RESEARCH ARTICLE Self-Reported Head Trauma Predicts Poor Dual Task Gait in Retired National Football League Players Brad Manor, PhD ,1,2,3 Junhong Zhou, PhD ,1,2,3 On-Yee Lo, PhD ,1,2,3 Hao Zhu, MA,1 Natalia A. Gouskova, PhD,1,2,3 Wanting Yu, ME,1 Ross Zafonte, DO ,3,4 Lewis A. Lipsitz, MD,1,2,3 Thomas G. Travison, PhD,1,2,3 and Alvaro Pascual-Leone, PhD1,3,5 Objective: Symptomatic head trauma associated with American-style football (ASF) has been linked to brain pathology, along with physical and mental distress in later life. However, the longer-term effects of such trauma on objective met- rics of cognitive–motor function remain poorly understood. We hypothesized that ASF-related symptomatic head trauma would predict worse gait performance, particularly during dual task conditions (ie, walking while performing an additional cognitive task), in later life. Methods: Sixty-six retired professional ASF players aged 29 to 75 years completed a health and wellness question- naire. They also completed a validated smartphone-based assessment in their own homes, during which gait was moni- tored while they walked normally and while they performed a verbalized serial-subtraction cognitive task. Results: Participants who reported more symptomatic head trauma, defined as the total number of impacts to the head or neck followed by concussion-related symptoms, exhibited greater dual task cost (ie, percentage increase) to stride time variability (ie, the coefficient of variation of mean stride time). Those who reported ≥1 hit followed by loss of con- sciousness, compared to those who did not, also exhibited greater dual task costs to this metric. Relationships between reported trauma and dual task costs were independent of age, body mass index, National Football League career dura- tion, and history of musculoskeletal surgery. Symptomatic head trauma was not correlated with average stride times in either walking condition. Interpretation: Remote, smartphone-based assessments of dual task walking may be utilized to capture meaningful data sensitive to the long-term impact of symptomatic head trauma in former professional ASF players and other con- tact sport athletes. ANN NEUROL 2020;87:75–83 R epeated exposure to biomechanical head trauma— especially during participation in contact sports such as American-style football (ASF)—has been linked to sub- current efforts to identify those at risk of developing chronic brain disease, as well as to design interventions aimed at assessing, preventing, and ultimately alleviating jective complaints of physical, mental, and emotional dis- the potential clinical disability. There is thus an urgent tress in life,1 as well as chronic traumatic encephalopathy need to identify and validate low-cost and widely accessi- upon autopsy.2 However, the longer-term effects of such ble cognitive–motor assessment tools, and subsequently, repeated head trauma on objective, clinically meaningful characterize and prospectively track large numbers of indi- metrics of cognitive–motor function remains largely viduals who have been exposed to repetitive head trauma unknown.3 This gap in knowledge significantly hinders in early life. View this article online at wileyonlinelibrary.com. DOI: 10.1002/ana.25638 Received Apr 16, 2019, and in revised form Sep 7, 2019. Accepted for publication Nov 3, 2019. Address correspondence to Dr Manor and Dr Zhou, Hinda and Arthur Marcus Institute for Aging Research, Hebrew SeniorLife, Boston, MA 02131. E-mail: bradmanor@hsl.harvard.edu and junhongzhou@hsl.harvard.edu From the 1Hinda and Arthur Marcus Institute for Aging Research, Hebrew SeniorLife, Boston, MA; 2Division of Gerontology, Department of Medicine, Beth Israel Deaconess Medical Center, Boston, MA; 3Harvard Medical School, Boston, MA; 4Spaulding Rehabilitation Hospital, Massachusetts General Hospital, and Department of Physical Medicine and Rehabilitation, Brigham and Women’s Hospital, Boston, MA; and 5Guttmann Neurorehabilitation Institute, Autonomous University of Barcelona, Badalona, Spain © 2019 The Authors. Annals of Neurology published by Wiley Periodicals, Inc. on behalf of American Neurological Association. 75 This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. ANNALS of Neurology Assessment of walking provides clinically meaningful iPhone also participated in “TeamStudy,” which used an app insights into cognitive–motor function and underlying designed for iPhones to gather health data that included a remote brain health.4 In addition to the peripheral and spinal assessment of dual task walking performance. neuromuscular systems, the complex task of walking depends upon “higher-level” sensory–motor integration Ethics Committee Approval and various cognitive functions.5 Moreover, in daily life, All TeamStudy procedures and its informed consent form were approved by the Beth Israel Deaconess Medical Center Institu- walking is almost always performed in the context of cog- tional Review