Two recent sham-controlled studies found that transcranial direct current stimulation (tDCS) was an effective treatment for depression. As tDCS is painless, relatively safe and inexpensive, its efficacy in treating depression warrants further investigation. This double-blind, randomized study tested tDCS at the same stimulation parameters as a previous positive study (1 mA current strength, five treatment sessions, active or sham, given on alternate days) in 40 depressed participants. Anodal stimulation was centred over the left dorsolateral prefrontal cortex, with the cathode placed on the lateral aspect of the contralateral orbit. tDCS was continued up to a total of ten active sessions per participant. Mood outcomes were measured by psychiatrist raters blind to treatment condition using the Montgomery–Asberg and other depression rating scales. Psychomotor speed was assessed immediately before and after a single tDCS session and attention, frontal executive function, working memory and verbal learning were assessed after each group of five sessions. Overall depression scores improved significantly over ten tDCS treatments, but there was no between-group difference in the five-session, sham-controlled phase. tDCS was found to be safe, with no adverse effects on neuropsychological function, and only minor side-effects. It is recommended that the efficacy of tDCS in depression be further evaluated over a longer treatment period, using enhanced stimulation parameters.
Transcranial direct current stimulation (tDCS) is a non-invasive form of brain stimulation in which a weak direct current is passed through the scalp into underlying cerebral tissue, with a resultant change in cortical excitability (Arul-Anandam & Loo, 2009a). tDCS has been applied in animal models and humans for many decades (Beveridge & Renvoize, 1988). Weak direct current stimulation shifts the resting membrane potential, as shown in animal experiments – anodal stimulation has been shown to depolarize the soma of pyramidal cells whereas cathodal stimulation hyperpolarizes them (Bindman et al.1964), and these changes outlast the duration of stimulation by minutes to hours. Recent studies using transcranial magnetic stimulation (TMS) to test cortical excitability before and after a period of tDCS stimulation have shown that similar changes occur in the human motor cortex with tDCS (Nitsche & Paulus, 2000).
For decades, there has been the suggestion that tDCS may have antidepressant effects. Beneficial effects of tDCS in depression were reported in the 1960s and 1970s (Arfai et al.1970; Costain et al.1964; Hordern & Weeks, 1965; Lippold & Redfearn, 1964; Lolas, 1977) but were not followed up by investigators. Outcomes in these studies were quite variable, probably reflecting the considerable diversity of tDCS methodology used, in terms of electrode placement (various parts of the head, trunk, limbs, etc.), electrode size, current amplitude (20–500 µA) and stimulation duration (2 min–8 h).
Recently, two double-blind, sham-controlled studies using left prefrontal anodal tDCS at higher stimulation intensities (1–2 mA) reported positive results in reducing depressive symptoms (Boggio et al.2008; Fregni et al.2006a). In the first study (Fregni et al.2006a), a 60% improvement in mean Hamilton Depression Rating Scale (HAMD) scores was reported after only five sessions of active tDCS (20 min at 1 mA), given on alternate weekdays, compared to 10% improvement in the sham group. The second study (from the same group of investigators) similarly reported robust results, with mean improvement in the active group on HAMD scores of 40.5%, compared to 10.4% in the sham group, after 2 wk (ten consecutive weekdays) of treatment (20 min at 2 mA) (Boggio et al.2008). The difference in outcomes between active and sham treatment groups was still evident at 1-month follow-up in the latter study. When these results were compared with those of depressed patients treated with 20 mg fluoxetine (open label), the authors reported that the response to tDCS was of a similar magnitude but of more rapid onset (Rigonatti et al.2008). Regarding safety, tDCS was well tolerated with minimal side-effects (transient headache, skin itching and redness).
These positive results suggest considerable potential for tDCS as an antidepressant treatment, which needs further testing and corroboration in clinical trials. We report the results of a double-blind, sham-controlled trial to test the antidepressant effects of left prefrontal anodal tDCS in depressed subjects.
Forty subjects with unipolar DSM-IV major depressive episode of up to 3 yr duration and a score ⩾20 on the Montgomery–Asberg Depression Rating Scale (MADRS; Montgomery & Asberg, 1979) were enrolled as outpatients. The diagnosis was based on a structured assessment using the MINI (MINI International Neuropsychiatric Interview; Sheehan et al.1997) and confirmed in a clinical interview by a study psychiatrist (C.K.L.). Subjects with bipolar disorder, drug or alcohol dependence or abuse, other Axis I disorders, neurological disorders, or who had failed to respond to electroconvulsive therapy in the current episode of depression, were excluded.
