T91. CHILDHOOD MALTREATMENT AND MENTAL HEALTH PROBLEMS: A GENETICALLY INFORMED APPROACH FROM AN ADMIXED POPULATION

Childhood maltreatment is a global public health concern due to its high prevalence and long-term mental health effects. Heritability estimates (6-62%) suggest shared genetic and environmental influences. Recent GWAS support gene–environment correlations, although the mechanisms underlying these correlations are uncertain. However, most studies focus on European ancestry. In LMICs like Brazil, levels of child maltreatment are higher than in many high-income countries, and socioeconomic adversity, a key environmental risk factor for child maltreatment, is widespread. This study examines whether polygenic scores (PGS) for susceptibility to childhood maltreatment predict mother-reported child maltreatment effects in a Brazilian birth cohort, considering ancestry effects, and links with child and maternal mental health. We used data from 3,053 participants in the 2004 Pelotas (Brazil) Birth Cohort, a prospective, population-based study in southern Brazil. Childhood maltreatment was assessed at ages 6 and 11 years via the Parent-Child Conflict Tactics Scales questionnaire; mean scores across timepoints were calculated for total, psychological and physical abuse, and corporal punishment. Saliva‐derived DNA from 6-year-olds (∼3.7K) was genotyped and imputed against 1000 Genomes, yielding ∼10 million high-quality SNPs. We used PRS-CS, a Bayesian genome-wide approach for PGS construction, based on Warrier, 2021 (N=185,414; mainly European ancestry). Child and maternal mental health were assessed at the ages of 6 and 11, respectively, using the Strengths and Difficulties Questionnaire and Edinburgh Postnatal Depression Scale (cutoff ≥10). Linear and Poisson models were adjusted for sex and the first 10 ancestry principal components (PCs). Analyses were by self-reported skin colour (White and Mixed-race Brazilians). Most participants were male (50.9%) and White (67.6%). Linear regression suggests that a 1SD increase in PGS was linked to a 0.24-point higher in total maltreatment scores (95%CI: 0.10;0.37, ΔR²= 0.4%). However, this association attenuated after adjustment (β: 0.15, 95%CI: -0.00;0.29, ΔR²= 0.1%), suggesting that sex and ancestry PCs may account for some of the initial effect. In skin colour-stratified models, a 1SD increase in PGS was associated with a 0.12-point higher corporal punishment score among White Brazilians (β: 0.12, 95%CI: 0.02; 0.22, ΔR²= 0.3%), but not among mixed‐race participants (p-value interaction: 0.083); no associations were found for total child maltreatment or psychological and physical abuse in skin colour-stratified models. No associations were found for child or maternal mental health in overall or skin colour-stratified models. GWAS in European populations have identified genetic variants associated with child maltreatment, but this is the first study linking genetic liability with observed child maltreatment and intergenerational mental health in an admixed population. An association between a PGS for child maltreatment and a score of child maltreatment was observed, but attenuated with adjustment for sex and ancestry PCs. Furthermore, we found some evidence of stronger associations between the PGS and some measures of child maltreatment in White Brazilians compared with mixed-race Brazilians. Our findings also highlight the complexity of polygenic scores in relation to complex social processes, such as child maltreatment, and the challenges of applying these scores in diverse populations.