Effects of neurostimulation for advanced Parkinson’s disease patients on motor symptoms: A multiple-treatments meta-analysas of randomized controlled trials

Deep brain stimulation (DBS) is the surgical procedure of choice for patients with advanced Parkinson disease (PD). We aim to evaluate the efficacy of GPi (globus pallidus internus), STN (subthalamic nucleus)-DBS and medical therapy for PD. We conducted a systematic review and multiple-treatments meta-analysis to investigate the efficacy of neurostimulation and medical therapy for PD patients. Sixteen eligible studies were included in this analysis. We pooled the whole data and found obvious difference between GPi-DBS versus medical therapy and STN-DBS versus medical therapy in terms of UPDRS scores (Unified Parkinson’s Disease Rating Scale). Meanwhile, we found GPi-DBS had the similar efficacy on the UPDRS scores when compared with STN-DBS. What is more, quality of life, measured by PDQ-39 (Parkinson’s disease Questionnaire) showed greater improvement after GPi-DBS than STN-DBS. Five studies showed STN-DBS was more effective for reduction in medication than GPi-DBS. Overall, either GPi-DBS or STN-DBS was an effective technique to control PD patients’ symptoms and improved their functionality and quality of life. Meanwhile, the UPDRS scores measuring parkinsonian symptoms revealed no significant difference between GPi-DBS and STN-DBS. STN-DBS was more effective for reduction in medication than GPi-DBS. Alternatively, GPi-DBS was more effective for improving the PDQ-39 score than STN-DBS.


Results
Study inclusion and basic characteristics of studies. We identified 561 references, from which we excluded 332 due to the duplicates. After screening the titles and abstracts, 131 were excluded because they failed to meet the inclusion criteria. By reading the full text of the remaining 98 articles, 58 studies were excluded because a result of without control groups, 14 were excluded because of not testing the effects of comparing STN-DBS with GPi-DBS or with medical therapy to treat idiopathic PD, 10 studies were eliminated due to no adequate data about UPDRS as the outcome measure. Ultimately, just leaving 16 qualified studies satisfied the pre-established inclusion criteria 10,[15][16][17][18][19][20][21][22][23][24][25][26][27][28][29] (Fig. 1).
Sixteen RCTs, with a total of 2186 participants, met the inclusion criteria and were included in this review. Studies were conducted in America (8/16, 50.0%), Italy (2/16, 12.5%), Germany (3/16, 18.6%), Switzerland (1/16, 6.3%), UK (1/16, 6.3%) and Netherlands (1/16, 6.3%), respectively. The mean age of those included studies was relatively elder at about 55 years, most were in Hoehn & Yahr stage II or more and there were slightly more males than females. The number of participants randomized into the 16   of follow up was ranged from 6 months to 3 years. In terms of outcome measures, UPDRS part I was observed in 6 studies; UPDRS part II (on phase or off phase) was observed in 10 or 6 studies, respectively; UPDRS part III (on phase or off phase) was reported in 16 and 9 studies, respectively; UPDRS part IV was showed in 7 studies. Moreover, PDQ-39 and Levodopa equivalent dose (LED, pre-and post-surgery) were reported as outcome measures in 9 and 11 studies, respectively. The basic characteristics of the 16 selected studies were summarized in Table 1. Table 2 showed the risk of bias of the included trials. All trials described the method of randomization used (e.g. random number table, computer generated). Fourteen trials gave information that allowed the assessment of whether an adequate concealment of allocation procedure was used. Twelve studies reported the blinding of participants. All trials described intention-to-treat analyses (ITT) and reported follow up data. Therefore, all of the included trials were deemed to have a low risk of bias.