Board. The consent form was integrated into the nitive tasks such as talking, reading signs, or making deci- app, and all participants reviewed the form, were quizzed on its sions. Within experimental paradigms, such dual tasking content to ensure understanding, and consented to the study as (eg, walking while counting backward) affects numerous part of the app’s setup procedures. Each participant completed metrics of gait.6 Acute, sport-related head trauma in youn- the FPHS health and wellness questionnaire and the TeamStudy ger adults has been linked to greater dual task cost (ie, per- walking assessment within 1 year of one another. centage decrement) to gait performance.7,8 At the same time, worse dual task gait performance in older adults has Self-Reported Health, Wellness, and Football been linked to executive cognitive dysfunction,9 chronic Exposures pain,10 increased likelihood of suffering a fall,11 increased Self-Reported Demographics, Medical History, and Health risk of converting from mild cognitive impairment to Status. We extracted data related to: (1) age at time of dementia,12,13 alterations in brain structure,14 and a questionnaire completion; (2) race; (3) history of musculo- reduced capacity to activate appropriate brain networks in skeletal surgery involving major joints that may significantly response to cognitive–motor tasks.15 influence gait (neck, spine, hips, knees, or ankles); (4) report Based upon the above evidence, we hypothesized of physician-diagnosed hypertension, arthritis, diabetes, that metrics of dual task walking performance would be stroke, Parkinson disease, dementia, and depression; sensitive to chronic brain dysfunction induced by repeated (5) chronic pain; (6) global physical function; and (7) global sports-related head trauma in young adulthood, and there- mental function. Chronic pain was assessed by the ques- fore, could ultimately be useful for the detection of conse- tion, “In the past 7 days, how would you rate your pain on quences of such head trauma, potentially before cognitive average on a scale of 0 (no pain) to 10 (worst pain imagin- complaints and other clinical symptoms arise. To this end, able).” Global physical health and global mental health were we developed and validated a smartphone app that utilizes queried using the Patient-Reported Outcomes Measure- the phone’s internal movement sensors and offers multi- ment Information System (PROMIS) global health instru- media instructions to enable remote assessment of dual ment and standard scoring procedures.17 task walking performance in larger numbers of individ- uals.16 The purpose of this study was to examine the Football-Related Exposures. We extracted the age at which cross-sectional relationship between dual task walking per- participants first played organized football, the number of formance and self-reported history of football-related seasons played in the National Football League (NFL), and symptomatic head trauma during their professional the position most often played professionally. For the latter, playing careers in a subsample of former professional ASF participants were classified as either linemen (including players participating in the Football Players Health Study offensive or defensive line) or nonlinemen (including line- (FPHS) at Harvard University. backer, defensive back, running back, wide receiver, tight end, quarterback, kicker/punter, special teams). Players were asked to report the number (with Subjects and Methods choices of 0, 1, 2–5, 6–10, and 11 or more) of times they Participants experienced a “hit or blow to the head, neck, or upper The FPHS is a portfolio of studies intended to further our body followed by…” each of the following symptoms: understanding of the benefits and risks of participation in ASF, headaches, nausea, dizziness, loss of consciousness, mem- identify risks that are potentially preventable or reversible, and ory problems, disorientation, confusion, seizure, visual develop interventions or approaches to improve the health and problems, weakness on one side of the body, and feeling well-being of former professional ASF players. At the time of this unsteady on your feet. We then created a continuous study, the FPHS comprised 3,506 former professional ASF players who had completed a health and wellness questionnaire “symptomatic head trauma” score for the total burden of probing demographics, football exposure including history of hits followed by concussion-related symptoms (by coding football-related symptomatic head trauma, medical history, phys- each response as 0, 1, 3.5, 8, and 13, and then summing ical function, cognitive health, and mental well-being. A self- the responses for all 11 symptoms). We also created a selected subset of this cohort (n = 66) of subjects who owned an yes/no dichotomous variable for at least 1 self-reported hit 76 Volume 87, No. 1 Manor et al: Head Trauma and Dual Task Gait followed by loss of consciousness, as we