EEG alpha power changes reflect response inhibition deficits after traumatic brain injury (TBI) in humans q

Neuroscience Letters 362 (2004) 1 5 www.elsevier.com/locate/neulet EEG alpha power changes reflect response inhibition deficits after traumatic brain injury (TBI) in humans q Richard A.P. Roche a, *, Paul M. Dockree a, Hugh Garavan a, John J. Foxe b, Ian H. Robertson a, Shane M. O Mara a a Department of Psychology and Trinity College Institute of Neuroscience, University of Dublin, Trinity College, Dublin 2, Ireland b The Nathan Kline Institute, 140 Old Orangeburg Road, Building 35, Orangeburg, NY 10962, USA Received 19 August 2003; received in revised form 28 November 2003; accepted 28 November 2003 Abstract Brain damage due to traumatic brain injury (TBI) has been associated with deficits in executive functions and the dynamic control of behaviour. Event-related brain potentials and spectral power data were recorded from eight TBI participants and eight matched controls while they completed a Go/NoGo response inhibition task. The TBI group was found to be significantly impaired at the task compared to controls, and exhibited abnormal N2 and P3 waveform components in response to NoGo stimuli relative to controls. Significant correlations were also found between alpha power, Go-trial RT and errors. We conclude that abnormal activity in the structures damaged in this group may render such patients less capable of maintaining a state of alpha desynchronisation compared to controls, resulting in poorer performance on the task. q 2004 Elsevier Ireland Ltd. All rights reserved. Keywords: Event-related potentials; Spectral power; Desynchronisation; N2/P3 complex; Inhibitory control A key capacity often compromised after frontal injury is response inhibition (RI), the ability to exert inhibitory control over motor output by withholding routine or reflexive behaviours. This capacity serves an important adaptive function, and may be deficient in such disorders as schizophrenia, attention deficit disorder, attention deficit hyperactivity disorder (ADHD) and obsessive-compulsive disorder (e.g. refs. [1 3]). Considerable imaging research has been carried out to determine the brain areas that underpin RI, and it is currently hypothesised that a circuit involving the cingulate cortex and dorsolateral areas of the prefrontal cortex are implicated in successful inhibitory control [4,6,7]. Garavan et al. [6] suggested that these areas may be used to differing extents in normal participants, depending on their self-rated level of absentmindedness; those more prone to cognitive failures seemed to rely on a rapid, last-gasp anterior cingulate mechanism to perform the same RI task whereas the less absentminded who used a more conservative, slow-and-steady prefrontal generator to accomplish the task. In the same experiment, strong activation related to error q Experimental subjects statement: Each subject s participation was obtained only with the understanding and written consent of that subject. The experiment was conducted in accordance with the Code of Ethics of the World Medical Association and the ethical standards of the APA. * Corresponding author. Tel.: þ353-1-608-3907; fax: þ353-1-671-2006. E-mail address: riroche@tcd.ie (R.A.P. Roche). detection was observed over the cingulate cortex. In the present experiment, we investigate the behavioural performance and electrophysiological activity of a group of patients with TBI relative to normal controls on an RI task. We used the same task as in Roche et al. [14], in which we observed P2 and P3 latency differences for correct versus error trials, as well as enlarged components for the highly absentminded, based on CFQ (Cognitive Failures Questionaire) score. In addition to event-related brain potential (ERP) components that relate to response inhibition, we also examine changes within the power spectrum that are associated with RI task performance. Specifically, we investigate power changes within the alpha band (7 13 Hz). Cortical areas involved in task-related sensory and cognitive processes show synchronisation