During the study, subjects were either medication-free or remained on antidepressant medications to which they had failed to respond, continued at stable doses which had not been altered for at least 4 wk prior to enrolment. Six subjects were also on antipsychotic medications, five subjects were on lithium and one subject was on an anticonvulsant medication (lamotrigine). No subjects were on benzodiazepines. After the study period, patients reverted to routine clinical management under their own treating psychiatrist.
The study was approved by the human research ethics committee of the University of New South Wales. After complete description of the study to the subjects, written informed consent was obtained.
Subjects were stratified by age and gender and then randomly assigned to active or sham treatment groups. Active or sham tDCS was given three times per week (Monday, Wednesday, Friday), for five treatment sessions with raters and subjects blind to treatment group assignment. All subjects then continued with active tDCS (at the same treatment frequency) for another five sessions. At enrolment, subjects were told that some of the ten treatment sessions may involve sham treatment. Thus subjects were not aware that treatment sessions 6–10 definitely involved active treatment. After the ten sessions, the integrity of the blinding was assessed by asking subjects to guess whether they had been assigned to the active or sham treatment group. Following this, the blind was broken and subjects who had only received five active treatments were given the option of receiving another five active treatments (see Fig. 1).
Anodal tDCS was administered over the left dorsolateral prefrontal cortex (DLPFC), identified as pF3 on the 10/20 EEG system. The first nine subjects were treated using a DC stimulator made by J. Lagopoulos. For subsequent subjects, an Eldith DC-stimulator (NeuroConn GmbH, Germany) was used. The current output of these devices was checked using an ammeter. The cathode was placed over the lateral aspect of the contralateral orbit. Conductive rubber electrodes (7×5 cm=35 cm2) covered by sponges soaked in saline were used, held in place by a head band. Stimulation was given at 1 mA for 20 min, with gradual ramping up of the current over 30 s. For sham stimulation, the procedure was identical, except that the current was gradually ramped down to zero, after the first 30 s, thus giving the same initial sensation of tDCS. The switching on and off of the current was programmed into the stimulator and did not require intervention by the operator. The machine was placed behind the subjects' heads so that they were unable to see the readout on the front panel of the stimulator.
Assessment of mood and cognition
The primary outcome measure for mood evaluation was the MADRS. Subjects were evaluated at baseline, after each five treatment sessions, 1 week and 1 month after completion of treatment. All ratings were conducted by a psychiatrist who was blinded to treatment condition, using the MADRS, 17-item HAMD (HAMD17; Hamilton, 1960) and Clinical Global Impression scale – Severity of Illness (CGI-S; Guy, 1976). Each subject was rated by the same psychiatrist throughout the study. At the same time-points, subjects rated their mood using the Beck Depression Inventory (BDI; Beck et al.1961) and Patient Global Impression scale – Improvement of Illness (PGI-I; Guy, 1976). Subjects were also assessed using the Core Measure of Psychomotor Disturbance (CORE; Parker et al.1990) at baseline, as a possible predictor of response.
Neuropsychological functioning was assessed at baseline and after each five treatment sessions using the following tests: Rey Auditory Verbal Learning Test (RAVLT; Rey, 1964), Trail Making Tests (TMT) A and B (Reitan & Wolfson, 1985), Wechsler Adult Intelligence Scale (WAIS) digit span (forwards and backwards; Wechsler, 1981), Controlled Oral Word Association Test (COWAT; Benton & Hamsher, 1989). Alternative test forms were used on subsequent occasions for RAVLT and COWAT. Immediate effects of tDCS on processing speed were also assessed after treatment sessions 1 and 5 using the Symbol Digit Modalities Test (SDMT; Smith, 1991), and simple and choice reaction-time tests.
The two treatment groups were analysed for differences in demographic and clinical variables at baseline using χ2 tests for categorical variables and t tests for dimensional variables (see Table 1). Statistical tests were two-tailed. Intention-to-treat last-observation-carried-forward scores were used for the analyses below.
