Implementing process analytical technology (PAT) tools for the analysis and control of manufacturing processes during hot-melt extrusion

In the research reported in this thesis, hot-melt extrusion (HME) was coupled with process analytical technology (PAT) tools in order to investigate drug interactions during processing, identify drug/excipient interactions and monitor the quality of the extruded products. In order to ensure the quality of the products, X-ray powder diffraction (XRPD), differential scanning calorimetry (DSC), near infrared (NIR) spectroscopy, variable temperature x-ray powder diffraction (VTXRPD) and dissolution studies were implemented in various drug delivery systems manufactured by HME. Paracetamol (PMOL) was processed with ethyl cellulose (EC) - Compritol 888 ATO (C888) polymer/lipid blends to produce sustained release drug formulations. The aim of this study was to optimize HME processing by using design of experiment (DoE) and to monitor the process using in-line NIR spectroscopy. Input variables such as drug loading, screw speed and feeding rate were investigated and in-line NIR spectra were collected during extrusion. A partial least squares calibration model and PCA plots were used to identify the optimum processing parameters. It was proved that drug loading has a significant effect on the final product while the PMOL dissolution rates were identified as a quality attribute for the development of sustained release formulations. Extruded pharmaceutical cocrystals were manufactured by using carbamazepine (CBZ) as a model drug with saccharine (SCH) and trans-cinnamic acid (TCA) as coformers at a 1:1 molar ratio. In-line NIR was employed to monitor the cocrystallization process along the HME barrel while off-line DSC and XRPD were used to characterise the extrudates. NIR analysis showed that cocrystal formation occurred gradually starting at the 2nd mixing zone and this was confirmed via DSC and XRPD analysis. Cocrystallization occurred via the formation of H-bonds between the drug and coformers facilitating enhanced drug release rates compared to the prototype cocrystals produced via solvent evaporation. Three sugar alcohols namely mannitol, sorbitol and xylitol were extruded at various weight ratios to produce sugar cocrystals. In-line NIR, employed for process monitoring, showed that cocrystal formation occurred by the formation of H-bonds. The extruded materials were compressed into orally disintegrating tablets and compared with tablets made using Starlac (as a binder) and the sugar alcohols. Characterization of the tablets showed that there is no significant difference - in terms of weight variation, thickness and hardness - of the tablets prepared from both extruded sugar cocrystals and the Starlac. However, tablets prepared from the extruded sugar cocrystals demonstrated faster disintegration times and better friability compared with those compressed with Starlac. In another study reflectance and transmission in-line NIR probes were coupled to HME to monitor the transformations of crystalline indomethacin (IND) to molecular solutions in the presence of Soluplus and Kollidon VA64 polymers. The analysis of the transmission and reflectance NIR data revealed that crystalline IND was completely transformed into molecular solutions while Raman analysis showed that it was homogeneously distributed in the polymer matrices. PCA analysis showed that the screw speed, during HME, affects the recorded spectra but not the homogeneity of the embedded drug in the films. Enhanced drug release, compared with bulk IND, was also attained using ground, hot-melt extruded films. VTXRPD was used as a predictive tool to monitor the polymorphic behaviour of physical blends of PMOL and Soluplus polymer in comparison to HME processed blends. By using VTXRPD it was found that at temperatures above 120°C Form I PMOL transforms to Form II and significant 2-theta value shifts were detected using VTXRPD. The in-line NIR analysis confirmed that the polymorphic transformations taking place during HME processing were identical to those observed in VTXPRD. Thus, VTXRPD can be used as a predictive tool for HME processing