Project Summary/Abstract Background Information and Relevance: Uveitis is an important cause of permanent vision loss that affects younger patients. Despite the human and economic impact of this disease, the risk factors for non-infectious uveitis are poorly understood. This is in part because epidemiologic studies of uveitis have been limited by insufficient numbers of participants. Newly available large health care claims databases provide an opportunity to increase the ability to detect uveitis risk factors. Hypotheses: Metformin, statin, angiotensin converting enzyme inhibitors are associated with a lower incidence of non-infectious uveitis, while female hormonal therapies are associated with a higher incidence of non-infectious uveitis. Specific Objectives: 1. To determine if non-infectious uveitis incidence varies in relation to putative protective medications, including metformin, statins and angiotensin converting enzyme inhibitors. 2. To determine if non-infectious uveitis incidence varies in relation the modifiable risk factor of female hormonal therapy, including hormonal replacement therapy and hormonal contraceptive therapy. Methods: The Clinformatics™ Data Mart Database contains medical claims on over 60 million beneficiaries from a large insurer in the United States. We will define non-infectious uveitis based on validated diagnosis codes recorded by an eye care provider twice within a 120-day period and exclusion of infectious or surgical causes of uveitis with diagnosis and procedural codes. Potential confounders including demographic (age, gender, race/ethnicity, education level, financial net worth) and clinical (smoking exposure) covariate information will be extracted from the database. Medication exposure will be rigorously captured based on the filling of outpatient prescriptions or coding of clinic-administered therapies. The cohort not exposed to the medication will be matched on age (±3 years), race/ethnicity, sex and date of plan entry and exit (±3 months) to the medication-exposed cohort. Propensity scores for each medication exposure will be estimated using multivariable logistic regression and the rich information on comorbid conditions and treatments available in the database. With multivariable Cox proportional hazards regression, we will calculate the hazard ratios for incident non-infectious uveitis based on exposure to each of the medications listed above. To account for the possibility of systematic differences between individuals with and without the exposures of interest, the Cox proportional hazards models will be weighted by the inverse of the predicted probability of their observed exposures using the propensity scores. We will interpret the results taking into the account the multiple comparisons being tested. Implications: The well-powered, rigorous analyses proposed here offer a unique opportunity to identify novel modifiable protective and risk factors for non-infectious uveitis, guide practice pattern...