HAMD17, 17-item Hamilton Depression Rating Scale; MADRS, Montgomery–Asberg Depression Rating Scale; CGI, Clinician Global Impressions; CORE, CORE Measure of Psychomotor Disturbance; BDI, Beck Depression Inventory; PGI, Patient Global Impressions.
Actual tally recorded.
Depression ratings over the first five treatment sessions were analysed for change with a repeated-measures design (ANOVA), testing for main effects of time and group as well as time×group interactions. A further repeated-measures analysis of covariance (ANCOVA) was performed on the same data, controlling for concurrent antidepressant treatment. Baseline CORE ratings were correlated (Pearson's correlation) with the percentage change in MADRS scores over the ten active treatment sessions (data pooled across both groups), to see if CORE scores predicted response.
Scores from neuropsychological tests examining for changes over the first five treatment sessions (active or sham), and scores from tests administered immediately before and after a single session of tDCS (active or sham) were also analysed with a repeated-measures design, testing for main effects of time and group as well as time×group interactions. Neuropsychological test scores were separately analysed for all subjects who received ten active treatments (either in the sham-controlled phase or open treatment phase), examining for changes across the ten sessions, using a repeated-measures ANCOVA, controlling for the percentage change in MADRS scores over the same treatment period.
To test the integrity of blinding, subjects' responses (active, sham) when asked to guess their treatment group were compared for active and sham groups using a χ2 test.
There were no significant differences between active and sham treatment groups at baseline, although there was a trend for a higher proportion of subjects to be on concurrent antidepressant medication in the sham treatment group (see Table 1). Thirty-five subjects completed the five-session sham-controlled phase and 34 subjects received ten active sessions of tDCS (over both sham-controlled and open treatment phases) (see Fig. 1).
There were significant main effects of time over the first five treatment sessions (sham-controlled phase) on the HAMD17 and MADRS, but there were no significant interactions between time and group (see Table 2, Fig. 2). Repeated-measures ANCOVAs on HAMD and MADRS scores, co-varying for concurrent antidepressants, also did not find any significant time×group interactions.
DCS, Direct current stimulation; HAMD17, 17-item Hamilton Depression Rating Scale; MADRS, Montgomery–Asberg Depression Rating Scale; CGI, Clinician Global Impressions; CORE, CORE Measure of Psychomotor Disturbance; BDI, Beck Depression Inventory; PGI, Patient Global Impressions.
Over the first ten treatment sessions, there was a significant difference between baseline and post-10 scores on all measures: HAMD17 (F=30.95, d.f.=33, p=0.00), MADRS (F=54.40, d.f.=33, p=0.00), CGI (F=29.53, d.f.=32, p=0.00), BDI (F=35.23, d.f.=31, p=0.00), PGI (F=46.23, d.f.=33, p=0.00) but again there were no significant interactions between time and group. Over the first ten treatment sessions, six participants in the active treatment group met the criterion for response (defined as ⩾50% reduction in MADRS scores from baseline) and five met the criterion for remission (final score ⩽10 on MADRS). In the sham treatment group, there were four responders and three remitters over the same time period.
Over the ten active treatment sessions, mean scores on MADRS decreased from 28.72 (s.d.=4.18) to 18.00 (s.d.=8.90) in the active treatment group, and from 22.45 (s.d.=8.09) to 14.47 (s.d.=10.04) in the group initially assigned to sham treatment. CORE scores did not correlate significantly with percentage change in MADRS scores over the ten active treatments for all subjects.
Sham vs. active tDCS
Over the first five treatment sessions, a significant effect of time was found on COWAT letter total whereby participants' performance was found to improve. In addition, a significant between-group effect was found on RAVLT learning (the active group having lower scores overall relative to sham) (see Table 3), but there were no significant time×group interactions.
DCS, Direct current stimulation RAVLT, Rey Auditory Verbal Learning Task; COWAT, Controlled Oral Word Association Test.
All analyses control for percentage change in Montgomery–Asberg Depression Rating Scale score from baseline.
Analyses of participants' performance immediately before and after tDCS sessions 1 and 5 in the sham-controlled phase showed a significant effect of time on the SDMT at both time-points (performance improved in both the active and sham conditions), and on Choice RT after DCS 1 (both groups improved similarly). There were no significant time×group interactions (see Table 4).
tDCS, Transcranial direct current stimulation; SDMT, Symbol Digit Modalities Test; RT, reaction time.