and desynchronisation within the alpha band [5,15]. Desychronisation refers to a phase disengagement of the dominant EEG alpha rhythm such that different neural populations start to oscillate at different frequencies in response to task demands; this is exhibited as a decrease in band-power measured from scalp electrodes. Modulation of alpha rhythmic activities depends on the interaction between thalamo-cortical and cortico-cortical networks [10]. The dynamic state of these networks depends, in part, on the modulating influence of cortical cholinergic inputs from the basal forebrain which contribute to the activation of frontoparietal regions during attentional control. It is predicted that 0304-3940/03/$ - see front matter q 2004 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.neulet.2003.11.064

2 R.A.P. Roche et al. / Neuroscience Letters 362 (2004) 1 5 disruption to distributed alpha systems following traumatic brain injury will lead to impaired top-down executive control resulting in poorer RI task performance. There were 15 participants. Seven (two female; mean age ¼ 35.9) had sustained TBI, while the control group consisted of eight normal participants matched to the TBI group for age (mean age ¼ 40), sex and performance intelligence (as measured by the National Adult Reading Test; NART). The participants were matched on emotional functioning using the Hospital Anxiety Depression (HAD) scale [16].In addition, all participants undertook neuropsychological tests that indexed attention (the Telephone Search and the Telephone Search While Counting subtests of the Test of Everyday Attention (TEA) [13]), memory (Logical Memory subtests I and II Wechsler Memory Scale (WMS-III uk )) and planning/strategy performance (revised Strategy Application Test (R-SAT) [9]). Each participant wore a Quikcap 32-channel EEG recording cap connected to the Neuroscan Synamps (Scan 4.1) ERP recording system (Medtech Systems Ltd., Horsham, UK) for the duration of the tasks, and EEG activity was recorded. Electrophysiological data were recorded in AC mode with a gain of 500 and a bandpass of 0.15 30 Hz. The A/D conversion rate was 1000 Hz, and the range was 11 mv. Sweeps were epoched from 250 ms prestimulus to 1000 ms poststimulus, and then averaged. For components of interest, a time-window was selected based on the grand average of all conditions and groups. Based on this window, the area under the curve for each component was taken as the dependent variable in a mixed factorial ANOVA. A spectral average was calculated as follows: the stimulus locked epochs were digitally bandpass filtered (7 13 Hz, 24 db/octave) and filtered values were converted to power by squaring (mv 2 ). Averaged spectra were obtained for all go-trials for each individual at the mid-line electrode sites (FZ, FCZ, CZ, CPZ and PZ). Subsequently, individual s averaged alpha power was correlated with their average response times to go-trials during the X-Y task. The task used was the X-Y response inhibition task used previously by Garavan et al. [6] and Roche et al. [14]. A stream of visual stimuli was presented at a rate of one per second, in black on a white background. The stimuli consisted of capital letters X or Y presented in alternating order. Participants were instructed to click the response box key with the index finger of the right hand for every stimulus when it followed a different stimulus (i.e. when an X followed a Y, or when a Y followed an X). When two identical stimuli followed each other (an X following an X, or a Y following a Y), participants were instructed to withhold their response (see Fig. 1A). A trial block consisted of 315 trials, of which 20 trials were critical lures, or trials which required a withhold. Two trial blocks were presented, with a short rest period (3 min) between the blocks. Stimuli remained on-screen for 700 ms, and were followed by a blank screen for 300 ms, giving an interstimulus interval of 1000 ms. For the neuropsychological test battery, no significant differences were found between the TBI and control groups for age (Fð1; 17Þ ¼0:03; P, 