Regional cerebral volumes in veterans with current versus remitted posttraumatic stress disorder

Abstract

We previously reported that hippocampal volume was associated with current, but not lifetime posttraumatic stress disorder (PTSD) symptom severity. In the present study, we test the hypothesis that like the hippocampus, the volumes of other brain regions previously implicated in PTSD, are also negatively related to current, but not lifetime PTSD symptom severity. One hundred ninety-one veterans underwent structural magnetic resonance imaging (MRI) on a 4 T scanner. Seventy-five veterans were trauma unexposed, 43 were trauma exposed without PTSD, 39 were trauma exposed with current PTSD, and 34 were trauma exposed veterans with remitted PTSD. Hippocampal, amygdala, rostral and caudal anterior cingulate, insula, and corpus callosum volumes, quantified with Freesurfer version 4.5, were analyzed by group using multivariate analysis of covariance. Veterans with PTSD had smaller hippocampal, caudal anterior cingulate, insula, and corpus callosum volumes than the unexposed controls (p≤0.009); smaller hippocampal, caudal anterior cingulate, insula (p≤0.009) and marginally smaller corpus callosum (p=0.06) than veterans with remitted PTSD; and smaller hippocampal and caudal anterior cingulate volumes than veterans without PTSD (p≤0.04). In contrast, there was no significant volume differences between veterans with remitted PTSD compared to those without PTSD or unexposed controls. The finding that current but not lifetime PTSD accounts for the volumes of multiple brain regions suggests that either smaller brain volume is a vulnerability factor that impedes recovery from PTSD or that recovery from PTSD is accompanied by a wide-spread restoration of brain tissue.

1. Introduction

Posttraumatic stress disorder (PTSD) is a constellation of disabling behavioral and emotional symptoms that occur in some individuals who experience severe psychological trauma. Imaging studies have shown that patients with PTSD have smaller hippocampi (Bremner et al., 1997; Hedges and Woon, 2007; Kitayama et al., 2005; Smith, 2005), amygdala (Karl et al., 2006), dorsal anterior cingulate (Chen et al., 2006; Kitayama et al., 2005; Woodward et al., 2006; Yamasue et al., 2003), insula (Chen et al., 2006; Corbo et al., 2005; Kasai et al., 2008), and corpus callosum (De Bellis et al., 1999, 2002; Villarreal et al., 2004) compared to individuals without the disorder. However, not all of these brain regions have shown consistent volume reduction in prior PTSD literature (Jelicic and Merckelbach, 2004; Shin and Liberzon, 2010; Woon and Hedges, 2009).

Animal research has provided compelling evidence that exposure to severe and chronic stress can damage the brain, particularly the hippocampal formation (McEwen and Sapolsky, 1995; Sapolsky et al., 1990). Such studies point to the neurotoxic effect of elevated levels of corticosteroids, which can cause atrophy/cell death. This has led to the proposal that psychological trauma may cause neurological damage via a similar mechanism in humans. However, not all individuals exposed to trauma develop PTSD (Kessler et al., 2005). Therefore controversy exists about whether the volumetric differences that have been observed between individuals with and without PTSD represent the consequence of traumatic exposure (Magarinos and McEwen, 1995; Sapolsky, 2000) or a pre-existing trait that predisposes people to pathological stress reactions to a traumatic event (Gilbertson et al., 2002).

We recently reported that smaller hippocampal volumes were associated with current, but not lifetime, PTSD symptoms (Apfel et al., 2011). The lack of hippocampal volume differences between veterans who recovered from PTSD and those who never developed PTSD may be an indication of the recovery of hippocampal volume that accompanied the decrease in PTSD symptoms. Evidence for this hypothesis comes from previous reports of adult hippocampal plasticity (e.g., negative correlations between hippocampal volume and the duration and severity of PTSD symptoms (Felmingham et al., 2009), alterations in hippocampal volume after long-term paroxetine therapy (Vermetten et al., 2003), other pharmacological intervention (Boldrini et al., 2009; Chen et al., 2010; Warner-Schmidt and Duman, 2006), and exercise (Pajonk et al., 2010; Pereira et al., 2007)). Alternatively, the veterans who recovered from PTSD may have always had larger hippocampal volumes than the veterans who failed to recover from PTSD and the larger hippocampus may have facilitated their recovery.

One goal of the present study was to replicate our previous finding of the inverse relationship between current, but not lifetime, PTSD symptom severity on hippocampal volume. Another goal was to examine the relationship between current and lifetime PTSD symptom severity on the volumes of other brain regions previously implicated in PTSD but that have less capacity for neurogenesis than the hippocampus (Ming and Song, 2011), such as the amygdala (Karl et al., 2006), dorsal anterior cingulate (Chen et al., 2006; Kitayama et al., 2005; Woodward et al., 2006; Yamasue et al., 2003), rostral anterior cingulate (Kasai et al., 2008; Rauch et al., 2003), insula (Chen et al., 2006; Corbo et al., 2005; Kasai et al., 2008), and corpus callosum (De Bellis et al., 1999,2002; Villarreal et al., 2004). We examined magnetic resonance imaging (MRI) data acquired on a 4 T scanner, which affords better gray/white matter contrast and smaller voxel size than the 1.5 T scanner on which data from our previous study (Apfel et al., 2011) was obtained. We hypothesize, that like the hippocampus, the volumes of these other brain regions will be negatively related to current, but not lifetime PTSD symptom severity.

