Neurophysiological markers of cognitive deficits and recovery in concussed adolescents

OBJECTIVE
The present study sought to determine: 1) whether concussed adolescents exhibited deficits in neurocognitive functioning as reflected by neurophysiological alterations; 2) if neurophysiological alterations could be linked to supplementary data such as the number of previous concussions and days since injury; and 3) if deficits in psychological health and behavioural tests increased during diagnosis duration.


METHODS
Twenty-six concussed adolescents were compared to twenty-eight healthy controls with no prior concussions. Self-report inventories evaluated depressive and concussive symptomatology, while behavioral tests evaluated cognitive ability qualitatively. To assess neurophysiological markers of cognitive function, two separate auditory oddball tasks were employed: 1) an active oddball task measuring executive control and attention as reflected by the N2b and P300, respectively; and 2) a passive oddball task assessing the early, automatic pre-conscious awareness processes as reflected by the MMN.


RESULTS
Concussed adolescents displayed delayed N2b and attenuated P300 responses relative to controls; showed elevated levels of depressive and concussive symptomatology; scored average-to- low-average in behavioral tests; and exhibited N2b response latencies that correlated with number of days since injury.


CONCLUSION
These findings demonstrate that concussed adolescents exhibit clear deficiencies in neurocognitive function, and that N2b response latency may be a marker of concussion recovery.


Introduction
Concussion has been defined as a consequence of traumatic biomechanical forces resulting in a complex pathophysiological process of biochemical changes in the brain (McCrory et al., 2009;Zhang et al., Recently, the effects of concussion within the adolescent population have garnered increased attention (Reddy et al., 2008;Grady, 2010;Master et al., 2012;Zhang et al., 2016) with a recent meta-analysis revealing that the largest increase of concussion incidence occurs in the adolescent population (Zhang et al., 2016). Adolescence, a period in human growth where significant changes in cognition, behaviour, and brain development occur (Blakemore & Choudhury, 2006), may be a time where both concussion incidence (Zhang et al., 2016) and neurological harm (Baillargeon et al., 2012) are maximal. Clearly, a better understanding of the neurophysiological effects of concussion on this population is in order.
In an effort to investigate the effects of concussion in adolescence, early research has focused primarily on behavioural measures such as self-report symptom inventories, behavioral tests, and neuropsychological assessments (e.g., Lovell et al., 2003;Hinton-Bayre, 2012;Echemendia et al., 2013). For instance, a study investigating memory dysfunction using the Immediate Post Assessment Concussion Tool (ImPACT) revealed that recently concussed high school athletes demonstrated significant deficits in memory function up to 7 days post injury . Furthermore, a study evaluating the agerelated differences associated with concussion between adolescent and young-adult populations revealed that adolescent athletes are more susceptible to prolonged concussion effects . These findings identify the detrimental behavioral effects of concussion; however, they fail to objectively identify any associated neurophysiological damage. Accordingly, in addition to behavioural measures, there is a need for objective measurement of the effects of concussion using, for example, neurophysiological recordings when assessing the effects of concussion in adolescent populations.
The P300 is a positive-deflecting neurophysiological response peaking~275 ms to 700 ms post stimulus onset (Polich, 2007). Depending on the cognitive tasks performed, the P300 has been shown to be sensitive to attention (Gray et al., 2004), memory (Polich, 2007), and cognitive workload (Allison and Polich, 2008). Concussion literature has predominantly investigated the modulation of the P3b component. The P3b has a centro-parietal scalp distribution that ap-pears~300 to 700 ms post stimulus onset (Baillargeon et al., 2012), and reflects attentional and working memory processes (Polich, 2007). The P3b has been shown repeatedly to be reduced and/or delayed in the adult concussion literature, thus, demonstrating its utility in assessing the pathological neurophysiological effects associated with concussion (e.g., Dupuis et al., 2000;Lavoie et al., 2004De Beaumont et al., 2012).
To date, only one study has used ERPs to identify the neurophysiological deficits resulting from concussion in an adolescent population. Baillargeon et al (2012) found significantly attenuated P3b responses in asymptomatic concussed adolescents (ages 13-16) approximately 6 months post-injury compared to healthy age-matched controls. In addition, only the adolescent group, relative to the child (ages 9-12) and adult (ages 18 + ) concussed groups, scored worse in the behavioural working memory task. Their study concluded that the adolescent group was most susceptible to working memory deficiencies following a concussion as reflected by both neurophysiological and neuropsychological evidence. This research supports the notion that concussion is likely to disrupt frontal lobe function; the brain region commonly associated with working memory (Thompson-Schill et al., 2002).
The Mismatch Negativity (MMN) is another component studied extensively in traumatic brain injury research (e.g., Daltrozzo et al., 2007;Fischer et al., 2010;Morlet and Fischer, 2014;Blain-Moraes et al., 2016;Connolly et al., 2019). The MMN, a negative-deflecting ERP occurring~150-200 ms post stimulus onset (Näätänen et al., 1978) that reflects an automatic attention function linked to a predictive coding process (Garrido et al., 2008) requiring a conscious state but not awareness (Atienza et al., 2002;Fischer et al., 2010;Dykstra and Gutschalk, 2015). Like the P300, the MMN is elicited in an oddball paradigm; however, unlike the P300, the MMN does not require active attention to be evoked. Historically, the MMN has been used solely in what can be characterized as "catastrophic" brain injury populations (e.g., coma, vegetative state, minimally conscious state). It was not until recently that the MMN was investigated in a concussed population and found to be significantly reduced in retired professional football players who had sustained their last concussion almost 30 years earlier (Ruiter et al., 2019); demonstrating the efficacy of the MMN in evaluating the long-lasting neurophysiological deficits of mild traumatic brain injuries (mTBI) such as concussion. However, the MMN has not been investigated in recently-concussed populations.
This study builds on research demonstrating neurophysiological abnormalities reflective of the cognitive consequences of concussion. In particular, it contributes to the small literature on adolescent concussion by extending the breadth of cognitive functions -and their neurophysiological manifestations -being assessed. It was hypothesized that amplitude reductions and/or latency delays would be seen in each of the assessed ERP components for the concussed participants compared to controls. Specifically, the present study investigated a full range of neurophysiological responses associated with memory, executive control, and attention by examining the MMN, N2b, and P300 components. To align with previous literature, the ImPACT, Child Depression Inventory (CDI), and Post-Concussion Symptom Scale (PCSS) were administered to gain a better understanding of symptom levels and behavioral functions associated with the concussed participants. The present study extends previous literature by being the first study to utilize ERPs to investigate the neurophysiological effects associated with symptomatic, acutely concussed adolescent participants, in addition to being the first to examine the MMN in an acutely concussed population.

