A study was undertaken aimed at developing a dynamic in vitro method based on the use of artificial membranes for screening and prediction of the absorption properties of drugs. The development of a suitable artificial membrane simulating the behaviour of the natural gastrointestinal one is the critical step for the fulfillment of this research and it has been the aim of the present work. A series of filters of different chemical nature and different characteristics (pore size, percent of porosity, thickness) were evaluated and compared in order to select the most suitable one to use as support for the artificial lipid membrane preparation. Permeation experiments were performed by using naproxen as reference model drug. The target was to achieve a naproxen apparent permeability (P(app)) value close to that previously obtained using a Caco-2 cell dynamic methodology (4.88 x 10(-5)cms(-1)), which related well to the fraction of dose absorbed in vivo in human. Filters of cellulose nitrate-acetate mixture with a porosity of 70% and 0.025 microm pore size were selected on the basis of the highest reproducibility of results and suitable drug P(app) value obtained. Mixture design experiments were then carried out in order to optimize the composition of the lipid mixture to use for porous support impregnation to obtain the desired P(app) value. Sixteen different lipid mixtures were prepared according to mixture experimental plan and evaluated by diffusion experiments. The contour plot obtained was used to select an optimum lipidic mixture (cholesterol 2.10%; Lipoid((R)) E 80 1.70%; n-octanol 96.2%) for the cellulose nitrate-acetate membrane impregnation. The experimental response obtained from diffusion experiments performed by utilising the optimised lipidic impregnation mixture, was 4.88 x 10(-5)cms(-1), matching exactly the prefixed target.