Childhood Trauma, Brain Structure and Emotion Recognition in Patients With Schizophrenia and Healthy Participants

Karolina I. Rokita; Laurena Holleran; Maria R. Dauvermann; David Mothersill; Jessica Holland; Laura Costello; Ruán Kane; Declan McKernan; Derek W. Morris; John P. Kelly; Aiden Corvin; Brian Hallahan; Colm McDonald; Gary Donohoe


Soc Cogn Affect Neurosci. 2020;15(12):1336-1350. 

In This Article



The current study was part of the 'Immune Response & Social Cognition in Schizophrenia' ('iRELATE') research project funded by the European Research Council examining the impact of the environment, genes and immune system on brain structure and function in SZ. Data were derived from 112 healthy participants (81 males and 31 females) and 46 patients (34 males and 12 females) either with a diagnosis of SZ (N = 35) or schizoaffective disorder (SZA) (N = 11). Patients with SZ and SZA were included in the same group as they exhibit a similar pattern of cognitive impairments and structural brain abnormalities (Amann et al., 2016; Hartman et al., 2019). Healthy participants were recruited through adverts placed in the Galway and Dublin areas, and patients were recruited from the local outpatient clinics and mental health services (e.g. day centres and day hospitals).

Inclusion criteria for patients were the following: (i) a diagnosis of SZ or SZA as confirmed by the Structured Clinical Interview for DSM-IV (Diagnostic and Statistical Manual of Mental Disorders, 4th Edition) (SCID) (American Psychiatric Association, 2000); (ii) being clinically stable at the time of assessment (i.e. medicated and being an outpatient) (iii) and no report of co-morbid psychiatric disorders. Healthy participants were only included in the study if they met the following criteria: (i) no report of mental or general health problems; (ii) no use of antipsychotic medication and (iii) having no first-degree relative with SZ or SZA. All participants were Caucasian and had to be aged between 18 and 65 years. Exclusion criteria for both groups were as follows: (i) a history of acquired brain injury causing loss of consciousness of >1 minute; (ii) substance abuse in the preceding 6 months; (iii) intellectual disability (i.e. IQ < 70); (iv) magnetic resonance imaging (MRI) contra-indicators (e.g. metal implants and claustrophobia) and (v) and a neurological disorder (e.g. epilepsy). Demographic and clinical characteristics of patients and healthy participants are presented in Table 1.


Following initial screenings conducted to identify eligible individuals to take part, participants were administered a comprehensive battery of measures over three study visits. The National University of Ireland Galway Research Ethics Committee, the Clinical Research Ethics Committee at University Hospital Galway and the Research Ethics Committee at Tallaght Hospital in Dublin reviewed and approved all study procedures. All participants provided written informed consent prior to the study visits.


Clinical Measures. The severity of symptomatology in patients was measured using the Positive and Negative Syndrome Scale [PANSS; (Kay et al., 1987)]. This 30-item semi-structured interview consists of three subscales: positive symptoms scale (7 items), negative symptoms scale (7 items) and general psychopathology scale (16 items). To reflect the 'absence' scores, we used the re-scaled Likert scale ranging from 0 (absent) to 6 (extreme). Ultimately, the total scores ranged from 0 to 42 for the positive and negative scales, from 0 to 96 for the general scale and from 0 to 138 for the total score.

Childhood Trauma Experiences. The Childhood Trauma Questionnaire [CTQ; (Bernstein et al., 2003)] was used assess traumatic experiences during childhood up to the age of 16. The CTQ is a 28-item self-report questionnaire investigating five different types of childhood trauma, including emotional abuse, physical abuse, sexual abuse, emotional neglect and physical neglect. Each scale consists of five items and a 5-point Likert scale is used for the responses, which range from 1 (never true) to 5 (very often true). However, in order to reflect the absence of trauma, we used the re-scaled Likert scale ranging from 0 (never true) to 4 (very often true). Ultimately, each subscale ranged from 0 to 20 and the total score ranged from 0 to 100, with higher scores indicating more traumatic experiences.

Emotion Recognition. Emotion recognition was assessed with the Emotion Recognition Task (ERT)—short version, which is a computerised task, implemented in the Cambridge Neuropsychological Test Automated Battery (Robbins et al., 1994) evaluating the ability to recognise six basic emotions: happiness, sadness, disgust, anger, surprise and fear. The task includes five practice trials and 48 assessed trials of male and female facial expressions of eight different intensities. Participants are instructed to correctly identify emotions presented on the screen by choosing one out of six names of emotional expressions. The overall number of correct responses (for all emotions) is the main outcome of this task that we assess in the current study.

