Using real world data from real world patients... this is clinical, not claims data...
"In the monotherapy cohorts, there were 12,440 users of rosiglitazone, 16,302 of
pioglitazone, 131,075 of metformin, and 48,376 of sulfonylureas. For the combined
endpoint of myocardial infarction plus coronary revascularizations, the hazard ratio forrosiglitazone versus pioglitazone was 0.97 (95%CI 0.78 â€“ 1.20), indicating essentially no difference between these thiazolidinediones. Both agents had somewhat less favorable outcomes than metformin, and both had somewhat better outcomes than sulfonylureas. In the dual therapy cohorts, there were 37,906 users of rosiglitazone in conjunction with metformin or sulfonylureas and 27,415 users of pioglitazone in similar combinations. Outcome rates in the rosiglitazone users versus the pioglitazone users were similar in combination with both metformin (HR 0.97, 95%CI 0.81 â€“ 1.17) and sulfonylureas (HR 1.12, 95%CI 0.89 â€“ 1.41). No combination with rosiglitazone or pioglitazone was meaningfully different in terms of outcomerates from a metformin-sulfonylurea combination.
The incidence of a combined endpoint of myocardial infarction and coronary
revascularization in users of rosiglitazone appears to be nearly the same as in users of pioglitazone, metformin and sulfonylureas. There is strong evidence against the
proposition that rosiglitazone in particular is associated with as much as a 40 percent
increase in risk. As in previous analyses of observational data in the US, the results
from the monotherapy and the dual-therapy comparisons, though not individually
significant, are consistent in suggesting that the risk of CHD events in patients using
thiazolidinediones may lie between the risks associated with sulfonylureas (higher
incidence) and metformin (lower incidence)."
http://www.fda.gov/ohrms/dockets/ac/07/briefing/2007-4308b1-00-index.htm (Appendix D)