Discussion
Main findings. In the present multiple-treatments meta-analyses, there were evidences to support the hypothesis of treatment difference between GPi-DBS versus medical therapy (p < 0.05) and STN-DBS versus medical therapy (p < 0.05) in the treatment of parkinsonian symptoms in PD patients in terms of UPDRS scores. Meanwhile, we found that GPi-DBS had the similar efficacy on the UPDRS scores when compared with STN-DBS, suggesting that GPi-DBS and STN-DBS improved the clinical symptoms of PD equally well. In addition, this study also confirmed DBS (including GPi-DBS and STN-DBS) could significantly reduce PDQ-39 score in these subjects compared with medical therapy. However, improvement in PDQ-39 was greater after GPi-DBS than STN-DBS. In addition to improve the motor function and quality of life, the effect size of DBS was evident for  the reduction of the required L-dopa medication dose compared with medical therapy. Consistent with previous results, STN-DBS allowed L-dopa medication dosages to be reduced to lower levels than GPi-DBS. Overall, the results from this paper showed that either GPi-DBS or STN-DBS was an effective technique to control patients' symptoms and improve their functionality and quality of life. Nevertheless, the UPDRS scores measuring parkinsonian symptoms revealed no significant difference between GPi-DBS and STN-DBS. These results were essentially identical to those described in previous review articles. Perestelo-Pérez et al. 14 reported DBS (mainly STN-DBS, six RCTs) significantly improved patients' symptoms, functionality and quality of life compared with best medical treatments. The effects sizes were intense for the reduction of motor signs and improvement of functionality in the off-medication phase. Moderate effects were observed in the case of motor signs and time in good functionality in the on-medication phase. In this meta-analysis, DBS (eight RCTs) of both targets (GPi and STN-DBS) was overall better than medical therapy in terms of UPDRS scores, PDQ-39 and LED with more participants. Liu Y et al. 30 indicated that GPi and STN-DBS improved motor function for PD patients (six RCTs). Differences in therapeutic efficacy for PD were not observed between the two procedures. STN-DBS allowed greater reduction in medication for patients, whereas GPi-DBS provided greater relief from psychiatric symptoms. Similarly, we also showed GPi-DBS had the equal efficacy on the UPDRS scores when compared with STN-DBS with eight RCTs. PDQ-39 improvements were greater after GPi-DBS than STN-DBS, and STN-DBS allowed L-dopa medication dosages to be reduced to lower levels than GPi-DBS. In addition, we carried out a comprehensive overview of all RCTs that compared GPi-DBS, STN-DBS and medical therapy in one paper to provide a clinically useful summary of the results of the multiple-treatments meta-analyses that can be used to guide treatment decisions.
Interpretation of the results. DBS is a well-established modality for the treatment of advanced PD. Recent studies have found DBS plus best medical therapy to be superior to best medical therapy alone for patients with PD and early motor complications, but its mechanism of action has been elusive 31 . To date, different mechanisms have been proposed to explain the therapeutic actions of DBS. Several studies in both experimental and clinical found that DBS could increase the output from the STN, as evidenced by the increased activity in the GPi, the downstream nucleus of the STN 32 . Therefore, a decoupling of the activities between the soma and the axon at the stimulation target may occur. In addition, recently, an important role of the motor cortex is becoming clear, and the emerging pictures have important implications for both the pathogenic process and treatment strategies for PD 5 . Moreover, site of surgery is a consideration when a surgical option is being considered. Our studies found no differences between GPi-DBS and STN-DBS in terms of improvement of UPDRS scores in line with previous meta-analyses 30 . STN-DBS allowed greater reduction in LED for patients, whereas GPi-DBS provided greater relief from quality of life (PDQ-39). To our knowledge, several studies of GPi-DBS versus STN-DBS also showed equivalent motor outcomes in a short-term follow up 33 . Concern has been expressed regarding the long-term durability of neurostimulation for PD. Meanwhile, the differences maybe due to the populations that were included, or differences related to the pharmacologic treatment of patients are treated. What is more, DBS is supposed to affect the firing rates and bursting patterns of neurons and, ultimately, the synchronized oscillatory activity of neuronal networks 34 . Both DBS targets can form such networks, which could explain why STN and GPi equally improved motor function. Increased understanding of the mechanism of DBS and the pathophysiology of the basal ganglia are needed to explain this issue.
Adverse effects of DBS consist of a wide variety of neurological and neuropsychological complications such as those related to surgery, hardware and stimulation. A recent extensive review suggests that the major surgery-related risk is intracranial hemorrhage and the overall incidence of hemorrhage was 5.0%, with symptomatic hemorrhage occurring in 2.1% of patients and hemorrhage resulting in permanent neurological deficit or death in 1.1% 35 . The most common hardware complications include infections, electrode migrations or misplacements, skin erosion, wire fractures and device malfunction, and the rate ranging from 4.3 to 17.8% between different studies. Stimulation-related adverse effects include muscle contractions, dysarthria, tremor, dyskinesia, headache and pain 36 . In general, DBS is a relatively safe approach associated with an encouragingly low rate of adverse effects. In addition, since DBS is expensive, cost must be considered when determining risk versus benefit 37 . A potential advantage reported in patients receiving DBS and medical therapy was significant reduction in medication use 20 . This would reduce both potential adverse effects and cost of the drugs used. What is more, long-term benefit of DBS must also be considered. Several studies have indicated that DBS does not alter the progression of PD. All in all, DBS is safe and soundness for advanced PD patients.
Implication from this research. A recent report had detailed patient characteristics that were likely to predict a positive outcome to DBS. These characteristics included responsiveness to L-dopa, younger age, and no cognitive dysfunction or psychiatric diseases et al. 38 . Namely, preoperative L-dopa responsiveness was found to be predictive of average improvement in UPDRS scores following DBS surgery. As such, though it is definite that the relationship between preoperative L-dopa responsiveness and surgical outcome exists, the actual magnitude may be obscured 39 . In addition, longer disease duration was correlated with a higher UPDRS scores at baseline 40 . Consequently, there were also higher absolute UPDRS scores following surgery that were expected. Patients receiving DBS and best medical therapy were significant reduction in medication use had been validated

