Table 2—

Observational studies of TZD exposure and association with heart failure

Study (ref.)Study type and data sourceStudy populationOutcome ascertainmentKey findings
Hartung 2005(19)Case control. Oregon Medicaid insurance claims database.Diabetic patients (288 case and 1,652 control subjects) with prescription claims data for OHAs within 60 days of hospital admission.ICD codes used to verify cases with first hospital admission due to heart failure. Control subjects had admission for other reasons.59 of 229 cases of heart failure had been exposed, whereas 216 of 1,436 controls were exposed.
Karter 2005 (20)Retrospective cohort. Kaiser Permanente Northern California Diabetic registry.23,440 patients initiating diabetes medication (“new user”) with no history of heart failure.ICD codes used to identify admissions for heart failure.Heart failure rates in 67 of 3,556 exposed vs. 253 of 19,884 unexposed.
Inzucchi 2005 (21)Retrospective cohort. Medicare National Heart Care Project.Diabetic patients with recent admission for myocardial infarction discharged on hypoglycemic agents.Readmission events (including heart failure) recorded for up to 1 year after discharge.Heart failure rates in 402 of 819 exposed compared with 3,294 of 7,914 unexposed.
Delea 2003 (22)Retrospective cohort, based on Pharmetrics Insurance database.Diabetic patients with pharmacy claim for oral hypoglycemic drugs. Patients with existing heart failure excluded.Follow-up data for 43 months to diagnose new heart failure cases. Verification through ICD codes for claims.Heart failure rates in exposed: 126 of 5,441 vs. unexposed: 397 of 28,103.
Masoudi 2005 (23)*Retrospective cohort. From a Medicare database.16,417 diabetic patients with a pre-existing history of hospital admission for heart failure.1-year follow-up to estimate risk of readmission for heart failure.Heart failure rates in 1,505 of 2,226 exposed vs. 8,912 of 13,093 unexposed.
Rajagopalan 2004 (24)*Retrospective cohort. Pharmetrics insurance database in U.S.1,668 matched pairs initiating pioglitazone or insulin between January 1999 and December 2001. Those with existing heart failure were excluded.Follow-up data of at least 3 months, using ICD codes for claims arising from heart failure.Heart failure rates in 33 of 2,226 on pioglitazone compared with 66 of 1,668 on insulin.
  • * Observational studies excluded from the meta-analysis. OHA, oral hypoglycemic agent.