POSTTRAUMATIC STRESS DISORDER (PTSD) is a debilitating psychiatric illness characterized by the reexperiencing of a traumatic event through nightmares and flashbacks, avoidance of stimuli related to the event, and hyperarousal (difficulty concentrating and sleeping).1 The pathogenesis is thought to involve increased activation of the amygdala and reduced inhibitory control by the ventromedial prefrontal cortex and the hippocampus.2 Direct, personal exposure to traumatic events, such as those experienced in combat, is necessary for the development of PTSD.
The prevalence of PTSD among Operation Enduring Freedom/Operation Iraqi (OEF/OIF) veterans ranges from 13% to 17% compared with 4% of the US adult population.3 The diagnosis of PTSD is based on symptoms alone. There are no objective diagnostic aids. These symptoms overlap considerably with those related to mild traumatic brain injury (mild TBI), which also commonly affects military personnel in Iraq and Afghanistan. Mild TBI is characterized by a brief loss or alteration of consciousness and a period of amnesia or change in mental state at the time of the injury.4 A RAND study estimated that 320 000 OEF/OIF service members had suffered a mild TBI as of January 2008,5 often as a result of exposure to blast.6,7 As with PTSD, the diagnosis of mild TBI is made clinically because there are no accepted diagnostic adjuncts.
Rates of PTSD seem to be higher among those exposed to mild TBI. A 2009 Institute of Medicine report concluded that there was evidence of an association between TBI and PTSD, based on observational studies conducted in military personnel who served in Iraq and Afghanistan.8 For example, one study reported the prevalence of PTSD to be more than 3 times higher after mild TBI, supporting the notion of an interaction between the them.6 The reason for this interaction between mild TBI and PTSD is not clear and is the focus of the current report.
Understanding the nature of this interaction may provide direction for future efforts to prevent and treat PTSD, especially when it co-occurs with mild TBI. However, it is also important for clinicians currently providing postdeployment medical care to veterans. Symptom overlap and lack of diagnostic aids makes it difficult to separate the effects of TBI from those of PTSD in clinical care of individual soldiers. This has given rise to controversy regarding the nature of the interaction between these 2 entities.9 Some maintain that postdeployment symptoms are due to both mild TBI and combat stress/PTSD.10 Others hold that combat stress is responsible for most postdeployment symptoms. Hoge et al6 found that PTSD and depression—but not mild TBI—were associated with poor neurobehavioral outcomes among 2525 army infantry soldiers following their return from Iraq. Two additional studies found similar associations.7,11 In all 3 studies, mild TBI was diagnosed by symptom self-report. However, even in controlled medical environments, a clinical diagnosis of mild TBI can be difficult to ascertain accurately,12,13 especially when a description of the event and associated symptoms are elicited months after the event.14,15 Mild TBI ascertainment in a combat environment is especially problematic because symptoms of loss of consciousness or altered mental state can also occur as a result of dissociation in the face of emotional trauma.16 Finally, exposure to blast, a frequent occurrence in the current conflicts, can cause subtle brain injury without frank loss of consciousness or amnesia.17,18
At the heart of this controversy, thus, is an incomplete understanding of the relations among blast, mild TBI, combat stress, and PTSD. The use of a symptom-based mild TBI diagnosis, without an objective measure of brain injury, may be obscuring the contribution of traumatically induced brain injury to the development of PTSD. When objective diagnostic measures of brain injury are used, TBI seems to play a more prominent role in the genesis of PTSD. Animal studies using histopathologic techniques to determine brain injury suggest that TBI19,20—or even just exposure to blast21—increases the vulnerability to PTSD. Until recently, the lack of an objective in vivo measure of brain injury has prevented studying the mild TBI-PTSD relation in humans.
Diffusion tensor imaging (DTI) has now matured to the point where it can potentially fill the role of an objective diagnostic adjunct to the detection of brain injury. This form of magnetic resonance imaging (MRI) has detected subtle changes in the white matter (WM) associated with civilian mild TBI22–24 and with mild TBI experienced by military personnel involved in OEF/OIF.25,26 However, DTI has not yet been used to study the TBI-PTSD relation, despite a 2009 recommendation from a working group of US government and civilian physicians and scientists published as the St. Louis Workshop Report.24
In the current study, we sought to understand the relations of mild TBI and blast exposure to brain WM structure on DTI and how these may interact to affect the development of PTSD.
We performed a nested cohort study of 52 OEF/OIF veterans who served in combat areas between 2001 and 2008. The parent cohort consisted of 500 OEF/OIF veterans from Veterans Integrated Service Network (VISN) 2 recruited between August 2008 and January 2010 as part of a study of cognitive, affective, and behavioral correlates of veterans with and without TBI funded by VA Health Services Research and Development. These subjects were enrolled through several VA medical centers in Upstate New York. A special modification to the original grant (Cognitive Assessment of Veterans After TBI: DTI Substudy) permitted DTI scanning of a subset of these individuals.
The subjects in this subset were chosen for their willingness to travel to the DTI site, to undergo DTI scanning, and for lack of contraindications to scanning (retained metallic foreign bodies and claustrophobia). At the time of study, DTI was not available at any VISN 2 VA facilities.