. Effect sizes of GPi-DBS vs STN-DBS for (A) UPDRS Part I (on phase); (B) UPDRS Part II (on phase); (C) UPDRS part III (on phase); (D) UPDRS part III (off phase) and (E) UPDRS Part IV (on phase).
by a multicenter study that evaluated the economic cost in PD subjects who underwent STN-DBS 41 . As efficacy of DBS is widely established, debate regarding when in the course of disease DBS surgery should be performed has arisen. The common approach is to provide surgery to patients when L-dopa has failed and all other options have been exhausted. Further studies would be required to evaluate when the optimum time window for DBS surgery to achieve maximum efficacy. Moreover, site of surgery is a consideration when a surgical option is being considered. Based on this paper, DBS can be employed to stimulate either STN or GPi in terms of improve UPDRS scores. The differences of PDQ39 and LED, if validated by other studies, could help determine the best site for DBS stimulation based on individual patient characteristics. One possibility reasons for this difference between the effects of DBS in GPi and STN over time is that the difference in doses of antiparkinsonian medications associated with STN and GPi stimulation is responsible 7 . The higher relative levels of dopaminergic medication taken by patients after GPi-DBS may enhance quality of life by improving aspects of PDQ-39 score. Thus, future study designed for research need to select suitable target for surgery according to the individual subject characteristics. Finally, stimulation-related side effects result mostly from diffusion of current into neighboring anatomic structures. In general, these side effects can be avoided by optimizing electrode placement. Hence, this needs to  be carefully balanced in the future studies to ascertain the narrow window between the threshold for therapeutic effect and the appearance of side effects.
Limitations. This paper has some potential weaknesses. Significant heterogeneity was observed in this analysis. A potential explanation may be that the measurements in each study were different. There were 3 studies included in the current study that contained larger numbers of patients (n > 200) than the other studies, which may create certain bias. Finally, we only included studies published in English, which may also create potential bias.

