Emotion Recognition and Adverse Childhood Experiences in Individuals at Clinical High Risk of Psychosis

Abstract

Objective

To investigate the association between facial affect recognition (FAR) and type of adverse childhood experiences (ACEs) in a sample of clinical high risk (CHR) individuals and a matched sample of healthy controls (HCs).

Methods

In total, 309 CHR individuals and 51 HC were recruited as part of an European Union-funded multicenter study (EU-GEI) and included in this work. During a 2-year follow-up period, 65 CHR participants made a transition to psychosis (CHR-T) and 279 did not (CHR-NT). FAR ability was measured using a computerized version of the Degraded Facial Affect Recognition (DFAR) task. ACEs were measured using the Childhood Experience of Care and Abuse Questionnaire, the Childhood Trauma Questionnaire, and the Bullying Questionnaire. Generalized regression models were used to investigate the relationship between ACE and FAR. Logistic regressions were used to investigate the relationship between FAR and psychotic transition.

Results

In CHR individuals, having experienced emotional abuse was associated with decreased total and neutral DFAR scores. CHR individuals who had experienced bullying performed better in the total DFAR and in the frightened condition. In HC and CHR, having experienced the death of a parent during childhood was associated with lower DFAR total score and lower neutral DFAR score, respectively. Analyses revealed a modest increase of transition risk with increasing mistakes from happy to angry faces.

Conclusions

Adverse experiences in childhood seem to have a significant impact on emotional processing in adult life. This information could be helpful in a therapeutic setting where both difficulties in social interactions and adverse experiences are often addressed.

Introduction

Social cognition alterations have received markedly increased attention in the recent years due to their possible relationship with increased liability to psychosis.¹ Patients with psychosis have difficulties in the interpretation of facial emotional expressions, and as these difficulties are detectable both in remission and in the acute phase of the illness^2–4 and in unaffected first-degree relatives,⁵ they may represent a trait rather than a state effect. In an attempt to clarify this, attention has focused on social cognition skills, including facial emotion recognition, both in patients at the early stages of the illness and in those at clinical high risk (CHR) of psychosis. Several studies have reported social cognition impairments in CHR individuals,^1,6–8 including alterations in facial affect recognition (FAR), which has been proposed as a possible endophenotype related to the genetic risk of development of psychosis.^9,10

Prior studies in the CHR population have reported alterations in FAR ability^7,8,11–13; however, findings are mixed^14–16 and the observed difficulties do not seem to be related to a specific type of emotion. For example, van Rijn et al¹¹ reported difficulties in the recognition of neutral facial expressions and misattribution of neutral faces as angry; Kohler et al¹⁷ reported difficulties in the recognition of angry and fearful expressions, but not of neutral ones, whereas Amminger et al.⁸ reported deficits in the recognition of fear and sadness. The severity of the observed FAR difficulties has also varied across different studies. For example, Thompson et al¹⁴ reported no impaired FAR in CHR subjects, whereas Leppänen et al¹⁸ identified pronounced difficulties similar to those seen in patients with established psychosis. FAR performance has been also studied in relation to clinical outcome. Recent studies^7,13 examined whether emotion recognition was predictive of transition to psychosis. Contrary to their initial hypotheses, Allott et al¹³ found that total face and prosody emotion recognition performance did not predict transition to psychosis whereas better recognition of fearful and worse recognition of neutral faces was predictive. Addington et al⁷ investigated face and prosody emotion recognition in a large sample (CHR = 172). This study found no differences in FAR across groups and no relationship with subsequent transition to psychosis.

The variability of results across studies could in part be explained by the employment of different tasks.^19,20 Tasks, in fact, varied in terms of the type of emotions examined, response time, and format and quality of the stimulus (eg, degraded vs nondegraded). In addition, with the exception of a few large studies,^7,8,21 most of the previous ones have used relatively small samples which may have limited their statistical power.²⁰ Differences in experienced adverse events between samples may also contribute to heterogeneity of findings. A further factor is the heterogeneity of the CHR population; inclusion criteria vary across studies^22,23 (including the inclusion of low social and occupational functioning as criterion)²²; only a small proportion of CHR individuals will develop psychosis²⁴; at the time of the assessment, they might be in different disease stages²⁵; and, even when they do develop psychosis, they might present different symptoms.

