Event-related potential studies of post-traumatic stress disorder: a critical review and synthesis

Although post-traumatic stress disorder (PTSD) is classified as an 'anxiety disorder', evidence of cognitive and information processing (IP) abnormalities in PTSD has been accumulating [1]. While many studies on emotional processing abnormalities in PTSD exist, event-related potentials (ERPs) studies focusing on early stages of IP abnormalities in PTSD are limited in number. The aim of this review was to summarize ERP findings in PTSD and determine whether there are consistent patterns of IP deviations reported in this disorder. We also sought to gain possible insight into clinical correlates of these differences. Another aim of this review was to assess if we could present suggestions for future research methods.

A detailed Medline search was performed. Keywords included PTSD or post-traumatic stress disorder and EEG, electrophysiology, P50, P100, N100, P2, P200, P3, P300, sensory gating, CNV (contingent negative variation), and MMN (mismatch negativity). The search was limited to human studies reported in the English language. Because of the very small number of electrophysiological studies in children with mental illnesses in general, three of the manuscripts which reported children studies were not included in this review. Papers which were solely EEG or sleep studies in PTSD without inclusion of ERPs were also not included. We considered only studies which included a healthy control group, enabling extraction of the differences between ERP responses in patients with PTSD and those of healthy participants. We then reviewed full texts of the selected manuscripts and summarized ERP responses from these studies in separate tables (Tables 1, 2, 3 and 4). We also extracted reported clinical correlates of the ERP results and summarized them in Table 5. This table shows the rating scales which were used in each study and the correlation of the scores of these tests with the ERP data.

Initial search yielded 57 papers of which 16 were solely EEG studies, another did not include a healthy control group, and three were children or adolescents studies, which were excluded from the current work. We excluded studies in children because of the small number of reports and a lack of homogeneity in ERP procedures used. Only a single study assessed CNV differences in PTSD patients [13] and thus it was not included in this review. After exclusion of these papers, 36 manuscripts had undergone comprehensive review. All of the presented data regarding the populations, their clinical characteristics (such as rating scales and scores when provided), and the utilized event-related response measures were collected from these 36 reports. Given the fact that most of the studies did not report possible comorbidities and medication regimens, these factors are not assessed or discussed in this review.

Of the 36 studies, 15 included veterans and/or military personnel with combat/war trauma and 17 included subjects exposed to civilian traumas such as motor vehicle accidents, rape and assault. Subjects of four studies were exposed to traumas of mixed etiology. All of the considered studies included healthy control subjects and most of them (especially combat related studies) included a third group of people who were exposed to trauma but did not develop PTSD based on the Diagnostic and Statistical Manual of Mental Disorders (DSM) III or IV diagnostic criteria.

A number of rating scales were used to assess PTSD symptomatology and comorbidities (Table 5). The most widely used scale was the Clinician-Administered PTSD Scale (CAPS), utilized in 10 of the 14 studies that evaluated clinical correlates. The CAPS is a well validated 'gold standard' 30-item structured interview that corresponds to the DSM-IV criteria for PTSD [14, 15]. CAPS subscores include re-experience (intrusiveness), avoidance and arousal. Other assessment instruments included the Mississippi scale for PTSD [16] (one study), a PTSD questionnaire [17] (one study), State-Trait Anxiety Inventory (STAI) [18] (three studies), brief symptom inventory [19, 20] (one study), Beck depression inventory [21] (two studies), Impact of Events Scale-Revised [22] (IES-R) (four studies), PTSD Check List-Military (PCL-M) [23] (one study), and Profile of Mood States (POMS) [24] (one study).

Standard tone sounds were used for the auditory oddball and gating studies. For the visual event-related potential paradigms, mostly three sets of images were presented to the subjects. These included neutral images such as nature scenes, trauma-specific aversive images and non-trauma-specific aversive pictures. Auditory stimuli were used in 22 studies, 13 used visual stimuli, and one study included both auditory and visual stimuli in their paradigms.

Most of the studies focused on the ERP component P300 (26 studies). P50, N100 and P200 components were also assessed by a fair number of studies (9 studies P50; 13 studies N100; and 10 studies P200). P100 (visual), N200 and the MMN were the least studied ERP components. Many studies probed more than one ERP component; this explains why the sum of the number of entries in tables 1, 2, 3 and 4 is larger than the total number of papers reviewed. The ERP components that were considered by at least four studies are summarized in the tables.

