History of Sport-Related Concussion and Long-Term Clinical Cognitive Health Outcomes in Retired Athletes: A Systematic Review

This review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (PRISMA; www.prisma-statement.org) and was recorded in PROSPERO, a registry of systematic reviews. The flow diagram is presented in Figure 1. Registration is available at https://www.crd.york.ac.uk/prospero/ (registration number: CRD42016050750).

Studies were included if the authors evaluated retired male or female athletes who participated in organized sport at the amateur to professional level. At least 1 form of cognitive testing must have been used as an outcome measure. Studies were excluded if the investigators explored only athletes still actively involved in sport or did not report data on retired athletes as a subgroup or if they were case studies with 5 or fewer participants. Neurocognitive testing must have been conducted with and the results obtained from the participant (ie, rather than from friends or family). The primary outcomes of interest were a variety of cognitive domains: attention, memory, executive function, intelligence, processing speed, visuospatial abilities, and psychomotor speed. The secondary outcome variable of interest was a history of sport-related concussion (SRC).

We searched the electronic databases EMBASE, PsycINFO, MEDLINE/PubMed, CINAHL, Cochrane Central Register of Controlled Trials, and Web of Science from their inception to April 2018 using the relevant database search engines. Common key words and medical subject headings were related to 3 components: (1) the participant (eg, retired athlete), (2) the primary outcome measure (eg, cognitive test used), and (3) the secondary outcome (eg, history of sport concussion). No search restrictions for date or language were imposed. The search strategy for each database and corresponding number of hits per database are presented in Appendix 1.

The electronic database searching was supplemented by searching the abstracts of the “International Conference on Concussion in Sports” consensus meetings (2001–2018) and conducting a gray-literature search and a hand search of the reference lists of the included studies. Two reviewers (J.C. and F.W.) independently screened the titles and abstracts to identify studies that potentially met the eligibility criteria. Full texts of these reports were retrieved and independently assessed for eligibility by the same 2 reviewers. Any disagreements on study inclusion were resolved through discussion and consultation with a third reviewer (S.P.B.) to reach a consensus. A total of 46 studies were included in this review.

A data-extraction template was used as a checklist of items that should be included in reports of cross-sectional studies based on the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.¹²  One reviewer (J.C.) recorded the study's aim, participant characteristics, details of concussion history, outcome measures used, and relevant outcome data (group means and standard deviations; Table 1).

Two reviewers (J.C. and F.W.) independently evaluated the methodologic quality of the studies using an adapted Downs and Black checklist.⁵⁹  Disagreements between the reviewers were resolved through discussion to achieve consensus. If agreement could not be reached, a third reviewer (S.P.B.) arbitrated. The checklist was modified to a maximum of 17 applicable questions that addressed the following methodologic components: reporting, external validity, internal validity (bias and confounding), and power. Seventeen items were rated on a 2-point scale (yes = 1 and no/unable to determine = 0), and 1 item was rated on a 3-point scale (yes = 2, partial = 1, and no = 0). The maximum achievable score was 18, with higher scores indicating better methodologic quality. Results were categorized according to the adapted Downs and Black checklist⁵⁹  from Hartling et al⁶⁰  and Hignett⁶¹  and were interpreted as follows: strong quality (≥14) represented the top 75%, moderate quality (scored 9–13) represented 50% to 74%, limited quality (5–8) represented 25% to 49%, and poor quality (<5) represented less than 25%.

After duplicates were removed, this review yielded a total of 2842 records. We screened the titles and abstracts of these records and identified 119 studies that potentially met the inclusion criteria and, hence, were subject to full review. Of these, 46 cross-sectional studies published between 1984 and 2018 met the criteria and were included in a quantitative synthesis. Given the heterogeneity of the study design and outcome measures, a meta-analysis could not be conducted for all studies. Where possible, we performed a meta-analysis depending on the homogeneity of the included studies.

Detailed information on the quality appraisal of the 46 studies is presented in the Supplemental Table (available online at http://dx.doi.org/10.4085/1062-6050-297-18.S1). A total of 11 (24%) studies had a methodologic quality score of poor or limited. The overall mean methodologic quality score was 10.3 ± 2.9 (out of a maximum of 18), which is considered moderate. A breakdown of the quality appraisal can be seen in Figure 2.

