Personalising the pharmacological treatment of bipolar disorder (360G-Wellcome-211085_Z_18_Z)

Background Prescribing for bipolar disorder is a major clinical dilemma as long-term pharmacological treatment is often necessary. Lithium is the most effective mood stabiliser. However, only 30% of individuals have a good therapeutic response. Presently, there is no reliable way to predict response or adverse event risk, or if an alternative treatment would be better for that patient. Aim To personalise prescribing for people with bipolar disorder via prediction models that quantify potential benefits and risks of existing treatments based on clinical phenotypic characteristics of the individual. Objectives Identify individualised clinical predictors of lithium and second-generation antipsychotic response. Determine clinical predictors of chronic kidney disease in individuals taking lithium. Determine clinical predictors of pathological weight gain in individuals taking second-generation antipsychotics. Methodology Data sources Swedish population registers, Hong-Kong health registers, Taiwanese health insurance database, UK primary care data linked to secondary care admission records, and UK mental health care data. Analyses Traditional epidemiological and machine learning methods; drawing on the strengths of each approach. Prediction model generation I will combine predictors from different datasets; resulting in models predicting drug response, chronic kidney disease and weight gain. Application Prediction models will be presented as online and smartphone application clinician decision aids.