Brain Imaging and Behavior DOI 10.1007/s11682-015-9371-y
MILITARY/VETERAN TBI
Personality and neuroimaging measures differentiate PTSD from mTBI in veterans Nicholas D. Davenport & Kelvin O. Lim & Scott R. Sponheim
# Springer Science+Business Media New York (outside the USA) 2015
Abstract Mild traumatic brain injury (mTBI) is common among recent veterans and often is associated with chronic post-concussive symptoms (PCS). Elevated PCS may also be a consequence of post-traumatic stress disorder (PTSD) which shares symptoms with PCS. Identification of personality, biological, and psychopathology factors that contribute to the relationship between mTBI and PCS could help isolate the sources of chronic post concussive syndrome in veterans. Clinician rated diagnoses (PTSD, Major Depression, Alcohol Dependence), personality characteristics (Multidimensional Personality Questionnaire [MPQ] subscales), white matter brain imaging measures (Mean Diffusivity, Generalized Fractional Anisotropy), and diagnoses of mTBI were collected from 125 American military veterans of Iraq or Afghanistan. Linear and logistic regression models were tested to determine contributions to PCS and whether there were similar contributors to PTSD and mTBI. PCS score was associated with personality characteristics of high Stress Reaction and Traditionalism and low Control as well as mTBI. A diagnosis of PTSD was associated with low Social Closeness, PCS, Alcohol Dependence, and abnormal white matter mean diffusivity. Diagnosis of mTBI was associated with fewer white matter mean diffusivity abnormalities, PCS, and number of deployments. As commonly observed clinically, both PTSD and mTBI were associated with higher rates of PCS, though the contribution of PTSD appears to be secondary to personality traits, particularly Stress Reaction. Furthermore, the observation of factors that are uniquely associated with Blast mTBI N. D. Davenport (*) : K. O. Lim : S. R. Sponheim Minneapolis Veterans Affairs Health Care System, VA Medical Center (B68-2), 1 Veterans Drive, Minneapolis, MN 55417, USA e-mail: [email protected] N. D. Davenport : K. O. Lim : S. R. Sponheim Department of Psychiatry, University of Minnesota, Minneapolis, MN, USA
(number of deployments) or with PTSD (Lifetime Alcohol Dependence and low Social Closeness), as well as a factor (region of abnormal MD) that had opposite effects on the likelihood of each diagnosis, indicates that the complex relationships between personality, psychopathology, and nature of mTBI need to be considered when interpreting chronic postconcussive symptoms. Keywords PTSD . mTBI . White matter . Personality . Military
Introduction Mild traumatic brain injury (mTBI) is among the most common injuries among American service members deployed to Iraq and Afghanistan (Taber et al. 2006; Warden 2006). In most cases, the neurological and cognitive symptoms associated with mTBI resolve within days to weeks (Karr et al. 2014), but approximately a third of soldiers continue to report substantial postconcussive symptoms (PCS), including headaches, memory problems, irritability, and insomnia, that persist beyond 5 months (Schneiderman et al. 2008). These symptoms are also commonly reported by non-deployed soldiers without mTBI (Miller et al. 2013) and overlap substantially with other conditions common in veterans, particularly post-traumatic stress disorder (PTSD), complicating their use in the identification of veterans experiencing chronic effects of mTBI. The potential confounding effect of PTSD on chronic PCS has been addressed in several recent studies. While some have concluded that PTSD symptoms explain a substantial portion of the relationship between mTBI and PCS (Hoge et al. 2008; Polusny et al. 2011; Schneiderman et al. 2008), others have
Brain Imaging and Behavior
determined that mTBI and PTSD contribute to PCS independently (Brenner et al. 2010). Given that PTSD is more common among veterans reporting mTBI than among those with other types of injuries (Hoge et al. 2008; Yurgil et al. 2014), further delineation of the potentially complicated relationship among mTBI, PTSD, and persistent PCS is of high clinical importance. In particular, the identification of additional biological and psychological factors specific to mTBI would be clinically useful in resolving diagnostic ambiguity in cases of soldiers exposed to multiple forms of trauma. Likewise, the identification of factors that relate to PCS beyond the effect of mTBI may help to characterize the subset of veterans at highest risk for experiencing chronic effects of mTBI. Biological factors that establish an underlying objective basis for symptom expression would be especially useful in resolving confounding effects of comorbid conditions. Specifically, evidence of structural brain abnormalities related to PCS expression beyond the static effect of mTBI itself could be interpreted as reflecting a biological basis for variability in long-term outcomes of mTBI. Based on evidence of diffuse axonal injury associated with more severe forms of TBI (Hammoud and Wasserman 2002), diffusion tensor imaging (DTI) has been used in recent studies of mTBI to determine whether measures of white matter integrity (WMI) have utility as biomarkers. In healthy white matter, water is constrained by myelin and cytoskeletal elements and thus able to diffuse more freely in the direction of the axon fibers than in perpendicular directions, creating a highly non-spherical, or anisotropic, pattern of diffusion (Beaulieu 2002). Therefore, WMI