Demographic, Behavioral, and symptomatology results
Results from the demographic data revealed that the concussed group's average age was 15.04, that they had sustained on average 1.88 previous concussions, and had participated in EEG testing on average 20.15 days after sustaining their most recent concussion (Table 1). According to ImPACT normative data , the results revealed that female concussed participants scored "Low-Average" in Verbal Memory, Visual Memory, and Motor Speed, as well as "Borderline" (almost "Impaired") in Reaction time (Table 2). Male concussed participants, on the other hand, scored "Average" in Verbal Memory and Visual Memory, "Low-Average" in Motor Speed, and "Borderline" in Reaction Time (Table 2). Normative values for Impulse Control and CEI scores were unavailable. Furthermore, on average, the concussed group scored 55.08 in concussive symptomatology and 56.07 in depressive symptomatology; demonstrating elevated levels of concussion symptoms and "Slightly above average" levels of depression according to the CDI (See Table 3).

MMN protocol (automatic attention)
When examining the waveforms and topographies in the MMN protocol ( Fig. 1), N1 waveforms can be seen across groups and conditions with clear topographical differences between the groups and across conditions. Response amplitudes, in particular, differed between groups as reflected by the fact that when data were scaled based on the N1 exhibited by the control group, the N1 representation in the concussed group was not observablean effect seen most clearly in the topographies of the response (thus, note scaling differences for the two groups in Fig. 1). MMN responses to stimulus onset revealed no discernible differences between groups. However, a drastic size difference between the two conditions was seen where the FT condition had significantly smaller amplitudes compared to the DT condition. Between the two groups, MMN scalp distributions in the FT condition showed a typical fronto-central distribution in the control group, whereas a more frontal-exclusive representation can be seen in the concussed group.
A Group main effect (P < 0.01) and a Group X Region interaction reflected the significantly reduced N1 amplitudes observed in the concussion group (P < 0.01). Post-hoc analyses (See Table 5) of the interaction emphasized the pervasiveness of the amplitude effect in the concussion sample with significant differences observed across 8 of the 9 ROIs for each condition. Further, a Group X Region interaction (P < 0.01) in N1 latency was observed where specific R-P was found to be significantly delayed in the concussed group. No significant amplitude or latency effects were observed for the MMN.