Adverse childhood experiences (ACEs), which are commonly reported by CHR individuals, are known factors that can affect FAR ability. For example, da Silva et al ²⁶ reported that FAR was impaired in children with a history of traumatic experiences of abuse and neglect. Other studies showed that even types of abuse considered “less severe,” ²⁷ such as emotional neglect, can lead to alterations of the neural and attentional systems involved in the processing of facial expressions. ²⁷ ACEs are highly prevalent in individuals with psychosis^28,29 as well as in the CHR population^30,31 and have also been associated with increased risk of transition from CHR to psychosis.^32,33 A recent meta-analysis³⁴ highlighted that CHR individuals report more frequent and severe ACEs than healthy controls (HCs). In a recent study, Kraan et al³⁵ reported that a history of emotional abuse in particular was associated with an increased risk of transition to psychosis. Independently, childhood adversities have also been associated with alterations in social cognition, including the processing and recognition of facial emotion expressions.^26,27,36 Based on the theory that children adjust their emotional perception through the learning of social experiences, childhood adversity has been suggested to change sensory thresholds, leading to less effective regulation, processing, and recognition of emotions.³⁷ This could confer greater vulnerability to psychosis as, according to the sociodevelopmental-cognitive model of psychosis, developmental alterations associated with enhanced genetic vulnerability, early brain insults, and ACEs might result in dysregulation of the dopaminergic system which in turn can lead to symptoms of psychosis.³⁸

The mechanisms underlying the observed impairments in facial emotional processing in CHR individuals have yet to be clarified. In particular, it remains unclear to what extent the observed FAR impairments could be associated with ACEs and if associations between ACEs and FAR are different in CHR individuals compared to the general population.

In the present study, we investigated the relationship between ACEs and FAR in a large sample of CHR individuals and a matched sample of HCs who were recruited as part of the European Union Gene–Environment Interactions (EU-GEI) Study,³⁹ a multicenter, prospective, naturalistic study. We predicted that (1) FAR ability in CHR individuals (CHR-T and CHR-NT) with a positive history of ACEs would be worse than in individuals without such a history, and that (2) the impact of ACEs on FAR ability will be more pronounced in CHR (CHR-T and CHR-NT) than in HC. To provide a complete picture, we also analyzed the direct associations between FAR and psychosis risk and predicted the following: (3) FAR ability would be worse in CHR individuals than in HC, and (4) worse FAR ability would be associated with the risk of subsequent transition to psychosis.

Methods

Sample

In total, 344 CHR participants and 67 HCs were recruited as part of EU-GEI study³⁹ from 11 centers (London, Amsterdam, Den Haag, Vienna, Basel, Cologne, Melbourne, Copenhagen, Paris, Barcelona, and Sao Paolo) from July 2010 to August 2015 and were clinically followed up for at least 24 months. The design of the study and the inclusion and exclusion criteria for both CHR and HC have been described elsewhere.³³ The study received ethical approval at each included site.

Measures

Sociodemographics and Clinical Data

Detailed sociodemographic characteristics were assessed using the modified Medical Research Council Sociodemographic Schedule.^39,40 The Comprehensive Assessment of At Risk Mental State (CAARMS)²² was used to measure subclinical psychotic-like symptoms and to determine the transition to psychosis. The Structured Clinical Interview for DSM Disorders (SCID)⁴¹ was used to establish the presence of other psychiatric disorders and to exclude the presence of current psychotic disorders.

Facial Affect Recognition Task

A computerized version of the Degraded Facial Affect Recognition (DFAR) Task ^19,42–44 was used to measure FAR ability. The task, which has been described in previous works,^19,42–44 shows images of four different actors (two males and two females) representing four emotions: angry, happy, fearful, and neutral. The task involves 64 trials consisting of 16 presentations on each of these emotion categories. Participants were asked to indicate the emotional expression of each image by a button press. To increase the task difficulty, images were passed through a filter resulting in a reduced visual resolution by 30%. Higher scores on the DFAR are indicative of a better ability to recognize facial expressions in that particular emotion. Results show that the proportion of images correctly recognized as neutral, happy, fearful, and angry, and also the overall proportion of correct answers. In addition, the direction of the misattribution for each emotion was also computed (eg, when a participant incorrectly attributes neutral to angry expressions).

Adverse Childhood Experiences Measures

The short version of the Childhood Experience of Care and Abuse questionnaire (CECA-Q)⁴⁵ and the brief version of Child Trauma Questionnaire (CTQ-B)⁴⁶ was used to measure the ACEs up to the age of 17. The CECA-Q assesses traumatic experiences such as the death of a parent, separation from parents (including being in foster care), parental discordance, lack of adult support, poverty, cruelty, and violence. These different measures of ACEs were categorized as present or absent.

The CTQ is a 25-item self-report questionnaire that assesses five domains: emotional abuse, emotional neglect, sexual abuse, physical abuse, and physical neglect. Each item uses a 5-point scale to identify the frequency or severity of the experience (from 1 [never] to 5 [almost always]). Validated cutoff scores were used to compute the presence or absence of specific traumatic experiences.⁴⁷ The CTQ subscales were dichotomized as present or absent using the following cutoff scores: physical abuse ≥8, sexual abuse ≥6, emotional abuse ≥9, physical neglect ≥8, and emotional neglect ≥10. The different types of trauma were considered as “present” when scores were above the cutoff.The Bullying Questionnaire⁴⁸ was used to measure the severity and frequency of bullying before the age of 17.

