Social cognitive performance in posttraumatic stress disorder: A meta-analysis

Background: Social support represents a key factor in the development of post-traumatic stress disorder (PTSD). Social cognition – the ability to perceive, interpret, and respond to other people – is considered fundamental in making use of social support. Gaining knowledge on the link between PTSD and social cognition is therefore essential. Whilst social cognitive difficulties in patients with PTSD are documented, an understanding of which particular social cognitive processes might be affected more than others, is lacking. The current meta-analysis was therefore aimed to examine social cognitive functioning in four underlying social cognitive domains (mentalization, emotion recognition, social perception, and attributional style) in PTSD diagnosed patients versus controls. Methods: Meta-analyzes were conducted on studies examining performance on at least one social cognitive domain in PTSD diagnosed patients compared to controls. Results: 19 studies were included, involving 565 patients and 641 controls. Relative to controls, the PTSD group scored lower on overall social cognitive functioning (SMD = − 0.42), specifically on mentalization (SMD = − 0.81) and social perception (SMD = − 0.30), whilst the latter should be interpreted with caution as only one study was found pertaining to this domain. No emotion recognition and attributional style differences were found. Limitations: There was evidence of moderate heterogeneity in the results of the included studies for overall social cognition and attributional style. Conclusions: Findings indicate that social cognition represents a potential important clinical factor in PTSD and underscore the importance of differentiating between underlying social cognitive processes in research and


Introduction
Social support represents a key factor in the development of posttraumatic stress disorder (PTSD) -characterized by intrusive thoughts or flashbacks, avoidance of trauma-related stimuli, negative alterations in mood and cognitions, and hyperarousal features (American Psychiatric Association, 2013). Social support can act as a buffer, as it can enhance coping with stress responses following trauma. Indeed, a lack of social support reflects a major risk factor for PTSD as shown in two meta-analyzes (Brewin et al., 2000;Ozer et al., 2003). Additionally, studies have shown that a lack of social support is linked to the maintenance of PTSD (Dirkzwager et al., 2003;Schnurr et al., 2004). Evidence hence indicates that social support plays a central role in PTSD.
Whilst the role of social support in traumatized individuals is wellestablished, it is less well understood if and how PTSD in turn affects the use of social support. It is plausible that the symptoms of PTSD itself, such as negative alterations in mood and cognition, decrease the ability to perceive, interpret, and respond to other people. These abilitiescollectively falling under the definition of social cognition (Green et al., 2008) are considered fundamental in effectively making use of social support (Sharp et al., 2012). Social cognitive processes can be subdivided into four domains (Green et al., 2008). These include the interpretation of behavior of others based on their mental states (mentalization or theory of mind); the identification and recognition of emotional states from social stimuli (emotion recognition); the recognition of social regulations, rules, and goals in social situations (social perception); and the explanation of social situations (attributional style). To understand how patients diagnosed with PTSD can use social supportive signals more effectively, it seems essential to further understand the association between PTSD and different social cognitive domainsrepresenting the primary goal of this study.

The association between PTSD and social cognitive domain functioning
Research suggests that core PTSD features like emotional dysregulation, emotional numbness, and feelings of interpersonal detachment may be accompanied by an inability to adequately perceive and infer social information (e.g. Lavoie et al., 2014;Mazza et al., 2012;Stevens and Javanovic, 2019).
Particularly, PTSD diagnosed individuals depict emotion dysregulation problems (Boden et al., 2013;Ehring and Quack, 2010;Shepherd and Wild, 2014), involving an inability to process emotional cues such as recognizing others' emotional expressions (Passardi et al., 2018;Poljac et al., 2011;Schmidt and Zachariae, 2009;Steuwe et al., 2014). As such, Poljac et al. (2011) compared war veterans with PTSD to a matched control group and found that traumatized veterans were less accurate in recognizing certain facial emotional expressions (fear and sadness) during a computerized emotion recognition task. Passardi et al. (2018) also found that the recognition of positive emotional expressions was impaired in a PTSD sample, in which larger impairments were linked to a higher number of reported traumatic events and moderate levels of trauma related dissociation. The latter converges with evidence showing that PTSD symptomatology, such as increased levels of dissociation, are associated with decreased performance on social cognitive tasks, including emotion recognition tasks (Nazarov et al., 2015;Nazarov et al., 2014).
Moreover, ample studies indicated decreased performances on mentalization tasks (e.g., Mazza et al., 2012;Mazza et al., 2015;Nazarov et al., 2014). PTSD related mentalization problems are for instance evidenced by PTSD diagnosed military police officers performing less compared to healthy participants in a task in which they had to describe what another person might feel in certain emotional contexts (Mazza et al., 2012). These authors also found that social cognition deficits were predicted by avoidance and numbing symptoms and they accordingly formulate that "central symptoms of PTSD as avoidance/emotional numbness or social withdrawal and emotional paralysis may underlie a deficit in the acquired inability to understand others' mental states and to understand/share emotions with others" (p. 254).
Furthermore, cognitive models of PTSD (Ehlers and Clark, 2000) postulate that traumatized individuals are biased towards appraising environmental social cues as negative and threatening. To illustrate, Elwood et al. (2007) studied how victims and non-victims of interpersonal violence interpretate short videos of social scenes. Results indicated that victims perceived positive valenced videos as risky and threatening videos as more likely to escalate to threat, as compared to non-victims. The latter finding is supported by accumulating evidence showing biased attributional processing in traumatized individuals (e.g., Boffa et al., 2018;Elwood et al., 2007;Kimble et al., 2012).
Lastly, a narrowly oriented attention bias towards social threat might impair the processing and interpretation of various contextual cues (Bomyea et al., 2017;Naim et al., 2015), which is considered essential in social perception (Green et al., 2008). Taken together, previous work indicates that patients with PTSD show difficulties in different domains of social cognition.

