Article Text

This article has a correction. Please see:

Download PDFPDF

Transcranial direct current stimulation for the treatment of Parkinson's disease
  1. David H Benninger1,
  2. Mikhail Lomarev1,2,
  3. Grisel Lopez1,
  4. Eric M Wassermann3,
  5. Xiaobai Li1,
  6. Elaine Considine1,
  7. Mark Hallett1
  1. 1Medical Neurology Branch, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, Maryland, USA
  2. 2St Petersburg VM Bekhterev Psychoneurological Research Institute, St Petersburg, Russia
  3. 3Brain Stimulation Unit, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, Maryland, USA
  1. Correspondence to Dr D H Benninger, Human Motor Control Section, Medical Neurology Branch, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Building 10 Room 7D37 (MSC1428), Center Drive, Bethesda, MD 20892, USA; benningerd{at}


Background Progression of Parkinson's disease (PD) is characterised by motor deficits which eventually respond less to dopaminergic therapy and thus pose a therapeutic challenge. Deep brain stimulation has proven efficacy but carries risks and is not possible in all patients. Non-invasive brain stimulation has shown promising results and may provide a therapeutic alternative.

Objective To investigate the efficacy of transcranial direct current stimulation (tDCS) in the treatment of PD.

Design Randomised, double blind, sham controlled study.

Setting Research institution.

Methods The efficacy of anodal tDCS applied to the motor and prefrontal cortices was investigated in eight sessions over 2.5 weeks. Assessment over a 3 month period included timed tests of gait (primary outcome measure) and bradykinesia in the upper extremities, Unified Parkinson's Disease Rating Scale (UPDRS), Serial Reaction Time Task, Beck Depression Inventory, Health Survey and self-assessment of mobility.

Results Twenty-five PD patients were investigated, 13 receiving tDCS and 12 sham stimulation. tDCS improved gait by some measures for a short time and improved bradykinesia in both the on and off states for longer than 3 months. Changes in UPDRS, reaction time, physical and mental well being, and self-assessed mobility did not differ between the tDCS and sham interventions.

Conclusion tDCS of the motor and prefrontal cortices may have therapeutic potential in PD but better stimulation parameters need to be established to make the technique clinically viable.

This study was publicly registered ( NCT00082342).

  • electrical stimulation
  • parkinson's disease
View Full Text

Statistics from


The progression of Parkinson's disease (PD) is therapeutically challenging. In the early stages of PD, motor deficits respond to dopaminergic therapy. This response diminishes and additional symptoms arise which result from progressive degeneration affecting non-dopaminergic neuronal systems.1 One hallmark of this progression is the emergence of difficulties with gait and postural control which eventually become refractory and critical causes of disability.

Surgical interventions, primarily deep brain stimulation (DBS) of various target nuclei, are the ultimate therapeutic options when conventional therapy fails. Yet DBS is limited to a small well defined patient population and carries the risk of serious surgical complications and significant neuropsychiatric side effects. Therapeutic alternatives are needed.

Therapeutic studies of non-invasive brain stimulation, foremost repetitive transcranial magnetic stimulation (rTMS), have yielded promising results in PD. Two meta-analyses concluded there was a modest therapeutic effect of rTMS in motor performance in PD.2 3

Transcranial direct current stimulation (tDCS) is another mode of non-invasive brain stimulation whereby a direct current is applied via surface electrodes on the head for a certain time in contrast with the electric impulse induced by the short lasting magnetic field in TMS. The possibility of modulating cortical excitability4 5 and promoting motor learning in healthy adults6 and motor recovery in chronic stroke7 has raised interest in tDCS as an intervention in PD. An open study reported improvement of gait and bradykinesia in PD.8 A recent crossover study found acute motor improvement after a single session.9 These findings are promising, especially since refractory gait disturbances might be improved. Thus these effects need to be confirmed in a controlled study and explored to determine if they persist for a longer period to exert a therapeutic benefit.

tDCS has some advantages over rTMS, including a favourable safety profile, tolerability, easier applicability and cost effectiveness. Thus tDCS could potentially complement the therapeutic armamentarium.

