The Effect of Prefrontal Transcranial Direct Current Stimulation on Attention Network Function in Healthy Volunteers

The effect of acute transcranial direct current stimulation (tDCS) on cortical attention networks remains unclear. We examined the effect of 20 min of 2 mA dorsolateral prefrontal cortex tDCS (bipolar balanced montage) on the efficiency of alerting, orienting and executive attention networks measured by the attention network test.


INTRODUCTION
The functional neuro-architecture of the attention system features three separable networks: alerting, orienting, and executive attention/control (1). The alerting system is spatially broad and facilitates distributed processing of temporally anticipated but not spatially localized events, and is associated with increased activity in thalamic, anterior, and posterior cortical sites networks. Orienting enables the selection/allocation of resources toward the spatial location of anticipated/salient stimuli and activates parietal cortex and frontal eye-fields. Executive control coordinates voluntary (over involuntary/automatic) responses and activates lateral and medial prefrontal cortex (PFC) and anterior cingulate cortices (2,3).
The PFC has long been considered to play a key role in the executive control of attention. Early animal behavioral studies (e.g., (4,5)) describe the effects of frontal damage as "a disruption of goal-directed behaviours" and recent clinical neuropsychological studies in humans with PFC lesions reveal selective deficits in executive attention (6)(7)(8)(9)(10). Neuroimaging studies reveal positive correlations between neuronal activity in subregions of PFC and executive control [for review, see (11)]. Taken together, extant evidence implicates PFC in executive attention control (12).
tDCS is a noninvasive brain stimulation method, which alters cortical tissue "excitability" through applying a weak (0.5-2 mA) constant direct current via scalp electrodes to the cortical region of interest (13). In contrast to other neurostimulation modalities, tDCS does not directly trigger action potentials in neuronal cells, but instead changes overall tissue excitability (14). Transcranial magnetic stimulation (TMS) of cortical attention networks selectively modulates corresponding behavioral performance measures (15). However, recent meta-analysis (16) suggests the effects of acute transcranial direct current stimulation (tDCS) remains unclear (16) and highlights the need for future studies to 1) include performance measures that can dissociate the selective effects of tDCS across attention networks and 2) examine state-dependent variables that might moderate the effects of tDCS on cognition [see (17)]. Stimulation of posterior parietal cortex modulates orienting networks (18), but a corresponding effect of PFC tDCS on executive attention has not been demonstrated (18). Other studies have examined the effects of tDCS on attention to emotional information and provide evidence that 20 min of active dorsolateral prefrontal cortex (DLPFC) tDCS with a bipolar balanced montage can reduce vigilance to threat (19), and that 17 min 1 mA monopolar stimulation of left PFC can enhance attention training procedures that direct attention away from threat stimuli (20).
We examined the effect of 20 min of 2 mA prefrontal DLPFC tDCS on the efficiency of alerting, orienting and executive attention networks measured by the attention network test (ANT) (21)-a cued reaction time flanker task that requires participants to make a swift response to central targets (flanked by distracter stimuli) cued by temporal-onset (alerting) and/or spatial (orienting) visual stimuli. The ANT is a well-established and widely used simple, computerized task that combines a flanker task and a cued-reaction time task. Participants make a speeded response to classify the direction of a central arrow, which is flanked by two pairs of arrows that either point in the same direction as the target arrow (congruent condition) or in the opposite direction (incongruent condition). The ANT has subsequently been used to characterize attentional deficits in alerting, orienting, and executive attention in healthy individuals, and at-risk/ clinical groups (e.g., (22)).
Consistent with findings from animal models, PFC lesion studies in humans (6)(7)(8)(9)(10)23) and functional neuroimaging evidence we predicted that a 20-min session of active prefrontal tDCS would selectively improve executive attention control compared to sham tDCS. Following recommendations from recent meta-analyses [e.g., 16] we used a robust physical and mental health screening procedure to match groups, examined subjective and autonomic state-dependent measures of mood/arousal at baseline and poststimulation, and examined retrospective expectancies about

