Article Text
PDF
Abstract
Objective This study investigated the longitudinal course of depressive symptom severity over 19 years in former American football players and the influence of concussion history, contact sport participation and physical function on observed trajectories.
Methods Former American football players completed a general health questionnaire involving demographic information, medical/psychiatric history, concussion/football history and validated measures of depression and physical function at three time points (2001, 2010 and 2019). Parallel process latent growth curve modelling tested associations between concussion history, years of football participation, and overall and change in physical function on the overall level and trajectory of depressive symptoms.
Results Among the 333 participants (mean(SD) age, 48.95 (9.37) at enrolment), there was a statistically significant, but small increase in depressive symptom severity from 2001 (48.34 (7.75)) to 2019 (49.77 (9.52)), slope=0.079 (SE=0.11), p=0.007. Those with greater concussion history endorsed greater overall depressive symptom severity, B=1.38 (SE=0.33), p<0.001. Concussion history, B<0.001 (SE=0.02), p=0.997 and years of participation, B<0.001 (SE=0.01), p=0.980, were not associated with rate of change (slope factor) over 19 years. Greater decline in physical function, B=−0.71 (SE=0.16), p<0.001, was predictive of a faster growth rate (ie, steeper increase) of depression symptom endorsement over time.
Conclusions Concussion history, not years of participation, was associated with greater depressive symptom severity. Neither factor was predictive of changes over a 19-year period. Decline in physical function was a significant predictor of a steeper trajectory of increased depressive symptoms, independent of concussion effects. This represents one viable target for preventative intervention to mitigate long-term neuropsychiatric difficulties associated with concussion across subsequent decades of life.
- concussion
- traumatic brain injury
- depression
Data availability statement
Data are available on reasonable request. The data used as part of this study that support the reported findings are available from the corresponding author (BLB), on reasonable request.
Statistics from Altmetric.com
Introduction
More lifetime concussions and repetitive head impacts (RHI) accumulated during contact sport participation have been associated with an elevated risk of being diagnosed with depression and/or endorsing elevated depressive symptom severity in select studies of former contact sport athletes.1–4 However, not all studies have observed this association.5–7 The largest limitation to better understanding this association and potentially resolving variability across studies is that most investigations have primarily involved cross-sectional studies at distinct points of the lifespan. This prohibits examination of the course of symptom endorsement and incorporation of factors that can influence longitudinal change in depression, such as physical comorbidities.8
One large-scale longitudinal study of older adults who played high school football in the 1950s failed to observe an association between likelihood of depression and participation in high school football.7 Those who participated in high school football did not endorse greater depressive symptoms across multiple time points (ages 54, 65, and 72) compared with well-matched groups. While this sole longitudinal study advanced our understanding of long-term outcomes associated with contact sport exposure, findings were limited by the fact that concussion history was not collected as part of the study and era of play differed from other published studies on the topic. Additionally, years of contact sport exposure (proxy measure for RHI) was not examined.
Longitudinal studies are essential for estimating risk of long-term outcomes associated with cumulative concussion/RHI, as prior work has shown that self-reported concussion history can be less reliable among individuals experiencing greater levels of difficulty later in life (ie, retroactive recall of concussions increased among those experiencing greater physical and mental health (MH) difficulties).9 Without true longitudinal studies, confounding factors, such as declines in physical function (PF) over the same time period, cannot be accurately measured or considered. Incorporation of PF as a potential influencing factor on depression is essential, given that poor PF predicts longitudinal depression in community dwelling adults,8 as well as the fact that prior studies have observed strong associations between physical and MH functioning among former contact sport athletes.10–12
The primary aims of the current study were to examine: (1) changes in depressive symptom severity over a 19-year period using prospectively collected data; (2) the relationship between concussion history and years of participation (proxy for RHI history) with overall depressive symptom severity, and the trajectory of these associations over the 19-year period and (3) the influence of PF on depression over time. We hypothesise that (1) depressive symptom severity will increase over the 19-year period, (2) concussion history, but not years of participation will be associated with overall and increases in depressive symptoms and (3) declines in PF will be associated with increased depression symptoms over time.