In this double blind, randomised, sham controlled study, we investigated whether anodal tDCS of the motor and prefrontal cortices improves gait and bradykinesia in PD and whether these effects persist for a longer time. The primary endpoint of interest was improvement of gait in the on state, chosen to see if any benefit would be beyond current best therapy.


Study population

Inclusion criteria were patients aged 40–80 years with PD according to UK PD Brain Bank criteria in a Hoehn and Yahr stage of 2–4 while ‘off’ medication. Patients had to have slowing of gait defined as a time of 6 s or more to walk 10 m. Patients with severe freezing or unable to walk 10 m were excluded. Patients were required to be on an optimal medication regimen with a total levodopa equivalent dose of ≥300 mg. Exclusion criteria were significant medical or psychiatric illness, and metal objects or stimulators in the head which might pose a hazard during tDCS.

A power analysis yielded a sample size of 21 participants per arm providing 80% power with two sided α=0.05, postulating a 20% improvement in gait time in the on state with tDCS compared with sham (primary outcome). After enrolment of more than 50% of the target population, we recalculated the power because of absence of subjective gait improvement. The rather small effect size of 0.256 would have required a sample size of 292. We opted to terminate the study. Thus in this study, we prospectively investigated 25 patients with mild to moderate PD (nine women, mean age 63.9±8.7 years, range 49–77 years, all right-handed; Hoehn and Yahr stage mean 2.4±0.2 in ‘on’ and 2.8±0.4 in ‘off’ medication). Demographic and clinical findings of the patients in the tDCS (n=13) and sham (n=12) intervention groups were comparable (see table 1).

Table 1

Demographic and clinical findings in patients with Parkinson's disease receiving tDCS (n=13) or sham (n=12)

The NIH Institutional Review Board approved the study and the early termination. We obtained written informed consent from all study participants. This study was publicly registered ( NCT00082342).

tDCS intervention

tDCS was applied in eight sessions within 2.5 weeks (Monday, Wednesday and Friday) when ‘on’ medication. Patients were at rest without concurrent cognitive or motor task. A battery driven stimulator, Phoresor II Model PM850 (Iomed, Salt Lake City, Utah, USA), delivered the tDCS through electrodes (saline soaked sponges). The device is FDA approved for transdermal iontophoretic drug delivery and for the purpose of tDCS considered no significant risk. We randomly assigned patients to a real or sham group according to a computer generated number with equal probability. All were naïve to tDCS. In the tDCS treatment group, anodal tDCS (2 mA) was delivered for 20 min through a large ‘3.5“×7” Rubber Pad W/Sponge Insert’ electrode (surface 97.5 cm2; current density 0.021 mA/cm2) that we placed symmetrically either over the premotor and motor (electrode centre 10 mm anterior to Cz) or prefrontal cortices (forehead above eyebrows). We stimulated a single target area during one session and alternated the position of the anode between sessions (starting with the motor area) so that each target area was stimulated four times. Cathodes (25 cm2 each) were positioned over the mastoids. These specific montages and approximate spatial distribution of the current density during tDCS have been reported elsewhere.10 In sham tDCS, we placed anode and cathode (each 9 cm2) 1 cm apart over the forehead and DC (1 mA) applied for 1–2 min which was short circuited through the skin creating the same temporary ‘tingling’ sensation without effects on the brain. We placed two additional electrodes inversely over the mastoids, not connected to the stimulator. We set up the stimulating apparatus out of sight of the patients and blinded investigators. In both sham and tDCS, the current was ramped up over 10 s and similarly decreased. A temperature sensor in the first three patients demonstrated that skin temperature did not increase during stimulation.

