Transcranial direct current stimulation and attention skills in burnout patients: a randomized blinded sham-controlled pilot study

Patients

Patients were recruited between January 2015 and December 2017 via a treatment center in Belgium specialized in the diagnosis and treatment of burnout (DIADIS NV, Oud-Turnhout). The definition of (Brenninkmeijer et al., 2001) was used to identify burnout patients, and a score of > 4 on the Dutch version of the Maslach Burnout Scale (MBS: Maslach-Pines, 2005) was considered an inclusion criterium. Patients with 1) excessive drug or alcohol use, 2) epilepsy, 3) depression, 4) bipolar syndrome, 5) chronic fatigue syndrome or any other history of psychiatric or neurological disorders, 6) implanted neurostimulator or pace-maker, 7) drugs interacting directly with the NMDA receptors, or 8) pregnancy were excluded. When new patients were diagnosed with burnout in the treatment center, they were asked whether they wanted to participate in the study. Included patients were pseudo-randomly assigned to a real atDCS or sham tDCS group using a pre-defined allocation code file in excel (to make sure that both groups were of equal size). Initially, 20 participants were targeted (10 per tDCS group) as a pilot study. This number was primarily based on practical issues, such as the average number of burnout patients that were treated every year at the treatment center, and the time the treating psychologist could devote to the study. All assessments were performed by the sole psychologist of the treatment center (PVN).

This study was approved by the ethical committee CME of the Vrije Universiteit Brussels (VUB) (B.U.N. 143201422009). All patients signed an informed consent. The trial was retrospectively registered at ISRCTN.com on 17/11/19 (ISRCTN94275121), since clinical trial registration was not explicitly required by the advising ethical committee for trials with an experimental device at the start of the trial. All protocol and trial details are available from the registration page.

Pretesting

After inclusion, baseline measures were taken to evaluate burnout, depression, quality of life, attention, and different components of the working memory. Burnout, depression, and overall quality of life were assessed by the MBS, the Beck’s Depression Inventory (BDI: Van der Does, 2002), and Question A of the Dutch version of the McGill Quality of Life Questionnaire (QoL: Cohen et al., 1997); translated by Kenniscentra Palliatieve Zorg) respectively.

Attention was measured by the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS: Randolph, 1998) Attention Index, and vigilance by the s-score of the D2 test (variability in processing speed).

The central executive of the working memory was evaluated with the following tests. Inhibition and shifting were assessed with Card III of the Stroop Color-Word test (Golden, 1978), the Trail Making Test part B (TMT: Reitan, 1958) and the Wisconsin Card sorting test (WCST: Heaton et al., 1993). Processing speed, i.e., updating and control, was assessed by the TMT part A, Cards I and II of the Stroop Color-Word test, and the D2 test (G_z: total number of tokens scanned; F%: error percentage relative to G_z; G_z – F: number of correctly identified tokens) (Brickenkamp, 1962).

As regards to the other components of the working memory: the phonological loop was tested by the Language Index of the RBANS, the Boston naming test (BNT: Kaplan et al., 1983; Flemish version BNT: Mariën et al., 1998) and semantic fluency tasks (naming as many animals, vegetables, means of transportation and clothes as possible within one minute). To determine the percentile of semantic fluency, Dutch non-published age-, gender-, and education-related norms were used (These data were obtained by master students in Linguistics at the VUB of 200 healthy participants in Belgium of varying age, gender, education, and geographic location and are available as extended data (van Dun, 2020)). These data were used to calculate the z-scores that were then converted to percentiles. The visuospatial sketchpad was assessed using the Raven’s progressive matrices (Raven, 1965), and the Visuospatial Index of the RBANS. Encoding was evaluated with the Immediate Memory Index and retrieval with the Recent Memory Index of the RBANS.

A categorized overview of the different tests is presented in Table 1.

After treatment, all tests were repeated to evaluate the impact of atDCS. Therapy always started on a Monday, and re-evaluation was completed the first Monday after the final atDCS session.