felt that this item derived from each walking condition separately. Only in particular may be more reliably determined by self- these specific metrics were chosen because each (1) can be report. accurately derived from smartphone motion data captured during walking with the phone in the pants pocket; and Remote Assessment of Dual Task Walking (2) is sensitive to cognitive status,9,12,13 chronic pain,10 Performance falls,11 and markers of brain health14,19 in older adults. Data Acquisition. Dual task walking performance was measured remotely in the participant’s home by the vali- Statistical Analysis dated TeamStudy smartphone app, created in collabora- Descriptive statistics were used to summarize cohort characteris- tion with Sage Bionetworks (Seattle, WA).16 Participants tics and gait outcomes, followed by model estimation and first watched an animation developed by Wondros (Los hypothesis testing. We examined relationships between self- Angeles, CA) that provided a general overview of the reported burden of hits followed by concussion-related symp- assessment. The user was then presented with several on- toms and primary and secondary gait outcomes using 3 separate screen text instructions. The gait assessment was designed approaches. First, 1-way analysis of variance (ANOVA) was used to examine the main effect of symptomatic head trauma score to evaluate walking under normal conditions, and again quartile on each gait outcome. Second, linear regression models while the subject performs verbalized serial subtractions of were used to examine the correlation between the symptomatic 3 from a random number between 200 and 999. head trauma score, modeled as a continuous variable (see above), Participants were encouraged to complete the walk- and each gait outcome. Third, 1-way ANOVAs were used to ing assessment in an empty room or hallways within their examine the influence of self-report of at least 1 hit to the head own home. Each assessment included one 45-second trial or neck followed by loss of conscious on each gait outcome. In of normal walking followed by one 45-second trial of dual each of these 3 analyses, models were completed both with and task walking. Trails were separated by at least 90 seconds. without adjustments for age, body mass index (BMI), exposure Trial start and end cues triggered acquisition of accelerom- to ASF defined as the number of NFL seasons played, and inju- eter, gyroscope, and magnetometer data, which was stored ries unrelated to head trauma that may influence gait as defined in the phone’s internal storage. Kinematic and question- by the total number of reported musculoskeletal injuries to the naire data were then automatically uploaded via wi-fi to a neck, spine, hips, knees, or ankles. Tests allowed for type I error probability of α = 0.05. Analyses were performed using JMP remote, cloud-based data server for offline analyses. software version 13 (SAS Institute, Cary, NC). Data Capturing and Preprocessing. We extracted 2 out- comes from each walking trial—average stride time and Results stride time variability—using procedures that have been Participant Characteristics shown to produce accurate results when compared to the Sixty-six participants of the FPHS joined TeamStudy in GAITRite mat within the laboratory, and high test–retest response to a broadly publicized advertisement and used reliability over multiple days in user homes.16 their smartphone to complete a remote dual task gait Walking trials completed within user homes typi- assessment. Two of these participants were excluded from cally included variable amounts of turning, which alters analysis because, after removal of periods of turning dur- stride timing.18 Turning also influences z-axis accelera- ing the assessment, <10 were identified in single and dual tions derived from the phone and thus the ability to accu- task trials. Within all trials of the included 64 participants, rately detect heel strikes. We therefore utilized a method a mean of 3.5 (standard deviation [SD] = 2.1) periods of we previously developed and validated to identify and turning were detected and removed. Twenty-seven phone- remove relatively sharp, rapid turns using rotated z-axis side leg strides (SD = 8, range = 12–35 strides) were iden- gyroscope data (ie, rotations about the global vertical axis). tified from the remaining, relatively straight periods of Stride times for the current analysis were thus calculated walking. Numbers of identified turns and strides did not from all z-axis acceleration data that occurred outside of differ between single and dual task conditions. detected periods of turning. To ensure reliability of Compared to the entire FPHS cohort, the TeamStudy results, we excluded any trials that contained <10 identi- subsample was similar in age, height, and body mass, but had fied strides. a greater proportion of white participants (Table 1). The Primary gait outcomes included the dual task cost to TeamStudy