Effect sizes of posttraumatic stress disorder (PTSD) diagnosis at parcels where statistically significant parcellated. Left lateral, medial, and inferior views of average parcellations of subjects with PTSD and subjects without PTSD with color encoding PTSD diagnosis effect sizes (η2) on cortical volume in those regions for which the probability of type I error was α<.0028. The presentation of the right hemisphere is arbitrary because statistics were limited to cortical volumes summed over both hemispheres (see text). Only 3 colored regions are obvious because the latter 2 are adjacent to one another and produced nearly equivalent effect sizes.
Scatterplots depicting the between-group and within-group distributions underlying findings of regionally smaller cortical volume at the parahippocampal gyrus, superior temporal cortex, lateral division of the orbital frontal cortex, and pars orbitalis of the inferior frontal gyrus. Least-squares means are indicated by vertical bars. PTSD indicates posttraumatic stress disorder.
Cortical thickness: effect of posttraumatic stress disorder (PTSD). Statistical difference maps of PTSD contrast on cortical thickness at P < .01 (uncorrected).
Customize your JAMA Network experience by selecting one or more topics from the list below.
Woodward SH, Schaer M, Kaloupek DG, Cediel L, Eliez S. Smaller Global and Regional Cortical Volume in Combat-Related Posttraumatic Stress Disorder. Arch Gen Psychiatry. 2009;66(12):1373–1382. doi:10.1001/archgenpsychiatry.2009.160
Two sets of findings predict smaller cerebral cortical gray matter volume in adult posttraumatic stress disorder (PTSD). Measures of intracranial tissue volume and cerebral tissue volume have been observed to be smaller in adolescents with maltreatment-related PTSD. Second, lower intelligence, a risk factor for PTSD, is associated with smaller cerebral tissue volumes. Nevertheless, to our knowledge, only 1 study has observed globally smaller cerebral tissue volume in adults with PTSD.
To apply a recently developed method providing improved estimates of cortical volume and to estimate associations between adult PTSD and selected regional cortical volumes not yet investigated.
Between-group comparison of global and regional cerebral cortical volumes in adult patients with combat-related PTSD and controls.
Two Department of Veterans Affairs medical centers with large inpatient and outpatient PTSD catchments.
Ninety-seven combat-exposed veterans of the Vietnam and Persian Gulf wars.
Main Outcome Measure
Global and regional cortical volumes determined using the FreeSurfer software program and the Desikan et al parcellation (modified).
Cerebral cortical volume, thickness, and area were observed to be smaller in association with adult combat-related PTSD. Robust associations were observed between PTSD and smaller cortical volumes in the parahippocampal gyrus, superior temporal cortex, lateral orbital frontal cortex, and pars orbitalis of the inferior frontal gyrus.
Cerebral cortical volume, thickness, and area may be smaller in adult chronic severe PTSD; however, the extracted structural variables did not mediate relations between intelligence and PTSD. The 4 regions exhibiting especially smaller cortical volumes in this sample share involvement in mechanisms subserving “top-down” facilitation of the identification of objects and words. Compromise of these regions may result in difficulty in relearning pretrauma schemata for interpreting the civilian physical and social environments.