DTI image data were correlated with information already gathered during the parent study, including mild TBI diagnostic data, blast exposure, severity of PTSD, and severity of combat experiences. These variables were defined in the following way:
PTSD severity was taken from the PTSD Checklist-Military. The PTSD Checklist-Military is a very widely used 17-item checklist developed at the National Center for PTSD that follows Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition) diagnostic criteria. Subject rate the extent to which they have been bothered over the last month by each of the 17 listed problems on a 5-point Likert scale ranging from 1 (“not at all”) to 5 (“extremely”). Thus, total scores can range from 17 to 85. The internal consistency coefficient (Cronbach α) for the total scale is 0.939. With a cutoff score of more than 44 used to define PTSD, overall diagnostic efficiency has been reported at 0.90, with a sensitivity of 0.94 and a specificity of =086,27 although a cutoff of 50 or more is also commonly used.28 As part of the parent study protocol, subjects underwent several assessments of PTSD severity at varying time points after return home. We analyzed the assessment that was administered nearest to the time of the DTI scan.
Mild TBI diagnosis was determined by in-person interview using a 22-item questionnaire developed to establish the nature, probability, and severity of deployment-related TBI among OEF/OIF veterans. The interview followed previously published TBI diagnostic criteria, which include confirmation of a possible TBI event, confirmation of alteration of consciousness, and confirmation of postconcussion symptoms.29 On the basis of the standardized clinical interview, interviewers rated the likelihood of mild TBI according to a 6-point scale: “not at all likely,” “very unlikely,” “somewhat unlikely,” “somewhat likely,” “very likely,” and “almost certainly.” These likelihood categories were used in all analyses. However, for descriptive purposes, subjects were defined as having mild TBI if interviewers rated them “very likely” or “almost certainly.”
Severity of exposure to traumatic events was taken from the Walter Reed Army Institute of Research Combat Experiences Survey. The Combat Experiences Survey is a 36-item scale that measures combat intensity based on frequency and type of combat experiences. The first 33 items are various deployment-related experiences that range from combat-related questions (“being attacked or ambushed”) to deployment duties (“handling or uncovering remains”) to possible deployment or combat-related events (“knowing someone seriously injured or killed,” “had a close call, dud landed near you,” and “provided aid to the wounded”). Item responses are on a 5-point scale related to how often the event was experienced, ranging from 0 (“never”) to 5 (“10 or more times”). The last 3 items of the scale are each scored differently than the first 33 items and pertain to how often a service member was in serious danger of being injured or killed, how many times one engaged in an enemy firefight, and whether 1 or more nights were spent in the hospital. These 3 items were excluded from the current study because of the differences in meaning and scaling. Total scores thus range from 0 to 165. Although formal psychometric data are not available for this instrument, an exploratory factor analysis conducted at the University of Buffalo suggested that 3 factors represented the scale well: exposure to combat environment, physical engagement, and proximity to serious injury and death. These 3 factors showed high internal consistency, with Cronbach α scores of 0.94, 0.81, and 0.82, respectively. Evidence for construct validity was shown by estimates convergent and discriminant validity of the factors with demographic data, other war experience scales, and cognitive, affective, and behavioral measures. (K. Donnelly, PhD, 2012, written communication)
Exposure to blasts
Blast exposure was determined by responses to interviews conducted during the parent study. The interviewers used a semistructured questionnaire to ask for a description of events in which they were exposed to any of the following: blast, improvised explosive device, bullet above the shoulder, rocket-propelled grenade, mortar, landmine, grenade, blow to the head, vehicular accident, fall, or assault. Among these, only events specifying exposure to a blast (blast, improvised explosive device, rocket-propelled grenade, mortar, landmine, and grenade) were considered. Events causing injury (ie, fall, assault, vehicular accident, or blow to head) that may have contributed to head injury—but did not involve exposure to a blast—were not considered blast-exposed. Responses were reviewed by the interviewing clinician for detail and clarity (ie, how occurred, the sequence of events, and the specifics of the blast exposure). Blast exposure was summarized as “yes” if an individual had 1 or more blast exposures. The number of events reported ranged from 0 to 3.
DTI data acquisition
Images were acquired on a Siemens 3T Trio MRI scanner (Siemens Medical, Erlangen, Germany) equipped with a 32-channel head coil. In addition to scout images for subject localizations, the following scans were acquired: (a) 3-dimensional structural T1-weighted images are collected using a magnetization prepared rapid gradient echo sequence with parameters such as time-of-repeat (TR) = 2350 ms, time-of-echo (TE) = 3.4 ms, 256 × 256 matrix, field-of-view (FOV) = 25.6 cm, 1-mm slice thickness; (b) 2-dimensional axial fast gradient-recalled echo (GRE) dual-echo sequence, with slice thick = 2 mm with no gap, FOV = 256 × 256 mm2, matrix = 128 × 128, echo times = 6.2/8.66 ms; (c) 2-dimensional axial DTI using a single-shot pulsed-gradient dual-echo spin-echo echo planar imaging sequence with TR/TE = 8900/81 ms, slice thick = 2 mm with no gap, matrix = 128 × 128, FOV = 256 × 256 mm2, integrated parallel acquisition techniques with generalized autocalibrating partially parallel acquisitions acceleration factor = 2, diffusion weighting direction = 60 with b = 700 s/mm2 and 1 average, and b = 0 images with 10 averages.30 For the whole-brain coverage of about 70 slices, total scan time for DTI was about 11 minutes. Standard T1/T2-weighted MRI images were acquired during the same session. GRE mages were reviewed for hemosiderin and fluid-attenuated inversion recovery (FLAIR) images for WM abnormalities. A FLAIR image is a T2-weighted image in which the fluid signal has been suppressed. Typically, fluid is bright white on a T2-weighted image, but in the FLAIR images, this signal is suppressed to further enhance the sensitivity. White matter signal abnormalities detected using FLAIR are considered relatively nonspecific and more common with advanced age. They may reflect gliosis resulting from small-vessel ischemia or demyelination such as that seen in multiple sclerosis. Other etiologies include hypertension and migraine.