Clinical assessment

Baseline and follow-up evaluations were performed before and (see online supplementary figure) 24 h, 1 and 3 months after the last tDCS intervention session. Primary outcome measures were the change in the timed test of gait in the on and off state 24 h after the intervention period compared with baseline. Secondary measures included changes 1 and 3 months after intervention completed. We assessed gait by measuring the time to walk 10 m. Patients were instructed to walk at a fast pace without taking the risk of falling wearing the same shoes and using assistive devices consistently if needed. We timed gait from initiation while standing and in the same conditions (location, lighting, etc). Secondary outcome measures included bradykinesia assessed in the hands and arms. We measured the time to perform the following sequence 10 times: (1) hand closing (squeezing a ball) and opening; (2) elbow flexion; (3) hand closing and opening; and (4) elbow extension. This is similar to a previously studied sequential task with elbow flexion and hand closure shown to correlate with bradykinesia.11 Before baseline assessment, patients practiced until performance appeared not to get faster and then abstained from further practice to minimise learning effects. We chose timed tests because they are more sensitive for detecting changes than scores such as the Unified Parkinson's Disease Rating Scale (UPDRS) and are independent from subjective assessment. These motor tests and the UPDRS were assessed in the ‘best on’ and ‘practically defined off state’ by the same blinded raters for the entire study on the same day. ‘Practically defined off state’ corresponded to overnight (≥12 h) withdrawal of dopaminergic medication and preceded, therefore, assessment in the ‘best on state’, considered by the patients and blinded rater to be the best response to their usual dopaminergic medication. Gait and bradykinesia were also timed before and after each intervention to evaluate acute effects. Additionally, unblinded investigators performed a short clinical assessment to monitor for the safety of tDCS.

The evaluation included the Beck Depression Inventory and a Health Survey (SF-12v2) addressing the subjective perception of health and well being. Patients appraised their state of mobility by checking boxes of defined states (‘on’ condition, ‘on’ with dyskinesias, ‘off’ condition, ‘off” with tremor or sleep) for each hour in a log (for 3 days). Visuomotor speed and procedural learning were tested in the Serial Reaction Time Task, as described previously.12

Statistical analysis

To compare the two groups on various outcome measures, a two sample t test or Wilcoxon ranked sum test, whichever was deemed appropriate, was used. Fisher's exact test was used to assess association between two categorical variables. For longitudinal continuous data, summary statistics (mean, SD) at each time point were reported. Linear mixed effects models (SAS, Proc Mixed,13) were used to analyse the serial gait time. The independent variables included group (tDCS or sham), condition (‘on’ or “off”), time (1 day, 1 month and 3 months after the intervention) and all the two-way and three-way interactions of the above three factors. The baseline gait measure was used as a covariate. The intercept and time variable were treated as the random effects to account for within subject variability. An unstructured variance–covariance matrix was adopted in the model. Similar approaches were used to analyse other outcome measures such as the sequential hand and arm movement time (the average of the left and right sides of the patient), UPDRS, UPDRS III and UPDRS bradykinesia. For session data on gait and hand and arm movement, treatment differences were assessed with linear mixed effects models in which post-session measures were the dependent variable and pre-session measures, session, treatment group and the interaction of these two were the independent variables. As for the analysis of the learning rate in the Serial Reaction Time Task, we used linear mixed effects models with autoregressive (1) covariance structure to take into account the correlations among the five measurements (blocks 2–6) per patient. The primary comparisons for this study were the between group differences in changes from baseline to 1 day after the treatment in gait time in the ‘on’ or “off” condition. A two sided significance threshold of 0.025 was set to adjust for the multiplicity. All other comparisons obtained from the model based contrasts were secondary. A p value less than 0.05 was considered significant. Statistical analysis was done using both SPSS (V.12.0.1) and SAS (V.9.1).


All 25 patients enrolled completed the study (figure 1). In a single patient, there were small first degree burns likely caused by accidentally malpositioned electrodes over the mastoids partially covering the earlobes with reduced contact surface resulting in an increased current density which healed completely within 3 days. We observed no other adverse events. All patients experienced occasional ‘tingling’, which was most commonly of short duration, but no pain or discomfort. Blinding appeared reliable based on patients' and blinded raters' reports. At baseline, primary and secondary outcome measures did not differ between the groups apart from a trend towards a lower score of physical health in the sham group (p=0.065) (tables 1–3).

