Neuropsychology

© 2020 American Psychological Association 2020, Vol. 34, No. 8, 881– 893
ISSN: 0894-4105 http://dx.doi.org/10.1037/neu0000690

Chronic White Matter Changes Detected Using Diffusion Tensor

Imaging Following Adult Traumatic Brain Injury and Their Relationship
to Cognition

Erica J. Wallace, Jane L. Mathias, and Lynn Ward Kerstin Pannek, Jurgen Fripp, and Stephen Rose
University of Adelaide Australian E-Health Research Centre, CSIRO, Brisbane,
Australia
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Objective: White matter (WM) changes detected using diffusion tensor imaging (DTI) are reportedly
This document is copyrighted by the American Psychological Association or one of its allied publishers.

related to cognitive outcomes following traumatic brain injury (TBI), but much existing research is
underpowered or has only examined general outcomes, rather than cognitive functioning. Method: A
large sample of adults who had sustained mild, moderate or severe TBIs seven months prior (N ⫽ 165)
and a control group (N ⫽ 106) underwent DTI and cognitive testing. Fractional anisotropy and mean
diffusivity were calculated for 5 regions (corpus callosum: genu, body, splenium; fornix; superior
longitudinal fasciculus) that recent meta-analyses identified as being affected by TBI and related to
cognition following TBI. Memory, attention and executive functioning, which are often affected by TBI,
were assessed. Results: Overall, mild TBI did not show significant WM or cognitive changes, relative
to controls, but moderate to severe TBI was associated with large WM alterations (all regions) and poorer
cognitive performance. No significant correlations were found between DTI findings and cognition in the
moderate to severe group. Conclusions: The findings have shown that moderate to severe TBI leads to
considerable WM and cognitive changes. Early and ongoing examination of mild TBI is needed to
determine whether WM and cognitive changes are initially present and, if so, when they resolve.

Key Points
Question: Are white matter (WM) changes, detected using diffusion tensor imaging (DTI), related
to cognitive outcomes following traumatic brain injury (TBI)? Findings: Relative to controls, mild
TBI did not show significant WM or cognitive changes, but moderate to severe TBI was associated
with large WM alterations and poorer cognitive performance. Importance: Moderate to severe TBI
leads to considerable cognitive and WM changes that can be detected using DTI. Next Steps: Early
and ongoing examination of mild TBI is needed to determine whether WM and cognitive changes are
initially present and, if so, when they resolve.

Keywords: traumatic brain injury, diffusion tensor imaging, adults, outcomes

Supplemental materials: http://dx.doi.org/10.1037/neu0000690.supp

Diffusion tensor imaging (DTI) better detects microstructural Barker, Greenwood, & Duncan, 2001). More particularly, DTI has
changes to brain tissue than conventional neuroimaging techniques been used to identify diffuse axonal injury (DAI), which is thought
(computed tomography, structural MRI) following traumatic brain to contribute to the cognitive problems that people experience after
injury (TBI; Oehr & Anderson, 2017; Rugg-Gunn, Symms, a TBI (Voelbel, Genova, Chiaravalotti, & Hoptman, 2012).

Erica J. Wallace, Jane L. Mathias, and Lynn Ward, Faculty of Health & Medical analysis, methodology, supervision, and writing–review and editing. Ker-
Sciences, University of Adelaide; X Kerstin Pannek, Jurgen Fripp, and Stephen stin Pannek contributed equally to methodology and served in a supporting
Rose, Australian E-Health Research Centre, CSIRO, Brisbane, Australia. role for conceptualization, resources, and writing–review and editing.
This work was supported by the National Health and Medical Research Jurgen Fripp served in a supporting role for formal analysis, resources, and
Foundation (Project Grant 519220). writing–review and editing. Stephen Rose served in a supporting role for
The authors have no competing financial interests to disclose. resources and writing–review and editing. Erica Wallace and Kerstin
Erica J. Wallace served as lead for methodology and writing– original Pannek contributed to formal analysis equally. Jane L. Mathias and Ste-
draft and contributed equally to conceptualization. Jane L. Mathias served phen Rose contributed to funding acquisition equally.
as lead for supervision and writing–review and editing and served in a Correspondence concerning this article should be addressed to Erica J.
supporting role for conceptualization, formal analysis, and methodology. Wallace, Faculty of Health & Medical Sciences, The University of Adelaide,
Lynn Ward served in a supporting role for conceptualization, formal South Australia, Australia, 5005. E-mail: [email protected]

881
 882 WALLACE ET AL.

DTI quantifies the diffusion of water molecules in the brain and Disparate findings have also been reported by studies that have
involves the calculation of a ‘diffusion tensor’ (see Mukherjee, examined the relationship between cognitive outcomes and DTI
Berman, Chung, Hess, & Henry, 2008 for a detailed explanation). following TBI. In particular, one review found that higher FA
The diffusion tensor can be visualized as a ‘diffusion ellipsoid,’ and/or lower MD were positively related to better cognitive func-
with its shape being dependent on the type of tissue that is being tioning in a number of studies but, in other studies, the relation-
examined (e.g., white or gray matter) and any physiological ships were reversed or nonsignificant (Hulkower, Poliak, Rosen-
changes caused by primary and secondary trauma (Hutchinson, baum, Zimmerman, & Lipton, 2013). Indeed, even studies that
Schwerin, Avram, Juliano, & Pierpaoli, 2018). Where there are no have investigated the relationship between the same cognitive
microstructural elements constraining diffusion, as is the case for domain and ROI have reported inconsistent findings. For example,
cerebrospinal fluid, diffusion occurs equally in all directions (Hu- the relationship between memory performance and FA in the
isman, 2010). This is known as isotropic diffusion and results in a cingulum has been found to be both positive (e.g., Geary, Kraus,
spherical diffusion ellipsoid (Amyot et al., 2015; Huisman, 2010). Pliskin, & Little, 2010; Kraus et al., 2007) and negligible (e.g.,
In contrast, the microstructural organization of healthy white mat- Jang et al., 2013) in different studies.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

ter (WM) restricts diffusion; diffusion occurs freely parallel to the Another meta-analysis, which examined the relationship be-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

axons, while being restricted in other directions. This causes the tween DTI and cognition following TBI, reported that WM integ-
diffusion ellipsoid to become elongated, with the long, principal rity (high FA, low MD) was associated with cognitive functioning
axis aligned with the white matter pathway (Huisman, 2010; following mild, moderate and severe TBI (Wallace, Mathias, &
Shenton et al., 2012; Suri & Lipton, 2018). This directional dif- Ward, 2018b). Specifically, WM integrity in the corpus callosum
fusion is known as anisotropic diffusion (Koerte, Hufschmidt, was related to better general cognition, memory, attention, exec-
Muehlmann, Lin, & Shenton, 2016). The DTI metric used to utive functioning, construction and motor performance, verbal and
quantify the directional dependence of diffusion is fractional an- language skills, and concept formation and reasoning. WM integ-
isotropy (FA), which is measured on a scale of 0 (indicating rity in the fornix was associated with better memory, attention, and
spherical/isotropic diffusion) to 1 (reflecting directional/anisotro- verbal/language skills and, in the superior longitudinal fasciculus,
pic diffusion; Suri & Lipton, 2018). In WM regions in adults, it was related to better general cognition, attention and executive
higher values may reflect WM integrity, while lower values may
functioning. Memory, attention and executive functioning were the
reflect WM damage (Koerte et al., 2016; Voelbel et al., 2012). In
most frequently examined of all of the cognitive domains, with
contrast, mean diffusivity (MD) or apparent diffusion coefficient
memory and attention most strongly related to the DTI findings in
measure the magnitude of diffusion, providing an average of
the corpus callosum, fornix and superior longitudinal fasciculus
diffusion along the three axes of the ellipsoid (Amyot et al., 2015;
(Wallace et al., 2018b). These findings were from studies that
Douglas, Muldermans, & Wintermark, 2018; Voelbel et al., 2012).
examined people with TBIs of all severities (mild to severe) at a
MD is also restricted by the microstructural organization of the
range of postinjury intervals (acute to chronic), based on initial
WM; low MD values may indicate healthy WM and high MD
subgroup analyses that suggested these findings could be com-
values may reflect WM damage (Niogi & Mukherjee, 2010).
bined. However, it is possible that this method led to important
DTI has increasingly been used in studies of TBI, with most
differences being missed, given that FA and MD findings may
reporting lower FA and higher MD following injury (Amyot et al.,
2015; Shenton et al., 2012). Specifically, higher FA and lower MD differ in mild compared to more severe injuries and when DTI is
have been found in a range of brain regions following mild TBI, performed in the acute compared to later time periods (see reviews
including the corpus callosum, internal capsule, external capsule, by Asken, DeKosky, Clugston, Jaffee, & Bauer, 2018; Shenton et
superior longitudinal fasciculus, corticospinal tract and sagittal al., 2012). In addition, there was limited overlap in the brain
stratum (e.g., Arfanakis et al., 2002; Dean, Sato, Vieira, McNa- regions and/or cognitive domains that were examined by the
mara, & Sterr, 2015; Inglese et al., 2005; Kraus et al., 2007; Zhu contributing studies, with most of the findings from this meta-
et al., 2014). These WM changes tend to be larger more following analysis based only on one or two studies (94%) and relatively
severe injuries (i.e., lower FA, higher MD; e.g., Kraus et al., 2007; small samples (Nparticipants ⬍26 in 60% of the studies). Thus, the
Matsushita, Hosoda, Naitoh, Yamashita, & Kohmura, 2011). How- current literature largely consists of studies with low statistical
ever, the changes to FA and MD may be reversed when DTI is power, which may contribute to the so-called ‘replication crisis’
performed soon after injury, with FA reported to be higher and MD whereby studies with insufficient statistical power fail to replicate
lower 72 hr after a mild TBI (Bazarian et al., 2007) and MD being expected findings (see Shrout & Rodgers, 2018).
lower within one week of sustaining a TBI (Huisman et al., 2004). More recently, there has been a large-scale study that examined
A recent meta-analysis examining the location and extent of the relationship between DTI performed in the subacute period
WM changes found that, following a TBI, there were widespread (median ⫽ 19 days postinjury) and general recovery in a large
WM alterations (lower FA, higher MD) in a large number of brain moderate to severe TBI sample (N ⫽ 217; Castaño Leon et al.,
regions (regions of interest: ROIs), regardless of when the DTI was 2018). This study found that lower FA in most of the examined
performed (Wallace, Mathias, & Ward, 2018a). These changes regions was associated with unfavourable outcomes (strongest
were evident even after mild TBI, which is notable given that this correlations: corpus callosum, cingulum, cerebral peduncles),
damage is rarely detected using conventional imaging (Shenton et which was measured using the extended Glasgow Outcome Scale
al., 2012; Strauss et al., 2015), despite 15% to 30% of people at six and 12 months postinjury. Although large in scale, this study
purportedly experiencing long-term cognitive impairments follow- did not examine specific cognitive outcomes or mild TBI, high-
ing mild TBI (McKee & Daneshvar, 2015; Niogi & Mukherjee, lighting the need for continued research to strengthen the knowl-
2010). edge base.
 WHITE MATTER CHANGES AND COGNITION FOLLOWING TBI 883