2. Methods

2.1. Subjects

The study sample included 191 veterans pooled from several previous studies; however, there was no subject overlap with our previous 1.5 T imaging study(Apfel et al., 2011). One hundred fifty-eight participants (83%) were Persian Gulf War (GW) veterans. Forty-four (28%) GW veterans met criteria for “Gulf War Illness” (Fukuda et al.,1998) while 60 (38%) had suspected low-level sarin exposure(Directorate for Deployment Health Support of the Special Assistant to the Under Secretary of Defense (Personnel and Readiness) for GulfWar Illness Medical Readiness, 1997). The remaining participants were veterans of Operation Iraqi Freedom (OIF; n=21), the Vietnam War (n=6), Operation Enduring Freedom (OEF, n=1), combat in Beirut (n=1), Bosnia (n=1), or multiple conflicts (i.e., Persian Gulf War, OIF, OEF, and Somalia, n=3). Thirty-nine participants met diagnostic criteria for current PTSD (PTSD+) according to the Clinician-Administered PTSD Scale (CAPS; Blake et al., 1995). Forty-three participants reported exposure to or witnessing an event involving threat for life or physical integrity and with experience of intense fear, helplessness, or horror during or outside the military service (i.e., criterion A of the DSM-IV PTSD diagnosis) and were free of current or lifetime PTSD (PTSD−). The remitted PTSD group (rPTSD) consisted of 34 GW veterans who did not meet diagnostic criteria for current PTSD but had lifetime combat-related PTSD. The non-exposed control group consisted of 75 GW veterans who reported no criterion A event. Eighteen of the 191 participants were part of previous publications on the effects of PTSD on hippocampal subfields (Neylan et al., 2010; Wang et al., 2010) and cortical perfusion and diffusion tensor imaging (Schuff et al., 2011). Fourteen participants were part of a previous publication on PTSD and occipital function and structure (Chao et al., 2012), while 103 were part of a previous publication on the effects of suspected sarin exposure and regional brain volume (Chao et al., 2011). Data from 56 veterans had not previously been published. There was no subject overlap with our previous 1.5 T imaging study (Apfel et al., 2011).

Lifetime and current PTSD were assessed with the CAPS (Blake et al., 1995), a structured interview measure that corresponds to DSM-IV criteria for PTSD. The CAPS assesses the frequency and intensity of re-experiencing, avoidance, and hyperarousal symptoms of PTSD. The diagnosis of current PTSD was based on symptoms experienced in the previous month associated with the subject's self-identified worst traumatic event. Other psychiatric disorders were assessed by administration of the Structured Clinical Interview for DSM-IV (First et al., 1995). All diagnoses were made by trained clinical interviewers who calibrated their assessments at weekly case consensus meetings, supervised by an experienced Ph.D.-level clinical psychologist. Depressive symptoms were assessed with the Beck Depression Inventory (BDI, Beck, 1961) and/or the Hamilton Depression Scale (HAMD, Hamilton, 1967).

Five items from the interview version of the Life Stressor Checklist (Wolfe et al., 1996) were used to assess exposure to severe childhood trauma before age 14 (O’Donovan et al., 2011; Otte et al., 2005; Pole et al., 2007). Participants were asked whether they had been exposed to any of the following experiences to the extent that they felt that they could die or be physically harmed: physical neglect, family violence, physical abuse, forced sexual touch, or forced sexual intercourse.

Psychiatric medications were not discontinued. Twenty-nine (15%) participants were on a stable dose of antidepressant. All participants provided written informed consent in accordance with procedures of the institutional review boards of the University of California, San Francisco and the San Francisco VA Medical Center.

2.2. Image acquisition and processing

Magnetic resonance images of the subjects' brains were acquired on a 4 T (Bruker/Siemens) magnetic resonance imaging (MRI) system with a birdcage transmit and eight channel receive coil. T1-weighted structural MRI were obtained with a 3D volumetric magnetization prepared rapid gradient echo (MPRAGE) sequence, TR/TE/TI=2300/3/950 ms, timing; 7° flip angle; 1.0 × 1.0 × 1.0 mm³ resolution; 157 continuous sagittal slices.