The key study characteristics and findings of the 46 cross-sectional studies are summarized and presented in Table 1. The sample consisted of 13 975 participants: 7387 collision-sport athletes, 662 contact-sport athletes, 3346 noncontact-sport athletes, and 2580 participants classified as controls. Collision sports were boxing, football, ice hockey, and rugby. All contact-sport athletes were soccer players. Sports classified as noncontact were archery, athletics, badminton, ballroom dancing, canoeing, cricket, fencing, field hockey, gliding, golfing, horseback riding, paragliding, pelota, rock climbing, running, sailing, skiing, squash, swimming, table tennis, track and field, triathlon, and weightlifting. Only 3 (7%) of the 46 studies included female participants.^35,54,56  Participants varied in age, medical history, socioeconomic background, concussion exposure, number of concussions reported, and types of sports played. A large proportion of these studies (n = 20) included retired NFL players.^* In 7 studies, retired rugby players were investigated.^{13,15,17,22,31,33,54}  Participants in the remaining studies included boxers (n = 3)^29,57,58  and soccer athletes (n = 4),^26,32,38,56  whereas other authors examined a combination of former professional and amateur athletes, including university and high school football and hockey athletes (n = 12).^† Of the 46 studies, 33 included a control group,^‡ and 13 did not.^§

Results from a wide variety of cognitive tests were reported. The tests and their corresponding cognitive domains are presented in Table 2. Objective neuropsychological tests alone were used in 31 studies^‖ to assess different aspects of cognition. Six studies^{15,32,35,46,52,55}  used subjective (self-reported) cognitive tests alone. The remaining 9 studies^¶ used a combination of objective and subjective tests. Cognitive tests were categorized according to the predominant cognitive domain assessed. The 9 cognitive domains assessed were global cognitive ability, attention, memory, executive function, language, psychomotor function, intelligence, perception, and self-reported cognitive functioning (Figure 3). Memory was the most commonly studied cognitive health outcome (n = 31, 67%), followed by attention (n = 17, 37%) and global cognitive ability (n = 16, 35%). The studies by Kuhn et al¹⁹  and Tremblay et al⁴²  were excluded from the analysis, as the cognitive results were duplicates of those reported by Solomon et al²⁴  and Tremblay et al,⁴⁸  respectively. The cognitive results are shown in Table 3.

Global cognitive ability was assessed in 16 studies.^# Individual cognitive tests are summarized in Table 2. Eleven of the 16 (69%) studies^** showed no evidence for increased global cognitive difficulties in retired athletes compared with control participants or normative data, whereas the other 5 (31%) studies did.^{30,33,47,50,51}  Decq et al³³  found that retired rugby players had a higher rate of mild cognitive disorders on a modified version of the Telephone Interview for Cognitive Status than players from a variety of other sports. However, the cognitive results were not associated with reported concussions. Amen et al⁵¹  and Willeumier et al⁵⁰  assessed the same cohort of retired NFL players and reported decreases from normal values in cognitive functioning and proficiency. Amen et al³⁰  also studied a large number of the same participants and demonstrated relationships between position and body mass measured by waist-to-height ratio and cognitive function. Randolph et al⁴⁷  observed that retired NFL players scored worse than the healthy control sample on the Repeatable Battery for the Assessment of Neuropsychological Status (P = .002). The meta-analysis of 3 studies measuring global cognition using the Mini-Mental State Examination showed no difference between groups for this outcome (mean difference [MD] = 0.09; 95% confidence interval [CI] = −0.35, 0.54; P = .68; Figure 4A). Overall, most studies did not support decreased global cognitive ability or proficiency in retired athletes.

Tests of attention were administered in 17 (37%) of 46 investigations.^†† In 3 studies,^23,28,31  investigators found that retired athletes had decreased attention scores. Alosco et al²³  reported that retired NFL players performed worse than controls on the Trail Making Test (TMT; P = .005). Wright et al²⁸  determined that retired NFL players had attentional and processing-speed deficits on the TMT and the Symbol Digit Modalities Test compared with normative values. Hume et al³¹  noted that the elite rugby group performed worse on the Online Computerized Neurocognitive Assessment Software–Vital Signs test of complex attention than the noncontact-sport athletes (effect size = −0.67; 95% CI = −1.07, −0.26). Based on the variety of tests administered to investigate attention, the balance of the evidence did not support decreased functioning in retired athletes. The results from 6 studies that assessed attention using the TMT are illustrated in Figures 4B and 4C. Neither the TMT–A (MD = −0.52; 95% CI = −3.40, 2.36; P = .72) nor the TMT–B (MD = −1.41; 95% CI = −4.58, 1.76; P = .38) revealed between-groups differences when the findings were pooled.