P300 Protocol/N2b (voluntary attention, memory, response inhibition/conflict monitoring)
The P300 oddball protocol evoked multiple ERP components differentially associated with sensory/perceptual and cognitive processes (Fig. 2). An observational summary of the findings shows a clear N1 sensory/perceptual response with a typical fronto-central distribution for both groups in each condition ( Fig. 2A, B). In contrast to the similarity between groups for the N1a fundamentally sensory responsethe cognitive responses exhibited a range of contrasts between adolescents who had been concussed and the healthy control population. The N2b differed markedly across groups and conditions in two different ways. In the FT condition, while typical N2b amplitudes can be seen in both groups, the response occurred significantly later (~25 ms) in the concussed group. Furthermore, N2b responses in the DT condition were larger than those in the FT condition, with no visually-discernible differences between the two groups in either amplitude or latency. While Table 1 Individual and mean (SD) values of age and sex for both the control group and concussed adolescent group, as well as the number of previous concussions and days since last concussion for the concussed adolescent group. the two groups exhibited very similar fronto-central distributions for the N2b in the DT condition, a more prominent frontal distribution was observed in the FT condition for the concussed group.
In terms of the late positivities, clear P3b responses were observed in both groups and in both conditions. However, while the distribution of the response in the control population was widespread in both conditions, the distribution exhibited in the concussed group showed a more centro-parietal distribution. Lastly, a late positive component (LPC) was found in the 500-700 ms in the control group for each condition, with the response being larger in the DT condition. The concussed group also exhibited a LPC in both the DT and FT conditions; however, the response had a more concentrated parietal distribution compared to the control group's more centro-parietal distribution. In addition, notable differences between the two groups can be seen in the topographies ( Fig. 2A, B).
Within the P300 protocol, statistical analyses (See Table 4) revealed no effect of amplitude or latency for the N1. A main effect of Group was found for the N2b where response latencies within the concussed group were significantly delayed (P < 0.05). Additionally, there was a Group X Condition interaction (P < 0.05) where post-hoc analysis (See Table 5) determined that response latencies within the FT condition were attributable to slower response latencies in the concussed group compared to the controls (P < 0.01). Moreover, a Group X Region interaction (P < 0.01) was revealed where the response latencies in the L-C, L-F, M-C, M-F, and R-F ROIs were found to be significantly slower in the concussed group relative to the control group. Although no main effect was found for N2b amplitude a Group X Region interaction was observed (P < 0.01). Post-hoc analysis revealed significantly smaller amplitudes in the R-F region of the concussed group.
A main effect of group (P < 0.01) was found for P3b amplitudes reflecting the concussed group's significantly attenuated response amplitudes compared to the control group. Additionally, a Group X Region interaction (P < 0.01) was found with post-hoc analyses showing significantly reduced amplitudes in the L-F, M-F, R-C, and R-F ROIs. No effects of P3b latency differences were observed. Finally, a Group X Region interaction (P < 0.05) for the LPC was found in the 500 to 700 ms time window; an effect reflecting the significantly decreased amplitudes for the concussed group at the L-F, M-F, and R-F frontal ROIs.

Post-hoc correlational analyses
A series of simple linear regression analyses applied to the demographic data demonstrated that N2b latencies in the FT condition within the aggregated frontal regions (L-F, M-F, R-F) were trending towards significance to the number of days since last concussion (F(1, 24) = 3.73, P = 0.06, R 2 = 0.13). However, when applying the analysis to aggregated M-F and R-F ROIs exclusively (See Fig. 3), results revealed that number of days since last concussion was predictive of N2b latency (F(1, 24) = 5.08, P < 0.05, R 2 = 0.17). This finding reveals that as the number of days since concussion increased, N2b response latencies decreased, (B = -0.57, P < 0.05). Furthermore, age was predictive of N2b latency (F(1, 24) = 4.47, P < 0.05, R 2 = 0.16). Thus, N2b response latencies decreased as a function of age (B = -4.49, P < 0.05). Accordingly, both age and days since injury were predictive of N2b latency in the concussed population. Finally, behavioral and symptom scores in the concussion group did not correlate significantly with the neurophysiological data.