We first grouped all the studies together for each ERP component and counted the number of studies which reported differences in the amplitudes in each direction. We then separated studies in auditory/visual modalities and amplitudes reported in response to the trauma related (TR) and trauma nonrelated (TNR) stimuli when applicable.

In this manuscript, we reviewed reports that assessed ERP components in subjects with PTSD and healthy subjects. The early stage of IP was not examined due to a paucity of studies. Most of the studies presented evidence for a difference in IP between patients with PTSD and control patients during the entire midlatency range and extending into the later stages of IP.

The earlier part of the midlatency range is almost entirely pre-attentive and reflects the automatic processes of stimulus registration (reflected by response amplitudes) and filtering processes (reflected by sensory gating measures). Processes occurring at this stage are likely substantially bottom-up in nature and might be clinically correlated with 'intrusiveness' and 'hyperarousal', as these symptoms are likely to be automatic and, to a degree, pre-attentive. On the other hand, the attentive top-down level of IP is usually probed by examining the later occurring ERPs like the P300 or P550. This level of information processing could be clinically related with volitional 'avoidance' symptoms.

Results of reviewed P50 gating studies show some evidence of impaired pre-attentive habituation in PTSD patients. Whether gating deficit results from exposure to stress or represents a pre-existing vulnerability to developing PTSD can only be determined through the conduct of careful longitudinal studies. Current literature shows that decreased gating is not uncommon in seemingly healthy individuals [44, 45]. On the other hand, studies also show that laboratory-induced stress (which, by definition, cannot be severe or chronic) can decrease gating in healthy individuals with normal baseline gating [46]. It is thus quite plausible to postulate that severe or chronic stress can be detrimental to the sensory gating function and possibly individuals with premorbid deficient sensory gating function are more susceptible to developing PTSD when subjected to this form of stress.

The noted possibility that the N100 and P200 components may be affected differently in association with PTSD is worthy of further investigation as the P200 along with the N100 form the vertex complex and are considered closely related. While experimentally dissociable, the two components share similar neural sources and topographical distributions [47]. A significant variance in the effects of stimulation on these two components would indeed be an interesting finding worthy of further exploration, as it could yield clues to the nature and timing of IP difficulties in this patient population. In fact, despite the small number of studies, a trend suggesting that the two components behave differently in association with PTSD is seen.

The most widely examined ERP component is the P300, which reflects task allocation of cognitive resources [3]. In a first glance at the reports that evaluated the P300 component, it is difficult to draw a conclusion as the results are contradictory. When divided into two groups of studies with presentation of visual and auditory stimuli, a higher number of auditory studies show reduced P300 amplitude while a larger number of visual studies show increased P300 amplitude in the PTSD patients. This could be due to the fact that most of the auditory studies are standard oddballs with presentation of neutral stimuli while a larger number of studies in the visual modality presented TR/aversive images as non-target stimuli. In other words, when results are examined in terms of relevance of the stimuli to the trauma, they appear to be more meaningful. Among 18 studies that reported P300 amplitude in response to the TNR/neutral stimuli in auditory/visual modalities, the majority - 13 studies - showed reduced amplitude while only two of the studies showed increased amplitude. Furthermore, among the eight studies which evaluated P300 responses to the TR/aversive stimuli, seven studies found increased P300 amplitude in patients with PTSD when compared to the control group. Of interest, the one study which showed reduced P300 amplitude in PTSD patients, in within group comparison, showed a smaller response to the neutral words as compared to the positive and negative words in PTSD patients. These findings suggest the existence of two possible P300-related abnormalities: a significant decrease in responses to TNR stimuli and a significant increase in TR stimuli. Whether these two abnormalities occur simultaneously or sequentially and whether these two abnormalities reflect different aspects of the PTSD syndrome are currently open questions. Based on the above, there might be an increased attentional resource allocation in subjects with PTSD to the cues of trauma at the expense of the neutral stimuli. In other words, subjects with PTSD show sensitization and impaired habituation to the stimuli which represent the traumatic stimuli. Considering the limited attentional resources (which might be even more limited in patients with PTSD due to causes such as traumatic brain injury), this shift in allocation of these resources (sensitization) takes place at the expense of reduced response to and processing of neutral stimuli. In a clinical language, this attentional phenomenon may be translated into hyperarousal to any reminder of the traumatic event at expense of the neutral and nonaversive stimuli. In a further speculative step, this sensitization to the TR cues/stimuli, may be extended to the top-down filtering of the TR memories and assumed to be responsible for impaired inhibition of internal TR stimuli/memories which present in the form of re-experiences, intrusive memories and flash backs.