Statistical Analyses

Analyses were performed using SPSS version 25.⁴⁹ Sociodemographic data were analyzed using means and standard deviations for continuous data and frequencies for categorical data. Analysis of the variance (ANOVAs) with Bonferroni correction for multiple comparisons was used to examine group differences in continuous variables, including the overall differences in the DFAR task. Chi-squares were used to assess differences in adverse experiences variables across groups. Two binary logistic regression models (the first including DFAR total and DFAR scores for each emotion; the second including all misattributions) were used to analyze the relationship between baseline FAR ability and transition to psychosis in the CHR group. Age, gender, IQ, ethnicity, and recruitment site were entered as covariates. In order to investigate the relationship between FAR and ACEs, firstly univariate analyses were performed considering FAR ability as a dependent variable and each ACEs variable as independent variables. Secondly, for each group (HC, CHR, CHR-T, CHR-NT), five generalized linear models (ie, total DFAR, neutral, happy, frightened, and angry conditions) were performed entering as independent variables those ACE variables (a complete list can be found in supplementary table S3) with a statistical significance of P < .15 in the univariate analyses⁵⁰ or those which have been found as significantly related to FAR ability in the literature (ie, physical and emotional abuse/neglect and sexual abuse⁵¹). In each analysis, age, gender, IQ, ethnicity, and recruitment site were entered as covariates. To analyze the relationship between DFAR performance and ACEs, clinical and sociodemographic variables, a gamma with log link linear distribution was assumed and Bonferroni correction was applied to the P-values of the marginal means derived from each tested model. When we analyzed the effect of emotional abuse, all the other types of abuse were entered as covariates.

Results

Sociodemographics and Clinical Data

A total of 411 individuals were assessed at baseline (CHR = 344; HC = 67); 16 HC and 35 CHR participants were excluded, as they did not complete the DFAR task (see supplementary tables S1 and S2 for a comparison between included/excluded samples). The final sample comprised 309 CHR and 51 HC. At 24-month follow-up, 58 (18.8%) CHR individuals had made a transition to psychosis (CHR-T) while 251 had not (CHR-NT). Baseline sociodemographics are detailed in table 1. There were no significant differences across groups except for IQ and employment status, which were significantly higher in HC compared to CHR-NT and CHR-T.

Table 1.

Sociodemographic And Clinical Characteristics

	HC (N = 51)	CHR-NT (N = 251)	CHR-T (N = 58)	Statistics
Mean age in years (SD)	23.37 (3.98)	22.63 (4.89)	22.67 (4.86)	F(2, 357) = 0.53, P = .59
Gender male, N (%)	27 (53%)	131 (52%)	33 (57%)	χ2 = 0.42, P = .81
Mean IQ (SD)	110.44 (17.84)a,b	98.87 (17.5)	96.96 (14.48)	F(2, 335) = 10.64, P < .0001
Ethnicity
White, N (%)	33 (65%)	187 (75%)	39 (67%)	χ2 = 14.26, P = .16
Ever employed, N (%)	48 (94%)a,b	190 (79%)	38 (70%)	χ2 = 9.59, P = .008
CHR inclusion group, N (%)
Genetic Vulnerability		37 (17%)	12 (22%)	χ2=.91, P = .43
APS		200 (87%)	57 (91%)	χ2=.78, P = .5
BLIPS		17 (8%)	5 (9%)	χ2=.12, P = .45

APS: attenuated psychotic symptoms; BLIPS: brief limited intermittent psychotic symptoms; CHR-NT: clinical high risk non transition; CHR-T: clinical high risk transition; HC: healthy controls.

^astatistically significant differences between HC and CHR-NT, after Bonferroni correction for multiple comparisons.

^bstatistically significant differences between HC and CHR-T, after Bonferroni correction for multiple comparisons.

Facial Emotion Recognition And Adverse Childhood Experiences

Below we report results for the CHR group (overall group and CHR-T and CHR-NT separately) and the HC group.

CHR group

The experience of emotional abuse in childhood was significantly associated with lower score in the DFAR total (β = −0.05, SE = 0.03, P = .04) and in the neutral condition (β = −0.07, SE = 0.03, P = .03) The experience of the death of a parent was significantly associated to worse neutral emotion recognition (β = −0.12. SE = 0.05, P = .01). Lack of adult support was significantly associated with worse angry emotion recognition (β = −.09, SE = 0.04, P = .02). The frequency of bullying was significantly associated to better recognition of frightened faces (β = 0.1, SE = 0.45, P = .02) and higher total DFAR score (β = 0.05, SE = 0.02, P = .01; table 2A).

Table 2.

Relationship Between Adverse Childhood Experience Variables and the DFAR Task (Adjusted Results).