Subset neurophysiological results
Separate analyses were conducted on the 17 concussed participants who sustained their concussion < 21 days (average 12) before date of testing. The delayed N2b response latencies for the concussed group compared to controls remained significant (F(1, 43) = 4.84, P < 0.05). Also a Group X Condition interaction (F(1, 43) = 6.51, P < 0.05) revealed slower response latencies in the FT (F(1, 43) = 10.44, P < 0.01) condition for the concussed group. Statistical analyses of this subset yielded similar results to what is reported above for those over 21 days (except for the P3b amplitude main effect).

Subset post-hoc correlational analyses
A series of simple linear regression analyses were also calculated on the subset of 17 concussed participants who sustained their last concussion < 21 days prior to testing. The subset (17) group results revealed that number of previous concussions remained unrelated  (P > 0.05) to N2b latency. Also, days since last concussion (See Fig. 4) was significantly correlated to N2b latency (F(1, 15) = 5.11.03, P < 0.05, R 2 = 0.25). Thus, like in the entire concussed group, as the number of days since concussion increased, the N2b response latencies decreased (B = −1.88, P < 0.05). Interestingly, age was no longer found to be significantly correlated to N2b response latency (P > 0.05). In summary, both age and days since concussion were predictive of N2b response latencies in the entire (26) concussed group, while only days since injury was found to be significant in the subset (17) group. Regressions were not calculated on the 9 participants who sustained their injury > 21 days at the time of testing as no main effect was found between the 9 subjects and the control group.

Behavioural findings
The concussed adolescent group self-reported numerous symptoms and high levels of symptom severity on the PCSS. Common symptoms included: headaches, sadness, difficulty concentrating, difficulty remembering, sensitivity to light and noise, and feelings of emotional instability. These results demonstrated the physical and emotional toll those who have recently sustained a concussion endure on a day-to-day basis until recovery, and are aligned with prior findings (e.g., Ryan and Warden, 2003;Lucas, 2011;Covassin et al., 2013). CDI results revealed that the concussed adolescents had higher average levels of depressive symptomatology than neurologically healthy controls; a finding that previous work has shown clearly in adolescence (Chrisman and Richardson, 2014) and other age groups (Chen et al., 2008;Kontos et al., 2012;Strain et al., 2013).
Behavioral results as assessed by the ImPACT proved unconvincing. The adolescent group performed at low-average to average levels in all categories except Reaction Time (RT), where both male and female results revealed what is referred to as a Borderline score (Table 2). Poor RT performance has been shown repeatedly to be a common impairment in concussed populations (Warden et al., 2001;Eckner et al., 2010;Kontos et al., 2012); however, it is noteworthy that other behavioral scores associated with performance such as Verbal and Visual Memory, and Motor Speed were unaffected. This finding may be attributable to behaviourally manifested cognitive deficits returning to pre-concussion performance in as little as 5 to 10 days post injury despite other lingering symptoms (Johnston et al., 2001).

