More Reasons to Say Goodbye to Glyburide

Abstract

Sulfonylureas have been used to control hyperglycemia in type 2 diabetes longer than any class of agents except insulins. Their use early in the course of diabetes is supported by evidence of long-term medical benefit (1), and although available in generic form and thus not commercially promoted, they are widely prescribed (2). Given these facts, it is surprising that much remains unknown about these agents, notably how the various members of the class should be deployed in current treatment algorithms. An article by Zeller et al. in this issue of JCEM (3) sheds considerable light on this question. The authors report analyses of prospectively collected data from a French registry concerning experience of patients with type 2 diabetes who were hospitalized for myocardial infarction. The question posed was what association, if any, might exist between prior use of sulfonylureas and death and other in-hospital outcomes in these patients. Analyses included comparisons between the individual sulfonylureas used in this population: glyburide (called glibenclamide in Europe), glimepiride, and gliclazide (in most cases given as an extended-acting formulation). The rationale for the study came from the known adverse effect of glyburide on cardiac ischemic preconditioning, the clinical relevance of which has been uncertain. One key finding of the French study was that, among the 1310 acutely ill patients studied, unadjusted mortality rates were lower for those whose treatment on admission included a sulfonylurea (3.9%) than for those taking insulin (9.4%), those on no antihyperglycemic drugs (8.4%), and those on oral therapies other than sulfonylureas (6.4%). Another was that when patients taking glyburide were compared with those taking gliclazide or glimepiride, two sulfonylureas that appear to lack cardiac effects, early mortality was almost 3-fold higher (7.5 vs. 2.7%) with glyburide. It is well known that epidemiological analyses of “real-world” experiences with prescribed pharmacotherapy may be confounded by incomplete collection of information on actual drug usage, concomitant therapies, and medical outcomes, in addition to biases introduced by physicians’ tendencies to assign different treatments to different kinds of patients. Many of these problems were reduced or eliminated by the prospective design of this study, which included systematic recording of detailed clinical information on all patients in a narrow window of time by a structured team of investigators. The problem of allocation bias in assignment of treatments, however, remains in this study as in others. Hence, the authors used several statistical methods to adjust for this kind of confounding, including an extensive multivariable model, further adjustment using a propensity score for assignment to glyburide treatment, and analysis of findings with glyburide vs. the other sulfonylureas by subgroups. These statistical adjustments did not alter the associations of sulfonylurea use with lower mortality than other forms of therapy, and of glyburide use with higher mortality than seen with the other sulfonylureas. The excess of mortality with glyburide vs. other sulfonylureas showed no tendency toward heterogeneity between subgroups. Thus, several approaches to testing the validity of the observations were supportive, and the magnitude of the effects was substantial. After multivariable adjustment, the risk of death accompanyinguseofa sulfonylureawashalf thatwhena sulfonylurea was not used (odds ratio, 0.50; 95% confidence interval, 0.27–0.94; P 0.03). After similar adjustment for covariates, the risk of early mortality was 85% lower with glimepiride or gliclazide than with glyburide (odds ratio, 0.15; 95% confidence interval, 0.04–0.56; P 0.005), and with the propensity score included in the model, the risk was 87%lower(oddsratio,0.13;95%confidence interval,0.21–