Should metformin be used to prevent diabetes in high-risk patients?

Lifestyle^1,2 and pharmacologic efforts^1,3,4 to reduce the risk of new-onset diabetes mellitus have been the subject of multiple investigations. Retrospective and prospective studies have suggested benefits with use of glucose-lowering agents,^1,4,5 angiotensin-converting enzyme inhibitors, and angiotensin receptor blockers.³ The

article by Salpeter and colleagues

is a meta-analysis of the metformin studies.⁶ As outlined by the authors' use of metformin is an appealing option because it is safe, produces few side effects, and is a cost-effective way to target some of the defects known to contribute to the metabolic defects associated with diabetes.

Limitations of such meta-analyses deserve comment. The major consideration in any study of glucose-lowering agents is whether the observed reduction in progression to diabetes mellitus is simply a function of glucose lowering or whether the effects of the agent, in this case metformin, are "durable" after discontinuing treatment. Only one of the studies included in the authors' meta-analysis, the Diabetes Prevention Program, tested durability of effect, but only for a 2-week period.^1,7 The trial showed that metformin reduced the incidence of diabetes by 31% as compared with placebo. The reduction after only a 2-week washout was 25%.⁷ Furthermore, data from the United Kingdom Prospective Diabetes Study (UKPDS) demonstrated quite clearly that metformin use does not slow the progressive loss of beta cell function, which is one of the major defects in diabetes.⁸ Based on these considerations, the authors' conclusions about the magnitude of the effect of metformin should be interpreted cautiously.

The authors have noted that metformin has a number of favorable effects on risk factors associated with coronary artery disease (CAD). Use of this drug in the UKPDS trial was associated with a 39% reduction in the risk for myocardial infarction in obese patients compared with conventional treatment.⁹ The UKPDS was clearly an "impact" trial, but it was carried out at a time when statin therapy was not in use, blood pressure goals were higher than currently recommended, and many of the strategies for risk reduction were not in place; thus, the inference that the favorable effects of metformin on obesity, insulin resistance, and lipids would likely reduce CAD risk may be correct. Nevertheless, in the absence of clinical outcomes data, cautious interpretation is advisable.

The authors contrast their study with recent thiazolidinedione (TZD) studies, which have shown a reduction in diabetes risk and suggest a "durable" effect.^4,5 However, the authors point out that this class of agents may have unacceptable adverse side effects, including an increased risk for CAD and heart failure. Durability of effect on glucose control also has been supported by data from A Diabetes Outcomes Progression Trial (ADOPT), in which glycemic control was maintained for longer periods of time with rosiglitazone than with metformin.¹⁰ The effects of TZDs on exacerbating heart failure are well known, but were very low in the Diabetes Reduction Assessment with Ramipril and Rosiglitazone Medication (DREAM) study; the patients in this trial were similar to those studied in the authors' meta-analysis.⁴ The meta-analysis by Nissen and Wolski¹¹ found that rosiglitazone was associated with a significantly increased risk of myocardial infarction and death from cardiovascular causes; however, this study has been criticized for the use of a random effects model in subjects from studies that were clearly heterogeneous (eg, a nondiabetic cohort in DREAM⁴ and a diabetic cohort in ADOPT¹⁰).

The recent Action to Control Cardiovascular Risk in Diabetes trial,¹² which assessed a population at high risk for CAD and carefully characterized each patient and had well-adjudicated end points, does not confirm that the small increase in all-cause mortality was attributable to any adverse effect of TZDs. In fact, the study showed a reduction in atherosclerotic disease events of about 10% (P = NS). In light of the recent clinical trial data, the somewhat negative picture of TZDs painted by the authors needs more cautious interpretation. I think there may be a role for the use of metformin in patients at risk of developing diabetes, but widespread use of this drug has not yet become the clinical norm.