Association between unremitting PTSD and smaller right hippocampal volume among veterans thirty years after combat

Abstract

Objective:

A significant percentage of individuals diagnosed with posttraumatic stress disorder (PTSD) show remission within the first year after traumatic exposure, with showing more gradual improvement over the next several years. However, a subgroup of individuals diagnosed with PTSD experience an unremitting course of illness that may persist for years or decades despite treatment efforts. The aim of the current study was to test the hypothesis that reduced hippocampal volume is associated with chronic, unremitting PTSD rather than with PTSD that has remitted over time.

Methods:

The current study included 44 Vietnam War veterans whose traumatic exposure had occurred over 3 decades prior to study participation. The authors studied the association between hippocampal volume and 3 PTSD diagnostic categories: veterans with unremitting PTSD (n=26), those with sustained remission from earlier PTSD (n=8), and those who never developed PTSD (n=10).

Results:

Compared to trauma-exposed veterans who never developed PTSD, those with unremitting PTSD, but not those who had recovered, showed reduced hippocampal volume. The authors further found these results to be anatomically lateralized to only the right hippocampus, consistent with previous neuroimaging studies of male combat veterans.

Conclusions:

This study’s findings support an emerging literature suggesting that the relevance of reduced hippocampal volume observed in PTSD may be related to failure to recover from rather than to the development of PTSD following trauma exposure. These findings may define a subpopulation of combat veterans who by virtue of their smaller right hippocampi are at risk for long-term failure to recover from PTSD.

Introduction

Posttraumatic stress disorder (PTSD) develops in a proportion of individuals exposed to severe traumatic events. Over 40% of those diagnosed with PTSD show remission within the first year after traumatic exposure, with a continued, more gradual remission rate over the next several years (1). However, a subgroup of individuals experiences an unremitting course that may persist for several decades despite treatment. Combat veterans form a significant component of this subgroup (2).

Hippocampal volume reduction has been one of the most consistent neuroanatomical findings associated with PTSD (3–6). In a twin study from our laboratory (4), we found that reduced hippocampal volume is a familial factor that likely pre-dates trauma exposure and may serve as a risk factor for the development of chronic PTSD. However, reduced hippocampal volume only emerged in individuals with severe symptoms over many years. An interpretation of this finding is that hippocampal volume differences predispose to more severe, unremitting PTSD.

Consistent with the above, correlational studies have demonstrated that smaller hippocampal size is associated with more severe and chronic PTSD (8). Of note, Apfel et al (9) examined hippocampal volume in three groups of trauma-exposed Gulf War veterans based upon lifetime and current CAPS scores: individuals with chronic PTSD (lifetime and current PTSD), individuals who recovered from PTSD (lifetime but not current PTSD), and individuals who never developed PTSD (no past or current PTSD). Apfel and colleagues found smaller hippocampal volume only in the chronic PTSD group, but not in the remitted or never-developed PTSD groups. These findings support the hypothesis that reduced hippocampal volume is unique to unremitting forms of PTSD (i.e., a failure to recover). This may have important treatment implications.

The first aim of the current study was to test the hypothesis that reduced hippocampal volume is associated with chronic, unremitting forms of PTSD rather than with a history of PTSD that has remitted over time. The second aim was to determine whether the right or the left hippocampal volume is associated with unremitted PTSD. The study attempted to replicate and expand Apfel et al.’s (9) findings above.