The current study therefore examined the DTI and cognitive Northopedic ⫽ 37, Nhealthy ⫽ 24) because they did not have usable
data from a large TBI sample, in addition to a control group MRI images (e.g., failed to have an MRI, excessive head move-
(comprising healthy persons and people with orthopedic injuries), ment, scanner artifacts; NTBI ⫽ 41, Northopedic ⫽ 15, Nhealthy ⫽ 8),
in order to assess the reliability and generalisability of the findings they had minor asymptomatic brain abnormalities on their MRI
from the aforementioned meta-analyses (Wallace et al., 2018a, (identified by the radiologist as “incidental findings of no conse-
2018b). To this end, it examined whether (a) WM integrity was quence”; NTBI ⫽ 11, Northopedic ⫽ 22, Nhealthy ⫽ 16), or the time
compromised (lower FA, higher MD) in the corpus callosum between the injury and the MRI was excessive (i.e., identified as
(genu, body, splenium), fornix and superior longitudinal fasciculus extreme outliers; more than 400 days; NTBI ⫽ 4; Tukey, 1977).
following mild and moderate to severe TBI, relative to controls One healthy control participant who was found to have previously
and whether damage was more notable following more severe sustained a head injury was also excluded (i.e., TBI occurred prior
injuries; (b) the mild and moderate to severe TBI groups per- to the study recruitment period). The final sample therefore con-
formed more poorly on tests of memory, attention and executive sisted of 165 people with TBI (Nmild ⫽ 134, Nmoderate ⫽ 15,
functioning; and (c) cognitive performance was related to WM Nsevere ⫽ 16), 47 orthopedic controls and 59 healthy controls. The
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

integrity/damage following TBI, and whether these relationships moderate and severe groups were both small, therefore they were
This document is copyrighted by the American Psychological Association or one of its allied publishers.

were equivalent in the controls. These five ROIs were chosen combined for all subsequent analyses.
based on of the findings from two recent meta-analyses (Wallace The two control groups were compared in terms of demo-
et al., 2018a, 2018b). Memory, attention and executive functioning graphic, cognitive and DTI variables, based on a recent study in
were examined because, in addition to being commonly affected which WM integrity differed in orthopedic compared to healthy
by TBIs (Cristofori & Levin, 2015), were frequently examined by controls (Wilde et al., 2019). Consistent with the findings from a
the meta-analyzed studies (Wallace et al., 2018b). WM integrity larger sample taken from this research project (Mathias, Denning-
was expected to be lower following mild and moderate to severe ton, Bowden, & Bigler, 2013), the orthopedic and healthy control
TBI (relative to controls), and related to cognitive functioning groups did not differ significantly in terms of their: age,
(memory, attention, executive functioning), which was expected to t(104) ⫽ ⫺.474, p ⫽ .637, education, t(102) ⫽ .312, p ⫽ .755,
be poorer following a TBI. proportion of males and females (␹2(1) ⫽ 3.628, p ⫽ .057),
cognitive performance, or FA and MD values from the five ROIs
Method (see Table S1 in online Supplemental materials for additional
summary descriptive data and statistical comparisons). Therefore,
Participants these two groups were combined to form a single control group
(Ncontrols ⫽ 106) for use in all subsequent analyses.
The study participants were all adults who were involved in a
larger research project that examined outcomes after TBI, which
Measures
was undertaken at the Royal Adelaide Hospital (2008 –2012).
Three groups were recruited for this study: TBI, orthopedic con- Cognitive tests. Memory, attention and executive functioning
trols and healthy controls. The TBI participants had all sustained a were assessed using two subtests from the Wechsler Memory
nonpenetrating TBI and the orthopedic controls had sustained an Scale-Third Edition (WMS-III; Wechsler, 1997), two computer-
orthopedic injury that did not involve the head or face (to avoid the ized reaction time (RT) tasks, and the Controlled Oral Word
possibility of a concussion or mild TBI). A separate group of Association Test (COWA; Spreen & Straus, 1998), respectively.
healthy controls was additionally recruited, consisting of family or These measures formed part of a larger battery of tests and self-
friends of the TBI sample and visitors to the Royal Adelaide report scales that were completed by all participants.
Hospital. Eligible participants (a) were aged between 18 and 80 Memory was assessed using the WMS-III Logical Memory and
years; (b) spoke English as their first language (necessary to Visual Reproduction subtests (immediate and delayed recall/tri-
complete the cognitive assessments); (c) did not have any known als). The Logical Memory task requires participants to verbally
psychiatric or neurological disorders, intellectual disabilities, or recall two stories, both immediately after hearing each story (im-
history of substance abuse; and (d) were able to undergo MRI (no mediate recall; LM-I) and after a delay of 25 to 35 min (delayed
contraindications) and cognitive testing. TBI severity was classi- recall; LM-II). The Visual Reproduction task involves participants
fied as mild, moderate or severe, based on their lowest Glasgow drawing five geometric designs from memory having seen each
Coma Scale (GCS) scores (mild: 13–15; moderate: 9 –12; severe: design for 10 s, both immediately (immediate recall; VR-I) and
ⱕ8) or, where GCS scores were not available, duration of loss of after a delay of 25 to 35 min (delayed recall; VR-II). Scores on
consciousness (LOC; mild: 0 –30 mins; moderate: 30 mins– 24 hr; both tests were age-scaled, to control for age-related differences in
severe: ⬎24 hr), and/or duration of posttraumatic amnesia (PTA; ability, and standardized to a mean of 10 (SD ⫽ 3) using the
mild: 0 –1 day; moderate: 1–7 days; severe: ⬎7 days; Amyot et al., normative data provided in the WMS-III manual (Wechsler, 1997).
2015). GCS scores were used as the primary method for determin- Attention was assessed using 4-choice compatible and 4-choice
ing injury severity because LOC and PTA were both based on incompatible visual RT tasks (Mathias, Beall, & Bigler, 2004;
self-report, which is often unreliable, with people tending to over- Mathias, Bigler, et al., 2004). Reaction time tasks assess informa-
estimate the length of their LOC and/or PTA (see Menon et al., tion processing speed, which is strongly associated with attention
2010; Sherer et al., 2015). (Chiaravalloti & DeLuca, 2008; Mathias, Bigler, et al., 2004;
A total of 221 people who had sustained a nonpenetrating TBI, Ponsford & Kinsella, 1992). Briefly, participants were presented
84 orthopedic and 84 healthy control participants were initially with four white rectangles (stimuli) on a computer screen, two on
recruited. In total, 117 participants were excluded (NTBI ⫽ 56, each side of a central fixation point. Participants were required to
 884 WALLACE ET AL.

respond as quickly and accurately as possible when one of the acquired for each participant (1 mm3 isotropic 3D T1 MPRAGE
rectangles turned red by pressing a button (response) with their sequence following ADNI recommendations: http://www.loni.ucla
middle (outer rectangle) or index (inner rectangle) finger. The .edu/ADNI/Research/Cores/ADNI_Siemens_Human_3TVB15_Trio
compatible task required participants to respond to the cue using .pdf). The following parameters were used: FOV ⫽ 24 ⫻ 25.6 ⫻
the corresponding finger of the hand on the same side as the 17.6 cm, TR/TE/TI ⫽ 2300/2.26/900ms, flip angle 9°. An opti-
stimulus (i.e., right-handed response to stimuli on the right side of mized diffusion sequence (Jones, Horsfield, & Simmons, 1999)
the screen). The incompatible task was designed to be more was used to acquire diffusion-weighted images along 64 noncol-
complex and to require interhemispheric processing, with partici- linear directions (b ⫽ 3000s/mm2), along with one non-diffusion-
pants responding using the hand on the opposite side to that of the weighted image. The following parameters were used: 60 axial
stimulus (i.e., right-handed response to a left-sided stimulus). slices, FOV ⫽ 25 ⫻ 25 cm, TR/TE ⫽ 9400/116ms, 2.5 mm slice
Median RTs were calculated from 60 trials for each of the two thickness, and acquisition matrix 100 ⫻ 100 with a 2.5 mm
tasks in order to control for lapses in attention and anticipatory isotropic image resolution. Two 2D gradient recalled echo images
responses (Mathias, Beall, et al., 2004; Mathias, Bigler, et al., (TE1/TE2 4.76/7.22ms) were used to acquire a field map for
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

