ARTIFICIAL NEURAL-NETWORK ANALYSIS OF A DIRECT COMPRESSION TABLETTING STUDY

Abstract

In this study, effects of lubricant type, tablet compression pressure, and the duration of mixing with the lubricants on drug release rates, and the tablet crushing strength were investigated. The artificial neural network (ANN) and the polynomial regression methods (PRM) were used for data analysis. Magnesium stearate (MS), glyceryl behenate (GB) (Compritol 888(R)), and MS-talc blend (BL) were used at 1% level in a direct compression formula containing hydrochlorothiazide (HCT) as a model drug. For ANN analysis, different input-hidden layer-output types were tried to determine the best models. For PRM, the quadratic model was selected. Both the ANN and the PRM models were found to be successful to characterize the process studied, and provided comparable results. Based on the model, it was found that when the critical mixing time of 10 min was exceeded, the mixing time became the predominant factor on tablet crushing strength with MS and the BL. Extended mixing with MS lowered the tablet crushing strength to a very low level. Tablets containing GB were not affected drastically by the mixing time, compression pressure played a more significant role in their case. When BL was used as a lubricant, tablets had significantly higher crushing strength values than the MS. ANN was found to be a flexible and accurate method to study process and formulation factors in pharmaceutical technology and can be used instead of traditional regression method in the cases of involving discrete variables and non-linear systems. Manipulating the models, some optimizations were attempted to maximize tablet crushing strength or to select the best lubricant type.