The amygdala is perhaps the most strongly implicated brain structure in the pathophysiology of posttraumatic stress disorder (PTSD). Prevalent models of anxiety have focused on an amygdalocentric neurocircuitry1
that is critical in fear response, conditioning, generalization2,3,4
, and facilitates the response to stressful experiences5
. Functional MRI studies6–9
have shown that individuals with PTSD have an exaggerated amygdala response to emotional stimuli when compared to controls. Animal studies have demonstrated changes in amygdala morphology with chronic stress10
, evident primarily in the growth of dendritic spines. Experimental studies of amygdala volume in mice and humans have shown an association between smaller amygdala volumes, increased levels of fear conditioning, and an exaggerated glucocorticoid response to stress11–13
. However, efforts to find evidence of an association between amygdala volume and PTSD in humans have produced equivocal results14–15
. Our goal was to reinvestigate amygdala volume changes in PTSD by addressing some of the potential methodological issues contributing to inconclusive findings.
The hippocampus has been the overwhelming focus of prior studies of morphological change in PTSD. These studies have demonstrated a clear association between smaller hippocampal volume and PTSD15–16
. In contrast, the relatively few studies of amygdala volume differences in PTSD have yielded mixed results (see ). Two meta-analyses that included amygdala volumetry showed inconsistent differences between trauma-exposed participants with and without PTSD. The first meta-analysis found lower volumes with small effect sizes17
in both the left (effect size= −.22) and right amygdala (−.18)15
, but only after restricting analysis to the subset of studies that produced a homogeneous sample. The second meta-analysis demonstrated only a trend association (p
= .06), with a small effect size (Hedges’ g
= −.29), between smaller left amygdala volume and PTSD patients as compared to trauma-unexposed healthy controls14
. The interpretation of these data is complicated by several methodological issues and limitations. For instance, the first meta-analysis15
included a group with intrusive memories but lacked a diagnosis of PTSD, the use of varied manual segmentation protocols, differences in raters across studies, scanners with lower spatial resolution and field strengths than current standards, heterogeneous trauma type (combat, sexual assault, intimate partner violence, etc.), and inclusion of children18
. Although these meta-analyses offer weak support for an association of amygdala volume and PTSD, several individual studies that failed to show significant differences19–26
suggest caution in this interpretation (see ). Furthermore, the number of studies reporting negative results may be an underestimation of the actual number given the disincentives for publishing negative results.
Our study had three main goals. The first was to assess the association of amygdala volume and PTSD in a large sample of trauma-exposed adults. The effect sizes from several prior studies that yielded marginal p
-values for amygdala differences21–22,24,26
suggest they were underpowered to detect real effects in the population. Therefore, we hypothesized that the PTSD group would show smaller amygdala volume than the trauma-exposed non-PTSD group when powered to adequately control for Type II error. We previously reported27
the sample size required to demonstrate a significant group difference in amygdala volume derived from automated segmentation using published effect sizes15
to be at least 55 subjects per group, which is substantially larger than sample sizes used in prior published studies (see ).
The second goal was to gain empirical evidence that might offer clues about factors contributing to PTSD. For example, if trauma exposure, trauma load, or illness chronicity were correlated with amygdala volume, this would be compatible with either an environmental cause for the volume change, or a pre-existing vulnerability that interacts with environmental factors such as trauma load and illness chronicity. Evidence of an association with PTSD in the absence of an association with trauma exposure would allow for the possibility that smaller amygdala volume may represent a vulnerability to PTSD. We hypothesized that a diagnosis of PTSD would be associated with smaller amygdala volume, but that volume would not be correlated with trauma exposure or illness chronicity based on evidence in animals and humans that smaller amygdalae constitute a risk for heightened fear and stress responses11–13
Our third and final goal was to confirm the established finding of smaller hippocampal volume14–15
in our sample of veterans serving after 11-September 2011 using the structural neuroimaging methods we have adopted. These methods included several enhancements such as an automated segmentation approach to better control variability in manual segmentation protocols by human raters, improved signal-to-noise ratio (SNR) provided by 3 Tesla field strength, and higher spatial resolution (1-mm isotropic voxels). For analyses of both structures, we used a multiple regression approach similar to Bremner and colleagues21
to control for variables such as symptoms of depression, trauma load, duration of PTSD, intracranial volume (ICV), age, medication, and alcohol abuse.