2004). diffusion data, which assists when correcting for distortion due to
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Executive functioning was assessed using the COWA (Spreen & susceptibility inhomogeneities.
Straus, 1998). Participants were required to generate as many Region of interest analysis. DTI images underwent quality
words as possible starting with the letters F, A and S within 1-min control through visual examination and automated detection and
intervals (per letter), with the requirement that the words not be removal of volumes that were affected by significant within-
proper nouns or the same word with a different ending (e.g., sit and volume head motion (Pannek et al., 2017), and were subsequently
sitting). The total number of correct responses (raw scores) was corrected for head movement, eddy current and susceptibility-
recorded. induced distortions, using FMRIB’s Diffusion Toolbox (FMRIB,
Oxford, U.K.; Andersson & Sotiropoulos, 2016; Jenkinson, Beck-
mann, Behrens, Woolrich, & Smith, 2012). Brain extraction was
Procedure
performed using a multiatlas approach, with each mask being
This study was approved by the Human Research Ethics Com- visually checked by the first author (EJW). Preprocessed diffusion
mittees of the University of Adelaide and the Royal Adelaide data were upsampled by a factor of 2, and FA and MD maps
Hospital. Participants for the TBI and orthopedic control groups generated for each participant. A study-specific template was
were identified on a prospective basis via hospital records and sent generated using FA maps from 80 randomly selected participants
a letter briefly describing the study and inviting them to partici- (NTBI ⫽ 40, Northopedic ⫽ 20, Nhealthy ⫽ 20) using Advanced
pate, while also providing a way to opt-out of the study. Those Normalization Tools (ANTs; Avants et al., 2011). This template
who did not opt-out within a 2-week period were contacted by the was registered to the John Hopkins University (JHU) atlas and
researchers. Healthy controls were friends/family of the TBI group individual brains were registered to this using ANTs. The regions
or were recruited via flyers, located throughout the hospital, which contained within the JHU atlas were transformed to each partici-
outlined the study and provided contact details for the researchers. pant’s diffusion space and visually checked by the first author
All three groups were screened (by phone) to establish their (EJW). For each selected region, FA and MD values were ex-
eligibility (using the aforementioned inclusion criteria) and obtain tracted voxelwise, and the region’s median FA and MD values
preliminary verbal consent. were calculated. The five ROIs that were examined for current
Participants were mailed information sheets containing back- purposes were the genu, body and splenium of the corpus callosum
ground information and several self-report questionnaires that (CC), the fornix, and the superior longitudinal fasciculus (SLF).
were relevant to the larger project (e.g., Rivermead Post-
Concussion Symptoms, Community Integration Questionnaire),
Statistical Analyses
which they completed prior to 2–3 hr of cognitive testing (partic-
ipants all tested individually). Written informed consent was ob- Independent samples t tests and chi-square tests were used to
tained at the beginning of the session, prior to being interviewed examine whether the TBI and control groups were demographi-
(regarding background and demographic information, and medical cally comparable (age, education, proportion of males and fe-
history) and undergoing cognitive testing. Participants additionally males). All assumptions for the t tests were met, with the exception
underwent MRI with DTI, as detailed below. MRI scans and of normally distributed data; however as the samples were large
cognitive testing were completed within one week of each other, (⬎50) these tests were deemed appropriate (see Lumley, Diehr,
an average of seven months postinjury (TBI: 200 days, SD ⫽ 41.4, Emerson, & Chen, 2002).
orthopedic controls: 218 days, SD ⫽ 41.8). An honorarium of $40 Next, the FA and MD values from the mild TBI (N ⫽ 134) and
was paid to all participants to assist with out-of-pocket expenses moderate-severe (N ⫽ 31) TBI subgroups and controls (N ⫽ 106)
when traveling to complete the cognitive assessment and MRI. were compared using Welch’s F tests (Welch, 1951) and Games-
Importantly, all data were collected exclusively for research pur- Howell post hoc comparisons (Toothaker, 1993)—the latter being
poses and the results of the cognitive testing were not provided to recommended when samples are unequal in size (Tomarken &
participants, thereby reducing the likelihood they would perform Serlin, 1986)—in order to determine whether WM integrity was
disingenuously or use their performance for litigation purposes. more compromised following more serious injuries (moderate-
MRI acquisition. All participants had an MRI at the Royal severe TBIs). Hierarchical regressions were conducted to deter-
Adelaide Hospital on a Siemens Tim Trio 3T scanner (Erlan- mine whether group differences remained after controlling for age,
gen, Germany), with a high-resolution structural image being sex and time postinjury, all of which may affect WM. Each ROI
 WHITE MATTER CHANGES AND COGNITION FOLLOWING TBI 885

was entered as the dependent variable, and age, sex, time postin- 134, 81.2%), with many fewer having moderate (N ⫽ 15) or severe
jury (step one) and group (TBI or control; step two) were entered (N ⫽ 16) injuries; the latter two being combined (moderate-severe
as independent variables. Hierarchical regressions, rather than TBI: N ⫽ 31, 18.7%) when examining the impact of injury
ANCOVAs, were performed because the groups differed on the severity (see Table 1 for summary details for these subgroups).
covariates (age and time postinjury), and group assignment was TBIs were most commonly caused by motor vehicle accidents,
not random (Miller & Chapman, 2001). followed by falls, bicycle accidents, assaults, and sporting injuries.
The cognitive scores of the mild TBI (N ⫽ 134) and moderate- When the demographic characteristics of the TBI (all TBI) and
severe (N ⫽ 31) TBI subgroups and the controls (N ⫽ 106) were control groups were compared, they were found to differ in terms
then also compared (Welch’s F tests and Games-Howell post hoc of sex (␹2(1) ⫽ 16.36, p ⫽ .000), age, t(269) ⫽ ⫺2.080, p ⫽ .038,
comparisons) to determine whether cognitive performance differed g ⫽ .26, education, t(262) ⫽ 2.724, p ⫽ .007, g ⫽ .37 and interval
according to injury severity. Hierarchical regressions were con- between injury and examination (i.e., compared to orthopedic
ducted to determine whether group differences in cognition re- controls; t(203) ⫽ 2.61, p ⫽ .010, g ⫽ .44), with the TBI group
mained after controlling for age, sex, time postinjury and years of having significantly more males, being older, having completed
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

education because these variables may have impacted on cognitive fewer years of education and having a shorter interval between
This document is copyrighted by the American Psychological Association or one of its allied publishers.

performance. Cognitive scores were entered as the dependent injury and examination (the TBI group was on average 4.4 years
variable, and age, sex, time postinjury, education (step one) and older, had completed 12 months less education than the controls
group (TBI or control; step two) were entered as independent and were examined 18.3 days earlier than the orthopedic controls).
variables. Given that WM and cognitive performance can be impacted by
For all group comparisons, standardized mean differences age, sex, time postinjury and education, these variables were
(Hedges’ g effect sizes) were calculated to evaluate the extent of examined further. The TBI and control groups did not differ in
any differences, with g ⫽ ⫺0.2, ⫺0.5, ⫺0.8 and ⫺1.3 correspond- terms of the number of volumes remaining after rejecting those for
ing to small, medium, large and very large effects, respectively motion, t(269) ⫽ ⫺.94, p ⫽ .347. Similarly, there were no group
(Cohen, 1992; Rosenthal, 1996). All effect sizes were calculated so differences in the number of volumes remaining between the mild
that a negative value indicated that the TBI group had reduced TBI, moderate-severe TBI and control groups, F(2, 94.28) ⫽ .56,
WM integrity (lower FA, higher MD) or poorer cognitive perfor- p ⫽ .573, thus this variable was not controlled for in the analyses.
mance, relative to controls.
Finally, partial correlation coefficients were calculated to deter-
Fractional Anisotropy
mine whether WM integrity (FA & MD) in five ROIs was related
to memory, attention and executive functioning, above what was FA values for the TBI subgroups (mild, moderate-severe) and
accounted for by age, sex, time postinjury and education. These controls were compared to determine whether WM was compro-
were only examined in the moderate to severe TBI group (the mild mised following TBI and whether more severe injuries led to
group did not show WM or cognitive changes, relative to controls), larger WM alterations in each ROI (see Table 2). The FA values
and only for those cognitive tests in which the moderate to severe for the mild TBI and control groups did not differ significantly,
group performed worse than the controls, after controlling for the with the associated effects sizes all being relatively small. In
effects of age, sex, time postinjury and education. The same contrast, all five FA values for the moderate-severe TBI and
associations were then examined separately in the control group, to control groups differed significantly (Bonferroni corrected p ⬍
determine whether the relationships were equivalent in these sub- .01), equating to medium (fornix: g ⫽ ⫺.62) to very large differ-
groups. All coefficients were calculated in such a way that a ences (CC genu: g ⫽ ⫺1.44). Moreover, these differences re-
positive correlation indicated that better cognitive functioning mained even after correcting for age, sex and time postinjury (see
(e.g., more accurate responses, faster RTs) was related to greater Table 3). Lastly, moderate and severe TBIs led to significantly
WM integrity (higher FA, lower MD). Correlations (r) of 0.1, 0.3 lower FA values than mild TBI in the CC (genu, body, splenium),
and 0.5 corresponded to weak, moderate/medium and strong rela- but not the fornix or SLF. Thus, more severe injuries led to less
tionships, respectively (Cohen, 1992). Bonferroni corrections directional/anisotropic diffusion, suggesting greater WM damage.
(Holland & Copenhaver, 1988) were used throughout to compen-
sate for the fact that multiple statistical analyses were performed.
Mean Diffusivity
Results MD values for the TBI subgroups (mild, moderate-severe) and
controls were compared to examine whether more severe injuries
Participants had a greater impact on the magnitude of diffusion (see Table 2).
The mild TBI and control groups did not differ significantly in any
Summary background, demographic and injury details for the ROI, but the moderate-severe TBI and control groups showed
TBI (provided for the full group [all TBI], mild and moderate- medium to large and significant differences in all five ROIs
severe TBI subgroups) and control groups are provided in Table 1. (g ⫽ ⫺.59 to ⫺1.16). Again, these differences could not be
The TBI and control participants were predominantly young to attributed to age, sex or time postinjury (see Table 3). Finally, the
middle-aged, right-handed males. On average, both groups had moderate-severe TBI group showed significantly higher MD than
completed more than 1-year postsecondary education. Most par- the mild TBI group in the CC (genu, splenium) and SLF, but not
ticipants reported that they had not sustained a previous TBI and the body of CC or fornix. Taken together, these findings suggest
only a limited number were involved in any litigation related to that more severe injuries led to a greater magnitude of diffusion,
their injury. The TBI group largely sustained mild injuries (N ⫽ indicative of more WM damage.
 886 WALLACE ET AL.

