Michael W. Collins, Scott H. Grindel, Mark R. Lovell, Duane E. Dede, David J. Moser, Benjamin R. Phalin, Sally Nogle, Michael Wasik, David Cordry, Michelle Klotz Daugherty, Samuel F. Sears, Guy Nicolette, Peter Indelicato, Douglas B. McKeag. Relationship Between Concussion and Neuropsychological Performance in College Football Players. JAMA. 1999;282(10):964–970. doi:10.1001/jama.282.10.964
Author Affiliations: Department of Behavioral Health, Henry Ford Health System, Detroit, Mich (Drs Collins and Lovell); Departments of Family Practice (Dr Grindel) and Family Medicine/Sports Medicine (Dr McKeag), University of Pittsburgh, Pittsburgh, Pa; Departments of Clinical and Health Psychology (Drs Dede, Moser, and Sears and Mr Phalin), Community and Family Health (Dr Nicolette), and Sports Health (Mr Wasik), and Department of Orthopedics and Rehabilitation, College of Medicine (Dr Indelicato), University of Florida, Gainesville; Departments of Athletic Training (Ms Nogle) and Psychology (Mr Cordry and Ms Daugherty), Michigan State University, East Lansing; and Department of Sports Medicine, St Vincent's Medical Center, Erie, Pa (Dr Grindel).
Context Despite the high prevalence and potentially serious outcomes associated
with concussion in athletes, there is little systematic research examining
risk factors and short- and long-term outcomes.
Objectives To assess the relationship between concussion history and learning disability
(LD) and the association of these variables with neuropsychological performance
and to evaluate postconcussion recovery in a sample of college football players.
Design, Setting, and Participants A total of 393 athletes from 4 university football programs across the
United States received preseason baseline evaluations between May 1997 and
February 1999. Subjects who had subsequent football-related acute concussions
(n=16) underwent neuropsychological comparison with matched control athletes
from within the sample (n=10).
Main Outcome Measures Clinical interview, 8 neuropsychological measures, and concussion symptom
scale ratings at baseline and after concussion.
Results Of the 393 players, 129 (34%) had experienced 1 previous concussion
and 79 (20%) had experienced 2 or more concussions. Multivariate analysis
of variance yielded significant main effects for both LD (P<.001) and concussion history (P=.009),
resulting in lowered baseline neuropsychological performance. A significant
interaction was found between LD and history of multiple concussions and LD
on 2 neuropsychological measures (Trail-Making Test, Form B [P=.007] and Symbol Digit Modalities Test [P=.009]),
indicating poorer performance for the group with LD and multiple concussions
compared with other groups. A discriminant function analysis using neuropsychological
testing of athletes 24 hours after acute in-season concussion compared with
controls resulted in an overall 89.5% correct classification rate.
Conclusions Our study suggests that neuropsychological assessment is a useful indicator
of cognitive functioning in athletes and that both history of multiple concussions
and LD are associated with reduced cognitive performance. These variables
may be detrimentally synergistic and should receive further study.
The management of mild traumatic brain injury (MTBI; eg, concussion,
defined as a traumatically induced alteration in mental status not necessarily
resulting in loss of consciousness) in athletics is currently one of the most
compelling challenges in sports medicine. Despite the high prevalence1 and potentially serious outcomes2,3
associated with concussion, systematic research on this topic is lacking.
Many sports medicine practitioners are not satisfied with current return-to-play
and treatment options, which do not appear to be evidence based.4- 6
There is also little research examining whether long-term cognitive morbidity
is associated with concussion. Past research with nonathletes revealed that
repeated concussions appear to impart cumulative damage, resulting in increasing
severity and duration with a second MTBI occurring within 48 hours.7 No data were presented which addressed more long-term
Although survey data have shown that a prior history of head injury
increases the risk for sustaining subsequent MTBI,8
other potential risk factors associated with sports-related concussion have
not been identified. Learning disability (LD), the etiology of which is presumably
secondary to central nervous system dysfunction,9
refers to a heterogeneous group of disorders manifested by difficulties in
the acquisition and use of listening, speaking, writing, reading, reasoning,
or mathematical abilities and which is traditionally diagnosed in early childhood.10,11 The incidence of diagnosed LD is
11.8% in the general university population.12
However, no study to date has addressed whether LD may represent a risk factor
(such as that seen with prior head injury) for poor outcome following sports-related
MTBI in college athletes.
