 3c). UPDRS-II scores (SCS-ON, Med-ON) were only reported by two studies [6, 14] (mean difference = 3.18 points [95% CI: −5.74; 12.11], p = 0.13, random-effects model) (online suppl. Fig. 3d).

Fig. 1.

Effect of tonic or burst stimulation on UPDRS-III score. Tonic stimulation in 43 patients from seven cohorts resulted in an estimated reduction of 4.52 points on the UPDRS-III score (95% CI: 1.35; 7.68, p < 0.05), and burst stimulation in 29 patients from four cohorts resulted in an estimated reduction of 4.2 points (95% CI: −2.53; 10.92, p = 0.14) when comparing the latest follow-up to baseline in the stimulation-ON/medication-ON state. Overall, 72 patients from eleven cohorts experienced an estimated decrease of 4.43 points (95% CI: 2.11; 6.75, p < 0.01) on the UPDRS-III score, equivalent to a 14% (6.6–21.3%) reduction. The exception was a report by Pinto de Souza et al. that used medication-OFF results for comparison. Total refers to the total number of patients; MD: mean difference (positive values indicate favorable outcomes); CI: confidence interval; df: degrees of freedom.

Fig. 1.

Effect of tonic or burst stimulation on UPDRS-III score. Tonic stimulation in 43 patients from seven cohorts resulted in an estimated reduction of 4.52 points on the UPDRS-III score (95% CI: 1.35; 7.68, p < 0.05), and burst stimulation in 29 patients from four cohorts resulted in an estimated reduction of 4.2 points (95% CI: −2.53; 10.92, p = 0.14) when comparing the latest follow-up to baseline in the stimulation-ON/medication-ON state. Overall, 72 patients from eleven cohorts experienced an estimated decrease of 4.43 points (95% CI: 2.11; 6.75, p < 0.01) on the UPDRS-III score, equivalent to a 14% (6.6–21.3%) reduction. The exception was a report by Pinto de Souza et al. that used medication-OFF results for comparison. Total refers to the total number of patients; MD: mean difference (positive values indicate favorable outcomes); CI: confidence interval; df: degrees of freedom.

Close modal

Effects on Pain

Five studies reported 56 patients with low back or leg pain by using VAS/NRS. The pooled effect size of SCS on VAS/NRS scores was calculated as 4.38 (95% CI: 2.67; 6.09, p < 0.001, random-effects model, equivalent to a 59% [36–82%] reduction) (Fig. 2a) [6, 7, 9, 11, 12]. Heterogeneity test revealed excess variation in the data (Q = 36.57, df = 6, p <0.001) with an I2 value of 84% (95% CI: 68–92%). Outlier identification algorithm did not detect any outlier study in random-effects model. Leave-one-out method identified the study by Agari et al. [6] as an influential study (influence effect size = 16.02) (suppl. Fig. 4a). Omitting this study from the pooled analysis resulted in a VAS/NRS score mean difference of 3.89 (95% CI: 2.29; 5.49, p < 0.01, random-effects model, n = 41) and decreased the I2 value to 49% (95% CI: 0–80%, Q = 9.87, p = 0.07) (online suppl. Fig. 4b).

Fig. 2.

Effect of stimulation on: pain (VAS/NRS), freezing of gait (FOG-Q score), mobility (TUG and SWS scores), and quality of life (PDQ-39 or PDQ-8 scores). a In 56 patients with low back or leg pain, the VAS/NRS was reduced by an estimated 4.38 (95% CI: 2.67; 6.09, p < 0.001), equivalent to a 59% (36–82%) reduction. Heterogeneity testing revealed excess variation in the data (Q = 36.57, df = 6, p <0.001) with an I2 value of 84% (95% CI: 68–92%). b The FOG-Q scores were reduced by an estimated 4.6 points in 14 patients across three studies (95% CI: −9.07; 18.28, p = 0.28) with significant heterogeneity among studies (Q = 27.92, df = 2, p <0.001, I2 = 93% [95% CI: 82–97%]). Data from Hubsch et al., which did not provide exact values, were excluded from the analysis. c The TUG and SWS scores were reduced by a standardized mean difference of 0.46 (95% CI: −0.33; 1.26, p = 0.19) in 37 patients. In the study by Chakravarthy et al., group 1 consisted of patients receiving continuous burst stimulation, while group 2 received cycling burst stimulation. Hubsch et al. used the SWS score, while other studies used the TUG score. d Quality of life, as assessed by PDQ-39 or PDQ-8, was reduced by an estimated 0.49 (95% CI: −0.29; 1.27, p = 0.13) in 19 patients across four studies. All assessments were conducted in the medication-ON state, except for the study by Pinto de Souza et al., which reported results in the medication-OFF state. Total refers to the total number of patients; (S)MD: (standardized) mean difference (positive values indicate favorable outcomes); CI: confidence interval; df: degrees of freedom.