Methods and Materials

Participants

Research participants were male veterans who self-identified as having served in combat role in Vietnam 3 or more decades prior to study participation who were recruited from the Manchester Veterans Affairs Medical Center (MVAMC) between 2006 and 2009. The participant pool comprised individuals who had previously participated in research and in many cases were outpatients receiving treatment. This study was carried out in accordance with the Declaration of Helsinki. The protocol was approved by the Institutional Review Board/Human Subjects Subcommittee at the Bedford, MA VAMC. All subjects provided informed written consent prior to participation. PTSD diagnostic status (DSM-IV), and their overall PTSD symptom severity, were determined by a doctoral-level psychologist using the Clinician-Administered Posttraumatic Stress Disorders Scale (CAPS) (7). Consistent with Apfel et al. (9), subjects were determined to have current PTSD if their total CAPS score was ≥40. Subjects were also interviewed using the Structured Clinical Interview for DSM-IV (SCID) (10) to determine the presence of other Axis I mental disorders. Subjects were excluded if they met criteria for a psychotic disorder, bipolar disorder, or for non-combat-related PTSD. Combat trauma severity was measured by the Combat Exposure Scale (CES) (11). Due to the high comorbidity of major depressive and substance use disorders with PTSD, these disorders were not exclusionary. Depression symptomatology severity was measures by the Beck Depression Inventory-II (BDI-II) (12). The Michigan Alcoholism Screening Test (MAST) (13) was completed by subjects as a measure of lifetime alcohol use severity. Additional exclusion criteria included a positive urine screen for substances of abuse (e.g., amphetamines, barbiturates, cocaine, opiates, benzodiazepines, cannabinoids, and others); a history of significant neurologic disorder (e.g., epilepsy, multiple sclerosis, encephalitis); or current inpatient status. Standard exclusion criteria for MRI participation were employed. The final sample comprised 44 combat veterans: 26 with Unremitting PTSD, 8 with PTSD in Sustained Remission, and 10 who Never Developed PTSD.

MRI Image Acquisition and Volumetric Analyses

MR scanning was performed at the Brigham and Women’s Hospital in Boston, MA with a 1.5 Tesla General Electric Signa System employing techniques previously described in detail and identical to those employed in our prior twin study (4, 14). Amygdala volume served as a control measure, given its general relevance to neurobiological models of PTSD. Hippocampus and amygdala were outlined manually on a Sun Microsystems workstation by a rater blind to diagnostic information, employing an established procedure for volumetric determination (3, 14). Two additional blind raters performed volumetric analyses of the hippocampus and amygdala on 5 random cases. Reliability assessment of the three raters resulted in the following intraclass correlation coefficients: right hippocampus = 0.96; left hippocampus = 0.96; right amygdala = 0.99; left amygdala = 0.99.

Statistical analysis:

We were interested in the association between categorical diagnosis of PTSD and hippocampal volume. Therefore, we started with a basic regression model with hippocampal volume as the dependent variable (DV) and diagnostic category as the independent variable (IV). Because intracranial volume might confound the relationship, we included it as a covariate in the basic model. Next, we considered including variables as potential confounders based on prior evidence of the relationship between them and the association of interest. Because depression and PTSD overlap substantially, we reasoned that including depression as a covariable would make the interpretation of the association between PTSD Diagnosis and hippocampal volume difficult to interpret and limit the generalizability of our findings; nevertheless, we tested it. We tested potential confounders by adding them one-at-a-time to the basic model and by calculating the percentage change in the estimate of our primary IV (diagnostic category) as well as its standard error. This is a well-established statistical approach known as change in estimate (CIE) (15–17). To determine whether a variable is a confounder, colinear, or neither, we used an a priori rule: if upon the introduction of the potential confounding variable the primary predictor estimate changes by more than 20%, the variable is considered a confounder and is kept in the model; if the standard error changes by more than 20%, it is considered colinear and not kept in the model. This strategy, and the 20% cut-point for the CIE strategy, yield an acceptable point estimator bias of about 10% (16). The final model included the basic model and any variable(s) which we believed to be potential confounders based on external evidence and which passed the a priori test for confounding during CIE analysis.

Results:

Those with unremitting PTSD, compared to either those with sustained remission or to those who never developed PTSD, were younger, had higher combat exposure, alcohol use, depression, current CAPS scores, and lifetime CAPS scores (Table 1). Race and marital status did not differ among groups (Supplemental Table 1).

Table 1:

Demographic characteristics of the PTSD diagnostic groups.

All participants				Those who never developed PTSD			Those in sustained remission from PTSD		Those with unremitting PTSD				ANOVA
Variable	N	Mean	Std Dev	N	Mean	Std Dev	N	Mean	Std Dev	N	Mean	Std Dev	F (DF)	P
Age	44	60.5	3.9	10	63.3	6.2	8	61.3	2.3	26	59.2	2.3	5.1 (2)	0.01
Years of Education	44	14.6	2.3	10	14.9	2.3	8	14.9	2.4	26	14.4	2.4	0.2 (2)	0.81
CES	44	22.9	9.2	10	20.6	7.8	8	21.9	4.7	26	24.2	10.7	0.6 (2)	0.56
MAST	43	15.2	15.0	10	5.6	5.6	8	15.8	12.1	25	18.8	16.9	3.1 (2)	0.06
BDI	39	17.0	13.8	10	2.0	2.5	7	9.9	7.7	22	26.0	10.9	27.4 (2)	<.0001
CAPS	44	43.9	30.3	10	4.6	7.6	8	18.4	7.4	26	66.9	12.9	140.4 (2)	<.0001
Lifetime CAPS	36	67.4	34.9	8	12.5	12.2	7	75.0	15.7	21	85.7	20.5	47.5 (2)	<.0001