Adult posttraumatic stress disorder (PTSD) has been associated with smaller hippocampal and anterior cingulate cortical (ACC) volumes,1-11 but to our knowledge, only 1 study has found smaller total brain volume.12 In contrast, 3 studies in children and adolescents with PTSD have found smaller intracranial (gray + white + cerebrospinal fluid) volume13,14 or smaller cerebral tissue (gray + white) volume.15,16 No explanation has been advanced for this discontinuity between children and adults. Analogous relationships between cerebral volumes and developmental adversity appear common to children and adults. In maltreated children, De Bellis et al13,14 twice found a direct linear relationship between intracranial volume and age at abuse onset. In adults with PTSD, Woodward et al17 found a similar relationship between a bone-based measure of cranial volume and age at first trauma, while Fennema-Notestine et al18 found an inverse relationship between intracranial volume and scores on the Childhood Trauma Questionnaire. However, neither of these studies found a main effect of PTSD on a measure of intracranial or cerebral tissue volume. The absence of such an effect in adult PTSD is also surprising in light of associations between PTSD and IQ19-24 and between IQ and cerebral tissue volume.25-29
There are challenges associated with comparing cerebral tissue volumes over groups. While studies of hippocampal volume routinely adjust for variation in body size by using a cerebral macrovolume as a covariate, this is not possible when that macrovolume is, itself, the focus of inquiry. Second, numerous studies have reported that diagnoses frequently comorbid with PTSD, such as alcoholism, are themselves associated with smaller cerebral tissue volumes.30-32 Controlling such effects requires enlarged samples. A further obstacle to settling this issue may be inherent in the phenomenon itself. Bossini et al33 found, in a relatively large sample of adult civilian trauma survivors meeting criteria for PTSD and free of comorbid psychopathology, that while whole-brain gray matter volume was smaller in PTSD, whole-brain white matter volume was larger. If this is the case, studies of unsegmented cortical tissue volumes in adult PTSD may be expected to produce ambiguous results.
Taking advantage of recent advances in the structural analysis of magnetic resonance images, this study revisited the status of cerebral cortical tissue volume in the same sample of 99 Vietnam and Gulf war veterans our group reported on previously.6,10,17 These methodological advances enabled fully automated extraction of the 2-dimensional cortical sheet and estimation of its thickness, area, and volume.34-36 A key innovation of this approach is its use of a priori knowledge of the geometry of the gray matter–white matter boundary, which is approximately planar over short distances. The validity of thickness estimates has been supported by comparisons with postmortem tissue.37 The thickness estimates have also replicated well-known global and local effects of normal aging on the cerebral cortex.38 Importantly, this method is relatively robust to variation in magnetic resonance manufacturer and software revision,39,40 an attractive feature in a study such as this that aggregates samples across sites.
In this study, we relied on the method of Dale et al35 and Fischl et al36 embodied in the FreeSurfer software package (version 4.0.1; Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Boston). A second component of the approach of Dale et al35 and Fischl et al36 incorporated in FreeSurfer is the use of multiresolution deformations to construct “inflated” or spherical transformations of the cortex that expose sulci. After reprojection, cross-subject registration can proceed by reference to cortical landmarks, such as fully exposed sulci,41,42 enabling the construction of more precise group average cortices than are obtainable with 3-dimensional methods.41 The precision of this coregistration is also carried over into any atlas-based parcellation. The analysis of cortical structure by parcels established a priori is attractive because it enables conventional methods of type I error control.43,44 This study used the “gyrographic” parcellation of Desikan et al.45 Parcelwise analyses were followed by an exploratory whole-brain vertex-wise analysis of cortical thickness to mitigate the limited ability of parcelwise analysis to detect group differences not spatially correspondent to predefined regions. Lastly, using FreeSurfer-derived parameters, we tested the proposition that cortical structure mediates relations between 2 Wechsler Adult Intelligence Scale (WAIS)46 subtest scores and PTSD severity.
The sample for this study included 97 of 99 participants reported on in 3 previous articles describing regional brain volumes in adult combat-related PTSD.6,10,17 Two participants were excluded because the cortical modeling algorithm failed to converge owing to artifact. Fifty participants with PTSD (37 Vietnam and 13 Gulf war veterans) met criteria for current PTSD as a result of experiencing 1 or more military traumas. Forty-seven subjects without PTSD (24 Vietnam and 23 Gulf war veterans) were free of diagnosable PTSD, current or lifetime. The PTSD diagnoses and severities were based on the Clinician-Administered PTSD Scale47 and remaining diagnoses, on the Structured Clinical Interview for DSM-IV.48 Approximately one-half of participants were classified as having lifetime (but not current) alcohol abuse/dependence. Also administered were the Combat Exposure Scale,49 the Beck Depression Inventory,50 and the vocabulary and digit symbol substitution subtests of the WAIS (Table 1).
