OBJECTIVE To assess the average bioequivalence of two formulations of metformin after single-dose administration of each treatment to healthy subjects under fasting conditions by assessing the pharmacokinetic measures of systemic exposure and evaluating the confidence intervals (CIs) for each pharmacokinetic parameter. METHODS Randomised, comparative, single-dose, open-label, balanced, two-period, two-treatment, crossover study under fasting conditions. METHODS 20 healthy volunteers (ten men and ten women) took part in the study. METHODS Subjects were investigated after a single dose of 850mg after a washout period of 7 days. Plasma samples were taken at regular time intervals according to the study protocol for measuring plasma metformin concentrations. Systemic exposure was estimated with the use of pharmacokinetic parameters (area under the curve of the plasma drug concentrations from time zero to the last sampling time [AUC(0-36)], area under the curve of the plasma drug concentrations from time zero to infinity [AUC(0-)(infinity)], time to peak drug concentration [t(max)], partial area under the concentration-time curve with a cut-off point at the t(max) of the reference product [AUC(p)], peak plasma drug concentration [C(max)], the ratio C(max)/AUC(0-)(infinity), and mean residence time [MRT]). The point estimates of pharmacokinetic parameters (geometric means of the ratios test [T]/reference [R] and the 90% CIs for the ratios of geometric means [T]/[R]), estimated by parametric and nonpara-metric analysis, were used in the statistical analysis. RESULTS The point estimates and the 90% CIs after parametric analysis of AUC(0-)(infinity) were 0.98 and (0.96-1.21), and after nonparametric analysis were 1.06 and (0.95-1.207), respectively. The two drug products were considered to be bioequivalent and with significant variability across subjects for the pharmaco-kinetic parameters AUC(0-36), AUC(0-)(infinity), C(max) and MRT according to analysis of variance of log-transformed data. CONCLUSIONS The two studied formulations of metformin were found to be bioequivalent. They showed similar extents and rates of absorption and similar exposure. However, analysis of variance of logarithmically transformed data revealed significant variability among individuals in AUC(0-36), AUC(0-)(infinity) and C(max), making careful individualisation of the metformin dosage important.