A
	CHR					HC
	β	SE	P **	Exp(ß)	95% CI	β	SE	P **	Exp(ß)	95% CI
DFAR total
Bullied frequency, P* =.02	.05	.02	.01	1.05	[1.01, 1.1]	.05	.03	.16	1.05	[.98, 1.12]
Death of a parent, P*= .067	−.07	.04	.08	.93	[.87,1.01]	−.18	.08	.02	.83	[.71, 0.97]
Taken into care, P*= .015	−.06	.50	.21	.94	[.85,1.03]	.13	.13	.29	1.14	[.89, 1.47]
Lack of adult social support, *P  = .015	−.03	.02	.15	.97	[.92, 1.01]	−.05	.06	.37	.95	[.85, 1.06]
Emotional abuse, P* = .048	−.05	.03	.04	.95	[.9, 0.99]	−.07	.05	.16	.93	[.83, 1.03]
Neutral
Death of a parent, P*= .015	−.12	.05	.01	.88	[.81, 0.98]	−.15	.14	.28	.86	[.66, 1.14]
Emotional abuse, P* = .1	−.07	.03	.03	.93	[.87, 0.99]	−.03	.08	.75	.97	[.83, 1.15]
Happy
Bullied severely, P  = .08	.04	.02	.11	1.04	[.98, 1.09]	.02	.03	.5	1.02	[.95, 1.1]
Death of a parent, P*= .057	−.08	.04	.06	.92	[.85, 1.01]	.14	.08	.09	1.15	[.98, 1.35]
Frightened
Bullied frequently, P*= .029	.1	.45	.02	1.1	[1.02, 1.22]	.02	.09	.79	1.02	[.81, 1.1]
Angry
Taken into care, P* = .033	−.12	.09	.18	.88	[0.73, 1.06]	−.27	.27	.33	.76	[.45, 1.31]
Lack of adult social support, P* = .001	−.09	.04	.02	.91	[.85, 0.99]	−.08	.13	.56	.92	[.71, 1.2]
B
	CHR-NT					CHR-T
	β	SE	P **	Exp(ß)	95%CI	β	SE	P **	Exp(ß)	95%CI
DFAR total
Bullied frequency, P*=.02	.05	.02	.004	1.05	[1.02, 1.09]	.3	.12	.02	1.35	[1.05, 1.72,]
Death of a parent, P*=.067	−.06	.03	.1	.94	[.88, 1.01]	−.08	.14	.59	.92	[.7, 1.22]
Taken into care, P*= 0.015	−.09	.04	.03	.91	[.85, 0.99]	.4	.34	.25	1.49	[.76, 2.94]
Lack of adult social support, *P  = .015	−.04	.02	.03	.96	[.92, 0.99]	.15	.13	.24	1.16	[.91, 1.49]
Emotional abuse, P* = .048	−.01	.02	.45	.99	[.94, 1.03]	−.37	.13	.004	.69	[.53, 0.89]
Neutral
Death of a parent, P*= 0.015	−.11	.06	.04	.89	[.8, 0.99]	−.07	.1	.48	.93	[.76, 1.14]
Emotional abuse, P* = .1	−.04	.04	.24	.96	[.89, 1.03]	−.27	.1	.006	.76	[.62, 0.92]
Happy
Bullied severely, P = .08	.02	.02	.11	1.02	[.99, 1.06]	.13	.14	.36	1.14	[.86, 1.49]
Death of a parent, P*= .057	−.06	.03	.05	.94	[.88, 1]	−.1	.2	.61	.9	[.61, 1.33]
Frightened
Bullied frequently, P* = .029	.11	.05	.02	1.12	[1.1, 1.22]	.27	.17	.1	1.3	[.95, 1.82]
Angry
Taken into care, P* = .033	−.17	.08	.04	.84	[.71, 0.99]	.56	.46	.22	1.75	[.71, 4.35]
Lack of adult social support, P* = .001	−.10	.04	.01	.9	[.84, 0.98]	.14	.14	.30	1.15	[.88, 1.51]

HC: healthy controls; CHR-NT: clinical high risk non transition; CHR-T: clinical high-risk transition; CI: confidence interval; SE: standard error

Β, standard error, expected β and P values are reported for all variables entered in the models. Interpretation of the exp(β): eg, in the CHR group who experienced bullying, the exp(β) for the DFAR total is 1.05. This means that the DFAR total of the group who experienced bullying is 1.05 times higher than the one who did not experience bullying. In other words, there is a 5% increase in accuracy on the DFAR total in the group who did experience bullying.

*P value in univariate analyses.

**P values after adjusting for gender, age, ethnicity, recruitment site, and IQ.

HC group

In the HC group, having experienced the death of a parent during childhood was significantly associated with lower DFAR total score (β = −0.18, SE = 0.08, P = .02). No other significant associations were found in this group (table 2A).