Psychotropic medications were not discontinued. Seventy percent of the participants with PTSD were taking some form of psychotropic medication vs 10.6% of participants without PTSD. Fifty-six percent of participants with PTSD were taking selective serotonin reuptake inhibitors vs 4 percent of participants without PTSD. Twenty-four percent of participants with PTSD were taking anticonvulsant/mood-stabilizing medications, while no participants without PTSD were taking a medication of this class.
Participants provided written informed consent in accordance with procedures of the institutional review boards of Stanford University Medical School/VA Palo Alto Health Care System or Boston VA Medical Center and the McLean Hospital.
Magnetic resonance images of subjects' brains were acquired with 1 of two 1.5-T GE Signa systems (GE Healthcare, Milwaukee, Wisconsin) located at the Diagnostic Radiology Center of VA Palo Alto Health Care System or the Brain Imaging Center of McLean Hospital (Belmont, Massachusetts). Coronal images were acquired with a 3-dimensional spoiled gradient recalled volumetric pulse sequence with the following parameters: repetition time = 35 milliseconds, echo time = 6 milliseconds, flip angle = 45°, number of excitations = 1, matrix size = 256 × 192, field of view = 24 cm2, slice thickness = 1.5 to 1.7 mm, and slices = 124.
Only brief summaries of the methods incorporated in FreeSurfer follow. Detailed descriptions may be found in the developers' original articles.35,36,51 Image preprocessing included resampling into cubic voxels, correction of inhomogeneity artifact, and skull stripping. Constrained region growing was used to create a unitary white matter volume for each hemisphere. The surfaces of the white matter volumes were overlain with tessellations constructed by assigning 2 triangles to the square face of each surface voxel, yielding approximately 160 000 vertices per hemisphere. FreeSurfer optimized this tessellation by minimizing an energy function of intensity and both normal and tangential smoothness.35,36 The pial surfaces of each hemisphere were computed by deforming the white matter surfaces toward the gray matter–cerebrospinal fluid boundary. Cortical thickness was computed as the distance between the inner and the outer cortical surfaces at each vertex.51
In FreeSurfer, cross-subject registration of hemispheric cortical surfaces was performed by projecting them onto spherical representations that minimized an energy function of concavity and cartographic distortion. This step is anchored to large-scale features of the cortex such as the central sulcus and sylvian fissure,36,41 which are also used to anchor the alignment of individual subjects' hemispheric surface models with an iteratively calculated study-specific average cortex.41Following alignment, participants' cortical thickness values were resampled into a common spherical coordinate system and smoothed using a gaussian kernel with full-width at half maximum of 10 mm.
The Desikan et al45 parcellation used to delineate cortical regions was based on a diverse training sample of 40 brains. These brains were first parcellated manually by a neuroanatomist using FreeSurfer-derived cortical surface representations with final sulcal delineations performed on “inflated” projections allowing full visualization of sulci. The probability of a cortical location falling into a specific cortical region was modeled by an anisotropic, nonstationary Markov random field incorporating global and local structural information and interregional constraints. A computational procedure determining the parcellation pattern with the highest probability given all priors ran iteratively until a stabile pattern emerged. Additional details regarding the mathematic underpinnings and extensibility of this approach can be found in Fischl et al.44
These automated procedures required approximately 24 hours of computation per brain on a 2 × 3-GHz quad-core Macintosh workstation. After processing, both lateral and medial surfaces of each parcellated cortical hemisphere were manually reviewed for gross artifacts. None were found.