Materials and methods
Participants
This longitudinal study spanning a 19-year period collected data at three time points (2001, 2010 and 2019). A general health survey (GHS; see below) including a depression symptom inventory, was sent to all living members of the NFL Retired Players Association (n=3,729), with 2,536 (68.7%) completing the survey. A second follow-up GHS was sent to 2102 former NFL players who had completed the 2001 GHS, had accurate contact information, and were not deceased. A total of 1291 (61.4%) completed the 2010 survey. In 2019, a third GHS was sent out to 15,025 former NFL players via hardcopy and electronic mail (based on availability of contact info) as part of the ongoing Neurologic Function across the Lifespan: A Prospective, LONGitudinal, and Translational Study for Former National Football League Players (NFL-LONG).13 Overall, 1,784 (11.9%) completed and returned the 2019 GHS.
Inclusion criteria for the study required at least 2 years of professional American football in the NFL and completion of measures at a minimum of two of the three assessment time points listed above. At time point 1, all longitudinal participants reported no history of Parkinson’s disease, Alzheimer’s disease, amyotrophic lateral sclerosis or schizophrenia. Rates of disease onset across the study period are provided in table 1.
Participant characteristics and mean (SD) of outcome variables of interest
Participants remotely completed a GHS at each time point that included validated instruments and underwent a rigorous validation process (pilot versions were tested and reviewed for face content validity). Participants provided demographic information, medical/psychiatric history, concussion history, American football participation history and completed measures of physical and MH functioning. A standard definition of concussion was presented to participants, which included: ‘a blow to the head that is followed by a variety of symptoms that may include any of the following: headache, dizziness, loss of balance, blurred vision, ‘seeing stars’, feeling in a fog or slowed down, memory problems, poor concentration, nausea or throwing up. Getting ‘knocked out’ or being unconscious does not always occur with a concussion.’14
Measures of depression and PF
Depressive symptom severity and PF were measured using the 36-Item Short Form Health Survey (SF-36) in 2001 and 2010 and the respective Patient-Reported Outcome Measurement Information System (PROMIS) Depression and PF Short-Forms V.2.0 in 2019.15 16 The use of PROMIS measures in 2019 was based on the fact that PROMIS measures allow for greater comparison across future studies (open access, validated across populations, developed in concert with measures assessing various other constructs). The SF-36 assesses overall health status and a number of functional domains, such as physical functioning, pain and MH. The PROMIS Depression and PF Short-Forms V.2.0 are comprehensive inventories rated on 5-point Likert scales that have been validated in older adults with multiple chronic conditions.
The PROsetta Stone tool was used for crosswalk conversion of scores on the MH and PF Indices of the SF-36 to standardised T-scores on PROMIS Depression and PF measures for comparability across time points.15 16 The rigorous methods employed to ensure valid conversion of instruments are described elsewhere.15 16 Briefly, measures were harmonised using an item response theory fixed parameter calibration linking method. This approach focuses on calibrating non-fixed items from one scale (SF-36 MH and PF Indices) onto a second measure with fixed parameters (PROMIS Depression and PF), in order to place both measures on a comparable metric. This harmonisation allowed for participant responses (ie, raw scores) for all measures to be converted to standardised T-scores (mean=50, SD=10) based on normative data obtained from the general population. Use of T-scores based on the referent population (US census matched sample) allows for examination of where the current sample scores fall along the normal distribution of that sample.
Statistical analysis
Latent growth curve modelling (LGCM) using full information maximum likelihood estimation examined the trajectory of depressive symptom severity at each study time point. The LGCM approach contains several strengths, as it allows for testing of the trajectory (ie, slope) of depression symptom severity from the ‘starting’ or overall measurement (ie, intercept) for individuals. LGCM allows for estimation of inter- and intra-individual variability by incorporating subject-specific variance within the models, resulting in trajectories of change that are more precisely estimated compared with other approaches (eg, repeated measures analysis of variance), which only examines mean change over time.