CHR-T and CHR-NT groups

In the CHR-NT group, having been taken into care and having experienced lack of social support in childhood were significantly associated with lower total DFAR (β = −0.09, SE = 0.04, P = .03; β = −0.04, SE = 0.02, P = .03) and with worse recognition of angry faces (β = −0.17, SE = 0.08, P = .04; β= −0.1, SE = 0.04, P = .01). Having experienced the death of a parent in childhood was significantly associated with worse neutral emotion recognition (β = −.11, SE = 0.06, P = .04) and worse happy emotion recognition (β = −.06, SE = 0.03, P = .05). CHR-NT individuals who experienced more frequent bullying performed better in the recognition of frightened facial expressions than those who did not (β = 0.11, SE = 0.05, P = .02; table 2B). In the CHR-T group, emotional abuse was significantly associated with worse DFAR total score (β = −0.37, SE = 0.13, P = .004) and worse recognition of neutral faces (β = −0.27, SE = 0.1, P = .006). In both CHR-T and CHR-NT groups, those individuals who experienced more frequent bulling obtained higher DFAR total score (β = 0.3, SE = 0.12, P = .02; β = 0.05, SE = 0.02, P = .004) than those who did not. No other significant associations were found in the CHR-T group (table 2B).

Facial Affect Recognition

There were no significant differences between groups (ie, HC; CHR-T and CHR-NT) in the number of mistakes in the DFAR tasks (table 3). There are some significant differences in the misattributions (supplementary table 4). CHR-T participants misattribute angry to happy faces significantly more than CHR-NT (F(2, 359₎ = 4.03, P = .02). CHR-T participants misattribute happy to frightened faces significantly more than HC participants (F(2, 359) = 3.28, P = .04).

Table 3.

DFAR Task Results Across Groups

DFAR task % of correct answers Mean (SD)	HC	CHR-NT	CHR-T	F (df), P
Total	75.7 (9.43)	76.52 (10.01)	73.76 (15.54)	F(2, 359) = 1.49, P = .23
Neutral	82.47 (15.31)	80.73 (16.04)	78.45 (16.89)	F(2, 359) = 0.87, P = .42
Happy	90.56 (9.47)	90.96 (11.52)	88.25 (19.2)	F(2, 359) = 1.05, P = .35
Frightened	60.54 (19.18)	61.01 (18.65)	60.56 (19.67)	F(2, 359) = 0.02, P = .98
Angry	69.24 (21.21)	73.38 (19.63)	67.78 (24.61)	F(2, 359) = 2.21, P = .11

HC: healthy controls; CHR-NT: clinical high risk non transition; CHR-T: clinical high-risk transition

Adverse Childhood Experiences

Descriptive data on ACEs are detailed in supplementary table 3 and have been already reported and discussed by Kraan et al.³³ A short summary for each instrument used is provided below.

CECA-Q

There were significant differences in the presence of parental discordance (HC < CHR-NT, χ2 = 5.86, P = .015), lack of adult support (HC < CHR-NT, χ2 = 12.11, P = .001; HC < CHR-T, χ2 = 15.31, P < .001); frequency of episodes of cruelty before age 11 (HC < CHR-NT, χ2 = 8.81, P = .003; HC < CHR-T, χ2 = 7.83, P = .005) and between age 12 and 16 (HC < CHR-NT, χ2 = 10.53, P = .001); and the frequency of episodes of violence before age 11 (HC < CHR-NT, χ2 = 12.33, P = .002; supplementary table 3).

C. TQ

There were significant differences in sexual abuse (HC < CHR-NT, χ2 = 9.84, P = .002; HC < CHR-T, χ2 = 6, P = .014), physical neglect (HC < CHR—T, χ2 = 17.55, P < .0001; HC < CHR-T, χ2 = 8.07, P = .005), emotional abuse (HC < CHR-NT, χ2 = 27.62, P < .0001; HC < CHR-T, χ2 = 19.25, P < .0001), and emotional neglect (HC < CHR-NT, χ2 = 39.29, P < .0001; HC < CHR-T, χ2 = 20.29, P < .0001; supplementary table 3).

Bullying

Bullying experiences were more severe in the CHR-NT group than in the HC group (χ2 = 9.86, P = .002; supplementary table 3).

Facial Emotion Recognition, Adverse Childhood Experiences, and Transition to Psychosis

Logistic regression analyses performed on the CHR sample only revealed a significant increase of transition risk with an increasing number of misattributions of happy to angry faces (β= 0.1, SE = 0.03; P = .006). No other significant associations between transition to psychosis and FAR ability were found.