Primary and secondary outcome measures

The primary outcome measure was attention. Secondary outcome measures were general measures (burnout, depression, and quality of life), and other components of the working memory (central executive, phonological loop, visuospatial sketchpad, encoding and retrieval).

Treatment

All patients received the standard behavioral therapy consisting of one session a week (for 3 weeks) focusing on 1) psycho-education and relaxation, 2) reducing mental overload, 3) defining and working to personal goals, 4) relapse prevention. 1) In the first session, the stress mechanism was explained, together with the characteristics that belong to it. Breathing exercises were taught to the patient through heart rhythm coherence, using EmWave2 software to visually guide the patients. 2) To reduce the mental overload, ‘don’t worry’-techniques were explained. Patients were advised to write down their worries and not get distracted by them continuously. Via cognitive behavioral therapy, using the ABCDE model (Ellis et al., 1997), they were taught to translate negative into positive thoughts. 3) During therapy, the patient’s life goals in different domains (e.g. work, personal relations, education, parenthood, friends, physical well-being, …) were established together with the therapist. In dialogue, priorities were established and possible (mental) barriers were discussed. This discussion primarily focused on rebalancing the different domains in the patient’s life. 4) Lastly, the therapy focused on reintegration on the work floor. Bad habits were identified and strategies were discussed to prevent the patients from falling back into these habits.

None of the patients had received psychotherapy before inclusion in this study.

In addition, patients received daily sessions of 2mA atDCS (TCT Research Limited, Hong Kong) over the left DLPFC (AF3 on the international 10/20 EEG system) (electrode size: 5x5cm²) and the reference electrode (5x7cm²) over the lateral aspect of the contralateral orbit (F8), as described in (Loo et al., 2012). The carbon electrodes were covered in sponges soaked in saline solution (0.9% NaCl) to improve conductivity. These were placed over the scalp using neoprene straps. Since these can absorb the saline solution, two different straps were used for both electrodes to avoid creating bridges, and throughout the sessions the absorption was monitored so that it would not spread beyond the surface of the electrodes. In the real tDCS group, stimulation lasted for 20min with a gradual ramp up over 30s. This resulted in a maximal current density of 0.08mA/cm² and a total charge of 0.096C/cm² per session. Impedance was continuously monitored during stimulation to stay below 10kOhm and was automatically disrupted for safety when it went above 15kOhm. During sham stimulation, the current was ramped up over 30s to 2mA after which it was immediately ramped down to simulate the cutaneous sensation of tDCS in the sham group. No therapy was given during stimulation. This resulted in 15 sessions in total (3 weeks, 5x/week). One group received real tDCS, the other received sham tDCS. The tDCS device was programmed by the therapist, but the patients did not know which type of tDCS they received. All test results were coded to blind the researcher who performed the analyses and data was unblinded only after the analyses were done. The protocol and electrode placement are illustrated in Figure 2.

Figure 2.

A. Visualization of the protocol, and B. electrode placement of the tDCS protocol.

All therapy and tDCS sessions were performed at the treatment center DIADIS NV in Oud-Turnhout, where the patients were recruited, by the same psychologist and co-author Pia Van Noppen.

Statistical analyses

Means and standard deviations (SDs) were reported to give a general overview of the results. An independent samples t-test was used to compare mean age between groups.

A full-factorial 2 (tDCS: sham, real) x 2 (time: pre, post) fixed effects linear mixed model with subject as a random effect was used to compare the results on all test measures between the sham and real tDCS groups before and after treatment. Normality and homoscedasticity of the residual data were checked via a normal quantile plot and residual plot, respectively. If model assumptions were violated, the outcome variable was transformed using the Box-Cox procedure (Box & Cox, 1964), as implemented in the MASS package in R version 7.3-51.4. Tukey HSD post-hoc pairwise comparisons were used to compare baseline scores between groups and to explore possible interaction effects. The level of significance was set at α = 0.05. All statistical analyses and figures were generated using the statistical software R version 3.6.0.