A custom software tool based on Matlab (The Mathworks, Natick, Massachusetts), C++, and various functions in the FSL package (FMRIB Analysis Group, Oxford University, Oxford, England,31) was used for image processing and statistical analysis. To ensure an accurate representation of data, the following processing steps were performed: (a) Motion and eddy current artifacts were simultaneously removed using the Eddy_Correct Tool of the FSL package. This step allows better alignment of images among the DTI data volumes scanned with different diffusion gradient directions. An additional step using the field map from dual-echo GRE images and the algorithm (FUGUE Toolbox from FSL package) removed the susceptibility artifacts from the original DTI data. (b) Diffusion tensor as well as 2 tensor-derived parametric maps, fractional anisotropy (FA) and mean diffusivity (MD), were estimated for individual subjects using the DTIFIT tool in FSL. (c) Using the 12-df affine registration with the FLIRT toolbox in FSL, nondiffusion-weighted images (b = 0) in each individual DTI data set were registered with the 3-dimensional T1-weighted image obtained by the magnetization prepared rapid gradient echo sequence, which has high image contrast between WM/gray matter (GM)/cerebrospinal fluid (CSF) and high image resolution.
Whole-brain WM analysis of FA and MD values
The segmentation of WM/GM/CSF based on the 3-dimensional T1-weighted image was performed for each individual using the FAST toolbox in FSL. The resultant high-resolution WM mask was then inversely projected into each individual's DTI space, using the inverse transformation matrix obtained in step c to generate the region of interest (ROI) mask for statistical analysis of FA and MD within the whole WM.
Automated ROI segmentation
The John Hopkins White Matter Parcellation Atlas Type II (JHU WMPM)32 was used to segment the whole brain into 56 WM regions, 22 deep GM regions, and 52 cortical GM regions. A 2-step image-processing approach was used to transform the JHU WMPM Atlas into individual subject's native space based on DTIStudio software package (cmrm.med.jhmi.edu, H. Jiang and S. Mori, Kennedy Krieger Institute, Johns Hopkins University, Baltimore, Maryland). In the first step, a 12-df affine linear registration was performed between each individual's FA map and the John Hopkins FA Atlas to achieve global adjustment of the size and orientation of the brain. In the second step, a nonlinear registration using Large Deformation Diffeomorphic Metrix algorithm was used to achieve more accurate transformation of the atlas into individual's native space. To further improve the transformation, the dual-contrast Large Deformation Diffeomorphic Metrix Mapping algorithm was adapted in which both FA image and the nondiffusion-weighted image (b = 0) were used for the nonlinear registration simultaneously. In the end, the JHU WMPM Atlas was transformed to each individual's native space and the binary masks for 130 ROIs were then automatically generated for ROI-based analysis.
The efficiency of the automated ROIs segmentation is illustrated in Figure 1, where the ROIs from transformed JHU WMPM Atlas are superimposed on the FA map from 4 study subjects. For each ROI, the mean value of FA/MD over all pixels within the ROI was calculated for statistical analysis. Each cortical GM region defined by the JHU WMPM Atlas is actually a mixture of cortical GM and a layer of “superficial” WM close to it. Unlike the deep WM structures inside which WM tracts are well organized, these superficial WM layers are the regions where WM fibers become smaller and branch into the GM. Therefore, fiber crossing and the partial volume effect make FA measurement in these WM layers less robust. Because of these concerns, the 52 GM cortical regions were not included, leaving 78 ROIs for analysis. The left (L) and right (R) hemispheric components of each structure were combined, yielding a total of 39 ROIs in the final analysis.
Multiple linear regression models were used to model PTSD severity as a function of mild TBI likelihood and whole-brain DTI indices, with control for blast exposure, severity of exposure to traumatic events, prior head injury, time since last tour of duty, age, and gender. All subsets and stepwise selection methods were used to refine and stabilize the models by removing candidate predictors that were not statistically significant at the 2-sided nominal 0.05 level. The effect of each predictor was summarized by its regression coefficient, standard error (SE), 95% confidence interval (CI), and 2-sided P value. Swapping the roles of PTSD severity and mild TBI likelihood, similar linear models were used to model the latter with the same set of variables.
Multiple binary logistic regression models were used to model the odds of exposure to at least 1 blast as a function of DTI indices, adjusted for PTSD severity, mild TBI likelihood, severity of exposure to traumatic events, time since last tour of duty, prior head injury, age, and gender. Stepwise selection was used to refine and stabilize the models by removing candidate predictors that were not statistically significant at the 2-sided nominal .05 level. The effect of each predictor was summarized by its odds ratio (OR), along with its associated 95% Wald CI, and likelihood ratio test P value. Profile likelihood CI values were used in place of Wald CI when parameter estimates were on the boundary (OR = 0). Nonparametric kernel density estimates, using the Silverman normal-based rule-of-thumb bandwidth selector, were used to graphically depict the distributions of DTI indices, stratified by blast exposure.