0:86), NART score (Fð1; 17Þ ¼3:38; P, 0:08) or HADS anxiety/depression scores (HAD-anxiety: Fð1; 17Þ ¼1:49; P, 0:24; HADdepression: Fð1; 17Þ ¼0:096; P, 0:34). Significant differences were found for CFQ score (lower score for controls; Fð1; 17Þ ¼7:66; P, 0:013), Logical Memory delayed recall (superior performance for controls; Fð1; 17Þ ¼11:32; P, 0:004) and R-SAT score (superior performance for controls; Fð1; 17Þ ¼19:29; P, 0:001), but not for Logical Memory recall (Fð1; 17Þ ¼3:98; P, 0:06), Logical Memory recognition (Fð1; 17Þ ¼1:08; P, 0:31) and the TEA (Fð1; 17Þ ¼2:84; P, 0:11). On the response inhibition task, the mean number of commission errors was 23.3 (^2.8) out of 40 lures for the TBI group (54.55% errors), compared to 14.7 (^2.8) out of 40 for controls (39.1% errors). This difference was significant (Fð1; 18Þ ¼4:84; P, 0:041; Fig. 1B). For the TBI patients, the mean reaction time to standard stimuli was 366.99 ms (^18.9), while for controls it was 348.38 ms (^ 18.9). This was not significant (Fð1; 20Þ ¼1:338; P, 0:261). Reaction times to errors of commission did not differ significantly (TBI: 336.12 ^ 18.5 ms; control: 313.22 ^ 18.5; Fð1; 20Þ ¼1:747; P, 0:201). Analysis of event-related potentials showed that the control group exhibited a typical Go/NoGo effect of enlarged N2 and P3 components for lures relative to standards (See Fig. 2, left panels). The N2 component did not differ significantly between the stimulus types (Fð1; 6Þ ¼2:55; P ¼ 0:15), but the P3a at FCz (Fð1; 6Þ ¼5:62; P ¼ 0:045) and the P3b at CPz (Fð1; 6Þ ¼5:35; P ¼ 0:049) were both larger for lures compared to standards. Stimulus X group interactions were non-significant. The Go/NoGo effect was almost completely absent for the TBI group (see Fig. 2, right panels). The N2 component was found to be maximal at the FCz electrode, but no main effect of Stimulus was found (Fð1; 6Þ ¼0:001; P ¼ 0:15). The same was true of the P3a component, also maximal at FCz (Fð1; 6Þ ¼0:30; P ¼ 0:60) and the P3b, which was largest at the posterior lead CPz (Fð1; 6Þ ¼4:18; P ¼ 0:09). There was some evidence in the TBI group of a lure-n2 component that failed to resolve into a P3 component. Between groups analyses revealed that, for standard stimuli and lures, controls and TBIs did not differ significantly for any of the three components. Analysis of the spectral data revealed no significant differences in alpha power between the groups (t ¼ 1:5; df ¼ 13, P ¼ 0:204). Importantly, the TBI group showed significantly greater variation in alpha power across midline electrodes compared to controls (Mann Whitney U ¼ 22:6; P ¼ 0:008). In addition, the standard deviation for Go-trial RTs in the TBI group was double that in the control group. When Go RTs were correlated with alpha power, a significant

R.A.P. Roche et al. / Neuroscience Letters 362 (2004) 1 5 3 Fig. 1. (A) Presentation sequence and response demands for stimuli in XY response inhibition task; a lure is defined as a stimulus that follows an identical stimulus (e.g. an X following and X, a Y following a Y). (B) Bar graph showing mean number of errors of commission in the TBI and control groups. negative correlation was observed whereby decreased alpha power was associated with longer RTs. This was only found at the Fz electrode for the controls, and more extensively at Fz, FCz and Cz for the TBI group (see Table 1). This negative correlation at Fz for both groups is shown in Fig. 3A. A nonsignificant negative correlation between RT and commission errors was found for the TBI group (r ¼ 20:66; P ¼ 0:11) but this correlation was weaker for controls (r ¼ 20:19; P ¼ 0:66; Fig. 3B); it is likely that lack of power accounted for the absence of a significant statistical effect. Conversely, there was a strong positive correlation between errors and alpha power at Pz (r ¼ 0:85; P ¼ 0:008) and CPz (r ¼ 0:82; P ¼ 0:013) in the control group, but no evidence of this relationship in the TBI group (Fig. 3C). The brain-injured group made significantly more errors of commission (false presses) than controls, but did not differ in their reaction times to standards