A two-factor unconditional linear LGCM, including the means and variances of intercept and slope (rate of change) factors were estimated. With LGCM, factors such as concussion history and years of participation can be included within the model to predict variation in overall levels and trajectories of depressive symptom severity when the variance of the intercept and slope factors are significantly greater than zero. Parallel process LGCM, which involves similar parameters and procedures, was performed to examine the parallel growth processes of depressive symptoms and PF (intercept and slope) over time points 1–3. Specifically, we examined whether overall levels and change (slope factor) in PF predicted the slope of change in depressive symptoms. Good model fit was determined based on conventional cut-offs for the following fit indices: root mean square error of approximation (RMSEA) <0.06 (acceptable if <0.08), standardised root mean square residual (SRMR) <0.08 and the Comparative Fit Index (CFI) and Tucker-Lewis Index (TLI) ≥0.950.17
Concussion history in high school, collegiate and professional football participation was provided at time point 1 and binned across five categories: 0, 1–2, 3–5, 6–9, 10 or more (10+). The analytical approach involving binning of reported concussions was selected to allow for testing a general linear trend, but also mitigating the potential for lower reliability and limiting error in recall bias. Total years of football participation were also reported at time point 1. Statistical analyses were performed in Mplus (V.8). Statistical significance was evaluated at the 0.05 level.
Results
Participants
Of the 351 eligible former players, 11 did not complete depression measures for at least two time points (returned GHS but did not complete depression measures) and 7 did not prospectively provide concussion history at time point 1, resulting in a final sample of 333. At time point 1, former players reported a mean age of 48.95±9.37 years old (table 1). Concussion history was as follows; 0 (21.6%), 1–2 (23.1%), 3–5 (30.9%), 6–9 (12.6%), 10+ (11.7%). Mean years of football participation was 15.90±4.11. The mean number of years since retirement for the sample at 2001 was 19.59±9.99. The percentage of the sample reporting a physician provided diagnosis of depression increased from time points 1–3 as 9.2%, 11.6% and 20.4%, respectively.
Of the 333 participants, completed visits were as follows; n=333 at time point 1, n=255 at time point 2, and n=333 at Time point 3. Given that time point 1 and time point 3 were comparable, differences in samples between time point 1 and time point 2 were assessed. Samples at time point 1 and time point 2 did not statistically differ in total years of football participation (t=0.28, p=0.778; mean difference=0.15), education (χ2=0.38, p=0.827), concussion history (χ2=5.70, p=0.22) or the presence of depression diagnosis at time point 1 (χ2=0.10, p=0.750).
Those missing visits at 2010 were significantly more likely to identify as non-white, (χ2=6.54, p=0.011) and be younger in age (t=3.53, p=0.001; missing=45.72; present=49.95). Of note, age (rs=−0.08 and −0.09, ps >0.05) and race (ts=1.24 and 0.23, ps >0.05) were not significantly associated with PROMIS Depression T-scores at time point 1 or Time point 3, respectively. Importantly, those who did not complete time point 2 did not report significantly different depressive symptom severity at time point 1 (t=1.72, p=0.086; missing=49.67; present=47.92) or at time point 3 (t=1.97, p=0.051; missing=51.60; present=49.19). Additionally, those missing at time point 2 had parallel mean T-score increases from time point 1 to Time point 3 (missing=1.93; present=1.27).
Unconditional LGCM
The two-factor linear LGCM of depressive symptoms produced adequate fit, χ 2(3)=6.96, p=0.073, RMSEA=0.063, CFI=0.977, TLI=0.977, SRMR=0.063. There was statistically significant growth increase in depressive symptom severity over the study period for the overall sample (slope growth factor mean=0.079, p=0.007). Not all former players exhibited uniform trajectories of depressive symptom severity over time and there was statistically significant interindividual variability within the sample, as indicated by the slope factor variance (σ=0.087, p=0.002; figure 1). Participants’ rate of change in depressive symptoms over time was not related their overall depressive symptom endorsement (ie, trajectory was not influenced by whether they endorsed higher or lower depressive symptoms; intercept and slope correlation, r=−0.048, p=0.772).
Group mean trajectory of depressive symptom severity across time points 1–3 (red line) and spaghetti plot of interindividual deviations from the group level mean. PROMIS, patient-reported outcome measurement information system.
Concussion history and years of participation
Those with a greater history of concussion were more likely to report greater overall depressive symptom severity (B=1.383, p<0.001; table 2; figure 2). Concussion history was not significantly associated with trajectory of depressive symptom severity change over time (B<0.001, p=0.997; figure 3). Years of participation was not associated with either overall depressive symptom severity endorsement (B=−0.09, p=0.385) or trajectory of depressive symptom severity change (B<0.001, p=0.980).
Conditional growth model of depressive symptom severity
Trajectory of depressive symptom severity scores across the three study time points (2001, 2010 and 2019) for each of the concussion history groups. PROMIS, patient-reported outcome measurement information system.