is typically characterized by high fractional anisotropy (FA) and low mean (i.e., overall) diffusivity (MD). Injury or disease is presumed to compromise white matter structures, leading to increased water diffusivity (i.e., higher MD) and a less directionally oriented pattern (i.e., lower FA). Investigations of civilian mTBI typically reveal a pattern of WMI changes mirroring the clinical course, namely acute disruption followed by normalization over a period of weeks or months and chronic disruption in only a subset of patients (Kou et al. 2013; Niogi and Mukherjee 2010; Yuh et al. 2014). However, because civilian mTBI is typically studied based on emergency room medical records, its characterization and diagnosis is less reliant on retrospective self-report than military mTBI. While several DTI studies in veterans have provided evidence of reduced WMI associated with mTBI (Davenport et al. 2012; Jorge et al. 2012; MacDonald et al. 2011; Morey et al. 2013; Yeh et al. 2014; Yuh et al. 2014), the potential effects of PTSD and PCS have not been meaningfully considered. While PTSD has received less attention in DTI studies, there are reports of decreased FA associated with PTSD in civilian (Fani et al. 2012; Kim et al. 2005; Kim et al. 2006) and military (Schuff et al. 2011) samples and one report of increased FA in left anterior cingulum (Abe et al. 2006). However, individuals with suspected concussion were explicitly excluded
from these studies. In a sample of veterans in which both mTBI and PTSD were represented, we previously reported higher Generalized FA (GFA), a measure analogous to FA that accounts for WMI across multiple fiber orientations within a voxel (e.g., crossing or diverging fibers), associated with a lifetime history of PTSD and relatively few abnormalities associated with mTBI (Davenport et al. 2015), suggesting that the chronic phase of mTBI may not involve systematic disruptions in WMI after accounting for PTSD. Moreover, given that GFA and FA are identical in regions modeled adequately by a single tensor, the observation of differences in GFA but not in FA demonstrates the additional sensitivity to WMI disruptions afforded by consideration of multiple fiber orientations. It is also possible that some of the variability in postconcussive symptom expression is due to psychological factors (e.g., psychopathology, personality traits) that reflect overlapping conditions (e.g., PTSD, depression) or differences in subjective experience and reporting of PCS. To date, most investigations of psychological factors contributing to PCS report have focused on dimensional measures of anxiety and depression (Clarke et al. 2012; Wood et al. 2014) or symptoms of PTSD and Major Depressive Disorder (Garden et al. 2010; Polusny et al. 2011; Wood et al. 2014). Most of these studies find strong associations between measures of psychopathology and PCS, even in non-clinical samples (Garden et al. 2010). The role of broader personality traits has been studied less fully, though the potential for neuroticism and negative affect to be associated with increased subjective perception and reporting of bodily sensations has been recognized (Fayol et al. 2009; Wood et al. 2014). One study looking at this relationship found that mTBI patients had higher levels of neuroticism than healthy controls, though neuroticism was elevated to a similar degree among patients with non-head injuries, indicating non-specificity to mTBI (Clarke et al. 2012). Given that PTSD is also associated with neuroticism (James et al. 2013; Kotov et al. 2010) and negative emotionality (Miller et al. 2003), it is likely that personality traits play an important role in its relationship with mTBI and PCS. The primary purpose of this investigation was to better characterize the potential contributions of white matter integrity, personality, and psychopathology to the relationship between mTBI and reported PCS within a sample of veterans recently returned from combat deployments to Iraq or Afghanistan . We adopted a broad approach that included objective biological measures of brain structure, self-report questionnaire data, and clinician-rated diagnoses and symptoms in order to more completely test the relationships of personality, brain abnormalities, and mental disorders to chronic postconcussive symptoms. Specifically, we tested the hypothesis that mTBI would remain a significant predictor of PCS after accounting for personality traits and the presence of PTSD as compared to the alternative hypothesis that psychological
Brain Imaging and Behavior
measures explain a substantial portion of the variability in PCS reporting, both of which have been supported in prior literature. As a corollary to this hypothesis, we tested whether factors related to PCS beyond the contributions of mTBI and PTSD were more indicative of variability in reporting (i.e., personality traits) or in underlying brain structure (i.e., WMI). A secondary purpose was to identify measures that independently relate to mTBI and PTSD diagnoses, namely personality traits, WMI indices, and history of mental health conditions that provide unique diagnostic information beyond the contribution attributable to PCS. Based on prior literature, we hypothesized that PTSD would be associated with psychological variables, especially Alcohol Dependence and negative emotionality traits (e.g., Stress Reaction). Likewise, we hypothesized that mTBI would be associated with biological measures, specifically reduced white matter integrity.