An automated, non-biased atlas-based Bayesian segmentation procedure, applied in Freesurfer v.4.5 (http://surfer.nmr.mgh.harvard.edu/), was used to derive quantitative estimates of brain structure and to label cortical and subcortical tissue classes (Dale et al., 1999; Desikan et al., 2006; Fischl et al., 1999). Freesurfer processing for volumetric T1-weighted images included: motion correction, brain extraction and removal of non-brain tissue using a hybrid watershed/surface deformation procedure (Segonne et al., 2004); automated spatial transformation and white matter (WM) segmentation of subcortical volumetric structures (Fischl et al., 2004a); intensity normalization, tessellation of gray/white matter boundary and automated topology correction (Segonne et al., 2007); and surface deformation following intensity gradients to optimally place gray/white matter and gray matter/cerebral spinal fluid borders at the location where the greatest shift in intensity defines the transition to the other tissue class (Dale et al., 1999). The reconstructed cortical surface models for each participant were visually inspected to ensure segmentation accuracy. Twenty subjects (nine non-exposed controls, 3 PTSD−, 2 rPTSD, 6 PTSD+) with poor segmentation accuracy (i.e., over or under-estimations) were excluded from statistical analyses.

Freesurfer spatially normalizes each cortical surface to a template cortical surface, allowing for the automatic parcellation of the cortical surface into 34 cortical (Fischl et al., 2004b). The volume of cortical structures is derived by multiplying the thickness and the surface are of the structure. The Freesurfer algorithm also performs an automatic subcortical segmentation where the non-cortical voxels in the normalized brain volume are assigned one of approximately 40 subcortical labels (e.g., hippocampus and amygdala). Because no thickness is calculated for subcortical structures, the volume of subcortical structures is derived by counting the number of voxels in each structure. Based on previous reports of the structural brain abnormalities associated with PTSD (Karl et al., 2006), we focused our analyses on the following a priori regions of interest (ROIs) from Freesurfer subcortical parcellations: the hippocampus, amygdala, corpus callosum, (see Fig. 1A-C for e.g.), insula, and anterior cingulate (see Fig. 1B and D for e.g.).

Fig. 1.

Examples of hippocampal (H, yellow) and amygdala (A, light blue) regions of interest (ROIs) from Freesurfer subcortical segmentation overlaid on a sagittal view of the brain (Panel A); hippocampal (H, yellow), amygdala (A, light blue) and insula (In, yellow) ROIs overlaid on a coronal view of the brain (Panel B); corpus callosum (CC, lavender) ROI overlaid on a signttal view of the brain (Panel C); and rostral anterior cingulate (RAC, purple) and caudal anterior cingulate (CAC, lavender) ROIs from Freesurfer cortical segmentation overlaid on a sagittal view of the brain (Panel D).

2.3. Statistical procedures

To account for individual differences in head size, all cortical volumes were expressed as percentage of intracranial volume (ICV), derived from Freesurfer (Buckner et al., 2004). To protect against type I error, we averaged volumes from homologous parcels across hemispheres because we had no a priori hypotheses about laterality. Significance was evaluated at p<0.05, 2-tailed for the six a priori ROIs.

The main analyses of regional brain volume by group (PTSD+, PTSD−, rPTSD, non-exposed) were carried out using multivariate analysis of covariance (MAN-COVA). Regions that yielded significant group effects (p<0.05) were followed by post-hoc planned contrasts that examined differences between the four groups. To determine which potentially confounding variables (i.e., age, gender, education, early life trauma, current major depressive disorder (MDD), past alcohol/drug use, use of antidepressants, Gulf War Illness status, and suspected sarin exposure) to include as a covariates in the MANCOVA, we first looped through all of the potential confounds, using multiple regression to test the association of the potential confounds with the dependent variables (i.e., the six a priori ROIs). Potential confounding variables that were associated with three or more dependent variables at p≤0.25 were included as covariates in the MANCOVA.

2.3.1. Post-hoc analyses

Because Woodward et al. (2009) previously reported smaller total cerebral cortex, parahippocampal gyrus, superior+ transverse temporal cortex, lateral orbital frontal cortex, and inferior frontal pars orbitalis volumes in veterans with chronic PTSD relative to those without PTSD using an earlier version of Freesurfer, we also examined the effects of current versus lifetime PTSD on these five regions in post-hoc analyses. Adjustments for multiple comparisons were made according to the number of ROIs (5) and the average intercorrelations among the ROIs (r=0.67; Sankoh et al. 1997). A 2-sided adjusted p<0.05 was considered statistically significant. Because this analysis revealed a significant group difference in total cortical volume, we re-examine the group effect on the volumes of the six a priori ROIs in a separate post-hoc MANCOVA.

Because Gulf War Illness status was significantly associated with hippocampal, caudal anterior cingulate, insula, and corpus callosum volumes, thus meeting the above criterion for inclusion as covariates in the MANCOVA, in post-hoc analysis, we tested the ability of Gulf War Illness status to predict the volumes of the a priori ROIs after accounting for age, gender, education, early life trauma, current MDD, antidepressant use, past alcohol/drug abuse/dependence, potential sarin exposure, and group status with linear regression. In the model, Gulf War Illness status was entered into the regression model in the second step, after all the other variables had been forced into the model in the first step.

