Miller DR (Center for Health Quality, Outcomes, & Economic Research, Bedford VAMC), Christiansen CL
(Center for Health Quality, Outcomes, & Economic Research, Bedford VAMC), Palnatti M
(Center for Health Quality, Outcomes, & Economic Research, Bedford VAMC), Lafrance JP
(Center for Health Quality, Outcomes, & Economic Research, Bedford VAMC), Pogach L
(Center for Healthcare Knowledge Management, East Orange VAMC), Cunningham F
(Pharmacy Benefits Management Medication Safety Center, Hines VAMC)
Meta-analyses of clinical trials published in 2007 reported increased risk of myocardial infarction (MI) with rosiglitazone use, but methodologic concerns and other studies cast doubt on these findings. In response to this safety signal, the Veterans Affairs (VA) Pharmacy Benefits Management Medication Safety Center (VAMedSAFE) and affiliated researchers studied this issue to inform and guide policy decisions.
To evaluate MI risks associated with rosiglitazone and pioglitazone, a rapid analysis of VA data was conducted in fall 2007 followed by a comprehensive new-user follow-up study completed in summer 2008. Using the Diabetes Epidemiology Cohorts, a national registry of VA patients with diabetes, we identified 162,215 patients first prescribed rosiglitazone or pioglitazone from 2000 to 2004. After excluding those with past MI or cardiac surgery, renal disease, or chronic heart failure and stratifying the sample by diabetes medication history (lines), we calculated propensity scores in each line and matched rosiglitazone and pioglitazone initiators to patients initiating other diabetes medication. MI risk was estimated using Cox regression models.
Rosiglitazone and pioglitazone were prescribed most often as 3rd line agents following metformin and sulfonylurea (41%). In this line, the hazard ratios (HR) and 95% confidence intervals for MI from the comprehensive study were 0.99 (0.88-1.11) for rosiglitazone and 0.80 (0.65-0.98) for pioglitazone relative to insulin initiators, and 1.24 (1.02-1.52) for rosiglitazone versus pioglitazone. Results were similar across the other lines (1st line, 2nd line, insulin supplement) with HRs for rosiglitazone close to 1.0 and for pioglitazone at or below 1.0, except for a small risk increase for rosiglitazone when prescribed with insulin. Results from the initial rapid analysis were similar to those from the comprehensive study and were used in revising VA Criteria for Non-formulary Use of Thiazolidinediones in November 2007.
These studies provide little evidence for increased MI risk with rosiglitazone use but suggest that there may be slightly lower risk with pioglitazone use.
This demonstrates how collaboration between VA researchers and VAMedSAFE in evaluating medication safety signals using VA data can provide both rapid and scientifically rigorous results to serve policy needs.