Demographics

In total, 16 patients (11F, 5M) were recruited and received either real (n = 8) or sham (n = 8) treatment. Of these, 15 (10F, 5M) were included in the analysis. One participant (pp01, F, sham) was excluded after analysis because she was diagnosed with sensory processing sensitivity (SPS). Mean age of our final sample (n = 15) was 44.8y ± 5.8y, with no significant difference between the real (42.5y ± 5.5y) and sham group (47.4y ± 5.3y) (t(12.86) = 1.76, p = 0.103). No participants reported serious adverse events. Only one complained about dizziness at the end of the stimulation.

An overview of the demographic characteristics and the initial scores on the MBS, the Dutch version of the BDI, and question A of the Dutch version of the QoL are given in Table 2A (see underlying data (van Dun, 2020)). Table 2B contains additional demographic characteristics and working-related information. A flow chart is provided in Figure 3.

Figure 3. Flow chart of the enrollment procedure (CONSORT 2010 Flow Diagram).

Impact of tDCS on burnout, depressive symptoms, and quality of life

The linear mixed model revealed a significant effect of Time for the MBS (F(1, 13) = 15.10, p = 0.002), but no effect of tDCS (F(1, 13) = 0.00, p = 0.971) or an interaction (F(1, 13) = 0.73, p = 0.408) (see Figure 4A). Tukey HSD post-hoc multiple comparisons indicated that only the real tDCS group improved significantly on the MBS (real: t(13) = 3.886, p = 0.009; sham: t(13) = 2.465, p = 0.113).

Figure 4.

Mean pre- and postscores on A. the Maslach Burnout Scale (MBS), B. Beck’s Depression Inventory (BDI), and C. McGill Quality of Life (QoL) questionnaire for the sham (dotted, triangles) and real (dashed, circles) group with 95% confidence intervals. Continuous lines indicate main effects, dashed and dotted lines indicate a significant difference between the pre- and postscores of the separate group (real: dashed; sham: dotted) as found by the post-hoc analyses. NS = non-significant; * = p≤0.05; ** = p≤0.01.

For the BDI, only an effect of Time was found ((F(1, 13) = 7.93, p = 0.015) (see Figure 4B). Post-hoc analyses revealed that only the sham group improved significantly after the intervention (t(13) = 3.58, p = 0.016) and the real group demonstrated a tendency towards improvement (t(13) = 2.82, p = 0.062).

The linear mixed model revealed no significant effects or interaction for the McGill Quality of Life (QoL) questionnaire. However, post-hoc analysis did reveal a significant improvement for the real group (t(13) = -3.21, p = 0.031) (see Figure 4C).

No significant differences were found at baseline for these three measures (MBS: t(13) = 0.04, p = 1.000; BDI: t(13) = -0.73, p = 0.883; QoL: t(13) = 0.00, p = 1.000). All means, SDs, and p-values of the post-hoc analyses are listed in Table 3. The results of the linear mixed model can be found in Table 7.

Impact of tDCS on attention and vigilance

Means, standard deviations, and p-values of the post-hoc analyses are shown in Table 4. The linear mixed model revealed a significant interaction between Time and tDCS for the RBANS Attention Index (F(1,13) = 14.80, p = 0.048), where the real group improved significantly more than the sham group (real: t(13) = -3.85, p = 0.010; sham: t(13) = -0.61, p = 0.929) (see Figure 5A). No significant difference was detected in the baseline scores (t(13) = 0.36, p = 0.984).

Figure 5. Mean pre- and postscores on the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS) Attention index for the sham (dotted, triangles) and real (dashed, circles) group with 95% confidence intervals.

Continuous lines with X indicate interaction effects, dashed and dotted lines indicate a significant difference between the pre- and postscores of the separate group (real: dashed; sham: dotted) as found by the post-hoc analyses. X = interaction effect; NS = non-significant; * = p≤0.05; ** = p≤0.01.

A significant interaction effect was also found for vigilance (F(1,13) = 12.15, p = 0.004), as measured by the s-score of the D2 test, with the sham group improving significantly. However, the assumptions of homoscedasticity and normality of the residuals of the model were doubtful, but did not improve using the Box-Cox transformation, which makes it difficult to interpret the results. In addition, the real and sham tDCS group tended to differ significantly at baseline (t(13) = 2.82, p = 0.061), with the sham group performing worse than the real tDCS group.

