The secondary analysis of the Diabetes Prevention Program (DPP) showed that withdrawal of metformin for 1–2 weeks resulted in a trend to a higher conversion rate from impaired glucose tolerance to diabetes as compared with the placebo group (1). This resulted in a reduction by 26% of the so-called prevention effect of metformin, i.e., from 31% in the primary analysis (2) to 25% in the secondary analysis, a reduction in the incidence of diabetes that, however, remained highly significant as compared with placebo. The same trend was observed in the Study to Prevent (STOP)- NIDDM trial with acarbose after a longer washout period of 3 months (3). As previously discussed (4) and emphasized in a previous Diabetes Care Editorial (5), one key question is to know whether the positive results with metformin in the DPP (2) or with acarbose in the STOP-NIDDM trial (3) could be interpreted as a real prevention of the disease or only as a delay in its progression, or even simply as a masking effect due to the metabolic effect of the drug. It was astonishing that a possible direct effect of metformin was not discussed in the original paper of the DPP, despite the fact that the results were initially presented without any washout period (2). Indeed, a significant improvement of insulin sensitivity and glucose metabolism was consistently demonstrated in placebo-controlled randomized clinical trials after only 48 h treatment with metformin (850 mg twice daily), both in insulin-resistant normoglycemic first-degree relatives of diabetic patients during a euglycemic insulin clamp (6) and in hyperglycemic patients with type 2 diabetes with the isotope dilution technique to measure hepatic glucose production (7).
Using the frequently sampled intravenous glucose tolerance test and the minimal model approach, we also reported such acute favorable effect of metformin in a population similar to that of the DPP, i.e., in subjects with impaired glucose tolerance with a significant increase of the insulin sensitivity index after only 48 h metformin therapy (+50%; P = 0.03) (8). That part of the overall effect seen in the DPP may be accounted for by a pure pharmacological effect of metformin that did not persist when the drug was stopped (1) is in agreement with those experimental results (6–8). Furthermore, the rather short washout period of only 1–2 weeks may be questionable. First, it did not allow a sufficiently high number of new cases of diabetes, which limited the statistical power of the analysis (5)—the P value was 0.098 despite a difference of 49% in diabetes rates between placebo and metformin groups during the washout period. In addition, while the washout period was beyond the classically requested period of five times the plasma half-life of the drug (9), it did not completely exclude the persistence of any biological effect of metformin, which has a particularly complex cellular mechanism of action. Thus, in our opinion and in agreement with Buchanan (5), the question remains open as to whether the results of the DPP with metformin correspond to preventing, delaying, or partially masking effects of type 2 diabetes.