Because suspected sarin exposure has been was linked to reduced total gray and white matter volumes (Chao et al., 2010, 2011; Heaton et al., 2007), we also tested the ability to suspected sarin exposure to predict volumes of the a priori ROIs after accounting for age, gender, education, early life trauma, current MDD, antidepressant use, past alcohol/drug abuse/dependence, Gulf War Illness, and group status with linear regression. In this model, sarin exposure status was entered into the regression in the second step, after all the other variables had been forced into the model in the first step.

3. Results

3.1. Sample characteristics

Demographic and clinical variables are reported by group in Table 1. The non-exposed control group was older than the PTSD+ (Tukey's post-hoc p<0.001) and PTSD− (Tukey's post-hoc p=0.001) groups and more educated than the PTSD+ (Tukey's post-hoc p=0.008) and rPTSD (Tukey's post-hoc p=0.03) groups. As expected, more veterans in the PTSD+ group had current (χ²=25.18, df=3, P<0.001) and lifetime (χ²=27.09, ^df=3, p<0.001) MDD, had higher Hamilton Depression scale (HAM-D, Tukey's post-hoc p<0.001) scores, and were on antidepressants (χ²=29.70, df=3, p<0.001) than veterans in the other three groups. HAM-D scores were also higher in the rPTSD group than veterans in the non-exposed group (Tukey's post-hoc p=0.006). The non-exposed control group had a lower incidence of early life trauma compared to the other three groups (χ²=25.28, df=3, p<0.001). More veterans in the rPTSD group fulfilled criteria for Gulf War Illness than veterans in the other three groups (^χ2=12.04, df=3, p=0.007). The PTSD+ group had fewer GW veterans (χ²=12.04, df=3, p=0.007) and therefore fewer veterans with suspected exposure to sarin compared to the other three groups (χ²=15.45, df=3, p=0.001). The 22 OEF/OIF veterans in the study sample were fairly evenly distributed among the PTSD− (n=9) and PTSD+ (n=13) groups (Fisher's exact test, p=0.22).

Table 1.

Demographic information and clinical variables.

	PTSD+	PTSD−	rPTSD	Nonexposed	F/χ2 value
N	39	43	34	75
Age (yrs)	42.7 (11.7)	44.2 (10.3)	47.6 (7.5)	50.8 (7.5)	8.51*
No. (%) female	2 (5%)	8 (19%)	3 (9%)	10 (13%)	3.60
No. (%) caucasian	24 (62%)	24 (56%)	16 (47%)	45 (60%)	2.00
Education (yrs)	14.3 (1.9)	15.6 (1.8)	14.4 (1.9)	15.6 (2.3)	5.68*
CAPS, total severity	60.1 (13.2)	8.7 (10.5)	15.3 (12.1)	0.0 (0.0)	377.88**
CAPS, intrusion	13.8 (6.8)	2.4 (4.5)	3.7 (4.1)	0.0 (0.0)	100.52**
CAPS, avoidance	23.7 (7.4)	2.7 (4.6)	3.5 (4.5)	0.0 (0.0)	268.65**
CAPS, hyperarousal	22.6 (5.5)	3.6 (4.4)	8.2 (6.2)	0.0 (0.0)	265.64**
Lifetime CAPSa	81.2 (23.4)	23.6 (10.7)	64.8 (19.9)	0.0 (0.0)	305.04**
Current MDD	15 (38%)	2 (5%)	6 (18%)	5 (7%)	25.18**
Lifetime MDD	17 (44%)	14 (33%)	15 (44%)	5 (7%)	27.09**
HAM-D scoreb	12.9 (5.5)	5.2 (4.5)	7.5 (5.1)	4.1 (4.8)	28.32**
Antidepressant use	16 (41%)	7 (16%)	4 (12%)	2 (3%)	29.70**
Early life trauma reported	18 (46%)	15 (35%)	9 (26%)	5 (7%)	25.28**
Past alcohol abuse/dependence	22 (56%)	15 (35%)	17 (50%)	27 (36%)	6.19
Past drug abuse/dependence	5 (13%)	4 (9%)	8 (24%)	5 (7%)	6.92
No. (%) Gulf War veterans	15 (38%)	34 (79%)	34 (100%)	75 (100%)	76.62**
No. (%) with Gulf War Illness	9 (23%)	12 (28%)	14 (41%)	9 (12%)	12.04*
No. (%) with suspected sarin exposure	9 (10%)	12 (33%)	14 (53%)	9 (32%)	15.45*

Numbers are means or numbers; standard deviations or percentages are in parentheses; PTSD, posttraumatic stress disorder; rPTSD, remitted PTSD; CAPS, Clinician Administered PTSD Scale; MDD, major depressive disorder; BDI, Beck Depression Inventory and HAM-D, Hamilton Depression Scale.