Fig. 2.

Effect of stimulation on: pain (VAS/NRS), freezing of gait (FOG-Q score), mobility (TUG and SWS scores), and quality of life (PDQ-39 or PDQ-8 scores). a In 56 patients with low back or leg pain, the VAS/NRS was reduced by an estimated 4.38 (95% CI: 2.67; 6.09, p < 0.001), equivalent to a 59% (36–82%) reduction. Heterogeneity testing revealed excess variation in the data (Q = 36.57, df = 6, p <0.001) with an I2 value of 84% (95% CI: 68–92%). b The FOG-Q scores were reduced by an estimated 4.6 points in 14 patients across three studies (95% CI: −9.07; 18.28, p = 0.28) with significant heterogeneity among studies (Q = 27.92, df = 2, p <0.001, I2 = 93% [95% CI: 82–97%]). Data from Hubsch et al., which did not provide exact values, were excluded from the analysis. c The TUG and SWS scores were reduced by a standardized mean difference of 0.46 (95% CI: −0.33; 1.26, p = 0.19) in 37 patients. In the study by Chakravarthy et al., group 1 consisted of patients receiving continuous burst stimulation, while group 2 received cycling burst stimulation. Hubsch et al. used the SWS score, while other studies used the TUG score. d Quality of life, as assessed by PDQ-39 or PDQ-8, was reduced by an estimated 0.49 (95% CI: −0.29; 1.27, p = 0.13) in 19 patients across four studies. All assessments were conducted in the medication-ON state, except for the study by Pinto de Souza et al., which reported results in the medication-OFF state. Total refers to the total number of patients; (S)MD: (standardized) mean difference (positive values indicate favorable outcomes); CI: confidence interval; df: degrees of freedom.

Close modal

Effects on Gait and Balance

Three studies provided information on changes in FOG-Q scores in 14 patients. Mean difference was 4.6 points (95% CI: −9.07; 18.28, p = 0.28, random-effects model), with significant heterogeneity among studies (Q = 27.92, df = 2, p < 0.001, I2 = 93% [95% CI: 82–97%]) (Fig. 2b). Five studies assessed gait by TUG and SWS tests in 37 patients and revealed a standardized mean difference of 0.46 (95% CI: −0.33; 1.26, p = 0.19, random-effects model) with low-to-moderate heterogeneity (Q = 6.7, df = 5, p = 0.24, I2 = 25.4% [95% CI: 0–68.7%]) (Fig. 2c). A quantitative gait analysis was reported by several studies [11, 13‒15]; however, a pooled estimate was not performed due to the substantial variability between the test methods such as distinct walking distance (6 m vs. 20 m), distinct control groups (comparison between age- and sex-matched control groups vs. comparison of baseline and follow-up scores within the same group), or medication status during testing (ON vs. OFF).

Effects on Quality of Life

Quality of life was assessed by PDQ-39 [10, 13, 14] or PDQ-8 [16] in 4 studies with 19 patients. Pooled estimate standardized mean difference was 0.49 (95% CI: −0.29; 1.27, p = 0.13, random-effects model) (Fig. 2d).