Memory was assessed in 31 studies^‡‡ using various tests (Table 2). The authors of 14 (45%) studies^§§ reported normal functioning on memory tests. Seventeen (55%) studies^‖‖ showed decreases in memory functioning among retired athletes compared with control participants or normative data. In 2 (6%) investigations,^39,43  researchers determined that retired NFL players performed worse than control participants on the California Verbal Learning Test. McMillan et al²²  identified worse performance on the Rey Auditory Verbal Learning Test by retired rugby players compared with the control group. Wilde et al²⁹  found that word-list recall on the Verbal Selective Reminding Test was worse in boxers than in control participants, whereas Koerte et al³⁴  observed that retired NFL players had decreased performance on a list-learning task. Tremblay et al⁴⁸  described decreased scores by retired American football athletes on the Rey Auditory Verbal Learning Test and Taylor Complex Figure Test. Using the Spatial Working Memory and Paired Associates Learning subtests to assess memory and reported differences, Pearce et al¹³  noted that retired rugby players performed more poorly than the control group. Researchers in the remaining 4 studies^28,36,37,58  reported deficits in memory tests; however, no control groups were included. The same group of retired NFL players studied by Amen et al⁵¹  and Willeumier et al⁵⁰  had decreases in memory on the MicroCog memory subset. Similar results were shown by Amen et al,³⁰  who studied a large proportion of the same players. In addressing relational memory impairments among retired athletes, Ford et al⁴⁹  described the multiple-concussion group as worse at recognizing intact pairs as “old” (intact-pair hits) than the age-matched group (P < .05). Overall, the results were mixed. However, preliminary evidence of memory decline exists in retired athletes. Our meta-analysis of 3 studies using the California Verbal Learning Test outcome measure indicated that retired athletes performed worse than control participants (MD = −6.48; 95% CI = −10.07, −2.88; P < .001; Figure 4D). Our meta-analysis of 5 studies that used the Rey-Osterreith Complex Figure Test outcome measure also showed that the control group outperformed retired athletes (Rey-Osterreith Complex Figure Test Copy and Immediate Recall: MD = −4.85; 95% CI = −7.15, −2.54; P < .001; and Rey-Osterreith Complex Figure Test Delayed Recall: MD = −5.36; 95% CI = −7.79, −2.94; P < .001; Figures 4E and 4F, respectively).

Tests of executive function were used in 11 investigations.^¶¶ The most commonly used instrument was the Wisconsin Card Sorting Test.^{16,34,37,54,56}  Six (55%) studies^{13,16,28,31,37,56}  demonstrated decreased executive function in retired athletes, whereas the remaining 5 (45%) studies did not.^{23,34,41,44,54}  Three studies^16,37,56  showed decreased performance on the Wisconsin Card Sorting Test in retired athletes compared with control participants. Wright et al²⁸  found that retired NFL players displayed deficits in executive ability compared with normative data (Heaton system: 37.5%; Wechsler system: 20.0%). Pearce et al¹³  reported that retired rugby players performed worse than the control group on the Cambridge Neuropsychological Test Automated Battery–Intra-Extra Dimensional Set Shift subtest. Hume et al³¹  determined that the retired elite-rugby group performed worse on tests of executive function (effect size = −0.41; 95% CI = −0.80, −0.02) on the Online Computerized Neurocognitive Assessment Software–Vital Signs. Overall, the results were mixed. However, evidence of a decline in executive function in retired athletes exists.