Conclusion
Based on current available information, either GPi-DBS or STN-DBS plus medical therapy was superior to best medical therapy alone in terms of reducing UPDRS scores and improving quality of life and decreased medication requirements. Meanwhile, our results showed no differences between GPi-DBS and STN-DBS in either of our primary outcomes (UPDRS scores). STN-DBS was more effective for reduction in medication than GPi-DBS. Alternatively, GPi-DBS was more effective for improving the PDQ-39 score than STN-DBS.

Methods
We undertook a systematic review and multiple-treatments meta-analysis based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement 42 . Meanwhile, to this study, we have no predefined protocol about the process and objective of this meta-analysis publication or available to the research community as an online appendix.
Study selection and data collection. For our analysis, we only included RCTs that compared GPi-DBS versus STN-DBS, or GPi versus medical therapy, or STN-DBS versus medical therapy in the advanced PD patients. To identify the relevant studies, we electronically searched databases of PubMed, Google scholar and Cochrane Central Register of Controlled Trials (CENTRAL) electronic databases up to June 2015 for all English language publications. The search terms were "Parkinson" and "stimulation" to include more comprehensive literatures. Reference lists from the resulting publications and reviews were used to identify further relevant publications. The following inclusion criteria were used: 1) randomized, controlled, properly concealed, clinical trials, comparing STN-DBS with GPi-DBS or with medical therapy to treat idiopathic PD; 2) studies describing advanced PD patients with certain degree response motor fluctuations, dyskinesia, painful dystonia, or bradykinesia despite optimal pharmacological treatment; 3) studies that used the UPDRS to measure the baseline disease and post-treatment results; 4) reports in which outcomes were measurable continuous variables; 5) reports that were published in English. Studies and patient populations were excluded for the following reasons: 1) not RCTs studies; 2) no adequate data about UPDRS as the outcome measure; 3) DBS was performed in pathologies other than PD; 4) duplicate publications; or 5) data could not be extracted. For each study, information was carefully extracted from all the eligible studies, including (a) the first author, year of publication, number of subjects, sex ratio (male/female), mean age of subjects, duration, diagnostic criteria of PD and country of centers; (b) outcome measures such as UPDRS scores, PDQ-39 and LED; (c) intervention characteristics of the trial groups and control groups. Two persons within the reviewing team independently reviewed references and abstracts retrieved by the search, assessed the completeness of data abstraction, and confirmed the quality rating. We used a structured data abstraction form to ensure consistency of appraisal for each study ( Table 1).

Risk of bias of RCTs.
The quality assessment of each RCT was assessed independently using the Cochrane Handbook for Systematic Reviews of Interventions 43 . Two investigators independently evaluated the methodological quality of the included studies. The tool classes studies as having low, moderate, or high risk of bias across six domains; sequence generation, allocation concealment, blinding, missing data, selective reporting and other biases. Disagreements were resolved through consensus or discussed with a third author.
Statistical analysis. UPDRS as the primary outcome measure was considered as continuous data, and then an estimate of the combined effect sizes utilizing WMD was given, and its standard error with 95% confidence interval. Given the potential clinical or methodologic heterogeneity between multi-studies, a random-effects model was used, which yields a more conservative estimate of the pooled effect. If outcomes were presented from the studies at different time points, we extracted data from the last time point. If means and standard deviations were not provided, we calculated them from standard errors, CI, or other statistical indices. Results from intention-to-treat analysis were preferred over results from completer analyses. For the assessment of heterogeneity, the I 2 statistic and chi-square tests were used. Standard meta-analyses were performed with Revman version 5.1. Probability value p < 0.05 were considered significant. The trials had to be conducted with populations of similar age and disease profile sharing a common treatment or placebo arm 44 . For example, a multiple-treatments meta-analysis to estimate the baseline risk or benefit associated with the administration of GPi-DBS versus medical therapy, STN-DBS versus medical therapy and GPi-DBS versus STN-DBS.