FA and MD values of the whole-brain WM were summarized with a quantile approach that uses FA and MD values from each WM voxel in a native, nonnormalized space. The adult human brain has approximately 30 000 WM voxels 2 mm3 in size. Each subject's WM FA and MD values were summarized by 3 preselected percentiles (1st, 50th, and 99th percentiles), covering both extremes as well as the center (median) of the distribution. Thus, each subject's WM voxels were represented by 6 DTI values. A seventh indicator was added by considering the findings of abnormal T1/T2-weighted MRI found in the process of DTI data. The rationale behind using relatively extreme percentiles (1st and 99th) as summary statistics is that we expect a head injury to affect only a relatively small proportion (1%-10%, for example) of the subject's brain, and the location could be different from subject to subject. Thus, when analyzing the whole brain, we would not necessarily expect the median to be strongly affected—unless most of the brain was injured. In a prior publication, we found the 1st percentile MD to be significantly decreased among civilian mild TBI subjects.23
In the ROI analysis, PTSD severity, mild TBI likelihood, and blast exposure were modeled similarly as in the whole-brain WM analyses, but each of the 234 ROI-DTI outcome-specific models contained only a single DTI measure: either mean FA or mean MD for 1 of the 39 WM and deep GM regions identified by the JHU WMPM Atlas. Furthermore, model selection was not performed separately for each model but rather each model was adjusted for those covariates significant in the whole-brain WM analyses. Since about 12 values of P < .05 would be expected by chance under the null hypothesis when performing 234 tests, ROI-based P values were adjusted for multiple comparisons using the Benjamini-Hochberg method to control the false discovery rate (FDR), assuming nonnegative dependence.
The characteristics of the 52 included subjects are shown in Table 1. Thirty subjects experienced at least 1 mild TBI, defined as a clinical interview rating of “almost certainly” or “very likely,” and 7 subjects had more than 1 mild TBI. Two subjects sustained mild TBI while on active duty but before being deployed; one of these subsequently suffered another mild TBI while deployed. Thirty-one subjects (60%) were exposed to 1 or more blasts. Fifteen participants met the diagnostic criteria of PTSD with a score of 50 or more. The number of subjects with and without PTSD and/or mild TBI is shown in Table 1. Standard T1/T2-weighted MRI was abnormal in 5 of 52 subjects (9.6%), as shown in Table 2, and 3 of those 5 subjects (60%) experienced mild TBI, versus 58% of those with normal T1/T2-weighted MRI.
Compared with the parent cohort, the 52 subjects in the study population had similar mean PTSD scores, mean age, and proportion of females, but a higher proportion with mild TBI as shown in Table 3.
Predictors of PTSD severity
PTSD severity was associated with severity of exposure to combat events, age, time since last tour, abnormal T1/T2-weighted MRI, and 1st percentile of MD on DTI (Table 4). PTSD severity increased with increasing whole-brain 1st percentile MD values. After adjustment for covariates, PTSD severity was an average of 13.3 points (SE = 6.5) higher in the 5 subjects with abnormal T1/T2-weighted MRI than the 46 subjects with normal scans (P = .046, with 1 subject omitted because of missing traumatic events exposure information). These clinical and DTI variables accounted for 33.4% of the variation in PTSD severity among the subjects, whereas the clinical variables alone explained only 20.6% of the variance. There was insufficient evidence that either mild TBI likelihood or self-report of exposure to blasts was related to PTSD severity.
Predictors of exposure to blast and of mild TBI likelihood
Blast exposure was associated with 1st percentile of FA on DTI (OR = 0.38 per SD; 95% CI, 0.15–0.92), abnormal T1/T2-weighted MRI (OR = 0.00; 95% likelihood ratio test CI, 0.00–0.09), and the severity of exposure to traumatic events (OR = 3.64 per SD; 95% CI, 1.40–9.43). Compared with unexposed subjects, those exposed to blast were more likely to have reduced 1st percentile FA values (see Figure 2). The odds of being exposed to at least 1 blast increased with severity of exposure to traumatic events and decreased with abnormal T1/T2-weighted MRI. None of the 5 subjects with abnormal T1/T2-weighted MRI was exposed to blast compared with 66% of subjects with normal T1/T2-weighted MRI. Removing these individuals from the logistic regression model did not substantially alter the OR for the 1st percentile of FA or the severity of exposure to traumatic events.
Modeling mild TBI likelihood as the outcome, those with a prior head injury had a mean mild TBI likelihood of 1.09 (SE = 0.48) points lower than those with none (P = .027) whereas subjects exposed to at least 1 blast had a mean TBI likelihood of 0.95 (SE = 0.44) points higher than those unexposed (P = .035). Since these effects cancel one another (and there is insufficient evidence of interaction, P = .25), the mild TBI likelihood for the 8 subjects with both a prior head injury and exposure to at least 1 blast (mean ± SD = 3.88 ± 1.89) was almost identical to that for the 14 subjects with neither (mean ± SD = 3.86 ± 1.88).
Specific brain regions differentially affected by PTSD and mild TBI
Of the 39 WM and deep GM regions analyzed, 10 seemed to be associated with PTSD, TBI, or blast after adjustment for covariates but before control for multiple comparisons (Table 5). DTI changes in the caudate nucleus and the inferior cerebellar peduncle were associated with both PTSD and TBI. After application of the Benjamini-Hochberg FDR procedure to adjust for multiple comparisons, none of these regions were significantly associated with PTSD, TBI, or blast exposure (FDR and Bonferroni/Holm-corrected P values >0.7).
In the current study, we examined the relations among combat-acquired PTSD, mild TBI, and several clinical and neuroimaging variables. As expected, PTSD severity was most strongly associated with the severity of exposure to traumatic combat events. However, PTSD severity was also associated with high 1st percentile of MD on DTI and WM lesions on MRI. Although the association of PTSD with combat intensity has been reported by prior investigators, we are the first to find an association between DTI abnormalities and PTSD acquired in a combat setting. Several authors have examined DTI images in veterans but either did not include PTSD measures or did not find a statistical relation to PTSD. Levin et al25 analyzed DTI scans in 37 OEF/OIF veterans with blast-related mild-moderate TBI, using fiber tracking and found an association between low FA in the corpus callosum and higher PTSD severity, but the results were not statistically significant (r = 0.30, P = .07). Similarly, Davenport et al33 did not find a relation between PTSD symptoms and DTI indices among 58 blast-exposed OEF/OIF veterans. MacDonald et al26 performed DTI scans on 63 OEF/OIF veterans with blast-related mild TBI but did not examine the effect of PTSD.
