Diabetes mellitus is a common, expensive, and burdensome chronic illness among Veterans. Many new therapies for management of hyperglycemia have emerged in recent years, but there continues to be concern and uncertainty regarding their safety and whether there are any incremental long term benefits. Most efficacy trials are restricted to relatively healthy diabetes patients who do not have the multiple co-morbidities often present in patients in actual VA clinic populations. This is further complicated by safety concerns such as hypoglycemia or increased risk of cardiovascular disease (CVD) events with use of certain agents, or increased mortality with intensive treatment of hyperglycemia. Further evidence on safety and effectiveness of diabetes medications from on-going clinical trials is likely to be limited, so high quality observational studies of comparative safety and effectiveness are needed. Currently, there are ten classes of drugs for diabetes, some with multiple agents in class. Although the overall benefits of maintaining glycemic control for avoiding metabolic complications and reducing risk of micro-vascular disease are well established, the evidence is incomplete and inconsistent for treatment benefits in reducing risk of cardiovascular disease and other macro-vascular complications, the most serious and fatal complications of diabetes. Furthermore, our understanding of the long-term safety of these agents in actual clinical populations is limited. The analysis of national VA data represents a unique opportunity for comparative safety and effectiveness evaluation specifically for diabetes research.
Our overall goal was to study the safety and effectiveness of diabetes therapies in patients served by the VA using the Diabetes Epidemiology Cohorts (DEpiC), a national registry of all VA diabetes patients (over 2 million) that was established by our research team. Specific aims of the project included evaluation of (1) metformin use, particularly in terms of lactic acidosis and CVD risks, relative to other alternative drugs; (2) sulfonylurea use, in terms of hypoglycemia and CVD risks, comparing to other medications and among the specific agents in class; (3) thiazolidinedione (TZD), in terms of fracture and CVD risks: (4) relative effectiveness of various agents in terms of glycemic control (lower HbA1C over time and longer time to subsequent new therapy), medication persistence, progression of microvascular complications (nephropathy, neuropathy/lower limb amputations, retinopathy) and time to long-term care.
This research used the Diabetes Epidemiology Cohorts (DEpiC), a national registry of all VA diabetes patients over more than a decade (about 2 million) that was established by our research team. Using linked longitudinal data from FY1998-2010, we applied previously developed methodologies to compare medications for diabetes patients in terms of their safety and effectiveness. Drug-to-drug and combination drug comparisons were made in new-user analyses with medications initially prescribed at different points in the course of diabetes treatment (first line, second line, third line, insulin initiation or supplementation). Within these cohorts, we stratified patients by medication and medical history to form homogeneous groups at similar stages of diabetes progression and conducted a series of new-user cohort studies. Patients were followed forward in time for outcomes of interest with comparisons made between patients initiating different drugs within class or group. Detailed information on course of treatment and disease status as well as other characteristics of patients and their care were included in the analyses. Rigorous epidemiologic methods were used to reduce potential bias and provide valid and meaningful results. This included careful selection of sample and exposure groups to replicate common clinical decisions in treatment, with continuous monitoring of patient records for study outcomes following drug initiation. We used Cox proportional hazards regression, propensity score matching, instrumental variables, and detailed sensitivity analyses to further reduce potential bias. All analyses were done in the overall sample and in subsamples of complex patients defined by concurrent chronic complex illnesses.
We obtained and linked study data, refined our methods, established diabetes cohorts, and conducted most of the analyses detailed in the proposal. In concurrent studies related to first and second line agents we have found that metformin use is associated with lower risk of cardiovascular disease events and death, unless it is prescribed in combination with sulfonylureas. In contrast, sulfonylurea use was associated with about a 50% increase in risks of CVD events (MI, stroke, cardiac surgery, and chronic heart failure) and death and this persisted in all parallel analyses for confounder adjustment. Among the sulfonylureas, glyburide was associated with about 20% higher risks of hypoglycemia and cardiovascular disease events but there were no differences in risk of death. TZDS were not associated with increases in CVD but we did find small increases in risks of fracture and chronic heart failure with their use. We have also conducted analyses evaluating risks for various outcomes with incretin drugs (e.g. sitagliptin, exenatide) but the numbers were not yet sufficient; we plan to continue these analyses with more recent data. In addition, we found important differences in rates of persistence, glycemic control, and time to onset of diabetes complications in comparisons among first and second line agents.
This research provides a better understanding of benefits and risks associated with treatments for diabetes in complex patients. In addition to peer-reviewed publications, we share our results with colleagues from the VA Pharmacy Benefits Management Service and the Specialty Care group from VA Patient Care Services. Application of its findings should be used rapidly in adjusting policy for safer and more effective prescribing practices, in veterans, and more broadly in all diabetes patients.
External Links for this Project
Grant Number: I01HX000391-01A1
- Powell WR, Christiansen CL, Miller DR. Meta-Analysis of Sulfonylurea Therapy on Long-Term Risk of Mortality and Cardiovascular Events Compared to Other Oral Glucose-Lowering Treatments. Diabetes therapy : research, treatment and education of diabetes and related disorders. 2018 Aug 1; 9(4):1431-1440. [view]