Search

Almost one in five UK adolescents are thought to self-harm. In the first birth cohort study of adolescent self-harm in the UK, the prevalence was 18.8% amongst 16-17-year-olds (Kidger et al., 2012). In the same study, 5.7% of participants self-harmed with suicidal intentions (suicidal self-harm – SSH). Self-harm is sometimes called deliberate self-harm and includes suicidal self-harm (SSH), but not exclusively. Self-harm as a term itself is more common in Europe and Australia, whilst non-suicidal self-injury and suicide attempts are terms used more often in North America (Gillies et al., 2018). Most importantly, some studies reporting on self-harm do not distinguish between SSH and non-suicidal self-harm (NSSH). Recent evidence suggests that SSH and NSSH share similar aetiologies as well as strong associations with depressive symptoms and thus might be part of a spectrum rather than two discrete entities (Lim et al., 2021). NSSH is also associated with later SSH (Hawton et al., 2003). SSH can lead to death by suicide, which is amongst the most common causes of young people’s deaths worldwide, accounting for over 150,000 15–29-year-olds’ deaths in 2019 (WHO, 2021). Prevention of self-harm should be informed by knowledge about risk factors (e.g., socio-demographic), characteristics (i.e., motivation for self-harm and help-seeking behaviours) and aetiological genetic and environmental processes. For the present study, we focus on a broad self-harm phenotype, which includes SSH and NSSH. Some socio-demographic risk factors for self-harm include parental socioeconomic status (e.g., Engstörm et al., 2011; Page et al., 2014) and being a sexual minority (i.e., not heterosexual, e.g., Miranda-Mendizábal et al., 2017). The nature and the rates of self-harm also seem to differ between ethnic groups in the UK, with lowest rates amongst Asian men and highest rates amongst Black women (Bhui et al., 2007; Al-Sharifi et al., 2015). One systematic review (Plener et al., 2015) and one meta-analysis (Gillies et al., 2018) concluded female gender to be associated with more self-harm, however some other studies found no evidence of gender differences (e.g., a study of nearly 1,200 adolescents recruited in the community; Kirchner et al., 2011), hence findings regarding the association between gender and self-harm during development remain mixed. Furthermore, a prevalent environmental adversity in adolescence – bullying victimisation has been shown to predict higher rates of self-harm (Fisher et al., 2012). Characteristics of self-harm include reasons, or motivations for self-harm as well as help-seeking behaviours. A recent meta-analysis reviewed both, finding the most common motivation to be ‘to obtain relief from feelings or thoughts’, and the second most common ‘to punish oneself’ (Gillies et al., 2018). With regards to help-seeking, majority of adolescents do not seek help for their self-harming (e.g., Rowe et al., 2014; Stallard et al., 2013). In the meta-analysis, the most common help-seeking was from friends, then family. With regards to medical help-seeking, 8% of adolescents from 17 studies sought it in general. Nine percent from 8 studies sought hospital treatment for self-harm-related injuries, and 18%, also from 8 studies, sought help from a mental health professional (Gillies et al., 2018). One suggestion in the light of high prevalence of self-harm and low help-seeking in adolescence is that risk factors, such as bullying, should gain more awareness and be tackled in the community setting (Stallard et al., 2013). It is not uncommon for self-harm to have an early age of initiation. For example, in a community-sample study of nearly 4,000 12- to 16-year-olds, 15% of them had engaged in self-harm. In sub-sample analyses, thirteen percent of Year 8 pupils (12- to 13-year-olds) and 13% of Year 9 pupils (13- to 14-year-olds) self-harmed (Stallard et al., 2013). A meta-analysis of studies across a 25-year period found the mean age of initiation of self-harm to be 13 years (Gillies et al., 2018). Importantly, early initiation may be associated with greater severity of self-harm (Ammerman et al., 2018). Nonetheless, there is a lack of research on risk factors and characteristics specifically in relation to earlier initiation rather than young people’s self-harm generally, as well as a lack of genetically informed research relating the age of initiation to the aetiology of self-harm. Genetically informed research on self-harm can help to highlight environmental risk factors, while controlling for genetic risk factors. Furthermore, it is important to understand nuances of genetic influence on self-harm, e.g., knowing whether genetic influence and environmental influence are independent will impact how genetic risk estimates, such as polygenic scores (PGS), are interpreted. Previous research on mixed-age (adolescent and adult) twin samples evidenced genetic influence on thoughts of self-harm (Althoff et al., 2012) and on SSH (Voracek and Loibl, 2007). Lim et al. (2021) found that additive genetic effects had influence on NSSH (55%) and on SSH (50%) in young adults in the Twins Early Development Study (TEDS). Additionally, this study investigated the aetiological overlap (i.e., bivariate twin models) between NSSH and SSH and compared the relationship of these two variables to other mental health measures. They found