There are several potential explanations for our observed associations between PTSD and abnormal DTI. The first is that combat stress produces changes on DTI that are unrelated to the physical forces to which the brain may be exposed. There is limited evidence to support this notion. Chronic stress in rodent models has been shown to induce the retraction and debranching of neuronal dendritic spines in medial prefrontal cortex,34 possibly via prolonged supranormal glucocorticoid levels.35 Reduced dentritic spine density could theoretically increase interaxonal water diffusion and lead to the increased MD observed in our subjects. Three authors reported changes in FA—but not MD—in adult human subjects who acquired PTSD in noncombat environments.36–38 In all 3 studies, these FA abnormalities were confined to the cingulum. In the current study, we did not find DTI abnormalities in the cingulum but we did observe a trend toward DTI abnormalities being associated with PTSD in other brain regions, several of which were also associated with a clinical diagnosis of mild TBI. These areas included the caudate nucleus and the inferior cerebellar peduncle. Thus, while chronic stress alone may be capable of inducing DTI changes, our results suggest that DTI changes associated with combat stress may be different from those associated with noncombat situations and possibly related to TBI.
An alternative explanation is that physical forces to the brain encountered during the course of combat operations—such as blast exposure or frank mild TBI—lead to brain injury that increases the vulnerability to PTSD. Prior evidence exists to support this notion as well. Several investigators have demonstrated that the brains of rodents exposed to sublethal blast display axonal swelling, disrupted axonal transport, glial reaction, and demyelination.19,20,39 This neuronal injury may provide the substrate for the development of prolonged periods of anxiety after exposure to stress. Kwon et al21 demonstrated that rats exposed to chronic stress and a single blast had a significantly longer period of anxiety and memory impairment than rats exposed to chronic stress alone. The blast plus stress-exposed rats showed evidence of neuronal and glial cell loss, whereas the stress-only rats did not.21 Our observation of a significant association between exposure to blast and DTI abnormalities suggests that increased PTSD vulnerability may be due in part to exposure to blast.
A final explanation involves the makeup of our subjects. Our subset is a slightly biased subset of the parent cohort compared with our subset that had similar mean PTSD scores but a higher proportion of subjects with mild TBI. Among our mild TBI subjects, the mean PTSD score was significantly lower than that of mild TBI subjects in the parent cohort (43.6 ± 15.4 vs 53.7 ± 16.3, P = .002). These differences may have obscured an association between mild TBI and PTSD, permitting DTI abnormalities to become relatively more significant in our sample. However, the mean PTSD score of 43.6 among our mild TBI subjects is similar to that in the study of Hoge et al6 (mean of 39 in those without loss of consciousness and 46 in those with loss of consciousness), showing a strong relation between PTSD and TBI.
Why PTSD would be associated with WM lesions on MRI is less clear, although we are not the first to report this. Five of 52 subjects in our cohort had lesions in periventricular and/or subcortical WM regions, and these individuals had significantly higher PTSD scores. Prior reports examining the relation between MRI and PTSD have found significant reductions in the volume of the hippocampus and, to a lesser extent, the amygdala.40,41 However, only one study reported an association with WM abnormalities. Canive et al42 identified WM lesions on MRI scans among 8 of 42 veterans with combat-related PTSD. These lesions were located in periventricular areas and at the WM/GM cortical junction. Similar to our findings, these lesions were identified only on the FLAIR images. The etiology of these WM lesions is uncertain. The GRE images failed to reveal hemosiderin deposits in these regions of FLAIR signal abnormalities, significantly reducing the likelihood that these WM lesions were due to trauma. In fact, in our multivariate analysis, neither mild TBI nor blast exposure was related to subjects with these lesions. White matter lesions in young adults can be associated with migraines,43 hypertension,44 and cocaine or opiate dependence.45 Lack of detailed information on medical history prevented us from adjusting our analyses for these conditions. Among older adults, WM lesions on MRI are often associated with demyelinating and cerebrovascular diseases.46 Although neither disease has not been shown to be linked to the development of PTSD, cerebrovascular disease has been associated with the reemergence of PTSD symptoms after initial resolution.47 Finally, WM lesions such as the ones we observed have also been found among 8% to 18% of asymptomatic, healthy adults.46,48 Regardless of etiology, these lesions may be useful in identifying those at increased risk of PTSD. However, the low proportion of subjects with abnormal MRI (9.6% in our cohort) may limit the practical application of MRI as a risk stratifier.
Mild TBI likelihood (linear, dichotomized ≥5, or otherwise) was not associated with PTSD severity. Objective evidence of brain injury on DTI, however, was associated with PTSD severity. Taken together, these findings suggest that self-report of mild TBI is either inaccurate or correlates poorly with cellular evidence of injury on DTI. Inaccurate self-report of mild TBI in the current study could certainly be possible, given the prolonged period of time (4 years on average) between the last tour of duty and recollection of mild TBI symptoms. Underreporting of mild TBI symptoms is also possible and has been reported by other researchers studying military cohorts.5
Even if self-report were accurate, however, a clinical diagnosis of mild TBI may not correlate well with underlying brain injury. Some mild TBI patients have no postconcussive symptoms, no demonstrable cognitive deficits, and normal neuroimaging, suggesting the absence of brain injury.49 On the contrary, some patients who suffer a blow to the head but do not meet the clinical definition of mild TBI do have objective evidence of brain injury.50,51 This poor correlation to brain injury likely confounds attempts to understand the relation between mild TBI and PTSD and underscores the urgent need to develop objective aids to the diagnosis through the use of biomarkers such as neuroimaging, electroencephalography, and serum protein tests.
