Introduction

With over 140,000 patients treated worldwide, subthalamic nucleus deep brain stimulation (STN-DBS) is an established treatment for motor complications in Parkinson disease (PD) [1]. The postSTN-DBS management, however, poses the challenge of identifying the optimal combination of dopaminergic therapies and stimulation settings.

Reduction in levodopa equivalent daily dose (LEDD) and other dopaminergic medications is widely endorsed after STN-DBS, and has become an “anticipated benefit” of this surgical modality. This paradigm stems from the rationale that STN-DBS might reduce PD cardinal symptoms to a similar extent than levodopa (L-dopa) and that decreasing medications reduces postoperative dyskinesia [1, 2]. On the other hand, medication reduction can elicit other problems, such as depression and apathy [3], which creates uncertainty as to the wisdom of aggressively lowering dopaminergic therapies in patients treated with STN-DBS [4]. While STN-DBS and L-dopa have been recognized as providing similar motor benefits, no systematic assessment of these two treatments combined has been performed in long-term studies [5]. In addition, the difference in motor complications and activities of daily living (ADL) between patients with high vs. low postsurgical LEDD reduction remains to be clarified.

In this meta-analysis, we sought to estimate the magnitude of difference between ON and OFF medication states and ON and OFF stimulation states to determine if STN-DBS and L-dopa may yield differential motor, dyskinesia, and ADL outcomes when combined compared to when stimulation and medication are administered alone.

Materials and methods

This meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) and the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines [6, 7]. Observational studies, randomized clinical trials (RCT), and non-randomized clinical trials (n-RCT) were included if meeting the following criteria: (a) surgical selection for bilateral STN-DBS, as per the Core Assessment Program for Surgical Interventional Therapies in PD (CAPSIT-PD) [8]; (b) presurgical assessment of motor symptoms in Medication-OFF (Med-OFF) and Medication-ON (Med-ON) conditions, as per the motor subscale of the Unified Parkinson’s Disease Rating Scale (UPDRS-III); (c) postsurgical assessment of motor symptoms in the following conditions: Stimulation-ON/Medication-ON (Stim-ON/Med-ON), Stimulation-OFF/Medication-OFF (Stim-OFF/Med-OFF), Stimulation-OFF/Medication-ON (Stim-OFF/Med-ON), and Stimulation-ON/Medication-OFF (Stim-ON/Med-OFF), using a supra-maximal L-dopa challenge dose to assess Medication-ON conditions (Supplementary Table 1), as per the CAPSIT-PD protocol [8].

Exclusion criteria were incomplete data reporting (i.e., lacking one or more of the four postsurgical CAPSIT-PD conditions) or sample sizes fewer than five patients. No restrictions were applied to gender, disease duration, disease severity, or DBS manufacturer.

Search methods

We searched for eligible studies in PubMed, Embase, Cochrane Movement Disorders Group Trials Register, Cochrane Central Register of Controlled Trials (CENTRAL), ClinicalTrials.gov, and the System for Information on Grey Literature in Europe (OpenGrey) up to 31 December 2017 using the following search terms: Parkinson disease, Parkinson, deep brain stimulation, DBS, and follow-up (Supplementary Table 2). No language restrictions were applied.

Meta-analysis design

We divided the analyses into short term (< 5 years) and long term (≥ 5 years) after surgery, comparing the change in UPDRS-III in the four possible Stimulation/Medication conditions. In addition to the motor outcome (UPDRS-III), we examined ADL (UPDRS-II) and the change in the proportion of the waking day spent with dyskinesia (UPDRS item 32).

Selection of studies and data extraction

Abstracts were reviewed for eligibility criteria by three investigators (J.A.V., M.S., and A.M.). Pertinent full-text articles were assessed and variables of interest extracted. Particular attention was paid to studies that shared the same population or published data from the same cohort at different time-points. In this scenario, the longest follow-up within each time interval (< 5 years and ≥ 5 years) was used for the analyses. Disagreements were anticipated to be settled by consensus.

Assessment of risk of bias and heterogeneity

Evidence quality was independently assessed by two investigators (J.A.V. and M.S.). For included studies, we used the Cochrane-validated “Quality Assessment Tool for Before–After Studies with No Control Group” [9]. Visual inspection of funnel plots was conducted to assess for publication bias [10]. Subsequently, the overall confidence in the effect for each outcome of interest was assessed following the Grading of Recommendations Assessment, Development and Evaluation (GRADE) methodology [11]. The degree of heterogeneity was deemed considerable if I2 statistic was ≥ 75% and significance test (p value) below 0.1 [12].