Impact of tDCS on the central executive

All means, standard deviations, and p-values of the post-hoc analyses are shown in Table 5.

The linear mixed model demonstrated a significant effect of Time for inhibition and shifting on the Stroop Color-Word test (card III) (F(1,13) = 5.96, p = 0.030) (Figure 6A) and the WCST (F(1,13) = 9.20, p = 0.010) (Figure 6B). Post-hoc comparisons only revealed a significant improvement in the real tDCS group on the WCST (real: t(13) = -3.03, p = 0.042; sham: t(13) = -0.54, p = 0.948). For the Stroop (card III) no significant improvements were found for either group post-hoc (real: t(13) = -2.44, p = 0.118; sham: t(13) = -0.53, p = 0.951). For the TMT B, the assumption of homoscedasticity of the residuals was violated and did not improve using the Box-Cox transformation, making interpretation of the model difficult. No significant effects or interaction were found with the non-transformed data.

Figure 6.

Mean pre- and postscores for inhibition and shifting on A. the Stroop Color-Word test Card III, and B. the Wisconsin Card Sorting Test (WCST), for the sham (dotted, triangles) and real (dashed, circles) group with 95% confidence intervals. Continuous lines indicate main effects, dashed and dotted lines indicate a significant difference between the pre- and postscores of the separate group (real: dashed; sham: dotted) as found by the post-hoc analyses. NS = non-significant; * = p≤0.05; ** = p≤0.01.

No significant differences were found in the baseline measures (Stroop card III: t(13) = -1.78, p = 0.327; WCST: t(13) = -1.79, p = 0.323; TMT B: t(13) = 0.54, p = 0.947).

For updating and control, a significant effect of Time was found for the TMT A (F(1,13) = 7.69, p = 0.016) (Figure 7A), the Stroop Color-Word test (card I) (F(1,13) = 6.30, p = 0.026) (Figure 7B) and the D2 (G_z: F(1,13) = 7.38, p = 0.018; and G_z – F: F(1,13) = 8.10, p = 0.014) (Figure 7C and 7D). The post-hoc tests only revealed trends towards improvement in the real group for the TMT A (real: t(13) = -2.77, p = 0.067; sham: t(13) = -2.00, p = 0.238), the Stroop Color-Word test (card I) (real: t(13) = -2.51, p = 0.105; sham: t(13) = -0.24, p = 0.995), D2 G_z (real: t(13) = -2.72, p = 0.074; sham: t(13) = -1.02, p = 0.741), and D2 G_z – F (real: t(13) = -2.85, p = 0.059; sham: t(13) = -0.89, p = 0.812). No significant effects or interaction was found for the Stroop Color-Word test (card II) or F% of the D2 test. However, the assumption of homoscedasticity of the residuals was violated in the Stroop Color-Word test (card II), which might have resulted in unreliable p-values.

Figure 7.

Mean pre- and postscores for updating and control on the A. Trail Making Test (TMT) part A, B. Stroop-Color Word test Card I, C. D2 G_z score, and D. D2 G_z – F score for the sham (dotted, triangles) and real (dashed, circles) group with 95% confidence intervals. Continuous lines indicate main effects, dashed and dotted lines indicate a significant difference between the pre- and postscores of the separate group (real: dashed; sham: dotted) as found by the post-hoc analyses. NS = non-significant; * = p≤0.05; ** = p≤0.01.

No significant differences were found in the baseline measures (TMT A: t(13) = 0.62, p = 0.925; Stroop card I: t(13) = -1.11, p = 0.689; Stroop card II: ; D2 G_z: t(13) = -0.97, p = 0.771; D2 G_z – F: t(13) = -1.27, p = 0.596; D2 F%: t(13) = -0.52, p = 0.952).

Impact of tDCS on other working memory components

All mean scores, standard deviations, and p-values of the post-hoc analyses are listed in Table 6.

