In the present work, the viability of the selective fractionation of carbohydrate mixtures (lactose-lactulose) using supercritical CO2 has been evaluated. As a first step towards the optimization of the fractionation process, the composition of the co-solvent used in the supercritical phase was studied by considering different ethanol:water mixtures, which undoubtedly determined the selectivity of the extraction process. The co-solvent which provided a purity higher than 95% in lactulose (a mixture ethanol:water 95:5) was selected to perform the optimization of the extraction conditions able to provide high extraction recoveries. Optimization was carried out using a full factorial (2 levels) design and considering as factors the extraction pressure (from 100 to 300 bar), the extraction temperature (from 60 to 100 degrees C and the modifier flow rate (from 0.2 to 0.4 mL/min, which corresponded to a total co-solvent percentage ranging from 4 to 18 vol.%). The responses evaluated were the amount (mg) of lactulose and lactose extracted and their recoveries (%). The statistical analysis of the results provided mathematical models for each response variable. The corresponding parameters were estimated by multiple linear regression (NILR) and high determination coefficients (higher than 0.98) were obtained. Then, the mathematical model was employed to predict the behavior of the different responses selected, as a function of the main variables involved in the process. The optimum conditions for selectively extracting lactulose with high recoveries were 100 bar, 100 degrees C and 0.2 mL/min of co-solvent (4% modifier in the supercritical phase); at these conditions, a 45% recovery was achieved with a purity higher than 95%. (c) 2006 Elsevier B.V. All rights reserved.