or erroneously pressed lures. Controls showed waveform differences between standards and critical lures, representing the typical Go/NoGo effect. TBIs, by contrast, did not show this pattern, with virtually identical waveforms for both stimulus types. A similar disparity characterised the groups when correct withholds and errors were compared, though the manifest differences in the control group failed to achieve statistical significance; as has been suggested, the low trial count for these averages may explain this surprising result. There were no differences in reaction times to Go-trials or errors between the TBI and control groups. This strongly suggests that the areas damaged in these TBI participants are necessary for the successful execution of the sort of top-down behavioural control required for response inhibition, and agrees with previous studies reporting impaired performance on sustained attention/ri tasks [8,11]. Given the hypothesised prefrontal-cingulate circuit suspected of involvement in RI, we may infer that the locus of damage in this group encompassed the cingulate and/or prefrontal cortices. This assertion is further supported by the results of the neuropsychological test battery, which showed impairments on frontally-dependent capacities such as strategy application and delayed memory recall. We may therefore conclude that the impairments in the TBI group tested here included frontal dysfunction, and this was associated with disrupted response inhibition performance. ERP waveform components also discriminated the TBI and control groups in their responses to stimuli. In the control group, a standard Go/NoGo effect consisting of enlarged N2 and P3 components was observed for lures over standards (Fig. 2, left panels). This pattern was effectively absent in the TBI group; no substantial N2 or P3 components were discernable (Fig. 2, right panels). This might be interpreted as evidence that TBI patients failed to attach any salience marker to lure stimuli when they were perceived, resulting in poorer performance of the task. When lure waveforms were compared across the groups, enlarged P3 amplitudes were also seen for the control group; this was in contrast to the standards waveform, in which the groups did not differ. The implication may be that, although lures might have been correctly identified (which is possible, given that TBIs were unimpaired on recognition memory in the Logical Memory Test), the resultant cortical activation was insufficient in the TBI group to allow the response to be withheld. Although the P3 peaked after the mean response latency occurred for errors of commission, it is possible that the onset of this component might signal some active response inhibition processes, and that if this onset does not occur within a certain latency window, the attempted inhibition will fail. The negative correlations between mean response time and mean alpha power demonstrate that longer RTs are associated with reductions in power for controls (at FZ) and Table 1 Correlation coefficients (with significance in parentheses) for Go RT and alpha power at electrode sites Fz, FCz, Cz, CPz, and Pz for controls and TBIs (* indicates correlation is significant at the 0.05 level (2-tailed)) Controls (n ¼ 8) TBIs (n ¼ 8) Electrode locations RT RT FZ 20.709 (0.049*) 20.789 (0.035*) FCZ 20.176 (0.677) 20.848 (0.016*) CZ 20.535 (0.172) 20.806 (0.029*) CPZ 20.575 (0.136) 20.660 (0.106) PZ 20.145 (0.732) 20.468 (0.290)

4 R.A.P. Roche et al. / Neuroscience Letters 362 (2004) 1 5 Fig. 2. Stimulus-triggered average ERP waveforms for standard (thin line) and lure (heavy line) stimuli at FCZ and CPZ electrode sites for the control (left panels) and TBI groups (right panels). Enlarged N2 and P3a and P3b components are visible for lures in the control group, but not the TBI group (these are shown at an increased scale in the insets, to allow waveform morphology to be identified). TBIs (at FZ, FCZ and CZ). Longer RTs to Go-trials during the X-Y task are associated with fewer errors of commission. Conversely, shorter RTs to Go-trials increase the