For the phonological loop, the linear mixed model revealed a main effect of Time for the BNT (F(1,13) = 12.92, p = 0.003) (Figure 8A) and the Language index of the RBANS (F(1,13) = 4.76, p = 0.048) (Figure 8B). Tukey HSD post-hoc comparisons revealed that only the real tDCS group improved significantly on the BNT (t(13) = -3.60, p = 0.015), a trend towards improvement was observed in the sham group (t(13) = -2.83, p = 0.060). No significant improvements were found for the groups separately for the RBANS Language Index (real: t(13) = -2.18, p = 0.180; sham: t(13) = 0.78, p = 0.863). No significant effects or interactions were seen for semantic fluency. Baseline scores did not differ significantly for these measures (BNT: t(13) = 0.23, p = 0.995; RBANS Language Index: t(13) = 1.72, p = 0.352; Semantic fluency: t(13) = 0.63, p = 0.921).

Figure 8.

Mean pre- and postscores for the phonological loop on the A. Boston Naming Test (BNT), and B. RBANS Language index for the sham (dotted, triangles) and real (dashed, circles) group with 95% confidence intervals. Continuous lines indicate main effects, dashed and dotted lines indicate a significant difference between the pre- and postscores of the separate group (real: dashed; sham: dotted) as found by the post-hoc analyses. NS = non-significant; * = p≤0.05; ** = p≤0.01.

No significant effects or interaction were observed for the visuospatial sketchpad (RBANS Visuospatial index and Raven). However, the assumptions for the linear mixed model of the Raven were violated, which could have affected the p-values. The Box-Cox transformation did not improve the data. Baseline scores did not differ significantly (RBANS Visuospatial Index: t(13) = 1.72, p = 0.452; Raven: t(13) = -0.16, p = 0.999).

A main effect of Time was observed for encoding, evaluated by the Immediate Memory index of the RBANS (F(1,13) = 11.93, p = 0.004) (Figure 9A). Post hoc analysis showed that this was mainly driven by a significant improvement of the real tDCS group (real: t(13) = -3.45, p = 0.020; sham: t(13) = -1.42, p = 0.508). Retrieval (RBANS Recent Memory index) also showed a significant effect of Time (F(1,13) = 8.58, p = 0.012) (Figure 9B). For retrieval, both groups trended towards significance (real: t(13) = -2.93, p = 0.051; sham: t(13) = -2.58, p = 0.094). No differences in baseline scores were observed for Immediate or Recent Memory (RBANS Immediate Memory Index: t(13) = 0.34, p = 0.986; RBANS Recent Memory Index: t(13) = -0.56, p = 0.943).

Figure 9.

Mean pre- and postscores for encoding on the A. Repeatable Battery for the Assessment of Neuropsychological Status (RBANS) Immediate Memory, and retrieval on the B. RBANS Recent Memory for the sham (dotted, triangles) and real (dashed, circles) group with 95% confidence intervals. Continuous lines indicate main effects, dashed and dotted lines indicate a significant difference between the pre- and postscores of the separate group (real: dashed; sham: dotted) as found by the post-hoc analyses. NS = non-significant; * = p≤0.05; ** = p≤0.01.

Underlying data

Harvard Dataverse: Transcranial direct current stimulation and attention skills in burnout patients: a randomized blinded sham-controlled pilot study. https://doi.org/10.7910/DVN/4VG2XS (van Dun, 2020)

This project contains the following underlying data:

Extended data

Harvard Dataverse: Transcranial direct current stimulation and attention skills in burnout patients: a randomized blinded sham-controlled pilot study. https://doi.org/10.7910/DVN/4VG2XS (van Dun, 2020)

This project contains the following extended data:

Reporting guidelines

CONSORT checklist and flow chart for “Transcranial direct current stimulation and attention skills in burnout patients: a randomized blinded sham-controlled pilot study”. https://doi.org/10.7910/DVN/4VG2XS (van Dun, 2020)

Data are available under the terms of the Creative Commons Zero “No rights reserved” data waiver (CC0 1.0 Public domain dedication).