Whether ERP findings in individuals with PTSD are the result of trauma, or are present in patients who are more susceptible to PTSD when subjected to a traumatic experience remains an open question. Hypersensitivity to aversive stimuli and lack of habituation to them could be a premorbid difference in people who are more susceptible to developing PTSD after a traumatic experience. This possibility becomes more important in light of the fact that most ERP components are heavily genetically influenced [48]. Whether the ERP findings are characteristic of people susceptible to PTSD or are results of the trauma needs to be further assessed in longitudinal cohort studies (for example, pre- and postdeployment studies in veterans). Twin studies may be utilized in assessing whether the different findings in subjects are inborn genetic characteristics or a result of traumatic experience. This review found only two twin studies, both of which reported differences in P200 and P300 components between the PTSD subjects and their non-PTSD identical twin [49, 33]. Although these results suggest that the information processing differences are acquired, more studies are needed to support these findings.

Tables 1, 2, 3 and 4 show that, although the amplitudes of the ERP components are examined in most of the reviewed studies, literature is very sparse in terms of the latency of these components. A negative correlation between the latency of the P300 and severity of avoidance has been found [50], in other words, the stronger the avoidance, the shorter the P300 latency. As latency of a component reflects the number of synaptic links leading to the development of the response [3], this observation might suggest that an altered process might be engaged in some patients with PTSD. Furthermore, the same study presents a positive correlation between latency and severity of the intrusiveness symptoms. Increased latency might thus suggest altered IP pathways in PTSD and support the need for close examination of latencies in ERP-PTSD studies. Differences in the direction of correlation between avoidance and latency, and intrusiveness and latency in the same patient population may point at different pathways being involved in the formation of different symptom clusters in PTSD. While highly speculative (based on a single report) these findings are of significant interest and deserve further exploration.

Given the large number of ERP variables examined as well as the large number of PTSD symptoms (and the varied PTSD assessment methods) no clear trends can be discerned (Table 5). If ERP components indeed reflect specific symptom type or severity, this can potentially be useful in both diagnostic evaluation and treatment monitoring. However, in those studies in which clinical correlations were assessed, comparable rating scales were not utilized. As a result, for the purpose of future ERP-PTSD studies, scales providing detailed accounts of symptom severity would be most suitable to examine any clinical-electrophysiological correlations. More importantly, it will be vital to analyze and include symptom-cluster information in the reports given the current debate over the make-up of symptom composition of the PTSD clusters. Whether ERP methodologies and measures can be used to guide the development of new or revised, empirically-based symptom clusters might also be a fruitful consideration in future studies.

In summary, our review points to a potential relationship between ERP measures and PTSD. The most consistent findings include diminished habituation to repetitive stimuli as evidenced by reduced P50 gating, and sensitization of the P300 response to TR stimuli. The P300 response to neutral stimuli, on the other hand, is diminished in PTSD patients, suggesting a potential 'trade off' between processing traumatic versus neutral stimuli. These differences in IP are consistent with the findings in emotional processing studies in PTSD, which reveal increased emotional response to the cues of trauma. Corresponding differences in IP and emotion responses can help in a more comprehensive understanding of PTSD. Although emotion regulation and processing is more extensively addressed in research, early cognitive processing of the information needs further exploration for a better understanding of the whole picture. The findings of this review may also translate to better understanding of the mechanisms involved in clinical symptoms development. Results on the N100 and P200 components are not as conclusive and other components are not addressed in the majority of the studies.

The inconsistency in the results of different studies can stem from differences in methodologies and patient populations or differences in illness-related variables. Other contributing factors can be physiological or external factors such as patients' personality traits, pre-existing factors (such as comorbid illnesses, presence of alcohol or drugs and tiredness) or contributing conditions (availability or lack of support, use of medications such as morphine, beta blockers, and benzodiazepines) prior to or immediately after the traumatic experience. Unfortunately, many of the reviewed literature failed to assess or report comorbid illnesses or medication regimens. Prospective investigations with more uniform methodologies, unified patient populations, and pre- and post-trauma approaches are necessary to further explore such relationships.