CES - Combat Exposure Scale; MAST - Michigan Alcoholism Screening Test; BDI - Beck Depression Inventory - II; CAPS - Clinician Administered PTSD Scale

Those with unremitted PTSD had smaller right hippocampal size compared to those who never developed PTSD (Figure and Table 2). The right hippocampus did not differ between subjects in sustained remission and those who never developed PTSD. The left hippocampus did not differ among diagnostic groups (see Figure, p=0.162). Neither the right, nor the left amygdala differed among diagnostic groups (p=0.132, p=0.639, resp.).

Figure:

Hippocampal volumes by diagnostic category

Footnote: Diagnosis: N: Never developed PTSD; S: PTSD in sustained remission; U: Unremitted PTSD

Table 2:

Final Model

Parameter	Estimate - Right Hippocampus	Standard Error	T Value	P Value	95% Confidence Limits
Diagnosis: PTSD in Sustained Remission	−0.1	0.2	−0.4	0.687	−0.6	0.4
Diagnosis: Unremitted PTSD	−0.6	0.2	−3.1	0.003	−1.03	−0.2
Diagnosis: Never developed PTSD (Reference category)	0
Intracranial Contents	0.0006	0.0007	0.96	0.345	−0.0007	0.002
Age	−0.03	0.02	−1.3	0.206	−0.07	0.02

CIE: Change in estimate; SE: Standard Error

The basic model results are presented in Supplemental Table 2. In the testing for confounding (Supplemental Table 3), adding depression to the basic model resulted in −13% CIE of diagnosis (i.e., Unremitted vs. Never developed), and 67% change in SE; thus, depression was colinear, not a confounder, and omitted from the final model (Supplemental Table 3). Adding age resulted in 22% CIE and 10% change in SE; therefore, it was a confounder and included in the final model. Adding CES resulted in 2% CIE and 2% change in SE; therefore, it was not included. Adding MAST resulted in −6% CIE and a 10% change in SE; therefore, it was not included. Race and marital status did not differ among diagnostic groups and were not included.

Our final model included the basic model plus age and intracranial volume as confounders (Table 2). Our main association of interest remained significant in this final model.

Discussion

Our results replicate and broaden those of Apfel et al. (9). We found reduced hippocampal volume in individuals with chronic, non-remitting PTSD, but not in those who recovered from or never developed PTSD. These findings bolster studies that have demonstrated an inverse correlation between hippocampal volume and the duration and severity of PTSD, i.e., smaller hippocampal size is associated with more severe and chronic PTSD (8). However, both our and Apfel’s results advance the literature by demonstrating that lower hippocampal size is found only in those who failed to recover from PTSD rather than in everyone who developed PTSD. Whereas the Apfel et al.(9) study included Gulf War veterans, our sample comprised Vietnam veterans 3 decades after the original trauma exposure. We further found these results to be lateralized to only the right hippocampus, consistent with previous MRI studies of male combat veterans with chronic PTSD (3, 4, 8). This association between smaller right hippocampal volume and unremitting PTSD may define a subset of combat veterans whose PTSD follows a decades-long, chronic, unremitting course (2).

The hypothesis that hippocampal volume may primarily be associated with a failure to recover from PTSD is consistent with the role of the hippocampus in fear conditioning and extinction. Animal studies have demonstrated that the hippocampus participates in context-based fear conditioning (18, 19), and in the extinction of conditioned fear responses (20). The hippocampus may influence extinction by encoding the relationship between distal cues in the contextual environment, i.e., ‘allocentric’ processing (21). That is, the hippocampus appears to integrate individual elements into a relational or configural representation of the external environment, which has been demonstrated in humans (22, 23).

In a previous study, we found that impairment on a paper-and-pencil measure of allocentric processing was related to both PTSD severity and hippocampal volume (23). The lateralized hippocampal findings of the present study are consistent with prior findings (24) that the right hippocampus is specifically involved in allocentric processing. Therefore, the right hippocampus may play an important role in the contextual processing of environmental cues (e.g., the ability to utilize safety cues in the environment), and ultimately in the failure to recover from PTSD. The validation of these specific mechanisms is, of course, beyond the scope of the current study.