Risk of Bias Assessment

The studies included in this analysis had a low risk of bias, as indicated by their Newcastle-Ottawa Quality Assessment Scale scores, which ranged from 5 to 6 stars out of a maximum of 6 stars (online suppl. Table 2). Analysis of the funnel plots and Egger’s regression test showed significant asymmetry in the VAS/NRS score (intercept: −3.42 95% CI: −5.87; −0.97, t = −2.7, p <0.05) but not in the UPDRS-III score (intercept: 0.18 95% CI: −0.85; 1.23, t = 0.3, p = 0.73) (online suppl. Fig. 5a, b). The findings may indicate publication bias in the reporting of VAS/NRS scores.

Our meta-analysis of eleven studies, which included a total of 76 patients, showed that SCS was associated with improvements in motor function and pain as measured by UPDRS-III and VAS/NRS scores, respectively. The treatment appeared to be safe, as no permanent irreversible neurological deficits due to insertion were reported for any of the patients. The pooled estimates for quality of life and gait tests did not reach statistical significance.

Our pooled data showed that UPDRS-III scores improved significantly by 4.43 points (−14%) overall. When focusing on just the axial subscores, the improvement was even more substantial at 2.35 points (−20%). However, when compared to other neuromodulation techniques like subthalamic DBS (which has been shown to improve UPDRS-III scores by 50.5% in a recent meta-analysis of 39 studies with 2,035 subjects [22]), the effect size of SCS on motor symptoms of PD can be considered very small. Moreover, the accuracy of these estimates in reflecting true effect size may be substantially impacted by various characteristics of the included studies. The mechanisms by which SCS improves motor symptoms are not yet fully understood, but it is possible that the relief of low back and leg pain may contribute to the improvement of motor function. To test this hypothesis, we performed a subgroup analysis of studies that included patients without pain. While the effect size was found to be higher in this subgroup (5.55 points), it did not reach statistical significance (p = 0.06), suggesting that pain relief may confound the relationship between SCS and motor improvement. It is worth noting that all of the included studies are open-label and do not include placebo or sham groups. Additionally, the initial disease severity of the patients and duration of the follow-up varies across studies (see Table 2). Three studies did not have a follow-up period of at least 1 year, and habituation in long term has been reported by some studies [11, 16]. The lack of blinding during assessments, the lack of long-term follow-up in some studies, and the potential for distinct SCS effect sizes between stages of the disease may introduce bias in the estimation of the true effect size.

The effect of SCS on supraspinal sensory networks and subsequent reduction in synchronized pathological low-frequency neural oscillations were demonstrated in both preclinical [17, 23] and clinical studies [24]. In addition to the proposed supraspinal mechanisms, spinal networks may also contribute to the locomotive effects of SCS. In patients with spinal cord injury, prolonged low levels of muscle activity can lead to the conversion of some motor units to the faster types, resulting in a mismatch between muscle recruitment and the functional demands of a motor task. SCS can activate interneuronal networks that functionally project to populations of motor neurons and muscle unit phenotypes that meet the work demands [25]. The location of the lead within spinal epidural space, as well as the extent and characteristics of the stimulated neurons may substantially influence the ability of SCS to facilitate the intrinsic automaticity of spinal networks. Epidural SCS at specific spatiotemporal resolution may have potential as a treatment option for patients with PD in the future, similar to its application in individuals with spinal cord injuries [26]. In this meta-analysis, we were unable to conduct a subgroup analysis based on the location of the leads due to the fact that all leads were placed in the lower thoracic region, except for one study that implanted the leads in the cervical region [11], another in the upper thoracic region [13], and one study in which leads were implanted in both regions [7]. However, a subgroup analysis based on lead type revealed that using paddle leads had a relatively large effect size of 12.63 points (95% CI: 7.63–17.64) in UPDRS-III scores. A paddle lead may result in a larger stimulation field compared to the linear leads, although it is important to note that the size of the stimulation field is dynamic in SCS and influenced by multiple factors including stimulation intensity, the number of active contacts, the position of the patient, and the distance between lead and spinal cord. It is also noteworthy that the extent of the stimulation field and the resultant paresthesia area may have a positive correlation with the magnitude of the placebo effect.