Another important finding in the current study is the strong association between lower values of 1st percentile FA on DTI and blast exposure. Although others have found low FA among veterans with blast induced-mild TBI, we are the first to demonstrate disruption of WM integrity independent of a clinical diagnosis of mild TBI. Warden et al52 found reduced FA in the cerebellum of a single blast-injured soldier, whereas MacDonald et al26 found reduced FA in several ROIs among 63 OIF/OEF veterans with blast-related mild TBI. Davenport et al33 reported reduced whole-brain FA among 25 OIF/OEF service members with blast-related mild TBI (and, just as in our study, FA reduction was located primarily in the 1st percentile region). In all 3 studies, FA changes were found among those with a clinical diagnosis of mild TBI; the independent effect of blast exposure was not explored.
The observed association of blast exposure with abnormal DTI, independent of mild TBI, suggests that blast induces subclinical brain injury. Two recent civilian studies support this idea. In the first, significant WM damage on DTI was found after a sports season among 6 high school football and hockey players who did not report concussion50 In the second, regional cortical activation on functional MRI was significantly altered in 4 of 8 high school football players who did not report concussion after a season of play.51 In both studies, the extent of brain injury on neuroimaging correlated with the number of subconcussive head blows. While these studies do not involve blast as the mechanism of injury, they both support the emerging concept that brain injury can occur in the absence of symptoms typically used to define mild TBI such as loss of consciousness, amnesia, or a brief period of confusion.
It is unclear why normal MRI was associated with blast exposure in our analysis. Of the 5 subjects with WM abnormalities, none were exposed to blast. It is very likely that these WM lesions predated exposure to blast and are not related to trauma.
In summary, PTSD severity seems to be related not only to the severity of combat stress but also to underlying structural brain changes on MRI and DTI. These brain changes may be due to neurochemical alterations induced by chronic stress or, more likely, by subclinical brain injury from blast exposure. A clinical diagnosis of mild TBI does not seem to play a role, either because it is inaccurately reported or because it correlates poorly with underlying brain injury. Our findings suggest that physical forces such as blast may play a role in the genesis of PTSD in a combat environment. Both MRI and DTI may be uniquely suited for the detection of these structural brain changes and the identification of veterans at risk for PTSD. While screening thousands of service members with MRI or DTI may not be feasible, our results highlight the pressing need to develop practical mild TBI biomarkers (blood, electroencephalogram, etc) that correlate with underlying brain injury.
Our results should be interpreted in light of several limitations. First, the relatively small number of analyzed subjects may have reduced our ability to detect small but significant relations among DTI changes in individual brain ROIs and PTSD, blast exposure, and mild TBI. Rigorous control for multiple comparisons resulted in nominally significant covariate-adjusted ORs of several regions, including the cerebellum and corpus callosum, to rise above the P value threshold of .05. Interestingly, others have found DTI abnormalities in the cerebellum associated with blast exposure26,52 and in the corpus callosum among those with mild TBI.53 These regions might serve as the focus of future studies.
There is no validated method for determining cumulative blast exposure from self-report. The physical forces experienced by the brain after an explosion depend on the strength of the blast and its proximity to the subject. The method of determining blast exposure in the current study captured neither of these variables. An objective measure of blast exposure would greatly facilitate our understanding of the relation between multiple blast exposures and PTSD. Although several blast pressure sensors are currently under development, multiple technical hurdles remain.54
Finally, while we found 1st percentile FA in WM to be significantly associated with blast, FA values in this region have potential to be contaminated by GM and WM fiber crossings. Low FA values are typically found at the boundaries between WM and GM, and the segmentation algorithm performs poorly in these regions. However, to mitigate against this, we used empirical threshold values of FA less than 0.2 and MD more than 1.2 × 10−3 mm2/s during the creation of the whole-brain WM mask to remove voxels potentially contaminated by CSF and/or GM. Unlike FA, the 1st percentile of MD is relatively safe since the difference of MD values between GM and WM is very small. The values of FA may also be decreased by WM lesions,55 such as the ones seen in 5 subjects. However, when these 5 subjects were excluded from the logistic regression model, we found no statistically significant change in the OR for prediction of blast exposure.
PTSD severity was associated with high 1st percentile of MD on DTI, WM lesions on MRI, and the severity of exposure to combat events, in addition to age and time since last tour of duty. Self-report of mild TBI was not significantly associated with PTSD severity. We speculate that brain changes detected on MRI and DTI may be due to neurochemical alterations induced by chronic stress or by subclinical brain injury from blast exposure. Our observation that blast exposure was associated with low 1st percentile of FA on DTI supports the latter. While MRI and DTI may be uniquely suited for identifying veterans at risk for PTSD, their high cost and limited availability may limit their use as screening tools. Nevertheless, accurate and object measures of brain injury, practical or otherwise, are clearly needed to better understand the complex relations among blast exposure, combat stress, and PTSD.
1. Yehuda R. Post-traumatic stress disorder. N Engl J Med. 2002;346(2):108–114.
2. Milad MR, Pitman RK, Ellis CB, et al. Neurobiological basis of failure to recall extinction memory in posttraumatic stress disorder
. Biol Psychiatry. 2009;66(12):1075–1082.