Table 1
Summary Background Information for the Traumatic Brain Injury and Control Groups

All TBI (N ⫽ 165) Mild TBI (N ⫽ 134) Moderate-severe TBI (N ⫽ 31) Controls1 (N ⫽ 106)
Variable N Mean SD Range N Mean SD Range N Mean SD Range N Mean SD Range

Age 165 43.6 16.8 19–80 134 43.8 17.0 19–80 31 42.5 15.9 20–72 106 39.2 16.7 18–77
Education (years) 160 13.1 2.6 7–22 129 13.3 2.5 8–20 31 12.6 3.2 7–22 104 14.1 2.8 7–20
Days since accident 160 200.0 41.4 98–338 129 194.3 38.2 98–338 31 223.6 46.5 135–336 452 218.3 41.8 136–344
Lowest GCS 143 13.2 3.1 3–15 114 14.5 0.7 13–15 29 8.1 3.6 3–12
LOC (hours) 126 1.7 15.0 0–168 111 0.3 1.3 0–10 15 12.3 43.2 0–168
PTA (hours) 133 31.8 107.0 0–576 107 7.4 49.2 0–504 26 132.4 192.64 0–576

N % N % N % N %
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Gender 165 134 31 106

This document is copyrighted by the American Psychological Association or one of its allied publishers.

Females 35 21.2 30 22.4 5 16.1 47 44.3

Males 130 78.8 104 77.6 26 83.9 59 55.7
Handedness 157 126 31 103
Right 138 87.9 111 88.1 27 87.1 96 93.2
Left 19 12.1 15 11.9 4 12.9 7 6.8
Previous TBI (self-report) 159 128 31 103
Yes 45 28.3 40 31.3 5 16.1 0 0
No 114 71.7 88 68.8 26 83.9 103 100
Involved in litigation 159 128 31 442
Yes 30 18.9 24 18.8 6 19.4 2 4.5
No 129 81.1 104 81.3 25 80.6 42 95.5
TBI severity 165 134 31
Mild 134 81.2 134 100 0
Moderate 15 9.1 0 15 48.4
Severe 16 9.7 0 16 51.6
Cause of injury 165 134 31 472
Motor vehicle 40 24.2 32 23.9 8 25.8 1 2.1
Fall 40 24.2 29 21.6 11 35.5 11 23.4
Bicycle 34 20.6 28 20.9 6 19.4 7 14.9
Assault 30 18.2 26 19.4 4 12.9 0 0.0
Sport 11 6.7 11 8.2 0 18 38.3
Pedestrian 4 2.4 2 1.5 2 6.5 1 2.1
Other 6 3.6 6 4.5 0 9 19.1
Note. TBI ⫽ traumatic brain injury; N ⫽ number of participants; SD ⫽ standard deviation; GCS ⫽ Glasgow Coma Scale; LOC ⫽ loss of consciousness;
PTA ⫽ post-traumatic amnesia; DTI ⫽ diffusion tensor imaging; orthopedic controls N ⫽ 47; healthy controls N ⫽ 59.
1
all controls (orthopedic ⫹ community controls). 2 orthopedic controls only.

Cognitive Outcomes (step one) and group (TBI or control; step two) entered as predic-
tors. These analyses revealed that group membership (moderate-
The cognitive scores for the TBI subgroups (mild, moderate- severe TBI or control) accounted for a significant amount of
severe) and control group were examined to determine the type variance on the three tests, even after controlling for differences in
and extent of the impairments following TBI and whether perfor- age, sex, time postinjury and education. Overall, the mild TBI
mance varied with injury severity (see Table 4). Although none of group performed comparably to the controls on all of the cognitive
the scores for the mild TBI and control groups differed signifi- tests, but those with moderate-severe TBI performed significantly
cantly (Bonferroni corrected p ⬎ .007), the moderate-severe TBI worse than controls on three of the tests; findings that were not
group performed significantly worse than controls (p ⬍ .007) on attributable to age, sex, time postinjury or education.
the delayed Visual Reproduction trial (VR-II; g ⫽ ⫺.70), the
compatible RT task (g ⫽ ⫺.86), and the COWA task (g ⫽ ⫺.75). Relationship Between Cognition and Fractional
However, as noted above, the TBI group was older, had a
Anisotropy
smaller interval between injury and examination (compared to
orthopedic controls), and had a significantly lower level of edu- Next, the relationship between cognitive performance and FA
cation than the control group, which may have contributed to these (in each of five ROIs) was examined in the moderate-severe TBI
differences. Three hierarchical linear regressions were therefore group (see Table S2 in online Supplemental materials; note that the
performed for those cognitive tests that differed between the mild TBI group was not examined because it did not show WM
moderate-severe TBI and control groups (VR-II, compatible RT, alterations or cognitive impairments, relative to controls). Corre-
COWA) in order to determine whether they differed significantly lations were examined for those cognitive tests in which the
after taking the differences in age, sex, time postinjury and edu- moderate-severe TBI group performed significantly worse than
cation into account (see Table 5). Cognitive scores were entered as controls (VR-II, compatible RT task, COWA), after taking into
the dependent variable, with age, sex, time postinjury, education account the effects of age, sex, time postinjury and education.
 WHITE MATTER CHANGES AND COGNITION FOLLOWING TBI 887

Medium or large significant and positive correlations (r ⱖ .3,

Note. TBI ⫽ traumatic brain injury; effect sizes; N ⫽ number of participants; SD ⫽ standard deviation; Games-Howell post-hoc comparisons used to generate p-values; Bonferroni corrected p ⬍
.002
.006
.002
.224
.016

.004
.022
.005
.141
.006
Moderate-severe

p
Bonferroni corrected p ⬍ .003) indicate that better cognitive
vs mild
TBI performance (more accurate responses, faster RTs) was related to

Hedges’ g
greater WM integrity (higher FA). Overall, none of the correla-

⫺.073
⫺1.10

⫺.95
⫺.23
⫺.90

⫺.71
⫺.71
⫺.62
⫺.30
⫺.45
tions were significant (p ⬎ .003).
These relationships were then compared to those seen in the
control group to see whether they were equivalent in people with
and without TBI (see Table S2 in online Supplemental materials).
<.001
.001
<.001
.003
.008

<.001
.003
<.001
.007
.001
Moderate-severe TBI

Again, none of the correlations were significant (p ⬎ .003). There-

p
vs controls

fore, after corrections for age, sex, time postinjury, and education,
no significant relationships were found between cognitive perfor-
Hedges’ g

mance and FA in any of the five ROIs in the moderate-severe TBI

⫺1.44
⫺1.03
⫺1.17
⫺.62
⫺.88

⫺1.12
⫺1.12
⫺1.16
⫺.59
⫺.88
or control groups.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Relationship Between Cognition and Mean Diffusivity

.021
.202
.159
.028
.881

.027
.260
.031
.217
.548
p
Mild TBI vs

The relationship between cognition (VR-II, compatible RT task,

controls

COWA) and MD in the five ROIs was also examined in the

Hedges’ g

moderate-severe TBI group. Medium to strong, significant and

⫺.22
⫺.22
⫺.25
⫺.35
.00

⫺.28
⫺.28
⫺.38
⫺.23
⫺.50

positive correlations (r ⱖ .3, Bonferroni corrected p ⬍ .003) were

again interpreted as indicating that better cognitive performance
Fractional Anisotropy and Mean Diffusivity Values for the Mild TBI, Moderate-Severe TBI, and Control Groups

was related to greater WM integrity (lower MD). As can be seen

in Table S2 in online Supplemental materials, none of the corre-
F(2, 77.54) ⫽ 13.87, p ⫽ .000

F(2, 79.19) ⫽ 10.71, p ⫽ .000

F(2, 77.78) ⫽ 13.03, p ⫽ .000

F(2, 77.43) ⫽ 12.34, p ⫽ .000

F(2, 76.57) ⫽ 8.41, p ⫽ .000

F(2, 89.56) ⫽ 7.10, p ⫽ .001

F(2, 78.67) ⫽ 5.01, p ⫽ .009

F(2, 76.73) ⫽ 6.52, p ⫽ .002

F(2, 90.96) ⫽ 5.12, p ⫽ .008

F(2, 78.85) ⫽ 7.63, p ⫽ .001

lations were significant (p ⬎ .003).