One of the most contentious issues in the field is the determining the optimal stimulation parameters and type (e.g., burst/tonic or continuous/cycling) for promoting motor improvement. According to a pioneering rodent study, 300 Hz stimulation had the most pronounced effect on locomotion, compared to lower frequency tonic stimulations [17]. A subsequent preclinical study in monkeys also showed significant motor improvement, with all tested frequencies (ranging from 4 to 300 Hz) being equally effective [23]. The finding that 300 Hz stimulation resulted in a more significant improvement in gait parameters compared to 60 Hz stimulation was subsequently replicated in 4 patients with PD using a double-blind design, with pulse width held constant at 90 μs while the frequency was varied [8, 13]. However, no subsequent human studies have tested 300 Hz stimulation. Similar efficacious gait improvements were also demonstrated in another study, with follow-up at 6 months, by using a combination of high pulse width (300–400 μs) and lower frequency (30–130 Hz). However, this effect was lost at 36-month follow-up [15, 16]. Our pilot trial did not find significant motor improvements at 12 months when using low frequencies (50–130 Hz) and low pulse widths [14]. Several studies used burst stimulation (continuous or cycling) rather than tonic stimulation, mostly in patients with accompanying low back pain [7, 9, 11]. When the results of these studies were pooled, there was no significant effect on UPDRS-III scores.

Pain in PD patients is multifaceted and can be caused by musculoskeletal and neuropathic radicular pain, which is often exacerbated by abnormal pain processing due to disease-specific neurodegenerative changes in the pain neuroaxis or by disease-related motor symptoms such as dystonia [27]. Our meta-analysis showed a significant improvement of 4.3 points (59%) in VAS/NRS scores in PD patients with low back pain. The reliability of this estimate had to be thoroughly investigated due to several factors, such as (1) significant between-study heterogeneity and the presence of influential studies, (2) study designs, and (3) the potential for publication bias. Herein, the influential study detection algorithms identified an influential study that decreased the effect size to 3.8 and reduced the heterogeneity from substantial to low-to-moderate levels when it was omitted from the analysis. Another factor that may have led to an overestimation of true effect size is the observational nature of the included studies and the lack of placebo (or sham) control groups. The placebo effect in SCS-neuropathic pain studies is well-documented, as shown by a recent meta-analysis, which found the mean VAS/NRS score change between active stimulation and placebo-/sham-controlled periods was only 1.15 on a 10-point scale [28]. Furthermore, the funnel plots indicate that there may be publication bias, with negative studies being underpublished.

This study also analyzed other outcome measures, such as gait, balance, and quality of life tests, but the limited number of studies that reported these measures made it difficult to draw clear conclusions about their outcomes. It is important to note that the absence of permanent complications in these patient cohorts may give clinicians the misleading impression that the procedure is completely safe. The number of patients in the included studies may not be sufficient to detect all types of complications, and the overall complication rates associated with SCS systems in the literature are not insignificant [29]. Additionally, the financial cost of SCS treatment for PD should be considered. Replacing well-established DBS treatments with experimental SCS treatments may not be a favorable and cost-effective option unless the DBS treatment has initially failed or is contraindicated. However, using both DBS and SCS synergistically may be a viable option. This study demonstrated that 30% of the patients in the included studies received implants for both DBS and SCS systems. Currently, such a treatment approach requires at least two implantable pulse generators, one for SCS and one for DBS [30]. In the future, it may be possible to design a single implantable pulse generator with at least four channels to control both DBS and cervical SCS systems simultaneously in PD patients with significant gait impairments. Such a design would also enable the recording of supraspinal effects of SCS through local field potential recordings and may offer mechanistic insights [24]. An even further advance would be the implementation of an adaptive SCS system utilizing sensory input from wearables or brain-computer interfaces and high spatiotemporal resolution SCS, as recently developed to be used in paralytic primates [31]. In future trials, it is essential to demonstrate the true effect size of SCS by employing a sham-controlled design that uses paresthesia-free stimulation parameters in larger and more homogenous cohorts, thereby allowing the size of the placebo effect to be determined. Subsequently, optimizing the stimulation parameters and varying the level of SCS can be explored.

