Relationships between aggression, sensation seeking, brain stiffness, and head impact exposure: Implications for head impact prevention in ice hockey

Abstract Objectives The objectives of this study were to (1) examine the relationship between the number of head impacts sustained in a season of men's collegiate club ice hockey and behavioral traits of aggression and sensation seeking, and (2) explore the neural correlates of these behaviors using neuroimaging. Design Retrospective cohort study. Methods Participants (n = 18) completed baseline surveys to quantify self‐reported aggression and sensation‐seeking tendencies. Aggression related to playing style was quantified through penalty minutes accrued during a season. Participants wore head impact sensors throughout a season to quantify the number of head impacts sustained. Participants (n = 15) also completed baseline anatomical and magnetic elastography neuroimaging scans to measure brain volumetric and viscoelastic properties. Pearson correlation analyses were performed to examine relationships between (1) impacts, aggression, and sensation seeking, and (2) impacts, aggression, and sensation seeking and brain volume, stiffness, and damping ratio, as an exploratory analysis. Results Number of head impacts sustained was significantly related to the number of penalty minutes accrued, normalized to number of games played (r = .62, p < .01). Our secondary, exploratory analysis revealed that number of impacts, sensation seeking, and aggression were related to stiffness or damping ratio of the thalamus, amygdala, hippocampus, and frontal cortex, but not volume. Conclusions A more aggressive playing style was related to an increased number of head impacts sustained, which may provide evidence for future studies of head impact prevention. Further, magnetic resonance elastography may aid to monitor behavior or head impact exposure. Researchers should continue to examine this relationship and consider targeting behavioral modification programs of aggression to decrease head impact exposure in ice hockey.


INTRODUCTION
There has been growing concern regarding the health-related consequences of the cumulative effect of repetitive head impact exposure (HIE) from sports participation, as athletes of collision sports can sustain hundreds of head impacts in a single season, and thousands over the course of a career (Mainwaring et al., 2018). The consequences associated with greater HIE include changes in brain structure and function, increased risk of concussion, and may even serve as a catalyst to later life neural diseases and disorders (Mainwaring et al., 2018;McKee et al., 2009;Montenigro et al., 2017). Therefore, reducing HIE may be critical to increasing athlete safety and long-term health.
Several interventions to reduce HIE have been studied to date, with most large-scale studies targeting exogenous variables such as when or how often contact can occur, or altering existing equipment (Emery et al., 2019;Stemper et al., 2020;Swartz et al., 2019). However, these interventions have not been consistently successful in reducing HIE, perhaps because they do not address athlete's specific behaviors.
Behavior modification programs encompass athlete-specific strategies to effectively change the way athletes play in order to reduce HIE, though they have only been studied on a smaller scale (Champagne et al., 2019;Combs et al., 2019;Swartz et al., 2019). To date, successful behavior modification programs have included interventions such as education through watching personalized film or receiving individual education by coaches about tackling techniques (Champagne et al., 2019). It is still unknown how athletes' personality traits may influence their HIE, though certain traits have been implicated in affecting the risk of sustaining sports-related concussion. However, modification of such traits has potential to be targeted through psychoeducation-based behavior modification programs; thus, understanding the relationships between these traits and HIE could also serve as a tool for identifying those who may be at risk for greater HIE and related neurological sequalae.
One such trait to consider is sports aggression, which can be defined as "any intentional behavior, not recognized as legal within the official rules of conduct, directed towards an opponent, official, team-mate or spectator who is motivated to avoid such behavior" (Tenenbaum et al., 1997). Existing research has described ice hockey athletes as among the most aggressive in relation to athletes of other sports, and has defined penalty minutes (PM) accrued as a marker of aggression within the hockey community (Pedersen, 2007). Players who accrue more PM are considered more aggressive and may be at greater risk for contact-related injury, including concussion (Agel et al., 2007;Cusimano et al., 2013;Emery & Meeuwisse, 2006). In fact, aggressive body checking has been identified as the primary act leading to concussions in hockey and accounts for the majority of "aggressive" PM assessed by referees (Bushman & Wells, 1998;Carré & McCormick, 2008;Emery et al., 2011;Gee & Leith, 2007). Based on the likelihood that accrual of more PM coincides with increased body contact, it is possible that more aggressive players, as measured by PM, may be more likely to sustain greater HIE.
Another behavioral trait closely related to aggression is sensation seeking, which is defined as "the need for varied, novel, and complex sensations and experiences, and the willingness to take physical and social risks for the sake of such experience" (Zuckerman, 1982). Interestingly, no studies have yet examined the role of sensation-seeking tendencies on HIE, but a recent study found that college athletes who reported higher levels of sensation-seeking tendencies at baseline were more likely to have a history of concussion and be more likely to sustain an in-season concussion (Liebel et al., 2020). As greater HIE has been identified as a potential risk factor for concussion, it is possible that athletes with more HIE may also exhibit greater sensation-seeking tendencies (Beckwith et al., 2013).
Although there seems to be an association between these behaviors and increased risk of physical contact or even head injury, no studies have directly studied how individuals' perceived aggression or sensation-seeking tendencies relate to HIE. Administration of psychometric inventories is one way to quantify these traits and allow athletes to self-report the degree to which they possess them. Several validated inventories exist that measure either aggression or sensation seeking, including the Competitive Anger and Aggressiveness Scale (CAAS) and the Brief Sensation Seeking Scale (BSSS) (Hoyle et al., 2002;Maxwell & Moores, 2007). The CAAS is a survey specifically designed and validated to assess sports-related aggression. The CAAS has previously been used to define aggression in youth ice hockey players, in a study that found players with higher CAAS scores sustained higher magnitudes of mean rotational acceleration per impact (Schmidt et al., 2016). The BSSS is a short survey validated to assess sensation-seeking tendencies. Although it is not sport specific like the CAAS, higher BSSS scores were found to increase the risk of sports-related concussion in a large cohort of collegiate athletes.
Both inventories thus show promise in quantifying athletes' baseline behavioral tendencies that may relate to HIE sustained individually. Therefore, the primary purpose of this study was to examine the relationships between HIE and outcomes from a novel behavioral assessment battery of aggression (CAAS and PM) and sensationseeking (BSSS) tendencies in a single season of collegiate ice hockey.
We performed a secondary, exploratory analysis to examine the relationships between these behaviors and MRI-based measures of neural structure, as the objective nature of neuroimaging can add valuable information on the underlying neural mechanisms that might explain aggression and sensation seeking. Both aggression and sensation seeking are linked to activation in structures of the limbic system and frontal cortex, particularly the amygdala, orbitofrontal cortex (OFC), anterior cingulate cortex (ACC), and ventromedial prefrontal cortex (vmPFC) (Cupaioli et al., 2021;Joseph et al., 2009). Structural neuroimaging analyses have identified that repetitive head impacts may affect structural connectivity and volume of the amygdala, but there is less evidence describing the other structures (Lepage et al., 2019;McAllister et al., 2014). We hypothesized that (1) there would be a positive relationship between the number of head impacts sustained in a season and our measures of aggression and sensation seeking, and (2) there would be a positive relationship between our MRI-based measures of brain structure and our measures of aggression and sensation seeking.

