Transcranial direct current stimulation (tDCS) combined with cognitive training in adolescent boys with ADHD: a double-blind, randomised, sham-controlled trial

Abstract Background Transcranial direct current stimulation (tDCS) could be a side-effect-free alternative to psychostimulants in attention-deficit/hyperactivity disorder (ADHD). Although there is limited evidence for clinical and cognitive effects, most studies were small, single-session and stimulated left dorsolateral prefrontal cortex (dlPFC). No sham-controlled study has stimulated the right inferior frontal cortex (rIFC), which is the most consistently under-functioning region in ADHD, with multiple anodal-tDCS sessions combined with cognitive training (CT) to enhance effects. Thus, we investigated the clinical and cognitive effects of multi-session anodal-tDCS over rIFC combined with CT in double-blind, randomised, sham-controlled trial (RCT, ISRCTN48265228). Methods Fifty boys with ADHD (10–18 years) received 15 weekday sessions of anodal- or sham-tDCS over rIFC combined with CT (20 min, 1 mA). ANCOVA, adjusting for baseline measures, age and medication status, tested group differences in clinical and ADHD-relevant executive functions at posttreatment and after 6 months. Results ADHD-Rating Scale, Conners ADHD Index and adverse effects were significantly lower at post-treatment after sham relative to anodal tDCS. No other effects were significant. Conclusions This rigorous and largest RCT of tDCS in adolescent boys with ADHD found no evidence of improved ADHD symptoms or cognitive performance following multi-session anodal tDCS over rIFC combined with CT. These findings extend limited meta-analytic evidence of cognitive and clinical effects in ADHD after 1–5 tDCS sessions over mainly left dlPFC. Given that tDCS is commercially and clinically available, the findings are important as they suggest that rIFC stimulation may not be indicated as a neurotherapy for cognitive or clinical remediation for ADHD.