To provide a brain tissue–based index that could be used to adjust for nuisance variance in body size, we imported estimates of cerebral white matter volume obtained from the same magnetic resonance image set using BrainImage and the protocols of the Stanford Psychiatric Neuroimaging Program (BrainImage 5.x; A. L. Reiss, Stanford University, Stanford, California). These estimates were independently derived and based on a nonoverlapping set of manual and automated procedures.6,52
To provide robust protection against type I error, the parcellated data were reduced in dimensionality to satisfy to the assumption of multivariate analysis of variance that the number of within-subjects factors be fewer than the number of cases in the smallest cell.53 Three strategies were used. Whereas prior analyses of these data used cohort (Vietnam vs Gulf war) as a factor, in this study we covaried for age. Second, homologous parcel volumes were summed across hemispheres. Third, 15 parcels were collated into a single superparcel. In the Desikan nomenclature, the constituents of this superparcel were as follows: lateral occipital cortex, inferior parietal cortex, superior parietal cortex, inferior temporal gyrus, middle temporal gyrus, banks of the superior temporal sulcus, supramarginal gyrus, postcentral gyrus, precentral gyrus, fusiform gyrus, lingual gyrus, pericalcarine cortex, cuneus cortex, precuneus cortex, and paracentral lobule. In addition, the transverse temporal cortex was merged with the surrounding superior temporal gyrus. This approach retains detailed parcellation of all cortex anterior to the motor strip and most limbic cortical regions, domains that have received the lion's share of empirical support for structural and functional cerebrocortical compromise in PTSD. In the context of a significant multivariate effect, between-subjects effects of PTSD at individual parcels were required to meet a Bonferroni-adjusted criterion of α<.0028 (0.05/18).53 Additional planned comparisons were performed on parcels corresponding to regions that have demonstrated structural and/or functional compromise in PTSD, the ACC,7,10,54-59 and the superior temporal gyrus.60 (The Desikan parcellation does not include the insular cortex or separately quantify the subgenual ACC.) Tests of adjustment for stature and cerebral white matter volume were accomplished with analysis of covariance and multivariate analysis of covariance. Neither stature (with PTSD, 172.5 cm; without PTSD, 173.25 cm) nor cerebral white matter volume (with PTSD, 512 mm3; without PTSD, 518 mm3) exhibited a significant effect of PTSD diagnosis (F's < 1).61
Neither cortical volume, thickness, nor area exhibited an effect of study site (all F's < 1).
Our analyses first considered parcellated cortical volume, which combines both thickness and area information, because volumetric indexes have been the subject of most prior research in this area. Total cortical volume summed over all parcels and analyzed with univariate analysis of variance covarying for age exhibited a significant main effect of PTSD (F1,92 = 10.66; P < .01; partial η2 = 0.104), being smaller in subjects with PTSD (350 744 vs 368 896 mm3). Cortical volume also tended to be smaller in subjects with lifetime alcohol abuse/dependence (F1,92 = 3.59; P = .06) but exhibited no PTSD × lifetime alcohol abuse/dependence interaction (F<1). The PTSD contrast was sharpened by adjustment for stature and cerebral white matter volume (F1,90 = 14.40; P < .001; partial η2 = 0.138), while effects of lifetime alcohol abuse/dependence (F1,90 = 3.04; P = .09) and the PTSD × lifetime alcohol abuse/dependence interaction (F<1) remained insignificant. Parcellated cortical volume exhibited a significant multivariate effect of PTSD (Wilks λ = 0.61; F18,75 = 2.69; P = .002; partial η2 = 0.39). There were no multivariate effects of lifetime alcohol abuse/dependence (Wilks λ = 0.83; F18,75 = 0.86; P = .63) or the PTSD × lifetime alcohol abuse/dependence interaction (Wilks λ = 0.85; F18,75 = 0.75; P = .75). The multivariate effect of PTSD (Wilks λ = 0.58; F18,73 = 2.95; P < .001; partial η2 = 0.42), lifetime alcohol abuse/dependence (Wilks λ = 0.81; F18,73 = 0.96; P = .25), and their interaction (Wilks λ = 0.85; F18,73 = 0.73; P = .33) was not substantially modified by adjustment for stature and cerebral white matter volume. Parcels meeting the Bonferroni-corrected significance criterion were, in order of effect size, the parahippocampal gyrus (P < .001; partial η2 = 0.181; approximating Brodmann area 36), superior temporal cortex (P = .001; partial η2 = 0.121; encompassing Brodmann areas 22, 41, and 42), lateral division of the orbital frontal cortex, (P = .002; partial η2 = 0.103; lateral aspect of Brodmann area 11), and pars orbitalis of the inferior frontal gyrus (P = .002; partial η2 = 0.101; approximating Brodmann area 47). All effects were in the direction of smaller cortical volumes in subjects with PTSD (Figure 1 and Figure 2) and (Table 2). The associations of PTSD with smaller volumes at the parahippocampal gyrus (P < .001; partial η2 = 0.192), superior temporal cortex (P < .001; partial η2 = 0.148), lateral division of the orbital frontal cortex (P = .001; partial η2 = 0.112), and pars orbitalis of the inferior frontal gyrus (P = .002; partial η2 = 0.098) all remained significant after adjustment for stature and cerebral white matter volume.