Our analysis suggests that SCS may provide significant motor and pain benefits for patients with PD, although this should be interpreted with caution due to several potential limitations including study heterogeneity, open-label designs, small sample sizes, and the possibility of publication bias. Further research using larger sample sizes and standardized, placebo-/sham-controlled designs as well as further investigation into optimal parameters for SCS is needed to gain a deeper understanding of the role of SCS in the management of PD.

An ethics statement is not applicable because this study is based exclusively on published literature.

O.Y., A.C.Y., A.U., Z.S., B.S., N.S., M.C., A.V., J.G., C.C., M.S., R.J., G.D., K.Y., D.H.A.‐P., J.N., and R.C. report no disclosures relevant to the manuscript. C.S. has been receiving fellowship grants from the Michael and Amira Dan Foundation and Turkish Neurosurgical Society. A.Z. was supported by the Henan Provincial People’s Hospital Outstanding Talents Founding Grant Project. A.M.L. is scientific director for Functional Neuromodulation and a consultant to Medtronic, Abbott, Boston Scientific, Insightec, and the Focused Ultrasound Foundation. A.F. reports the following: consultancies from AbbVie, Medtronic, Boston Scientific, Sunovion, Chiesi Farmaceutici, UCB, and Ipsen; membership in advisory boards of AbbVie, Boston Scientific, and Ipsen; receiving honoraria from AbbVie, Medtronic, Boston Scientific, Sunovion, Chiesi Farmaceutici, UCB, and Ipsen; and receiving grants from the University of Toronto, Weston Foundation, AbbVie, Medtronic, and Boston Scientific. S.K.K. receives honoraria, consulting, and/or speaker fees from Abbott, Boston Scientific, inBrain, Medtronic, Novo Nordisk, Parkinson Canada, and Movement Disorders Society; and research support from Parkinson Canada, CIHR, MJFF, FUS Foundation, MitoO2, Toronto Western Hospital Foundation, Weston Foundation, and RR Tasker Chair in Stereotactic and Functional Neurosurgery.

No funding was received for this study.

Conception of the work: Can Sarica and Ajmal Zemmar; acquisition of the data: Can Sarica, Omid Yousefi, Ayse Uzuner, Zhiyuan Sheng, and Mohammadmahdi Sabahi; data interpretation: Can Sarica, Ajmal Zemmar, Artur Vetkas, Cletus Cheuyo, Ghazaleh Darmani, Kazuaki Yamamoto, David H. Aguirre-Padilla, Joseph Neimat, Robert Chen, Suneil K. Kalia, Alfonso Fasano, and Andres M. Lozano; statistical analysis: Can Sarica and Jurgen Germann; writing the first draft: Can Sarica, Ajmal Zemmar, Andrew C. Yang, Brendan Santyr, Nardin Samuel, Michael Colditz, Raja Jani, and Ghazaleh Darmani; and critical revision and approval of the manuscript: all authors.

Additional Information

Can Sarica and Ajmal Zemmar equally share first author contribution.

All data generated or analyzed during this study are included in this article. Further inquiries can be directed to the corresponding author.

