Emotion processing in maltreated boys and girls: Evidence for latent vulnerability

Evidence of alterations in emotion processing in maltreated youth has been hypothesized to reflect latent vulnerability for psychopathology. However, previous studies have not systematically examined the influence of psychopathology on the results. Here, we examined emotion recognition and learning in youth who differed in terms of presence vs. absence of maltreatment and psychopathology and tested for potential sex effects. Maltreatment and psychopathology were assessed in 828 youth (514 females) aged 9–18 years using diagnostic interviews and self- and parent-report questionnaires. Emotion recognition was assessed via identification of morphed facial expressions of six universal emotions. For emotion learning, reward and punishment values were assigned to novel stimuli and participants had to learn to correctly respond/withhold response to stimuli to maximize points. A three-way interaction of maltreatment by psychopathology by emotion indicated that when psychopathology was low, maltreated youth were less accurate than non-maltreated youth for happy, fear and disgust. A three-way interaction of sex, maltreatment and emotion indicated that maltreated girls and boys were impaired for fear, but girls showed an impairment for happy, while boys for disgust. There were no effects of maltreatment, psychopathology, or sex on reward learning. However, a two-way interaction between sex and maltreatment showed that maltreated girls were worse at learning from punishment relative to non-maltreated girls, while maltreated boys were better than non-maltreated boys. The study provides the first clear evidence of latent-vulnerability in emotion recognition in maltreated youth and suggests that girls and boys might be characterized by distinct profiles of emotion recognition and learning following maltreatment. Supplementary Information The online version contains supplementary material available at 10.1007/s00787-022-02132-1.


Supplement 1: Recruitment and Distribution of Participants per Group and Sex by Site
Recruitment took place between January 2014 and February 2018.In total, 1827 youths participated, including 880 with CD (61% females) and 947 TD youths (65% females).Data collection included questionnaires, behavioural, neuro-physiological and genetic measures, and structural and functional MRI data.Participants were recruited at 12 sites across Europe (Table 1).Sources included local clinics, mental health services, youth offending services, mainstream and special schools, youth clubs, other community services, outreach events and word of mouth.Supplement 6: Data Collection Procedure Referrals to the project were initially screened by telephone to assess suitability.
Participants judged likely to be eligible, and their parents or guardians, were then invited to a more comprehensive screening session, either at the local site (i.e., university or associated clinic), or at the participant's home.The first session generally lasted 1-3 hours.Informed consent was provided at the beginning of the session.Participants and their parents/guardians were interviewed separately by trained researchers to assess for the presence of any mental disorders.An IQ test was also administered.The caregivers responded to standardized probe questions aimed at investigating whether the child/adolescent had experienced i) harm on purpose, ii) harm in the course of discipline, iii) sexual harm, or iv) harm that led to the involvement of an agency (e.g., social services).Examples of probe questions included "Do you remember any time when the child was disciplined severely enough that he or she may have been hurt?" or "Has there been any other situation where your child has been harmed by someone (not accidents)?This could include sexual harm, such as flashing".Following each discussion, interviewers rated privately the probability that the child had been severely harmed.For instance, in the case of physical harm, using a criterion of intentional strikes to the child by an adult that left visible marks for more than 24 hours or that required medical attention, a score of 0 was assigned if maltreatment had definitely not or probably not occurred, and a score of 1 was assigned if maltreatment had probably occurred, definitely occurred, or if authorities had been involved.(Lansford et al., 2002).Each rater's judgement (i.e., no/probable/definite maltreatment) was compared with scores from a neutral rater and agreement between the independent raters was reported as 90% ((κ = 0.56) in previous studies (Dodge, Pettit, Bates, & Valente, 1995) and in ours.
Supplement 8: Schedule for Affective Disorders and Schizophrenia for School-Age Children: Present and Lifetime Version (K-SADS-PL) The K-SADS-PL (Kaufman et al., 1997)  administered separately to participants and parents (or another responsible adult informant) by trained researchers, and combined parent and child summary ratings of all symptoms (past, present and lifetime) were then generated.Where assessors gave discrepant ratings for a symptom, they discussed all available information until an agreement was reached for the summary rating.Except for CD, ODD, and ADHD, where DSM-5 criteria were used, all diagnoses were generated based on the DSM-IV-TR diagnostic criteria, which were current at the outset of the project (APA, 2000).Inter-rater reliability for current CD diagnoses was high (94.7%agreement across raters, Cohen's kappa=0.91).
Supplement 9: Inter-Rater Reliability Inter-rater reliability (IRR) values were calculated for comorbid disorders, notably ADHD), MDD, ODD and GAD (Cohen's kappas >= 0.84 and agreement across trained raters >= 92% for all diagnostic categories).The severity of CD and comorbid disorders was defined based on the number of symptoms endorsed across informants (i.e. a symptom was considered present if endorsed by either informant).The inter-rater reliability was conducted on a subsample of participants (n=75) from across all of the FemNAT-CD research sites.For ethical reasons, at the UK sites two interviewers attended the sames interviews but coded the interviews seperately; at all other sites, interviews were videotaped and scored by different raters.