Planned comparisons at parcels corresponding to the rostral and caudal ACC (generally corresponding to pregenual and dorsal subregions of ACC, respectively) both found significant but relatively modest effects of PTSD (rostral division: F1,92 = 4.83; P = .03; partial η2 = 0.050; caudal division: F1,92 = 6.55; P = .01; partial η2 = 0.066). Findings at both the rostral and caudal ACC remained significant after adjustment for stature and cerebral white matter volume (rostral division: F1,90 = 4.85; P = .03; partial η2 = 0.051; caudal division: F1,90 = 6.26; P = .01; partial η2 = 0.065). A planned comparison of cortical volume at the superior temporal cortex was moot.
Weighted mean cortical thickness calculated over the whole brain was analyzed with univariate analysis of variance covarying for age and found to be significantly lower in subjects with PTSD than in subjects without PTSD (2.384 mm vs 2.428 mm, respectively; F1,92 = 4.69; P = .03; η2 = 0.049). There were no effects of lifetime alcohol abuse/dependence or the PTSD × lifetime alcohol abuse/dependence interaction on cortical thickness (F's <1). Analyses of cortical thickness by parcel found no multivariate effect of PTSD (Wilks λ = 0.71; F18,75 = 1.66; P = .07; partial η2 = 0.285), lifetime alcohol abuse/dependence (Wilks λ = 0.835; F18,75 = 0.82; P = .67), or their interaction (Wilks λ = 0.78; F18,75 = 1.20; P = .28) (Table 3). Planned comparisons at the rostral and caudal divisions of the ACC found significant effects in the direction of thinner cortex in PTSD (rostral division: F1,92 = 7.72; P = .007; partial η2 = 0.077; caudal division: F1,92 = 4.97; P = .03; partial η2 = 0.051). Planned comparisons also found thinner cortex in participants with PTSD at the superior temporal cortex (F1,92 = 4.264; P = .04; partial η2 = 0.044). These results were unchanged by adjustment for stature and cerebral white matter volume.
Adjusted for age, total cortical area tended to be smaller in PTSD (F1,92 = 3.79; P = .06) but exhibited no effect of lifetime alcohol abuse/dependence (F1,92 = 1.03; P = .16) and no PTSD × lifetime alcohol abuse/dependence interaction (F<1). Adjustment for stature and cerebral white matter volume rendered the effect of PTSD significant (F1,90 = 7.29; P = .34; partial η2 = 0.075), but not effects of lifetime alcohol abuse/dependence or the PTSD × lifetime alcohol abuse/dependence interaction (lifetime alcohol abuse/dependence: F1,90 = 1.72; P = .34; PTSD × lifetime alcohol abuse/dependence: F<1). Cortical area exhibited no multivariate effects of PTSD (Wilks λ = 0.78; F18,75 = 1.16; P = .32), lifetime alcohol abuse/dependence (Wilks λ = 0.78; F18,75 = 1.15; P = .32), or the PTSD × lifetime alcohol abuse/dependence interaction (Wilks λ = 0.78; F18,75 = 1.9; P = .29), a result not modified by adjustment for body size. Among planned comparisons, only the superior temporal cortex exhibited an effect of PTSD (F1,92 = 6.14; P = .02; partial η2 = 0.063), being smaller in participants with PTSD, a result strengthened by adjustment for body size (F1,92 = 9.44; P = .003; partial η2 = 0.095).