Measures of treatment effect

Random-effects meta-analyses using generic-inverse variance were used to pool the mean differences and standard errors of the following outcomes: (a) motor score; (b) dyskinesia duration; (c) ADL, at the prespecified follow-up intervals, with 95% confidence intervals (C.I.) for these pooled estimates. Subgroup analyses were conducted to compare studies with high (≥ median) vs. low (< median) postsurgical LEDD reduction from baseline, within, and beyond 5 years. All the analyses were performed in Review Manager® (RevMan, version 5.3. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014).

Results

Out of the 1632 records derived from the initial search strategy (Supplementary Fig. 1), 12 observational and non-randomized studies met full criteria [2, 5, 13,14,15,16,17,18,19,20,21,22], which underwent data extraction (Table 1) and assessment for individual risk of bias (Supplementary Table 3). Pooled studies were assessed to determine the overall quality of evidence (Supplementary Table 4). The agreement was met between evaluators in all cases, and no signs of publication biases were observed in funnel plots (Supplementary Fig. 2).

Table 1 Characteristics of included studies
Full size table

The total study population consisted of 401 PD patients treated with STN-DBS (n = 366 with < 5 years and n = 196 with ≥ 5 years of follow-up). Six patients had undergone the previous neurosurgical procedures (n = 4 in Ostergaard and Sunde [13]; and n = 2 in Schupbach et al. [21]).

Follow-up < 5 years

The Stim-ON/Med-ON condition reduced (improved) UPDRS-III by − 35.7 points (95% CI − 40.4, − 31.0) compared with Stim-OFF/Med-OFF, but also by − 11.2 [− 14.0, − 8.4] compared with Stim-OFF/Med-ON and − 9.5 [− 11.0, − 8.0] compared with Stim-ON/Med-OFF. No difference was observed between Stim-ON/Med-OFF and Stim-OFF/Med-ON conditions (Fig. 1).

Fig. 1
figure 1

STN-DBS and L-dopa effect on UPDRS-III at < 5 years. Inverse variance method was used to calculate mean differences and data were pooled using a random-effects model. Results are shown as point estimates and 95% confidence interval. STN-DBS + L-dopa Stimulation-ON/Medication-ON, STN-DBS Stimulation-ON/Medication-OFF, L-dopa Stimulation-OFF/Medication-ON, OFF state Stimulation-OFF/Medication-OFF, UPDRS-III Unified Parkinson’s Disease Rating Scale part III, IV inverse variance, CI confidence interval

Full size image

High vs. low postsurgical LEDD reduction (Fig. 2) resulted in a similar improvement in dyskinesia duration [− 1.4 (95% CI − 1.5, − 1.2) vs. − 1.0 (95% C.I. − 1.7, − 0.4); p = 0.33; I2 = 0] and no significant differences in the ADL outcomes [0.6 (95% CI − 0.5, 1.6) vs. − 0.01 (95% CI − 4.1, 4.1); p = 0.79; I2 = 0].

Fig. 2
figure 2

Dyskinesia duration and ADL in higher vs. lower LEDD reduction subgroups. Inverse variance method was used to calculate mean differences and data were pooled using a random-effects model. Results are shown as point estimates and 95% confidence interval. STN-DBS + L-dopa Stimulation-ON/Medication-ON, STN-DBS Stimulation-ON/Medication-OFF, L-dopa Stimulation-OFF/Medication-ON, OFF state Stimulation-OFF/Medication-OFF, UPDRS-II Unified Parkinson’s Disease Rating Scale part II, UPDRS-IV #32 Unified Parkinson’s Disease Rating Scale-Item 32, IV inverse variance, CI confidence interval

Full size image

Follow-up ≥ 5 years

The Stim-ON/Med-ON condition reduced UPDRS-III by − 28.6 points (95% CI − 32.8, − 24.4) compared to Stim-OFF/Med-OFF, but also by − 8.1 [− 10.2, − 5.9] compared with Stim-OFF/Med-ON and − 8.0 [− 10.3, − 5.6] compared with Stim-ON/Med-OFF. No difference was observed between Stim-ON/Med-OFF and Stim-OFF/Med-ON conditions (Fig. 3).