Neurophysiological findings
The neurophysiological data demonstrated the neurocognitive  consequences associated with concussion. The data obtained in the P300 protocol was particularly clear in demonstrating differences between the healthy control participants and those adolescents who had sustained a concussion with the latter showing significantly delayed N2b latencies and reduced P3b amplitudes.
The N2b, a response associated with inhibitory executive functions (Heil et al., 2000), such as response inhibition and conflict monitoring (Folstein & Van Petten, 2008), has been shown in previous literature to be affected by concussion ). In the present study, a significant delay of~25 ms was found in the N2b response latency for  the concussed adolescent group providing evidence of a disruption in executive control processes -a finding complementary to previous concussion research (Howell et al., 2013). This N2b response delay in the concussed group was also seen in the subset containing only those group members who sustained their concussion ≤ 21 days prior to testing. Results such as these add to the evidence that claims of symptom resolution in 90% of concussion cases within 21 days Guskiewicz et al., 2003;McKeon et al., 2013; inter alia) are dependent on how symptoms are assessed or measured. Some of the neurophysiological manifestations of cognitive dysfunction after concussion clearly are capable of lasting years (e.g., De Beaumont et al., 2009;Ruiter et al., 2019). However, the current data set suggests that some measures, such as the neurophysiological marker for executive function employed in this study, may reveal a recovery trend that begins within the 21-day window but also continues (meaning that symptoms had not yet resolved) when data from up to 58 days is included.
Post-hoc regression analyses on the N2b data sets provided an additional layer of insight on relating the neurophysiological responses of the concussed group to their demography, symptomatology, and behavioral data. Specifically, days since injury was found to be predictive of N2b response latencies as was, in a marginally nonsignificant effect, age. When the complete sample of concussed participants was considered, N2b latencies decreased as the number of days since concussion increased. Similarly, N2b latencies decreased as a function of age. No effects on the N2b were found as a function of number of concussions; although the total number of prior concussions was < 2 per person and the range of prior concussions was 0 -6 and clearly skewed.
When similar linear regression analyses were conducted on the subsample of 17 concussed participants who were tested within 21 days of their injury, results were comparable to the larger analysis with all of the concussed participants, with decreases in N2b latencies being significantly related to an increased number of days since the time when they sustained their concussions. This relationship between decreased N2b latencies with increased number of days since being concussed in this subsample was not accompanied by a significant relationship with age.
Although there is evidence demonstrating that both amplitude and latency decrease with increasing age (e.g., Amenedo and Dıaz, 1998;Lamm et al., 2006; for a review; Lewis et al., 2006) there is a wealth of evidence that is either contradictory or more nuanced. For example, the literature provides evidence indicating that N2b latency continues to shorten in healthy population samples over time until maturity at~25 years of age (Arain et al., 2013) at which point, response latencies become more delayed as age increases. Lamm et al. (2006) found latency decreases across a small age span (ages 7 -16 yrs)a finding that can then be integrated with Amenedo and Dıaz (1998) evidence to provide a trajectory of N2b latency across an age span overlaps with some of the participants in the current study. In contrast, to these findings of shortening latencies until young adulthood, other work has provided compelling evidence that N2b latency remains unaffected between childhood and adolescence (e.g., Johnstone et al., 1996). The most important piece of information to remember is that the N2b latencies observed in the concussed group may have exhibited a decreasing latency related to time since injury but in all analyses, the latency of the concussed group remained significantly delayed even when compared to a slightly older control sample. If there was any type of age confound in the current study, we argue it was minimal as apparent by the age effect disappearing in the subset analysis.
Taken together the analyses of the neurophysiological response associated with the executive function(s) involved in the task used in this study revealed that while the concussed group exhibited significantly delayed latencies compared to the healthy controls, there remained a new discovery showing that as days since injury increased N2b response latency decreased. This result indicates that as recovery from concussion progresses, the N2b latency begins its return to time periods reflective of typical cognitive performance. It is particularly important to note that this effect remained significant in the subset of the concussed group tested within 21 days of their injury. This finding emphasizes the importance of this particular ERP component as a measure sensitive to the recovery of important cognitive functions; and also demonstrates its reliability in different sample sizes. These findings also emphasizes the utility of ERP components as clinical state "trackers" of various cognitive functions after brain injury, as well as recovery of those functions. In the current investigation, the N2b latency stands out as being able to track neurophysiological markers of cognitive abnormalities associated with concussion and subsequent improvement over time.