Though the analysis of a priori parcels permitted a conventional approach to type I error control, type II–like errors may have resulted when regional differences between groups did not correspond closely to parcels because of spatial restriction, spanning parcel boundaries, or both. Figure 3 presents cortical thickness difference maps contrasting participants with and without PTSD (at P < .01 uncorrected). Foci of cortical thinning were apparent within the boundaries of the parahippocampal and lateral orbital frontal parcels and dorsal anterior cingulate, bilaterally, within the superior temporal gyrus on the left and within the pars orbitalis of the inferior frontal gyrus on the right. These locations were generally correspondent with the parcelwise effects on volume and thickness. It is also apparent, however, that the superparcel spanned multiple regions of thinning in the parietal cortex on the right and in the lateral occipital/posterior temporal cortex bilaterally. As well, parcels containing foci of apparent thinning in the left lateral frontal cortex were summed with homologous parcels in the right hemisphere absent thinning.
In light of data relating intelligence to both PTSD and brain volume, we examined whether parameters extracted with FreeSurfer were compatible with the possibility that cortical structure mediates relations between intelligence and PTSD severity. Linear regression was first used to establish the direct path between WAIS subtest scores and PTSD severity. Here, as in prior studies, estimators of intelligence accounted for additional variance in PTSD (Fchange2,92 = 17.42; P < .001) after entry of combat exposure. In the final model, all predictors were significant (Combat Exposure Scale score: β = 0.346; t = 4.01; P < .001; WAIS vocabulary score: β = −0.311; t = 3.75; P < .001; WAIS digit symbol substitution score: β = −0.279; t = 3.07; P = .003; Fegression4,91 = 18.03; P < .001; adjusted R2 = 0.418). Second, we examined correlations between predictors and proposed mediators, finding that cortical volume correlated with both vocabulary (r = 0.220; P = .03) and digit symbol substitution (r = 0.351; P < .001) score. Next, we performed a Sobel test62 of the reduction in predictive strength resulting from entry of cortical volume into the model separately for WAIS vocabulary and digit symbol substitution scores. In both cases, results of the test were negative (vocabulary score: Sobel test statistic = 1.72; P = .09; digit symbol substitution score: Sobel test statistic = 1.67; P = .10). Noting that the correlation between mean cortical thickness (weighted by parcel area) and total cortical area was mildly negative in this sample (r95 = −0.21; P = .04), we reconsidered these parameters separately. Examination of correlations between these predictors and proposed mediators disclosed that cortical thickness correlated with digit symbol substitution score (r = 0.276; P = .006), but not vocabulary score (r = −0.008; P = .94), while total area correlated with vocabulary score (r = 0.257; P = .01), but not digit symbol substitution score (r = 0.192; P = .06). Results were comparable with those obtained with volume (vocabulary score − > cortical area − > PTSD severity: Sobel test statistic = 0.204; P = .84; digit symbol substitution score − > cortical thickness − > PTSD severity: Sobel test statistic = 1.73; P = .79).