Methods and materials Participants Participants consisted of 133 veterans of Operations Enduring and Iraqi Freedom described in a prior report (Davenport et al. 2015) who had reported either traumatic combat experiences or exposure to explosive blasts during their most recent deployment (2–5 years prior to participation) on a research or clinical instrument. All reports of mTBI and clinical symptoms were independently assessed during study participation. Exclusionary criteria included native language other than English, current or predeployment unstable medical condition that would reasonably be expected to significantly affect brain function (e.g., anoxic episode >10 s, stroke, seizures, multiple sclerosis, etc.), uncorrected visual problems or hearing loss, moderate or severe TBI not due to blast, any predeployment Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV-TR; American Psychiatric Association 2000) Axis I psychotic or mood disorder, pre-deployment PTSD according to the Clinician Administered PTSD Scale (CAPS; Blake et al. 1995), current or past substance dependence other than nicotine or alcohol, and contraindications to MRI (e.g., metallic implants, shrapnel, claustrophobia). Participants completed an informed consent process that included complete description of the study, and participants were provided monetary compensation for participation after each study procedure. The study protocol was reviewed and approved by the University of Minnesota and Minneapolis Veterans Affairs Medical Center Institutional Review Boards and the U.S. Army Medical Research and Materiel Command (USAMRMC).
Clinical assessment The presence of mTBI was determined by doctoral-level neuropsychologists based on consensus review of information obtained from the Minnesota Blast Exposure Screening Tool (MN-BEST; Nelson et al. 2011), including detailed descriptions of events and reports of altered consciousness (e.g., confusion, disorientation), loss of consciousness (LOC) less than 30 min, post-traumatic amnesia (PTA) up to 24 h, and neurological symptoms (e.g., headache, tinnitus, nausea, sensitivity to light or noise) immediately after each event. Blast-related injuries were defined as those in which the individual felt a blast wave and attributed the resultant concussion to its effects, though physical impacts (e.g., hitting head on ground, being hit by debris) were allowed to be present provided that a blast wave was the prominent source of injury. The three most significant potential blast-related and impact-related TBI events were considered. One participant reported LOC lasting 30–60 min but was included because the duration of LOC could not be verified and, therefore, mild TBI could not be ruled out. Impact-related injuries generally occurred in a civilian context several years prior to military deployment (e.g., childhood accidents, participation in youth contact sports), whereas blast-related injuries occurred during military deployment and were most proximal to the report of post-concussive symptoms. Therefore, Blast mTBI (i.e., recent deploymentrelated injuries involving exposure to explosive blasts) was considered the index event most likely related to current PCS, and Impact mTBI (i.e., history of civilian non-blast injuries) was considered as an independent contributing factor. Blast exposure was determined by asking participants whether they had ever been close enough to an explosion to feel the blast wave. Reports of post-concussive symptoms were obtained by asking participants whether they had experienced in the past month any of eight features consistent with PCS (memory problems, poor balance, irritability, tinnitus, sensitivity to light, sensitivity to noise, headaches, insomnia). The total number of symptoms reported was used as the primary measure of PCS. Twelve participants did not have PCS ratings completed at the time of clinical interview. Four of these participants had provided PCS ratings during the initial phone screen within one month prior to participation, while the other 8 were excluded from all analyses. DSM-IV-TR diagnoses, including PTSD, MDD, and Alcohol Dependence, were finalized by a consensus review process involving advanced doctoral students and doctoral-level licensed psychologists based on information obtained from the Structured Clinical Interview for DSM-IV-TR (SCID; First et al. 2002) and Clinician-Administered PTSD Scale (CAPS; Blake et al. 1995) administered by trained study interviewers. Because individuals with pre-deployment PTSD or Major Depressive Disorder were excluded from participation, these diagnoses by rule had onsets during or after the
Brain Imaging and Behavior
