2026 ECNP Congress

PS04-4045

Investigating childhood trauma and interacting risk and resilience factors to predict depression, psychosis, and high-risk individuals

Margarita David — T. Haidl, S. Mena Ortega, N. Penzel, R. Upthegrove, E. Meisenzahl, R.K.R. Salokangas, J. Hietala, A. Bertolino, R. Lencer, C. Pantelis, S. Wood, G. Romer, D. Dwyer, J. Kambeitz, L. Kambeitz-Ilankovic, F. Schultze Lutter, S. Ruhrmann, D. Popovic, P. Lalousis, N. Koutsouleris, G.R. Jacobs · Ludwig-Maximilian-University, Department of Psychiatry and Psychotherapy, Munich, Germany

2026 ECNP Congress

Background

Childhood trauma (CT) is common, and a transdiagnostic risk factor for mental health disorders. However, exposure to early-life adversity does not uniformly lead to psychopathology, as individual differences in post-traumatic outcomes are linked to multisystem resilience mechanisms, including individual, social, family and cultural factors. Previous findings suggest that CT data alone are insufficient to distinguish between diagnostic groups such as recent-onset depression (ROD), recent-onset psychosis (ROP) and individuals at clinical high-risk for psychosis (CHR). Our study aimed to investigate risk and resilience factors interacting with CT experiences to determine if it is possible to predict which individuals develop psychosis versus depression following exposure to CT.

Methods

Our sample included 1083 individuals (ages 15 to 40) with ROD, ROP, CHR, and healthy controls (HC) from the multisite Personalised Prognostic Tools for Early Psychosis Management (PRONIA) study. We applied support vector machine learning classification models (L2-regularized logistic regression), within a nested 10-fold cross-validation structure. This framework was implemented to predict diagnostic and HC groups using NeuroMiner software. Algorithm performance was measured using balanced accuracy. Feature preprocessing included imputing missing data and regressing out effects of age and sex. Features used to predict groups included data from self-report measures such as the Childhood Trauma Questionnaire, Bullying Scale for Adults, Resilience Scale for Adults, the NEO Five-Factor Inventory for personality, Coping Inventory for Stressful Situations, Multidimensional Scale of Perceived Social Support and the Level of Expressed Emotion Questionnaire.

Results

Models were able to differentiate clinical groups from HC with high levels of accuracy (BAC= 85% for HC vs ROD; BAC= 88.1% for HC vs CHR; BAC= 81.2% for HC vs ROP). Models were also able to differentiate diagnostic groups. The most significant predictive features differentiating ROP and ROD (BAC=65.1%) included elevated overall CT, physical neglect, bullying stress, and sexual harassment predictive of ROP. Additionally, higher task-oriented coping strategies and perception of self resilience contributed to ROP prediction, while higher emotional-oriented coping strategies and neuroticism contributed to ROD prediction. Comparing ROP and CHR (BAC=65.2%), a higher perception of self resilience was significantly associated with ROP, while higher emotional-oriented coping strategies and neuroticism were significant features of CHR. In comparison, model performance predicting ROD vs CHR was lower (BAC=53.5%) and there were no significant predictive features.

Conclusions

By integrating resilience factors, such as coping strategies and personality, together with CT, we demonstrated that a multisystemic approach allows for a more accurate classification and understanding of different psychiatric trajectories, as compared to utilizing CT data alone. This highlights that the interaction between CT and individual psychological factors and resources is crucial for clinical differentiation. Specifically, we found that ROP differed from ROD and CHR in terms of both CT and resilience factors, while ROD and CHR were similar. These findings may provide and enhance detection of roadmaps for personalized, targeted prevention and intervention for psychopathology informed by individual CT and resilience profiles.

References

No references