The current study
Whilst social cognitive difficulties in patients with PTSD are documented, an understanding of which particular domains might be affected more than others, is still lacking. To our best knowledge, only Lavoie et al. (2014) compared functioning on all four domains in PTSD and other anxiety disorder groups versus non-clinical controls. They assessed studies on social cognition in PTSD and findings indicated deficits in the domains of mentalization and emotion recognition, suggesting that different social cognitive processes might be differentially affected in PTSD. Addressing domain functioning in PTSD is relevant, as it will inform which specific social cognitive processes should be targeted clinically, which may then help traumatized individuals to perceive and use social support more effectively, as a protective factor against PTSD symptomatology. Moreover, interventions targeting specific social cognitive processes are shown to be more effective compared to broad-based or comprehensive interventions, that target multiple social cognitive processes (see Roelofs et al. 2016 for a meta-analysis). Social cognitive aspects may also be targeted through preventive practices. Specifically, as conceptualized by the latent-vulnerability model (McCrory and Viding, 2015), already existing problems in social cognitionfor instance due to early maltreatment which may distort the perceptions of oneself and others (see also Sharp et al., 2012) -may represent a latent vulnerability factor suited for prevention programs insofar as problems with social cognition may complicate the use of social support following traumatic experiences later in life, potentially increasing the risk for developing manifest clinical conditions such as PTSD.
The aim of the current review was to examine functioning among traumatized individuals diagnosed with PTSD in the four social cognitive domains. Domain functioning was compared to non-clinical controls using a meta-analytic approach. In line with previous work, we hypothesized that the PTSD group will demonstrate deficits in all four domains compared to non-clinical controls. Hypotheses about the degree as to which domains were more affected than others in terms of effect sizes were exploratory of nature.

Method
This study was conducted in line with the Preferred Reporting Items for Systematic Reviews and Meta-Analyzes (PRISMA) guidelines (Moher et al., 2009).

Eligibility criteria
Articles eligible for inclusion were: (1) English written peer-reviewed articles, (2) that reported or provided data on request to calculate between-group standardized mean difference (SMD) estimates (i.e., sample size, means, and standard deviations), (3) in which the task performance was compared on at least one social cognitive domain, (4) between two adult age groups (≥18 years old) including one clinical group primarily diagnosed with PTSD based on formal criteria of the Diagnostic and Statistical Manual of Mental Disorders (DSM-III, DSM-VI, DSM VI-R, DSM-5; American Psychiatric Association, 2013) versus a control group without such a diagnosis.
In line with Lavoie et al. (2014), social cognitive domain tasks were operationalized as follows. In mentalization or theory of mind (TOM) tasks, mental states or beliefs of a target person were identified or inferred (response based on information about both the target person and the context). For example, in the Strange Stories test (Mazza et al., 2012), short stories describing emotional situations were presented in which participants were instructed to infer what the target person might feel given the defined context.
Emotion recognition tasks involve the identification or discrimination of emotional states (response based on information of a target person without any contextual information). For example, emotional states were identified based on facial expressions (e.g., Steuwe et al., 2014), vocal cues (e.g., Nazarov et al., 2015), or eye regions (e.g., Mazza et al., 2012;Nazarov et al., 2014).
Social perception tasks included the detection of social roles, rules, or goals that characterized particular social situations (response based on contextual information). For example, the Interpersonal Perception Task-15 (Nazarov et al., 2014) consisted of short vignettes of social interactions accompanied by objective multiple-choice questions concerning the interactions (e.g., judging the relationship between two interacting people as "acquaintances" or "lovers for four years").
Attributional style tasks measured how participants tended to explain events. There were two types of attribution tasks. First, tasks indicating whether participants assessed social events as positive or negative (e.g., rating the valence of hypothetical events such as thinking of the moment that "a past memory suddenly comes to mind"; e.g., Boffa et al., 2018). Second, tasks measuring to what extent hypothetical events were explained according to a depressive bias style, in which events were attributed to internal, stable, and global causes (Gray et al., 2003;McCormick et al., 1989).