METHOD Participants
Thirty healthy volunteers were recruited through online adverts and randomized (by gender) to receive either 2 mA active tDCS (n 5 15; 10 females, 5 males; mean age 5 20.8 years, SD 5 1.8) or nonactive sham stimulation (n 5 15; 11 females, 4 males; mean age 5 21.5 years, SD 5 2.9). Participants underwent a telephone health screen prior to the testing session, and a mental health screen on the day of the study session, using a structured diagnostic interview [Mini International Neuropsychiatric Interview] (24). A physical health checklist screened participants against current and lifetime physical illness exclusion criteria. Eligible participants were required to be aged 18-55 years. All participants were right handed. Exclusion criteria included metal or electronic implants, epilepsy, recent medication (past eight weeks bar topical treatment, paracetamol, oral, injectable, or skin patch contraception), pregnancy, elevated blood pressure (>140/90 mm Hg), cardiovascular disease, lifetime history of psychiatric illness/alcohol/drug dependence, current smoker, body mass index <18 or P28 kg/m 2 , and recent use of alcohol (confirmed by breath test). The research was approved by the Ethics and Research Governance Committee, Department of Psychology, University of Southampton and conducted in experimental laboratories with controlled lighting and temperature in the Department of Psychology, Highfield Campus, Southampton. All procedures complied with the Helsinki Declaration of 1975, as revised in 2008. Participants provided informed consent, were debriefed upon completion and compensated with either "study participation credits" (if university students) or with financial compensation at the rate of £6/hour.
The intervention was 20-min double-blind 2 mA stimulation with a bipolar-balanced montage (anode centered over left DLPFC,* F3, cathode centered over right DLPFC, F4) or nonactive sham tDCS (HDCkit, Magstim, UK). To increase the power of our study to detect predicted effects we utilized a comparatively high current intensity of 2 mA with 4 3 4 cm electrodes encased in saline soaked sponge pads to achieve a current density 5 0.125 mA/cm 2 [akin to (19,27) and larger than current densities reported in recent meta-analyses (16)]. In the active condition the stimulator supplied the 2 mA current for 20 min. In the sham condition, 2 mA stimulation was ramped up and delivered for the first 15 sec only. The participant was instructed to remain seated, relaxed, and refrain from any motor activity for the duration of the stimulation. Immediately after stimulation (active or sham) physiological and subjective measures were collected (peak effects). Participants then completed the ANT (commencing within 2 min of stimulation offset), which lasted approximately 20 min. *Though electrodes were centred over F3 and F4 to maximize stimulation of DLPFC we acknowledge that stimulation likely extended beyond these sites to broader PFC. Attention Network Test A central fixation cross was presented for 400-1600 msec, followed by a cue for 100 msec (except on no-cue trials) (Fig. 2). Four hundred milliseconds after cue offset (500 msec in no-cue trials), target and flanker arrows were displayed until response. Center and double cues (appearing above and below fixation) signaled target onset. Spatial cues (displayed either above or below fixation) signaled the onset and spatial location of target arrows. Targets were flanked by a pair of distracter arrows that pointed in the same (congruent) or opposite (incongruent) direction as the target. Participants completed 8 practice trials and 128 randomized experimental trials (16 counterbalanced trials/cue-type condition). Stimuli were presented using Inquisit 2.
See Figure 1 for study overview and Figure 2 for summary of trial events in the Attention Network Test.

DATA ANALYSIS
Incorrect responses (0.8%) and RTs greater than 1000 msec (1.6%, identified as outliers using boxplots) were removed (with no difference between groups, F's < 1). The alerting effect was calculated by subtracting the mean RT from double-cue trials from the mean RT on no-cue trials (RT[no-cue]-RT[double-cue]). The orienting effect was calculated by subtracting the mean RT on spatial-cue trials from the mean RT on center cue trials (RT[center-cue]-RT[spatial-cue]). The executive control effect was calculated by subtracting the mean RT of all congruent trials, from the mean RT of incongruent trials (RT[incongruent]-RT[congruent]). Large positive alerting and orienting scores reflect increased alerting and orienting network function respectively. Conversely, low positive execute control scores reflect improved executive control (i.e., reduced effect of flanker distractors on incongruent vs. congruent trials).
Independent samples t-tests compared active and sham tDCS groups on baseline and poststimulation anxiety, mood, heart rate, and blood pressure. Separate 2(Group: active vs. sham) 3 2 (Time: baseline vs. poststimulation) mixed design analyses of variance (ANOVA)s tested for group differences in anxiety, mood, heart rate, and blood pressure over time.
Reaction time data was entered into an omnibus 2(Group: active vs. sham) 3 4(Cue-type: spatial, double, central, no cue) 3 2(Congruence: congruent vs. incongruent) ANOVA. This omnibus analysis tests for group differences in global reaction time (i.e., main effect of Group) and interactions that would reflect a selective effect of group on task performance. Three separate independent samples t-tests compared active vs. sham tDCS groups on alerting, orienting and executive control attention network scores. All analyses were performed using SPSS software v23.
Consequently, any observed differences in attention network function cannot be attributed to unanticipated group differences in mood or arousal at baseline nor poststimulation.
A series of three separate independent samples t-tests compared active vs. Sham tDCS groups on alerting, orienting, and executive control attention network scores. Executive attention control network scores were lower (reflecting greater executive attention control) following active vs. sham stimulation t(28) 5 2.06, p 5 .04, d 5 0.76, see Figure 3. Alerting and orienting attention network function were unaffected by stimulation condition, see Table 2 and Figure 3. Group differences were not moderated by baseline/change in mood, alertness, heart rate, or blood pressure.
Supplementary analyses examined whether group differences in alerting and orienting were moderated by "target congruence" (i.e., whether group differences were observed on easier congruent trials but not more challenging incongruent trials). Separate 2 (Group) 3 2(congruence) ANOVAs on alerting and orienting scores did reveal significant effects of congruency (F(1,28) 5 17.28, p < .001; F(1,28) 5 11.71, p < .01) on alerting and orienting respectively, with better performance on congruent as compared to incongruent trials (consistent with the main effect of Congruence revealed in the omnibus ANOVA). There were no effects of tDCS group nor congruency 3 group interactions (F's < 0.01, p's > .95) and consequently there was no evidence that tDCS affected alerting and orienting attention network function (on easier congruent, or more challenging incongruent trials).