Effects of ten active DCS sessions on neuropsychological test performance
In the active group no significant changes were found across the following measures: RAVLT total (F=0.37, d.f.=16, p=0.55), RAVLT immediate (F=1.09, d.f.=16, p=0.31), RAVLT delay (F=0.04, d.f.=16, p=0.85), RAVLT learning (F=0.02, d.f.=16, p=0.89), TMT A (F=0.62, d.f.=16, p=0.44), Digit span forwards (F=0.88, d.f.=16, p=0.36), COWAT letter (F=2.92, d.f.=16, p=0.11), and COWAT category (F=0.24, d.f.=16, p=0.63). Improvements in performance were found on Digit span backwards (F=6.57, d.f.=16, p=0.02) and TMT B (F=5.01, d.f.=16, p=0.04).
Similarly, in the group which commenced with sham tDCS but went on to receive ten active treatment sessions, no significant changes were found following the active treatment phase on the following measures: RAVLT immediate (F=0.06, d.f.=13, p=0.81), RAVLT delay (F=0.43, d.f.=13, p=0.52), RAVLT learning (F=0.09, d.f.=13, p=0.77), TMT A (F=0.05, d.f.=13, p=0.83), TMT B (F=0.79, d.f.=12, p=0.39), Digit span forwards (F=0.99, d.f.=13, p=0.34), Digit span backwards (F=0.08, d.f.=13, p=0.78), COWAT letter (F=0.03, d.f.=13, p=0.86), and COWAT category (F=0.45, d.f.=13, p=0.51). Participants' performance in this group improved on RAVLT total (F=8.35, d.f.=13, p=0.01).
Adverse outcomes and side-effects
One participant in the sham treatment group committed suicide the day after receiving the first active treatment session. No clinical changes, including negative emotional reactions, had been apparent after the active treatment session. The timing of the suicide coincided with the single occasion when he was briefly left alone, as he had been continuously in the company of family members during the preceding few weeks. Suicidal thoughts had been present for months, although the risk of suicide had not been rated as high, and had been noted and closely monitored during the sham-controlled phase, by both the treating psychiatrist and study psychiatrist. There were no other serious adverse outcomes.
Side-effects occurring during active treatment were mild to moderate skin redness (n=32), itchiness (n=13) or tingling (n=6) at the electrode sites (primarily at the anode), mild headache (n=8), light-headedness (n=4), ringing in the ears (n=3), blurred vision (n=2), brighter or illuminated vision (n=2), tingling in face/body (n=2), transient hypomania (n=1) (see Arul-Anandam et al.2009a), alleviation of neuropathic pain (n=1) (see Arul-Anandam & Loo, 2009 b), reduced concentration (n=1), slight ‘eyelid jolt’ (n=1), nausea (n=1), mild euphoria (n=1), disorientation (n=1), transient tiredness (n=1), constriction when swallowing (n=1), insomnia (n=1) and anxiety (n=1). Side-effects occurring during sham treatment were mild redness (n=12), tingling (n=8) or itchiness (n=7) at the electrode sites, mild headache (n=6), nausea (n=2), light-headedness (n=2), clearer vision (n=1) and a pulsating sensation at the start (n=1). In both groups some subjects experienced multiple side-effects.
Integrity of blinding
When asked to guess their treatment group at the end of the ten-session double-blind phase, responses of the active group were: active (n=5), sham (n=7), unsure (n=5). Sham group responses were: active (n=7), sham (n=10), unsure (n=0). The difference in active/sham guesses between the two groups was not significant (χ2=0.00, d.f.=1, p=0.98). Subjects were equally likely to base their guesses on either their change in mood or side-effects experienced.
Unlike Fregni et al. (2006a), this study in a larger sample of patients did not find any difference between active and sham tDCS, despite using the same treatment parameters (1 mA, 20-min stimulation, same electrode montage, five sessions given on alternate days). The response to active tDCS over ten treatment sessions is comparable to that reported by Boggio et al. (2008). In the Boggio et al. (2008) study, tDCS was given at a higher intensity (2 mA), for 20 min each weekday, over a 2-wk period, in an attempt to optimize efficacy. However, the magnitude of clinical improvement was no greater than that of the Fregni et al. (2006a) study, possibly because a treatment period of >2 wk is necessary for a full antidepressant response, as seen in clinical trials of TMS (Loo & Mitchell, 2005).