The P3b reduction found in the present study is consistent with previous concussion literature (Lavoie et al., 2004;Theriault et al., 2009;Baillargeon et al., 2012) and provides further evidence that specific neurophysiological markers of attention and working memory function can track commonly reported cognitive symptoms of concussion (Gronwall, 1989;Broglio et al., 2009;Theriault et al., 2009;Ozen et al., 2013).
Our previous study observed both a P3a and P3b response (Ruiter et al., 2019). In contrast, the present study interpreted the current findings as being a different pairing of responses, a P3b and LPC complexdue primarily to differences in response topographies. In the DT condition of the earlier study, both the concussed and control participants exhibited typical fronto-central topographies characteristic of the P3a. However, in the present study, both the FT and DT conditions in each group revealed scalp topographies characteristic of P3b responses ( Fig. 2A, B). In addition, the LPC reflected a type of recollected information process that appeared in a continuous memory-based processing task of button presses to more (the standard) or less (the deviants) frequently occurring stimuli. Within this context, the current findings can be seen as reflecting decisional factors including accuracy of response decisions and, in particular, confidence in response selection (Finnigan et al., 2002). The response did not differ between the two groups, but was not seen at all in the prior study and has not been reported in the primarily adult concussion literature. As a result, the role of the LPC in concussion remains to be determined.
The P3b amplitude reduction found in the present study is consistent with the extensive literature reporting smaller P300 amplitudes associated with concussion (Lavoie et al., 2004;Theriault et al., 2009;Baillargeon et al., 2012); and the current study provides further evidence that specific neurophysiological markers of attention and working memory function can track commonly reported symptoms of cognitive dysfunctions linked to concussion (Gronwall, 1989;Broglio et al., 2009;Theriault et al., 2009;Ozen et al., 2013). P3b amplitudes were most notably smaller than those of controls at frontal and central topographical locations.
Amplitudes of the LPC at frontal sites were found to be significantly reduced when compared with controls ( Fig. 2A, B). Uncovering a LPC in the current study was not anticipated but fits other aspects of the observed data in its reflection of executive function as well as its consequent abnormality in concussed individuals. Based on work examining late positivities and executive functions including work on the frontal selection positivity (FSP) (Kenemans et al., 1993) and the frontal P3 (P3f) (Makeig et al., 1999) (see Perri & Di Russo, 2017 for review), what we refer to as a LPC occurs in the context of memory functions and decision structures that are, like FSP and P3f, linked to executive functions. In the current context, the LPC reflected a type of recollected information process that appeared in a continuous memory-based processing task of button presses to more (the standard) or less (the deviants) frequently occurring stimuli. Within this context, the current findings can be seen as reflecting decisional factors including accuracy of response decisions and, in particular, confidence in response selection (Finnigan et al., 2002).
The hypothesis for the current study, based on Ruiter et al. (2019) that the MMN would differ between the two groups was not supported. This failure to replicate the first examination of the MMN in older individuals with a history of multiple concussions is important. Like the N2b and LPC effects found in the present study, the failure to identify differences in the MMN represents a new piece of information in the examination of concussion from a neurophysiological perspective. The current study is the first examination of the MMN in an acute concussion adolescent population just as our examination of the MMN in the earlier study was the first examination of the response in an adult concussion population. Putting together these two novel sets of data highlights that 1) the older population with a history of many concussions and repeated blows to the head Ruiter et al., 2019) exhibited abnormalities in the MMN while, 2) the adolescent population with fewer than two previous concussions showed no MMN abnormalities at all. These two facts raise the question of whether concussion-related MMN abnormalities represent a biomarker for having reached the point of irreversible neurophysiological and cognitive dysfunction; or, might MMN abnormalities occur earlier at some mid-point between the two samples we have tested and thus serve as a warning of an impending point of irreversibility. To answer this question an examination of the MMN in concussion is needed in different age groups and with different histories of concussion incidences.