Latent growth model of depressive symptom severity across time points 1–3, with concussion history and years of football participation predicting the intercept and slope factors. Greater history of concussion was significantly associated with overall depressive symptom severity (B=1.38, SE=0.33, p<0.001). Concussion history was not significantly associated with trajectory (slope) of depressive symptom severity change over time (B<0.001, SE=0.02, p=0.997). Years of participation was not associated with either overall depressive symptom severity endorsement (B=−0.088, SE=0.10, p=0.385) or trajectory of depressive symptom severity change over time (B<0.001, SE=0.01, p=0.980). Correlation between intercept and slope factor, (r=−0.05, p=0.772) within the model not displayed; *p<0.001.
Depression and PF parallel process growth model
Within the parallel process growth model, overall PF and depression symptom severity were correlated (r=−0.47, p<0.001). Rate of PF decline (B=−0.71, p<0.001), but not overall PF (B=−0.01, p=0.061), was predictive of a greater growth rate or steeper trajectory of depression symptom endorsement increases over time (figure 4). Better overall PF was associated with less decline in PF decline over the study period (intercept and slope factor correlation, r=−0.28, p=0.011). Greater concussion history was associated with higher overall depressive symptom severity (B=1.33, p<0.001) and worse overall PF (B=−0.85, p=0.003).
Parallel process latent growth model of depression symptom severity and physical function over 19-year period (time point 1–3), which includes the physical function intercept and slope factors as predictors of the depression slope factor, as well as concussion history as a predictor of depression and physical function intercept factors. Phys. Func.=PROMIS Physical Function Short From V.2.0. Not displayed within the model: correlation between physical function intercept and slope factor, r=−0.28, p=0.011. Within the model, overall physical function and depression symptoms severity were significantly correlated (r=−0.47, p<0.001). Rate of physical function decline was predictive of a greater growth rate or steeper trajectory of depression symptom endorsement increases over time, (B=−0.71, SE=0.16, p<0.001). Overall physical function was not significantly associated with trajectory of depression symptom severity (−0.01, SE=0.01, p=0.061). Greater concussion history was associated with higher overall depressive symptom severity (B=1.33, SE=0.28, p<0.001) and worse overall physical function (B=−0.85, SE=0.28, p=0.003); *p<0.001. PROMIS, patient-reported outcome measurement information system.
Discussion
Results from this prospective, longitudinal study indicated that collectively, small increases in depressive symptom severity were observed over a 19-year period in former NFL players that were consistent with those found in the general population (ie, the mean and distribution of standardised scores were similar to normative data derived from the general population).18 Concussion history was significantly associated with overall greater depressive symptom severity, but not predictive of increased depressive symptoms over the study period. Greater PF decline was significantly predictive of a steeper increase in depressive symptoms over the study period. Cumulative years of football participation were not significantly predictive of overall depressive symptom severity or changes in severity over time. These findings suggest that associations between depressive symptom severity and concussion history do not inherently progressively worsen over time, and that other factors, such as PF, may play an important role in neurobehavioural changes (ie, depression) that occur later in the life of former professional contact sport athletes.
