Prediction of the mutual diffusion coefficient for controlled drug delivery devices

Abstract

Diffusion coefficients of drugs in polymers were predicted using models based on modified free volume theory of diffusion. The descriptors used were chosen to uniquely relate the structure with desired properties. Model parameters were obtained using quantitative structure property relationships (QSPRs) developed using multiple linear regression and artificial neural networks with Bayesian regularization. Viability of the approach was established by predicting solvent diffusion coefficient in polymer for four polymer–solvent systems (polystyrene–toluene, polyvinylacetate–toluene, polystyrene–ethylbenzene and polystyrene–tetrahydrofuran) and comparing with the experimental values. The model was subsequently used on three polymer–drug systems (paclitaxel–polycaprolactone, hydrocortisone–polyvinylacetate and procaine–polyvinylacetate). The predicted diffusion coefficient for Paclitaxel–Polycaprolactone was used to study the release of Paclitaxel from Polycaprolactone under perfect sink condition. It is envisaged that the proposed model could be used in a reverse engineering framework to select polymers for designing the optimally controlled drug release devices.