Eight groups^{16,17,21,23,24,39,43,49}  reviewed language tests, including tests of naming, speech production, and verbal fluency. The Boston Naming Test was used in 4 (50%) studies.^21,39,43,49  The mean Boston Naming Test T score of the retired athletes was lower than that of the control group in 2 of the 4 studies,^21,43  and no differences were found in the other 2 studies.^39,49  Esopenko et al¹⁶  used the Phonemic Word List Generation (verbal phonemic fluency test) and found no differences between retired contact-sport athletes and control participants. The Controlled Oral Word Association Test was used in 5 investigations,^{17,23,24,43,49}  and no differences were identified between the retired athlete and control groups. Overall, our review showed mixed evidence of language deficits in retired athletes.

Psychomotor function was assessed by 6 groups,^{13,17,22,31,45,56}  with 4 sets of researchers^13,17,22,45  reporting decreases in retired athletes compared with control participants. Pearce et al¹³  used the Cambridge Neuropsychological Test Automated Battery–Visuomotor Reaction Time subtest and found that retired rugby players reacted more slowly to the stimulus than the control group (P < .01). McMillan et al²²  observed that retired rugby players had decreased fine motor coordination in the dominant hand on the Grooved Pegboard test, a measure of visual-motor coordination, whereas Gardner et al¹⁷  reported worse scores for the nondominant hand (P = .03). Pearce et al⁴⁵  assessed fine motor control and learned that reaction time (both reaction to stimulus and movement time on the O'Connor Finger Dexterity Test) was better in the healthy control group than in the retired elite and amateur Australian football players (P = .003). The other 2 studies^31,56  did not support decreases in psychomotor function in retired athletes compared with control individuals. Evidence of difficulties in psychomotor functioning exists in retired athletes.

Premorbid intelligence was assessed in 12 studies.^## A total of 9 studies^*** showed no difference in intellectual ability in retired athletes compared with control groups. In 2 investigations,^16,34  retired athletes performed worse than control participants on intellectual function tests. Stamm et al³⁷  demonstrated that former NFL players exposed to tackle football before age 12 years performed worse on the Wide Range Achievement Test 4 than former NFL players exposed at age 12 or later. Overall, we did not find evidence of decreased intellectual functioning in retired athletes.

Alosco et al²³  used the Brief Smell Identification Test and reported scores that were lower among former NFL players than among noncontact control athletes. Visuospatial perception was assessed by 2 sets of researchers^16,22  who used the Judgment of Line Orientation test. No group differences were detected between retired athletes and control individuals.

In 14 studies,^††† a variety of subjective self-reported cognitive functioning tests were used to compare retired athletes and control groups or normative data (Table 2). Eleven (79%) studies^‡‡‡ identified increased subjective reports of cognitive difficulties experienced by former athletes. Authors of the remaining 3 studies^15,17,20  described no increase in self-reported concerns among retired athletes. We found evidence of increased sport-related cognitive concerns among retired athletes.

Investigators in 3 studies^32,47,55  implemented self-reports rather than formal diagnoses to examine mild cognitive impairment (MCI) in retired athletes. Guskiewicz et al⁵⁵  determined that 35% of the retired NFL players self-reported cognitive difficulties, which were deemed consistent with MCI. Randolph et al⁴⁷  noted indications of possible cognitive impairment in retired NFL players: a subsample of the players was compared with a clinical sample of patients with MCI, revealing similar profiles of impairments. Conversely, Vann Jones et al³²  stated that the prevalence of possible MCI or dementia in former professional soccer players was not different from a large control sample. No association was demonstrated between low- and high-risk playing positions or length of playing career and a positive MCI screening result. Age was the only risk factor across both groups.

Meta-analyses were conducted where possible for outcome measures using different scales and tools to assess the cognitive domains (Figure 5). The data were pooled using the standardized MD (SMD) and a random-effects model according to the Cochrane guidelines.⁶²  Global cognitive ability was almost different between groups (SMD = −0.14; 95% CI = −0.29, 0.00; P = .05; Figure 5A). The attention domain demonstrated no between-groups difference (SMD = 0.05; 95% CI = −0.14, 0.23; P = .63; Figure 5B). Retired players performed worse than control participants on tests of memory (SMD = −0.43; 95% CI = −0.59, −0.27; P < .001; Figure 5C). Assessment of executive function also revealed that control individuals outperformed retired athletes (SMD = −0.53; 95% CI = −0.78, −0.28; P < .001; Figure 5D). Tests of language (SMD = −0.38; 95% CI = −0.74, −0.02; P = .04; Figure 5E) and psychomotor function (SMD = −0.30; 95% CI = −0.50, −0.10; P = .003; Figure 5F) also favored control groups. The intelligence domain showed no between-groups difference (SMD = 0.01; 95% CI = −0.24, 0.26; P = .91; Figure 5G). Pooling of the self-report test results revealed that retired athletes reported more cognitive difficulties than control participants (SMD = −0.43; 95% CI = −0.66, −0.19; P < .001; Figure 5H).

Our review focused on a qualitative synthesis of the included studies. The authors of 14 studies^§§§ used a subjective, self-reported cognitive functioning test in mixed sporting populations, with 11 groups^‖‖‖ reporting increased subjective cognitive difficulties experienced by former athletes compared with control individuals or normative data. A synthesis of studies demonstrated more evidence for cognitive deficits in the areas of memory, executive function, psychomotor function, and self-reported cognitive functioning.

The balance of evidence to date did not support an association between cognitive deficits in retirement and concussion history and exposure to head impacts. Most studies (n = 33) did not identify any association between concussion history or exposure to head impacts and cognitive deficits. However, investigators in 13 (28%) of 46 studies^¶¶¶ described a frequency-response relationship, with greater cognitive impairments (subjective or objective) in athletes with greater levels of exposure to head impacts or concussions. The cross-sectional nature of the studies included in this review and an insufficient number of longitudinal studies limited our ability to make causal inferences about the relationship between concussion and long-term cognitive outcomes.

We appraised the literature regarding SRC and cognitive health outcomes in retired athletes. Our review was unique because we focused on clinical cognitive outcomes in living retired athletes. Findings suggested that certain areas of cognition may be affected by an SRC history. However, bias appeared to exist toward assessments of certain areas of cognition, such as global cognitive ability, attention, and memory, with neglect of other aspects of cognition, including language, psychomotor function, and perception. Furthermore, epidemiologic studies were confounded by individual differences in susceptibility to age-related neurocognitive decline or dementia-related pathologic conditions. Therefore, lifetime exposure to concussion is more than likely one of a myriad of environmental and predetermined risk factors for diminished cognitive reserve and early expression of neurocognitive decline. The researchers did not control for premorbid intellectual function and factors related to cognitive reserve. An important factor was the substantial overlap between a normal age-related neurocognitive downturn and the manifestation of SRC in later life.⁶³ 

From this review, evidence of increased self-reported cognitive difficulties emerged. A substantial number of studies identified self-reported cognitive concerns among retired athletes (11 of 14 [79%] studies that tested for self-reported symptoms). However, conclusions drawn from self-reported data should be cautious, particularly in the absence of clear associations between self-reported symptoms and previous head-impact exposure or reported number of concussions and the potential recall bias that continues to be a limitation of all retrospective studies in this area. Furthermore, external factors, such as the media, may influence former players' reporting.^64,65  Among the investigations of objective and subjective cognition in retired athletes,^### a lack of clear agreement existed between the measures. Most authors (n = 6 studies) did not support^{16,17,20,22,25,54}  the subjective reports with respect to neuropsychological test results. The clinical importance of the findings should be carefully considered in the overall context of the individual's performance on neuropsychological and cognitive testing and symptom self-report. The emerging disparity between subjective and objective tests may indicate that more sensitive cognitive test measures are required to recognize changes, which may then result in self-reported difficulties and translate to anomalies on cognitive tests.

Our meta-analysis of the cognitive domains of memory, executive function, language, psychomotor function, and self-reported cognitive functioning revealed that retired athletes performed worse than control participants. However, the magnitudes of the effect sizes were small; therefore, whether the effect sizes were clinically meaningful is unclear. An additional methodologic shortcoming of the reviewed studies was a paucity of studies that used a prospective, longitudinal design and biased recruitment. The study designs also raised important concerns along with retrospective recall bias and failure to control for confounding variables: namely, failure to control for a history of non–sport-related concussion. A further factor was the lack of suitable control groups. Important aspects associated with professional team sports, such as high levels of physical fitness, high income, and potential drug use (eg, opioid analgesics), were not considered. The late effect upper limb orthopaedic injuries may have had on psychomotor tests was also overlooked. A total of 13 studies^**** did not include a control group, which greatly limited the conclusions that could be drawn because of a lack of context with respect to population normative data. Among the investigations that included a control group (n = 33), only a small number (n = 8) included an appropriate noncontact-athlete control cohort^{15,26,31,34,35,38,56,57} ; most researchers (n = 22 studies)^†††† did not accurately match control individuals with retired noncontact athletes. In addition, a large proportion of investigators (n = 19) did not provide a definition of concussion,^‡‡‡‡ which reduced the validity of player recall.⁶⁶  Others^67,68  noted both underreporting (eg, lack of understanding about concussion) and exaggeration of head impacts. The inability to accurately quantify participants' exposure to subconcussive blows was a further difficulty.

Just over 50% of the studies (n = 24)^§§§§ controlled for a history of neurologic or psychiatric conditions. A history of head-impact exposure or concussion is one of myriad factors that may lead to neurocognitive decline in retired athletes; therefore, inclusion and exclusion criteria must be robust, and these factors must be appropriately controlled in research. Individuals with premorbid psychiatric or other health problems or life stressors are more likely to experience postconcussion syndrome.^69,70  Similarly, only a minority of authors (n = 14 studies)^‖‖‖‖ reported past or present alcohol or drug use. Downs and Abwender⁵⁶  described screening for alcohol abuse but did not provide these findings in their “Results” section. Some investigators^17,22,33,35  noted higher alcohol consumption among retired athletes than among control individuals, which could have negatively influenced cognitive performance on certain cognitive tests. Substance abuse has been associated with sustained deficits in executive functioning, especially inhibition.⁷¹  Long-term, high-dose anabolic steroid exposure may cause cognitive deficits, notably in visuospatial memory.⁷² 

Authors of epidemiologic and intervention studies^73,74  have suggested that overall physical activity preserves or improves cognitive function during aging; therefore, the failure to control for current activity levels among retired players versus control groups in most research presented a potential bias. However, only a minority of studies^{26,35,38,42,48,52}  controlled for the modifiable risk factors, such as diet and physical activity, that accounted for physical activity engagement and exercise frequency. Similarly, only a small proportion of investigators^{22,23,28,30,50–52}  examined factors such as body mass index, weight-to-height ratio, or cardiovascular health. These factors may affect cognitive functioning: a meta-analysis⁷⁵  indicated that being categorized in the overweight or obese range in midlife was a risk factor for dementia later in life. A large proportion of the studies (n = 20) included retired NFL players. Given the propensity toward being overweight in this population,⁷⁶  the risk of cognitive impairment may be elevated.⁷⁷  Hinton et al⁵²  found that dietary fat intake was more associated with self-reported cognitive difficulties than was exposure to football alone among former collegiate football players. Wright et al²⁸  reported that body mass index was associated with cognitive-reserve outcomes in retired NFL players.

Aside from the proposed negative relationship between concussion and cognitive function, the potential causes of cognitive concerns in retired athletes are diverse. Factors including genetics,⁷⁸  diet and nutrition,⁷⁹  exercise,⁸⁰  obesity,⁸¹  chronic pain and life stress,⁸²  childhood adversity,⁸³  personality factors,⁸⁴  family history of neurologic conditions,⁸⁵  steroid use,⁸⁶  drug and alcohol use,⁸⁷  depression and anxiety,⁸⁸  general medical history (eg, hypertension, diabetes, heart disease),⁸⁹  and neurodegenerative diseases (eg, Alzheimer disease, Parkinson disease, and amyotrophic lateral sclerosis)^90–92  have been implicated in exacerbated cognitive decline with aging. Most authors did not control for these variables. Parental socioeconomic status, race, and medical history independently predicted baseline memory scores among collegiate athletes, whereas concussion history and years exposed to sport did not.⁹³  None of the investigators accounted for socioeconomic status during childhood, which could affect cognitive reserve later in life and may account for differences among athlete groups with a history of concussion or head-impact exposure versus control participants or normative data. The importance of premorbid information; intellectual level; and learning disabilities, such as attention-deficit/hyperactivity disorder, which is known to exist at a high level among athletes, should also be considered. Athletes with a history of multiple concussions and a premorbid learning disability are vulnerable to neurocognitive impairment.⁹⁴ 