Previous research has outlined the reliability, validity, and sensitivity
of neuropsychological tests in assessing the specific cognitive areas associated
with MTBI in the general population.13- 15
To date, 3 published studies have examined the use of neuropsychological testing
in US football players.16- 18
The only multicenter study16 was conducted
in the mid-1980s and was designed to address the acute effects of concussion.
The current study was designed to address 2 issues: first, to investigate
whether a relationship exists between prior concussion and diagnosed LD among
college football players and determine the influence of these variables, in
isolation and combination, on baseline neuropsychological performance; and
second, to evaluate the use of a neuropsychological test battery in diagnosing
concussion and delineating recovery of cognitive function following MTBI in
Participants in this study consisted of 393 male college football players
from 4 Division IA football programs: Michigan State University, East Lansing
(n=119); the University of Florida, Gainesville (n=106); the University of
Pittsburgh, Pittsburgh, Pa (n=85); and the University of Utah, Salt Lake City
At the initial preseason baseline session, the following self-reported
data were collected: age, playing position, SAT/ACT scores (Scholastic Aptitude
Test/American College Testing, ie, college entrance examination scores), history
of LD, neurological history (eg, central nervous system neoplasm or epilepsy),
history of psychiatric illness (eg, depression and/or mania or anxiety), history
of alcohol and/or drug abuse, prior sports played, and history of concussion.
Educational records at each institution were used to verify a documented history
of diagnosed LD. A standardized concussion history form was administered at
baseline to obtain detailed information regarding previous concussions, year
of concussion, description of incident, nature and duration of relevant symptoms
(eg, confusion and/or disorientation, retrograde and/or anterograde amnesia,
and loss of consciousness), neuroimaging results (if any), and days lost from
participation in football (if any). Athletes who reported amnesia were asked
to provide any known collateral information from the athletic trainer, sports-medicine
physician, or other source familiar with the details of the incident. All
previous concussions were classified using the practice parameter of the American
Academy of Neurology.19
Appropriate review for research with human subjects was granted separately
from the 4 institutions at which the participants were enrolled. Each participant
provided written informed consent for voluntary participation. All data collection
was completed by the research team of clinical neuropsychologists (clinicians
with PhDs or doctoral-level students) or team physicians or athletic trainers
who were thoroughly trained in the use of the measures. Each examiner was
required to attend a 2-hour workshop and was supervised during test adminstration
(by M.W.C.) to facilitate the appropriate standardized administration of the
test battery. All measures were administered and scored in a standardized
manner to minimize differences between test administrators and institutions.
Project investigators trained in neuropsychological assessment completed all
data scoring and interpretation.
Baseline data collection at 3 universities (Michigan State University,
University of Pittsburgh, and University of Florida) was completed prior to
the 1997/98 and 1998/99 football seasons during the months of May to August.
Baseline data collection at the University of Utah occurred during February
1999 for the 1999/2000 season (only baseline data from the University of Utah
were used for analyses). Approximately 95% of all roster football players
(scholarship and scout team players) voluntarily participated in the project.
At these baseline sessions, demographic and player history information was
obtained via interview.
Each athlete was then administered a battery of neuropsychological tests
(approximately 30 minutes in length) that is used by the National Football
League.17,20 Tests in the battery
were the Hopkins Verbal Learning Test (HVLT; verbal learning, delayed memory);
Trail-Making Tests, Forms A and B (Trails A and Trails B; visual scanning
and executive functioning); Digit Span Test (attention and concentration);
Symbol Digit Modalities Test (SDMT; information processing speed); Grooved
Pegboard Test, dominant and nondominant hand (bilateral fine motor speed);
and the Controlled Oral Word Association Test (COWAT; word fluency). This
test battery, described in detail elsewhere,17
was constructed to evaluate multiple aspects of cognitive functioning. In
addition to neuropsychological testing, athletes also completed the Concussion
Symptom Scale17 to assess a baseline level
of self-reported symptoms. This Likert scale consists of 20 symptoms commonly
associated with concussion (eg, headache, dizziness, and trouble falling asleep),
with symptoms ranging from none (score, 0) to severe (score, 6).
Athletes who sustained a concussion during the course of a season underwent
serial neuropsychological evaluations following the incident (within 24 hours
of the incident, and at days 3, 5, and 7 postinjury). Concussion was defined
according to the American Academy of Neurology practice parameter.19 Thus, players experiencing a traumatically induced
alteration in mental status, not necessarily resulting in a loss of consciousness,
were included. Athletic trainers initially identified the majority of suspected
concussions, and respective team physicians performed the examinations and
made the final decisions. Once the diagnosis was established, neuropsychological
testing was administered as soon as possible following injury (within 24 hours
in all cases). The neuropsychological tests and self-report inventory used
in the postinjury phase were identical to those used at baseline, although
alternate and reliable forms of the HVLT and COWAT were administered to minimize
learning effects associated with these measures.