Figure 1

Flow diagram of patients with Parkinson's disease enrolled in this therapeutic study.

Table 2

Gait and sequential hand and arm movement time at baseline, 1 day, and at 1 and 3 months after the tDCS and sham intervention

Table 3

Secondary outcome measures at baseline, 1 day, and at 1 and 3 months after the tDCS and sham intervention


Compared with sham intervention, the decrease in walking time in those receiving tDCS in the “off” condition 1 day after tDCS barely missed significance (−22.6% vs −19.6%; p=0.03). Since walking times in the off state alone in one patient (sham stimulation) were extreme outliers (baseline 87.8 s and a day, at 1 and 3 months after tDCS 46.3, 24.6 and 9.7 s) and the marked decrease remained inexplicable but for the patient's declared desire that the intervention be successful, we repeated the analysis of the off state excluding those measurements, and this analysis indicated a significant decrease in walking time with tDCS (−22.6% vs 3.6%; p=0.002). No differences were seen when ‘on’ (−17.4% vs −12.7%; p=0.44) or beyond the immediate post-intervention period at 1 or at 3 months thereafter (table 2, figure 2A). Comparing post-interventional performance (1 day after the last intervention) with baseline in each group, the decrease in walking time was significant with tDCS and sham when ‘on’ (−17.4%, p<0.01, and –12.7%, p=0.03) and with tDCS when “off” (−22.6%, p<0.01) and remained significant 1 month later in the tDCS group when ‘on’ (−19.3%, p=0.02). Walking time decreased with tDCS compared with sham intervention when looking at the sessions (treatment×time interaction, p=0.0007) (figure 2B; online supplementary table 4), being significant at the first session (p=0.014), while the opposite was found in the fourth session (p=0.011).

Figure 2

(A) Walking time before, 1 day and at 1 and 3 months after the intervention (mean±SE). The figure shows the decrease in time needed to walk 10 m in the ‘on’ and ‘off’ conditions. Abscissa indicates the time of measurement; ordinate indicates the walking time. The solid lines indicate the transcranial direct current stimulation (tDCS) (n=13) and the broken lines the sham group (n=11, patient with outliers excluded). (B) Walking time before and after each intervention (mean±SE). The figure shows the time needed to walk 10 m Abscissa indicates the time of measurement; ordinate indicates the walking time. The solid lines indicate the tDCS (n=13) and the broken lines the sham group (n=12). Walking time decreased with tDCS compared with sham intervention when looking at the sessions (treatment×time interaction; p=0.0007) being significant at the first session (p=0.014) while the opposite was found in session 4 (p=0.011). (C) Sequential hand and arm movement test before, 1 day and at 1 and 3 months after the intervention (mean±SE). The figure shows the decrease in time needed to execute the sequential hand and arm movement test in the ‘on’ and ‘off’ conditions which persist for 3 months after tDCS. Measurements for the left and right hands were pooled. Abscissa indicates the time of measurement; ordinate indicates the execution time. The solid lines indicate tDCS (n=13) and the broken lines sham group (n=12). (D) Sequential hand and arm movement test before and after each intervention (mean±SE). The figure shows the time needed to execute the sequential hand and arm movement test. Measurements for the left and right hands were pooled. Abscissa indicates the time of measurement; ordinate indicates the execution time. The solid lines indicate tDCS and the broken lines sham group. Sequential hand and arm movement time decreased with tDCS compared with sham intervention when looking at the sessions (treatment×time interaction; p=0.015), being significant after the first two and sixth session (p=0.035, 0.005 and 0.034). *p<0.05; **p<0.01;***p<0.0001.


Bradykinesia decreased significantly more with tDCS than sham intervention (−28.4% vs −11%, p=0.002, when ‘on’ and −36.0% vs −17.8%, p<0.0001, when “off”) (table 2, figure 2C). Comparing post-interventional performance with baseline in each group, the decrease in bradykinesia was significant in the tDCS and sham group (at all time points in the on and off condition, p<0.001 and p<0.05, respectively, figure 2C). Bradykinesia decreased with tDCS compared with sham intervention when looking at the sessions (treatment×time interaction, p=0.015; figure 2D; online supplementary table 5), being significant after the first two and sixth session (p=0.035, 0.005 and 0.034).