Data extraction
The first and second author independently screened the titles and abstracts of all retrieved articles for possible inclusion. Potentially eligible articles were evaluated full-text using the inclusion criteria. Disagreements were resolved through discussion and if this was not possible by consultation with the last author. The following information was extracted from all included articles: authors; year of publication; title; sample size, age, and gender composition of the PTSD group and the control group; social cognitive domain(s); tasks used; and task outcome data (M and SD) to calculate between-group standardized mean differences (SMD).

Data analysis
Review Manager 5 software (Version 5.3, The Nordic Cochrane Centre, Copenhagen, Denmark) was used to conduct meta-analyzes and to generate forest plots. SMDs including 95% confidence intervals (CI) were calculated using a random effects model in which the SMD represented the between-group mean difference divided by the pooled standard deviation. SMDs were weighted using the inverse variance method. SMDs were interpreted as small (0.2-0.5), moderate (0.5-0.8), or large (>0.8) based on Cohen (2013).
Five pooled SMD estimates were calculated: one for each social cognitive domain separately and one for all four domains combined. If needed, SMDs were inverted to ensure that a negative SMD signified lower PTSD performance relative to controls. For studies utilizing more than one measure to assess the same domain (e.g., two mentalization tasks were used in Mazza et al., 2012), those measure outcomes were pooled before analyzes, resulting in a single domain estimate per study. When task outcomes were not presented as a total score, but as several subscale scores (e.g., recognition rates of happy, fearful, and angry faces in Bell et al., 2017), those subscale scores were also pooled before analyzes. When studies examined multiple control groupsfor instance, comparing a mission related PTSD group versus both with/without mission exposed control groups without PSTD (Mazza et al., 2012) the control group that corresponded the most to the PTSD group was included in the analyzes (i.e., mission exposed without PTSD).
Between-study statistical heterogeneity was analyzed using the χ 2 test of significance and the I 2 statistic. The I 2 statistic indicates the percentage of variance between studies that is due to differences between studies rather than chance (Higgins et al., 2003). The degree of heterogeneity was interpreted as low (I 2 = 25%), moderate (I 2 = 50%) or substantial (I 2 = 75%). When study outliersdefined as when a study's confidence interval did not overlap with the overall pooled effectappeared, they were excluded and analyzes were run again (Viechtbauer and Cheung, 2010).

Study selection
A flowchart of the selection process is shown in Fig. 1. The literature search yielded 2262 records. The references of these studies yielded eight additional records. When duplicates were removed 1734 records remained. Of these, 1684 records were excluded based on title and abstract. The full text of the remaining 50 records was examined. Of these, 19 studies met the inclusion criteria and were included in this metaanalysis.

Meta-analytic results
Meta-analyzes were conducted to analyze between-group differences on overall social cognition, mentalization, emotion recognition, and attributional style in PTSD diagnosed versus controls. It was not possible to run a meta-analysis for social perception, as only one study was found in this domain.

Social perception
One of the 19 studies compared social perception between PTSD diagnosed and controls (Nazarov et al., 2014). This study reported a significant lower social perception performance in PTSD diagnosed participants compared to controls with a moderate SMD of − 0.30, 95% CI [− 0.62, 0.02].