METHODS
Eighteen Division I male collegiate club ice hockey players partici- ). The CAAS is a 12-item questionnaire that measures athletes' perceived aggressiveness and anger during athletic competition (Maxwell & Moores, 2007). Participants self-reported their answers on a 7-point Likert scale, ranging from 1 (strongly disagree) to 7 (strongly agree), for a total score between 12 and 72, in which higher scores denoted more self-reported anger and aggression. The BSSS is a valid and reliable eight-question measure of trait sensationseeking tendencies, adapted from the Sensation-Seeking Scale-Form V (Hoyle et al., 2002). Questions were scored on a 5-point scale ("strongly disagree" to "strongly agree"), where higher scores indicated greater sensation-seeking tendencies.
To quantify HIE, participants wore Smart Impact Monitors (SIM, Triax, Norwalk, CT) at all practice and home games in a season (Cummiskey et al., 2017). Head impacts >10 g were recorded and wirelessly transmitted to a computer with Triax software installed and overseen by a member of the research staff. Each session of each season was video recorded and the film was used to confirm true head impacts and remove false impacts (Patton et al., 2020 ) and damping ratio (Manduca et al., 2001;McGarry et al., 2012;Van Houten et al., 2001). Stiffness and damping ratio reflect the composition and organization of brain microstructure, and both have been shown to be sensitive to brain health (Hiscox et al., 2021;McIlvain et al., 2020;Schwarb et al., 2016).
For both volumetric and MRE analyses, we chose a priori to evaluate the bilateral amygdala, thalamus, hippocampus, OFC, vmPFC, and ACC as our regions of interest (ROIs), as these are structures that have been consistently reported as neural structures that contribute to both sensation seeking and aggression. Volume of ROIs was extracted from T1-weighted images using Freesurfer following a visual inspection of fit quality and completion of the automatic segmentation (Buckner et al., 2004;Fischl, 2012;Fischl & Dale, 1999). Images of the Freesurfer segmentations for each subject are included as Figure S1.   (Buckner et al., 2004). Additionally, we incorporated a priori spatial information during MRE mechanical property estimation using NLI with soft prior regularization (SPR), which has been used to reduce uncertainty in measures McGarry et al., 2013). Masks of our ROIs were provided separately to NLI for use in SPR by registering