Fig. 3
figure 3

STN-DBS and L-dopa effect on UPDRS-III at ≥ 5 years. Inverse variance method was used to calculate mean differences and data were pooled using a random-effects model. Results are shown as point estimates and 95% confidence interval. STN-DBS + L-dopa Stimulation-ON/Medication-ON, STN-DBS Stimulation-ON/Medication-OFF, L-dopa Stimulation-OFF/Medication-ON, OFF state Stimulation-OFF/Medication-OFF, UPDRS-III Unified Parkinson’s Disease Rating Scale part III, IV inverse variance, CI confidence interval

Full size image

High vs. low postsurgical LEDD reduction (Fig. 2) resulted in a similar improvement in dyskinesia duration [− 1.1 (95% CI − 1.3, − 0.9) vs. 1.1 (95% CI − 1.5, − 0.7); p = 0.99; I2 = 0] and no significant differences in the ADL outcomes [5.6 (95% CI 1.0, 10.3) vs. 6.8 (95% CI 3.0, 10.6); p = 0.71; I2 = 0].

Discussion

The results of this meta-analysis demonstrated that while there was similar individual efficacy of STN-DBS and L-dopa, their combined effect on motor severity was additive within and beyond 5 years of follow-up, with a UPDRS-III differential between 9.5 and 11.2 points in the short term and between 8.0 and 8.1 points in the long term. These values are above the 3.25 point threshold considered the minimal clinically important difference (MID) for the UPDRS-III [23]. In addition, no difference was observed in the extent of dyskinesia duration improvement or in the ADL outcome between studies with high vs. low postsurgical LEDD reduction.

Taken together, these data argue against the paradigm of invariably aiming at reducing the dopaminergic tone as part of the postsurgical management of STN-DBS patients. In fact, there is no evidence that greater reduction in dopaminergic therapies might lead to better control of dyskinesia, while harnessing an additive effect between STN-DBS and L-dopa may be particularly relevant at advanced disease stages, in which the main sources of disability are relatively resistant to the conventional medical and surgical therapies alone, such as gait, balance, speech, swallowing, and cognitive impairments [5, 19, 24, 25]. Furthermore, lower reduction in dopaminergic medications, as reported after unilateral STN-DBS [4], might result in lower incidence of apathy and depression [26].

The underlying mechanism behind the additive effect of stimulation and medication might reflect the complementary effects of both intervention, modulating both dopaminergic and non-dopaminergic pathways including, but not limited to, cholinergic and adrenergic circuits [27]. Furthermore, there may be a differential modulation of nigro-striatal dopaminergic pathways between STN-DBS and L-dopa in advanced PD, when L-dopa response may be limited by aberrant synaptic plasticity, reduced density in D3 striatal dopamine receptors [28], and progressive loss of dopaminergic neurons in the caudal putamen [29]. Whether STN-DBS effectiveness might be hampered by advanced degeneration of dopaminergic and non-dopaminergic pathways remains unclear [30].

Some limitations attenuate the strength of our conclusions. First, we included uncontrolled, non-randomized clinical studies of small sample sizes, which lower the confidence in the overall effect. Although heterogeneity was present, a meta-regression was not performed due to the limited number of studies included in the analyses [12]. To minimize these shortcomings, we carefully assessed the individual and overall quality of included studies as per the Cochrane and GRADE handbook recommendations [11]. Second, although standard in the presurgical evaluation of patients, the comparison between Med-OFF and a supra-therapeutic Med-ON condition may not represent an accurate estimate of the patient daily response to L-dopa therapy. Relatedly, not all measurements of UPDRS-III postoperative response to management may have accurately represented ecologically valid settings, such as their functioning at home. In addition, the data on dyskinesia based on UPDRS item 32 lack characterization of semiology, severity, and functional impairment, and may not be sensitive enough to treatment. While the assessment of dyskinesia duration and ADL reflected information gathered from daily-living clinical experience, the possibility exists that subgroup analyses might be underpowered to detect small differences between high and low postsurgical LEDD reduction subgroups. Unfortunately, data from the full UPDRS-IV and from quality of life scales were not consistently available, and would have prevented the construction of a pooled estimate with meta-analysis.

In conclusion, our data confirm the comparable efficacy of STN-DBS and L-dopa, but also suggest an additional benefit to be attained by their combined application, which is greater than each treatment alone. While a postsurgical reduction in dopaminergic therapies may be necessary to ameliorate dopaminergic side effects such as sedation, hallucinations, impulsivity, and orthostatic hypotension, our findings suggest that, for any other reasons, including “simplification” of the daily therapeutic schedule, a significant reduction in dopaminergic tone might preclude the potentially additive effect of STN-DBS and levodopa on PD motor symptoms. Further controlled, prospective studies will be needed to clarify optimal therapeutic strategies, but the available evidence supports the notion that clinicians may be missing an important source of outcome optimization in PD by aggressively reducing medications after bilateral STN-DBS.

Data access and responsibility statement

Drs. Merola and Vizcarra had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.