strong genetic and non-shared environmental correlations between the two, as well as lack of significant differences in relationships of SSH and NSSH to other mental health phenotypes. The strongest genetic correlations of SSH and NSSH were with depressive symptoms and with insomnia. To date, there have been no studies investigating the aetiology of self-harm over development. However, many young people who experience self-harm also have depressive symptoms (Tuisku et al., 2006) – a relationship that may be partly explained by shared genetic aetiology (Lim et al., 2021). Studies looking at the developmental aetiology of depression have shown that genetic influences in paediatric depression increase over time from childhood to adolescence, in designs where two age groups are compared (Eley and Stevenson, 1999; Thapar and McGuffin, 1994) as well as in longitudinal designs (Silberg et al., 1999). Furthermore, it has been suggested that new genetic effects on depression may emerge during adolescence (genetic innovation; Nivard et al., 2015). In adulthood, genetic influences on depression appear to become stable (Trovik et al., 2019; Nivard et al., 2015). Considering the overlap in aetiology of depression and self-harm, it could be that, similarly to depression, genetic influence on self-harm increases across adolescence. One explanation put forward for the increase in heritability in psychopathology across adolescent development is the role of gene-environment interplay (Thapar and Rice, 2006) as environmental influences which correlate or interact with genetic influences are in some cases captured by heritability. Furthermore, as adolescents get older, they experience more environments and thus are more likely to be exposed to environmental risks, which can inflate the estimates of genetic influence. Conclusions of a recent meta-analysis on self-harm support this theory, as the authors suggest that older adolescents may be exposed to a greater number of self-harm risk factors (Lim et al., 2019). One such environmental risk factor could be bullying victimisation (Fisher et al., 2012). We plan to first explore the socio-demographic and environmental factors associated with self-harm and age of initiation of self-harm in the TEDS sample. We will also explore the effect of age of initiation on self-harm motivation and help-seeking. Then, we plan to apply the classical twin model to TEDS data to investigate the relative genetic and environmental influences on self-harm across development. In this pre-registration, a detailed outline of this planned study is provided. Aims The present study aims to investigate: 1) the prevalence of self-harm at different developmental stages and its incidence in young adulthood, 2) the relationship between socio-demographic, environmental factors, and self-harm as well as its age of initiation, 3) the relationships between the age of initiation, motivations for self-harm, and help-seeking behaviour, 4) developmental changes in aetiological influences on self-harm.

Cardiometabolic diseases (CMDs) are among the leading causes of death worldwide (WHO, 2022). CMDs like diabetes, high blood pressure and hyperlipidaemia have seen a dramatic increase during the past decade. In Sri Lanka, CMDs have been increasing and this could be influenced by changes in activity levels and dietary patterns (Ediriweera et al., 2018; De Silva et al., 2019). The World Health Organization estimates that more than 1 billion people across the world have obesity, including 650 million adults and 379 million children (WHO, 2022). Obesity is known to significantly elevate a person’s risk of developing CMDs, independent of other cardiometabolic risk (CMR) factors. An island-wide study in Sri Lanka showed an alarmingly high prevalence of obesity, particularly abdominal obesity, among Sri Lankan adults. The study also predicted an increase in obesity over time, and called for immediate action to reverse this trend through effective public health interventions (Katulanda et al., 2010). Abdominal obesity, measured by waist circumference, is particularly known to increase the risk of cardiometabolic events. This relationship appears to be independent of general obesity measured by body mass index (BMI), which is the most common anthropometric measure used to investigate obesity-morbidity associations (Thu Tran et al., 2018). However, differences exist in which anthropometric measures best predict obesity related CMDs. Accumulating evidence also suggests that both environmental and genetic factors are important in explaining the development of obesity and associated CMDs (Nelson et al., 2000; Heianza and Qi, 2019). However, the relative impact of genetic and environmental factors on the association between obesity indices and CMR factors have not been yet explored in a Sri Lankan population. We aim to examine the genetic and environmental contributions to obesity indices (waist circumference and BMI) and CMR factors and their covariation in a Sri-Lankan population twin sample, using multivariate twin model-fitting analysis. More specifically, we want to establish whether the phenotypic overlap between the variables is due to correlated latent genetic and/or environmental factors. Moreover, we will be able to establish if the environmental correlations are due to family or individual specific environmental effects.