most recent deployment but were not necessarily present at time of participation. Eleven primary personality trait scales were derived from the Multidimensional Personality Questionnaire, Brief Form (MPQ-BF; Patrick et al. 2002) completed by the participant directly. Each score was converted to a t-score (mean 5.0, standard deviation 1.0) based on the norms provided with the scoring software. Although a standard deviation of 10 is commonly used when reporting MPQ trait scores, a standard deviation of 1.0 was selected to provide more straightforward interpretation of beta weights and odd ratios in subsequent regression analyses. No participants had elevated scores on validity scales that would indicate invalid reporting.
correlation between GFA and FA, yet superior sensitivity of GFA to effects of PTSD and mTBI (Davenport et al. 2015). Six variables were computed for both GFA and MD, producing a total of 12 white matter integrity (WMI) variables for each individual: average value across all white matter voxels (1); the presence of at least one region of interest (ROI), out of the 20 major tracts of the JHU Atlas (Mori et al. 2005), with a value that is abnormally high (2), abnormally low (3), or abnormal in either direction (4); and the total number of voxels with abnormally high values (5) or abnormally low values (6). Abnormal values were defined as those 2 standard deviations away from the mean of the individuals without PTSD or Blast mTBI using bias-corrected thresholds according to the algorithm of Watts and colleagues (Watts et al. 2014). Regions of interest investigated are shown in Fig. 1.
Image acquisition and processing Statistical analyses Details of the MRI acquisition and processing are described in detail elsewhere (Davenport et al. 2015). Briefly, MRI data were acquired on a 3 Tesla Siemens Trio (Erlangen, Germany) scanner using a 12-channel birdcage head coil. Head movements were minimized by placing pads around the participant’s head. The diffusion imaging sequence (TR/TE=9000/ 84 ms, 72 oblique axial slices, 128×128 matrix, 256 mm FOV, 2.0 mm thickness) acquired images in each of 30 noncollinear directions at b=800 s/mm2, along with 5 images with b=0 evenly distributed throughout the sequence, and this sequence was run twice. A field map was collected immediately following the DTI sequence for offline correction of distortion artifacts. Tools from the FMRIB software library (FSL; Smith et al. 2004; Woolrich et al. 2009) were used to remove artifacts due to movement, eddy currents, and field inhomogeneity distortions. The FSL Diffusion Toolbox (Behrens et al. 2003) was used to compute FA and MD, and custom Matlab (The Mathworks, Natick, Massachusetts) scripts were used to compute GFA (Assemlal et al. 2007). The FA maps of all participants were nonlinearly aligned to a standard template and averaged to create a study-specific global mean FA image. To further reduce anatomic variability, the aligned FA maps were subsequently nonlinearly registered to the global mean FA image, and the two resultant warp fields were concatenated and applied to the native FA, MD and GFA images. A white matter mask defined as the set of voxels in which FA>0.20 across all individuals and FA>0.25 in at least 90 % of individuals was applied to all images. Conceptually, MD represents the overall amount of diffusion within a voxel regardless of its orientation, whereas both GFA and FA represent the anisotropy (i.e., degree to which diffusion is directionally specific). To reduce potential redundancy of information across measures, we excluded FA from analyses based on our prior report demonstrating a high
Data reduction To address potential colinearity among measures and to reduce the number of personality variables considered in subsequent analyses to those most likely to be related to Blast mTBI, PTSD, or PCS, the 11 MPQ measures were entered as predictors in logistic regression models of PTSD and Blast mTBI, and a linear regression model of PCS. Forward selection with an entry criterion of p
MILITARY/VETERAN TBI
Personality and neuroimaging measures differentiate PTSD from mTBI in veterans Nicholas D. Davenport & Kelvin O. Lim & Scott R. Sponheim
# Springer Science+Business Media New York (outside the USA) 2015
Abstract Mild traumatic brain injury (mTBI) is common among recent veterans and often is associated with chronic post-concussive symptoms (PCS). Elevated PCS may also be a consequence of post-traumatic stress disorder (PTSD) which shares symptoms with PCS. Identification of personality, biological, and psychopathology factors that contribute to the relationship between mTBI and PCS could help isolate the sources of chronic post concussive syndrome in veterans. Clinician rated diagnoses (PTSD, Major Depression, Alcohol Dependence), personality characteristics (Multidimensional Personality Questionnaire [MPQ] subscales), white matter brain imaging measures (Mean Diffusivity, Generalized Fractional Anisotropy), and diagnoses of mTBI were collected from 125 American military veterans of Iraq or Afghanistan. Linear and logistic regression models were tested to determine contributions to PCS and whether there were