Discussion
The aim of this meta-analysis was to examine overall social-cognitive functioning and performance differences on four underlying social cognitive domains (mentalization, emotion recognition, social perception, attributional style) in PTSD compared to controls. As expected, the analyzes demonstrated a deficit with a medium effect size for overall social cognition in the PTSD group compared to controls, not varying as a function of trauma type (comparing studies involving samples exposed to mostly war trauma versus interpersonal trauma), gender or age as suggested by the results of the post-hoc analyzes. It should be noted that post-hoc analyzes were conducted for explorative reasons as they were probably underpowered. A moderate amount of between-study heterogeneity was found with respect to overall social cognition. The latter could potentially be a consequence of the notion that social cognition represents an overarching construct that is built on separate yet related underlying domains. Put differently, it could hence support the idea that probing partitioned social cognitive domains in PTSDwhich is our primary study interestmight be essential. To this end, our hypothesis with respect to observing impaired domain functioning in PTSD was only partially supported. The PTSD group displayed lower performances with large effect sizes for mentalization and social perception, yet the latter should be interpreted with caution as only one study pertaining social perception was found. No group differences were found for emotion recognition and attributional style. When overall social cognition was partitioned into four domains, only the latter domain showed heterogenetic study findings (which is elaborated on below).

Social cognitive functioning and domain performances
Given that overall social cognitive performance was lower in the PTSD group and that social cognitive skills might be considered vital in the ability to perceive and respond to social supporta welldocumented protective factor for the PTSD development (Brewin et al., 2000;Ozer et al., 2003) our findings underscore that social cognition is an important clinical target in PTSD (Sharp et al., 2012). To target PTSD related social cognitive deficits effectively, it is key to identify which specific social cognitive domains are affected.
Accordingly, our results revealed that the ability to mentalize was impaired in PTSD versus controls, which converges with existing  (Boffa et al., 2018;emotion recognition in Mazza et al., 2012, Schmidt and Zachariae 2009, Solomon et al., 1991, as the CI's of these studies and the overall estimate did not overlap. evidence (Mazza et al., 2012;Nazarov et al., 2014). On the one hand, problems in mentalizing may be explained by the PTSD symptom profile. Specifically, evidence suggests that core PTSD symptoms, such as emotional numbing, being emotionally unresponsive and feelings detached form others (American Psychiatric Association, 2013), can potentially make it difficult to process and integrate social information (Mazza et al., 2012;Lavoie et al., 2014). Indeed, Mazza et al. (2012) for instance depicted that social cognition deficits were predicted by avoidance and numbing symptoms. Furthermore, neuroimaging findings showed that PTSD symptoms and impaired mentalization are linked to overlapping disrupted brain regions, including hyperactive amygdala reactivity and a less activation in the medial prefrontal cortex (Pitman et al., 2012;Zoladz and Diamond, 2013). These neuroimaging findings further strengthen the idea that mentalization might be explained as a consequence of PTSD symptomatology. On the other hand, problems with the ability to infer self and others' behavior based on mental states might be seen as a predisposing factor placing individuals at risk of developing PTSD symptomatology following trauma. For instance, in their social cognitive model of PTSD, Sharp et al. (2012) postulate that, based on experiences with attachment figures, individuals build attachment schemes that helps to understand the self and others. As such, insecure attachment patterns can result in basic negative reflections of the self and others, which is proposed to adversely affect social cognitive capacities. The latter subsequently make it difficult for an individual to make use of social support, potentially increasing the risk for developing PTSD following trauma (Sharp et al., 2012). In line  (Mazza et al., 2012;Schmidt and Zachariae, 2009) and one on attributional style (Boffa et al., 2018) were excluded from the forest plots, as the CI's of these studies and the overall estimate did not overlap. No study outliers were found on mentalization.
with the latent-vulnerability model (McCrory and Viding, 2015), social cognitive deficits such as problems with mentalization may hence pose a latent vulnerability factor for developing psychiatric symptoms following trauma later in life. Taken together, current findings point out that PTSD is accompanied by a mentalization deficit, yet follow-up research has to determine the direction of causality between these two constructs.
Our findings further demonstrated a diminished social perception ability in PTSD relative to controls. However, these results must be interpreted with caution as only one study was found for this domain. The finding nevertheless suggests that traumatized individuals experience problems in perceiving and understanding their social context, emphasizing the importance of future research on PTSD and social perception. Processing various contextual cues is suggested to be impaired due to a narrowly oriented attentional bias to environmental threat (Bomyea et al., 2017;Naim et al., 2015). Addressing the latter is relevant; given that studies have shown that a lack of perceived social support is related to PTSD symptom severity (Hofman et al., 2016;Jankowski et al., 2004).
Contrasting our hypothesis, no emotion recognition impairment was found in the PTSD group. This is consistent with some studies (e.g., Bell et al., 2017;Nazarov et al., 2014;Simmons et al., 2011), but diverges from others (e.g., Poljac et al., 2011;Steuwe et al., 2014;Schmidt and Zachariae 2009). However, these latter studies mostly incorporated individuals with interpersonal trauma (e.g., childhood abuse, interpersonal combat). It may be that these clinical groupsgiven their trauma aetiologyrespond differently to the social-emotional cues that were used in the emotion recognition tasks. Another explanation for finding no emotion recognition deficit might be that in some traumatized samples exposure to traumatic events or threatening situations lead to no reduced or even enhanced emotion recognition. For example, Masten et al. (2008) showed that maltreated children showed faster emotion recognition capabilities, interpreted as being functional for detecting threat in dangerous environments. Our current findings thus suggest no emotion recognition deficit in PTSD, but future studies should determine to what extent the recognition of emotional states depends on trauma aetiology.
Finally, the PTSD group did not differ in attributional style from controls. This was not anticipated, as cognitive models of PTSD predict that ambiguous stimuli or events are interpreted negatively (Ehlers and Clark, 2000) and biased processing is likely to be disorder-specific (e.g., van den Heuvel et al. 2005). One explanation for the lack of a difference may thus be that the included studies primarily examined causal explaining of general events. However, it may be essential to investigate trauma-specific causal attributions in PTSD (Gonzalo et al., 2012). Indeed, several studies have shown that internal, global, and stable attributions of particular traumatic events were associated with PTSD symptoms (e.g., Gray et al., 2003;Wenninger and Ehlers, 1998). For example, it has been consistently shown that internally focused self-blame attributions following sexual assault were related to PTSD symptom development (e.g., Berman et al., 2018;Boeschen et al., 2001;Filipas and Ullman 2006;Najdowski and Ullman 2009). Trauma-specific attributions was also shown to be more predictive of PTSD symptoms compared to a more broad-based dispositional attribution style (Gray et al., 2003). The focus on trauma-specific attributions may also explain why only Boffa et al. (2018) reported a negative attributional bias with a large effect size in PTSD, as this was the only study that used a PTSD specific measure of attributional style. Future research should further explore the role of general versus trauma-specific attributions in PTSD.

