Comparative Effectiveness of Dynamic Patterns of Glucose Lowering Therapies. In type 2 diabetes, the maintenance of intensive glucose control over time typically requires the use of multiple glucose lowering agents. Over the past decade, the availability of new glucose lowering agents has created more options for patients and providers and potentially more uncertainty regarding the optimal approach to achieving intensive glucose control. Uncertainty regarding treatment choices has been heightened by concerns regarding the possible adverse effects of glucose lowering agents and associated patterns of intensive glucose lowering. The glucose control portion of the Action to Control Cardiovascular Risk in Diabetes trial was terminated because of excessive deaths in the very intensive glucose control arm. The increase in deaths has been hypothesized to be related to the combinations of drugs employed as well as the overall pace with which glucose lowering was achieved. At present, little is known regarding the longitudinal nature of glucose lowering therapies in clinical practice. There is in fact no standardized approach to describing longitudinal treatment patterns or modeling their selection. During this 4 year award, we propose to utilize the entirety of the Veterans Health Administration type 2 diabetes cohort (N~900,000, years 2003-2011) to 1) describe patterns of care seeking, medication choices, and glucose control among patients with type 2 diabetes; 2) identify the clinical and non-clinical predictors of distinct patterns of care seeking and medication choices among type 2 diabetes patients; 3) perform comparative effectiveness on 9-year diabetes outcome rates between different patterns of glucose-lowering therapies and glucose lowering; 4) and develop forward- looking algorithms that will identify optimal individualized treatment at any point in time during nine years of glucose management. As part of this award, we will develop a micro-simulation model of dynamic diabetes treatment decisions, utilize novel instrumental variables to account for treatment selection bias, and validate the model both internally and externally. The completion of these aims will provide us with a useful taxonomy of dynamic patterns of glucose lowering therapies as well as an innovative approach to modeling the longitudinal nature of treatments that will be relevant for multiple chronic conditions. For type 2 diabetes, the studies will help to identify patterns of glucose lowering therapies that are particularly harmful and in which subpopulations the risk for adverse events is highest.

R01HS018542

2012-08-01

2017-05-31