Our results mainly differed from these two previous studies in that the same degree of improvement occurred in the group receiving sham treatment. There are several considerations relevant to this. First, approximately two thirds of the participants in the sham treatment group in our study were on concurrent antidepressant medications, whereas all subjects in the Boggio et al. (2008) study had been off antidepressant medications for at least 2 months prior to tDCS. While it is possible that this accounted for improvement during sham treatment in our study, this is unlikely, given that these were medications to which the participants had previously failed to respond. The medications were continued at stable doses due to concerns over clinical deterioration in the event of medication withdrawal. Second, the magnitude of improvement during sham treatment is comparable to that typically seen in antidepressant treatment trials (Walsh et al.2002). Our sample was only moderately treatment resistant as judged by the number of antidepressant medication trials failed prior to study enrolment (predominantly stages 0–III as defined by Thase & Rush, 1995), and a smaller placebo response may be expected if tDCS were trialled in a more treatment-resistant population. However, our sample had a similar level of treatment resistance to that of the Boggio et al. (2008) study. On the other hand, debriefing after the sham-controlled period confirmed our observations that subjects found it difficult to differentiate active and sham treatment, a factor that is likely to increase the placebo response. The difficulty of differentiating a true response from placebo response has also been evident in clinical trials of TMS in depression (Loo & Mitchell, 2005), possibly due to the raised expectations arising from these new brain stimulation technologies.
Apart from the issue of concurrent medications, our sample also differed from that of the Boggio et al. (2008) study in that patients with personality disorders were not excluded. Others have demonstrated that the presence of personality disorder reduced the likelihood of response to antidepressant treatment (Black et al.1988). In other respects (age, gender ratio, depression severity), our sample was comparable to that of Boggio et al. (2008).
The CORE measure of psychomotor disturbance has been proposed as a predictor of response to other physical antidepressant treatments (e.g. electroconvulsive therapy, Hickie et al.1996) but did not predict response to active treatment in this study.
tDCS given at 1 mA on alternate days over ten sessions was shown to be safe, with minor side-effects and no adverse effects on neuropsychological functioning. Given the circumstances of the suicide, it is unlikely that it was related to the single session of active tDCS stimulation, although this possibility cannot be excluded. Numerous other studies with anodal prefrontal and/or cathodal supraorbital stimulation at similar parameters have not reported negative emotional effects (e.g. Boggio et al.2008; Fregni et al.2005, 2006a; Kincses et al.2004; Nitsche & Paulus, 2000). Improvements found in neuropsychological test performance were probably due to practice effects. These results are consistent with those of the clinical trial by Fregni et al. (2006b) and studies in which prefrontal tDCS at similar parameters in healthy volunteers was actually found to enhance neuropsychological function (Fregni et al.2005; Kincses et al.2004).
The main limitation of this study is that the sham-controlled period only spanned five treatment sessions given over 1½ wk. It is likely that a longer treatment period is necessary to distinguish true response from placebo effects which tend to be immediate but transient (Papakostas et al.2006). Strengths of our study include the detailed characterization of participants' depressive illness, and comprehensive assessments by blinded raters of the effects of tDCS on mood and neuropsychological function.
In conclusion, this study found that active tDCS had no adverse effects on neuropsychological function but was not superior to a robust placebo response, which occurred in the context of the excellent blinding that is possible with sham tDCS. Future studies may be more likely to demonstrate a difference between active and sham tDCS by the use of higher stimulation parameters (e.g. >1 mA current amplitude, daily treatment sessions) and testing tDCS in a more treatment-resistant sample.
The authors thank Mr Dusan Hadzi-Pavlovic for advice in statistical analysis. Funding for this study was provided by a National Health and Medical Research Council (Australia) Project Grant (no. 510142) and Program Grant (no. 510135). Trial registry name: Depression and Transcranial Direct Current Stimulation (tDCS). Registration identification number: NCT00256438 (http://clinicaltrials.gov/ct2/show/NCT00256438).
Statement of Interest