Similarly, there were no significant correlations between cog-
Fractional anisotropy (FA)

nition and MD in the five ROIs in the control group. Therefore,

Welch’s F

Mean diffusivity (MD)

cognitive performance was not related to MD in any of the five

ROIs in the moderate-severe TBI and control groups, after con-
trolling for the effects of age, sex, time postinjury and education.

Discussion
The current study examined WM alterations in the CC, fornix
.04
.04
.04
.08
.02

.03
.03
.02
.12
.02
SD
(N ⫽ 106)
Controls

and SLF, and the relationship between WM integrity and cognitive

functioning seven months after people had sustained a mild, mod-
Mean

.49
.54
.62
.33
.44

.52
.51
.50
.83
.47

erate or severe TBI. These regions were investigated because they

appear to be most affected following a TBI (Wallace et al., 2018a)
and are also strongly related to postinjury cognitive outcomes
.07
.07
.05
.08
.03

.05
.05
.04
.11
.03
SD
severe TBI

(Wallace et al., 2018b). Most existing research has used relatively

Moderate-

(N ⫽ 31)

small samples (e.g., Brandstack, Kurki, Hiekkanen, & Tenovuo,

Mean

2011; Matsushita et al., 2011; Ubukata et al., 2016), with the

.42
.49
.57
.28
.42

.56
.55
.53
.90
.49

resulting low statistical power potentially limiting the reliability

and generalisability of the findings. As noted, appropriate analyses
.05
.05
.04
.09
.02

.04
.04
.03
.14
.02
SD

must be conducted to ensure that studies have the necessary

(N ⫽ 134)
Mild TBI

.01 considered significant (highlighted in bold).

statistical power to replicate expected findings (Shrout & Rodgers,

2018). One exception to this is a recent study that examined a large
Mean

.48
.53
.61
.30
.44

.53
.52
.51
.86
.48

sample, but only investigated moderate to severe TBI and general

outcomes, which were classified as favorable or unfavourable,
rather than specific cognitive outcomes (Castaño Leon et al.,
Superior longitudinal fasciculus

Superior longitudinal fasciculus

2018). Thus, a large-scale investigation was undertaken in order to

Region of interest (ROI)

Corpus callosum—splenium

Corpus callosum—splenium

determine whether, and to what extent, WM changes were related

Corpus callosum—body

Corpus callosum—body
Corpus callosum—genu

Corpus callosum—genu

to cognition following mild, moderate and severe TBI.

Overall, at seven months postinjury, mild TBIs did not result in
significant alterations in WM when compared to controls. Al-
though other DTI studies that used similar postinjury periods have
reported WM damage following mild injuries (for reviews see
Table 2

Fornix

Fornix

Niogi & Mukherjee, 2010; Shenton et al., 2012), a number of

reviews have noted that the findings are inconsistent (Asken et al.,
 888 WALLACE ET AL.

Table 3
Regression Analyses of Fractional Anisotropy and Mean Diffusivity (Moderate to Severe TBI Compared to Controls)

Fractional anisotropy (FA) Mean diffusivity (MD)

Variables Adjusted R2 ⌬R2 Standardized ␤ p Adjusted R2 ⌬R2 Standardized ␤ p

Corpus callosum—genu
Step 1 .099 .099 .080 .080
Age ⫺.372 .001 .331 .004
Sex ⫺.049 .660 ⫺.048 .674
Time since injury ⫺.019 .864 .011 .925
Step 2 .351 .252 .380 .300
Group (TBI, control) ⫺.522 <.001 .567 <.001

Corpus callosum—body
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Step 1 .120 .120 .102 .102

This document is copyrighted by the American Psychological Association or one of its allied publishers.

Age ⫺.373 .001 .325 .005

Sex ⫺.003 .977 ⫺.080 .478
Time since injury .120 .275 ⫺.114 .305
Step 2 .327 .207 .307 .205
Group (TBI, control) ⫺.475 <.001 .474 <.001

Corpus callosum—splenium
Step 1 .016 .016 .004 .004
Age ⫺.222 .061 .118 .316
Sex ⫺.084 .473 ⫺.053 .653
Time since injury .080 .490 ⫺.149 .202
Step 2 .229 .213 .273 .269
Group (TBI, control) ⫺.483 <.001 .539 <.001

Fornix
Step 1 .397 .397 .342 .342
Age ⫺.647 ⬍.001 .563 ⬍.001
Sex ⫺.016 .860 ⫺.116 .229
Time since injury .064 .479 ⫺.099 .298
Step 2 .479 .082 .489 .147
Group (TBI, control) ⫺.304 .001 .401 <.001

Superior longitudinal fasciculus

Step 1 .093 .093 ⫺.008 ⫺.008
Age ⫺.331 .004 .101 .395
Sex .094 .406 ⫺.128 .282
Time since injury ⫺.017 .876 ⫺.023 .846
Step 2 .205 .112 .226 .234
Group (TBI, control) ⫺.359 .001 .505 <.001
Note. TBI ⫽ traumatic brain injury; bold values indicate that group membership accounted for a significant amount of variance after controlling for
variables.

2018; Shenton et al., 2012). Indeed, TBI is increasingly being with poorer cognitive performance, despite reviews finding that 15%
recognized as leading to heterogeneous patterns of injury (Bigler to 30% of people with mild TBI have long-term cognitive deficits
& Stern, 2015; Cristofori & Levin, 2015), raising the possibility (McKee & Daneshvar, 2015; Shenton et al., 2012). However, it is
that the location and extent of WM damage varied between indi- possible that any cognitive problems may have recovered, given that
viduals, but was overlooked at a group level. Alternatively, WM the sample was assessed an average of seven months post injury and
damage in those with mild TBI may have resolved prior to the that most people return to preinjury cognitive levels three to six
study, which was conducted seven months postinjury. For in- months following a mild TBI (Cristofori & Levin, 2015). Those with
stance, partial recovery of WM alterations has been reported in moderate to severe TBIs performed more poorly than controls on a
several longitudinal studies (e.g., between one and nine months test of visual memory (VR-II), attention (compatible RT) and exec-
after injury; Arfanakis et al., 2002; Grossman et al., 2013; Mayer utive functioning (COWA), even after controlling for differences in
et al., 2010). In contrast to those with mild TBI, people with age, sex, interval between injury and examination (i.e., compared to
moderate to severe TBI displayed WM damage, indicated by less orthopedic controls), and education. These findings support previous
directional/anisotropic and a greater magnitude of diffusion (lower research, which has found that approximately 60% of people with
FA, higher MD) in all five regions, relative to controls. moderate TBI and only 15% to 20% of those with severe injuries
In terms of cognition and consistent with some previous research return to preinjury cognitive levels (for a review see Cristofori &
(Kraus et al., 2007; Mayer et al., 2010), mild TBI was not associated Levin, 2015).
 WHITE MATTER CHANGES AND COGNITION FOLLOWING TBI 889

The relationship between cognitive performance and DTI find-

Note. TBI ⫽ traumatic brain injury; SD ⫽ standard deviation; WMSIII ⫽ Wechsler Memory Scale-third edition (Nmild ⫽ 128, Nmoderate-severe ⫽ 31, Ncontrols ⫽ 103); RT ⫽ reaction time; 4-choice
compatible visual RT task (Nmild ⫽ 127, Nmoderate-severe ⫽ 30, Ncontrols ⫽ 102); 4-choice incompatible visual RT task (Nmild ⫽ 126, Nmoderate-severe ⫽ 29, Ncontrols ⫽ 102); COWA ⫽ Controlled Oral
.373
.433
.014
.012
.017
.107
.045
Moderate-severe

p
ings was examined in the moderate to severe TBI group to deter-

vs mild
mine the most promising relationships for further examination.

Hedges’ g

⫺.28
⫺.29
⫺.64
⫺.57
⫺.74
⫺.46
⫺.52
The specific relationships that were examined were between DTI
findings (FA, MD) in the CC, fornix and SLF (i.e., WM changes
were detected in all five ROIs) and one test each of memory,

.025
.036
.021
.002
.003
.008
.003
Moderate-severe

p
processing speed and executive functioning (VR-II, compatible
vs controls

RT, COWA). These tests were chosen because the moderate to

Hedges’ g

severe TBI group performed more poorly in these tests than the
⫺.59
⫺.60
⫺.60
⫺.70
⫺.86
⫺.76
⫺.75
controls, even after accounting for differences in age, sex, the
interval between injury and examination, and education. No cor-
relations were significant. The same associations were then exam-
.065
.032
.982
.562
.304
.114
.147
p

ined in the control group, to determine whether they were equiv-

Mild TBI vs
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

controls

alent in people with and without TBI. Again, no correlations were

Word Association Test (Nmild ⫽ 128, Nmoderate-severe ⫽ 31, Ncontrols ⫽ 103); Bonferroni corrected p ⬍ .007 considered significant (highlighted in bold).
Hedges’ g
This document is copyrighted by the American Psychological Association or one of its allied publishers.

significant, although more were positive in the moderate to severe

⫺.29
⫺.33
⫺.02
⫺.13
⫺.20
⫺.26
⫺.25

TBI group compared to the controls. Thus, it is possible that the

group differences in age, sex, time postinjury and education ac-
counted for any relationship between cognition and WM integrity.
5.011, p ⫽ .009
4.72, p ⫽ .012

4.50, p ⫽ .014
6.53, p ⫽ .002
6.34, p ⫽ .003
5.73, p ⫽ .005
6.47, p ⫽ .002

Limitations and Directions for Future Research

Welch’s F

This study has several limitations that should be considered.