similar contributors to PTSD and mTBI. PCS score was associated with personality characteristics of high Stress Reaction and Traditionalism and low Control as well as mTBI. A diagnosis of PTSD was associated with low Social Closeness, PCS, Alcohol Dependence, and abnormal white matter mean diffusivity. Diagnosis of mTBI was associated with fewer white matter mean diffusivity abnormalities, PCS, and number of deployments. As commonly observed clinically, both PTSD and mTBI were associated with higher rates of PCS, though the contribution of PTSD appears to be secondary to personality traits, particularly Stress Reaction. Furthermore, the observation of factors that are uniquely associated with Blast mTBI N. D. Davenport (*) : K. O. Lim : S. R. Sponheim Minneapolis Veterans Affairs Health Care System, VA Medical Center (B68-2), 1 Veterans Drive, Minneapolis, MN 55417, USA e-mail: [email protected] N. D. Davenport : K. O. Lim : S. R. Sponheim Department of Psychiatry, University of Minnesota, Minneapolis, MN, USA
(number of deployments) or with PTSD (Lifetime Alcohol Dependence and low Social Closeness), as well as a factor (region of abnormal MD) that had opposite effects on the likelihood of each diagnosis, indicates that the complex relationships between personality, psychopathology, and nature of mTBI need to be considered when interpreting chronic postconcussive symptoms. Keywords PTSD . mTBI . White matter . Personality . Military
Introduction Mild traumatic brain injury (mTBI) is among the most common injuries among American service members deployed to Iraq and Afghanistan (Taber et al. 2006; Warden 2006). In most cases, the neurological and cognitive symptoms associated with mTBI resolve within days to weeks (Karr et al. 2014), but approximately a third of soldiers continue to report substantial postconcussive symptoms (PCS), including headaches, memory problems, irritability, and insomnia, that persist beyond 5 months (Schneiderman et al. 2008). These symptoms are also commonly reported by non-deployed soldiers without mTBI (Miller et al. 2013) and overlap substantially with other conditions common in veterans, particularly post-traumatic stress disorder (PTSD), complicating their use in the identification of veterans experiencing chronic effects of mTBI. The potential confounding effect of PTSD on chronic PCS has been addressed in several recent studies. While some have concluded that PTSD symptoms explain a substantial portion of the relationship between mTBI and PCS (Hoge et al. 2008; Polusny et al. 2011; Schneiderman et al. 2008), others have
Brain Imaging and Behavior
determined that mTBI and PTSD contribute to PCS independently (Brenner et al. 2010). Given that PTSD is more common among veterans reporting mTBI than among those with other types of injuries (Hoge et al. 2008; Yurgil et al. 2014), further delineation of the potentially complicated relationship among mTBI, PTSD, and persistent PCS is of high clinical importance. In particular, the identification of additional biological and psychological factors specific to mTBI would be clinically useful in resolving diagnostic ambiguity in cases of soldiers exposed to multiple forms of trauma. Likewise, the identification of factors that relate to PCS beyond the effect of mTBI may help to characterize the subset of veterans at highest risk for experiencing chronic effects of mTBI. Biological factors that establish an underlying objective basis for symptom expression would be especially useful in resolving confounding effects of comorbid conditions. Specifically, evidence of structural brain abnormalities related to PCS expression beyond the static effect of mTBI itself could be interpreted as reflecting a biological basis for variability in long-term outcomes of mTBI. Based on evidence of diffuse axonal injury associated with more severe forms of TBI (Hammoud and Wasserman 2002), diffusion tensor imaging (DTI) has been used in recent studies of mTBI to determine whether measures of white matter integrity (WMI) have utility as biomarkers. In healthy white matter, water is constrained by myelin and cytoskeletal elements and thus able to diffuse more freely in the direction of the axon fibers than in perpendicular directions, creating a highly non-spherical, or anisotropic, pattern of diffusion (Beaulieu 2002). Therefore, WMI is typically characterized by high fractional anisotropy (FA) and low mean (i.e., overall) diffusivity (MD). Injury or disease is presumed to compromise white matter structures, leading to increased water diffusivity (i.e., higher MD) and a less directionally oriented pattern (i.e., lower FA). Investigations of civilian mTBI typically reveal a pattern of WMI changes mirroring the clinical course, namely acute disruption followed by normalization over a period of weeks or months and chronic disruption in only a subset of patients (Kou et al. 2013; Niogi and Mukherjee 2010; Yuh et al. 