Predicting Drug Release Rate of Implantable Matrices and Better Understanding of the Underlying Mechanisms through Experimental Design and Artificial Neural Network-Based Modelling.

There is a growing interest in implantable drug delivery systems (DDS) in pharmaceutical science. The aim of the present study is to investigate whether it is possible to customize drug release from implantable DDSs through drug-carrier interactions. Therefore, a series of chemically similar active ingredients (APIs) was mixed with different matrix-forming materials and was then compressed directly. Compression and dissolution interactions were examined by FT-IR spectroscopy. Regarding the effect of the interactions on drug release kinetics, a custom-made dissolution device designed for implantable systems was used. The data obtained were used to construct models based on artificial neural networks (ANNs) to predict drug dissolution. FT-IR studies confirmed the presence of H-bond-based solid-state interactions that intensified during dissolution. These results confirmed our hypothesis that interactions could significantly affect both the release rate and the amount of the released drug. The efficiencies of the kinetic parameter-based and point-to-point ANN models were also compared, where the results showed that the point-to-point models better handled predictive inaccuracies and provided better overall predictive efficiency.