subsample was similar to the FPHS cohort in age stride time and dual task cost to stride time variability; at first exposure to organized ASF, number of NFL seasons that is, the percentage change in each outcome from nor- played, and proportion of linemen to nonlinemen. As com- mal walking to dual task walking conditions. Secondary pared to the FPHS cohort, TeamStudy participants reported outcomes included stride time and stride time variability greater symptomatic head trauma scores and were more likely January 2020 77 ANNALS of Neurology TABLE 1. Self-Reported Demographic, Health, and Football-Related Characteristics of the TeamStudy Subsample Compared to the Total Population of the FPHS at Harvard University Characteristic FPHS TeamStudy Subsample Sample size 3,506 64 Age, yr, mean (SD) 52.8 (14.2) 52.7 (12.4) Height, in, mean (SD) 74.3 (2.5) 74.9 (2.7) Body mass, lb, mean (SD) 239.6 (42.0) 240.7 (41.3) Race, n (%) white 2,053 (58.6) 47 (73.4) Age at first organized football, yr, mean 11.8 (3.1) 11.7 (3.0) (SD) Professional seasons, n, mean (SD) 6.8 (3.8) 6.5 (4.5) Primary position played, n (%) Lineman 1,248 (35.6) 28 (43.8) Nonlineman 2,258 (64.4) 36 (56.2) Concussion-related symptoms Symptomatic head trauma score, 32.0 (29.2) 21.9 (14.8) mean (SD)a Loss of consciousness, n (%)b 1,763 (50.2) 31 (48.4) Physician-diagnosed medical conditions, n (%) Hypertension 1,325 (37.8) 23 (35.9) Diabetes 313 (8.9) 2 (3.1) Parkinson disease 24 (0.7) 1 (1.6) Stroke 99 (2.8) 2 (3.1) Arthritis 1,785 (50.9) 30 (46.9) Dementia 125 (3.6) 0 (0.0) Depression 312 (8.9) 1 (1.6) History of musculoskeletal surgery, n 2,115 (60.3) 42 (65.6) (%)c Global pain, mean (SD)d 4.8 (2.5) 3.75 (2.3) PROMIS physical health, mean (SD)e 48.4 (8.1) 51.8 (6.8) e PROMIS mental health, mean (SD) 43.4 (11.3) 49.4 (9.5) a Symptomatic head trauma score: sum of reported number of hits to the head or neck followed by 1 of 11 concussion-related symptoms. b Loss of consciousness: report of at least 1 hit to the head or neck followed by loss of consciousness. c Musculoskeletal surgery: at least 1 reported surgery for the neck, spine, hips, knees, or ankles. d Global pain: response to the question, “In the past 7 days, how would you rate your pain on average on a scale of 0 (no pain) to 10 (worst pain imaginable)”. e PROMIS global physical and mental health were t-scored using standard procedures. FPHS = Football Players Health Study; PROMIS = Patient-Reported Outcomes Measurement Information System; SD = standard deviation. 78 Volume 87, No. 1 Manor et al: Head Trauma and Dual Task Gait to report depression as a medical condition. On the other Table 2). The dual task cost to stride time variability in hand, the TeamStudy and FPHS cohorts reported approxi- the highest quartile was 9.8 (95% confidence interval mately similar PROMIS physical and mental health scores. [CI] = 3.5–27.9) times higher than in the first quartile The range of symptomatic head trauma score was and 4.4 (95% CI = 2.0–9.4) times higher than in the sec- broad (4–75) across participants. This score was not ond quartile. A significant effect of quartile was also pre- related to participant age, number of NFL seasons played, sent for stride time variability specifically within the dual or the PROMIS physical health score (r2 < 0.02, task walking condition (unadjusted: F = 4.6, p = 0.03; p = 0.30–0.65). TeamStudy participants who reported adjusted: F = 4.2, p = 0.03). The highest 3 burden quar- greater symptomatic head trauma score reported worse tiles exhibited greater dual task stride time variability than PROMIS mental health score (r2 = 0.09, p = 0.01). the lowest burden quartile by 1.3 (95% CI = 1.1–1.6), 1.4 (95% CI = 1.2–1.7), and 1.8 (95% CI = 1.5–2.2) Effects of Self-Reported Head Trauma on Gait times, respectively. Symptomatic head trauma quartile was Performance not associated with the dual task cost to stride time, stride We examined relationships between gait outcomes and time within the dual task walking condition, or either gait self-reported burden of hits followed by concussion-related metric within the single task walking condition. symptoms by separating TeamStudy participants into Linear regression analysis similarly revealed that par- quartiles of symptomatic head trauma score (Table 2). An ticipants with higher symptomatic head trauma scores effect of quartile was observed for the dual task cost to exhibited greater dual task costs to stride time variability stride time variability (F = 6.9, p = 0.005). This effect was (Figs 1 and 2). This relationship remained significant independent of age, BMI, the number of NFL seasons, when models were adjusted for age, BMI, the number of and the number of reported musculoskeletal surgeries NFL seasons, and the number of reported musculoskeletal (F = 6.0, p = 0.001). Tukey post hoc testing of the surgeries (r2 = 0.20, p = 0.003). Relationships between adjusted model indicated that those in the highest burden higher symptomatic