⫽
⫽
⫽
⫽
⫽
⫽
⫽

Group comparisons (i.e., ROI analysis) were used to examine WM

82.52)
79.10)
81.30)
86.52)
74.47)
75.58)
82.18)

changes following TBI, however the location and extent of WM

damage is likely to have varied between individuals (Cristofori &
F(2,
F(2,
F(2,
F(2,
F(2,
F(2,
F(2,

Levin, 2015). Although useful for identifying broad patterns,

group comparisons largely overlook the interindividual variability
3.05
2.95
3.09
3.37
90.86
174.61
12.43

that is a hallmark of TBI (Bigler & Stern, 2015; Cristofori &

SD
Controls

Levin, 2015). This highlights the importance of examining indi-

vidual differences in WM changes following TBI, possibly by
10.63
11.27
10.83
12.55
436.59
637.99
43.11
Mean

comparing individuals to normative databases (Hulkower et al.,

Summary Cognitive Data for Mild TBI, Moderate-Severe TBI, and Control Groups

2013; Niogi & Mukherjee, 2010). Longitudinal studies should also

be completed to examine the trajectories of WM and cognitive
3.34
3.83
3.46
3.08
123.84
219.43
13.14
SD
severe TBI
Moderate-

alterations following TBI. Early and continued examination of

mild TBI would help to determine whether initial WM and cog-
8.77
9.35
8.90
10.23
522.83
779.34
33.68
Mean

nitive changes are present and, if so, when they resolve.

Despite an initial sample size considerably larger than that used
in much existing DTI and cognition research (NTBI ⫽ 165), it is
3.25
3.01
3.02
3.29
83.43
194.91
11.89

worth noting that the primary findings came from the moderate to
SD
Mild TBI

severe TBI group (Nmoderate-severe ⫽ 31), which had a sample size

that was not much greater than those used elsewhere in the liter-
9.70
10.27
10.90
12.10
453.79
687.07
40.07
Mean

ature. In addition, previous research has found that the extent of

WM damage differs for moderate and severe injuries (Castaño
WMSIII - Visual Reproduction immediate (VR-I)

Leon et al., 2018), but it was not possible to examine the three
WMSIII - Visual Reproduction delayed (VR-II)
WMSIII - Logical Memory immediate (LM-I)

injury categories separately in the current study because very few

WMSIII - Logical Memory delayed (LM-II)

participants sustained moderate (9.1%) or severe (9.7%) TBIs. The

two groups were necessarily collapsed into one, with all analyses
4-choice incompatible visual RT task

based on a combined moderate to severe TBI group. Additional

4-choice compatible visual RT task

large-scale studies are needed in which mild, moderate, and severe

Cognitive test

injuries are more evenly represented.

Multiple previous concussions may also affect WM and/or
cognitive performance and should have been included as a cova-
riate in the analyses, however this information was not collected
from participants. A head injury exposure guided interview was
not administered to participants, thus it is possible that previous
head injuries were not reported accurately by participants. Further,
Table 4

COWA

the underlying cause of TBI and orthopedic injuries was not

controlled for in the analyses, but may reflect lifestyle differences
 890 WALLACE ET AL.

Table 5
Regression Analyses of the Cognitive Tests (Moderate to Severe TBI Compared to Controls)

Standardized Standardized
Variables Adjusted R2 ⌬R2 ␤ p Adjusted R2 ⌬R2 ␤ p

WMSIII - Logical Memory 4-choice compatible visual RT task

immediate
Step 1 .307 .307 .349 .349
Age .122 .218 .548 ⬍.001
Sex ⫺.218 .035 ⫺.090 .368
Education .459 ⬍.001 ⫺.312 .002
Time since injury .083 .392 .052 .581
Step 2 .321 .014 .423 .074
Group (TBI, control) ⫺.160 .120 .295 .002
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

WMSIII - Logical Memory delayed 4-choice incompatible visual RT

task
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Step 1 .245 .245 .383 .383

Age .153 .139 .573 ⬍.001
Sex ⫺.273 .012 ⫺.064 .511
Education .343 .001 ⫺.324 .001
Time since injury .044 .661 .031 .738
Step 2 .260 .015 .459 .076
Group (TBI, control) ⫺.166 .122 .299 .002
WMSIII - Visual Reproduction Controlled Oral Word Association
immediate Test
Step 1 .130 .130 .097 .097
Age ⫺.285 .011 .187 .098
Sex ⫺.161 .160 .019 .870
Education .259 .023 .230 .047
Time since injury .100 .360 ⫺.224 .046
Step 2 .144 .014 .186 .089
Group (TBI, control) ⫺.170 .140 ⫺.329 .004
WMSIII - Visual Reproduction
delayed
Step 1 .104 .104
Age ⫺.222 .050
Sex ⫺.176 .131
Education .255 .027
Time since injury .096 .386
Step 2 .186 .082
Group (TBI, control) ⫺.317 .006
Note. TBI ⫽ traumatic brain injury; WMSIII ⫽ Wechsler Memory Scale-third edition; RT ⫽ reaction time; COWA ⫽ Controlled Oral Word Association
Test; bold values indicate that group membership accounted for a significant amount of variance after controlling for variables.

between the groups. In addition, the time between injury and MRI a voxel contains crossing fibers and/or more than one fiber tract
varied considerably, although a recent meta-analysis showed that (Jeurissen, Leemans, Tournier, Jones, & Sijbers, 2013; Raffelt et
DTI findings did not differ depending on postinjury interval (Wal- al., 2015). ROI analysis has proven useful in studies with specific
lace et al., 2018a). Finally, the DTI measures (e.g., FA, MD) hypotheses about where differences will be found (a priori hypoth-
obtained from the fornix may be inaccurate because the fornix is a eses), but recently developed methods—such as fixel-based anal-
very thin structure, surrounded by cerebrospinal fluid. This struc- ysis (FBA; Raffelt et al., 2015)— can evaluate individual fiber
ture may therefore suffer from partial volume effects, especially populations in regions where fibers cross and may better determine
where there was atrophy (Jones & Cercignani, 2010). Partial the nature and extent of WM changes following TBI.
volume effects can occur when voxels contain more than one type The current findings suggest that moderate to severe TBI leads
of tissue (e.g., WM, gray matter, cerebrospinal fluid), each of to WM and cognitive changes, but no association was found
which have different diffusion properties (Vos, Jones, Viergever, between the two. However, participants were only examined at one
& Leemans, 2011). Several techniques have been developed to time-point (i.e., 7 months postinjury). Large-scale, longitudinal
help mitigate this problem, including the suppression of cerebro- studies are now needed to determine whether early DTI findings
spinal fluid contamination either in the acquisition or analysis of for the CC, fornix and SLF predict long-term cognitive outcomes
diffusion data (Jones & Cercignani, 2010). (e.g., years postinjury). It is possible that early DTI may help to
Although widely used, ROI analysis calculates measures (e.g., identify individuals who are more likely to experience long-term
FA, MD) that are averaged across all of the fiber tracts that are cognitive problems in these domains, potentially allowing for early
present within each voxel and, therefore, may be inaccurate when intervention and rehabilitation (e.g., cognitive and/or skills train-
 WHITE MATTER CHANGES AND COGNITION FOLLOWING TBI 891

ing, group therapy) in order to optimize outcomes and decrease the White matter integrity and its relationship with outcome. Journal of
levels of TBI-related disability. Neurotrauma, 35, 2365–2376. http://dx.doi.org/10.1089/neu.2018.5691
Chiaravalloti, N. D., & DeLuca, J. (2008). Cognitive impairment in mul-
tiple sclerosis. The Lancet Neurology, 7, 1139 –1151. http://dx.doi.org/
Conclusions 10.1016/S1474-4422(08)70259-X
Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155–159.
This study found that moderate to severe TBI leads to WM
http://dx.doi.org/10.1037/0033-2909.112.1.155
damage in the CC, fornix and SLF—as reflected in less directional/ Cristofori, I., & Levin, H. S. (2015). Traumatic brain injury and cognition.
anisotropic and a greater magnitude of diffusion (higher FA, lower Handbook of Clinical Neurology, 128, 579 – 611. http://dx.doi.org/10
MD)—and impairments to visual memory, attention and executive .1016/B978-0-444-63521-1.00037-6
functioning, when compared to healthy and orthopedic controls. In Dean, P. J. A., Sato, J. R., Vieira, G., McNamara, A., & Sterr, A. (2015).
contrast, mild TBI was not associated with WM alterations or Multimodal imaging of mild traumatic brain injury and persistent post-
cognitive impairments seven months postinjury, suggesting a lack, concussion syndrome. Brain and Behavior, 5, 45– 61. http://dx.doi.org/
or potential resolution, of WM damage and/or cognitive impair- 10.1002/brb3.292
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

