Building Artificial Neural Networks for the Optimization of Sustained-Release Kinetics of Metronidazole from Colonic Hydrophilic Matrices

Abstract

Introduction: Drug development has traditionally used mathematical models to predict formulation behavior. Objective: Building artificial neural networks for the drug release evaluation of drug delivery systems using sustained-release metronidazole-coated colonic hydrophilic matrices as a model. Methods: The technological factors associated with the biopharmaceutical performance of hydrophilic metronidazole matrices were evaluated using a quality by design approach (QbD). The developed neural network includes variables related to the technological process for producing the matrices. These are related to the materials used, such as the type and viscosity of core polymers, the type of coating agent, or the matrix production process, such as the mixing time of core materials or the percentage of the coating agent. The output variables of the neural network were the percentages of drug released in vitro at 1, 6, 12, and 24 h and the mean dissolution time of the matrix. An iterative quasi-Newton method was used to train the artificial neural network. Results: A neural network with excellent prediction capacity allows selecting the technological variables with the greatest influence on the % of drug dissolved: the type of coating agent used and the percentage of the total weight increase after coating for 1 h and 6 h of drug release and also the viscosity of the HPMC for 12 and 24 h. Conclusions: The optimized neural network demonstrated an excellent predictive capacity for in vitro drug dissolution profiles, allowing the use of this type of methodology based on artificial intelligence methods in the optimization of drug delivery systems.