Discussion

To our knowledge, this is the largest study to date that has examined the relationship between FAR ability and ACEs in CHR individuals. We found that emotional abuse and lack of adult social support in childhood were significantly associated with poor FAR ability in CHR participants. The experience of the death of a parent in childhood was significantly associated with poor FAR in both CHR and HC. In addition, the number of happy to angry misattributions was related to the incidence of later transition to psychosis.

Our first and second hypotheses were partially confirmed. Several associations between FAR ability and ACE were found. These were statistically significant in the CHR group but not in HC. This might suggest the possible presence of resilience mechanisms in HC or a possible interaction between ACE and other factors, such as genetic vulnerability,⁵² in CHR individuals leading to compromised FAR ability. The latter would further support the integrated sociodevelopmental-cognitive model.³⁸ However, the HC group was smaller than the CHR one and we cannot exclude that other associations would have been evident in the HC group if the sample had been as large as the CHR one. Yet, having experienced the death of a parent during childhood was associated with poor FAR ability in both the clinical and nonclinical groups. Early bereavement is considered as one of the most severe life events⁵³ and has been associated with several adverse outcomes, including a higher risk of developing mental and physical illness.^54–57 Interestingly, Fernández-Alcántara et al⁵⁸ did not find a significant association between the experience of the death of a parent and emotion recognition; however, participants included in that study experienced parental loss after age 18. The fact that we did find a significant association suggests that age of parental loss may be a key variable influencing social cognitive processes. If parental loss happens during the early developmental period, before the age of 17, it may have a significant impact on the ability to recognize emotions in others. To confirm this result, future studies should investigate this association by looking at different age spans.

Other forms of ACEs were associated with poorer performance on the FAR task in the CHR group. Thus, in the CHR group alone, emotional abuse was associated with a worse global FAR ability. Both emotional abuse and having experienced the death of a parent were associated with worse recognition of neutral faces. Lack of adult social support was associated with worse recognition of angry faces. Generalized difficulties in FAR have been widely reported in children with trauma antecedents,⁵⁹ and these difficulties seem to remain stable in adulthood.^43,60 This is in line with other studies which have reported that living in a neglected environment has an impact on the accurate recognition of others’ emotions.^37,61,62

When considering the CHR-NT and CHR-T groups separately, the larger CHR-NT group seems to be driving the significant results seen in the CHR group as a whole. Interestingly, in the CHR-T group but not in the CHR-NT nor in the HC one, emotional abuse was associated with worse total and neutral emotion recognition. This strengthens our previous findings³³ and confirms that emotional abuse seems to be an important risk factor for the subgroups who go on to develop psychosis. Except for bullying, associations between ACEs and FAR ability that are statistically significant in the CHR-NT group are not significant in the CHR-T group. There are at least two possible explanations for these negative findings. Firstly, it is possible that these ACEs are actually not related to the development of psychosis. Secondly, the CHR-T group is relatively smaller compared to the CHR-NT group hence the lack of significant results might be due to lack of statistical power. Post hoc power analyses that performed to test this hypothesis confirmed low statistical power (19%–38%) in these categories (supplementary table 5).

The frequency of bullying in the CHR group was associated with an overall better DFAR total score (in both CHR-NT and CHR-T) and better recognition of frightened expressions (CHR-NT). Previous studies investigating ACEs highlighted how bullying, compared to other ACEs, seems to have a distinct association both with psychosis⁶³ and with FAR.⁶⁴ Bullying usually happens in a school environment and therefore in middle and late childhood.⁶⁵ On a speculative level, the fact that this usually happens later in life might have resulted in a milder impact on the FAR skills that develop early in life as a result of the child relationship with the main caregivers.⁶⁶ The significant relationship between bullying and increased ability to recognize frightened faces might be interpreted as the result of increased interpersonal sensitivity associated with the bullying experience⁶⁷ or as a protective mechanism.

Contrary to our third hypothesis, FAR ability was not significantly different in CHR individuals and HC. This was true for both overall accuracy and each individual emotion. Although previous studies have reported differences in the ability to recognize facial emotions between CHR and HC,^6,8,68 others found no differences between CHR and HC,^14,16,69 or between CHR-T and CHR-NT.⁷ These inconsistent findings could, at least in part, be attributed to the use of different tasks to measure the FAR ability. For example, other large studies investigating FAR in the CHR population (eg, Addington et al.²¹ and Amminger et al.⁸) did not use a degraded-face task. In these studies, to increase task difficulty, response time was limited. In the present study, while there was no time limit and participants were instructed to be as accurate as possible, the stimulus was degraded. Although both type of studies manipulated the paradigm to increase task difficulty, one by manipulating time and the other one by manipulating the quality of the stimulus, the fact that this was done in different ways could have had an impact on the observed results. Inconsistencies could be also due to lack of adjustment for confounding variables (such as IQ⁶⁸), or the use of small samples.