This study of cerebral cortical structure in combat veterans using the method of Dale et al35 and Fischl et al36embodied in FreeSurfer combined with the Desikan parcellation found evidence that cerebral cortical volume is smaller in adults with PTSD. The cerebral cortex of participants with PTSD was also significantly thinner, the estimated difference of 0.041 mm being similar to that found in a comparison of subjects with animal phobia with controls.63 Total cortical area was also smaller in the subjects with PTSD after adjustment for stature and cerebral white matter volume. These results are in general agreement with studies of adolescents with PTSD that have found smaller macrostructural brain volumes compared with matched controls.13-15 Their divergence from our prior findings of a trend only in favor of smaller cerebral tissue volume in PTSD6,10 is likely related to the current focus on cortical gray matter exclusively. A similar argument might apply to the absence of effects of comorbid alcoholism when PTSD and age were controlled. The finding of a smaller cortex is also broadly compatible with studies implicating lower intelligence as a risk factor for PTSD; however, the parameters and psychometrics assessed herein did not support the proposition that cortical structure mediates the relationship between intelligence and PTSD as part of a causative scenario.
This study found evidence that the association of PTSD with smaller cortical volume is not uniform across the brain. Especially strong effects of PTSD were observed at the parahippocampal gyrus, superior temporal cortex, pars orbitalis of the inferior frontal gyrus, and adjacent orbital frontal cortex. This pattern is an intriguing one. The parahippocampal gyrus is juxtaposed to the hippocampus, the region of the brain first identified as abnormal in adult PTSD.1 In humans, functional imaging studies have linked the parahippocampal gyrus to the representation of space,64 a role it may have taken over from the hippocampus, in which “place” cells were first identified in rodents.64 Recent results suggest that in humans the hippocampus and parahippocampal gyrus collaborate in encoding the interrelations of salient objects within space65 and in contextualization more generally.66,67 Misperception of safe vs threatening contexts has been proposed to contribute to PTSD.68-71 Relevant rodent models have also invoked the concept of context as represented in the hippocampus proper.72-74
These observations may be translatable into concrete predictions regarding the performance of persons with PTSD on a set of cognitive tasks not previously used in this area. In a rich series of studies, Bar and Aminoff75,76 developed the proposition that the representation of context, that is, of regularities in object co-occurrence and location in the environment, contributes to facilitated visual object identification in a “top-down” fashion. Bar77 also identified a second substrate for facilitated object identification in the ventrolateral and orbital frontal cortex, regions also observed herein to be smaller in subjects with PTSD. Bar et al78 suggest this second substrate generates an “initial guess” as to object identity that is based on a low–spatial-frequency version of the stimulus transmitted rapidly to the ventrolateral orbitofrontal cortex over the dorsal visual pathway. These “guesses” are then projected back to the inferotemporal cortex where they converge with “bottom-up” information resulting in facilitated identification. These 2 mechanisms appear to cooperate,67,79 and the parahippocampal gyrus and orbital frontal cortex are known to be heavily interconnected.80-83 In summary, the current results are suggestive of structural compromise in PTSD of multiple components—hippocampal, parahippocampal, and orbital frontal—of a distributed system subserving facilitated object identification. Behavioral tasks assessing facilitated object identification may therefore be sensitive to PTSD. The cognitive psychology literature contains a number of tasks, such as the primed object/nonobject discrimination of Gronau et al,79 that may be used in testing this hypothesis.84-86
