Author	J.F. Hayes
Author	D.P.J. Oosborn
Author	K. Walters
Author	M.B. King
Author	L. Marston
Abstract	Objectives To determine 1) mortality and morbidity in people with bipolar disorder, and 2) the impact of maintenance medication on relapse/reoccurrence and adverse events. Methods Objective 1: I conducted a meta-analysis of studies examining mortality in bipolar disorder populations. I then carried out a cohort study in United Kingdom primary care electronic health records to understand rates of mortality and morbidity in bipolar disorder relative to the general population. Objective 2: I completed a network meta-analysis of the efficacy of maintenance mood stabiliser medications (lithium, valproate, olanzapine and quetiapine) in preventing relapse. I then carried out a series of cohort studies in primary care electronic health records. These studies examined 1) the effectiveness and tolerability of these medications, 2) the rates of renal, endocrine, hepatic and metabolic adverse events, and 3) the rates of self-harm, accidental injury and suicide. Propensity score methods were used to address issues of confounding. Results Objective 1: All-cause and cause specific mortality was elevated in people with bipolar disorder (summary standardised mortality ratio 2.05; 95% CI 1.89 to 2.23). In a cohort of 17,341 with bipolar disorder, mortality rates increased from the mid-2000s relative to the general population (hazard ratio increased by 0.14 per year; 95% CI 0.10 to 0.19). Objective 2: Trials comparing lithium, valproate, olanzapine, quetiapine and placebo did not show superiority of one drug. In the electronic health records cohort studies individuals prescribed lithium went for longer before treatment failure (for example valproate had hazard ratio 1.20; 95% CI 1.10 to 1.32 compared with lithium), had increased mild (but not severe) renal failure (hazard ratio for valproate: 0.56; 95% CI 0.45 to 0.69 compared with lithium), hypo- and hyperthyroidism and hypercalcemia rates. However, they had lower rates of clinically significant weight gain (hazard ratio for >15% weight gain with valproate: 1.62; 95% CI 1.31 to 2.01 compared with lithium) and there was no difference in hepatotoxicity, cardiovascular events or diabetes mellitus rates. Additionally, people taking lithium had lower self-harm (hazard ratio for alternatives: 1.51; 95% CI 1.21 to 1.88 compared to lithium) and accidental injury rates. Conclusions Bipolar disorder is associated with increased mortality and morbidity, and the mortality gap with the general population has widened in recent years. Despite limited trial evidence, lithium appears to offer the best opportunity for mood stabilisation. Lithium is associated with increased renal and endocrine dysfunction, but these risks are offset by the potential of more frequent weight gain with alternative drugs. Furthermore, lithium may be associated with specific anti-suicidal effects. These risk and benefits should be considered when individual treatment decisions are made.