2014). However, because civilian mTBI is typically studied based on emergency room medical records, its characterization and diagnosis is less reliant on retrospective self-report than military mTBI. While several DTI studies in veterans have provided evidence of reduced WMI associated with mTBI (Davenport et al. 2012; Jorge et al. 2012; MacDonald et al. 2011; Morey et al. 2013; Yeh et al. 2014; Yuh et al. 2014), the potential effects of PTSD and PCS have not been meaningfully considered. While PTSD has received less attention in DTI studies, there are reports of decreased FA associated with PTSD in civilian (Fani et al. 2012; Kim et al. 2005; Kim et al. 2006) and military (Schuff et al. 2011) samples and one report of increased FA in left anterior cingulum (Abe et al. 2006). However, individuals with suspected concussion were explicitly excluded
from these studies. In a sample of veterans in which both mTBI and PTSD were represented, we previously reported higher Generalized FA (GFA), a measure analogous to FA that accounts for WMI across multiple fiber orientations within a voxel (e.g., crossing or diverging fibers), associated with a lifetime history of PTSD and relatively few abnormalities associated with mTBI (Davenport et al. 2015), suggesting that the chronic phase of mTBI may not involve systematic disruptions in WMI after accounting for PTSD. Moreover, given that GFA and FA are identical in regions modeled adequately by a single tensor, the observation of differences in GFA but not in FA demonstrates the additional sensitivity to WMI disruptions afforded by consideration of multiple fiber orientations. It is also possible that some of the variability in postconcussive symptom expression is due to psychological factors (e.g., psychopathology, personality traits) that reflect overlapping conditions (e.g., PTSD, depression) or differences in subjective experience and reporting of PCS. To date, most investigations of psychological factors contributing to PCS report have focused on dimensional measures of anxiety and depression (Clarke et al. 2012; Wood et al. 2014) or symptoms of PTSD and Major Depressive Disorder (Garden et al. 2010; Polusny et al. 2011; Wood et al. 2014). Most of these studies find strong associations between measures of psychopathology and PCS, even in non-clinical samples (Garden et al. 2010). The role of broader personality traits has been studied less fully, though the potential for neuroticism and negative affect to be associated with increased subjective perception and reporting of bodily sensations has been recognized (Fayol et al. 2009; Wood et al. 2014). One study looking at this relationship found that mTBI patients had higher levels of neuroticism than healthy controls, though neuroticism was elevated to a similar degree among patients with non-head injuries, indicating non-specificity to mTBI (Clarke et al. 2012). Given that PTSD is also associated with neuroticism (James et al. 2013; Kotov et al. 2010) and negative emotionality (Miller et al. 2003), it is likely that personality traits play an important role in its relationship with mTBI and PCS. The primary purpose of this investigation was to better characterize the potential contributions of white matter integrity, personality, and psychopathology to the relationship between mTBI and reported PCS within a sample of veterans recently returned from combat deployments to Iraq or Afghanistan . We adopted a broad approach that included objective biological measures of brain structure, self-report questionnaire data, and clinician-rated diagnoses and symptoms in order to more completely test the relationships of personality, brain abnormalities, and mental disorders to chronic postconcussive symptoms. Specifically, we tested the hypothesis that mTBI would remain a significant predictor of PCS after accounting for personality traits and the presence of PTSD as compared to the alternative hypothesis that psychological
Brain Imaging and Behavior
measures explain a substantial portion of the variability in PCS reporting, both of which have been supported in prior literature. As a corollary to this hypothesis, we tested whether factors related to PCS beyond the contributions of mTBI and PTSD were more indicative of variability in reporting (i.e., personality traits) or in underlying brain structure (i.e., WMI). A secondary purpose was to identify measures that independently relate to mTBI and PTSD diagnoses, namely personality traits, WMI indices, and history of mental health conditions that provide unique diagnostic information beyond the contribution attributable to PCS. Based on prior literature, we hypothesized that PTSD would be associated with psychological variables, especially Alcohol Dependence and negative emotionality traits (e.g., Stress Reaction). Likewise, we hypothesized that mTBI would be associated with biological measures, specifically reduced white matter integrity.
