Of the seven loci at which combat-exposed twins with PTSD had lower gray matter density than combat-exposed twins without PTSD at a liberal threshold of p<0.001, four were located in predicted brain regions, viz., right hippcocampus, pregenual ACC, and left anterior and right mid- insulae, even though the predicted brain regions occupy less than 10% of total gray matter volume. This regional specificity supports the validity of the present results and implicates limbic and paralimbic structures as the major sites of gray matter density reductions in combat-related PTSD. Gray matter reductions in pregenual ACC and both insulae significantly correlated only with the cluster B “re-experiencing” symptoms of PTSD.
The ACC, especially its pregenual, or “affective,” division, and insula are components of the anterior “paralimbic belt,” are strongly interconnected to each other and to the amygdala, and are highly involved in emotional aspects of brain function (30
). Impaired pregenual ACC function is one of the most robust neuroimaging findings in PTSD (4
). A neurocircuitry model of PTSD posits that the ventromedial prefrontal cortex, including pregenual ACC, inhibits the expression of classically conditioned fear responses by the amygdala (33
). Thus impairment in this brain region might be expected to most affect the DSM-IV symptoms that are putatively most closely related to conditioned fear, viz., the cluster B symptoms (especially B.4 and B.5, viz., intense psychological distress [and/or] physiological reactivity on exposure to internal or external cues that symbolize or resemble an aspect of the traumatic event). To the extent that diminished structure implies diminished function, reduced pregenual ACC gray matter density is consistent with this neurocircuitry model.
Functional neuroimaging studies of the hippocampus in PTSD are less common, but they too support impairment in this brain region (34
). Hippocampal impairment may contribute to PTSD by reducing the ability to construct declarative narratives that bind the affect associated with the traumatic event (37
), or by the ability to recognize safe contexts (33
), or by other unknown mechanisms. The reduced gray matter density found in bilateral insulae is paradoxical in light of studies that have generally found hyperactivity in this brain region in PTSD (38
) and other anxiety conditions (39
). One model of anterior insula function posits that this structure detects the difference between an observed and expected body state and generates an interoceptive prediction signal that triggers anxiety (39
). A structurally compromised insula may be less inhibited in generating such signals in PTSD, but this is in the realm of speculation.
The most interesting result from the present study is the significant Diagnosis × Exposure interaction in the pregenual ACC, with combat-exposed PTSD twins having lower gray matter density than their own combat-unexposed co-twins as well as than the combat-exposed twins without PTSD and their co-twins, supporting the inference that pregenual ACC gray matter reduction is an acquired sign of PTSD. In animals, exposure to chronic stress has been shown to damage not only the hippocampus in rodents (40
) and primates (41
), but also the ACC in rodents (42
) and primates (44
). It has been hypothesized that such damage may provide a basis for structural changes observed in PTSD (42
). A recent study of mentally healthy persons that used automated segmentation found that those who reported early life stressors had smaller ACCs than those who did not (46
). However, causal inferences are difficult to draw from the cross-sectional study of non-twins.
