Understanding and Predicting Pharmaceutical Excipient Variability and Performance Using Spectroscopic and Multivariate Analysis Approaches

Abstract

The overall objective of this study is to investigate whether raw material near-infrared (NIR) spectra can be used to understand and predict the excipient performance. The first aim is to investigate the extent that NIR can be used to capture excipient physiochemical properties. NIR spectra of microcrystalline cellulose (MCC) sourced from multiple manufacturers were obtained. Partial least square discriminate analysis (PLSDA) was used to build classification models of MCC manufacturers. The model prediction of manufacturer was significant. The classification can be attributable to the differences in the content of oxidized cellulose groups, water content and states, hydrogen bonding, and degree of polymerization of MCC. Second aim is to understand excipient variability and the impacts of their variability on performance in a multivariate manner. Thirteen magnesium stearate (MgSt) samples from multiple lots, grades and manufacturers were obtained and extensively characterized in micrometrics, particle size distribution, specific surface area, thermal properties, crystalline structure, NIR and Raman spectra. The excipient performances were assessed in a tablet direct compression process. Multivariate modeling method was applied to correlate excipient physiochemical properties, key formulation and processing variables to excipient performance. The results showed that MgSt property variations within grades are smaller than that of grades and manufacturers. Physiochemical properties of MgSt are manufacturer dependent. Excipient performance can be modeled and predicted from raw material characterizations. The prediction depends on formulation factors and performance of interest. The effects of lubricant properties on lubrication sensitivity to prolonged mixing were assessed. It was found that the particle morphology was responsible for the lubrication sensitivity. The third aim of the study is to test if NIR and Raman spectra can be used as a surrogate tool for excipient characterization and performance prediction. The correlation between excipient performance and spectra has been examined. It was shown that through capturing key physical and chemical properties of materials, NIR spectroscopy can be used as an alternative tool to predict excipient performance in some cases. The prediction of excipient performance through Raman spectra was not successful. These findings demonstrated the application of NIR in conjunction with multivariate models as an effective and promising tool to understand raw material variability and predict pharmaceutical excipient performance.