Psychostimulants are the gold-standard treatment for improving ADHD symptoms, but have associated side-effects [10], poor adherence in adolescence [11,12], and are not indicated for all individuals with ADHD [10]. Evidence of longer-term efficacy is also limited [12,13], possibly due to brain adaptation [14]. Meta-analyses indicate small to moderate efficacy of behavioural therapies, cognitive training (CT), neurofeedback, or dietary interventions on ADHD symptoms [15]. Non-invasive brain stimulation techniques, however, are promising given their potential to stimulate key dysfunctional brain regions associated with ADHD, with potentially longer-term neuroplastic effects that drugs cannot offer [3,[16][17][18][19]. Transcranial direct current stimulation (tDCS) is particularly well-suited for paediatric populations as it is user-friendly, well tolerated with minimal side effects [20], and is cheaper relative to other techniques, such as transcranial magnetic stimulation [21].
In tDCS, a weak direct electrical current is delivered through two electrodes placed on the scalp (one anode, one cathode), generating subthreshold, polarity-dependent shifts in resting membrane potentials in underlying brain regions. The resulting net increase (predominantly under the anode) or decrease (predominantly under the cathode) in neuronal excitability leads to modulation of the neuronal network [22], with neurophysiological effects persisting after stimulation, presumably by potentiating mediators of practice-dependent synaptic plasticity, including GABA, glutamate [23,24], dopamine, and noradrenaline [25][26][27].
Evidence of cognitive performance and clinical improvement following tDCS is, however, limited. Two meta-analyses of tDCS studies stimulating mainly the left dlPFC in 1-5 sessions in children and adolescents with ADHD indicate a modest improvement in inhibitory control measures [28,29] with the later one showing non-significant improvement on processing speed and inhibitory measures with no effects on attention measures [19]. Only two sham-controlled studies tested ADHD symptoms, reporting improvement in inattentive symptoms, but not impulsivity/hyperactivity, immediately, one [30,31], and two weeks [31] after anodal tDCS over left or right dlPFC.
There is evidence that tDCS effects can be enhanced when combined with CT by functionally priming the brain regions that mediate the cognitive function being trained [32][33][34][35]. Multi-session anodal tDCS combined with CT has also been shown to elicit longer-term cognitive improvements in healthy controls for up to one [36,37], nine [38], or 12 months [34], and clinical improvements in psychiatric disorders for at least one month [39][40][41], suggesting longer-term neuroplastic effects.
Most studies in ADHD have targeted the left dlPFC. However, one of the most consistent findings of meta-analyses of functional magnetic resonance imaging (fMRI) studies is the underactivation of rIFC during tasks of cognitive control [3,[7][8][9]42]. The rIFC is a cognitive control "hub" region, playing a key role in cognitive control [43,44], sustained attention [45][46][47], and timing networks [48], mediating key functions of impairment in ADHD [2,49]. The rIFC dysfunction has also been shown to be disorder-specific to ADHD compared to several neurodevelopmental disorders in comparative fMRI meta-analyses of . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.07.20245456 doi: medRxiv preprint cognitive control [3,8,9]. Further, its upregulation is the most consistent effect of single-dose and longer-term psychostimulant medication [9,50].
Only three published studies applied tDCS over rIFC in ADHD, in 1 or 5 sessions.
One found improvements after anodal relative to sham tDCS in Flanker task errors and intrasubject response variability in 21 adolescents with ADHD [51], but no effect in 14 children with ADHD on a combined n-back and Stop task with high definition (HD) or conventional anodal tDCS [52]. In 20 undiagnosed high-school students with ADHD symptoms, anodal compared to sham tDCS improved Go accuracy in a Go/No-Go (GNG) task but no other inhibitory or Stroop task measures [53]. To our knowledge, no shamcontrolled study so far has measured clinical and cognitive effects of multi-session anodal tDCS over rIFC in combination with CT in patients with ADHD In this double-blind, sham-controlled, randomised controlled trial (RCT), 50 children and adolescents with ADHD were administered 15 consecutive weekday sessions of anodal or sham tDCS over rIFC combined with CT of EF typically impaired in ADHD. The primary outcome measures were parent-rated clinical symptoms and cognitive performance in an inhibition and a sustained-attention task. Secondary outcome measures were other clinical, safety, and EF measures.
Given some evidence of clinical and cognitive improvements with anodal tDCS over the dlPFC [30,31] or rIFC [51,53] and prolonged effects when tDCS is paired with CT, we hypothesised that, compared to sham, at posttreatment, multi-session anodal tDCS of rIFC with multi-EF training would show greater improvement in ADHD symptoms and/or performance on EF tasks mediated by rIFC and targeted by the CT. We also hypothesised that clinical and EF task improvement would persist 6-months after posttreatment. Finally, we hypothesised no side or adverse effects.

MATERIALS AND METHODS
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.07.20245456 doi: medRxiv preprint

Trial Design
This pre-registered (ISRCTN: 48265228), double-blind, sham-controlled, parallel RCT compared multi-session anodal versus sham tDCS over the rIFC combined with multi-EF training. Randomisation to stimulation group was stratified by age (10 to 14.5 and 14.5 to 18 years) and medication status (naïve and non-naïve) using Smith randomisation [54,55] conducted by Innosphere Ltd (Haifa, Israel). Experimenters, participants, and parents/caregivers were blind to stimulation group. This trial received local research ethics committee approval (REC ID: 17/LO/0983) and was conducted in accordance with the Declaration of Helsinki and Consolidated Standards of Reporting Trials (CONSORT) guidelines [56].

Participants
Fifty male participants (10 to 18 years) were recruited from South London clinics, social media, and parent support groups from February 2018 to 2020. All participants had a clinical DSM-5 diagnosis of ADHD established by a clinician, and validated by the Schedule of Affective Disorders and Schizophrenia for School-Age Children-Present and Lifetime version (K-SADS-PL) [57] and the Conners 3rd Edition-Parent (Conners 3-P, cut-off t-score > 60) [58]. Autism spectrum disorders (ASD) was excluded using a combination (both required) of both the parent-rated Social Communication Questionnaire (SCQ, cut-off > 17) [59] and the pro-social scale of the Strengths & Difficulties Questionnaire (SDQ, cut-off < 5) [60]; for participants falling outside these criteria, the absence of ASD was confirmed by is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.07.20245456 doi: medRxiv preprint to tDCS (e.g., metallic implants [except dental braces], previous neurosurgical procedures, history of migraine, diseased/damaged skin on the scalp), treated with drugs that lower seizure thresholds (e.g., antipsychotics, tricyclic antidepressants), and an inability to provide consent from the legal caregiver for children under 16-years or from participants over 16years ( Figure 1). Participants received £540 plus travel expenses for participating.
Baseline assessment was scheduled at least two weeks after medication titration.
Thirty-two participants received stable ADHD medication (non-psychostimulants: 2; psychostimulants: 30; between 21 weeks and 10 years). To help avoid medication effects from masking tDCS effects, 14 of the 32 participants followed our optional request to abstain from psychostimulants at least 24 hours before each assessment time point. A further seven participants abstained for trial duration (i.e., baseline to posttreatment). One participant stopped taking medication 3 months before follow-up. Before commencing each game, participants chose a game to play from three options; the . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