Heritability of aspects of the externalising spectrum such as aggressive antisocial behaviour, behavioural disinhibition and risk taking range from moderate to high (Dhanani et al., 2018) suggesting a role for genetics in the development of externalising symptoms. Linnér et al. (2021) recently published a multivariate analysis of 1.5 million people which investigated the genetic links to externalising-related traits. Using the results from their genomic structural equation model, researchers created polygenic risk scores (PRS) for their externalising behaviour factor. The study reported associations between their PRS for externalising and a broad range of externalising related phenotypes including substance use initiation, antisocial personality disorder, ADHD diagnosis and rule breaking. Associations found in PRS analyses in unrelated individuals can be confounded by several factors, such as population stratification, assortative mating (Selzam et al., 2019) and indirect genetic effects of parental genotype on child phenotype (genetic nurture, Kong et al., 2018). Using sibling pairs in genomic analyses can allow for the estimation of genetic effects free from such confounds and biases, as the siblings are matched for parental and family environment. DZ twin pairs are also matched for prenatal environment and time-variant factors. By comparing within-family and between-family effects, we can investigate the extent to which passive genotype-environment correlation (rGE, I.e. genetic nurture, population stratification, and assortative mating) inflates the association between genomic variation and externalising phenotype. Little is known about PRS prediction of externalising symptoms in different developmental periods. Costantini et al., (2021) reported an association between higher polygenic scores for neuroticism, a personality trait associated with externalising behaviours, and increasing levels of externalising problems from early to late childhood (4-11 years). Our work will extend on this by including later developmental periods, when children become more autonomous and so parenting may have less of an impact. We also will be using a polygenic score specifically developed for externalising. No study to-date has investigated how within- and between-family differences in PRS prediction of externalising behaviour may change across development. Twin analyses have previously shown that common environmental effects (C – those shared between twins), are larger in childhood. Conversely, heritability (h2) estimates get larger as children age. Arguably, genetic nurture effects may be captured in estimates of the shared environment. If so, genetic nurture effects would follow the same developmental trends as C and reduce in magnitude across development. We will test whether this is the case in the TEDS sample, calculating within/between-family PGS estimates at multiple time points across development. Following the twin literature, we expect to see larger genetic nurture effects (captured by the between-family genomic analyses) in childhood than in adulthood, with the expectation that direct genetic effects will be smaller in childhood and increase with age. We will complement our genomic analysis with twin analyses using the same data.

There is convincing evidence from a meta-analysis of large-scale register and other studies that the rate of ADHD diagnoses is higher among children who are young-in-class, compared to their older peers. As ADHD diagnosis is made relative to behaviours expected of children of the same age (i.e. children have to show inattentive and hyperactive-impulsive behaviours significantly more than other children of the same age to obtain the diagnosis), it is possible that young-in-class children are rated as having more ADHD symptoms than they would if correctly compared to those accurately matched on age. If so, a proportion of the young-in-class children with ADHD may actually have less severe ADHD or a lower genetic loading for ADHD, if they have crossed the threshold to diagnosis partly due to their relative immaturity. Apart from the baseline evidence for the relative age effect in ADHD, this issue remains largely unexplored. Our study will test if children who were young-in-class are rated, on average, higher on ADHD symptoms than their older peers. If we observe a relative age effect on ADHD symptoms, we will proceed to test further possible reasons for this observation

While many children with ADHD continue their inattentive or hyperactive behaviors into adulthood, the symptoms as well as comorbidities change during developmental course. Comorbidities of ADHD, for example with conduct disorders and neurodevelopment disorders, are more frequent during childhood, while emotional disorders become more common during adulthood (Franke et al., 2018; Sadek, 2018). Therefore, the nature of the comorbidity between emotional problems and ADHD at different developmental stages is worth exploring. Differential genetic effects between the two domains of ADHD symptoms have been found. As a result, comorbidities of other disorders with inattention and impulsivity may be due to different mechanism (Ghirardi et al., 2019; Plourde et al., 2017; Michelini et al., 2015). Splitting the environmental and genetic effects of inattention and impulsivity when exploring the association between ADHD and emotional problems is important. Previous studies have investigated the influence of common genetic influences in the association between ADHD and emotional problems (Brooker et al., 2020; Chen et al., 2016). However, the mechanism of the increasing association between emotional problems and ADHD with age is not clear. We will use data from the Twins Early Development Study (TEDS). Using these data, we will be able to apply structural equation models and investigate the association between ADHD and emotional problems over time. This study will explore the co-occurrence of ADHD and emotional symptoms during childhood, adolescent and adulthood as well as the causal effect between the two traits.