When the Diagnosis × Exposure interaction at [8 50 12] was adjusted for MAST score, its statistical significance level was reduced to corrected p=0.10, which falls short of statistical significance. We did not obtain data regarding recent alcohol consumption. This is a limitation considering that imaging findings related to alcohol may be more sensitive to recent as opposed to more remote intake. On the other hand, the likelihood that increased alcohol use by the PTSD veterans accounts for the gray matter density reduction in their ACCs is diminished by the consideration that if the PTSD subjects studied here had consumed enough alcohol to damage their brains, evidence for this should have been found in other brain regions known to be affected by alcohol, including superior, motor, and other areas of the frontal cortex and cerebellum (47
), none of which (except for a small cluster in left inferior frontal cortex) showed volumetric reduction in the PTSD compared to the non-PTSD combat veterans at even the liberal threshold of uncorrected p<0.001. Thus, the specificity of volumetric diminution to our predicted brain regions argues against a global effect such as alcohol-induced brain damage. Finally, a recent manual tracing study found comparably (and significantly) reduced ACC volume in subgroups of PTSD veterans with and without a history of lifetime alcohol abuse or dependence, in comparison to non-PTSD veterans (49
When the Diagnosis × Exposure interaction in pregenual ACC was adjusted for number of potentially traumatic lifetime non-combat events, its statistical significance level also was reduced to corrected p=0.10. This means we cannot be fully confident that stressful events other than military combat do not account for the reduced ACC gray matter density in the PTSD veterans. However, even if such events did contribute, this would still not be inconsistent with stress-induced diminution of this structure. When the Diagnosis × Exposure interaction in pregenual ACC was adjusted for depression, it was no longer significant. This is not surprising given the high association between depression and PTSD in the present sample, in which self-reported depression appears to have been acquired along with PTSD, making the two likely facets of the same post-traumatic psychopathology.
A limitation of the design employed here is that it cannot identify the specific environmental difference(s) between the combat-exposed PTSD twins and their non-combat-exposed, identical co-twins that is responsible for an acquired abnormality. However, because the most salient, common difference in the present study was the presence of combat-related PTSD in the former, and because as noted above the observed effects remained significant or nearly significant after considering the contributions of several important potentially confounding variables, it is reasonable to attribute this lower gray matter density to the presence of combat-related PTSD.
Combat-exposed twins with PTSD also had lower gray matter density than combat-exposed twins without PTSD in right hippocampus and left anterior and right mid- insulae, as well as at another site within pregenual ACC, replicating previous studies. These results could not be explained by group differences in age, combat severity, number of potentially traumatic lifetime non-combat events, alcoholism, or child abuse. Unfortunately, the analyses that included the data from the combat-unexposed co-twins were unable to shed light on the origin of these gray matter reductions in the combat-exposed twins with PTSD, because they failed to yield either a significant difference between high- and low-risk combat-unexposed co-twins (which would support a pre-trauma vulnerability factor) or a significant Diagnosis × Exposure interaction (which would support an acquired abnormality). Finally, the present results were unable to replicate previously reported segmentation and voxel-based morphometric findings of gray matter reduction in subcallosal cortex and dorsal ACC.
In the same twin sample studied here, we previously found manual tracing evidence that diminished hippocampal volume represents a pre-trauma vulnerability factor for PTSD, rather than an acquired PTSD sign (15
). In contrast the present VBM results suggest that diminished volume in pregenual ACC is acquired as a result of the combat exposure that led to PTSD and/or the PTSD itself. We have no ready explanation as to why diminutions in these two structures should have different origins. As noted, above, however, the origin of gray matter density reduction in the pregenual ACC site other than the one that showed the significant interaction could not be explicated by the present data; it is possible that it represents a PTSD vulnerability factor. Additional techniques that may help to clarify this uncertainty in future studies include cortical parcellation (segmentation) and magnetic resonance spectroscopy.
It has been suggested that VBM may not detect very small, localized gray matter volume reductions, since false-negative VBM findings may arise from the changes in the shape or displacement of structures in the course of spatial normalization (7
). Additionally, VBM may be biased against finding group differences in areas that are spatially complex (50
). Inversely, we cannot rule out the possibility that the abnormalities detected by VBM in the present study reflected group differences in the shape of brain structures, rather than their volume (11
), although even shape differences may have functional consequences. The failure of VBM to find a significant hippocampal gray matter reduction in the high- vs. low-risk, combat-unexposed co-twins contrasts with our positive result in the same sample using manual segmentation of hippocampus (15
) suggests that the latter technique may be more sensitive to reduced volume in this structure than the voxel-based approach. Similarly, we are unable to rule out the possibility that subtle group differences in other brain regions in this study remained below the sensitivity of VBM, or the detection power conferred by our sample.