Unified Parkinson disease rating scale

Compared with sham, tDCS had no effects on the total and motor UPDRS scores (table 3). However, a composite UPDRS bradykinesia score (items 23–25: finger tapping, opening/closing and pro-/supination of the hands) indicated an improvement with tDCS in the immediate post-intervention period in the off state but this decrease was not significant in the on state. Comparing UPDRS scores in each group, there were significant decreases in the total scores when ‘on’ (p<0.05) in both groups and UPDRS III motor scores when ‘on’ in the sham group (p<0.05). There was a trend to a decreased UPDRS III motor scores when ‘on’ and “off” (p=0.07 and p=0.06).

Serial reaction time task

Comparing tDCS and the sham groups, there were no significant changes from baseline to any time after the intervention in reaction time, error rate (ER) and sequence specific learning (table 3). There was no significant block effect during sequence repeating blocks 2–6 for per cent change of reaction time and ER in the tDCS and sham groups at any time point.

Changes from baseline to any post-intervention time in the Beck Depression Inventory, health survey and self-assessment of mobility (log) did not differ between groups (table 3).


The principal finding of this first double blind, randomised, sham controlled study is that anodal stimulation of the motor and prefrontal cortices improved upper extremity bradykinesia. The effects on gait are somewhat ambiguous: tDCS increased walking speed in the off state when excluding the patient with excessive walking times we considered factitious but statistical significance was barely missed when including this patient and correcting for multiplicity. Gait did not improve when medicated but session data suggest that tDCS might have had a short lived beneficial effect.

This study supports findings of efficacy of tDCS on bradykinesia and potentially on gait,8 9 encouraging further research into its therapeutic potential. Larger studies could confirm these findings as the observed effect is small, which emphasises the need for a more powerful stimulation for clinical impact.

Gait disturbances arise from various pathophysiological mechanisms which might differ in their response to tDCS. So far, reports of gait improvement with tDCS8 contrast with reports not showing benefit.9 Likewise, rTMS improved gait14 15 but not in all studies.16 DBS of the pedunculo-pontine nucleus17 18 is reported to improve gait disturbances refractory to conventional therapy. Since the pedunculo-pontine nucleus connects with the cortico-striato-thalamo-cortical circuit, its activity could, theoretically, be modulated by cortical stimulation. We evaluated gait by speed alone but may have missed qualitative gait improvement which needs to be addressed in future studies.

The improvement in bradykinesia in the best on state suggests that effects of anodal tDCS may exceed the best response of certain symptoms to dopamine substitution. Thus tDCS may act on mechanisms eluding dopaminergic medication and complement conventional therapy. There was also an improvement in the sham group. An explanation is motor learning produced by the repeated assessment sessions as we kept patients from further practice otherwise after enrolment. If true, we expect a similar learning effect in the tDCS group which provides another rationale for controlled studies. Motor learning could explain why the proper effect of tDCS on bradykinesia might be smaller as the composite UPDRS bradykinesia score suggests. tDCS might enhance learning6 and thus combining tDCS and rehabilitation training could carry a potential benefit. The decrease in walking time with sham most plausibly reflects familiarisation with the task.

Other than what has been discussed, no superior effects of tDCS could be discerned from sham stimulation. There was improvement in most measures with sham intervention substantiating a placebo effect. The persuasive concept of stimulation ‘boosting’ under active brain areas may have heightened the expectancy and emphasises the importance of controlled studies, which are feasible since blinding appears reliable.19

The efficacy of tDCS is enhanced when repeated,8 as with rTMS,14 but the number of sessions for the optimal response remains unknown. There might be a larger effect of tDCS within the first sessions, particularly for bradykinesia. Yet, potential effects could have been masked by the larger variability in motor performance since assessments before and after the intervention could not control for fluctuations, in contrast with assessments in the well defined best on state and ‘practically defined off state’. Additionally, the placebo effect appeared larger immediately after the intervention.