likelihood of a commission error to the target, as was evident in the TBI group, where a correlation between RT and errors was found. The control group, by contrast, had longer Go RTs (with significantly reduced variance) which were accompanied by fewer commission errors. The implications are that longer RTs are indicative of an optimal strategy during the task whereas shorter RTs reflect poor top-down control. Our findings suggest that desynchronisation within the alpha range may underlie heightened attentional control during the task. This claim is supported by the positive correlation between number of commission errors and alpha power that was seen in the control group. The absence of this relationship in the TBI group raises the possibility that other factors (for example, lower-order stimulus processing) may interact with topdown control processes to determine the quality of performance in this group. Successful task performance in the TBI group may also be more sensitive to a speed-accuracy trade-off, as evidenced by the correlation between RT and errors in this group. In addition, alpha power variance was found to be significantly larger for the TBI group, possibly explaining the lack of a between-groups effect for alpha power. This may imply that, for TBIs, frontal alpha generators may be less efficient at stabilizing RTs during task performance. By contrast, controls, who show less variance of RT and alpha power, may be more adept and maintaining a desynchronised alpha state that is associated with a more optimal response strategy. The implication is that, rather than being unable to achieve alpha desynchronisation, sufferers of TBI may simply be unable to maintain alpha desynchronisation, leading to poorer performance. Increased variation in alpha power and RT in the TBI group suggests that their sustained attention capacity is prone to fluctuations of efficiency throughout the RI task. Sustained attention has been localised to the right prefrontal and parietal cortices in imaging studies using the Sustained Attention to Response Task (SART; [12]), a similar Go/NoGo task to the present RI paradigm. Traumatic brain injured patients also show performance deficits on the SART [12], making significantly more errors of commission on No-Go trials compared to controls. In addition, TBIs showed increased variation in RTs to Go-trials, and shorter RTs were predictive of subsequent errors on the NoGo trials. It is possible that the proposed right-hemisphere circuit that supports sustained attention in these kinds of Go/NoGo tasks is disrupted following brain injury and that these impairments may result in a drift of controlled processes making the inhibition of a response at the critical moment more difficult, as was observed here. The presence of a normally functioning sustained attention system may be a prerequisite condition for successful response inhibition performance. We conclude that in healthy controls, alpha desynchronisation indexes efficient top-down control, which is manifested in longer RTs and fewer errors. By contrast, for TBI patients this relationship between alpha power and performance is less straightforward. TBIs show greater variation in alpha power and RT, suggesting an inability to maintain desynchronisation. In addition, the absence of a

R.A.P. Roche et al. / Neuroscience Letters 362 (2004) 1 5 5 Special thanks go to our participants for their patience and cooperation, to Miguel Angel Rodriguez at Medtech for technical assistance, to Kevin Murphy (Trinity College) for writing the program, and to those at Headway Ireland. References Fig. 3. (A) Scatter plot depicting significant negative correlation between Go RT and alpha power at Fz electrode site for controls (filled squares; r ¼ 20:709; P ¼ 0:049) and TBIs (unfilled circles; r ¼ 20:789; P ¼ 0:035). (B) Scatter plot depicting negative correlation between Go RT and errors of commission for controls (filled squares; r ¼ 20:18; P ¼ 0:66) and TBIs (unfilled circles; r ¼ 20:66; P ¼ 0:11). (C) Scatter plot depicting significant negative correlation at CP 3 electrode between