Combined Application of Experimental Design and Artificial Neural Networks in Modeling and Characterization of Spray Drying Drug: Cyclodextrin Complexes

Abstract

The aim of this study was to investigate the usefulness of combined application of quality by design tools such as central composite design (CCD), response surface methodology (RSM), and artificial neural networks (ANN) in the characterization, modeling, and optimizaton of spray drying of a poorly soluble drug : cyclodextrin complex. Models were developed by RSM and ANN from different pools of data. The model with best predictability was the ANN multilayer perceptron (MLP)1 model developed from the largest group of data (R-2 for response yield 0.854, moisture content 0.886). On the other hand, analysis of equations derived from the application of RSM contributed in better understanding the complex relationships between input and output variables. By application of a desirability function approach, optimal process parameters that resulted in the best process yield (86%) and minimal moisture content in the powder (3.3%) were established (25% feed concentration, 180 degrees C inlet air temperature, 10% pump speed).