Conclusion
In summary, the present study provides a detailed investigation of neurophysiological markers of cognitive dysfunction in concussion as manifested in acutely injured adolescents. The findings of this study highlight the range of cognitive dysfunctions consequent to concussion as reflected by the MMN, N2b, P300 (P3b) and the LPC. The results also support and extend previous literature in demonstrating the unreliability of subjective psychological health tests and evaluations.
The current study provides evidence for possible markers of recovery processes with the finding that N2b response latencies changed (decreased) toward more normal response times as days since injury increased; a finding indicating that decreasing N2b latencies and associated cognitive improvements in executive control, appear to be a marker of, or a prognostic for, concussion recovery. Also, the absence of automatic attention abnormalities (as manifested by the MMN) in this adolescent population contrasts sharply with earlier findings we observed in individuals with more significant and longer histories of concussion. This finding suggests that the MMN abnormalities seen in concussion could serve as a potential marker of irreversible cognitive dysfunction linked to concussion. This proposal could be more than hypothetical given recent positron emission tomography work demonstrating higher tau levels in brain regions affected by chronic traumatic encephalopathy (CTE) (Stern et al., 2019) including brain regions associated with neural populations known to generate the MMN (Alho, 1995;Jemel et al., 2002;Opitz et al., 2002).

Demographic data
Collected through participant self-reporting, the demographic data consisted of each participant's sex and age for both the control and concussed groups, while concussed participants also reported number of previous concussions, and number of days since the most recent concussion to the date of the EEG testing (Table 1).

Behavioral tests
The ImPACT was administered to the concussed population prior to EEG testing ( Table 2). The ImPACT is a computerized neurocognitive test designed to measure sports-related concussions . It is comprised of 6 independent tests providing 5 composite scores (Verbal Memory, Visual Memory, Motor Speed, Reaction Time, and Impulse Control) and a Cognitive Efficiency Index (CEI) score. Scores for the ImPACT were compared against age and gender matched normative data (percentiles) provided by the developers .

Self-Reported symptomatology inventories
The PCSS and CDI (Kovacs, 1992) were used to evaluate concussion and depression symptomatology, respectively (Table 3). The PCSS was used to measure the severity of concussive symptoms such as: fatigue, headaches, and sensitivity to light and noise, whereas, the CDI was used to assess depressive symptoms exclusively.

EEG Task, Stimuli, & experimental conditions
5.5.1. P300 -Active The first protocol employed a P300 active auditory oddball task consisting of 4 tones: 1) Standard Tone (ST, 1000 Hz, 80 dB SPL [sound pressure level], 50 ms duration), 2) Frequency Tone (FT, 1200 Hz, 80 dB SPL, 50 ms), 3) Duration Tone (DT, 1000 Hz, 80 dB SPL, 100 ms), and Intensity Tone (IT, 1000 Hz, 90 dB SPL, 50 ms). Due to technical issues, responses to IT (presented 6% of the time) were discarded from all analyses in this study. The ST was presented 492 times (82% of the stimulus set) while the deviant tones (FT and DT) were presented 36 times each (12% [6% each] of the stimulus set). Tones within the protocol had an inter-stimulus interval (ISI) of 1000 ms. Throughout the duration of the protocol, participants were asked to left-click to every ST and right-click to every deviant tone to ensure they were actively attending to the presented stimuli; response side was counterbalanced within participants.

MMN -Passive
The second protocol administered was a longer version of the same auditory oddball task used in the P300 protocol; however, participants were instructed to ignore the tones and to focus solely on the visuallyneutral silent film presented on the screen in front of them. This protocol was designed to elicit automatic attention /predictive coding processes manifested by the MMN. The ST was presented 1968 times (82% of the stimulus set) while the deviant tones were presented 144 times each (12% [6% each] of the stimulus set). Tones within the protocol had an ISI of 500 ms.
To create a distraction between the two oddball tasks, the protocols were separated by a 10-minute language comprehension task where participants judged the semantic congruity of spoken sentences.

Procedure
Prior to EEG testing, all participants completed the Edinburgh Handedness Inventory (Oldfield, 1971) and a general pre-screen form regarding characteristics such as age, sex, and general medical history, while the concussed group also completed the ImPACT, PCSS, and CDI.
Participants wore noise-cancelling headphones while seated in a comfortable chair facing a computer screen in a sound-attenuated room. In the first protocol, participants were instructed to look at a fixation cross located in the center of the computer screen while they actively listened -and differentially responded -to a series of standard and deviant tones. In the second protocol, participants were instructed to focus solely on a visually-neutral silent film, and that the auditory tone sequence being presented during the film was of no importance to the study. The experiment took approximately 50 min.

Neurophysiological recordings
EEG data were recorded online from 64 Ag/AgCl active electrodes (BioSemi ActiveTwo system) inserted into a flexible cap in accordance with the International 10-20 System. Analog data were recorded at 0.01-100 Hz bandpass and digitized at a sampling rate of 512 Hz with a 60 Hz notch filter. Five Ag/AgCl external electrodes were placed on the nose, each mastoid, as well as above and beside the outer canthus of the left eye. Using the same bandpass and sampling rate, eye movements (electrooculography) were recorded from the two external electrodes placed near the left eye. During EEG acquisition, data were referenced to the DRL (driven right-leg) and were subsequently re-referenced offline to the linked (averaged) mastoids.

