In-line solid state prediction during pharmaceutical hot-melt extrusion in a 12 mm twin screw extruder using Raman spectroscopy
Lien Saerens,Dima Ghanam,Cedric Raemdonck,Kjell Francois,Jürgen Manz,Rainer Krüger,Susan Krüger,Chris Vervaet,Jean Paul Remon,Thomas De Beer
DOI: https://doi.org/10.1016/j.ejpb.2014.03.002
Abstract:The aim of this research was to use Raman spectroscopy for the in-line monitoring of the solid state of materials during pharmaceutical hot-melt extrusion in the die head of a 12 mm (development scale) twin-screw extruder during formulation development. A full factorial (mixed) design was generated to determine the influence of variations in concentration of Celecoxib (CEL) in Eudragit® E PO, three different screw configurations and varying barrel temperature profiles on the solid state, 'melt temperature' and die pressure of continuously produced extrudates in real-time. Off-line XRD and DSC analysis were used to evaluate the suitability of Raman spectroscopy for solid state predictions. First, principal component analysis (PCA) was performed on all in-line collected Raman spectra from the experimental design. The resulting PC 1 versus PC 2 scores plot showed clustering according to solid state of the extrudates, and two classes, one class where crystalline CEL is still present and a second class where no crystalline CEL was detected, were found. Then, a soft independent modelling of class analogy (SIMCA) model was developed, by modelling these two classes separately by disjoint PCA models. These two separate PCA models were then used for the classification of new produced extrudates and allowed distinction between glassy solid solutions of CEL and crystalline dispersions of CEL. All extrudates were classified similarly by Raman spectroscopy, XRD and DSC measurements, with exception of the extrudates with a 30% CEL concentration extruded at 130 °C. The Raman spectra of these experiments showed bands which were sharper than the amorphous spectra, but broader than the crystalline spectra, indicating the presence of CEL that has dissolved into the matrix and CEL in its crystalline state. XRD and DSC measurements did not detect this. Modifications in the screw configuration did not affect the solid state and did not have an effect on the solid state prediction of new produced extrudates. Secondly, the influence of variations in die pressure on the Raman spectra was examined. The applied drug concentration, processing temperature and feeder performance influence the die pressure, which is reflected in the Raman spectra as a change in spectral intensity. When applying PCA on the raw spectra from the experimental design, the first principal component describes the influence of die pressure on the spectra, which was seen as a decrease in Raman intensity of the whole spectrum when the pressure in the sample increased. Clustering according to processing temperature was found, although the temperature in the die remained constant, indicating that a difference in viscosity, resulting in a changed die pressure, was detected. When the feeder was stopped, the score values of the first principal component almost simultaneously decreased, and only stabilized once the die pressure became stable. Since Raman spectra collected in the extrusion die are influenced by changes in die pressure, disturbances upstream of the extrusion process can be observed and identified in the Raman measurements.