Our fourth hypothesis was also only partially confirmed. Although no significant differences were found in the accuracy rate between groups, the number of errors from happy to angry was associated with subsequent transition to psychosis. This suggests that individuals who go on to develop psychosis are more likely to interpret happy degraded faces as angry. While this finding warrants replication in another CHR study, it would be in line with data from studies which have used other methods, such as virtual reality or imaging, to study mechanisms underlying paranoid ideation⁷⁰ and psychosis risk,⁶⁹ and with the notion that psychosis involves the attribution of salience to non-salient stimuli.⁷¹

Strengths and limitations

This study has several strengths. Firstly, this is the first study investigating the relationship between FAR and ACEs in CHR individuals, and the sample size was large. Secondly, the fact that the sample was recruited in 11 different centers in and outside Europe also suggests that results are likely to be generalizable. Finally, the emotion recognition task used has been previously used in a number of studies investigating emotion recognition in psychiatric and nonpsychiatric populations,^19,42–44 and the data analysis minimized the potential confounding effects of to age, sex, IQ, and ethnicity.

This study has a number of limitations. Firstly, the size of the HC group did not match that of the CHR sample, which may have reduced our power to detect statistically significant associations within the HC group. Secondly, ACEs were assessed retrospectively, hence the recall might have been subjected to bias.⁷² Thirdly, ACEs might lead to other changes, for example in stress-induced hormones and neurotransmitters,⁷³ which in turn might have conferred greater likelihood of social cognition/emotion recognition difficulties. These possible associations remain to be tested in future studies. Fouthly, we could not establish causality between ACEs and FAR ability. It is indeed possible that, for some, but not all ACE (eg, loss of a parent), impaired FAR could preceed and be related to ACE.⁷⁴ Fiftly, our results, especially the emotion-specific associations, could be specific to our sample and the type of ACEs experinced by our partecipants. Finally, study samples with and without DFAR information were significantly different in terms of age, ethnicity, group status, and IQ (supplementary table S1). This could have had an impact on the results.

Future studies

To further clarify the underlying mechanisms, future studies should investigate the relationship and possible mediating effect of other variables, such as genetic information. This might help in understanding some of the inconsistencies found in the studies investigating FAR. Genetic information has been collected and will be analyzed as part of the EU-GEI study.

Conclusions

ACEs are associated with emotional processing in adult life, particularly in individuals at CHR of psychosis. These findings could inform the delivery of therapeutic interventions aimed at the social cognition sequelae of early adversity.

Appendix

*EU-GEI High Risk Study Group Author

Philip McGuire¹

Lucia R. Valmaggia²

Mathilde Antoniades¹

Maria Calem¹

Matilda Azis¹

Sara Pisani¹

Lieuwe de Haan³

Mark van der Gaag^4,5

Eva Velthorst^3,6

Tamar C. Kraan³

Daniella S. van Dam³

Nadine Burger⁵

Barnaby Nelson⁷

Patrick McGorry⁷

G Paul Amminger⁷

Christos Pantelis^8,9

Athena Politis⁷

Joanne Goodall⁷

Anita Riecher-Rössler¹⁰

Stefan Borgwardt¹⁰

Charlotte Rapp¹⁰

Sarah Ittig¹⁰

Erich Studerus¹⁰

Renata Smieskova¹⁰

Rodrigo Bressan¹¹

Ary Gadelha¹¹

Elisa Brietzke¹²

Graccielle Asevedo¹¹

Elson Asevedo¹¹

Andre Zugman¹¹

Neus Barrantes-Vidal¹³

Tecelli Domínguez-Martínez¹⁴

Anna Racioppi¹⁵

Manel Monsonet¹⁵

Lídia Hinojosa¹⁵

Thomas R. Kwapil¹⁶

Mathilde Kazes^17,18

Claire Daban^17,18

Julie Bourgin^17,18

Olivier Gay^17,18

Célia Jantac^17,18

Marie-Odile Krebs^17,18

Dorte Nordholm,¹⁹

Lasse Randers,¹⁹

Kristine Krakauer¹⁹

Louise Birkedal Glenthøj¹⁹

Birte Glenthøj¹⁹

Merete Nordentoft¹⁹

Stephan Ruhrmann²⁰

Dominika Gebhard²⁰

Julia Arnhold²¹

Joachim Klosterkötter²⁰

Gabriele Sachs²²

Iris Lasser²²

Bernadette Winklbaur²²

Philippe A Delespaul^23,24

Bart P. Rutten²³

Jim van Os^1,25

1. Department of Psychosis Studies, Institute of Psychiatry, Psychology & Neuroscience, King’s College London, De Crespigny Park, Denmark 458 Hill, London, United Kingdom SE5 8AF.

2. Department of Psychology, Institute of Psychiatry, Psychology & Neuroscience, King’s College London, De Crespigny Park, Denmark Hill, 456 London, United Kingdom SE5 8AF.