^aLifetime CAPS score sample sizes for PTSD+, PTSD−, rPTSD, and non-exposed group are 12, 33, 34, and 75 respectively.

^bHAMD score sample sizes for PTSD+, PTSD−, rPTSD, and non-exposed group are 38, 42, 34, and 75 respectively.

^*p<0.01.

^**p<0.001.

3.2. Consideration of potentially confounding variables

When we tested for an association between the potentially confounding variables and the a priori ROIs, age (model p≤0.001, 0.06<R²<0.15 for all ROIs except the corpus callosum), gender (model p<0.11, 0.01<R²<0.04 for amygdala, hippocampus, caudal anterior cingulate, and corpus callosum), current MDD (model p<0.17, 0.01<R²<0.03 for amygdala, hippocampus, insula, and corpus callosum), and Gulf War Illness status (model p<0.18, 0.01<R²<0.01 for the hippocampus, caudal anterior cingulate, insula, and corpus callosum) met the criterion for inclusion as covariates in the subsequent MANCOVA (i.e., were associated with at least three of dependent variables at p≤0.25).

3.3. Group differences in the volumes of the a priori ROIs

The volumetric results are summarized in Tables 2 and 3. A MANCOVA that accounted for age, gender, current MDD, and Gulf War Illness yielded a significant main effect of group (Wilks λ=0.83, F_18,504=1.86, p=0.02, partial η²=0.059). There were significant group effects on hippocampal (F_3,183=4.25, p=0.006, partial η²=0.065), caudal anterior cingulate (F_3,183=3.04, p=0.03, partial η²=0.047), insula (F_3,183=3.05, p=0.03, partial η²=0.048), and corpus callosum (F_3,183=3.62, p=0.01, partial η²=0.056) volumes. Planned contrasts revealed that the PTSD+ group had smaller hippocampal (p=0.002), caudal anterior cingulate (p=0.006), insula (p=0.009), and corpus callosum (p=0.004) volumes than the non-exposed control group; smaller hippocampal (p=0.002), caudal anterior cingulate (p=0.03), insula (p=0.01), and marginally smaller corpus callosum (p=0.06) volumes than the rPTSD group; and smaller hippocampus (p=0.04) and caudal anterior cingulate (p=0.01) compared to the PTSD− group. Compared to non-exposed controls, the PTSD− group had smaller corpus callosum volume (p=0.02). There were no significant differences between rPTSD and non-exposed control groups (p>0.37) or between the rPTSD and PTSD− groups (p>0.13).

Table 2.

Least squared means volume^a by region and group.

	PTSD+	PTSD−	rPTSD	Nonexposed	F-value	Effect sizes (partial η2)
A priori regions
Amygdala	0.28 (0.03)	0.28 (0.02)	0.29 (0.03)	0.29 (0.02)	1.60	0.026
Hippocampus	0.68 (0.07)	0.73 (0.06)	0.76 (0.07)	0.75 (0.05)	4.25**	0.065
Rostral Ant Cingulate	0.32 (0.04)	0.33 (0.03)	0.32 (0.04)	0.34 (0.03)	1.29	0.021
Caudal Ant Cingulate	0.32 (0.04)	0.36 (0.04)	0.36 (0.04)	0.36 (0.03)	3.04*	0.047
Insula	1.10 (0.09)	1.14 (0.08)	1.18 (0.09)	1.18 (0.07)	3.05*	0.048
Corpus callosum	0.28 (0.04)	0.29 (0.03)	0.31 (0.04)	0.31 (0.03)	3.62*	0.056
Post-Hoc Woodward et al. regions
Total cerebral cortex	41.01 (3.22)	42.87 (2.90)	44.05 (3.23)	44.35 (2.27)	3.86**	0.059
Parahippocampal gyrus	0.40 (0.05)	0.41 (0.04)	0.43 (0.05)	0.42 (0.03)	1.60	0.026
Lateral orbitofrontal cortex	1.19 (0.10)	1.22 (0.09)	1.25 (0.10)	1.28 (0.07)	2.55	0.040
Inferior frontal cortex−pars orbitalis	0.41 (0.04)	0.42 (0.04)	0.43 (0.04)	0.44 (0.03)	1.78	0.028
Superior+transverse temporal cortex	2.17 (0.21)	2.30 (0.19)	2.38 (0.21)	2.37 (0.15)	3.47*	0.054

All volumes expressed as a percentage of intracranial volume (ICV) with standard deviations in parentheses.

^aControlling for age, gender, current major depressive disorder, and Gulf War Illness status.

^*p<0.05.

^**p≤0.01.

Table 3.

Results of planned contrasts and effect size for regions exhibiting significant main group effects.