head trauma score and greater stride quartile exhibited greater dual task costs to stride time var- time variability specifically within the dual task walking iability than the 2 quartiles with lowest burden (see condition trended toward significance in both unadjusted TABLE 2. Relationship between Self-Reported Symptomatic Head Trauma Quartile and Remotely Assessed Walking Performance in Retired NFL Players Symptomatic Head Trauma Score (quartile thresholds) <11 11–19 20–28 >28 p Sample, n 16 16 16 16 Stride time, s Single task 1.1 (0.1) 1.2 (0.1) 1.2 (0.1) 1.2 (0.1) 0.89 Dual task 1.2 (0.1) 1.2 (0.1) 1.3 (0.1) 1.3 (0.1) 0.87 Stride variability, COV Single task 2.6 (0.8) 3.4 (1.2) 3.2 (1.1) 3.1 (1.7) 0.06 a b b b Dual task 2.7 (0.6) 3.6 (1.2) 3.8 (1.2) 4.9 (1.3) 0.03* Dual task cost, % Stride time 1.7 (4.6) 3.2 (4.2) 4.4 (6.2) 3.8 (5.1) 0.19 a a,b b,c c Stride variability 3.3 (6.8) 7.4 (10.9) 22.9 (13.4) 32.4 (17.6) 0.003* All values reflect group mean (standard deviation). Probability values reflect 1-way analyses of variance adjusted for age, body mass index, number of NFL seasons, and number of reported musculoskeletal surgeries. a,b,c Within each row, quartile means with different superscript letters were significantly different from one another based on Tukey post hoc testing of adjusted models. *Statistically significant. COV = coefficient of variation; NFL = National Football League. January 2020 79 ANNALS of Neurology FIGURE 1: Stride times derived from remote smartphone-based dual task gait assessment for 2 retired National Football League (NFL) players. Symptomatic head trauma score was defined as the number of self-reported hits to the head or neck followed by 1 of 11 queried symptoms related to concussion (see Subjects and Methods). Each data point represents a validated stride time derived from consecutive heel strikes of the phone-side leg during all nonturning periods of walking within each 45-second trial. The dual task cost to stride time variability, as defined by the change (ie, percentage increase) from single to dual task walking conditions, was noticeably higher in the player who reported higher symptomatic head trauma. (r2 = 0.09, p = 0.09) and adjusted (r2 = 0.06, p = 0.10) p = 0.007; adjusted: F = 4.5, p = 0.001; Table 3). Self- regression models. The symptomatic head trauma score reported loss of consciousness did not have a significant was not associated with the dual task cost to stride time or effect on any other gait outcome. stride time within the dual task walking condition. Nei- ther stride time nor its variability during the single task Relationships between Gait Performance and walking condition was related to self-reported symptom- Participant Demographics, ASF Exposure, and atic head trauma. Health and Wellness Self-report of experiencing hits to the head or neck Older participants tended to have longer stride times when specifically followed by loss of consciousness was also asso- walking in both single task (r2 = 0.30, p = 0.04) and dual task ciated with worse gait performance. Participants who (r2 = 0.32, p = 0.03) conditions. Those who reported at least reported experiencing at least 1 hit followed by loss of 1 surgery to the neck, back, hips, knees, or ankles also tended consciousness had 2.5 times (95% CI = 1.6–3.9) greater to have longer stride times in both walking conditions (single dual task cost to stride time variability than those who did task: r2 = 0.25, p = 0.02; dual task: r2 = 0.18, p = 0.05). Age not report loss of consciousness (unadjusted: F = 4.3, and history of musculoskeletal surgery did not significantly FIGURE 2: The relationship between self-reported symptomatic head trauma and the dual task cost to stride time variability. The symptomatic head trauma score was defined as the number of self-reported hits to the head or neck followed by 1 of 11 queried symptoms related to concussion (see Subjects and Methods). The dual task cost was calculated as the percentage change in the coefficient of variation (COV) in mean stride time from single to dual task walking conditions. 80 Volume 87, No. 1 Manor et al: Head Trauma and Dual Task Gait participant age, BMI, NFL career duration, and reported TABLE 3. Relationship between Self-Reported Loss history of musculoskeletal surgery, this preliminary evi- of Consciousness following a Hit to the Head or dence indicates that remote assessments of dual task walk- Neck and Remotely Assessed Walking Performance ing may be particularly sensitive to chronic brain in Retired NFL Players dysfunction associated with repetitive sports-related head Loss of Consciousness trauma. No Yes p Performance decrements associated with dual tasking are theorized to arise as the 2 tasks compete for shared Sample, n 32 32 brain resources.20 The observation that participants who reported greater total symptomatic head trauma, or at least Stride time, s 1 hit followed by loss of consciousness, also exhibited Single task 1.2 (0.1) 1.2 (0.1) 0.80 greater dual task costs suggests that these individuals pos- Dual task 1.2 (0.1) 1.2 (0.1) 0.68 sessed diminished capacity to differentially recruit the required brain networks or enlist the alternative cognitive Stride variability, COV strategies needed to maintain performance in both tasks.21 Single task 3.2 (1.1) 3.0 (1.1) 0.83 Whereas symptomatic head trauma was linked to worse Dual task 3.5 (1.1) 3.7 (1.2) 0.41 gait performance specifically during dual task walking, it was not correlated with gait performance during single Dual task cost, % task walking—a task that, as compared to dual-tasking, is Stride time 2.9 (5.7) 2.9 (5.4) 0.94 less cognitively demanding and associated with relatively Stride variability 8.6 (10.2) 21.4 (12.7) 0.007a lower levels of cortical brain activation.22 The results of this study therefore suggest the possibility that the accu- All values reflect group mean (standard deviation). Probability values reflect 1-way analyses of variance adjusted for age, body mass index, mulation of biomechanical head trauma severe enough to number of NFL seasons, and number of reported musculoskeletal induce concussionlike symptoms may manifest in later life surgeries. as diminished “cognitive reserve.”21 However, future stud- a Statistically significant. ies that assess and control for serial subtraction perfor- COV = coefficient of variation; NFL = National Football League. mance are needed. Such work should also consider capturing common proxies of cognitive reserve such as eduation,23 examine compensatory strategies that may correlate with stride time variability or dual task cost metrics. enable some individuals to maintain function in the pres- Gait metrics were not significantly correlated with participant ence of brain pathology, and consider the possible benefi- BMI, race, age at first exposure to organized football, number cial protective impact of cognitive remediation, physical of NFL seasons played, or primary position played within the exercise, and/or other interventions aimed at promoting NFL. Participants who reported worse PROMIS global phys- cognitive reserve into senescence.24 ical health walked with longer stride times in the single task Although walking entails repetitive cyclical move- condition (r2 = 0.19, p = 0.05), but not during dual task ments, no two gait cycles are exactly the same. In the cur- walking. Gait metrics were not sensitive to general pain sever- rent study, self-reported symptomatic head trauma did not ity or PROMIS global mental health. influence single or dual task average stride time, a metric closely related to walking speed. However, participants Discussion within the 3 highest quartiles of symptomatic head trauma This cross-sectional study utilized a validated smartphone score, as compared to the lowest quartile, walked with ele- app to remotely assess dual task walking performance in vated dual task stride time variability. Mounting evidence former professional ASF players participating in the FPHS indicates that dual task stride time variability may be a at Harvard University. Participants who reported relatively sensitive predictor of even subtle changes in cognitive high burden of hits followed by concussion-related symp- function.9,13,25 Beauchet et al,13 for example, reported in toms during their professional playing careers exhibited a sample of 56 nondemented older adults that those with significantly greater stride time variability when dual greater dual task stride time variability at baseline were tasking, as well as greater dual task cost to this metric of more likely to experience global cognitive decline as mea- gait performance. Those who reported suffering at least sured by the Mini-Mental State Examination over a 1 hit to the head or neck followed by loss of consciousness 5-year follow-up period. Recently, Lo et al19 provided also exhibited greater dual task costs to stride time vari- resting-state functional magnetic resonance imaging evi- ability. As each of these results were independent of dence from a small sample of older adults that greater January 2020 81 ANNALS of Neurology stride time variability correlated with weaker functional adjusted analyses focused on symptomatic head trauma connectivity between 2 large-scale brain networks; namely, score quartiles—together with corroborating results of the dorsal attention and default networks. Intriguingly, models focused on self-report of suffering at least 1 hit the functional connectivity between these 2 networks has followed by loss of consciousness—provide preliminary separately been linked to sustained attention defined as evidence that suffering relatively large amounts of symp- the capacity to maintain performance on a continuous tomatic head trauma during one’s professional playing cognitive task over time.26 The observed link between career is associated with worse gait performance in later symptomatic head trauma and stride time variability