Balance Markers and Saccadic Eye-Movement Measures in Adolescents With Postconcussion Syndrome

Two groups of participants were recruited for this study. The concussion group consisted of 25 adolescents (10 boys, 15 girls; age range = 11–16 years; median [interquartile range (IQR)] age = 14 years [11.5–16 years]; time since concussion = 1–3 months) receiving care for postconcussion syndrome at a regional concussion clinic. Postconcussion syndrome was diagnosed if they experienced 1 or more symptoms for at least 4 weeks in 1 or more of the following clinical domains outlined by the 5th International Conference on Concussion in Sport²:

• Symptoms: somatic (eg, headache), cognitive (eg, feeling like in a fog), or emotional (eg, lability);

• Physical signs (eg, loss of consciousness, amnesia, neurologic deficit);

• Balance impairment (eg, gait unsteadiness);

• Behavioral changes (eg, irritability);

• Cognitive impairment (eg, slowed reaction time);

• Sleep or wake disturbance (eg, somnolence, drowsiness).

The control group consisted of 33 uninjured adolescents (18 boys, 15 girls; age range = 11–16 years; median age = 13 years [IQR = 11.5–14]) who reported no concussive symptoms and had not sustained head trauma in the year before the study. This group was recruited from 2 local private schools. To be included in the study, participants in both groups were required to be able to read text on a board or computer tablet placed 60 cm in front of them without prescription glasses. Corrective contact lenses were permitted. Baseline characteristics for participants in both groups are summarized in Table 1. All participants and their parents provided written informed assent and consent, respectively, and the study was approved by the Children's Hospital of Eastern Ontario Research Ethics Board and the Health Sciences and Sciences Research Ethics Board at the University of Ottawa.

The concussion group completed the Post-Concussion Symptom Inventory before performing the balance protocol.¹⁷  This inventory consists of a list of concussive symptoms for which participants are instructed to rate their preinjury and postinjury degree of symptoms. The 11- and 12-year-old participants completed the version for children aged 8 to 12 years (17 symptoms, 3-point graded scale), and participants between ages 13 and 16 years completed the version for adolescents aged 13 to 18 years (20 symptoms, 7-point graded scale).

All participants completed 3 dual-task balance conditions while standing on a Wii Balance Board (Nintendo Co, Ltd, Kyoto, Japan) that recorded the movement of the COP under their feet. While completing these conditions, they wore a pair of Tobii Pro Glasses 2 (Tobii Pro, Stockholm, Sweden) that recorded their eye movements. Participants were free to move their heads while completing these conditions. For each dual-task balance condition, the secondary task consisted of a different version of a Stroop task that included primarily a cognitive component (cognitive condition), primarily a gaze-shifting component (visual condition), or both cognitive and gaze-shifting components (combined condition). The experimental setup is illustrated in Figure 1. The concussion group completed the protocol once between 1 and 3 months postinjury (median number of days = 39.5 [IQR = 30–56.75]). The control group completed the protocol once.

For the cognitive condition, a visual Stroop task was displayed on a tablet (Windows 8 Surface Pro; Microsoft Corp, Redmond, WA) placed at eye level 60 cm in front of the participant. One hundred words were presented one at a time in the center of the participant's visual field. The words red, yellow, blue, and green were presented in a random order and written in an incongruent ink color (eg, red printed in blue). The participant was instructed to name the color of the ink when the word appeared on the tablet.

For the visual condition, the Stroop task was displayed on a board placed 60 cm in front of the participant. The board contained 20 rows of 5 words consisting of a random series of the words red, yellow, blue, and green written in a congruent ink color (eg, red printed in red). The participant was instructed to read the words from left to right, beginning with the first row, at a comfortable pace.

For the combined condition, the Stroop task was displayed on a board placed at 60 cm in front of the participant. The board contained 20 rows of 5 words consisting of a random series of the words red, yellow, blue, and green written in an incongruent ink color (eg, red printed in blue). The participant was instructed to name the color of the ink of each word from left to right, beginning with the first row, at a comfortable pace.

The board used for the visual and combined conditions was 40-cm wide, which translated to a visual range of 36.9° when placed 60 cm in front of the participant. For all 3 conditions, we stopped participants at 3 minutes, regardless of whether they had completed the task. For the cognitive and combined conditions, we calculated an accuracy percentage by dividing the number of correct responses by the number of total responses and then multiplying by 100.

The visual and combined conditions contained a gaze-shifting component in which the participants were required to shift their point of gaze from 1 word to the next. The cognitive condition did not include a gaze-shifting component because the words were presented in the center of the participants' visual field. In contrast to the visual condition for which the participants were required to read the words displayed on the board, the cognitive and combined conditions increased the cognitive load by requiring them to suppress the meaning of the word while naming the color of the ink.

The raw pressure data from the Wii Balance Board were directly recorded on a laptop computer (MacBook Pro; Apple, Inc, Cupertino, CA) via a Bluetooth (Bluetooth SIG, Inc, Kirkland, WA) device. The raw pressure data were transformed to the COP in the lateral and anterior-posterior (AP) directions using a custom MATLAB (The MathWorks, Inc, Natick, MA) script. The COP data were resampled at a frequency of 60 Hz using a sliding window average with relevant interval interpolation.¹⁸  The velocity of the COP in the AP and medial-lateral (ML) directions and a COP 95% ellipse area were calculated for each dual-task balance condition for each participant.

