Abstract
Self-harm can be defined as any act of self-injury regardless of motivation or intention. Being an important predictor for future suicide death, a better understanding of its aetiology is needed in designing interventions to prevent or reduce its risks. There is also an ongoing debate whether self-harm should be further categorised along the dimension of suicidal intention into non-suicidal selfharm (NSSH) and suicidal self-harm (SSH), so that they can be treated as two separate conditions.Genetically informed designs are methods that utilise genetic information such as known genetic relationships (e.g., twins) or genetic variation data. These designs can be exploited for causal inference to paint a clearer picture of causality and help reveal the complex aetiology of self-harm. Using a range of genetically informed designs, this thesis intended to achieve two main aims: (1) to strengthen causal inference and identify plausible causal risk factors for self-harm; (2) to investigate to what extent NSSH and SSH are aetiologically similar (or different). The genetically informed designs employed in this thesis are: polygenic scoring, Mendelian randomisation (MR), classical twin design, and Mendelian randomisation - Direction of Causation (MR-DoC) model.
In Chapter 2, using polygenic scoring and Mendelian randomisation (MR), I identified plausible causal risk factors for self-harm in a systematic manner. Polygenic scores (PS) for known risk factors of self-harm were generated in the UK Biobank sample. MR was then conducted to further strengthen the causal inference. I found that schizophrenia, major depressive disorder (MDD), and attention deficit/hyperactivity disorder (ADHD) are the plausible causal risk factors for self-harm, suggesting that psychiatric conditions are prominent in its aetiology. Polygenic scoring analyses did not find evidence for aetiological difference between NSSH and SSH in this chapter.
In Chapter 3, I employed classical twin modelling to disentangle the genetic and environmental influences for NSSH and SSH. Using data from Twins Early Development Study (TEDS), I found that both NSSH and SSH have similar aetiological structure, with high genetic and non-shared environmental correlations, suggesting high aetiological similarity between them. I also found that NSSH and SSH share similar aetiological relationships with a range of mental health problems. Among them, child-rated insomnia and MDD symptoms have the strongest phenotypic and genetic correlations with both NSSH and SSH. Results from Chapter 3 suggest that NSSH and SSH are aetiologically similar.
In Chapter 4, I integrated MR within twin modelling to combine their strengths in causal inference using MR-DoC models. Potential causal risk factors identified from Chapter 2 (ADHD, MDD and schizophrenia) and Chapter 3 (MDD symptoms and insomnia) were included as exposures of interest in this chapter, whereas NSSH and SSH were the outcomes. Applying the MR-DoC model to TEDS data, I found that there were significant putative causal effects from child- and parentrated depressive symptoms to both NSSH and SSH, besides significant unmediated pleiotropy. This chapter further supported the role of MDD symptoms in the aetiology of self-harm and found no evidence of aetiological difference between NSSH and SSH.
Using a range of genetically informed methods, this thesis identified several plausible causal risk factors for self-harm and did not find evidence that NSSH and SSH are aetiologically different. This thesis also demonstrated that genetically informed designs are a worthy toolbox for causal inference. Findings from this thesis may be helpful in designing interventions for self-harm and in considering whether NSSH and SSH should be labelled as two separate conditions. Nonetheless, limitations concerning representativeness of the samples, measurement issues, and assumptions of the genetically informed methods used should be acknowledged in interpreting the findings of this thesis.
| Date of Award | 1 Apr 2023 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Cathryn Lewis (Supervisor), Fruhling Rijsdijk (Supervisor) & affiliated academic (Supervisor) |