Limitations
The main limitation reflects that there was evidence of moderate heterogeneity in the results of the included studies for overall social cognition, which might be explained by between-study variability of the studies examining attributional style, as only this domain showed a moderate level of heterogeneity after social cognition was partitioned into four domains. Heterogenetic findings on attributional style might be a consequence of variations in used measures. Specifically, the inclusion of studies using indices to probe attributions pertaining to a general spectrum of events, instead of probing trauma-specific attributions, may be the reason of observed variations across these studies.
Second, while some studies controlled for neurocognitive functioning (e.g., executive functioning in Nazarov et al., 2014, intelligence in Nietlisbach et al., 2010 and psychiatric comorbidity (e.g., secondary depressive disorder in Mazza et al., 2012), potential confounding effects of such or other unmeasured factors could not be excluded. Future studies should take such factors into account.
Third, while the present study included data concerning all social cognitive domains, the separate meta-analyzes of the domains included a relatively small number of studies. The meta-analysis could not be conducted for social perception and the results for the other domains may have been underpowered, implying that significant effects may have been undetected. Hence, there is a need for studies comparing PTSD groups with non-clinical controls on social cognitive indices in order to further elucidate functioning across social cognitive domains in traumatized individuals. The incorporation of more studies would make it possible to further probe and detect potential effects of partitioned social cognitive domains, which would lead to an increased understanding with regard to the social cognitive profile in traumatized individuals. An increased understanding is of clinical importance, given that it could provide insights as to which domains could be seen as targets to intervene on.
Fourth, the findings were based on cross-sectional between-group comparison studies. Future longitudinal research should determine whether social cognitive deficits are a risk factor for developing PTSD symptomatology following trauma or to what extent traumatic events and the presence of PTSD contribute to problems in social cognition.

Conclusion
Regardless of these limitations, this study indicates that PTSD is accompanied by social cognitive impairments. The findings demonstrated that underlying social cognitive domains were differentially affected, indicating impairments in mentalization and social perception specifically. No deficiencies were found for emotion recognition and attributional style, yet future research should determine the role of trauma aetiology. Social cognition is an important clinical factor in the aetiology and maintenance of PTSD. The current study underscores the importance of differentiating between underlying social cognitive domains in research and treatment of PTSD.