A comparison of all-cause mortality with pioglitazone and insulin in type 2 diabetes: an expanded analysis from a retrospective cohort study

Objectives: The objectives of this study were to assess and compare all-cause mortality rates between pioglitazone (PIO) and insulin (INS). Research design: The study population included 56,536 patients with type 2 diabetes aged >= 45 years who were first-time users of PIO or INS. Data from 1 May 2000 until 30 June 2010 from the i3 InVision Data Mart database were linked to death records of the US Social Security Administration obtained in March 2012, with approval from the Institutional Review Board and in full compliance with the Health Insurance Portability and Accountability Act of 1996. Main outcome measures: Kaplan-Meier curves were generated and hazard ratios (HRs) were estimated for the occurrence of deaths in the PIO and INS cohorts using Cox proportional hazards models adjusted with inverse probability weights derived from propensity scores. Results: After adjustment for >40 covariates through inverse probability weights derived from propensity scores, the PIO group showed a significantly lower risk of all-cause mortality (HR 0.33; 95% confidence interval, 0.31, 0.36). The risk of all-cause mortality was also significantly lower in the PIO cohort than the INS cohort among subgroups based on baseline variables such as sex, age (<55 years, >= 55 years), antidiabetic medication use (sulfonylureas or metformin), lipid-altering medication use, and congestive heart failure status. The study has some limitations. Use of a claims database means a potential bias toward a younger cohort. Disease-specific mortality was not identified because of no recorded cause of death. Reliable information regarding the differences in disease deterioration rate and some clinical and lab results were not available, which limits the statistical adjustment of baseline variables. Conclusion: PIO was associated with a lower risk of all-cause mortality than INS.