alpha power and errors for controls (filled squares; r ¼ 0:82; P ¼ 0:013) and non-significant correlation for TBIs (unfilled circles; r ¼ 0:333; P ¼ 0:466. correlation between alpha power and errors in this group suggests that other factors, such as inadequate stimulus processing (as revealed by abnormal NoGo-N2 and P3 components) may interact with the top-down effect in influencing task performance. Acknowledgements This work was supported by Enterprise Ireland; S.M. O Mara was in receipt of a Berkeley Fellowship from Trinity College (2000 2001). H. Garavan was in receipt of NIDA support: DA14100. Thanks go to Deirdre Foxe and Glenn Wylie at NKI for their assistance with ERP analyses. [1] C.S. Carter, A.W. MacDonald 3rd, L.L. Ross, V.A. Stenger, Anterior cingulate cortex activity and impaired self-monitoring of performance in patients with schizophrenia: an event-related fmri study, Am. J. Psychiat. 158 (2001) 1423 1428. [2] B.J. Casey, F.X. Castellanos, J.N. Giedd, W.L. Marsh, S.D. Hamburger, A.B. Schubert, Y.C. Vauss, A.C. Vaituzis, D.P. Dickstein, S.E. Sarfatti, J.L. Rapoport, Implication of right frontostriatal circuitry in response inhibition and attention-deficit/hyperactivity disorder, J. Am. Acad. Child Adol. Psychiat. 36 (1997) 374 383. [3] C.E. Curtis, M.E. Calkins, W.G. Iacono, Saccadic disinhibition in schizophrenia patients and their first-degree biological relatives. A parametric study of the effects of increasing inhibitory load, Exper. Brain Res. 137 (2001) 228 236. [4] E.C. Dias, J.J. Foxe, D.C. Javitt, Changing plans: a high density electrical mapping study of cortical control, Cereb. Cortex 13 (2003) 701 715. [5] J.J. Foxe, G.V. Simpson, S.P. Ahlfors, Cued shifts of intermodal attention: parieto-occipital,10 Hz activity reflects anticipatory state of visual attention mechanisms, NeuroReport 9 (1998) 3929 3933. [6] H. Garavan, T.J. Ross, K. Murphy, R.A.P. Roche, E.A. Stein, Dissociable executive functions in the behavioural control: Inhibition, error detection and correction, NeuroImage 17 (2002) 1820 1829. [7] H. Garavan, T.J. Ross, E.A. Stein, Right hemispheric dominance of inhibitory control: an event-related functional MRI study, Proc. Natl. Acad. Sci. 96 (1999) 8301 8306. [8] K. Konrad, S. Gauggel, A. Manz, M. Schöll, Inhibitory control in children with traumatic brain injury (TBI) and children with attention deficit/hyperactivity disorder (ADHD), Brain Injury 14 (2000) 859 875. [9] B. Levine, D. Dawson, I. Boutet, M.L. Schwartz, D.T. Stuss, Assessment of strategic self-regulation in traumatic brain injury: its relationship to injury severity and psychosocial outcome, Neuropsychology 14 (2000) 491 500. [10] F.H. Lopes da Silva, J.E. Vos, H. Mooibroek, A. van Rotterdam, Relative contributions of intracortical and thalamo-cortical processes in the generation of alpha rhythms, revealed by partial coherence analysis, Electroenceph. clin. Neurophysiol. 50 (1980) 449 456. [11] T. Manly, A.M. Owen, L. McAvinue, A. Datta, G.A. Lewis, S.K. Scott, C. Rorden, J. Pickard, I.H. Robertson, Enhancing the sensitivity of a sustained attention task to frontal damage. Convergent clinical and functional imaging evidence, Neurocase 9 (2003) 340 456. [12] I.H. Robertson, T. Manly, J. Andrade, B.T. Baddeley, J. Yiend, Oops! : performance correlates of everyday attentional failures in traumatic brain injured and normal subjects, Neuropsychologia 35 (1997) 747 758. [13] I.H. Robertson, T. Ward, V. Ridgeway, I. Nimmo-Smith, The structure of normal human attention: The Test of Everyday Attention, J. Int. Neuropsychol. Soc. 2 (1996) 525 534. [14] R.A.P. Roche, H. Garavan, J.J. Foxe, S.M. O Mara, Individual differences discriminate event-related potentials but not performance during response inhibition, Exp. Brain Res (under review). [15] M.S. Worden, J.J. Foxe, N. Wang, G.V. Simpson, Anticipatory biasing of visuospatial attention indexed by retinotopically specific alpha-band EEG increases over occipital cortex, J. Neurosci. 20 (2000) 1 6. [16] A.S. Zigmond, R.P. Snaith, The hospital anxiety and depression scale, Acta Psychiatr. Scand. 67 (1983) 361 370.