Methods and materials Participants Participants consisted of 133 veterans of Operations Enduring and Iraqi Freedom described in a prior report (Davenport et al. 2015) who had reported either traumatic combat experiences or exposure to explosive blasts during their most recent deployment (2–5 years prior to participation) on a research or clinical instrument. All reports of mTBI and clinical symptoms were independently assessed during study participation. Exclusionary criteria included native language other than English, current or predeployment unstable medical condition that would reasonably be expected to significantly affect brain function (e.g., anoxic episode >10 s, stroke, seizures, multiple sclerosis, etc.), uncorrected visual problems or hearing loss, moderate or severe TBI not due to blast, any predeployment Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV-TR; American Psychiatric Association 2000) Axis I psychotic or mood disorder, pre-deployment PTSD according to the Clinician Administered PTSD Scale (CAPS; Blake et al. 1995), current or past substance dependence other than nicotine or alcohol, and contraindications to MRI (e.g., metallic implants, shrapnel, claustrophobia). Participants completed an informed consent process that included complete description of the study, and participants were provided monetary compensation for participation after each study procedure. The study protocol was reviewed and approved by the University of Minnesota and Minneapolis Veterans Affairs Medical Center Institutional Review Boards and the U.S. Army Medical Research and Materiel Command (USAMRMC).
Clinical assessment The presence of mTBI was determined by doctoral-level neuropsychologists based on consensus review of information obtained from the Minnesota Blast Exposure Screening Tool (MN-BEST; Nelson et al. 2011), including detailed descriptions of events and reports of altered consciousness (e.g., confusion, disorientation), loss of consciousness (LOC) less than 30 min, post-traumatic amnesia (PTA) up to 24 h, and neurological symptoms (e.g., headache, tinnitus, nausea, sensitivity to light or noise) immediately after each event. Blast-related injuries were defined as those in which the individual felt a blast wave and attributed the resultant concussion to its effects, though physical impacts (e.g., hitting head on ground, being hit by debris) were allowed to be present provided that a blast wave was the prominent source of injury. The three most significant potential blast-related and impact-related TBI events were considered. One participant reported LOC lasting 30–60 min but was included because the duration of LOC could not be verified and, therefore, mild TBI could not be ruled out. Impact-related injuries generally occurred in a civilian context several years prior to military deployment (e.g., childhood accidents, participation in youth contact sports), whereas blast-related injuries occurred during military deployment and were most proximal to the report of post-concussive symptoms. Therefore, Blast mTBI (i.e., recent deploymentrelated injuries involving exposure to explosive blasts) was considered the index event most likely related to current PCS, and Impact mTBI (i.e., history of civilian non-blast injuries) was considered as an independent contributing factor. Blast exposure was determined by asking participants whether they had ever been close enough to an explosion to feel the blast wave. Reports of post-concussive symptoms were obtained by asking participants whether they had experienced in the past month any of eight features consistent with PCS (memory problems, poor balance, irritability, tinnitus, sensitivity to light, sensitivity to noise, headaches, insomnia). The total number of symptoms reported was used as the primary measure of PCS. Twelve participants did not have PCS ratings completed at the time of clinical interview. Four of these participants had provided PCS ratings during the initial phone screen within one month prior to participation, while the other 8 were excluded from all analyses. DSM-IV-TR diagnoses, including PTSD, MDD, and Alcohol Dependence, were finalized by a consensus review process involving advanced doctoral students and doctoral-level licensed psychologists based on information obtained from the Structured Clinical Interview for DSM-IV-TR (SCID; First et al. 2002) and Clinician-Administered PTSD Scale (CAPS; Blake et al. 1995) administered by trained study interviewers. Because individuals with pre-deployment PTSD or Major Depressive Disorder were excluded from participation, these diagnoses by rule had onsets during or after the
