In a carefully conducted randomized clinical trial (RCT), Hong et al. (1) report that 3 years of metformin treatment was associated with significantly lower cardiovascular (CV) risk after an additional 2 years of postdrug follow-up versus glipizide (a sulfonylurea) in 304 type 2 diabetic (T2D) patients with coronary artery disease. The study is unique because UK Prospective Diabetes Study (UKPDS), the landmark study creating metformin’s cardioprotective reputation, did not report CV outcomes comparing metformin versus sulfonylurea (2). Thus, the study by Hong et al. is probably the first report of CV outcomes comparing metformin versus sulfonylurea from a dedicated CV trial. The lack of previous such comparison is remarkable, and Hong et al. should be applauded for their contribution.
Some caveats may apply, however. First, several previous larger-scale RCTs in T2D patients, including dedicated CV trials, do not necessarily support metformin as cardioprotective—maybe the contrary. For example, the significantly increased mortality of metformin (17.5%) versus usual care (11.5%) on sulfonylurea background in a UKPDS trial (n = 537; median follow-up: 6.6 years) (2); the significantly lower risk of serious CV events with sulfonylurea (1.8%), but not with metformin (3.2%), versus rosiglitazone (3.4%) in the A Diabetes Outcome Progression Trial (ADOPT) (n = 4,360; median treatment duration: 4 years) (3); and the numerical higher mortality with metformin (7.6%) versus rosiglitazone (6.2%) in sulfonylurea users in the Rosiglitazone Evaluated for Cardiac Outcomes and Regulation of Glycaemia in Diabetes (RECORD) study (n = 2,225; mean follow-up: 5.5 years) (4). Such results, not observational but from larger-scale RCTs, may question metformin’s safety. Does Hong et al. prove otherwise? Without a placebo (or untreated) group, an active comparator study, as from Hong et al., can probably not prove the safety (or efficacy) of either intervention—only versus each other. Both may induce harm (or benefits), possibly to various degrees. Accordingly, most ongoing diabetes drug CV trials are placebo-controlled. Unfortunately, and probably contrary to common belief, adequately powered placebo-controlled RCTs of metformin do not exist to demonstrate its long-term safety.
Second, in the study by Hong et al., apparently as predefined (though the protocol was not provided), follow-up was extended for about 2 years postdrug without attempts to maintain previous therapies; i.e., treatments (any) in that not-short period were potentially confounded between groups (not shown). For unclear reasons, and unlike information at www.clinicaltrials.gov (1,5), the authors reported primary CV outcomes after the postdrug follow-up period (median 5 years; range 3.7–5.7), but not after the 3-year treatment period as done for HbA1c, lipids, secondary CV outcomes, etc. (1). Somewhat confusing, some of the 3-year data were reported as “end of follow-up,” while, except CV outcomes, corresponding data at the actual end of follow-up (5 years) were apparently not collected (1).
Finally, for survival data analysis usually a Kaplan-Meier plot shows events during the entire study period allowing for judgment if event rates were uniformly distributed throughout. The authors did not show such plot, unfortunately. Thus, we do not know if risk was similar while on the trial drugs versus off the trial drugs. In clinical practice, glucose-lowering treatment often is chronic. Hence, risk on treatment may be as important as any off-treatment legacy.
For better understanding of the unique data from Hong et al. (1), we hope the authors will provide additional CV (and mortality) risk estimates for the 3-year treatment period and Kaplan-Meier plots for the whole study period.
Duality of Interest.
S.S.L. and Y.G. are employees of Boehringer Ingelheim and have stocks in Novo Nordisk A/S. No other potential conflicts of interest relevant to this article were reported.