in life. We therefore believe that a remote smartphone-based the current study is thus supported by reports that retired assessment of dual task walking may provide important NFL players who report a history of concussion, as com- information for the evaluation of present and incipient pared to those who do not, tend to have difficulties in dysfunction in former professional ASF players and possi- attentional control,27 structural abnormalities within bly others exposed to high-risk repetitive head trauma. frontoparietal tracts and the corpus callosum,28 and reduced global resting-state functional connectivity.29 However, reported links between sports-related head Acknowledgment trauma, brain structure, brain function, and/or health out- This work was primarily supported by the FPHS at Harvard comes have been derived largely from cross-sectional stud- University, the Harvard Catalyst j Harvard Clinical and ies and/or comparisons to suboptimal control groups. Translational Science Center (NIH/National Center for Longitudinal studies that include multimodal mobile Research Resources and National Center for Advancing imaging of numerous aspects of dual task gait perfor- Translation Sciences, UL1 RR025758), and the NIH/NIA mance30 and/or habitual physical activity coupled with Boston Claude D. Pepper Older Americans Independence physiologic, biomarker, and neurocognitive testing are Center (OAIC) (P30 AG031679). B.M. and J.Z. were also thus needed to establish whether a history of sports-related supported by grants from the NIH/NIA (R01 head trauma alters the time course of changes in brain AG059089-01), the United States-Israel Binational Science health, and whether such changes underline reductions in Foundation, and the Marcus-Applebaum Foundation at gait and other cognitive–motor functions over time. Hebrew SeniorLife. A.P.-L. was also supported by grants Metrics of gait performance were obtained via from the NIH, National Science Foundation, Sidney smartphone assessment without supervision from trained R. Baer, Jr. Foundation, and Defense Advanced Research personnel in nonlaboratory environments that undoubt- Projects Agency (DARPA; HR001117S0030). The content is edly varied from participant to participant. Although it is solely the responsibility of the authors and does not necessar- thus impossible to confirm whether participants com- ily represent the official views of Harvard Catalyst, Harvard pleted the assessment correctly in the absence of environ- University and its affiliated academic health care centers, the ment perturbations, only those steps identified from National Institutes of Health, NeuroNEXT, the Sidney periods meeting specific validated criteria were included in R. Baer, Jr. Foundation, the FPHS, or DARPA. analysis. Moreover, we observed that participant age, his- tory of musculoskeletal surgery, and self-report of worse physical function were each correlated with longer average Author Contributions stride times under both single and dual task walking con- B.M., R.Z., and A.P.-L. contributed to study concept and ditions. The presence of these relationships, which were design. B.M., J.Z., O.-Y.L., H.Z., W.Y., N.A.G., and expected and corroborated by several previous reports in T.G.T. contributed to data acquisition and analysis. B.M., other populations,15,31 provides additional support that J.Z., O.-Y.L., H.Z., R.Z., L.A.L., T.G.T., and A.P.-L. gait assessments were completed successfully as directed. contributed to manuscript and figure preparation. Symptomatic head trauma scores were derived from self-report of events that occurred quite remote in time for Potential Conflicts of Interest many participants. The accuracy of such recall may also Nothing to report. have been influenced by multiple factors, including partic- ipant age, NFL career duration, and current health status. Sampling biases may also have been present; for example, older players who were willing and able to participate in References the study despite relatively high symptomatic head trauma 1. Montenigro PH, Alosco ML, Martin BM, et al. Cumulative head impact exposure predicts later-life depression, apathy, executive scores may not be representative of all players of similar dysfunction, and cognitive impairment in former high school and col- age and self-reported burden. Nevertheless, results of lege football players. J Neurotrauma 2017;34:328–340. 82 Volume 87, No. 1 Manor et al: Head Trauma and Dual Task Gait 2. Mez J, Daneshvar DH, Kiernan PT, et al. Clinicopathological evalua- 16. Manor B, Yu W, Zhu H, et al. Smartphone app-based assessment of tion of chronic traumatic encephalopathy in players of American gait during normal and dual-task walking: demonstration of validity football. JAMA 2017;318:360–370. and reliability. JMIR Mhealth Uhealth 2018;6:e36. 