The Tobii Pro Glasses 2 consist of a binocular eye-tracking system. Eye movements were captured by the eye tracker at a frequency of 100 Hz. The recorded data were imported into the Tobii Pro analysis software and segmented into fixations and saccades using a velocity threshold identification classification algorithm. The velocity threshold was set at 30°/s, and the window length was set at 20 milliseconds. Fixations shorter than 60 milliseconds were discarded, and gaps in fixations lasting less than 75 milliseconds were filled in using a gap fill-in algorithm. In addition, consecutive fixations separated by 75 milliseconds or less and with a maximum angle of 0.5° between the last sample of the first fixation and the first sample of the second fixation were merged. Both voluntary saccades and saccades produced to stabilize gaze in response to head movement were captured. Although microsaccades were likely produced when fixating on words, the identification classification algorithm categorized these minor eye movements as part of a fixation. The number of saccades completed by each participant during each balance condition was identified and was expressed as the number of saccades completed divided by the number of words completed in the condition.

Sample size was calculated based on a 2-tailed analysis with the α level set at .05 and power set at 80%. It was calculated using data from a study⁹  in which the researchers measured balance using COP measures in an adolescent population with concussion. A sample size of 22 participants in each group was needed to detect group differences in COP measures.

For each COP dependent variable and the number of saccades completed, separate 2-way (condition × group) repeated-measures analyses of variance were used to test for differences between conditions (within-subject factor) and groups (between-subjects factor). When a main effect was observed for group, we performed post hoc comparisons using independent-samples t tests to identify the conditions in which a between-groups difference was present. For the independent-samples t tests, the α level was set at .017 to adjust for multiple comparisons. We also used independent-samples t tests to determine the presence of differences in the accuracy percentages on the Stroop tasks between groups for both the cognitive and combined conditions. We calculated Cohen d effect sizes for each comparison between groups for the COP measures, the number of saccades completed, and the accuracy percentage on the Stroop tasks. All analyses were completed using SPSS (version 25; IBM Corp, Armonk, NY).

The concussion group was older and experienced more previous concussions than the control group (P values = .04 and .008, respectively). The incidence of specific postconcussion symptoms present when the concussion group completed the protocol is summarized in Table 2.

We observed a main effect of condition for ML velocity (F_2,112 = 12.1, P < .001), the 95% ellipse area (F_1.86,104.31 = 7.25, P = .001), and the number of saccades completed (F_1.51,84.5 = 25.03, P < .001). No main effect of condition was noted for AP velocity (F_2,112 = 1.39, P = .25). A main effect of group was identified for the 95% ellipse area (F_1,56 = 12.5, P = .001). Post hoc comparisons showed that the concussion group swayed over a larger ellipse area than the control group during the visual (P = .02) and combined (P = .005) conditions but not during the cognitive condition (P = .07; Figure 2A). No main effect of group was demonstrated for AP velocity (F_1,56 = 2.57, P = .12; Figure 2B), ML velocity (F_1,56 = 0.157, P = .69; Figure 2C), or the number of saccades completed (F_1,56 = 2.04, P = .16; Figure 3). Effect sizes for each comparison between groups are presented in Table 3. No group difference was observed for the percentages of accuracy on the Stroop task for the cognitive (P = .07) or combined (P = .35) condition (Table 4).

The purpose of our study was to compare balance measures and saccades between adolescents 1 to 3 months postconcussion and healthy uninjured adolescents while they completed 3 dual-task balance conditions involving primarily a cognitive component, primarily a gaze-shifting component, or both cognitive and gaze-shifting components. In agreement with our initial hypothesis, the concussion group swayed over a larger COP ellipse area than the control group during the visual and combined conditions. However, in contrast to our hypothesis, the concussion group did not sway over a larger COP ellipse area during the cognitive condition. The concussion group also did not perform the visual and combined conditions with greater COP ML velocity and did not display more saccades under these conditions. Thus, the visual and combined conditions elicited greater amounts of sway in the concussion group, but these larger amounts of sway were not associated with more saccades.

Researchers^13,14  have shown that individuals with traumatic brain injuries, including concussion, exhibit more saccades than necessary when completing an eye-tracking task. In these studies, participants were instructed to complete a simulated reading task that required saccades to track a dot that moved from left to right on a screen. Although they were told to complete a single saccade to shift their point of gaze toward the dot's new position, they demonstrated an excessive number of saccades to accomplish this task. In our study, the concussion group did not display more saccades than the control group while completing any of the dual-task balance conditions, although the visual and combined conditions required participants to shift their point of gaze from 1 word to the next. These diverging results may be related to the presentation of the eye-movement task. In our investigation, the words were separated by equal distances, whereas in the simulated reading task used in the previous studies,^13,14  the dot moved across the screen, and the distance between the dot's displacements varied. Therefore, it is possible that our concussion group quickly learned to adjust the amplitude of their saccades because the distance remained consistent, and all of the words were visible at the start of the trial. In addition, in the previous work,^13,14  the head was fixed while participants completed the simulated reading tasks, whereas in our study, participants could move their heads while performing the visual and combined conditions. Therefore, the concussion group in our study may have compensated with head rotation to shift their point of gaze from 1 word to the next rather than relying on saccades to accomplish the task.