Football players from within the sample served as controls. Control
athletes were matched with athletes who sustained concussion according to
ACT/SAT scores, history of LD, history of previous concussion, institution,
and playing position. In addition, to control for exertion, each control athlete
was tested within the same time frame as the athletes who experienced concussion
(eg, following a game or practice). Within the context of these variables,
it was possible for controls to be matched to more than 1 player with concussion.
No control athlete experienced a concussion during the course of the study.
Controls were excluded from further study.
Data from the 4 universities were pooled and analyzed using Statistica Version 5.1 statistical software for Windows.23
To explore the relationship between prior history of concussion, diagnosis
of LD, and neuropsychological baseline performance, multiple analysis of variance
(MANOVA) was performed. Concussion history (no prior concussion vs 1 vs ≥2
concussions) and LD (positive or negative diagnosis) were entered as independent
variables, and cognitive and symptom total scores were entered as dependent
measures. The MANOVA design was selected to allow an analysis of performance
differences between the athletes with different concussion and LD histories,
across multiple neuropsychological domains. This design also permitted an
analysis of possible interaction effects between concussion and LD histories.
For in-season (postconcussion) data, a discriminant function classification
analysis was conducted to determine the accuracy of the neuropsychological
test battery in separating athletes with concussions from control athletes
within 24 hours of concussion. The 8 tests constituting the neuropsychological
test battery were used as predictor variables, and membership in the group
with concussions or control group was used as the dependent (grouping) variable.
To provide preliminary information regarding the recovery pattern of
athletes with concussions relative to the control group and to their own baseline
performance, standard scores were created to convert the selected neuropsychological
test scores to a common metric. These standard scores were constructed so
that baseline performance for each group would have a mean of 100 and SD of
15.21 Group differences of one-half SD (7.5
standard score units) are considered to reflect at least a moderate difference
between the means.22 Any deviation from 100
indicates a change in performance relative to baseline for each group. The
recovery pattern of players who sustained concussion across different time
intervals was evaluated by standardizing all neuropsychological test results
and comparing performance of the athletes with concussion with controls' performance
within 24 hours, and at 3, 5, and 7 days postinjury.
The multiuniversity sample included 393 male football players with a
mean (SD) age of 20.4 (1.7) years and 2.6 (1.3) mean (SD) years in college.
Forty-six percent of the sample was African American, 48% European American,
4% Polynesian American, 1% Asian American, and 1% Hispanic American. Of the
393 players, 6% (n=25) were quarterbacks; 8% (n=33), running backs; 13% (n=52),
wide receivers; 16% (n=64), offensive linemen; 6% (n=23), tight ends; 17%
(n=67), defensive backs; 16% (n=61), defensive linemen; 13% (n=48), linebackers;
and 5% (n=20), kickers.
Of the players completing the ACT examination to qualify for college
admission (n=180), the mean (SD) score was 20.0 (1.7). Of those qualifying
with the SAT (n=200), the mean (SD) score was 952.9 (149.1). College admission
scores were missing for 13 individuals. Three players in the sample reported
a documented history of diagnosed psychiatric illness (eg, bipolar disorder
and major depression). These players completed the baseline evaluation, but
were excluded from further study. No player in the sample reported a diagnosis
of major neurological disorder or history of abuse of alcohol or other drugs.
Forty-six percent (n=179) of the sample reported no prior history of
concussion, 34% (n=129) reported experiencing 1 concussion of any grade, and
20% (n=79) reported a history of 2 or more sustained concussions (range, 2-10)
of any grade. A significant relationship was found between total years participating
in football and total number of concussions sustained (r=0.15; P≤.02). Quarterbacks (17 of 25)
and tight ends (15 of 23) had the the highest rates of prior concussion (68%
and 65%, respectively). Running backs–fullbacks (11 of 33) and kickers-punters
experienced the lowest rates of prior concussion (33% and 46%, respectively).
The prevalence of LD within the total sample of 393 athletes was 13.5%
(n=53). Of the players with no history of concussion (n=179), 10.6% (n=19)
had a diagnosed LD; of those who had experienced 1 prior concussion (n=129),
14.7% (n=19) had diagnosed LD, and of those who had experienced multiple concussions
(n=79), 19.0% (n=15) had a diagnosed LD. Although these data suggest a possible
trend between history of LD and history of multiple concussions, this relationship
was not statistically significant (χ2=3.74; P=.15).