	Planned contrasts						Effect size (d)
	PTSD+ vs. CN	PTSD+ vs. PTSD−	PTSD+ vs. rPTSD	rPTSD vs. CN	rPTSD vs. PTSD−	PTSD− vs. CN	PTSD+ vs. CN	PTSD+ vs. PTSD−	PTSD+ vs. rPTSD	rPTSD vs.CN	rPTSD vs. PTSD−	PTSD− vs. CN
A priori regions
Hippocampus	**	*	**	ns	ns	ns	−1.15	−0.77	−1.14	0.16	0.46	−0.36
Caudal ant cingulate	**	*	*	ns	ns	ns	−1.13	−1.00	−1.00	0.00	0.00	0.00
Insula	**	ns	**	ns	ns	ns	−0.99	−0.47	−0.89	0.00	0.47	−0.57
Corpus callosum	**	ns	*	ns	ns	*	−0.85	−0.28	−0.75	0.00	0.57	−0.67
Woodward et al. regions
Total cerebral Ctx	**	ns	**	ns	ns	ns	−1.20	−0.61	−0.94	−0.11	0.38	−0.10
Superior+transverse temporal Ctx	**	ns	**	ns	ns	ns	−1.10	−0.65	−1.00	0.05	0.40	−0.41

ns, not significant.

^*p<0.05.

^**p≤0.01.

3.4. Group differences in the Woodward et al. regions

The post-hoc MANCOVA that examined volumes of the Woodward et al. (Woodward et al., 2009) brain regions, with age, gender, current MDD, and Gulf War Illness as covariates, yielded no overall effect of group (Wilks λ=0.91, F_15,495=1.17, p=0.29, partial η²=0.32). However, there were significant group effects on total cortical volume, derived by summing all the Freesurfer parcellated cortical volumes (F_3,183=3.86, p=0.01, partial η²=0.059) and superior+ transverse temporal cortical volume (F_3,184=3.47, p=0.02, partial η²=0.054). Planned contrasts revealed that the PTSD+ group had smaller total cortical volume than the non-exposed control (p=0.002), and the rPTSD (p=0.009) groups and smaller superior+ transverse temporal volumes than the non-exposed controls (p=0.004) and rPTSD (p=0.007) groups. There were no significant differences between the rPTSD and non-exposed control (p>0.76), the rPTSD and PTSD− (p>0.26), the PTSD− and non-exposed control (p>0.12), or the PTSD− and the PTSD+ (p>0.09) groups.

The significant difference in total cortical volume across groups prompted us to re-examine the group effect on the volumes of a priori ROIs in a separate, post-hoc MANCOVA that took into account total cortical volume, along with age, gender, and current MDD. As expected, there was a significant multivariate effect of total cortical volume (Wilks λ=0.20, F_6,177=119.00, p<0.001, partial η²=0.801) and the effect of group was no longer significant (Wilks λ=0.88, F_18,501=1.26, p=0.21, partial η²=0.041).

3.5. Effects of Gulf War Illness and potential sarin exposure

Linear regression showed that neither Gulf War Illness status nor potential sarin exposure were significantly associated with the volumes of the six a priori ROIs after accounting for group membership and other potentially confounding variables (see Table 4).

Table 4.

Post-hoc regression analyses examining effects of Gulf War Illness and potential sarin exposure.

	Effects of Gulf War Illness					Effects of potential sarin exposure
	Step 1a		Step 2b			Step 1c		Step 2d
	R 2	F 10,180	R 2	Δ R2	Δ F1,179	Δ R2	F 10,180	R 2	Δ R2	Δ F1,179
Amygdala	0.13	2.74*	0.13	0.00	0.04	0.13	2.69*	0.13	0.00	0.43
Hippocampus	0.15	3.21**	0.16	0.01	1.61	0.15	3.28**	0.16	0.01	1.06
Rostral anterior cingulate	0.16	3.42**	0.16	0.00	0.05	0.16	3.40**	0.16	0.00	0.17
Caudal anterior cingulate	0.13	2.67*	0.11	0.01	1.22	0.13	2.74*	0.13	0.00	0.55
Insula	0.18	4.04**	0.20	0.01	2.55	0.19	4.33**	0.19	0.00	0.20
Corpus callosum	0.13	2.57**	0.14	0.01	3.32	0.14	2.89**	0.14	0.00	0.56

^aVariables entered, age, gender, education, early life trauma, current MDD, antidepressant use, potential sarin exposure, past alcohol abuse/dependence, past drug abuse/dependence, group.

^bVariable entered, Gulf War Illness status.

^cVariables entered, age, gender, education, early life trauma, current MDD, antidepressant use, Gulf War Illness, past alcohol abuse/dependence, past drug abuse/dependence, group.

^dVariable entered, potential sarin exposure.

^*p<0.01.

^**p≤0.001.