Our findings do not appear to be explained by the comorbid variables common to PTSD. Despite the potentially confounding nature of age, depression, and lifetime alcohol use with severity of PTSD, we were able to rule-out these variables as relevant explanatory factors, finding them either to be collinear rather than confounding (depression), or not impacting our results when included as a covariate (age).

Our findings potentially shift the relevance of smaller hippocampal volume from ‘development’ of PTSD to ‘recovery’ from PTSD. This may also alter the field’s perspectives on the nature of the debate concerning hippocampal volume as a ‘pre-existing versus acquired’ feature in PTSD. Utilizing a similar population sample and identical MRI techniques, our current findings of right hippocampal diminution closely mirror those reported in our previous twin study of discordant monozygotic twins (4), in which a pre-existing vulnerability pattern was observed. Furthermore, our subsequent report (23) that the association between allocentric processing, PTSD, and hippocampal volume was consistent with pre-existing vulnerability supports a similar interpretation of the current findings, particularly if we hypothesize that allocentric operations are relevant to the recovery/extinction process. Of particular note, Ben-Zion et. al. (25) found evidence for smaller right hippocampal volume in “non-remitted” (but not “remitted”) PTSD patients over the course of a 18-month longitudinal study, and concluded that their results support the ‘vulnerability trait’ hypothesis. Ben-Zion’s (25), Apfel’s (9), our twin study (4), and our current results support the vulnerability hypothesis. Animal studies support this hypothesis as well. Frankland et al. (18) found that animals that received hippocampal lesions before fear conditioning were unable to fully differentiate aversive versus non-aversive contextual environments and failed to effectively extinguish conditioned fear responses in a new chamber that had no history of aversive conditioning. These results were replicated in mice with genetic mutations associated with compromised hippocampal function (18). Therefore, to the extent that hippocampal diminution is relevant to recovery from rather than development of PTSD, we suggest that the most parsimonious interpretation of our findings supports pre-existing vulnerability, viz., a pre-existing smaller hippocampus predisposes to a failure to recover from PTSD rather than to the development of PTSD per se.

A few limitations of the current report deserve mentioning. The first major limitation is that MRI results were done cross-sectionally, 30 years after the trauma. The acquisition of MRI hippocampal volume shortly following combat trauma and then 30 years following trauma was unavailable in the current sample. Without measuring the hippocampal volume prior to the trauma, it is impossible to draw conclusions about the causal associations between it and PTSD. For example, the hippocampal volume may have decreased overtime due to the influences of hormonal or other neurobiological consequences of chronic unremitting PTSD. In opposition to this hypothesis, Ben-Zion et al. (25) observed stable hippocampal volumes in both remitting and non-remitting patients during their 18-month longitudinal follow-up, failing to find any evidence for progressive, stress-related atrophy due to PTSD during first 2 years after trauma when the PTSD symptomatology is most severe. Nevertheless, a causal inference from our findings can only be drawn from future longitudinal research that includes at least 2 timepoints of hippocampal measurement, the first one before or close to the traumatic event. A second limitation is that we did not collect information on possible confounders such as treatment history or socioeconomic status. For example, it is possible that different treatment histories between the remitted and unremitted PTSD veterans resulted in both different clinical outcomes and hippocampal volumes. Although we did not characterize specific treatments received by our study subjects, our sample comprised veterans receiving treatment as usual in the VA comprising medications and psychotherapy. A third limitation is the relatively small and uneven sample size across 3 groups, although such sample sizes are usual in imaging studies. An additional limitation is that our sample was almost exclusively White (97.7%) and male.

In summary, findings of the current study support an emerging literature suggesting that the relevance of reduced hippocampal volume observed in PTSD may be related to failure to recover from rather than to the development of PTSD following trauma exposure, and that this failure to recover can extend for many decades. We further demonstrate that these results are specific to smaller right hippocampal size, consistent with existing literature on extinction-relevant allocentric processing. These findings may define a subpopulation of combat veterans who by virtue of their smaller right hippocampi are at risk for long-term failure to recover from PTSD. These findings may be used to create novel treatments targeting failure to recover. One example of such pathway is teaching allocentric extinction skills which are compromised by reduced hippocampal capacity.