ments within this time frame. An examination of the relationship Douglas, D. B., Muldermans, J. L., & Wintermark, M. (2018). Neuroim-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

between cognition and WM integrity in the moderate to severe TBI aging of brain trauma. Current Opinion in Neurology, 31, 362–370.
group revealed no significant associations, suggesting that any http://dx.doi.org/10.1097/WCO.0000000000000567
potential relationship may have been accounted for by group Geary, E. K., Kraus, M. F., Pliskin, N. H., & Little, D. M. (2010). Verbal
differences in age, sex, interval between injury and examination learning differences in chronic mild traumatic brain injury. Journal of
(i.e., relative to orthopedic but not healthy controls), and/or edu- the International Neuropsychological Society, 16, 506 –516. http://dx
.doi.org/10.1017/S135561771000010X
cation. Large-scale, longitudinal studies are now needed to deter-
Grossman, E. J., Jensen, J. H., Babb, J. S., Chen, Q., Tabesh, A., Fiere-
mine whether early examination of the CC, fornix and SLF can
mans, E., . . . Grossman, R. I. (2013). Cognitive impairment in mild
help to identify people who are most likely to exhibit long-term traumatic brain injury: A longitudinal diffusional kurtosis and perfusion
cognitive problems. imaging study. AJNR. American Journal of Neuroradiology, 34, 951–
957, S1–S3. http://dx.doi.org/10.3174/ajnr.A3358
Holland, B. S., & Copenhaver, M. D. (1988). Improved Bonferroni-Type
References
Multiple Testing Procedures. Psychological Bulletin, 104, 145–149.
Amyot, F., Arciniegas, D. B., Brazaitis, M. P., Curley, K. C., Diaz- http://dx.doi.org/10.1037/0033-2909.104.1.145
Arrastia, R., Gandjbakhche, A., . . . Stocker, D. (2015). A review of the Huisman, T. A. (2010). Diffusion-weighted and diffusion tensor imaging
effectiveness of neuroimaging modalities for the detection of traumatic of the brain, made easy. Cancer Imaging, 10 Spec no A, S163–S171.
brain injury. Journal of Neurotrauma, 32, 1693–1721. http://dx.doi.org/ http://dx.doi.org/10.1102/1470-7330.2010.9023
10.1089/neu.2013.3306 Huisman, T. A., Schwamm, L. H., Schaefer, P. W., Koroshetz, W. J.,
Andersson, J. L. R., & Sotiropoulos, S. N. (2016). An integrated approach Shetty-Alva, N., Ozsunar, Y., . . . Sorensen, A. G. (2004). Diffusion
to correction for off-resonance effects and subject movement in diffu- tensor imaging as potential biomarker of white matter injury in diffuse
sion MR imaging. NeuroImage, 125, 1063–1078. http://dx.doi.org/10 axonal injury. AJNR. American Journal of Neuroradiology, 25, 370 –
.1016/j.neuroimage.2015.10.019 376.
Arfanakis, K., Haughton, V. M., Carew, J. D., Rogers, B. P., Dempsey, Hulkower, M. B., Poliak, D. B., Rosenbaum, S. B., Zimmerman, M. E., &
R. J., & Meyerand, M. E. (2002). Diffusion tensor MR imaging in Lipton, M. L. (2013). A decade of DTI in traumatic brain injury: 10
diffuse axonal injury. AJNR: American Journal of Neuroradiology, 23, years and 100 articles later. AJNR. American Journal of Neuroradiology,
794 – 802. 34, 2064 –2074. http://dx.doi.org/10.3174/ajnr.A3395
Asken, B. M., DeKosky, S. T., Clugston, J. R., Jaffee, M. S., & Bauer, Hutchinson, E. B., Schwerin, S. C., Avram, A. V., Juliano, S. L., &
R. M. (2018). Diffusion tensor imaging (DTI) findings in adult civilian,
Pierpaoli, C. (2018). Diffusion MRI and the detection of alterations
military, and sport-related mild traumatic brain injury (mTBI): A sys-
following traumatic brain injury. Journal of Neuroscience Research, 96,
tematic critical review. Brain Imaging and Behavior, 12, 585– 612.
612– 625. http://dx.doi.org/10.1002/jnr.24065
http://dx.doi.org/10.1007/s11682-017-9708-9
Inglese, M., Makani, S., Johnson, G., Cohen, B. A., Silver, J. A., Gonen,
Avants, B. B., Tustison, N. J., Song, G., Cook, P. A., Klein, A., & Gee,
O., & Grossman, R. I. (2005). Diffuse axonal injury in mild traumatic
J. C. (2011). A reproducible evaluation of ANTs similarity metric
brain injury: A diffusion tensor imaging study. Journal of Neurosurgery,
performance in brain image registration. NeuroImage, 54, 2033–2044.
103, 298 –303. http://dx.doi.org/10.3171/jns.2005.103.2.0298
http://dx.doi.org/10.1016/j.neuroimage.2010.09.025
Bazarian, J. J., Zhong, J., Blyth, B., Zhu, T., Kavcic, V., & Peterson, D. Jang, S. H., Kim, S. H., Kim, O. R., Byun, W. M., Kim, M. S., Seo, J. P.,
(2007). Diffusion tensor imaging detects clinically important axonal & Chang, M. C. (2013). Cingulum injury in patients with diffuse axonal
damage after mild traumatic brain injury: A pilot study. Journal of injury: A diffusion tensor imaging study. Neuroscience Letters, 543,
Neurotrauma, 24, 1447–1459. http://dx.doi.org/10.1089/neu.2007.0241 47–51. http://dx.doi.org/10.1016/j.neulet.2013.02.058
Bigler, E. D., & Stern, Y. (2015). Traumatic brain injury and reserve. Jenkinson, M., Beckmann, C. F., Behrens, T. E., Woolrich, M. W., &
Handbook of Clinical Neurology, 128, 691–710. http://dx.doi.org/10 Smith, S. M. (2012). FSL. NeuroImage, 62, 782–790. http://dx.doi.org/
.1016/B978-0-444-63521-1.00043-1 10.1016/j.neuroimage.2011.09.015
Brandstack, N., Kurki, T., Hiekkanen, H., & Tenovuo, O. (2011). Diffu- Jeurissen, B., Leemans, A., Tournier, J. D., Jones, D. K., & Sijbers, J.
sivity of normal-appearing tissue in acute traumatic brain injury. Clinical (2013). Investigating the prevalence of complex fiber configurations in
Neuroradiology, 21, 75– 82. http://dx.doi.org/10.1007/s00062-011- white matter tissue with diffusion magnetic resonance imaging. Human
0058-5 Brain Mapping, 34, 2747–2766. http://dx.doi.org/10.1002/hbm.22099
Castaño Leon, A. M., Cicuendez, M., Navarro, B., Munarriz, P. M., Jones, D. K., & Cercignani, M. (2010). Twenty-five pitfalls in the analysis
Cepeda, S., Paredes, I., . . . Lagares, A. (2018). What can we learn from of diffusion MRI data. NMR in Biomedicine, 23, 803– 820. http://dx.doi
diffusion tensor imaging from a large traumatic brain injury cohort? .org/10.1002/nbm.1543
 892 WALLACE ET AL.

Jones, D. K., Horsfield, M. A., & Simmons, A. (1999). Optimal strategies Physical Medicine and Rehabilitation, 98, 2308 –2319. http://dx.doi.org/
for measuring diffusion in anisotropic systems by magnetic resonance 10.1016/j.apmr.2017.03.019
imaging. Magnetic Resonance in Medicine, 42, 515–525. http://dx.doi Pannek, K., Fripp, J., George, J., Boyd, R., Colditz, P., & Rose, S. (2017).
.org/10.1002/(SICI)1522-2594(199909)42:3⬍515::AID-MRM14⬎3.0 Automatic detection of volumes affected by subvolume movement. Paper
.CO;2-Q presented at the ISMRM 25th Annual Meeting & Exhibition, Honolulu,
Koerte, I. K., Hufschmidt, J., Muehlmann, M., Lin, A. P., & Shenton, M. E. HI.
(2016). Advanced neuroimaging of mild traumatic brain injury. In D. Ponsford, J., & Kinsella, G. (1992). Attentional deficits following closed-
Laskowitz & G. Grant (Eds.), Translational research in traumatic brain head injury. Journal of Clinical and Experimental Neuropsychology, 14,
injury (pp. 277–297). Boca Raton, FL: CRC Press. 822– 838. http://dx.doi.org/10.1080/01688639208402865
Kraus, M. F., Susmaras, T., Caughlin, B. P., Walker, C. J., Sweeney, J. A., Raffelt, D. A., Smith, R. E., Ridgway, G. R., Tournier, J. D., Vaughan,
& Little, D. M. (2007). White matter integrity and cognition in chronic D. N., Rose, S., . . . Connelly, A. (2015). Connectivity-based fixel
traumatic brain injury: A diffusion tensor imaging study. Brain: A enhancement: Whole-brain statistical analysis of diffusion MRI mea-
Journal of Neurology, 130, 2508 –2519. http://dx.doi.org/10.1093/brain/
sures in the presence of crossing fibres. NeuroImage, 117, 40 –55.
awm216
http://dx.doi.org/10.1016/j.neuroimage.2015.05.039
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Lumley, T., Diehr, P., Emerson, S., & Chen, L. (2002). The importance of
Rosenthal, J. A. (1996). Qualitative descriptors of strength of association
This document is copyrighted by the American Psychological Association or one of its allied publishers.