RESULTS
Descriptive statistics for CAAS, BSSS, PM norm , and number of head impacts are reported in Table 1  (range: 0-38). A significant relationship was found between (r = .65, p < .01, p-corrected < .01) PM norm and number of head impacts sustained in our cohort, such that more PM norm were related to more head impacts sustained (Figure 1). Only nonsignificant relationships between CAAS, BSSS, and number of impacts were found (all p > .05; Figure 1).
We used preseason values of volume, stiffness, and damping ratio of our ROIs to explore the relationships between neural structure and CAAS, BSSS, PM norm , and number of head impacts (Table 2).
Descriptive statistics for our neuroimaging measures are presented in Table 1. All participants' volumetric data are represented, but MRE data from five of the 15 participants were excluded due to poor signalto-noise ratio (McGarry et al., 2011). We found significant relationships between brain mechanical properties-both stiffness and damping ratio-in several of our ROIs and the CAAS, BSSS, and number of impacts ( Figures S2 and S3). Though these relationships were not still statistically significant at p < .05 after correction for multiple comparisons, these are the first data to describe how brain stiffness may relate to behavioral traits in athletes. BSSS scores were related to stiffness of

DISCUSSION
This study utilized a novel battery of behavioral inventories to assess the relationships among aggression, sensation-seeking tendencies, brain structure, and HIE over a season in men's collegiate club ice hockey players. The main finding of this study was a strong, statistically significant correlation (r = .62) between head impact frequency and PM sustained, suggesting a relationship between more aggressive playing style and sustaining a greater number of head impacts over a season. Our secondary, exploratory analysis revealed that viscoelastic mechanical properties of several neural structures known to regulate aggression and sensation seeking, especially the thalamus, were related to components of our behavioral battery. We did not observe any significant relationships between volume of these structures and our battery, suggesting that measuring the mechanical properties may be a more sensitive measure to use in further understanding the link between personality traits, behavior, and brain structure.
This was the first study to describe relationships between a behavioral assessment battery including a measure of aggression directly  (Bushman & Wells, 1998;Carré & McCormick, 2008;Emery et al., 2011;Gee & Leith, 2007). Given that checking is the primary mechanism for sustaining RHI in hockey, this evidence supports the relationship found between aggression and HIE in this study.
We did not find any significant relationship between a player's perceived competitive aggressive tendencies (CAAS) and HIE, suggesting that using the CAAS may not be useful for understanding behaviors that result in HIE. However, there are several alternate explanations to consider as to why we did not find a relationship between CAAS scores and HIE. In general, there is a high base rate of aggression found in hockey players and therefore there may be a ceiling effect to the CAAS (Bushman & Wells, 1998). It is possible that our participants perceived themselves as similarly aggressive, and thus not enough variability was reported through CAAS scores to have elucidated a relationship to HIE.
The frustration-aggression hypothesis may also help explain why we did not find a significant relationship between HIE and CAAS, considering that we did find a significant relationship between HIE and PM norm .
It states that frustration will occur when an individual cannot achieve a particular goal, and increase the chances he or she will vent frustration in an aggressive manner that is usually directed toward the source of frustration (Berkowitz, 1989). In hockey terms, this means a player would need to be placed in a situation that causes frustration in order to display aggression, which would not be captured through a measure of perceived aggression, like the CAAS, though it could be captured with a direct measure from playing, such as PM norm (Gee & Leith, 2007).
The effects of sensation seeking have not been as frequently studied as aggression in hockey players, but we chose to include this measure because higher reported sensation seeking may be a risk factor of sports-related concussion in collegiate athletes (Liebel et al., 2020;Osborn et al., 2009). We did not find any significant relationship between BSSS and HIE in our cohort, suggesting, like the CAAS, that the BSSS may not be a useful tool in this application. In this case, the BSSS was not specific to sports competition and instead quantified the perceived desire to seek out certain experiences, but not engagement in directed acts. In a hockey context, this may or may not include engaging in body contact and therefore support why we did not find a relationship between BSSS scores and HIE. However, it may still be worth monitoring BSSS scores in hockey athletes despite this lack of relationship, as higher BSSS scores have been reported as related to increased risk of sustaining a concussion (Liebel et al., 2020).
In addition to assessing these relationships, we also chose to explore how differences in MRI-measured brain structure may affect both scores on our behavioral battery and HIE in a season of ice hockey.
The objective nature of neuroimaging, as compared to self-reported inventories or behavioral assessments, can aid in solidifying our understanding of the mechanisms linking aggression or sensation seeking and HIE. Furthermore, it can help us to understand who may be more likely to display such traits, and therefore at risk of the consequences associated with those traits. We found that baseline values of both stiffness and damping ratio were related to both aggression and sensation seeking, but structural volume was not. Though these relationships did not exhibit significant corrected p-values after correction of multiple comparisons, the results from this exploratory analysis are consistent with previous brain MRE literature, which has reported that changes in mechanical properties of neural tissue can occur in the absence of changes in volume or thickness (Hiscox et al., 2021).