The MANOVA yielded significant main effects for both LD (F=4.57; P<.001) and concussion history (F=1.91; P=.009) on neuropsychological test results, which indicated that both
of these variables were significantly related to overall neuropsychological
performance. The interaction of LD and concussion history was not significant
(F=1.17; P=.28). A follow-up series of univariate
F tests was completed to identify the specific neuropsychological measures
that accounted for the significant MANOVA. Tests for the LD main effect were
Trails B (F=15.98;P<.001); SDMT (F=22.9; P<.001); COWAT (F=11.6; P<.001);
and Hopkins delayed memory (F=11.8; P<.001). For
the history of concussion main effect, significant tests included Trails B
(F=6.1; P=.002); SDMT (F=7.8; P<.001); and total symptoms reported (F=4.6; P=.01).
To evaluate concussion group differences on the neuropsychological tests,
additional post hoc analyses were conducted using the Tukey Honest Significant
Difference test for unequal subjects.24Table 1 presents the group means (SDs)
for athletes. The group with no history of concussion reported fewer symptoms
than both the single concussion group (P=.04) and
the multiple concussion group (P<.001) on the
concussion symptom inventory. Baseline symptoms increased as the number of
concussions increased. On Trails B, the multiple concussion group performed
significantly worse at baseline than the group with no history of concussion
(P=.02) and the single concussion group (P<.001). Baseline data also differed significantly on the SDMT with
the multiple concussion group performing worse than both the group with no
history of concussion (P=.008) and the single concussion
group (P<.001). These findings are not attributed
to preexisting group differences in terms of aptitude as the multiple concussion
group had higher SAT and ACT scores than did the group with no history of
concussion and the single concussion group. The table presents demographic
and neuropsychological test data for the group with LD and the group without
To investigate the interplay between concussion history and LD on baseline
neuropsychological test performance, a concussion history and LD interaction
term was constructed. Univariate F tests for all 10 neuropsychological variables
demonstrated statistically significant interactions for Trails B (F=4.99; P=.007) and SDMT (F=4.74; P=.009).
In both cases, athletes with a history of multiple concussions and LD performed
significantly worse than did athletes with no history of LD who had experienced
multiple concussions (Figure 1).
Nineteen players in the study sample were diagnosed by team medical
staff as sustaining a concussion during the course of the 1997-1999 seasons.
Thirteen individuals sustained a grade 1 concussion (mental status abnormalities
resolved within 15 minutes), 4 athletes sustained a grade 2 concussion (mental
status abnormalities that lasted longer than 15 minutes, but resolved within
45 minutes), and 2 athletes sustained a grade 3 concussion (brief [approximately
5-10 seconds] loss of consciousness). The time between baseline testing and
in-season concussions ranged from 5 weeks to nearly 18 months.
The group with concussions consisted of 16 athletes who had completed
all of the neuropsychological measures (3 athletes were excluded because they
either missed a testing session or failed to complete at least 1 of the tests
at 1 session). The control group consisted of 10 matched control athletes.
Given the different numbers of subjects in the 2 groups, a priori probabilities
for group membership were set at 62% for the group with concussions and at
38% for the control group. Based on testing conducted 24 hours after injury,
the neuropsychological test battery resulted in an overall 89.5% correct classification
rate for the 2 groups (87.5% correct classification for players with concussions;
90% correct classification for the control subjects). Current data provide
preliminary support regarding the sensitivity of our test battery in classifying
athletes with concussions vs controls. Although the number of subjects in
each group is small, comparison of the magnitude of change in performance
for each group is presented in Figure 2.
Performance of the athletes with concussions was noticeably worse (approximately
1 SD) than the control group on Hopkins Total (trials 1-3 of the HVLT) and
delayed memory (total delay score on the HVLT) within 24 hours after injury.
Moderate differences in performance persisted between groups until at least
day 5 postconcussion.
The results of this study suggest that history of concussion and LD
are independently related to lower baseline cognitive performance within a
large, multiuniversity sample of football players. The negative impact of
LD on neurocognitive functioning in a general population is well established.25 Within our sample, the domains of executive functioning
(ie, ability to plan and execute a nonverbal behavior), speed of information
processing, speeded word fluency, and memory appear to be attenuated in those
athletes with LD. We also found that a history of concussion is significantly
and independently associated with long-term deficits in the domains of executive
functioning and speed of information processing, as well as an increase in
self-reported symptoms. Such defects appear to be present in those individuals
who have sustained 2 or more prior concussions. In this study, a history of
1 concussion does not appear to result in the long-term cognitive morbidity
associated with 2 or more episodes of concussion.