The mechanisms by which tDCS and other stimulation modalities improve motor performance in PD are not known and they probably differ. The best evidence supporting clinical efficacy of brain stimulation comes from DBS. DBS supposedly interferes with pathological activity and induces changes in activity20 21 and excitability22 23 of the motor cortex suggesting a possible mechanism that acts trans-synaptically along cortico-striato-thalamo-cortical circuits. Direct stimulation of the motor cortex in small series support this concept24 but findings are not uniformly positive and changes in brain activity are not always found.25 High frequency rTMS with modest efficacy2 3 increases excitability that presumably correlates with improvement in bradykinesia.14 No other controlled study explored changes in brain physiology and behaviour and their interaction. In contrast with pulsed stimulation, tDCS delivers a continuous current that modulates membrane excitability and induces shifts in cortical excitability without rapid depolarisation sufficient to produce an action potential.4 5 Anodal stimulation increases excitability and, thereby, firing of active neurons,4 and supposedly reverses decreased activity in motor and prefrontal cortices in PD.8 9

tDCS may cause release of dopamine as does rTMS of prefrontal and motor cortices in the caudate and putamen in healthy26 27 in PD28 and even sham rTMS, highlighting a possible mechanism of placebo.29 The widespread activation with anodal tDCS30 may release dopamine and could be the mechanism for acute improvement.9 Further evidence for an involvement of dopamine in tDCS effects comes from the observation that anodal tDCS of M1 prolongs the cortical silent period 31 shown to reflect dopaminergic action in PD.22 32 Both DBS22 and rTMS33–35 modulate the cortical silent period, suggesting similar mechanisms on motor cortex excitability.

This persistence of effects implies functional and structural changes in synaptic strength, which constitutes the basic mechanism in plasticity. Pharmacological blocking of N-methyl-d-aspartate receptors prevents long lasting effects of tDCS on cortical excitability36 37 suggesting tDCS may recruit N-methyl-d-aspartate receptor dependent plasticity, an action thought to depend on dopamine.38 This role of dopamine in plasticity in PD is demonstrated by the effects of 5 Hz39 and 1 Hz rTMS40 on cortical excitability in the on but not the off state. Thus induction of longer lasting effects with tDCS might depend on dopamine. This would explain the absence of enduring effects with tDCS when unmedicated.8

This study suggests a therapeutic potential of tDCS but stimulation must be more powerful to improve functional status and quality of life along with motor performance.


The authors thank David Prosper for help in the research, Sungyoung Auh for statistical analysis and Devera Schoenberg for skilful editing.


View Abstract


  • Linked articles: 205112.

  • Funding This research was supported by the Intramural Research Program of the NINDS, NIH and in part by a grant from the USAMRMC (W81XWH-06-1-0534).

  • Competing interests MH has received personal compensation or travel expenses for activities with Neurotoxin Institute, John Templeton Foundation, Parkinson's and Ageing Research Foundation, University of Pennsylvania, Thomas Jefferson University, Baylor College of Medicine, American Academy of Neurology, Medical University of South Carolina, Northshore-Long Island Jewish Hospital, American Clinical Neurophysiology Society, Columbia University, University of Alabama, Blackwell Publisher, Cambridge University Press, Springer Verlag, Taylor and Francis Group, Oxford University Press, John Wiley and Sons and Elsevier as an advisory board member, an editor, a writer or a speaker. MH has received licence fee payments from the National Institutes of Health for the H-coil, a type of coil for magnetic stimulation. MH and his wife held stock in Agilent Technologies, Amgen, Amylin Pharmaceuticals, Merck and Co, Monsanto Co New Del, Sanofi Aventis Adr, Coventry Health Care Inc, Sigma Aldrich Corp, Warner Chilcott Ltd, Pfizer Inc, Genentech, Inc, United Health Group, St Jude Medical and Eli Lilly and Company.

  • Ethics approval This study was conducted with the approval of the Institutional Review Board, NINDS, NIH.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles