Article Text

Download PDFPDF

Research paper
Transcranial direct current stimulation for the treatment of post-stroke depression: results from a randomised, sham-controlled, double-blinded trial
  1. Leandro C L Valiengo1,2,
  2. Alessandra C Goulart1,
  3. Janaina F de Oliveira1,
  4. Isabela M Benseñor1,
  5. Paulo A Lotufo1,
  6. Andre R Brunoni1,2
  1. 1Center for Clinical and Epidemiological Research, University Hospital, University of São Paulo, São Paulo, Brazil
  2. 2Service of Interdisciplinary Neuromodulation, Laboratory of Neurosciences (LIM-27), Department and Institute of Psychiatry, University of São Paulo, São Paulo, Brazil
  1. Correspondence to Dr Andre R Brunoni, Av. Prof Lineu Prestes, 2565, 3o andar, CEP 05508-000, Hospital Universitário da Universidade de São Paulo, São Paulo 1.40302e+006, Brazil; brunoni{at}usp.br

Footnotes

  • Contributors LCLV and ARB 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. LCLV, ACG and ARB designed the study. LCLV, JFO and ARB collected the study data. LCLV and ARB analysed the data and wrote the first draft of the paper. All authors discussed the results and commented on the manuscript at all stages.

  • Funding This work was supported by a research grant from São Paulo Research State Foundation (Fundação de Amparo à Pesquisa do Estado de São Paulo, FAPESP Grant Number 2011/22872-4; 2012/20911-5), awarded to ACG. FAPESP is an independent public foundation and had no role in any aspect of the study, including: design and conduct of the study; collection, management, analysis and interpretation of the data; and preparation, review or approval of the manuscript. ARB was supported by the following grants: 2013 NARSAD Young Investigator from the Brain & Behavior Research Foundation (grant number 20493), 2013 FAPESP Young Researcher from the São Paulo State Foundation (grant number 20911-5), and National Council for Scientific and Technological Development (CNPq, grant number 470904). ARB also received equipment from Soterix Medical (not used in this study). LCLV was awarded a research grant from Stanley Medical Foundation.

  • Competing interests None declared.

  • Ethics approval Brazilian National Ethics Committee.

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

  • Data sharing statement Owing to the ethically sensitive nature of the research, supporting data cannot be made openly available. Anonymised data can be provided on request.

View Full Text

Statistics from Altmetric.com

Introduction

Stroke is prevalent worldwide, with significant morbidity and mortality. Post-stroke depression (PSD) occurs in about 40% of patients with acute stroke.1 These patients experience cognitive deficits, poor recovery in daily living activities, long hospitalisations and high mortality.1 Pharmacological treatments for stroke have shown mixed findings; a recent meta-analysis suggested that antidepressant drugs were only marginally superior to placebo.2 Also, pharmacotherapy might be limited by contraindications, pharmacokinetic interactions and severe adverse effects, particularly in older individuals.3 Alternatively, transcranial direct current stimulation (tDCS) might be a potential non-invasive treatment for PSD.

The tDCS technique alters neuronal resting membrane potentials to facilitate (anodal) or inhibit (cathodal) neuronal firing rates.4 ,5 The antidepressant effects of tDCS are based on findings that, in depression, the left dorsolateral prefrontal cortex (DLPFC) is hypoactive and the right DLPFC is hyperactive.6 To achieve antidepressant effects, anodal tDCS is delivered over the left region for increasing cortical excitability, whereas cathodal tDCS is delivered over the right DLPFC for decreasing cortical excitability. TDCS might also act by increasing cortical activity, which may reverse the ‘bottom-up’ pattern of cortical hypoactivity/subcortical hyperactivity observed in some forms of depression.7 Since the pathophysiological basis of PSD might involve a disruption of several frontostriatal neural circuits involved in mood regulation,1 we reasoned that tDCS over the DLPFC may ameliorate post-stroke depressive symptoms.

Consistent with that notion, repetitive transcranial magnetic stimulation (rTMS) has been used to treat PSD8 and vascular depression.9 High-frequency rTMS, a technique that also increases (although by different mechanisms) cortical excitability, ameliorated depressive symptoms when applied over the left DLPFC.10 Furthermore, controlled clinical trials11 ,12 and a meta-analysis13 showed that tDCS improved depression symptoms. We also previously reported that tDCS was associated with mood improvements in PSD.14 ,15

Here, we conducted a randomised, sham-controlled trial (RCT) to assess the safety and efficacy of tDCS for treating PSD. Our main hypothesis was that tDCS would produce a significant reduction in depressive symptoms compared to sham stimulation.

Material and methods

This study, reported according to the CONSORT guidelines,16 was conducted at the University Hospital, University of São Paulo, São Paulo, Brazil. Participants were actively recruited from October 2012 to September 2014 (clinicaltrials.gov identifier NCT01525524). All participants provided written informed consent. The local institutional review board approved the study. We used a randomised, sham-controlled, double-blind trial design, with a 1:1 permuted block randomisation. Allocations were concealed with a central randomisation method. Raters, operators and patients were blinded to treatment allocations. Contact between participants was avoided to enhance study blinding.

This 6-week trial comprised two phases. First, in the acute treatment phase, patients received daily tDCS sessions on weekdays for 2 weeks (10 sessions). Second, in a follow-up phase, patients received one tDCS session every other week (12 total sessions). The study end point, chosen a priori based on our previous study,12 was 4 weeks after the 10th day of stimulation, that is, at the end of week 6.

No more than two non-consecutive missed visits were allowed during the acute treatment phase when a session was missed, an extra tDCS session was performed to complete the number of sessions. This procedure does not reduce tDCS efficacy according to previous data.17 However, when stimulation was disrupted for more than three consecutive days, the patient was excluded from the study (ie, a dropout).

Participants

Eligibility criteria (table 1) were chosen to include a relatively homogeneous and representative sample. Stroke was diagnosed by a trained physician and confirmed by both an anamnesis of a neurological condition (stroke) and a physical examination. Clinical stroke syndrome was defined as sudden neurological symptoms, compatible with ischaemic or haemorrhagic stroke, with possible neurological deficits, determined by physical examination and neuroimaging,18 which were unlikely to result from a transient ischaemic attack or non-stroke pathology. The diagnosis of stroke was confirmed with brain CT or MRI. Depression was diagnosed by a trained psychiatrist with the Mini-International Neuropsychiatry Interview (MINI) for Diagnostic and Statistical Manual Fourth Edition (DSM-IV) psychiatric disorders.19 Patients were recruited by referrals from neurological and rehabilitation centres and from a naturalistic stroke surveillance cohort study.20

Table 1

Study inclusion and exclusion criteria

Interventions

Stimulations were performed with a standard tDCS device (DC-Stimulator, Neuroconn, Ilmenau, Germany). We used a ‘bifrontal’ montage, where the anode and cathode were placed over the areas corresponding to, respectively, the left and right DLPFC (F3 and F4 areas, according to the international 10–20 electroencephalography system). Rubber electrodes were inserted into 25 cm2 saline-soaked sponges and fixed with a headband. The current intensity was 2 mA. All patients were at rest, seated in an isolated room without external stimuli, and did not perform any specific task during the sessions.

The sham condition consisted of a brief stimulation period (60 s), to mimic common skin effects experienced just after stimulation, followed by no stimulation during the remaining period, similar to sham conditions described in previous depression trials.11 ,12 Randomisation was conducted with an automated device that produced sham or active stimulation, according to a number code. Number codes were randomised by a research assistant not involved in any other aspect of the trial, and typed out by the study nurse, who was blinded to the group condition.

Assessments

The primary outcome was the HDRS-17 score. Secondary outcomes were clinical response (categorical, defined as ≥50% reduction from the baseline HDRS score), remission (categorical, defined as an end point HDRS score <8), and scores on the Montgomery-Åsberg Depression Rating Scale (MADRS), the Rankin scale and the Barthel index. Response and remission were estimated only at the end point (week 6). Treatment-resistant depression was assessed with the Massachusetts General Hospital Staging method.21

Safety was assessed with a tDCS adverse effects questionnaire;22 the Young Mania Rating Scale, to measure treatment-emergent mania and hypomania; and cognitive assessments (Frontal Assessment Battery, Mini-Mental State Examination, Montreal Cognitive Assessment, Digit Span forward and backward tests, Stroop tests, Trail Making A and B, and Symbol Digit test). The Barthel index assessed the degree of patient independence. The Rankin scale assessed the level of disability.

Blinding efficacy was assessed at the end point by asking raters and participants to guess the allocation group.

Mood scales were applied at all time points. The Rankin and Barthel scales were applied at all time points, except week 4. The tDCS adverse effects questionnaire was applied at the end of weeks 2 and 6 (end point). Cognitive assessments were applied at baseline (week 0) and at the end point.

Statistical analysis

All analyses were performed with Stata V.12 (StataCorp, College Station, Texas, USA). We implemented two-sided significance tests at the 5% significance level. Missing data were considered to occur at random. We performed an intention-to-treat analysis; missing values were imputed with the last observation carried forward (LOCF) method. The robustness of LOCF for the primary outcome was examined by comparing results with those from a mixed-model repeated-measures analysis (MMA).

The sample size was estimated from Jorge et al8 who applied rTMS for PSD. They observed 38.3% vs 13.3% reductions in depressive symptoms for the active and sham groups, respectively. For that effect size and α and β values of 5% and 20%, respectively, we estimated a sample of 40 participants. Assuming an attrition rate of 20%, we increased the total sample size to 48.

Baseline clinical and demographic variables were compared between groups with t-tests and χ2 tests for continuous and categorical variables, respectively. The primary outcome was analysed with a mixed, repeated measures analysis of variance model, with one dependent within-subject variable (HDRS score), one within-subject variable (time, 4 levels) and one between-subject variable (group, 2 levels). We compared the slopes of depression scores across time between groups.

Secondary continuous outcome analyses were similar to the HDRS analysis, except that logistic regressions were performed for categorical variables. We also estimated the number needed to treat (a measurement used to assess and compare the effectiveness of clinical interventions), based on the ORs for response and remission. The frequency of adverse events was compared between groups with the χ2 test or the Fisher exact test. To assess changes in cognitive performance, we used analyses of covariance, with the group as the independent variable, the baseline cognitive performance as the covariate, and the end point cognitive performance as the dependent variable. To identify predictors of response, we performed general linear models, with the difference between baseline and end point scores as the dependent variable and the group and the putative predictor as independent variables. We performed post hoc Bonferroni's correction for multiple comparisons. Finally, we dichotomised (above vs below the median) the variables ‘duration of stroke’, ‘duration of depressive episode’ and ‘depression onset after stroke’. Then we performed ANOVAs with the depression score change (baseline minus end point) as the dependent variable, and the group and each dichotomised variable as independent variables.

Results

Participants

Of ∼194 volunteers, 48 patients were included and 43 completed the study (figure 1).

Figure 1

Flow diagram for participant selection. ITT, intention-to-treat; HMDRS, Hindi adaptation of Mattis Dementia Rating Scale.

The groups had similar baseline clinical and demographic characteristics. Only 16.6% of patients were previously using antidepressants and required a drug washout (table 2).

Table 2

Baseline demographic and clinical characteristics of 48 patients with post-stroke depression assigned to receive active or sham tDCS

Primary outcome—HDRS-17

We observed a significant main effect of time (LOCF: F3,138=45.3; p<0.001; MMA: F3,138=66.7; p<0.001) and a significant interaction between time and group (LOCF: F3,138=4.57; p=0.004, MMA: F3,138=6.74; p<0.001). Active tDCS produced superior HDRS-17 scores compared to sham at the end point (LOCF: mean difference, 4.7 points; 95% CI 2.1 to 7.3; p<0.001; MMA: 3.35 (0.07 to 6.63), p=0.04), but not at 2 weeks (LOCF: mean difference, 1.6 points; 95% CI −1.9 to 5.3; p=0.35; MMA: 0.48 (−3.2 to 4.8), p=0.79) or at 4 weeks (LOCF: mean difference, −0.08 points; 95% CI −3.5 to 3.4; p=.96; MMA: 1.56 (−4.6 to 1.5), p=0.31; figure 2).

Figure 2

Primary study outcome. Active transcranial direct current stimulation (tDCS) showed a significantly greater improvement in depressive symptoms compared to sham tDCS. Depression was measured with the Hamilton Depression Rating Scale, 17-items (HDRS-17). The scores in the active group were significantly lower than in the sham group at the end point, but not at baseline, week 2 (T2), or week 4 (T4). Bars represent 95% CI.

Secondary outcomes

Montgomery-Åsberg Depression Rating Scale

We observed a significant interaction between time and group (F3,138=5.72; p=0.001) and a main effect of time (F3,138=52.63; p<0.001). Active tDCS produced superior MADRS scores compared to sham after 6 weeks (mean difference, −4.5 points; 95% CI −8.8 to −0.2; p=0.04), but not after 2 weeks (mean difference, −4 points; 95% CI −9.0 to 1.0; p=0.11) or 4 weeks (mean difference, −2.3 points; 95% CI −7.2 to 2.5; p=0.33; see online supplementary table S1).

supplementary data

Remitters and responders

The response rates were higher in the active group than in the sham group (37.5% vs 4.1%, respectively, OR=13.8, 95% CI 1.6 to 120, number needed to treat=3). Also, remission was achieved by 5 (20.8%) patients in the active group and none in the sham group (OR=7.9, number needed to treat=5, p=0.049).

Clinical global impression (CGI)—severity

For the CGI, a significant main effect of time was observed (F3,183=22.04; p=0.001). No significant effect was observed for group or interaction (F3,183=1.82; p=0.14).

Rankin and Barthel

No significant main or interaction effect was observed for group or time with the Rankin scale (Fs<0.86, Ps>0.35) or with the Barthel index (Fs<0.37, Ps>0.604).

Safety outcomes

There was no change in cognitive performance due to tDCS (see online supplementary table S2). The frequency of adverse events was not significantly different between active and sham groups at all time points (see online supplementary table S3). No serious adverse events were reported.

Predictor variables

No variable was identified as a potential predictor of response (see online supplementary table S4). Also, depression onset after stroke, duration of the depressive episode, and duration of the stroke episode, when handled as binary variables, did not influence the outcome (see online supplementary figure S1).

Integrity of blinding

Neither patients nor raters correctly guessed the allocation group beyond chance (see online supplementary table S5).

Discussion

This study was the first randomised, sham-controlled trial to evaluate the efficacy of tDCS for PSD. Active tDCS was superior to sham in improving depressive symptoms and also presented higher response and remission rates.

Among the strengths of this study were the effective blinding and low attrition rate. The use of an antidepressant-free sample increased the internal validity of our study, because we could exclude the confounding effects of pharmacotherapy. However, this also limited study generalisability in some scenarios. For instance, although this might not be the case in low-income and middle-income countries (eg, only 16.6% of our sample was using antidepressants prior to the drug washout), physicians in high-income countries are more likely to evaluate patients who initiated selective serotonin reuptake inhibitors (SSRIs) immediately after a stroke, in accordance with a recent study that showed that motor recovery was enhanced in patients with stroke after 3 months of early administration of fluoxetine combined with physiotherapy.23 In fact, the effects of tDCS seem to be enhanced when combined with SSRIs. In our previous study,12 the use of sertraline combined with tDCS improved depression symptoms more effectively than either treatment alone. Also, in mechanistic studies, acute and chronic administration of citalopram boosted the efficacy of tDCS, which resulted in strengthening LTP-like glutamatergic plasticity.4 ,5

Although antidepressants are the first choice to treat major depression, 30% of the patients will not respond to four different pharmacological interventions.24 Meta-analyses have shown that the NNTs of antidepressant treatments are 5 for response and 7 to 9 for remission25–27 and, for rTMS, 6 for response and 8 for remission.28 In addition, a recent individual patient data meta-analysis for tDCS in primary depression showed NNTs of 7 and 9 for response and remission, respectively.29 In the present study, tDCS antidepressant effects were higher as the NNTs were lower. Therefore, tDCS can be a therapeutic option for patients with PSD as it seems to be at least as effective as other treatments, and was also well tolerated. TDCS can also be used as an adjuvant with other pharmacological treatments to enhance the clinical efficacy of both interventions.

Our sample was relatively homogeneous in the duration of depressive and stroke episodes and in depression onset after stroke. One explanation for this homogeneity might be that we recruited only patients with no prior depressive or stroke episodes. The relatively long duration of the depressive episode and the fact that only a few patients had previously taken antidepressant drugs might be explained by a lack of adequate healthcare in Brazil. For example, there may be barriers to healthcare access, underdiagnosis of depressive episodes or possibly mistaken diagnosis of symptoms as non-psychiatric stroke symptoms.

The median depression onset was 3 months after stroke; therefore, our cases exhibited an aetiology that was more likely associated with vascular damage, rather than with disability and lifestyle changes.30 However, that distinction was limited, because it is difficult to disentangle ‘biological’ and ‘psychological’ causes. Moreover, previous studies showed that the anatomical correlates of PSD changed over time,30 based on short-term (3–6 months) compared to long-term (1–2 years) outcomes. In addition, the natural history of PSD is dynamic; 50% of patients recover from depression in the first year, and 38% of cases experience recurrence after the second year.31 Nonetheless, in this study, depressive episodes lasted a median of 7 months, and strokes lasted a median of 12 months; thus, we believe that, under these conditions, our results were unlikely to be confounded by de novo post-stroke depressive episodes after remission from a previous episode. Moreover, we excluded potential de novo cases, based on our psychiatric interview.

Finally, our analyses showed that tDCS was effective, regardless of the duration of depression or stroke episodes or the time of depression onset after the stroke.

In an RCT of 20 patients, Jorge et al8 demonstrated that high-frequency rTMS was an effective treatment. However, compared to rTMS, tDCS does not require specialised personnel for operation, it is more affordable, and its maintenance costs are lower;32 therefore, tDCS might be more available than rTMS. Another advantage of tDCS is its lack of association with the risk of inducing seizures in adults, which was previously reported for rTMS. This might be particularly important for patients with stroke.

Our rates of tDCS efficacy were slightly higher than previously observed,11 ,13 but similar to our earlier study,12 which also used a bifrontal tDCS montage. In contrast, in other RCTs, the cathode was placed either over the contralateral supraorbital area (SO) or in F8 (which is more lateral to the right DLPFC than F4). Theoretically, the bifrontal montage should induce greater depression improvement than the SO or F8 approaches, because the bifrontal montage modulates both sides of the prefrontal cortex simultaneously. Nonetheless, a modelling study showed that both F4 and F8 cathode positioning could strongly stimulate this brain region.33 Moreover, another RCT did not find superior efficacy with the bifrontal montage, although that result was more likely due to the high treatment resistance of that sample.34

Consistent with findings in our previous study,12 we observed a significant active-to-sham difference at week 6, but not at week 2. Both studies also found that the placebo response decreased and the active response increased from weeks 2 to 6. We speculate that the higher placebo effect at week 2 compared to week 6 might be due to patient interactions with staff during daily tDCS sessions, which stopped after 2 weeks. Moreover, the increasing active response over time suggested that tDCS antidepressant effects might have an onset similar to that of standard antidepressant drugs; that is, it might take several weeks for full manifestation. In fact, the antidepressant effects of tDCS are probably related to neuroplastic changes in brain areas involved in MDD pathophysiology, including modulation of the serotonin system, enhancement of neuroplasticity in cortical areas and top-down neuromodulation in subcortical areas. These potential mechanisms involve long-term changes in plasticity, which could explain why tDCS required a long time frame for the manifestation of antidepressant effects.

We have not followed up our patients after the study end point due to their physical limitations in returning to our clinical centre. Therefore, we were not able to verify the extent to which the tDCS antidepressant effects lasted in those responding to treatment. In fact, there are only two studies addressing this issue. Valiengo et al35 followed up, for 6 months, 42 patients who responded to tDCS, showing a relapse rate of 53% at the end of the follow-up period. A similar relapse rate was observed in another study that followed 26 tDCS responders for a similar period.36

A recent meta-analysis of individual patient data showed that higher tDCS ‘doses’ (including the number of sessions, current density and session duration) were associated with greater depression improvement.29 Therefore, a longer treatment period (more than 10 consecutive sessions) might have achieved greater improvement. However, in this proof-of-concept study, daily sessions were conducted for only 2 weeks to avoid a high dropout rate. We reasoned that more returns to the clinical centre could have been particularly difficult for patients with physical disabilities. An alternative approach might be to develop devices that allow domiciliary tDCS use.

Although our patients were at rest during tDCS sessions, a recent study suggested that tDCS combined with online cognitive control therapy might be more effective than either treatment alone.37 Since this was the first PSD study to investigate tDCS efficacy, we used the approach used in previous tDCS trials to facilitate comparisons. However, the combination intervention should be investigated in future studies.

No predictors of response were identified, including pharmacotherapy use and stroke location. However, tDCS is a relatively non-focal technique; although the targeted area is the left DLPFC, other regions in the prefrontal cortex are also modulated.33 Also, cathodal-right tDCS could have induced antidepressant effects in patients with left-sided lesions, possibly reducing the influence of stroke location (in terms of laterality) as a predictor of response.

We observed no improvement in motor symptoms. However, we did not assess motor cortex neuroplasticity in our sample. In contrast, Player et al38 showed that frontal tDCS increased neuroplasticity in the motor cortex of patients with depression. Interestingly, we observed overall improvement in CGI over time in both groups. The CGI scale provides a single value for the clinician's subjective impression on the global severity of the condition. Since these patients also exhibited non-depressive symptoms related to the stroke, the CGI might have been underpowered for detecting depression-specific improvements. Moreover, at baseline, our patients were only moderately severe on the CGI scale; therefore, a ceiling effect could have occurred.

Limitations

First, we could not perform MRI scans in all patients at baseline, because it was not available. Therefore, we could not investigate whether the tDCS response could be predicted by predictors identified previously for rTMS efficacy in PSD, such as the degree of frontal atrophy, the severity of subcortical ischaemic changes, or changes in grey and white matter volumes.8 Second, we did not simulate the current distribution in computer models to investigate whether the predicted current flow might be associated with depression improvement. Third, our sample size was small; therefore, some secondary analyses might have been underpowered.

Conclusion

TDCS was an effective, safe treatment for PSD. It did not induce cognitive impairment or manic symptoms. Our results warrant further replication in a larger sample with a longer follow-up period to delineate more clearly the effects of treatment.

Acknowledgments

The authors thank Cibele Soares (study nurse) and Roberta Ferreira de Mello (study coordinator) for their indispensable help in conducting this study.

References

View Abstract

Footnotes

  • Contributors LCLV and ARB 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. LCLV, ACG and ARB designed the study. LCLV, JFO and ARB collected the study data. LCLV and ARB analysed the data and wrote the first draft of the paper. All authors discussed the results and commented on the manuscript at all stages.

  • Funding This work was supported by a research grant from São Paulo Research State Foundation (Fundação de Amparo à Pesquisa do Estado de São Paulo, FAPESP Grant Number 2011/22872-4; 2012/20911-5), awarded to ACG. FAPESP is an independent public foundation and had no role in any aspect of the study, including: design and conduct of the study; collection, management, analysis and interpretation of the data; and preparation, review or approval of the manuscript. ARB was supported by the following grants: 2013 NARSAD Young Investigator from the Brain & Behavior Research Foundation (grant number 20493), 2013 FAPESP Young Researcher from the São Paulo State Foundation (grant number 20911-5), and National Council for Scientific and Technological Development (CNPq, grant number 470904). ARB also received equipment from Soterix Medical (not used in this study). LCLV was awarded a research grant from Stanley Medical Foundation.

  • Competing interests None declared.

  • Ethics approval Brazilian National Ethics Committee.

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

  • Data sharing statement Owing to the ethically sensitive nature of the research, supporting data cannot be made openly available. Anonymised data can be provided on request.

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.