Superior temporal cortical volume was also found to be smaller in participants with PTSD in this study. Tests of associations between PTSD and cortical structure in this region were planned a priori because De Bellis et al60,87 reported larger cortical volumes in this region in adolescents with maltreatment-related PTSD. These 2 results are not necessarily incompatible, as the cortex exhibits both linear and quadratic trends during maturation such that a specific region could be initially larger and ultimately smaller than normal in a group undergoing atypical development.88 Moreover, the superior temporal parcel overlaps temporal cortical regions that have exhibited inverse correlations between IQ and cortical thickness during childhood.88 It is especially relevant, here, that the superior temporal cortex has also been implicated in top-down facilitation of stimulus identification, in this case, of words. Lexical decision, the discrimination of words from nonwords, a verbal analog of the Gronau et al task, is facilitated on trials preceded by primes semantically related to the words. Both functional magnetic resonance imaging and electroencephalographic studies have implicated the superior temporal cortex, especially in the left hemisphere, in such effects.89,90 The obverse of semantic priming, what might be termed semanticinterference, has been repeatedly demonstrated in PTSD in studies using the emotional Stroop task.91 These studies have found evidence that, in participants with PTSD, trauma-related semantic content interferes with the performance of a cover task. Vythilingam et al92 recently extended this literature by demonstrating that exaggerated Stroop interference is accompanied, in PTSD, by facilitated lexical decisions when prime and test words share negative or trauma-relevant content. How might we rationalize the co-occurrence of facilitated semantic priming of trauma-related content in PTSD despite structural compromise of a brain region subserving this function? One possibility is that structural compromise results in reduced plasticity impeding the recovery of non–trauma-related semantic contextualizations. By analogy, structural/functional compromise of the hippocampus, parahippocampal gyrus, and ventrolateral orbital frontal cortex could impede the recovery of pretrauma function of top-down mechanisms facilitating object identification. One result would be the persistence of trauma-conditioned top-down influences on the identification of objects and words in non–trauma-related settings, a phenomenologically and clinically salient feature of PTSD.
A limitation of a priori parcellation is its insensitivity to regional findings not corresponding to parcel boundaries. Herein, an exploratory vertex-wise analysis disclosed “missed” foci of cortical thinning in the left frontal, right parietal, and bilateral occipital/posterior temporal cortex. These regions may deserve future consideration. A second limitation of a priori parcellation is that different regions of the cortex are characterized by more or less consistent landmarks. The sulcal determinants of ACC are notably variable93,94 and may represent a special challenge to automated parcellation. Our inspection of ACC parcels confirmed such variability, especially in anterior-posterior extents. The coefficients of variation of the areas of the right and left rostral ACC were the highest and third highest in their respective hemispheres. The absence of larger effects of PTSD at parcels corresponding to ACC should be considered in this light. Another important limitation of this study is the fact that the PTSD and non-PTSD groups differed in their rates of psychotropic medication use. Studies have not yet addressed the effects of the medications (principally selective serotonin reuptake inhibitors) administered to these participants with PTSD on cortical structure. In schizophrenia, results of 3 studies of the effects of antipsychotic use on cortical thickness have been negative95; however, some earlier studies quantifying volume have found reductions attributable to drug use.96
Correspondence: Steven H. Woodward, PhD, Dissemination and Training Division, National Center for PTSD, Mail Code 334 PTSD, VA Palo Alto Health Care System, 3801 Miranda Ave, Palo Alto, CA 94304 (firstname.lastname@example.org).
Submitted for Publication: March 17, 2008; final revision received May 29, 2008; accepted May 30, 2008.
Financial Disclosure: None reported.
Funding/Support: Major funding for this work was provided under a Department of Veterans Affairs/Department of Defense assistance agreement to Drs Woodward and Kaloupek from the US Army Medical Research and Materiel Command (Department of Defense) administered through the Institute for Medical Research (Department of Veterans Affairs). Additional support was provided by Swiss National Research Funds grants 323500-111165 (Dr Schaer) and 3200-063135, 3232-063134, and PP00B-102864 (Dr Eliez).
Additional Contributions: Stephen Grate and Lori Moll facilitated the necessary arrangements at the Department of Veterans Affairs/Department of Defense. Allan L. Reiss, MD, Perry K. Renshaw, MD, Wendy Stegman, BA, Lorraine Leskin, PhD, Kelly Teresi, MA, Catherine Kutter, PhD, Rebecca Prestel, BA, Rosemond Villafuerte, MA, Ann Smith, Eileen Connolly, and Michael A. Dove, MBA, provided institutional and technical support. We thank the Research Services of the VA Palo Alto and VA Boston health care systems for their support.