(676b) Characterization and Modelling Reactive Protein Crystallization: Defining the End of a Process

This is a case study on how PAT’s can be used for enhanced process understanding to improve a crystallization process and thus increase capacity in pharmaceutical manufacturing. The emergence and improvement of new PAT’s enables insights into crystallization processes, that were not previously possible. Therefore, these technologies can be used to characterize and optimize crystallization processes that were introduced into manufacturing decades ago.

The studied crystallization process is a reactive crystallization used in the formulation of a drug product, where x% of the API is crystallized and (100-x)% of the API is dissolved. Upon formulation some of the API precipitates into amorphous particles from which the crystals grow. It had been assumed that the concentration of both API and preservatives in the dissolved phase stayed constant throughout the process. After crystallization the product is filled.

This study consists of three parts:
The first part is calibration and validation of a single in-line PAT probe to predict critical quality attributes (CQA) such as amount of API crystallinity, API and preservative concentration in solution and particle size distribution. The probe uses three different measurement techniques: Raman spectroscopy, high-dynamic range turbidity and advanced chord-length-distributions. The calibrations are performed with chemometrics with reference values from off-line HPLC for API and preservative concentration, on-line small-angle x-ray scattering for degree of crystallinity and automated microscopy for particle size distribution. Segregation of the crystal size distribution from the particle size distribution is difficult, as amorphous particles and small crystals can be difficult to distinguish due to similar aspect ratios and size. However, modelling using the crystallinity amount from the orthogonal experimental techniques will improve this segregation.

The second part explore the crystallization dynamics during the current formulation process. The data showed that the concentration of preservatives in the dissolved phase increased until the formation of the first crystals. Furthermore, the API concentration in the dissolved phase also increased during crystallization then reached a plateau after the crystallization start, but before the end of the conversion of amorphous to crystals. The particle size distribution shows the different phases in the crystallization process, but also that the crystals keep growing beyond the end of conversion of amorphous to crystals. This is most likely due to Ostwald Ripening. Furthermore, cryogenic transmission electron microscopy images were taken at critical timepoints for qualitative assessment of the particles at very high resolution. The data indicates that different subprocesses within the process have different endpoints and therefore raises the question of whether the process could be defined as finished earlier without affecting the CQA’s. In extension of this another interesting question arises: “what happens if the product is filled if some processes, but not all, are finished?”.

Thus, the third part investigates the effect of the filling process on the CQA’s, as the product requires constant recirculation, which exerts a large amount of shear stress on the product. Shear stress can affect the crystallization process, and future experiments will investigate if the CQA’s are affected when the drug product is filled even though one or several sub-processes have not ended yet. Preliminary results indicate that the amount of API crystallinity is increased while decreasing API concentration in the dissolved phase when adding shear stress. Furthermore, the aspect ratio of the particles increases as well. However, the particle size distribution does not appear to be affected if the shear stress is only applied before the onset of crystallization growth.

Furthermore, hybrid predictive modelling of the processes will be used to understand differences in nucleation, growth, agglomeration and breakage phenomena when applying shear stress during the different phases of crystallization. Also, the model can potentially be used for in silico development of crystallization process control strategies during manufacturing.