Resolving drug effects from class effects among drugs for type 2 diabetes mellitus: more support for cardiovascular outcome assessments.

As illustrated most recently by the controversy surrounding the myocardial infarction risk associated with rosiglitazone, substantial uncertainty remains regarding the effects of different glucoselowering drug classes and, importantly, different drugs within each class on cardiovascular (CV) risk and mortality outcomes in patients with type 2 diabetes (T2DM). In the setting of the ongoing proliferation of anti-hyperglycaemic therapeutic classes and formulations with myriad therapeutic options for the treatment of T2DM presently available, this uncertainty has prompted regulatory agencies in both Europe and the USA to reassess the approval process for new T2DM medications, with changes focused primarily on excluding with a specified degree of statistical certainty incremental CV risk prior to new drug approval. Longterm randomized clinical outcome trials with both new and presently available medications are recommended, but not mandated. In the absence of definitive CV risk assessment from randomized trials for presently available drug classes and individual drugs within each class, critical analyses of existing databases are both imperative and informative. In this context, Schramm et al. have now reported the results of a nationwide registry-based observational analysis of clinical outcomes associated with various insulin secretagogues each compared with metformin. While this is not the first study to evaluate outcomes with these drug classes comparatively, the observations are among the most robust published based on the very large sample of patients with drug choices largely free of selection bias, sufficient numbers of events ascertained to yield substantial statistical power to analyse outcomes for each insulin secretagogue individually, with additional stratification by history of previous myocardial infarction. The overall results of the study suggest that most but not all insulin secretagogues (sulphonylureas and meglitinides) are associated with worse outcomes compared with metformin. Tolbutamide, glibenclamide (known as glyburide in the USA and Canada), glipizide, and glimepiride were all associated with significantly increased mortality and CV risk compared with metformin, but outcomes with gliclazide and repaglinide were not statistically different from those with metformin. In interpreting these data, it is of key importance to note that the observation of less benefit with most sulphonylureas in the study compared with metformin should not be interpreted as causing harm. Given the fact that metformin has an estimated risk reduction of 40% for major adverse cardiac events and death compared with placebo, when comparing outcomes associated with other drugs against metformin, hazard ratios of up to 1.7 would suggest treatment effects similar to or better than placebo, especially when considered in the context of favourable effects on microvascular disease risk associated with improved glucose control. Therefore, beyond the direct comparisons with metformin of each secretagogue, the most important and novel finding of the present study is the variability of the estimates of hazard associated with individual insulin secretagogues, suggesting that some may be better than others with regard to the outcomes assessed. Of course, as noted by the investigators, such interpretations are limited by the non-randomized observational nature of the present analyses deriving from an administrative database, with some variance in the propensity to prescribe the specific secretagogues analysed that may confound associations beyond the ability to adjust completely for differences in patient mix between the secretagogue groups. The apparent paradox of superior outcomes with metformin, a drug with modest glucoselowering properties, compared with sulphonylureas that are approximately twice as potent raises the possibility that some