, Volume 56, Issue 9, pp 1934-1943

Mortality among veterans with type 2 diabetes initiating metformin, sulfonylurea or rosiglitazone monotherapy

Rent the article at a discount

Rent now

* Final gross prices may vary according to local VAT.

Get Access

Abstract

Aims/hypotheses

Despite oral hypoglycaemic medications being the most commonly used pharmacological treatments for type 2 diabetes, research is limited on their comparative safety, particularly their effects on overall mortality. We compared mortality risk with monotherapy initiation of four oral hypoglycaemic medications in a nationwide cohort of US veterans with type 2 diabetes.

Methods

We identified new users of oral hypoglycaemic medication monotherapy between 2004 and 2009 who received care for at least 1 year from the Veterans Health Administration. Patients were followed until initial monotherapy discontinuation, addition of another diabetes pharmacotherapy, death or end of follow-up. Mortality HRs were estimated using Cox regression adjusted for potential confounding factors.

Results

Among new users of metformin, sulfonylureas and rosiglitazone (185,360 men, 7,812 women), 4,256 (2.2%) died during follow-up. Average duration of medication use ranged from 1.4 to 1.7 years. Significantly higher mortality risk was seen for glibenclamide (known as glyburide in the USA and Canada) (HR 1.38, 95% CI 1.27, 1.50) or glipizide (HR 1.55, 95% CI 1.43, 1.67) compared with metformin monotherapy, and for glipizide compared with rosiglitazone (HR 1.27, 95% CI 1.01, 1.59) or glibenclamide monotherapy (HR 1.12, 95% CI 1.02, 1.23). A significant sex–rosiglitazone interaction was seen (p = 0.034) compared with metformin monotherapy, with women having a higher HR (HR 4.36, 95% CI 1.34, 14.20) than men (HR 1.19, 95% CI 0.95, 1.49).

Conclusions/interpretations

Significantly higher mortality was associated with glibenclamide, glipizide and rosiglitazone use compared with metformin, and with glipizide use compared with rosiglitazone or glibenclamide. The potential for residual confounding by indication should be considered in interpreting these results.