Intervention
The copyright holder for this preprint this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.07.20245456 doi: medRxiv preprint ACTIVATE TM programme tracked choices and constrained future options so that multiple EF were twice as likely to be trained as working memory [62]. The training tracking Stop task trained motor response inhibition by requiring participants to cancel a prepotent motor response to a go signal that is followed shortly by an unexpected and rare (30% of trials) stop signal [63]. The delay between go and stop signals is dynamically adjusted to the participant performance, with better inhibition resulting in longer stop signal delays thus increasing the difficulty to withhold a response. Participants were encouraged to wait for the stop signal before responding to the go signal, thereby training inhibition and waiting behaviour/response delay (see Supplement).  The saline solution was held in tufts by capillary forces between strands, which prevented it from spreading/dripping. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 9, 2020.  Table 1-5 and Figures 1-3).

ADHD symptoms and related impairments. Caregiver-rated Conners 3-P ADHD index [58] measured ADHD symptoms, Weekly Parent Ratings of Evening and Morning
Behaviour-Revised scale (WREMB-R) [78] and the Columbia Impairment Scale-Parent version (CIS) [79] measured ADHD related difficulties and functional impairments, the . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 9, 2020. child-and caregiver-rated Affective Reactivity Index (ARI) measured irritability [80], and the child-rated Mind Excessively Wandering Scale (MEWS) measured mind-wandering [81].

Safety, feasibility, and tolerability measures. Safety was measured with parent-
rated questionnaires on side effects [82] and adverse events [83]. Participants rated their mood, wakefulness [84], and the tolerability of tDCS [85]. Parents and participants rated the overall impression of tDCS and CT using a rating scale from 0 (not at all/never) to 4 (definitely/always) designed specifically for this study.
EEG. EEG was measured at rest and during the GNG task at each assessment timepoint but will be reported in a separate publication.
Training outcome measures. The ACTIVATE TM DV was the highest game level reached each week averaged across three games that loaded on multiple EF (i.e., Treasure Trunk, Magic Lens, Printer Panic). The Stop task DV was PI on average for each week.

Assessment Time Points
Primary and secondary outcome measures were collected at baseline, posttreatment, and six-month follow-up. Adverse events and overall impression of tDCS and CT were measured at posttreatment. Ratings of mood, wakefulness, and the tolerability of tDCS were measured in each stimulation session. Baseline and posttreatment sessions were scheduled up to three weeks before the first and after the last stimulation session, respectively.

Statistical Methods
Confirmatory analysis. Repeated measures analysis of covariance (ANCOVA) tested group differences across posttreatment and follow-up, covarying for age in months, medication status (naïve, on-, or off-medication), and baseline value of outcomes where applicable. Sensitivity analyses were also conducted to remove statistical outliers on . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.07.20245456 doi: medRxiv preprint cognitive tasks, participants who changed medication between posttreatment and follow-up, or participants whose treatment spanned over four instead of three weeks (see Supplement Benjamini-Hochberg p-value adjustment [86] was applied to control for multiple testing. Only results based on adjusted p-values are reported, see Tables 3-4  Medication status. Medication status was coded as a categorical covariate with three groups: medication-naïve, on-medication, and off-medication. Participants who abstained for the treatment trial period were coded as off-medication, participants who abstained at assessment time points only were coded as off-medication for cognitive outcomes and onmedication for clinical outcomes (which can cover three weeks).