Brain Imaging and Behavior
most recent deployment but were not necessarily present at time of participation. Eleven primary personality trait scales were derived from the Multidimensional Personality Questionnaire, Brief Form (MPQ-BF; Patrick et al. 2002) completed by the participant directly. Each score was converted to a t-score (mean 5.0, standard deviation 1.0) based on the norms provided with the scoring software. Although a standard deviation of 10 is commonly used when reporting MPQ trait scores, a standard deviation of 1.0 was selected to provide more straightforward interpretation of beta weights and odd ratios in subsequent regression analyses. No participants had elevated scores on validity scales that would indicate invalid reporting.
correlation between GFA and FA, yet superior sensitivity of GFA to effects of PTSD and mTBI (Davenport et al. 2015). Six variables were computed for both GFA and MD, producing a total of 12 white matter integrity (WMI) variables for each individual: average value across all white matter voxels (1); the presence of at least one region of interest (ROI), out of the 20 major tracts of the JHU Atlas (Mori et al. 2005), with a value that is abnormally high (2), abnormally low (3), or abnormal in either direction (4); and the total number of voxels with abnormally high values (5) or abnormally low values (6). Abnormal values were defined as those 2 standard deviations away from the mean of the individuals without PTSD or Blast mTBI using bias-corrected thresholds according to the algorithm of Watts and colleagues (Watts et al. 2014). Regions of interest investigated are shown in Fig. 1.
Image acquisition and processing Statistical analyses Details of the MRI acquisition and processing are described in detail elsewhere (Davenport et al. 2015). Briefly, MRI data were acquired on a 3 Tesla Siemens Trio (Erlangen, Germany) scanner using a 12-channel birdcage head coil. Head movements were minimized by placing pads around the participant’s head. The diffusion imaging sequence (TR/TE=9000/ 84 ms, 72 oblique axial slices, 128×128 matrix, 256 mm FOV, 2.0 mm thickness) acquired images in each of 30 noncollinear directions at b=800 s/mm2, along with 5 images with b=0 evenly distributed throughout the sequence, and this sequence was run twice. A field map was collected immediately following the DTI sequence for offline correction of distortion artifacts. Tools from the FMRIB software library (FSL; Smith et al. 2004; Woolrich et al. 2009) were used to remove artifacts due to movement, eddy currents, and field inhomogeneity distortions. The FSL Diffusion Toolbox (Behrens et al. 2003) was used to compute FA and MD, and custom Matlab (The Mathworks, Natick, Massachusetts) scripts were used to compute GFA (Assemlal et al. 2007). The FA maps of all participants were nonlinearly aligned to a standard template and averaged to create a study-specific global mean FA image. To further reduce anatomic variability, the aligned FA maps were subsequently nonlinearly registered to the global mean FA image, and the two resultant warp fields were concatenated and applied to the native FA, MD and GFA images. A white matter mask defined as the set of voxels in which FA>0.20 across all individuals and FA>0.25 in at least 90 % of individuals was applied to all images. Conceptually, MD represents the overall amount of diffusion within a voxel regardless of its orientation, whereas both GFA and FA represent the anisotropy (i.e., degree to which diffusion is directionally specific). To reduce potential redundancy of information across measures, we excluded FA from analyses based on our prior report demonstrating a high
Data reduction To address potential colinearity among measures and to reduce the number of personality variables considered in subsequent analyses to those most likely to be related to Blast mTBI, PTSD, or PCS, the 11 MPQ measures were entered as predictors in logistic regression models of PTSD and Blast mTBI, and a linear regression model of PCS. Forward selection with an entry criterion of p