Neuropsychological Assessment. Three subtests from the Wec-hsler Adult Intelligence Scale—Third Edition [WAIS-III-R; (Wechsler, 1997)] including Vocabulary, Digit Symbol and Block Design were administered to estimate Full Scale Intelligence Quotient. The shortened versions of the WAIS-III are frequently used to reduce administration time and were previously found to highly correlate (r > 0.9) with the full version consisting of 11 subtests (Axelrod, 2002; Wymer et al., 2003).

MRI Acquisition. Imaging for all participants was performed on a 3T Philips Achieva MR scanner (Philips Medical Systems, Best, the Netherlands), located in the Centre for Advanced Medical Imaging at St. James's Hospital, Dublin, Ireland. Three-dimensional T1-weighted structural scans (FFE (Fast Field Echo) pulse sequence, TR (Repetition Time)/TE (Echo Time) = 8.5/3.9 ms, FOV (Field-of-view) = 256 × 256 × 160 mm3, a spatial resolution of 1 mm3, TI (Inversion Time) = 1060 ms, flip angle = 8°, SENSE (Sensitivity Encoding) factor = 1.5 and acquisition time = 7 min 30 s) of the whole brain were acquired for each participant. Foam padding was used to stabilise the head, and headphones were provided to minimise the scanner noise.

MRI Analysis. The FreeSurfer image analysis software suite [Version 6.0.0; (Fischl et al., 2002)] was used for processing of the T1-weighted images, including cortical surface reconstruction, cortical volume parcellation and subcortical volume segmentation. The Desikan-Killiany atlas was used to label 34 cortical structures per hemisphere (Desikan et al., 2006). Since the study focuses on three a priori defined regions, including the amygdala, hippocampus and ACC, we extracted bilateral volumes for these specific regions. Total ACC volume for each hemisphere was calculated by adding the volume areas of the caudal and rostral sub-regions to avoid multiple testing issues. Intracranial volumes (ICV) were also generated by FreeSurfer and chosen as a covariate in the analyses to account for head size variability among participants. All T1 images were visually inspected by two independent researchers and images with clear motion artefacts (N = 3; patients) or incidental findings (N = 2; HCs (healthy controls)) were not included in the analysis. The exclusion of images with motion artefacts was also supported by the assessment of outliers using a standardised protocol provided by the ENIGMA (Enhancing Neuro Imaging Genetics through Meta-Analysis) consortium ( as the images with significant motion artefacts returned a high number of errors in reconstruction (varying between 20 and 30).

Statistical Analysis

Statistical analyses were conducted with Statistical Package for Social Sciences (SPSS) Version 25 (SPSS IBM Corp., 2017) and the mediation analyses were performed with PROCESS macro Version 3.3 (Hayes, 2013). Log transformations were applied to reduce skewness of the childhood trauma variables that were not normally distributed. Sexual abuse and physical abuse were converted into dichotomous variables (0 = absent, 1 = present) since the majority of the sample (i.e. 86% and 80%, respectively) did not report these experiences. Where specified, years of education was controlled in the analyses due to significant group differences on this variable and its possible impact on brain structure and cognition (Cox et al., 2016; Guerra-Carrillo et al., 2017).

Between-group Differences on Childhood Trauma, ROI Volumes and Emotion Recognition. To compare both groups on childhood trauma and emotion recognition measures, we first conducted independent sample t-tests and chi-square (χ 2) tests, where appropriate. ANCOVAs (Analyses of covariance) were used to assess between-group differences on the specified brain regions with age, years of education and ICV as covariates in these analyses.

Associations Between Childhood Trauma, ROI Volumes and Emotion Recognition in all Participants. In the next step, partial correlations were conducted between childhood trauma (i.e. physical neglect and total childhood trauma), emotion recognition task and ROI volumes for all participants in order to assess the relationships between these measures and select the variables for the mediation analyses. Additional correlations were subsequently conducted for both groups separately (healthy participants and patients). Age, years of education and ICV were controlled for.

Moderated Mediation Analyses. To explore whether regional brain volumes in the amygdala, hippocampus and ACC mediate the association between childhood trauma and performance on the ERT task, we conducted moderated mediation analyses using Model 59 from the PROCESS macro (Hayes, 2013), which allowed the inclusion of a mediator and a moderator in the same model. This particular model was used as it considers moderation effects on all paths (i.e. direct and indirect), compared to other more common models. We examined both direct and indirect (mediating) effects between childhood trauma and the ERT task, with the ROI volumes as mediators in these associations and a group type as a moderator, controlling for age, years of education and ICV. The reliability of the associated indirect effect was estimated using 5000 bootstrapping iterations to obtain bias-corrected and accelerated bootstrap 95% confidence intervals (CIs). When a 95% bootstrapped CI does not include zero, it indicates the parameter is statistically significant.