Blinding
Participants could retrospectively identify their stimulation condition (Active 12/15 correct; Sham 12/15 correct). The effect of tDCS vs. sham on executive attention was greater in participants who were aware of their stimulation condition (d (aware, n 5 24) 5 0.96 vs. d (all, n 5 36) 5 0.76). The perceived stimulation group had a smaller effect on executive control t(28) 5 1.82, p 5 .08, d 5 0.69, and no effect on other performance measures, mood, alertness and heart rate/blood pressure t's < 1.

DISCUSSION
Executive attention on the ANT was superior following 20 min of 2 mA anodal tDCS of left DLPFC using a balanced montage, when compared to sham stimulation. These findings support evidence from neuroimaging studies that implicate PFC in executive attention during the ANT [e.g., (2)]. The effects of tDCS on executive control complement findings from other studies in healthy volunteers, including evidence that 10 min of anodal 1.5 mA tDCS over the presupplementary motor area improves inhibitory control in a stopsignal task [vs. cathodal tDCS and no stimulation control; (29)]; that 10 min of 1 mA left anodal tDCS of DLPFC reduces distraction in a Sternberg memory task; and that anodal tDCS of the left DLPFC enhances computerized attentional bias training toward and away   from target stimuli (20) and reduces eye fixation durations to threat (27). Our findings also extend recent findings from studies that have specifically examined the effect of tDCS on ANT performance. Roy et al. (18) found tDCS of the right parietal cortex increased spatial reorienting, while tDCS of the other cortical targets (left parietal and dorsolateral) did not modulate alerting nor executive attention control-perhaps reflecting ceiling effects in high functioning healthy participants. Consequently, positive effects in our study [and (19,27)] might reflect the higher current intensity and density required to modify attention in healthy adults [2 vs. 1.5 mA in Roy et al. (18)]. We did not observe any effects of tDCS on mood or anxiety, consistent with null effects observed in other acute administration tDCS studies (30). Large standard deviations (Table 1) might reflect large individual differences in mood and arousal that mask small effects of tDCS group. We selected short established measures of anxiety, mood and arousal and administered these immediately after stimulation in fixed order (and typically completed within 2 min before the ANT commenced). These instruments might lack sensitivity required to detect small changes in mood over time, however it is more likely the evidence to date, together with our own findings, suggest that acute tDCS administration may not achieve reliable changes in mood to the extent seen following repeated administration in clinical groups (31). We did not include an active control site or low current control condition to test for dose response, and participants were able (when asked) to retrospectively discriminate stimulation condition. The lack of an off-target active control condition is a limitation particularly if unsuccessful blinding increases demand characteristics (e.g., increased motivation to perform well). In our study, though demand characteristics cannot be excluded, participants in the active condition did not display significant improvements in global reaction times or error rates in any of the trial types, nor any improvements in mood thus suggesting the effect of anodal tDCS on the executive control occurred independently of mood or motivation. Consequently, we offer evidence that 2 mA DLPFC stimulation selectively modulates the executive attention subtest of the ANT. Future research should examine whether the effect of tDCS of DLPFC on executive attention is enhanced during concomitant stimulation [see (32)], repeated administration and higher current density, and whether these positive effects generalize across behavioral measures of executive attention, and varying cognitive loads. Previous studies have used the ANT to reveal deficits in executive attention in anxiety, depression, and chronic pain for example, fibromyalgia [e.g., (33)]. Interestingly, recent evidence suggests 20 min of anodal tDCS (1 mA) over left DLPFC (vs. sham) can selectively improve executive attention and orienting performance in patients with fibromyalgia (34) raising the possibility that our stimulation protocol might improve executive attention deficits in populations characterized by impaired executive attention on the ANT, including patients with mood and anxiety disorders.