(58e) Impurity Purging through Systematic Process Development of a Continuous Two-Stage Crystallization | AIChE

(58e) Impurity Purging through Systematic Process Development of a Continuous Two-Stage Crystallization

Continuous downstream processing of pharmaceutical products is often conducted in parallel with upstream development and optimization, leading to scarce and ever-changing input material. In order to efficiently make progress, meet deliverable deadlines, and match FDA requirements for API analysis, a systematic development plan must be imposed. Ciprofloxacin is included in the essential lists of medicines, and is the drug used to validate the POD (Pharmacy on Demand) system discussed in this report. Purification in this system is performed in a two-stage continuous MSMPR set-up with a first-stage pH-mediated precipitation, followed by a traditional seeded, anti-solvent crystallization second stage. Development of both stages was vastly conducted in batch to conserve material, while flow experiments were used for validation. OFAT and DOE approaches were used throughout to optimize the impurity purging. During process development a large change in input quality was achieved by switching a reagent in the final step of synthesis, leading to a different and improved impurity profile. To re-assess the purging ability of our system, more OFATs were conducted across temperature, solvent ratios, pH, and flow rates, to help re-optimize with the changed input across both stages. Upon optimization, it was confirmed that FDA specifications (0.07 % area by LC or less per impurity, 98-102% by mass assay) could be met. In finding this, one individual impurity remained at all close to the threshold and a majority of its purging was achieved in the first stage. To assess purging power of this stage, pH and temperature were varied as OFATs and in a 2 factor DOE and solid and liquid composition of API and this impurity were assessed. This impurity was found to have a linear relationship between purging and temperature that could be modeled. This linear model was then tested in flow to show errors of less then 40% consistently in predicting the purging levels of this individual impurity. By general process optimization this system was narrowed to necessitate control of a single impurity, which was found to be controlled quantitatively by temperature, ensuring product quality every time.