Concussion history, RHI and depression symptoms
The proportion of athletes endorsing depressive symptoms 1 SD or above the normative mean at time point 1 (6.7%) was lower than would be expected within the general population. Biopsychosocial factors associated with elevated depression in the general population (eg, chronic disease, older age (>75 years old), lower education attainment) were not well represented in the current sample at time point 1 and may account for the lower than expected symptom endorsement.19 20 Lower depression severity at time point 1 may also be due to potential benefits of participation in sport. This would be consistent with a prior meta-analysis of retired professional and collegiate athletes from diverse sports that showed higher levels of positive MH functioning within athletes compared with general population normative data.21 Conversely, similar to other cross-sectional studies of samples across the lifespan, a positive association between concussion history and overall depression symptom severity was observed in the current study.4 21–23
A number of clinical and retrospective postmortem autopsy studies have observed an association between RHI via contact sport exposure and behavioural/mood symptoms.2 24 25 Conversely, a number of studies have failed observe an association between contact sport participation and worse MH or depression-related outcomes.5–7 13 It has been suggested that behavioural and mood symptoms associated with contact sport exposure represent the clinical correlate of chronic traumatic encephalopathy (CTE), traumatic encephalopathy syndrome (TES). The direct relationship between behavioural/mood related symptoms and CTE-neuropathological change is not clear at this point. A recent autopsy study using the 2014 TES criteria failed to observe a significant relationship between behavioural/mood symptoms and CTE pathology.26 The more recently published National Institute of Neurological Disorders and Stroke Consensus Diagnostic Criteria for TES has attempted to address this issue by designating psychiatric-related symptoms as supportive, but not core components of the diagnostic criteria.27
Longitudinal trajectory of depression symptoms
Within the current study, there was a small (ie, mean T-score increase of 1.43), but significant increase in depressive symptoms over the 19-year study period. Additionally, the number of individuals who endorsed depressive symptoms at or above 1 SD above the population based normative mean (T-Score ≥60) increased from 6.7% to 16% of the sample over the 19-year period. The rate and degree to which individuals increased in depressive symptom severity over a 19-year period was not significantly associated with concussion history. Additionally, the current study did not observe a significant association between years of participation, therefore exposure to RHI and depressive symptom severity or trajectory. The mean age of the current sample at time point 1 (48.95 years) is particularly germane to this consideration, given that it is approximately at this point in the lifespan where individuals with concussive or RHI history have been thought to develop progressive worsening of psychiatric difficulties based on previous studies.24 27 28
The importance of longitudinal studies in accurately estimating the association between prior concussion and depressive symptom severity, as well as other neurobehavioural outcomes, is underscored by the results of the current study. Simply querying about concussion history and depressive symptoms concurrently at any of the three time points 9 years apart (as opposed to longitudinally) would have yielded a statistically significant association between the two variables. Further, had the current study not collected concussion history at time point 1 and only at time point 3, we might expect parallel increases in the number of concussions reported along with elevations in depressive symptoms over time. This potential expectation is based on a prior study showing that individuals tend to increase in the number of retrospective concussions recalled as they experience greater physical and MH difficulties at follow-up.9
PF decline predicts depressive symptom change
Decline in PF predicted the slope (ie, rate of increase) in depressive symptoms over the 19-year period within this study. When accounting for the effects of concussion history on overall depression and PF, only decline in PF, but not overall PF was significantly associated with the rate of increase in depression symptoms. The dynamic relationship between concussion history, physical symptoms, and depressive symptom severity in former NFL players has been previously reported cross-sectionally.10 29 Further, lower PF and activity as a predictor of elevated depressive symptoms at longitudinal follow-up in the general population has been well documented.30–32 Improvement of PF represents a notable target area for intervention as a potential means to mitigate risk of depressive symptom elevations across the subsequent decades of life within this population.33
Growing evidence from longitudinal and cross-sectional studies suggests that a host of psychosocial factors, such as living alone/social connectedness, subjective social status and perceived isolation, are predictive of longitudinal depression among adults in mid-life to later-life.34–36 Within a cohort 686 former NFL players between the ages of 50–70, level of depressive symptom severity endorsement was significantly related to psychosocial factors such as participation in social roles and activities, emotional support, and meaning and purpose. As part of a neurobiopsychosocial approach, future studies should look to include these and related factors when examining longitudinal trajectories of depression in former contact sport athletes.
Strengths
The current study has a number of strengths. Primarily, this is the first study to longitudinally assess the association between concussion history and contact sport participation with depression symptom trajectory over approximately two decades. Longitudinally measuring PF also allowed us to evaluate the influence of this critical factor in the trajectory of depressive symptoms. Third, the advanced analytical method allowed for more precise estimations of change in depression symptoms through consideration of interindividual and intraindividual variability in trajectories of change over time.
Limitations
Although concussion history was queried in a prospective nature in regard to the 19-year longitudinal period of investigation, the current study is limited by the fact that concussion history was self-reported postretirement as opposed to recorded at the time of injury. Concussions sustained in sport prior to high school were not collected at baseline. However, there is still a notable range in the number of concussions reported by players and there are currently no studies suggesting that concussions sustained during pre high school participation are fundamentally different than higher levels of play or have a differential effects on long-term outcomes. Additionally, players included within the study were recruited via the National Football League Player’s Association-Retired Section, and in theory, recruitment could have missed a select ‘disconnected’ subgroup of former NFL players not involved in anyway with the organisation. Although years of participation was not significantly associated with depression endorsement or trajectory, more comprehensive indices of cumulative exposure, which were not available for use in the current study, have shown differential associations with neurobehavioural outcomes.38 Finally, with a 19-year study period, there is increased risk for confounding factors that may have not been accounted for in the current study, such as genetic influences/mutations, changes in substance use or lifestyle changes.
