(539a) Scientific Considerations on Continuous Crystallization and Its PAT Method Validation
- Conference: AIChE Annual Meeting
- Year: 2017
- Proceeding: 2017 AIChE Annual Meeting
- Group: Pharmaceutical Discovery, Development and Manufacturing Forum
Wednesday, November 1, 2017 - 12:30pm-12:52pm
Continuous manufacturing (CM) is an emerging technology in the pharmaceutical manufacturing sector. As the final purification/isolation step, crystallization can significantly affect the drug substance critical quality attributes (CQAs) and is one of the critical steps in an integrated continuous manufacturing (CM) process. Continuous crystallization has the potential to improve manufacturing efficiency and product quality, but both industry and FDA currently lack commercial experience with the technology. Here, we focus on the application of quality risk management to the development of a continuous crystallization system. Specifically, we designed and built an automated two-stage Mixed Suspension Mixed Product Removal (MSMPR) crystallization platform for a model compound (Carbamazepine, CBZ) that exhibits multiple polymorphs. The crystallization process includes integration of Process Analytical Technology (PAT) tools (on-line Raman microscopy and Focused Beam Reflectance Microscopy, FBRM) for real-time monitoring. Material transport, process control, and product collect approaches were investigated as the three key areas for risk identification and mitigation during continuous crystallization process development. Our proof of concept continuous crystallization system uses feedback controls to achieve constant levels in crystallizers, a centralized automation program coded in LabView, and PAT monitoring for polymorphs and particle size distribution (Raman and FBRM). We developed a PAT method utilizing Raman Spectroscopy to monitor solute concentration and polymorphs of CBZ. A calibration model was developed and validated; and the limit of detection for a metastable polymorphic form was quantified. The impact of multiple factors, such as solid content, and temperature, on the model prediction was studied. The model was validated following the principles described in USP <1225>. The results demonstrated that a linear model can predict the solute concentration with less than 10 percent error. The repeatability and intermediate precision was evaluated and the relative standard deviation was below 5 percent. The limit of detection and quantification were found to be 3.8 and 7.4 mg/ml, respectively. The limit of detection for the metastable form was determined by monitoring the ratio of characteristic peaks when increasing the percentage of the metastable form in relation to the total amount of crystals in the solution. A statistically significant change in the peak ratio was observed after the concentration of the metastable form reached 25 percent.