Little is known about the relationship between brain stiffness and personality traits, but adolescents who engaged in more risky behaviors trended toward having greater stiffness in neural structures related to risk taking (McIlvain et al., 2020). We found thalamic stiffness was related to sensation seeking, such that greater stiffness was related to higher BSSS scores (r = .71). In general, higher stiffness is thought to reflect higher tissue integrity and neural health, as it results from a denser glial matrix or healthier myelin sheaths (Johnson & Telzer, 2018). Our results suggest that a thalamus with greater integrity may be more likely to exhibit a greater degree of sensationseeking traits. We also found a significant relationship between amygdala stiffness and BSSS scores (r = .65), which may be explained by similar reasoning. The amygdala communicates with the cortex in order to make decisions regarding emotion, which could explain why both greater stiffness of the amygdala and thalamus related to more pronounced perceptions of sensation seeking in our participants (Baas et al., 2004;Bell & Shine, 2016;Sherman, 2007).
In addition to stiffness, we also found significant relationships between BSSS scores and damping ratio. While stiffness is thought to reflect microstructural composition, MRE literature suggests that damping ratio reflects microstructural organization and is more correlated with brain function (Johnson et al., 2018;Sack et al., 2013;Schwarb et al., 2016). BSSS scores were inversely related to vmPFC damping ratio, such that higher BSSS was related to lower vmPFC damping ratio. The vmPFC has a more targeted role in controlling an individual's response to emotion, whereas the amygdala evaluates emotions from stimuli, and the thalamus communicates with the two (Baas et al., 2004;Bell & Shine, 2016;Sherman, 2007). It is fitting that we would find a relationship with damping ratio in the more functionally specific vmPFC and BSSS, as it has been suggested that damping ratio is generally more correlated with brain function than stiffness due to heightened sensitivity to individual differences (Hiscox et al., 2021).
Altogether, our results provide initial support toward the clinical utility of monitoring PM to determine which hockey players may be at risk for greater HIE due to heightened aggressive behaviors. Finding interventions that lower aggression, such as behavioral modification through psychoeducation, may be a novel approach to lower HIE and preventing subsequent neuropathologies (Bazarian et al., 2014;McKee et al., 2009;Montenigro et al., 2017). No studies have yet examined how using psychoeducation to modify traits such as aggression affects HIE or head injury risk in collegiate athletes, but behavioral modification programs targeting football-specific physical behavior (e.g., tackling) have been successful in reducing HIE in youth athletes (Champagne et al., 2019;Combs et al., 2019). Interestingly, despite belief that aggression will benefit individual or team performance, this relationship has not been proven (Gee & Leith, 2007). This is an exciting finding, as lowering aggression may help keep athletes safer without impairment or negatively affecting performance.
Despite our novel findings, there are several limitations to this study that may affect generalizability of our results. First, we were limited to a small sample based on the size of the ice hockey team studied, but our sample size was comparable to existing studies that have also utilized our more advanced measures of MRI and head impact monitoring. We were limited in our ability to collect head impact data from games away from the home stadium, although it is likely that individual's HIE would trend in a similar direction, thus we expect our measure of HIE to accurately reflect the total exposure also incorporating away games. Additionally, our results may have been affected by the inherent limitations associated with self-reported measures. Due to the subjective nature of such measures, it is not possible to know whether participants responded honestly or accurately to the questions on the CAAS or BSSS, though we were able to include an objective measure of aggression (PM norm ) to partially combat this limitation. Although we ran our secondary analysis as exploratory, we were left with a low MRE sample due to insufficiencies in data quality and thus unequal samples between the two analyses. However, this is the first study to describe brain viscoelasticity in relation to either athletes or repetitive head impacts and may serve as preliminary evidence for further use of MRE in assessing the causes and effects of HIE. As brain MRE is a relatively new neuroimaging modality, future studies should also consider how the relationships between MRE and behavior compare to those between behavior and other imaging methods sensitive to tissue microstructure (e.g., diffusion tensor imaging) to develop a comprehensive understanding of how brain structure may influence expression of personality traits.

CONCLUSION
This study provides important results that help our understanding of HIE risk, as it is the first study to examine how personality traits, brain structure, and HIE are related. Our results suggest that monitoring PM may be useful in determining athletes at risk for greater HIE, and that behavior modification targeting aggression may be useful in reducing HIE. However, our design and methods should be considered as preliminary evidence for future research with larger cohorts on this topic.

CONFLICT OF INTEREST
The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT
Data available on request from the authors.