Comparative Effectiveness and Safety of Newer and Older Antihyperglycemic Medications

Background: People diagnosed with diabetes are initially advised on lifestyle changes and started on metformin, but often require the addition of 2nd line antihyperglycemic medications. The choice of 2nd line antihyperglycemic therapy is complex because the multiple available drugs, distributed across several drug classes, have different benefits and risks. Newer 2nd line antihyperglycemics (sodium-glucose co-transporter-2 inhibitor (SGLT2i), and glucagon-like peptide-1 receptor agonist (GLP1)) have been shown to reduce risk of cardiovascular events compared to placebo in individuals at high risk of cardiovascular disease. Evidence also suggests that these newer agents may reduce risk of kidney disease in people with relatively preserved kidney function. Whether the newer agents offer benefits in cardiovascular and kidney outcomes compared to older (less costly) 2nd line agents including dipeptidyl peptidase-4 inhibitors (DPP4) and sulfonylureas, and whether these benefits extend to individuals with intermediate or low cardiovascular risk and people with reduced kidney function is not known. Significance/Impact: Results will provide real-world evidence to guide the selection of antihyperglycemic agents by cardiovascular risk status, kidney function category, and will provide evidence on the risk of adverse events. The proposal falls under several Health Services Research priorities including Primary Care Practice (diabetes is highly prevalent in veterans) and Health Care Informatics (using big data to advance care of veterans); and Office of Research & Development priorities as our approach will highlight VA data as a national resource and the results will have direct and substantial real-world impact in informing care. Innovation: The proposal will leverage the power of the VA’s large-scale electronic health records and recent methodologic innovations in causal inference and pharmacoepidemiology — specifically in the use of real- world observational data to emulate a target randomized trial — to provide much needed evidence of the comparative effectiveness and safety of newer vs. older antihyperglycemic agents. Specific Aims: To use observational healthcare data from the Department of Veterans Affairs to emulate four- arm randomized trials of the comparative effectiveness of incident use of newer (SGLT2i, GLP1) and older (DPP4, sulfonylureas) 2nd line antihyperglycemics—among metformin users—on cardiovascular outcomes (aim 1), kidney outcomes (aim 2), and evaluate the risk of adverse events associated with these drug classes (aim 3). Methodologies: For each specific aim we will define a target randomized trial protocol, including eligibility criteria, treatment assignment, treatment initiation, treatment strategy follow-up, outcome assessment, and analytic plan. We then will use electronic medical record data from the VA to construct aim-specific cohorts to emulate the specifications of the target trial for the related aim, estimating the differences in risk of cardiovascular disease, kidney disease, and adverse events between the studied antihyperglycemics. Randomization will be emulated by inverse probability of treatment weighting based on predefined variables and a high dimensional variable selection algorithm. Intention-to-treat effects will be estimated using discrete time survival analyses, and adjusted intention-to-treat-effects will be estimated after accounting for loss to follow-up. Per-protocol effects (of a specified treatment strategy) will be estimated after accounting for non- adherence to assigned treatment strategies. Differences in risk of outcomes between the second-line antihyperglycemics will be reported as hazard ratios and adjusted incidence rates. Implementation/Next Steps: The results from this proposal will inform clinical practice guidelines for the treatment of diabetes. Future studies will leverage advances in machine learning to create unique individualized precision care plans for each person with diabetes.