The concussion and control groups performed the Stroop task during the combined and cognitive conditions with similar accuracy percentages. The Stroop Color and Word Test measures executive function,¹⁹  and individuals may demonstrate declines in executive function postconcussion.^20,21  Whereas we observed no difference in the accuracy percentages for the Stroop tasks, the concussion group's executive functions may have recovered by the time the participants completed the protocol between 1 and 3 months postinjury. It is also possible that the concussion group allocated more attention to the Stroop task than to the balance task. According to the capacity-sharing model of dual tasks, the concurrent performance of a cognitive and motor task depends on a general attentional resource pool that allows simultaneous performance of both tasks. When the attentional demands of performing both tasks concurrently exceed the attentional resource pool, performance decrements in 1 or both tasks are observed.²²  Therefore, the concussion group may have allocated more attention to the Stroop task, leading to decreased performance on the balance task.

The concussion group did not complete the Stroop tasks with more errors, exhibit more saccades, or sway faster in the ML direction than the control group during the dual-task balance conditions; however, the participants swayed over larger COP ellipse areas during the visual and combined conditions. In addition, although the concussion group did not sway over larger COP ellipse areas during the cognitive condition, they showed a trend toward larger ellipse areas under this condition. A larger sample size might have been better powered to detect a group difference for this variable and condition. Together, these results demonstrate that completing a secondary task that involved a gaze-shifting component affected the concussion group's balance to a greater extent than completing a secondary task that did not include a shift in their point of gaze.

The Stroop task during the visual and combined conditions may have interfered with the concussion group's balance to a greater extent due to head rotation. The board used to display the Stroop task under these conditions had a visual range of 36.9,° which fell outside of the 5° of central vision that individuals use to focus on words while reading.²³  Hence, participants were required to shift their point of gaze from 1 word to the next by either executing saccades or rotating their heads. Whereas these results were not statistically different, the concussion group on average displayed fewer saccades than the control group in all dual-task balance conditions. This suggests that, when given the option, participants with concussion may choose to compensate with head rotation instead of saccades. In addition, given that the vestibular system is sensitive to rotational accelerations of the head,²⁴  the head rotation involved in the visual and combined tasks would have increased the amount of vestibular information to be processed and integrated by the central nervous system. Murray et al⁵  suggested that balance deficits postconcussion were due to the inability to efficiently organize and integrate sensory information. Thus, the possible head rotation involved in the visual and combined conditions may have increased the amount of vestibular information to be processed, leading to less efficient balance control in the concussion group. Furthermore, the larger group difference observed in the combined versus visual condition may reflect increases in both vestibular feedback and cognitive load.

Our study had limitations. First, the sampling frequency of the Tobii Pro Glasses 2 was low. Although the low sampling frequency means that the system may fail to detect every saccade,²⁵  it has the advantage of allowing individuals to move freely within their environments. In contrast, eye trackers with higher sampling frequencies are screen based and require individuals to keep their heads fixed while sitting in front of a screen. It would not have been possible to complete our study with one of these eye trackers, as we needed participants to remain standing and be free to move naturally so that we could accurately capture the balance data. Second, the concussion and control groups were not matched based on age, sex, or other covariates, such as the number of previous concussions and baseline levels of physical activity. The concussion and control groups were recruited simultaneously, so it was not possible to match the participants based on these characteristics. Third, not all participants with concussion had prolonged impairment of balance or saccadic eye movement postconcussion. Therefore, potential group differences in the COP velocity variables and number of saccades could have been “washed out” due to this limitation. More group differences in these measures might have emerged if participants completed the protocol closer to the time of injury. Despite these limitations, our results highlight important information regarding dual-task balance testing after concussion.

To our knowledge, our investigation is one of only a few that have measured balance and saccades simultaneously in a population with concussion. Our finding of no differences in the number of saccades between groups may be explained by the concussion group's possible preference for using head rotation rather than eye movements to shift the point of gaze. These results have important implications for the safe practice of sports, as well as participation in everyday activities, because sports and other common activities often require the simultaneous integration of balance, saccades, head rotation, and cognitive abilities. Future researchers should focus on distinguishing the effects of saccades from the effects of head rotation on balance performance in a population with concussion.

Performing dual-task balance conditions for which the secondary task involved a gaze-shifting component (ie, visual condition) or both a gaze-shifting component and a high cognitive load (ie, combined condition) resulted in greater sway amplitude in adolescents with concussion relative to healthy uninjured adolescents. However, these larger amounts of postural sway observed in the adolescents with concussion were not associated with increased saccadic eye movements.

This research was supported by the Ontario Graduate Scholarship award (Dr Rochefort).

We thank the parents and children for their participation.