Childhood maltreatment and social cognition deficits across the psychosis spectrum: Evidence from the PREGAP study

Background: Childhood maltreatment (CM) is common in early psychosis and may contribute to social‐cognitive difficulties beyond illness-related processes. We examined whether CM explains unique variance in social cognition across psychosis stages, whether specific CM subtypes map onto distinct social‐cognitive domains, and whether associations vary by clinical stage. Methods: We conducted a cross-sectional study including 107 participants: 39 with first-episode psychosis (FEP), 35 at clinical high risk for psychosis (CHR-P), and 33 healthy controls (HC). CM was assessed using the Childhood Trauma Questionnaire (CTQ). Social cognition was evaluated across multiple domains, including facial emotion recognition (Penn Emotion Recognition Test [PERE]), theory of mind (Movie for the Assessment of Social Cognition [MASC] errors and the Hinting Task), and social perception (Faux Pas Test). Self-reported social functioning was measured with the GEOPTE scale. Adjusted correlation analyses and linear regression models were performed, followed by comparison of 3 nested models to assess incremental explained variance: model A included CM only; model B included clinical group only; and model C included both CM and clinical group. Childhood maltreatment–by–group interaction terms were examined to explore stage-dependent effects. All analyses were exploratory, with 2-sided p < .05 considered statistically significant and no correction for multiple comparisons applied. Results: CM showed domain-specific links with social cognition. Physical abuse related to more hypomentalization errors (MASC) and poorer disgust recognition (PERE); sexual abuse also predicted poorer disgust recognition. Emotional neglect was associated with lower Hinting performance and reduced recognition of neutral expressions; physical neglect related to reduced neutral and increased joy recognition. Higher total CTQ scores were tied to poorer neutral and better joy recognition. Model comparisons indicated that adding CM improved prediction beyond clinical status, particularly for neutral recognition associated with physical neglect (ΔR²; 0.18; p < 0.01), emotional neglect (ΔR²; 0.15; p < 0.01), and CTQ total (ΔR²; 0.10; p < 0.01). CM-by-group interactions suggested stronger links between (physical/sexual) abuse and impaired disgust recognition in HC than in CHR-P/FEP, and a specific emotional-neglect–by–FEP effect on neutral recognition; these were underpowered and interpreted cautiously. GEOPTE (social functioning) was higher with emotional abuse. Conclusions: CM contributes independently and additively to social-cognitive variation in early psychosis, with subtype–domain specificity most evident for theory of mind and facial emotion recognition. Findings support trauma-informed assessment within early intervention services and motivate longitudinal, mechanistic, and interventional work to test whether targeting CM-related mechanisms improves social cognition and functional outcomes.