Capitalizing on prescribing pattern variation to compare medications for type 2 diabetes

Area of Science:

• Pharmacology and Pharmacoepidemiology
• Health Services Research
• Biostatistics

Background:

• Clinical trials often focus on glycemic control, neglecting long-term outcomes like mortality for hypoglycemic medications.
• Assessing long-term drug effects typically requires lengthy and costly clinical trials.
• This study explores an alternative methodology for evaluating long-term medication impacts.

Purpose of the Study:

• To employ observational, quasi-experimental methods with instrumental variables (IVs).
• To compare the long-term outcomes of sulfonylureas (SUs) versus thiazolidinediones (TZDs) as second-line diabetes treatments.
• To evaluate effects on mortality, hospitalizations, and major cardiovascular events.

Main Methods:

• Utilized administrative data from Veterans Health Administration and Medicare (2000-2010).
• Included US veterans prescribed metformin who then initiated SUs or TZDs.
• Employed IV models using regional SU prescribing variations to mitigate selection bias; survival models predicted outcomes.

Main Results:

• Initiating sulfonylureas (SUs) versus thiazolidinediones (TZDs) significantly elevated the risk of all-cause mortality (HR 1.50, 95% CI 1.09-2.09).
• SUs also increased the likelihood of ambulatory care sensitive condition (ACSC) hospitalizations (HR 1.68, 95% CI 1.31-2.15).
• No significant difference was found for acute myocardial infarction or stroke (HR 1.15, 95% CI 0.80-1.66).

Conclusions:

• Findings indicate a higher risk of major adverse events with SUs as a second-line hypoglycemic agent.
• Quasi-experimental IV methods offer a valuable alternative to extended clinical trials for long-term outcome assessment.
• This approach can enhance the evaluation of medication safety and effectiveness over time.