3. Seal KH, Bertenthal D, Miner CR, Sen S, Marmar C. Bringing the war back home: mental health disorders among 103,788 US veterans returning from Iraq and Afghanistan seen at Department of Veterans Affairs facilities. Arch Intern Med. 2007;167(5):476–482.
4. Centers for Disease Control and Prevention, National Center for Injury Prevention and Control. Report to Congress. Mild Traumatic Brain Injury in the United States: Steps to Prevent a Serious Public Health Problem. Atlanta, GA: Centers for Disease Control and Prevention; 2003.
5. Tanielian T, ed. Invisible Wounds of War: Psychological and Cognitive Injuries, Their Consequences and Services to Assist Recovery. Washington, DC: The RAND Centre for Military Health Policy Research; 2008.
6. Hoge CW, McGurk D, Thomas JL, Cox AL, Engel CC, Castro CA. Mild traumatic brain injury in U.S. Soldiers returning from Iraq [comment]. N Engl J Med. 2008;358(5):453–463.
7. Schneiderman AI, Braver ER, Kang HK. Understanding sequelae of injury mechanisms and mild traumatic brain injury incurred during the conflicts in Iraq and Afghanistan: persistent postconcussive symptoms and posttraumatic stress disorder
. Am J Epidemiol. 2008;167(12):1446–1452.
8. Institute of Medicine. Gulf War and Health, Volume 7: Long-term Consequences of Traumatic Brain Injury. Washington, DC: The National Academies Press; 2009.
9. Bryant RA. Posttraumatic stress disorder
and mild brain injury: controversies, causes and consequences. J Clin Exp Neuropsychol. 2001;23(6):718–728.
10. Vasterling JJ, Verfaellie M, Sullivan KD. Mild traumatic brain injury and posttraumatic stress disorder
in returning veterans: perspectives from cognitive neuroscience. Clin Psychol Rev. 2009;29(8):674–684.
11. Bryant RA. Disentangling mild traumatic brain injury and stress reactions. N Engl J Med. 2008;358(5):525–527.
12. Powell JM, Ferraro JV, Dikmen SS, Temkin NR, Bell KR. Accuracy of mild traumatic brain injury diagnosis. Arch Phys Med Rehabil. 2008;89(8):1550–1555.
13. Rutland-Brown W, Langlois JA, Bazarian JJ, Warden D. Improving identification of traumatic brain injury after nonmilitary bomb blasts. J Head Trauma Rehabil. 2008;23(2):84–91.
14. Iverson GL. Clinical and methodological challenges with assessing mild traumatic brain injury in the military. J Head Trauma Rehabil. 2010;25(5):313–319.
15. Wessely S, Unwin C, Hotopf M, et al. Stability of recall of military hazards over time. Evidence from the Persian Gulf War of 1991. Br J Psychiatry. 2003;183:314–322.
16. Bremner JD, Southwick S, Brett E, Fontana A, Rosenheck R, Charney DS. Dissociation and posttraumatic stress disorder
in Vietnam combat veterans. Am J Psychiatry. 1992;149(3):328–332.
17. Cernak I, Noble-Haeusslein LJ. Traumatic brain injury: an overview of pathobiology with emphasis on military populations. J Cereb Blood Flow Metab. 2010;30(2):255–266.
18. Park E, Gottlieb J, Cheung B, Shek PN, Baker AJ. A model of low-level primary blast brain trauma results in cytoskeletal proteolysis and chronic functional impairment in the absence of lung barotrauma. J Neurotrauma. 2011;28:343–357.
19. Cernak I, Wang Z, Jiang J, Bian X, Savic J. Ultrastructural and functional characteristics of blast injury-induced neurotrauma. J Trauma. 2001;50(4):695–706.
20. Cernak I, Wang Z, Jiang J, Bian X, Savic J. Cognitive deficits following blast injury-induced neurotrauma: possible involvement of nitric oxide. Brain Inj. 2001;15(7):593–612.
21. Kwon SC, Kovesdi E, Gyorgy AB, Wingo D, et al. Stress and traumatic brain injury: a behavioral, proteomics, and histological study. Front Neurol. 2011;2(12):1–14.
22. Arfanakis K, Haughton VM, Carew JD, Rogers BP, Dempsey RJ, Meyerand ME. Diffusion tensor MR imaging in diffuse axonal injury. Am J Neuroradiol. 2002;23(5):794–802.
23. Bazarian JJ, Zhong J, Blyth B, Zhu T, Kavcic V, Peterson D. Diffusion tensor imaging detects clinically important axonal damage after mild traumatic brain injury: a pilot study. J Neurotrauma. 2007;24(9):1447–1459.
24. Benzinger TL, Brody D, Cardin S, et al. Blast-related brain injury: imaging for clinical and research applications: report of the 2008 St. Louis workshop. J Neurotrauma. 2009;26(12):2127–2144.
25. Levin HS, Wilde E, Troyanskaya M, et al. Diffusion tensor imaging of mild to moderate blast-related traumatic brain injury and its sequelae. J Neurotrauma. 2010;27:683–694.
26. MacDonald CL, Johnson AM, Cooper D, et al. Detection of blast-related traumatic brain injury in U.S. military personnel. N Engl J Med. 2011;364(22):2091–2100.
27. Blanchard EB, Jones-Alexander J, Buckley TC, Forneris CA. Psychometric properties of the PTSD Checklist (PCL). Behav Res Ther. 1996;34(8):669–673.
28. Forbes D, Creamer M, Biddle D. The validity of the PTSD Checklist as a measure of symptomatic change in combat-related PTSD. Behav Res Ther. 2001;39(8):977–986.