Depression and anxiety, defined as psychological distress when manifesting as a comorbidity, are common mental health conditions (Arvidsdotter et al., 2016; Choi et al., 2020). Psychological distress can burden the affected individual, in addition to their family and wider society (Othman et al., 2019). This is demonstrated by the inverse relationship which exists between psychological distress and educational attainment, full-time employment, and quality of life (Brenes, 2007). The global increase in psychological distress prevalence disproportionately affects low- and middle-income countries (LMICs; Javed et al., 2021). This burden is largely attributed to the existing treatment gap (of up to 93%) in LMICs, which is caused by inadequate treatment availability and access (Mascayano et al., 2015). For example, Sri Lanka is a LMIC that has experienced natural disasters, civil conflict and political unrest in recent decades (Kathriarachchi et al., 2019). In the context of these, a high proportion (almost 20%) of children aged 13 to 18 years in Sri Lanka are reported to have experienced a psychological disorder (Perera, 2009). Similarly, adults who were exposed to the civil war reported increasing levels of psychological distress for up to 5 years post-conflict (Jayasuriya et al., 2016). Despite the introduction of a decentralised public health system, a significant mental health treatment gap persists (van den Broek et al., 2021). Unique barriers to accessing mental healthcare in Sri Lanka include: a failure to recognise symptoms or available services, social stigma and/or a negative perception to treatment efficacy (Fernando et al., 2017). The existing mental health treatment gap necessitates a review of alternative risk and protective factors for psychological distress that can be targeted for preventive intervention. Socioeconomic status (SES) is strongly associated with psychological distress (Maselko, 2017). Two underlying mechanisms facilitate this relationship. Social causation suggests that mental disorders arise due to the adverse social and financial stress of lower SES (Lund et al., 2010; Wickham et al., 2017). Conversely, social selection finds individuals with mental disorders to have a greater likelihood of drifting into a low SES (Chiu et al., 2017; Saraceno et al., 2005). A recent longitudinal study of SES and depression in China (a LMIC) found these mechanisms to work concurrently (Jin et al., 2020). Modifiable protective factors that mediate the link between low SES and psychological distress can be targeted for preventive interventions to help break the debilitating cycle of low SES and psychological distress (Jin et al., 2020). One possible proactive intervention is dietary diversity. A growing body of research in LMICs links dietary diversity with SES and mental health outcomes, whereby low SES is associated with reduced dietary diversity (Mozaffari et al., 2021; Onyango, 2003) while transitioning from a homogeneous diet to one with a variety of food groups is protective against psychological distress (Poorrezaeian et al., 2015, 2017; Yin et al., 2019). Thus, low dietary diversity may mediate the association between low SES and psychological distress, however, this has not been previously investigated. Although we suggest dietary diversity as a mediator of the relationship between low SES and psychological distress, we recognise the possibility of alternative relationships. For example, psychological distress may be associated with compensatory hyperphagia (carbohydrate craving which is associated with decreased fruit and vegetable consumption; Razolli et al., 2017; Matsungo & Chopera, 2020), and thus mediate the association between low SES and reduced dietary diversity. An alternative mechanism is that SES, low dietary diversity and psychological distress are jointly influenced by a common factor. For example, the civil conflict and consequent economic hardship in Sri Lanka may increase poverty, serve as a source of stress with consequent psychological distress, disrupt food supply which would make a varied diet expensive and inaccessible (Conklin et al., 2016; Siche, 2020). The evidence base for these possibilities is relatively slim, hence the decision to investigate dietary diversity as a mediator of low SES and psychological distress. However, the direction of causal effects may be better tested in future studies. Furthermore, considering that genetic and environmental factors influence individual differences in depressive and anxiety symptoms (Nas et al., 2021), and SES (e.g., through active gene-environment correlations, Scarr & McCartney, 1983; Trzaskowski et al., 2014); it is possible that their covariance and the proposed mediation by dietary diversity are confounded by correlated genetic and environmental influences. In which case, interventions targeting dietary diversity to reduce the risk for psychological distress among those from low socioeconomic backgrounds will not be sufficient. In contrast, the absence of confounding by correlated aetiological (genetic and environmental) influences will further reinforce the utility of targeting dietary diversity for mental health preventive efforts. The present research, thus, aims to deepen the understanding of the interplay between SES, dietary diversity and psychological distress and disentangle overlapping genetic and environmental influences on these relationships. Considering that most mental health studies using genetically sensitive designs are carried out in higher-income countries which may not generalise to LMICs, our findings can serve as a first step in identifying targets for preventive mental health interventions in LMICs to reduce the existing treatment gap. Furthermore, the recent history of civil conflict in Sri Lanka may make our findings generalisable to LMICs with high levels of environmental stress. The present study will test 1) whether SES and dietary diversity are associated with psychological distress, 2) whether the association of SES with psychological distress symptoms is mediated by dietary diversity, 3) whether the direct phenotypic mediation effects are a better explanation for the observed covariances of the latent factors.