The study findings also suggest an interaction of prior concussion and
LD on select neuropsychological measures. Players with a history of 2 or more
concussions and LD performed significantly worse on tests of executive functioning
(Trails B) and speed of information processing (SDMT) relative to players
with concussions who were not diagnosed with LD, suggesting an additive effect
of LD and multiple episodes of concussion on lowered functioning. When compared
with accepted normative values,26 athletes
with LD and multiple concussions performed in the brain impairment range on
these 2 measures.
Three potential hypotheses may account for these findings. First, athletes
with LD who experienced concussion may have had less brain reserve capacity27 than athletes without LD. The margin of cognitive
reserve may be less in athletes with LD and the threshold for manifesting
neurobehavioral morbidity may be lower. A second hypothesis is that LD may
have made the initial diagnosis of concussion more complex and confusing.
For example, an athlete without LD, on initial postinjury assessment, may
have revealed discernable and unequivocal cognitive and neurobehavioral impairment
when compared with baseline presentation. Conversely, the athletes with LD
may have had difficulty performing in the requisite skill areas, both before
and after concussion, which were assessed during the acute phase of injury
(eg, serial 7s, digits backward, months of year in reverse order). Since cursory
examination did not reveal overt impairment relative to baseline presentation,
these athletes may have returned to play earlier than athletes without LD.
Subsequent MTBI may have occurred prior to full neurological recovery. A third
hypothesis is that the athletes with LD may have had difficulty learning the
proper techniques for safe play or could have neurobehavioral characteristics
(eg, impulsivity and attentional impairment) leading to increased risk of
Our data potentially have strong academic implications. Athletes with
LD10 have difficulty in areas of academic achievement,
especially within the context of excessive academic and athletic demands.
Our data suggest that experiencing 2 or more prior concussions is associated
with an attenuation of cognitive skills, which, when combined with the deficits
associated with an LD, leads to even further compromised functioning. It is
logical to assume that the cognitive domains selectively affected (ie, executive
functioning and speed of information processing) are prerequisites for academic
success. Thus, academic achievement may become even more difficult for these
athletes. This issue is important given that 20% of our sample experienced
Our study has several limitations. First, although educational records
confirmed a diagnosis of LD, concussion history was obtained via structured
clinical interview and thus represents nonverifiable self-reported data. However,
concussion history is not typically presented in formal medical records and
many concussions go unrecognized by trainers or team medical staff.8 Thus, a self-reported history is essential when examining
concussion history and usually represents the only available source of information.
Second, although we attempted to gather information regarding alcohol and
drug use within our sample, the data obtained were self-reported and may be
subject to underreporting.
Data from in-season concussions, although limited by a small sample
size, suggest that comprehensive neuropsychological evaluation may prove valuable
in the acute and follow-up evaluation of the athlete with a concussion. Any
symptoms (cognitive deficit or self-reported) should alert the clinician that
full neurological recovery has yet to occur and return to play is contraindicated
until these symptoms resolve. Furthermore, mental status screening instruments
are variable in nature and may assess only cursory cognitive functioning (eg,
no speed of processing or comprehensive memory component), thus missing the
subtle deficits associated with injury. In this study, verbal learning and
memory appeared to be the most sensitive components in discriminating athletes
with concussions from control athletes at the within-24-hours testing session.
Less striking differences were found across other cognitive domains. However,
careful examination of postconcussion data at days 3, 5, and 7 revealed greater
improvement in performance by control subjects relative to athletes with concussion.
Specifically, athletes with concussions failed to demonstrate the magnitude
of "learning effect" manifested by control subjects, especially across the
domains of speed of information processing and executive functioning. This
failed learning effect has been observed in previous research.16
These types of subtle cognitive impairments can be detected through formal
neuropsychological testing procedures. With the development and implementation
of similar neuropsychological programs, this type of objective information
may prove useful in helping athletes, their families, and physicians make
more informed decisions about returning to play, thus protecting the athlete
from potential adverse effects of concussion.
Finally, identification of other factors associated with risk of concussion
may be found through the study of genetic markers such as apolipoprotein E
polymorphism. This marker has only recently been linked to poor outcome at
6 months following head injury28 and has been
identified as a risk factor for boxer's dementia.29
Future studies of sports-related concussion should more systematically investigate
the importance of genetic factors in recovery.