29. Lew HL, Poole JH, Vanderploeg RD, et al. Program development and defining characteristics of returning military in a VA Polytrauma Network Site. J Rehabil Res Dev. 2007;44(7):1027–1034.
30. Zhu T, Liu X, Gaugh MD, et al. Evaluation of measurement uncertainties in human diffusion tensor imaging (DTI)-derived parameters and optimization of clinical DTI protocols with a wild bootstrap analysis. J Magn Reson Imaging. 2009;29(2):422–435.
31. Smith SM, Jenkinson M, Woolrich MW, et al. Advances in functional and structural MR image analysis and implementation as FSL. Neuroimage. 2004;23(suppl 1):S208–S219.
32. Oishi K, Zilles K, Amunts K, et al. Human brain white matter atlas: identification and assignment of common anatomical structures in superficial white matter. Neuroimage. 2008;43(3):447–457.
33. Davenport ND, Lim KO, Armstrong MT, Sponhein SR. Diffuse and spatially variable white matter disruptions are associated with blast-related mild traumatic brain injury. NeuroImage. 2012;59(3):2017–2024. doi:10.1016/j.neuroimage.2011.10.050.
34. Radley JJ, Rocher AB, Rodriguez A, et al. Repeated stress alters dendritic spine morphology in the rat medial prefrontal cortex. J Comp Neurol. 2008;507(1):1141–1150.
35. Cerqueira JJ, Taipa R, Uylings HB, Almeida OF, Sousa N. Specific configuration of dendritic degeneration in pyramidal neurons of the medial prefrontal cortex induced by differing corticosteroid regimens. Cerebral Cortex. 2007;17(9):1998–2006.
36. Abe O, Yamasue H, Kasai K, et al. Voxel-based diffusion tensor analysis reveals aberrant anterior cingulum integrity in posttraumatic stress disorder
due to terrorism. Psychiatry Res. 2006;146(3):231–242.
37. Schuff N, Zhang Y, Zhan W, et al. Patterns of altered cortical perfusion and diminished subcortical integrity in posttraumatic stress disorder
: an MRI study. Neuroimage. 2011;54(suppl 1):S62–S68.
38. Kim SJ, Jeong DU, Sim ME, et al. Asymmetrically altered integrity of cingulum bundle in posttraumatic stress disorder
. Neuropsychobiology. 2006;54(2):120–125.
39. Saljo A, Bao F, Haglid KG, Hansson HA. Blast exposure causes redistribution of phosphorylated neurofilament subunits in neurons of the adult rat brain. J Neurotrauma. 2000;17(8):719–726.
40. Hull AM. Neuroimaging findings in post-traumatic stress disorder. Systematic review. Br J Psychiatry. 2002;181:102–110.
41. Karl A, Schaefer M, Malta LS, Dorfel D, Rohleder N, Werner A. A meta-analysis of structural brain abnormalities in PTSD. Neurosci Biobehav Rev. 2006;30(7):1004–1031.
42. Canive JM, Lewine JD, Orrison WW, et al. MRI reveals gross structural abnormalities in PTSD. Ann N Y Acad Sci. 1997;821:512–515.
43. Kruit MC, van Buchem MA, Hofman PA, et al. Migraine as a risk factor for subclinical brain lesions. JAMA. 2004;291(4):427–434.
44. van Dijk EJ, Breteler MM, Schmidt R, et al. The association between blood pressure, hypertension, and cerebral white matter lesions: cardiovascular determinants of dementia study. Hypertension. 2004;44(5):625–630.
45. Lyoo IK, Streeter CC, Ahn KH, et al. White matter hyperintensities in subjects with cocaine and opiate dependence and healthy comparison subjects. Psychiatry Res. 2004;131(2):135–145.
46. Katzman GL, Dagher AP, Patronas NJ. Incidental findings on brain magnetic resonance imaging from 1000 asymptomatic volunteers. JAMA. 1999;282(1):36–39.
47. Cassiday KL. Recall of traumatic memories following cerebral vascular accident. J Traumat Stress. 1992;5(4):627–631.
48. Hoggard N, Darwent G, Capener D, Wilkinson ID, Griffiths PD. The high incidence and bioethics of findings on magnetic resonance brain imaging of normal volunteers for neuroscience research. J Med Ethics. 2009;35(3):194–199.
49. Bazarian JJ, Blyth B, Cimpello L. Bench to bedside: evidence for brain injury after concussion—looking beyond the computed tomography scan. Acad Emerg Med. 2006;13(2):199–214.
50. Bazarian JJ, Zhu T, Blyth B, Borrino A, Zhong Z. Subject-specific changes in brain white matter on diffusion tensor imaging after sports-related concussion. Magn Reson Imaging. 2012;30(2):171–180.
51. Talvage TM, Nauma E, Breedlove EL, Yoruk U, et al. Functionally-detected cognitive impairment in high school football players without clinically-diagnosed concussion [published online ahead of print October 1, 2011]. J Neurotrauma. doi:10.1089/neu.2010.1512.
52. Warden DL, French LM, Shupenko L, et al. Case report of a soldier with primary blast brain injury. Neuroimage. 2009;47(suppl 2):T152–T153.
53. Niogi SN, Mukherjee P. Diffusion tensor imaging of mild traumatic brain injury. J Head Trauma Rehabil. 2010;25(4):241–255.
55. Iverson GL, Hakulinen U, Waljas M, et al. To excluded or not to exclude: white matter hyperintensities in diffusion tensor imaging research. Brain Inj. 2111;25(13/14):1325–1332.