EEG data analysis
EEG data were analyzed offline using Brain Vision Analyzer (v2.01) software. Data were filtered with a bandpass of 0.1-30 Hz (24 dB/oct; Duncan et al., 2009). After filtering, visual raw data inspection was conducted to remove the segments containing artifacts not related to eye movements (e.g., muscle movements). The criteria for data removal was +/-50 ÂµV. Following visual inspection and data removal, ocular Independent Component Analysis (ICA) was applied to correct for vertical and horizontal eye movements. Across participants, an average of 31.3 (87%) epochs for the FT condition and 29.5 (82%) epochs for the DT condition were processed in the P300 protocol. In the MMN protocol, an average of 131.8 (92%) and 130.7 (91%) epochs were processed for the FT and DT conditions, respectively. Figs. 1 and 2 were generated using Brain Vision Analyzer by averaging across subjects for each condition per protocol. Representative sites Fz and Cz for the MMN and P300 protocols respectively were used because historically that is where the neurophysiological responses are largest. Data from the P300 protocol were segmented into −200 ms pre-stimulus and 1000 ms post-stimulus onset intervals for all experimental conditions. Similarly, data from the MMN protocol were segmented, changing the interval to −200-600 ms. Following segmentation, data were averaged per condition for each protocol. Only correct responses from the P300 protocol were used for analysis.
Difference waves were calculated only for the MMN by subtracting standard condition ERPs from each of the deviant condition ERPs. For both protocols, automatic peak detection (Barr et al., 1978) was performed within the respective time windows of each ERP component: N1 (75 -125 ms), N2b (170 -270 ms), and the P300 (275 -700 ms) for each condition (ST, FT, and DT) in the P300 protocol, and on the N1 (75 -125 ms) and the MMN (150 -250 ms) in the MMN protocol.
5.9. Statistical analysis 5.9.1. Demographic, Behavioral, and symptomatology data Average age, number of previous concussions, and the average number of days since concussion at the time of EEG testing were tabulated (Table 1) as were average behavioral test scores (Table 2), and levels of concussive and depressive symptomatology (Table 3).

EEG data
In an effort to reduce the inflation of Type 1 errors, we clustered electrodes into nine (9) (Ruiter et al., 2019). The electrode clustering method used in our analyses was conducted in accordance with Luck & Gaspelin's (2017) statistical recommendations for ERP analyses to reduce the likelihood of Type 1 errors in the results. Specifically, the clusters were created where specific ERP responses should be present based on decades of previous research. This step is critical in significantly reducing the familywise error rate (~50%) and acts as a firstpass multiple comparisons correction prior to applying a conservative Bonferroni correction to the statistically significant ANOVA outcomes (Luck & Gaspelin, 2017). Peak amplitude was acquired by taking the average value in a −50 ms to +50 ms time-window around the detected peak; latency was defined as the time from stimulus onset to the maximal point (positive or negative depending on the ERP component) within the defined component windows (see above). Statistical analyses were conducted using R statistical software (R, Version 3.3.3).
Mixed-effects analyses of variance (ANOVAs) were performed to examine the effect of ROI (9 levels: as defined above), Group (2 levels: control and concussed), and Condition (2 levels: FT and DT) on the amplitude and latency of the analyzed ERPs. ANOVAs were conducted independently for each ERP component for both amplitude and latency (independently within each protocol) with an alpha level of p < 0.05. To minimize Type 1 errors, Greenhouse-Geisser adjusted degrees of freedom was applied when the sphericity assumption was violated (Greenhouse & Geisser, 1959). In cases where results showed a significant interaction (Fisher's least significant difference method: Williams & Abdi, 2010), post-hoc analyses were applied to examine the origin of the effects. Bonferroni correction was applied to correct for multiple comparisons within each set of measure [e.g., latency] and single protocol [e.g., P300]). For example, if there exists a significant ROI × Group interaction in the latency of a particular component (e.g., P300), the significance threshold is divided by a factor of 9 (for each ANOVA conducted on each ROI). Notably, post-hocs were examined only when a significant interaction was found.

Correlational analyses
When ANOVAs revealed a main effect of concussion between the two groups, subsequent linear regressions were calculated to assess the relationship between the concussed populations' ERP components' amplitudes and/or latencies and their behavioral, symptom, and demographic data (Baillargeon et al., 2012). All statistical analyses were conducted using R statistical software (R, Version 3.3.3).

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.