Although various quality by design (QbD) approaches have been used to establish a design space to obtain robust drug formulation and process parameters, the effect of excipient variability on the design space and drug product quality is unclear. In this study, the effect of microcrystalline cellulose (MCC) variability on drug product quality was examined using a design space for immediate-release tablets of amlodipine besylate. MCC variability was assessed by altering the manufacturer and grade. The formulation was developed by employing the QbD approach, which was optimized using a D-optimal mixture design. Using 36 different MCCs, the effect of MCC variability on the design space was assessed. The design space was shifted by different manufacturers and grades of MCC, which resulted in associations between the physicochemical properties of MCC and critical quality attributes (CQAs). The correlation between the physicochemical properties of MCCs and CQAs was assessed through a statistical analysis. A predictive model correlating the physicochemical properties of MCCs with dissolution was established using an artificial neural network (ANN). The ANN model accurately predicted dissolution with low absolute and relative errors. The present study described a comprehensive QbD approach, statistical analysis, and ANN to comprehend and manage the effect of excipient variability on the design space.
Keywords: artificial neural networks; critical quality attributes; design space; excipient variability; microcrystalline cellulose; quality by design.