Missing data & statistical outliers.
In primary and secondary outcomes, only posttreatment WM task data for one participant were missing. Missing stop signal task data (1.33%) were random and replaced by group averages.

Baseline demographics
Compared to sham tDCS, the anodal tDCS group was significantly younger, had fewer years in education, higher ADHD-RS Total scores and ODD symptoms, and worse performance on the Macworth Clock, Time Discrimination, and list sort working memory tasks and during CT spent significantly more time playing the Peter's Printer Panic game (Tables 1 and 2).

Primary outcome measures
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 9, 2020. ; Cognitive. There were no significant effects on primary cognitive outcome measures after adjusting for multiple testing (see Table 3 and Figure 3).  Table 3 and Figure 4).

Secondary outcomes measures
Cognitive. The were no significant effects on secondary cognitive outcome measures after adjusting for multiple testing (See Table 4).  Table 4). is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 9, 2020. ; Commissions, Mackworth Clock % Omissions, WM, Response Variability, and Prematurity (baseline versus follow-up only) (see Table 9 in Supplement). In clinical measures, both groups improved in ADHD-RS and Conners 3-P Index (baseline versus posttreatment or follow-up); and in ARI Parent and Child, WREMB-R, and CIS (baseline versus posttreatment only).
Group differences in CT performance across the three weeks were explored with repeated measures ANCOVAs covarying for baseline, age in years, medication status (naïve, on-, or off-medication) and total time spent playing each game. There were significant effects in ACTIVATE TM or Stop task after adjusting for multiple testing (see Table 10 in Supplement).
We also explored if outcome changes that showed a significant time effect across both groups from baseline to posttreatment or follow-up were correlated with changes in CT performance scores (week 3 minus week 1). No correlations were significant (Table 11 in Supplement).
Given that age was not matched between groups, we conducted an post-hoc moderation analysis [87], predicting a change in ADHD-RS (baseline minus posttreatment) in a regression analysis by stimulation group, age, and a stimulation by age interaction. While the stimulation by age interaction was not significant (β=0.2, SE=.12, t(46)=1.7, p=.096), the simple effects showed a significant reduced improvement for anodal tDCS versus sham in older participants (1SD above mean age; Safety, feasibility, tolerability, and blinding integrity.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 9, 2020. ; There were no significant group differences in ratings of mood, wakefulness, overall impression of tDCS and CT (Tables 12 and 13 in Supplement), and side effects (Table 4).
Adverse effects were significantly higher in the anodal versus sham tDCS group at posttreatment (F 1,44 =4.09, p=.05, η p 2 =.08), driven mainly by higher parent-ratings for the items "he seems more grumpy and irritable", "has little appetite", and "has more problems falling asleep" (Table 4). Tolerability ratings showed that stimulation was well tolerated, with only significantly higher reports of burning sensation during anodal than sham tDCS (see Table 14 in Supplement). Group assignment guesses did not exceed chance level for Side effects did not differ, but at posttreatment, higher adverse effects relating to mood, sleep and appetite were reported following anodal compared to sham tDCS.
The lack of an observable clinical or cognitive effect extend previous meta-analytic evidence of no significant cognitive effects and limited evidence of clinical effects in ADHD with 1-5 sessions of predominantly left dlPFC anodal tDCS [19]. These findings are unexpected given that rIFC underactivation is consistently associated with poor cognitive control, attention and clinical symptoms in ADHD [3,[7][8][9]. While the findings of no clinical effect of tDCS of rIFC are novel, the negative effects on cognition extend evidence from . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 9, 2020. ; prior 1 or 5 session sham-controlled tDCS studies stimulating rIFC in ADHD which showed no or moderate effects (see introduction) [51][52][53].
Findings are furthermore unexpected given evidence of a synergistic effect of combined CT and tDCS on improving cognition [32,34,88]. Although we covaried for age, one possible explanation for the negative findings on clinical symptoms and cognition is that the anodal tDCS group were significantly younger with larger baseline clinical and cognitive impairments compared to sham, both of which could have impaired learning [89]. This would be supported by evidence that ADHD children with worse neurocognitive skills at baseline show less CT gains [90] and neurofeedback learning [91][92][93], while in healthy controls poorer cognitive performance at baseline can lead to null and even detrimental effects of tDCS [94,95].