the normality assumption in large public health data sets. Annual Review
and effect size. Journal of Social Service Research, 21, 37–59. http://
of Public Health, 23, 151–169. http://dx.doi.org/10.1146/annurev
dx.doi.org/10.1300/J079v21n04_02
.publhealth.23.100901.140546
Rugg-Gunn, F. J., Symms, M. R., Barker, G. J., Greenwood, R., & Duncan,
Mathias, J. L., Beall, J. A., & Bigler, E. D. (2004). Neuropsychological and
J. S. (2001). Diffusion imaging shows abnormalities after blunt head
information processing deficits following mild traumatic brain injury.
Journal of the International Neuropsychological Society, 10, 286 –297. trauma when conventional magnetic resonance imaging is normal. Jour-
http://dx.doi.org/10.1017/S1355617704102117 nal of Neurology, Neurosurgery, and Psychiatry, 70, 530 –533. http://
Mathias, J. L., Bigler, E. D., Jones, N. R., Bowden, S. C., Barrett- dx.doi.org/10.1136/jnnp.70.4.530
Woodbridge, M., Brown, G. C., & Taylor, D. J. (2004). Neuropsycho- Shenton, M. E., Hamoda, H. M., Schneiderman, J. S., Bouix, S., Pasternak,
logical and information processing performance and its relationship to O., Rathi, Y., . . . Zafonte, R. (2012). A review of magnetic resonance
white matter changes following moderate and severe traumatic brain imaging and diffusion tensor imaging findings in mild traumatic brain
injury: A preliminary study. Applied Neuropsychology, 11, 134 –152. injury. Brain Imaging and Behavior, 6, 137–192. http://dx.doi.org/10
http://dx.doi.org/10.1207/s15324826an1103_2 .1007/s11682-012-9156-5
Mathias, J. L., Dennington, V., Bowden, S. C., & Bigler, E. D. (2013). Sherer, M., Sander, A. M., Maestas, K. L., Pastorek, N. J., Nick, T. G., &
Community versus orthopaedic controls in traumatic brain injury re- Li, J. (2015). Accuracy of self-reported length of coma and posttrau-
search: How comparable are they? Brain Injury, 27(7– 8), 887– 895. matic amnesia in persons with medically verified traumatic brain injury.
http://dx.doi.org/10.3109/02699052.2013.793398 Archives of Physical Medicine and Rehabilitation, 96, 652– 658. http://
Matsushita, M., Hosoda, K., Naitoh, Y., Yamashita, H., & Kohmura, E. dx.doi.org/10.1016/j.apmr.2014.10.024
(2011). Utility of diffusion tensor imaging in the acute stage of mild to Shrout, P. E., & Rodgers, J. L. (2018). Psychology, science, and knowledge
moderate traumatic brain injury for detecting white matter lesions and construction: Broadening perspectives from the replication crisis. An-
predicting long-term cognitive function in adults. Journal of Neurosur- nual Review of Psychology, 69, 487–510.
gery, 115, 130 –139. http://dx.doi.org/10.3171/2011.2.JNS101547 Spreen, O., & Straus, E. (1998). Compendium of neuropsychological tests.
Mayer, A. R., Ling, J., Mannell, M. V., Gasparovic, C., Phillips, J. P., New York, NY: Oxford University Press.
Doezema, D., . . . Yeo, R. A. (2010). A prospective diffusion tensor Strauss, S., Hulkower, M., Gulko, E., Zampolin, R. L., Gutman, D.,
imaging study in mild traumatic brain injury. Neurology, 74, 643– 650. Chitkara, M., . . . Lipton, M. L. (2015). Current clinical applications and
http://dx.doi.org/10.1212/WNL.0b013e3181d0ccdd future potential of diffusion tensor imaging in traumatic brain injury.
Mckee, A. C., & Daneshvar, D. H. (2015). The neuropathology of trau- Topics in Magnetic Resonance Imaging, 24, 353–362. http://dx.doi.org/
matic brain injury. Handbook of Clinical Neurology, 127, 45– 66. http:// 10.1097/RMR.0000000000000071
dx.doi.org/10.1016/B978-0-444-52892-6.00004-0 Suri, A. K., & Lipton, M. L. (2018). Neuroimaging of brain trauma in
Menon, D. K., Schwab, K., Wright, D. W., Maas, A. I., & the Demograph-
sports. In B. Hainline & R. A. Stern (Eds.), Handbook of clinical
ics and Clinical Assessment Working Group of the International and
neurology (Vol. 158, pp. 205–216). New York, NY: Elsevier.
Interagency Initiative toward Common Data Elements for Research on
Tomarken, A. J., & Serlin, R. C. (1986). Comparison of anova alternatives
Traumatic Brain Injury and Psychological Health. (2010). Position state-
under variance heterogeneity and specific noncentrality structures. Psy-
ment: Definition of traumatic brain injury. Archives of Physical Medi-
chological Bulletin, 99, 90 –99. http://dx.doi.org/10.1037/0033-2909.99
cine and Rehabilitation, 91, 1637–1640. http://dx.doi.org/10.1016/j
.1.90
.apmr.2010.05.017
Toothaker, L. E. (1993). Multiple comparison procedures. London, Eng-
Miller, G. A., & Chapman, J. P. (2001). Misunderstanding analysis of
covariance. Journal of Abnormal Psychology, 110, 40 – 48. http://dx.doi land: Sage. http://dx.doi.org/10.4135/9781412985178
.org/10.1037/0021-843X.110.1.40 Tukey, J. W. (1977). Exploratory data analysis (Vol. 2). Boston, MA:
Mukherjee, P., Berman, J. I., Chung, S. W., Hess, C. P., & Henry, R. G. Addison Wesley.
(2008). Diffusion tensor MR imaging and fiber tractography: Theoretic Ubukata, S., Ueda, K., Sugihara, G., Yassin, W., Aso, T., Fukuyama, H.,
underpinnings. AJNR. American Journal of Neuroradiology, 29, 632– & Murai, T. (2016). Corpus callosum pathology as a potential surrogate
641. http://dx.doi.org/10.3174/ajnr.A1051 marker of cognitive impairment in diffuse axonal injury. The Journal of
Niogi, S. N., & Mukherjee, P. (2010). Diffusion tensor imaging of mild Neuropsychiatry and Clinical Neurosciences, 28, 97–103. http://dx.doi
traumatic brain injury. The Journal of Head Trauma Rehabilitation, 25, .org/10.1176/appi.neuropsych.15070159
241–255. http://dx.doi.org/10.1097/HTR.0b013e3181e52c2a Voelbel, G. T., Genova, H. M., Chiaravalotti, N. D., & Hoptman, M. J.
Oehr, L., & Anderson, J. (2017). Diffusion-tensor imaging findings and (2012). Diffusion tensor imaging of traumatic brain injury review:
cognitive function following hospitalized mixed-mechanism mild trau- Implications for neurorehabilitation. NeuroRehabilitation, 31, 281–293.
matic brain injury: A systematic review and meta-analysis. Archives of http://dx.doi.org/10.3233/NRE-2012-0796
 WHITE MATTER CHANGES AND COGNITION FOLLOWING TBI 893

Vos, S. B., Jones, D. K., Viergever, M. A., & Leemans, A. (2011). Partial Wilde, E. A., Ware, A. L., Li, X., Wu, T. C., McCauley, S. R., Barnes, A.,
volume effect as a hidden covariate in DTI analyses. NeuroImage, 55, . . . Levin, H. S. (2019). Orthopedic injured versus uninjured comparison
1566 –1576. http://dx.doi.org/10.1016/j.neuroimage.2011.01.048 groups for neuroimaging research in mild traumatic brain injury. Journal
Wallace, E. J., Mathias, J. L., & Ward, L. (2018a). Diffusion tensor of Neurotrauma, 36, 239 –249. http://dx.doi.org/10.1089/neu.2017.5513
imaging changes following mild, moderate and severe adult traumatic Zhu, Y., Li, Z., Bai, L., Tao, Y., Sun, C., Li, M., . . . Zhang, M. (2014).
brain injury: A meta-analysis. Brain Imaging and Behavior, 12, 1607– Loss of microstructural integrity in the limbic-subcortical networks for
1621. http://dx.doi.org/10.1007/s11682-018-9823-2 acute symptomatic traumatic brain injury. BioMed Research Interna-
Wallace, E. J., Mathias, J. L., & Ward, L. (2018b). The relationship tional, 2014, 548392. http://dx.doi.org/10.1155/2014/548392
between diffusion tensor imaging findings and cognitive outcomes fol-
lowing adult traumatic brain injury: A meta-analysis. Neuroscience and
Biobehavioral Reviews, 92, 93–103. http://dx.doi.org/10.1016/j
.neubiorev.2018.05.023
Wechsler, D. (1997). Wechsler Memory Scale (3rd ed.). San Antonio, TX:
Psychological Corporation.
Received October 23, 2019
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Welch, B. L. (1951). On the comparison of several mean values: An

alternative approach. Biometrika, 38, 330 –336. http://dx.doi.org/10 Revision received May 29, 2020
This document is copyrighted by the American Psychological Association or one of its allied publishers.

.1093/biomet/38.3-4.330 Accepted June 4, 2020 䡲