Search

The pharmaceutical industry is currently implementing new manufacturing principles and modernizing the related processing solutions. A key element in this development is implementation of process analytical technologies (PAT) for measuring product quality in a real-time mode, ideally for a continuously operating processing line. Near-infrared (NIR) spectroscopy is widely used for this purpose, but has limited use for low concentration formulations, due to its inherent detection limit. Light-induced fluorescence (LIF) spectroscopy is a PAT tool that can be used to quantify low concentrations of active pharmaceutical ingredient, and recent development of instrumentation has made it available for in-line applications. In this study, the content of tryptophan in a dynamic powder flow could be measured as low as 0.10 w/w % with LIF spectroscopy with good accuracy of RMSEP = 0.008 w/w %. Both partial least squares regression and support vector machines (SVM) were investigated, but we found SVM to be the better option due to non-linearities between the calibration test and the in-line measurements. With the use of SVM, LIF spectroscopy is a promising candidate for low concentration applications where NIR is not suitable.

There is a need for an improved process understanding of solid dosage pharmaceuticals. In the present study, Raman spectroscopy together with partial least squares (PLS) regression was used to monitor the solid-state composition of piroxicam during processing in the presence of excipients. It was found that including variable selection in PLS regression offers improved quantitative models in terms of predictive performance, easier interpretation of results and reduced experimental workload relative to full-spectrum PLS regression. By means of the applied interval PLS (iPLS) regression model, it was observed that excipients with high water-absorbing potential (microcrystalline cellulose and hydroxypropyl cellulose) and a water-activity-reducing solvent (ethanol) delayed the onset of monohydrate formation during processing in aqueous environments. An alkalizing excipient (sodium bicarbonate) decreased the onset time of monohydrate formation during wet granulation and decreased the dehydration rate during a drying operation. In this study it is demonstrated that the physical stability of hydrate–anhydrate systems in process environments is complicated by a multitude of factors on both macroscopic level and molecular level, and that variable selection for PLS regression is a valuable tool for screening the effects of excipients on the solid-state properties of pharmaceuticals.