4. Discussion

The first aim of this study was to replicate our previous finding that hippocampal volume is related to current but not lifetime PTSD symptoms severity. We found smaller hippocampal volumes in the PTSD+ group compared to the non-exposed control, PTSD−, and rPTSD group. In contrast, there were no significant hippocampal volume differences between the rPTSD and non-exposed control group or between the rPTSD and PTSD− group. This extends our previous 1.5 T MRI finding to a different cohort of veterans scanned on a 4 T MR scanner.

Consistent with previous reports, we also found that current, chronic PTSD was associated with smaller anterior cingulate (Chen et al., 2006; Kitayama et al., 2005; Woodward et al., 2006; Yamasue et al., 2003), insula (Chen et al., 2006; Corbo et al., 2005; Kasai et al., 2008), and corpus callosum (De Bellis et al. 1999,2002; Villarreal et al., 2004) volumes compared to the non-exposed control and PTSD- groups. We also found reduced total cortical volume in veterans with current, chronic PTSD, although our results, like those of Woodward et al. (2009) suggest that the association of PTSD with smaller cortical volume is not uniform across the brain.

There is controversy concerning the nature and origin of reduced regional brain volumes in PTSD. At issue is whether these reductions represent pre-existing vulnerability factors for developing PTSD following exposure to trauma, acquired signs of PTSD due to the traumatic stress that caused PTSD, or the chronic stress of having PTSD (or both). Thus far the debate has focused on the development of PTSD rather than recovery from PTSD. Because the association of PTSD to structural brain volumes should also consider symptomatic recovery, another aim of this study was to examine the relationship between current versus lifetime PTSD symptom severity and the volumes of other brain regions implicated in PTSD (e.g., amygdala, corpus callosum, insula, and anterior cingulate cortex). We found no significant brain volume differences between veterans who had remitted PTSD, veterans who never developed PTSD, and those who had not been exposed to trauma.

If we assume that PTSD can develop in individuals following trauma exposure regardless of pre-trauma brain size, then it is plausible that individuals with larger pre-trauma brains have a greater capacity to recover from PTSD than individuals with smaller pre-trauma brains. Therefore, the smaller brain volumes that we observed in veterans with chronic PTSD may be a pre-existing vulnerability, either due to genetic or environmental circumstances that hinder recovery from PTSD. On the other hand, if the smaller brain volumes that we observed in veterans with chronic PTSD were a consequence of having persistent PTSD over many years, then the lack of brain volume difference between veterans with remitted PTSD, those who never developed PTSD, and those who had not been exposed to trauma may be due to morphological restitution in the recovered veterans. Although the caudal anterior cingulate, insula, and corpus callosum are generally considered to have less capacity for neurogenesis than the hippocampus, there have been reports of suggestive of morphological restitution in non-hippocampal brain regions in elderly adults after six months of aerobic exercise training (Colcombe et al., 2006), in adults with anorexia nervosa following weight restoration (Roberto et al., 2011; Swayze et al., 1996), and in recovered alcoholics following abstinence (Bartsch et al., 2007; Gazdzinski et al., 2005,2010; ) that is not solely attributed to rehydration (Schroth et al., 1988; Trabert et al., 1995).

Although it has been proposed that trauma exposure, even in the absence of PTSD, may be associated with reduced brain volume (Hedges and Woon, 2010; Sapolsky, 2000), one longitudinal study found no baseline hippocampal volume differences and no differences in the 6-month hippocampal atrophy rates of 10 trauma survivors who developed PTSD relative to 27 trauma survivors who did not develop PTSD. Bonne et al. (2001) cited the older age at trauma exposure, the length of trauma exposure, and chronicity of symptoms as possible explanations for this negative finding. Thus, within the context of the current study, a third possibility is that there were never any significant brain volume differences between veterans in the rPTSD, PTSD−, or the non-exposed control groups. Because the first Gulf War lasted less than a year, resulted in fewer than 150 battle deaths, and was fought by individuals who volunteered for military service, these factors may explain why there were no significant brain volume differences between Gulf War veterans in the rPTSD and the non-exposed control groups, and why some of the Gulf War veterans in the current study were able to recover from PTSD.

A somewhat unexpected finding in the current study is the smaller corpus callosum volume in the PTSD− group relative to the non-exposed control group. However, there was also a higher incidence of childhood trauma in the PTSD− compared to the non-exposed control group (35% versus 7%; χ²=18.18, p<0.001). Childhood trauma has been shown to alter brain structure (Bremner et al., 1997; Dannlowski et al., 2012; De Bellis and Kuchibhatla, 2006; Richert et al., 2006; Teicher et al. 1997, 2012, 2004), particularly the corpus callosum (Andersen et al., 2008; De Bellis and Keshavan, 2003; De Bellis et al., 1999, 2002).