1.
Poewe
W
,
Seppi
K
,
Tanner
CM
,
Halliday
GM
,
Brundin
P
,
Volkmann
J
.
Parkinson disease
.
Nat Rev Dis Primers
.
2017 Dec
3
1
17013
.
2.
Paff
M
,
Loh
A
,
Sarica
C
,
Lozano
AM
,
Fasano
A
.
Update on current technologies for deep brain stimulation in Parkinson’s disease
.
J Mod Dynam
.
2020 Sep
13
3
185
98
.
3.
Chen
J
,
Zhong
Y
,
Wei
H
,
Chen
S
,
Su
Z
,
Liu
L
.
Polyethylene glycol recombinant human growth hormone in Chinese prepubertal slow-growing short children: doses reported in a multicenter real-world study
.
BMC Endocr Disord
.
2022
;
22
(
1
):
201
27
.
4.
Rodriguez-Oroz
MC
,
Moro
E
,
Krack
P
.
Long-term outcomes of surgical therapies for Parkinson’s disease: long-term surgical therapy outcomes in PD
.
Mov Disord
.
2012 Dec
27
14
1718
28
.
5.
Flouty
O
,
Yamamoto
K
,
Germann
J
,
Harmsen
IE
,
Jung
HH
,
Cheyuo
C
.
Idiopathic Parkinson’s disease and chronic pain in the era of deep brain stimulation: a systematic review and meta-analysis
.
J Neurosurg
.
2022 Apr
137
6
1821
30
.
6.
Agari
T
,
Date
I
.
Spinal cord stimulation for the treatment of abnormal posture and gait disorder in patients with Parkinson’s disease
.
Neurol Med Chir
.
2012
;
52
(
7
):
470
4
.
7.
Chakravarthy
KV
,
Chaturvedi
R
,
Agari
T
,
Iwamuro
H
,
Reddy
R
,
Matsui
A
.
Single arm prospective multicenter case series on the use of burst stimulation to improve pain and motor symptoms in Parkinson’s disease
.
Bioelectron Med
.
2020 Dec
6
1
18
.
8.
de Lima-Pardini
AC
,
Coelho
DB
,
Souza
CP
,
Souza
CO
,
Ghilardi
MG
,
Garcia
T
.
Effects of spinal cord stimulation on postural control in Parkinson’s disease patients with freezing of gait
.
Elife
.
2018 Aug
7
e37727
.
9.
Furusawa
Y
,
Matsui
A
,
Kobayashi-Noami
K
,
Kojima
Y
,
Tsubouchi
A
,
Todoroki
D
.
Burst spinal cord stimulation for pain and motor function in Parkinson’s disease: a case series
.
Clin Park Relat Disord
.
2020
;
3
:
100043
.
10.
Hubsch
C
,
D’Hardemare
V
,
Ben Maacha
M
,
Ziegler
M
,
Patte-Karsenti
N
,
Thiebaut
JB
.
Tonic spinal cord stimulation as therapeutic option in Parkinson disease with axial symptoms: effects on walking and quality of life
.
Parkinsonism Relat Disord
.
2019 Jun
63
235
7
.
11.
Mazzone
P
,
Viselli
F
,
Ferraina
S
,
Giamundo
M
,
Marano
M
,
Paoloni
M
.
High cervical spinal cord stimulation: a one year follow-up study on motor and non-motor functions in Parkinson’s disease
.
Brain Sci
.
2019 Apr
9
4
78
.
12.
Nishioka
K
,
Nakajima
M
.
Beneficial therapeutic effects of spinal cord stimulation in advanced cases of Parkinson’s disease with intractable chronic pain: a case series
.
Neuromodulation
.
2015 Dec
18
8
751
3
.
13.
Pinto de Souza
C
,
Hamani
C
,
Oliveira Souza
C
,
Lopez Contreras
WO
,
dos Santos Ghilardi
MG
,
Cury
RG
.
Spinal cord stimulation improves gait in patients with Parkinson’s disease previously treated with deep brain stimulation: spinal Cord Stimulation in PD
.
Mov Disord
.
2017 Feb
32
2
278
82
.
14.
Prasad
S
,
Aguirre-Padilla
DH
,
Poon
Y
,
Kalsi-Ryan
S
,
Lozano
AM
,
Fasano
A
.
Spinal cord stimulation for very advanced Parkinson’s disease: a 1-year prospective trial
.
Mov Disord
.
2020 Jun
35
6
1082
3
.
15.
Samotus
O
,
Parrent
A
,
Jog
M
.
Spinal cord stimulation therapy for gait dysfunction in advanced Parkinson’s disease patients: spinal cord stimulation for gait in PD