Supplement 10: Imputation of Missing Data
Missing values of the PDS score were imputed based on the whole FemNAT-CD sample.
It has been shown that missing data in a multi-item instrument is best handled by imputation at the item level (Eekhout, de Vet, Twisk, Brand, de Boer, & Heymans, 2014).Thus, missing values of the single items were imputed first and the scores were calculated based on the imputed items.The imputation was done in SAS ® version 9.4 using the procedure PROC MI.Imputation by fully conditional specification (FCS) is used, which offers a flexible method to specify the multivariate

Supplement 13: ER and EL Results by largest data collection sites
The FemNAT-CD is a multi-site study and to account for potential confounds related to participants being tested at different European locations, we have re-run the analyses by only including the sites with the biggest numbers of participants (i.e., Germany and UK, N= 576).
For emotion recognition, the analysis of covariance showed main effects of psychopathology, sex, emotion and morph.There were also significant interactions between emotion and sex, and emotion, psychopathology and maltreatment.These results are in line with the ones reported in the main manuscript, where data from all 11 sites were analyzed.However, there were also some discrepancies, in that the emotion by maltreatment interaction was no longer significant (p=.072), and neither the emotion by sex by maltreatment (.055).For emotion learning, re-running the analysis with these two sites only, resulted in the same pattern of results.For avoidance errors, we found main effects of psychopathology, sex and block, twoway interactions between sex and maltreatment and sex and psychopathology, and the same sex by maltreatment by block, and sex by psychopathology by block (Table 2).

Notes:
Covariates evaluated in the model were SES = 0.125, mean-centered IQ=1.245, and pubertal status (1 = pre/early puberty; 2 = mid/late/post puberty).The adjustment for multiple comparisons was obtained using the Bonferroni correction.

FIGURE D. Interactions between emotion, maltreatment, and psychopathology in the female sample
Supplement 17: The Passive Avoidance Learning Task.Results by Block In the passive Avoidance Learning task, participants completed 10 blocks, the first one being for practice and therefore not included in the analysis.For simplification reasons, we have left these results out of the manuscript, reporting them here instead.The analysis indicated a significant three-way interaction between block, sex and maltreatment (F(8,5656)=4.38,p<.001, ηp 2 =.006) and between block, sex and psychopathology ((F(8,5656)=2.93,p=.003, ηp 2 =.004).For the former, pairwise comparisons indicated that maltreated females made more avoidance errors compared to maltreated males during blocks 3,6,7, and 8.For the latter, low

FIGURE A .
FIGURE A. Example of Permutation List with Study-ID, Subject-ID and the six possible task orders more avoidance errors than low psychopathology males during the same blocks (Figure E).

FIGURE
FIGURE E. Interactive Effects of Maltreatment, Psychopathology, and Sex on Avoidance Errors.Results by Block.

Notes:
The panels on the right show mean % of anger and disgust responses between the control and resilient groups.Errors bars show SEM.

TABLE 1 . Distribution of participants per Group and Sex by Site
Parents and guardians, and participants above the required age, gave written Supplement 3 : Participants' Demographic Characteristics and Psychopathology Subscales Scores.The Passive Avoidance Sample.
Note: f/m=female/male; Control = No Psychopathology, No Maltreatment; Resilient = Probable/Definite Maltreatment only; PT= High Psychopathology only; MT+PT=Probable/Definite Maltreatment, and High Psychopathology.Notes: Control = No Maltreatment, Low Psychopathology; Resilient= Probable/Definite Maltreatment, Low Psychopathology; High Psychopathology=No Maltreatment, High Psychopathology; Psychopathology + Maltreatment = High Psychopathology, Probable/Definite Maltreatment; SES = Socioeconomic Status (SES was computed based on parental income, education level and occupation); CD = Conduct Disorder (Diagnosis of CD was based on the Schedule for Affective Disorders and Schizophrenia for School-Age Children-Lifetime and Current versions (KSADS-L/C); PDS = Pubertal Developmental Status; CBCL = Child Behaviour Checklist; The CBCL scores for the Internalising scale were computed using the Anxiety/Depression, Withdrawal and Somatic Complaints subscales, whereas the scores for the Externalising scale were computed using the Rule-breaking and Aggressive Behaviour subscales; EMM = Estimated Marginal Mean; SEM = Standard Error of the Mean; Post-hoc tests are

TABLE 4 . Main and Interactive Effects of Maltreatment, Psychopathology and Sex on Emotion Recognition and Learning with age replacing pubertal category as a covariate Emotion Recognition
Emotion by age groups interaction in the Emotion Hexagon Task.
FIGURE B. FIGURE C. Emotion by age category interaction in the Emotion Hexagon Task.Supplement 16: Emotion Recognition Results -Female sample

TABLE 5 . Main and Interactive Effects of CM and Psychopathology on Emotion Recognition. Female sample only. Emotion Recognition
Supplement 18a: Confusability Matrix of Emotion Eesponses by Emotion Depicted by GroupsBayes Factor Analysis showed that the resilient group was different compared to the control group for responses of anger and disgust.Specifically, resilient youth showed more response bias towards anger when happiness was depicted, and more disgust responses when fear was depicted (FigureF).18c:Bayes Factors (BF10) for the Confusability Matrix Data.Comparisons Between Control vs. Resilient and PT vs. MT +PT Groups.
Notes:This table shows BF10 factors for the comparisons of emotion responses between the control and maltreated group (i.e., resilient) and between the high psychopathology versus high psychopathology and maltreatment group, for each of the six emotions depicted.