Another unexpected, ancillary finding is the absence of an effect of past alcohol abuse/dependence on cortical volume, despite the moderately high incidence of past alcohol dependence/abuse in the current sample (35–56% among the four groups). Although alcoholism has been associated with smaller cerebral tissue volumes (Jernigan et al., 1991; Pfefferbaum et al., 1995; Pfefferbaum et al., 1998), it is worth noting that others have also reported no effect of lifetime alcoholism on brain volume after controlling for the effects of PTSD and age (Woodward et al. 2007, 2009). Furthermore, the effects of alcohol abuse on the brain have been shown to be, at least in part, reversible within the first few weeks to months of sobriety (Gazdzinski et al., 2005, 2010; Pfefferbaum et al., 1995). In our previous 1.5 T study, we reported that hippocampal volume was significantly related to current but not lifetime alcohol use (Apfel et al., 2011). The current average number of drinks consumed per month by the veterans in the present study was 16±27. Because there have been no reports of such an amount of alcohol consumption appreciably affecting brain volume, this may also account for why we did not find an effect of lifetime alcoholism on cortical volume.

Consistent with a recent meta-analysis that concluded there were no significant amygdala volumes differences between PTSD patients and comparison subjects (Woon and Hedges, 2009), we did not find any amygdala volume differences between the groups. Functional imaging studies have reported heightened amygdala reactivity among individuals with PTSD in response to generic fear and trauma-related stimuli (Etkin and Wager, 2007; Lanius et al., 2006), in accordance with animal studies that find trophic structural alterations in the amygdala as a consequence of severe and/or prolonged stress (Cui et al., 2008; Vyas et al., 2003, 2002). However, the recent reports of both smaller (Morey et al., 2012) and larger (Kuo et al., 2012) amygdala volumes in combat-exposed veterans/military service members with current PTSD highlights the inconsistency of the literature concerning amygdala volume in PTSD. A further obstacle to settling this issue may be inherent in the phenomenon itself. Some have found that early life exposure to trauma (Lupien et al., 2011; Mehta et al., 2009; Tottenham et al., 2010) increases amygdala volume while others have reported that severity of criterion A trauma (Mollica et al., 2009) decreases amygdala volume in adults. Thus, studies of amygdala volume in adult PTSD may be expected to produce ambiguous results in individuals who have experienced both early life and adult trauma.

The present findings should to be considered in the context of a number of limitations: First, the sample, while large, was inhomogeneous and included subgroups of veterans with Gulf War Illness and with suspected exposure to sarin. However, regression analysis revealed no effect of Gulf War Illness on the volumes of the six a priori ROIs after accounting for group membership. This is in line with our previous imaging study in a non-overlapping veteran sample that found no effects of Gulf War Illness on structural volumes but significantly higher rates of PTSD among veterans with Gulf War Illness (Weiner et al., 2010). Similarly, there was no overall effect of suspected exposure to sarin. Furthermore, because there was a lower incidence of suspected sarin exposure among veterans in the PTSD+ group relative to the other three groups and a higher incidence of suspected exposure to sarin (53% versus 32%; χ²=4.33, p=0.04) in the rPTSD group compared to the non-exposed control group, even if there had been a potential confounding effect of suspected sarin exposure on brain volume, it would have likely introduced a negative bias away from finding differences in the relationship brain volume measures to current PTSD. There was also a higher incidence of current and lifetime MDD, higher HAMD scores, antidepressant use among veterans in the PTSD+ group compared to other three groups. However, assembling a PTSD group that is free of depressive symptoms is unlikely to generalize. Moreover, there has recently been evidence that calls into question whether PTSD and depression are distinct entities among individuals exposed to trauma given the common criterion symptoms (Elhai et al., 2011). Another limitation is the cross-sectional nature of the study, which limits our ability to determine causality. Therefore, future prospective studies will be necessary to investigate the proposed hypotheses concerning recovery from PTSD and the possible restoration of brain tissue. A third limitation is the current study design, which restricted our ability to consider other pre- and post-trauma factors (e.g., heredity, socioeconomic status) that likely influences one's ability to recover from PTSD. Thus, additional studies will be needed to tease apart the contributions of genetic and environmental pre- and post-trauma factors that affect recovery from PTSD and possible recovery of brain tissue. A fourth limitation is the absence of an objective measure of pre-morbid intelligence, which has been shown to be moderately correlated with brain volume (McDaniel, 2005) and a factor in PTSD risk (McNally and Shin, 1995). However, education, which is a reasonable proxy for IQ, was not significantly associated with the dependent variables when we initially examined the relationships between the potentially confounding variables and volumes of a priori ROIs. The subjective nature of patient history and clinical scores, which may be biased by under- or over-reporting, is another potential limitation. Finally, the low percentage of women in the current sample of veterans limits generalization of these findings. These limitations notwithstanding, the present study found that current, but not lifetime PTSD, is associated with reduced volumes of regional and total cortical volumes in adults, suggesting that either smaller brain volume is a vulnerability factor that impedes recovery from PTSD or that recovery from PTSD is accompanied by a wide-spread restoration of brain tissue.