Individual differences in baseline intelligence and education status have not been addressed in most studies. Stamm et al³⁷  reported that 14% of the group younger than 12 years and 0% of the group older than 12 years had a learning disability and displayed differences in premorbid intellectual functioning (ie, Wechsler Test of Adult Reading scores), with the younger group representing those who were exposed to tackle football before age 12 and the older group representing those who were not exposed until a later age. This casts doubt on whether the group differences reflected premorbid impairment as opposed to the effects of concussion. Many authors included in control groups participants who were exposed to considerable concussion risk, albeit a lower risk of head impacts. The control groups of Alosco et al²³  and Murelius and Haglund⁵⁷  included participants with a history of playing soccer, and Alosco et al²³  included 1 participant with a history of amateur wrestling; McMillan et al²²  included a control group in which 34% of participants had a history of concussion and rugby participation. The inclusion of current and retired athletes in some studies, without distinguishing between them, may have skewed the results,^46,54  as the results may have reflected the effects of recent concussions or current head-impact exposure.

The assessment of cognition should ideally capture all of its domains. The tests used varied greatly among studies, and only a small proportion of the authors used a comprehensive battery that explored all aspects of cognitive functioning. Most studies focused on specific domains, such as tests of attention, memory, and executive function. Furthermore, many of the assessments used were designed to detect gross cognitive impairments and may have failed to uncover subtle changes in cognitive function. Given the media interest in concussion and public perceptions, studies with negative findings are potentially less likely to be published. We did not assess publication bias in this review; it was not possible to perform a funnel plot due to the heterogeneity of the outcome measures used. Large-scale, prospective longitudinal studies with a high level of control of confounding factors are required to confirm the effects of aging with a history of concussion on cognitive functioning in retired athletes.

The self-selected convenience samples limited the conclusions that could be drawn. This concern was exacerbated in some investigations because the inclusion criteria were limited to retired NFL players with a minimum 6-month history of self-reported complaints of cognitive, behavioral, and mood symptoms^23,34,37  and players presenting to memory clinics with cognitive or behavioral symptoms.²⁷  Self-selected participants may not have represented the retired athlete population. Given the retrospective nature of the studies, the possible long-term sequelae of concussion were influenced by methodologic biases, making it difficult to draw conclusions. A large proportion of the studies (n = 43) were based on retired male athletes; only 3 groups of authors^35,54,56  recruited retired female athletes. Alumni of the NFL accounted for a large number of participants in many of the studies (n = 20). The career paths and levels of head-impact exposure make it impossible to infer results beyond this unique cohort. Aside from the level of head-impact exposure, a host of factors separate NFL players from the population at large, including income, education level, and various lifestyle aspects. Large gaps exist in our knowledge of the effect of concussion on female athletes (because of potential sex influences on concussion recovery^95,96 ) and on people who participate at other sporting levels.

A further limitation was that concussions were often not well documented in the past. Therefore, all researchers relied on the athletes' self-reported history of concussion, making this information subject to retrospective recall bias. Only Wright et al²⁸  corroborated patients' self-reported history with verifiable reports. Relying solely on players' self-reported history of concussion and retired athletes' responses to a survey-based questionnaire regarding subjective memory difficulties is potentially unreliable. This is compounded by the fact that SRCs that occurred in the past may have been overlooked by clinicians unless loss of consciousness occurred.⁹⁷  A history of non–sport-related concussion, which accounts for most concussions,^98,99  needs to be considered in future study designs.

A total of 46 studies evaluated 9 aspects of cognitive functioning. Relative to the control groups or normative data, 5 areas showed declines: memory, executive function, language, psychomotor function, and self-reported cognitive functioning. The 4 other areas—global cognitive ability, attention, intelligence, and perception—were, on balance, neutral. The preliminary evidence of a dose-response association between cognitive health outcomes and past concussion exposure warrants further examination to determine the complex interaction between previous head-impact exposure and factors that may influence cognitive health in the aging retired athlete. As detailed throughout this review, confounding variables, case ascertainment, recall bias on behalf of the participants, and publication bias in the SRC field at large may have inflated these findings.

We thank David Mockler (medical librarian), Trinity College Dublin.

Supplemental Table. Detailed quality appraisal for the 46 studies.

Found at DOI: http://dx.doi.org/10.4085/1062-6050-297-18.S1