.
Mov Disord
.
2018 May
33
5
783
92
.
16.
Samotus
O
,
Parrent
A
,
Jog
M
.
Long-term update of the effect of spinal cord stimulation in advanced Parkinson’s disease patients
.
Brain Stimul
.
2020 Sep
13
5
1196
7
.
17.
Fuentes
R
,
Petersson
P
,
Siesser
WB
,
Caron
MG
,
Nicolelis
MAL
.
Spinal cord stimulation restores locomotion in animal models of Parkinson’s disease
.
Science
.
2009 Mar
323
5921
1578
82
.
18.
Thevathasan
W
,
Mazzone
P
,
Jha
A
,
Djamshidian
A
,
Dileone
M
,
Di Lazzaro
V
.
Spinal cord stimulation failed to relieve akinesia or restore locomotion in Parkinson disease
.
Neurology
.
2010 Apr
74
16
1325
7
.
19.
Moher
D
,
Liberati
A
,
Tetzlaff
J
,
Altman
DG
PRISMA Group
.
Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement
.
BMJ
.
2009 Jul
339
jul21 1
b2535
.
20.
Wells
GA
,
Shea
B
,
O’Connell
D
,
Peterson
J
,
Welch
V
,
Losos
M
The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses
Ottawa, Canada
Ottawa Health Research Institute
. Available from: http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp. 2011.
21.
Luo
D
,
Wan
X
,
Liu
J
,
Tong
T
.
Optimally estimating the sample mean from the sample size, median, mid-range, and/or mid-quartile range
.
Stat Methods Med Res
.
2018 Jun
27
6
1785
805
.
22.
Lachenmayer
ML
,
Mürset
M
,
Antih
N
,
Debove
I
,
Muellner
J
,
Bompart
M
.
Subthalamic and pallidal deep brain stimulation for Parkinson’s disease: meta-analysis of outcomes
.
NPJ Parkinsons Dis
.
2021 Sep
7
1
77
.
23.
Santana
MB
,
Halje
P
,
Simplício
H
,
Richter
U
,
Freire
MAM
,
Petersson
P
.
Spinal cord stimulation alleviates motor deficits in a primate model of Parkinson disease
.
Neuron
.
2014 Nov
84
4
716
22
.
24.
Darmani
G
,
Arora
T
,
Drummond
NM
,
Cortez Grippe
T
,
Saha
U
,
Munhoz
RP
.
Thalamocortical spectral and coherence characteristics for clinically effective and ineffective spinal cord stimulation in chronic pain: a case study
.
Clin Neurophysiol
.
2023 Feb
146
18
20
.
25.
Edgerton
VR
,
Gad
P
.
Spinal automaticity of movement control and its role in recovering function after spinal injury
.
Expert Rev Neurother
.
2022 Aug
22
8
655
67
.
26.
Rowald
A
,
Komi
S
,
Demesmaeker
R
,
Baaklini
E
,
Hernandez-Charpak
SD
,
Paoles
E
.
Activity-dependent spinal cord neuromodulation rapidly restores trunk and leg motor functions after complete paralysis
.
Nat Med
.
2022 Feb
28
2
260
71
.
27.
Beiske
AG
,
Loge
JH
,
Rønningen
A
,
Svensson
E
.
Pain in Parkinson’s disease: prevalence and characteristics
.
Pain
.
2009 Jan
141
1–2
173
7
.
28.
Duarte
RV
,
Nevitt
S
,
McNicol
E
,
Taylor
RS
,
Buchser
E
,
North
RB
.
Systematic review and meta-analysis of placebo/sham controlled randomised trials of spinal cord stimulation for neuropathic pain
.
Pain
.
2020 Jan
161
1
24
35
.
29.
Bendersky
D
,
Yampolsky
C
.
Is spinal cord stimulation safe? A review of its complications
.
World Neurosurg
.
2014 Dec
82
6
1359
68
.
30.
Sarica
C
,
Iorio-Morin
C
,
Aguirre-Padilla
DH
,
Najjar
A
,
Paff
M
,
Fomenko
A
.
Implantable pulse generators for deep brain stimulation: challenges, complications, and strategies for practicality and longevity
.
Front Hum Neurosci
.
2021 Aug
15
708481
.
31.
Lorach
H
,
Charvet
G
,
Bloch
J
,
Courtine
G
.
Brain–spine interfaces to reverse paralysis
.
Natl Sci Rev
.
2022 Sep
9
10
nwac009
.