Aims: The initial aim of this thesis was to understand the relationship between non-psychotropic medication and risk of suicidality. This was achieved by conducting a systematic review, which, among other conclusions, identified the need for improved estimation of risk of suicide and attempted suicide associated with antiepileptic drugs (AEDs). This stimulated this programme of research which sought to estimate the risk of suicide and other causes of unnatural death in people with epilepsy, the role of AEDs in fatal poisonings, the risk of self-harm in people with epilepsy and factors associated with self-harm amongst those people with epilepsy. Methods: Cohorts of individuals with prevalent epilepsy were identified separately in two population-based linked-primary care datasets: the Clinical Practice Research Datalink (CPRD) in England and the Secure Anonymised Information Linkage (SAIL) in Wales. Individuals were matched on age, gender and general practice to up to 20 people without epilepsy. The risks of cause-specific types of unnatural death (e.g. suicide, accident) were estimated using stratified Cox proportional hazards models, adjusted for level of deprivation. From each of the prevalent epilepsy cohorts, individuals with incident epilepsy, no history of self-harm and who were new users of the AEDs; carbamazepine, lamotrigine or valproate, were identified. The risk of first self-harm event associated with each AED compared to valproate was estimated using Inverse Probability of Treatment Weighting propensity score analysis. Estimates from each dataset were combined in a random effects meta-analysis. In the CPRD, the risk of self-harm in the incident epilepsy cohort versus a comparison cohort was estimated using a stratified Cox proportional hazards model. From this cohort, a nested case-control study was constructed. Individuals with a first self-harm event (cases) were matched to up to 20 people with no history of self-harm (controls). Conditional logistic regression was used to estimate the risk of self-harm associated with various factors including history of mental illness diagnoses, referrals and AED utilisation. Results: There were 44,678 and 14,051 people in the prevalent epilepsy cohorts and 891,429 and 279,365 in the comparison cohorts, in the CPRD and the SAIL respectively. Increased risks of suicide (HR 2.15, 95%CI 1.51-3.08) and accidental death (HR 2.97, 95%CI 2.54-3.48) were observed for people with epilepsy versus the comparison cohort, from the deprivation-adjusted meta-analysed estimates. Overall, AEDs were involved in 9.7% (95%CI 3.6%-19.9%) of the 62 poisoning deaths in people with epilepsy. There were 5,107 new users of carbamazepine, lamotrigine or valproate with incident epilepsy in the CPRD and 2,654 in the SAIL. No increased risk of self-harm was evident for carbamazepine (HR 1.53, 95%CI 0.89-2.64) or lamotrigine (HR 1.35, 95%CI 0.79-2.29), compared to valproate, from the meta-analysed estimates. In the CPRD, there were 11,690 individuals with incident epilepsy and 215,569 in the comparison cohort. The deprivation-adjusted hazard ratios for first self-harm event were 5.31 (95%CI 4.08-6.89) in the year following diagnosis and 3.31 (95%CI 2.85-3.84) in subsequent years. The nested-case control study derived from this incident epilepsy cohort included 273 cases of first self-harm and 3,790 controls. An increased risk of self-harm was associated with history of a mental illness diagnosis (OR 4.08, 95% CI 3.06-5.42) or referral to specialist psychiatric services (OR 3.41, 95% CI 2.63-4.43), compared to none; or being prescribed no AEDs (OR 1.47, 95% CI 1.01-2.12) or two AEDs (OR 1.84, 95% CI 1.33-2.55) in the 90 days prior to index date, compared to a single AED. Augmentation of AED treatment carried an elevated risk (OR 2.12, 95% CI 1.38-3.26) whereas there was no evidence to indicate that switching from one AED to another altered risk (OR 0.69, 95% CI 0.21-2.23). Conclusions: Compared to those without the condition, people with epilepsy are at an elevated risk of unnatural death, including suicide and accidental death, and nonfatal self-harm. The risk of self-harm is particularly elevated in the year following diagnosis of epilepsy but persists beyond this. Factors associated with increased risk of self-harm within the epilepsy population include prior mental illness and referral to psychiatric services. There was no evidence of difference in the risk of self-harm associated with carbamazepine or lamotrigine compared to valproate, but further replication of this result would be beneficial. However, treatment with multiple AEDs and augmentation of AED treatment increase the risk of self-harm within this population. These may be markers of uncontrolled epilepsy.