Alternatively, given the stronger electric field strengths in children than in adults [96,97], multiple sessions of tDCS may have triggered a homeostatic plasticity responsei.e., the amount and direction of plasticity was attenuated in response to excessive increases in neuronal excitability [98][99][100] -thereby temporarily disrupting the excitability of rIFC [99][100][101]. This is in line with our post-hoc moderation analysis that revealed older but not younger participants improved less in the ADHD-RS Total Scores in the anodal versus sham tDCS group at posttreatment. Future studies should verify if tDCS has differential effects depending on current strength and age of participants with ADHD.
Another possibility is that rIFC stimulation downregulated neighbouring dorsal prefrontal or parietal regions part of the dorsal attention network [7,102], or left hemispheric prefrontal regions that mediate positive emotions [103,104].
Interestingly, however, impulsiveness/hyperactivity symptoms, which are most closely associated with rIFC activation [3,105], were lower at follow-up in the anodal relative to the sham tDCS group. The finding -that needs replication -could suggest longer-term . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 9, 2020. ; neuroplastic consolidation effects as have been shown in other neurotherapies, such as neurofeedback [106,107].
Both groups improved in symptoms and cognitive performance from baseline to posttreatment or follow-up, which could suggest gains due to CT [62]; however, given the lack of correlation with CT performance, placebo effects cannot be ruled out.
The negative findings from this trial are crucial given that tDCS is being increasingly incorporated into clinical practice, is considered an acceptable alternative to medication by parents, and is already commercially available [17,108]. Particularly alarming is that parentrated ADHD symptoms and adverse effects were higher at posttreatment after anodal tDCS relative to sham.
Findings are not encouraging for the efficacy of multi-session tDCS of rIFC combined with CT in ADHD. However, there are limitations to the study. Although our sample of 50 participants is the largest sample of any tDCS study in children and adolescents with ADHD, larger studies may be more adequately powered to detect effects. We cannot rule out that positive results are achievable with other study designs and stimulation parameters. Computational current flow models suggest that higher stimulation intensities might be required to modulate clinical symptoms and cognitive functions mediated by rIFC given that this it is a deeper region compared to the commonly stimulated dlPFC [97].
Another limitation is that we could not test for weekly dose effects as ADHD symptoms were . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 9, 2020. ; only measured at baseline, posttreatment and follow-up; yet weekly changes in ACTIVATE TM game performance and stop task PI did not show dose effects.
Larger, double-blind, randomised-controlled trials should systematically investigate optimal and ideally individualised stimulation protocols (e.g., different stimulation sites, intensity, duration, number of sessions, etc.) measuring clinical, cognitive, and possible non-  . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 9, 2020. ; . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 9, 2020. ; is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 9, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 9, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 9, 2020.  is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 9, 2020. ADHD-RS, ADHD Rating Scale; PI, Probability of Inhibition; SD, Standard Deviation *Adjusted values as predicted by the repeated-measures ANCOVA testing group differences at posttreatment and follow-up, adjusting for baseline, age at entry and medication status (naïve, off-medication, on-medication). †Benjamini-Hochberg adjustment was applied to p-values for time, group, and time by group interaction effect separately and was applied separately to primary cognitive, secondary cognitive, and secondary clinical outcome measures separately. ‡Benjamini-Hochberg adjustment was not applied to these measures .

CC-BY 4.0 International license
It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 9, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 9, 2020. Behaviour-Revised *Adjusted values as predicted by the repeated-measures ANCOVA testing group differences at posttreatment and follow-up, adjusting for baseline, age at entry and medication status (naïve, off-medication, on-medication). †Benjamini-Hochberg adjustment was applied to p-values for time, group, and time by group interaction effects separately and was applied separately to primary cognitive, secondary cognitive, and secondary clinical outcome measures separately. ‡Benjamini-Hochberg adjustment was not applied to these measures . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 9, 2020. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted December 9, 2020. ; is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted December 9, 2020.   is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.07.20245456 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted December 9, 2020. ;