Child maltreatment is a significant global public health issue that encompasses physical, sexual, and psychological abuse, as well as emotional and physical neglect of children under 18 years of age (Alkema et al., 2024; Tolliver et al., 2023; World Health Organization, 2022). The World Health Organization reports that three out of four children aged 2 to 4 experience regular physical punishment or psychological violence from their parents or caregivers (World Health Organization, 2022). Alarmingly, around 20 % of girls and 10 % of boys have suffered sexual abuse (World Health Organization, 2022). The repercussions of child maltreatment are profound, and could have lasting effects on individuals' social, emotional, and physical well-being throughout their lifetime (Olson et al., 2024). Furthermore, child maltreatment incurs a high lifetime cost per victim, contributing significantly to economic and healthcare burdens at the population level (Peterson et al., 2018).

Extensive research has identified child maltreatment as a critical risk factor for a spectrum of internalising problems, including depression, anxiety, and post-traumatic stress (Barboza and Dominguez, 2017; Gardner et al., 2019). A meta-analysis of 96 observational studies indicated that various forms of child maltreatment were consistently linked to a 1.65- to 2.48-fold increase in the likelihood of developing depressive disorders (Gardner et al., 2019). This association may be explained by the impaired capacity for emotional regulation resulting from maltreatment and alterations in the hypothalamic-pituitary-adrenal axis that regulates stress response (Chen et al., 2024). Additionally, exposure to child maltreatment heightened the risk of frequent and severe externalising behaviours during childhood and adolescence, such as aggression, violence, conduct problems, and risky sexual behaviours (Bauer et al., 2021; Vachon et al., 2015). The presence of these internalising and externalising behavioural disorders can impede educational and occupational achievement, perpetuating cycles of poverty and social isolation (Zoellner and Maercker, 2006). Promoting children's psychological and behavioural health is essential for healthy development and health equity across the lifespan.

However, the relationship between child maltreatment and child behavioural problems may be more complex. Emerging research suggests that high levels or changes in child behavioural problems over time may result in future maltreatment exposure (Font and Berger, 2015). For instance, children exhibiting frequent behavioural issues (e.g., hyperactive or delinquent behaviours) may provoke caregivers to implement harsher disciplinary strategies, potentially constituting maltreatment through practices such as spanking or physical aggression (Olson et al., 2024). Similarly, caregivers may resort to yelling or harsh criticism to address persistent worrying or mood instability. A cohort study employing the cross-lagged model found a robust association between child maltreatment and subsequent child behavioural problems, while increases in externalising behaviours also predicted later maltreatment exposure (Olson et al., 2024). This evidence suggests a potential feedback loop between child maltreatment and behavioural problems.

Despite these insights, prior research linking child maltreatment to mental/behavioural problems has largely relied on observational data (e.g., cross-sectional, cohort, or case-control studies) characterized by considerable heterogeneity. Such methodology limitations, often stemming from ethical and practical constraints, complicate causal inferences, as potential confounding by genetic, socioeconomic and environmental factors cannot be eliminated. It is conceivable that maltreatment and behavioural problems are jointly determined by a similar set of underlying factors (e.g., deprivation and parental functioning), rather than one affecting the other (as visualized in Fig. 1a). In addition, the directionality of causal effects between child maltreatment and behavioural problems has not been ascertained, with both uni-directional and bi-directional hypotheses receiving some degree of support in observational studies. However, even using a longitudinal design with temporal precedence (e.g., behavioural problems measured after child maltreatment assessment), the outcome status collected during follow-up could be a continuation of pre-existing traits that were not adequately measured at baseline. Therefore, the detected association might misleadingly suggest causation from exposure to outcome when it actually reflects a reverse effect from an earlier, unobserved state of the outcome variable to the baseline exposure status (Fig. 1a). Statistical models (e.g., cross-lagged panel models mentioned above) attempted to disentangle this with repeatedly measured data but were still vulnerable to confounding and distortion of directionality by uncaptured traits prior to data collection.

Mendelian randomization (MR), which leverages genetic variants as instrumental variables to mitigate confounding and reverse causality, offers a robust solution to enhance the causal inference capabilities of observational studies, without animal studies or randomized controlled trials (Lawlor et al., 2008). Since genetic alleles are randomly assigned at conception (Mendel's Law of Segregation), they are largely independent of socio-environmental confounders that develop after birth (Fig. 1b). In addition, the inherent temporal precedence assures the causal direction between genetic instruments (i.e., proxies of exposure variable) and the outcome variable. Therefore, under several assumptions, this MR approach can effectively address confounding bias and reverse causality because genetic information cannot be influenced by confounders and the development of outcome events.

Although maltreatment is an environmental exposure, genetic studies show that heritable traits in the child can influence its likelihood through gene–environment correlation—a process by which genetic predispositions shape the environments individuals experience (Micalizzi et al., 2017). Twin and family studies suggest genetic factors may account for up to 60 % of the variation in reported maltreatment (Pezzoli et al., 2019), and a recent GWAS meta-analysis (N = 185,414) identified 14 loci associated with child maltreatment experience (Warrier et al., 2021). Importantly, these findings do not imply biological determinism or victim culpability; rather, they highlight how genetically influenced characteristics may unintentionally evoke adverse responses in caregiving environments.

Using the genetic predispositions of child maltreatment as instrumental variables, two previous MR studies investigated the effect of child maltreatment on neurodevelopmental disorders or severe externalising spectrum disorders, including attention deficit hyperactivity disorder (ADHD), autism, schizophrenia, conduct disorder, and substance use disorder (Warrier et al., 2021; Konzok et al., 2024). However, these studies mainly focused on clinical disorders instead of commonly observed child behavioural problems (e.g., aggressive behaviour), thus had relatively limited generalisability and implications for daily parenting practices and child protection. In addition, associations between child maltreatment and internalising behavioural problems remain under researched. To address these knowledge gaps, our study aimed to use a bi-directional MR design with up-to-date genetic databases to elucidate the causal relationship between maltreatment experience and a more general spectrum of internalising and externalising child behavioural problems, thereby providing nuanced insights for developing preventive intervention strategies to mitigate adverse sequelae of childhood maltreatment and behavioural problems.