3. AMC, Academic Psychiatric Centre, Department Early Psychosis, Meibergdreef 5, 1105 AZ Amsterdam, The Netherlands.

4. VU University, Faculty of Behavioural and Movement Sciences, Department of Clinical Psychology and EMGO+ Institute for Health and Care Research, van der Boechorststraat 1, 1081 BT Amsterdam, The Netherlands.

5. Parnassia Psychiatric Institute, Department of Psychosis Research, Zoutkeetsingel 40, 2512 HN The Hague, The Netherlands.

6. Icahn School of Medicine at Mount Sinai, department of Psychiatry, 1425 Madison Ave, New York, NY 10029.

7. Centre for Youth Mental Health, University of Melbourne, 35 Poplar Road (Locked Bag 10), Parkville, Victoria 485 3052, Australia; Orygen, The National Centre of Excellence in Youth Mental Health, The University of Melbourne, Australia.

8. Center for Neuropsychiatric Schizophrenia Research (CNSR) and Center for Clinical Intervention and Neuropsychiatric Schizophrenia Research (CINS), University of Copenhagen, Mental Health Centre Glostrup, Denmark.

9. University of Copenhagen, Faculty of Health and Medical Sciences, Dept. of Clinical Medicine, Copenhagen, Denmark.

10. University of Basel, Basel, Switzerland.

11. LiNC – Lab Interdisciplinar Neurociências Clínicas, Depto Psiquiatria, Escola Paulista de Medicina, Universidade Federal de São Paulo – UNIFESP.

12. Depto Psiquiatria, Escola Paulista de Medicina, Universidade Federal de São Paulo – UNIFESP.

13. Departament de Psicologia Clínica i de la Salut (Universitat Autònoma de Barcelona), Fundació Sanitària Sant Pere Claver (Spain), Spanish Mental Health Research Network (CIBERSAM).

14. CONACYT-Dirección de Investigaciones Epidemiológicas y Psicosociales, Instituto Nacional de Psiquiatría Ramón de la Fuente Muñiz (México).

15. Departament de Psicologia Clínica i de la Salut (Universitat Autònoma de Barcelona).

16. Department of Psychology, University of Illinois at Urbana-Champaign, IL.

17. INSERM, IPNP UMR S1266, Laboratoire de Physiopathologie des Maladies Psychiatriques, Université Paris Descartes, Université de Paris, CNRS, GDR3557-Institut de Psychiatrie Paris, France.

18. Faculté de Médecine Paris Descartes, GHU Paris – Sainte-Anne, Service Hospitalo-Universitaire, Paris, France.

19. Mental Health Center Copenhagen and Center for Clinical Intervention and Neuropsychiatric Schizophrenia Research, CINS, Mental Health Center Glostrup, Mental Health Services in the Capital Region of Copenhagen, University of Copenhagen.

20. Department of Psychiatry and Psychotherapy, University of Cologne, Cologne, Germany.

21. Psyberlin, Berlin, Germany.

22. Department of Psychiatry and Psychotherapy, Medical University of Vienna, Vienna, Austria.

23. Department of Psychiatry and Neuropsychology, School for Mental Health and Neuroscience, South Limburg 463 Mental Health Research and Teaching Network, Maastricht University Medical Centre, P.O. Box 616, 6200 MD 464 Maastricht, The Netherlands.

24. Mondriaan Mental health Trust, P.O. Box 4436 CX Heerlen, The Netherlands.

25. Department of Psychiatry, UMC Utrecht Brain Center, Utrecht University Medical Centre, 3584 CG Utrecht, The Netherlands.

Funding

The European Network of National Schizophrenia Networks Studying Gene–Environment Interactions (EU-GEI) Project is funded by grant agreement HEALTH-F2-2010–241909 (Project EU-GEI) from the European Community’s Seventh Framework Programme. Additional support was provided by a Medical Research Council Fellowship to M. Kempton (grant MR/J008915/1). S. Tognin is supported by by a Brain and Behavior Young Investigator award (NARSAD YI, 24786) and by a Maudsley Charity Grant (1510). G. Modinos is supported by a Sir Henry Dale Fellowship, jointly funded by the Wellcome Trust and the Royal Society (#202397/Z/16/Z). B. Nelson was supported by an NHMRC Senior Research Fellowship (1137687). N. Barrantes-Vidal was supported by the Ministerio de Ciencia, Innovación e Universidades (PSI2017-87512-C2-1-R) and the Generalitat de Catalunya (2017SGR1612 and ICREA Academia Award).

Conflict of Interest

G. Sachs is the president of the Austrian Society of Neuropsychopharmacology and Biological Psychiatry, which is partially financed by the support from pharmaceutical companies. G. Sachs received consulting fees and/or honoraria for speeches within the last three years from Angelini, AOP Orphan, Alkermes, Janssen, Lundbeck, Pfizer.