Conclusion
Higher depressive symptom severity endorsement was associated greater history of concussion, consistent with prior studies. Depressive symptom severity increases, commensurate with the normative data from the general population, do not appear to be inherently progressive over a 19-year period in this sample of former NFL players. Decline in PF, not concussion history or years football participation, played an important role in predicting worsening of depressive symptoms over time. This represents an important and viable area of prophylactic intervention in order to mitigate the potential for elevated depressive symptoms later in life in former NFL players.
Data availability statement
Data are available on reasonable request. The data used as part of this study that support the reported findings are available from the corresponding author (BLB), on reasonable request.
Ethics statements
Patient consent for publication
Ethics approval
This study was approved by an Institutional Review Board (IRB) at the Medical College of Wisconsin and University of North Carolina at Chapel Hill #19-1065 and participants provided written informed consent prior to any study activities.
Acknowledgments
We are grateful for the participation of the athletes, without whom this research would not be possible. The authors would like to thank Hope Campbell (Department of Neurosurgery at the Medical College of Wisconsin), Candice Goerger (Centre for the Study of Retired Athletes at the University of North Carolina at Chapel Hill), Caprice Hunt (Centre for the Study of Retired Athletes at the University of North Carolina at Chapel Hill), and Gregory Kobelski (Department of Orthopaedics at Boston Children’s Hospital) for study coordination and management.
References
Footnotes
Twitter @BenjaminBrett1
Contributors BLB: designed and conceptualised the study; major role in acquisition of the data; analysed data; drafted and revised the manuscript for intellectual content. As the guarantor, he accepts full responsibility for the finished work and/or the conduct of the study, had access to the data, and controlled the decision to publish. SW: major role in acquisition of the data; analysed data; revised the manuscript for intellectual content. ZK: major role in acquisition of the data; analysed data; revised the manuscript for intellectual content. RM: designed and conceptualised study; analysed data; drafted the manuscript for intellectual content. AC: revised the manuscript for intellectual content. JDD: revised the manuscript for intellectual content. RE: major role in acquisition of the data; revised the manuscript for intellectual content. KMG: major role in acquisition of the data; analysed data; revised the manuscript for intellectual content. WM: major role in acquisition of the data; analysed data; revised the manuscript for intellectual content. MM: designed and conceptualised study; major role in acquisition of the data; analysed data; revised the manuscript for intellectual content.
Funding This study (NFL-LONG) was funded by the National Football League (NFL) with support from the National Institute on Aging (BLB-Award Number K23 AG073528).
Competing interests BLB acknowledges support from the National Institute of Neurological Disorders and Stroke (Award Number L30NS113158-02) and National Institute on Aging (Award Number K23 AG073528).
ZYK reports grants from National Institutes of Health; grants from Centers for Disease Control and Prevention; and grants from National Football League. AC discloses funding from the National Collegiate Athletic Association. RM reports grants from U.S. Department of Defense; grants from NFL Foundation; and grants from National Institute of Neurological Disorders and Stroke. RE is a paid consultant for the NHL and co-chair of the NHL/NHLPA Concussion Subcommittee. He is also a paid consultant for Major League Soccer and Princeton University Athletic Medicine and occasionally provides expert testimony in matters related to MTBI and sports concussion. WM receives royalties from (1) ABC-Clio publishing for the sale of his books, Kids, Sports and Concussion: A guide for coaches and parents, and Concussions; (2) Springer International for the book Head and Neck Injuries in Young Athlete and (3) Wolters Kluwer for working as an author for UpToDate. His research is funded, in part, by philanthropic support from the National Hockey League Alumni Association through the Corey C. Griffin Pro-Am Tournament and a grant from the National Football League.Dr Guskiewicz reports compensation from National Collegiate Athletic Association for other services and grants from Boston Children's Hospital (sub-award from the National Football League). MM acknowledges researching funding from the NIH, US Dept. of Defense, CDC, NCAA and NFL to the Medical College of Wisconsin.
Provenance and peer review Not commissioned; externally peer reviewed.