3. Montenigro PH, Baugh CM, Daneshvar DH, et al. Clinical subtypes 17. Cella D, Riley W, Stone A, et al. The Patient-Reported Outcomes of chronic traumatic encephalopathy: literature review and proposed Measurement Information System (PROMIS) developed and tested research diagnostic criteria for traumatic encephalopathy syndrome. its first wave of adult self-reported health outcome item banks: Alzheimers Res Ther 2014;6:68. 2005-2008. J Clin Epidemiol 2010;63:1179–1194. 4. Montero-Odasso M, Almeida QJ, Bherer L, et al. Consensus on 18. Strike SC, Taylor MJ. The temporal-spatial and ground reaction shared measures of mobility and cognition: from the Canadian Con- impulses of turning gait: is turning symmetrical? Gait Posture 2009; sortium on Neurodegeneration in Aging (CCNA). J Gerontol A Biol 29:597–602. Sci Med Sci 2019;74:897–909. 19. Lo OY, Halko MA, Zhou J, et al. Gait speed and gait variability are associated with different functional brain networks. Front Aging Neu- 5. Yogev-Seligmann G, Hausdorff JM, Giladi N. The role of executive rosci 2017;9:390. function and attention in gait. Mov Disord 2008;23:329–342; quiz 472. 20. Pashler H. Dual-task interference in simple tasks: data and theory. Psychol Bull 1994;116:220–244. 6. Ruffieux J, Keller M, Lauber B, Taube W. Changes in Standing and Walking Performance Under Dual-Task Conditions Across the 21. Tucker AM, Stern Y. Cognitive reserve in aging. Curr Alzheimer Res Lifespan. Sports Med 2015;45:1739–1758. 2011;8:354–360. 7. Parker TM, Osternig LR, Lee HJ, et al. The effect of divided attention 22. Mirelman A, Maidan I, Bernad-Elazari H, et al. Increased frontal brain on gait stability following concussion. Clin Biomech (Bristol, Avon) activation during walking while dual tasking: an fNIRS study in 2005;20:389–395. healthy young adults. J Neuroeng Rehabil 2014;11:85. 23. Grotz C, Seron X, Van Wissen M, Adam S. How should proxies of 8. Howell DR, Osternig LR, Chou LS. Detection of acute and long-term cognitive reserve be evaluated in a population of healthy older effects of concussion: dual-task gait balance control versus comput- adults? Int Psychogeriatr 2017;29:123–136. erized neurocognitive test. Arch Phys Med Rehabil 2018;99: 1318–1324. 24. Stern Y, Arenaza-Urquijo EM, Bartres-Faz D, et al. Whitepaper: Defining and investigating cognitive reserve, brain reserve, and 9. Lee J, Park S. Effects of a priority-based dual task on gait velocity brain maintenance. Alzheimer Dement 2018;pii:S1552-5260(18) and variability in older adults with mild cognitive impairment. J Exerc 33491-5. Rehabil 2018;14:993–997. 25. Hausdorff JM, Schweiger A, Herman T, et al. Dual-task decrements 10. Hamacher D, Hamacher D, Schega L. A cognitive dual task affects in gait: contributing factors among healthy older adults. J Gerontol gait variability in patients suffering from chronic low back pain. Exp A Biol Sci Med Sci 2008;63:1335–1343. Brain Res 2014;232:3509–3513. 26. Esterman M, Noonan SK, Rosenberg M, Degutis J. In the zone or 11. Herman T, Mirelman A, Giladi N, et al. Executive control deficits as a zoning out? Tracking behavioral and neural fluctuations during prodrome to falls in healthy older adults: a prospective study linking sustained attention. Cereb Cortex 2013;23:2712–2723. thinking, walking, and falling. J Gerontol A Biol Sci Med Sci 2010;65: 1086–1092. 27. Alosco ML, Koerte IK, Tripodis Y, et al. White matter signal abnor- malities in former National Football League players. Alzheimers 12. Montero-Odasso MM, Sarquis-Adamson Y, Speechley M, et al. Asso- Dement (Amst) 2018;10:56–65. ciation of dual-task gait with incident dementia in mild cognitive 28. Tremblay S, Henry LC, Bedetti C, et al. Diffuse white matter tract impairment: results from the Gait and Brain Study. JAMA Neurol abnormalities in clinically normal ageing retired athletes with a his- 2017;74:857–865. tory of sports-related concussions. Brain 2014;137(pt 11):2997–3011. 13. Beauchet O, Launay CP, Sekhon H, et al. Association of increased 29. Hampshire A, MacDonald A, Owen AM. Hypoconnectivity and gait variability while dual tasking and cognitive decline: results from hyperfrontality in retired American football players. Sci Rep 2013;3: a prospective longitudinal cohort pilot study. Geroscience 2017;39: 2972. 439–445. 30. Malcolm BR, Foxe JJ, Butler JS, De Sanctis P. The aging brain shows 14. Allali G, Montembeault M, Brambati SM, et al. Brain structure covari- less flexible reallocation of cognitive resources during dual-task walk- ance associated with gait control in aging. J Gerontol A Biol Sci Med ing: a mobile brain/body imaging (MoBI) study. Neuroimage 2015; Sci 2019;74:705–713. 117:230–242. 15. Jor’dan AJ, Manor B, Iloputaife I, et al. Diminished locomotor control 31. Slater LV, Hart JM, Kelly AR, Kuenze CM. Progressive changes in is associated with reduced neurovascular coupling in older adults. walking kinematics and kinetics after anterior cruciate ligament injury J Gerontol A Biol Sci Med Sci 2019; https://doi.org/10.1093/gerona/ and reconstruction: a review and meta-analysis. J Athl Train 2017;52: glz006. 847–860. January 2020 83