Crystallization is a cornerstone of pharmaceutical manufacturing because it strongly influences product quality attributes.1 However, identifying suitable crystallization conditions through trial and error requires exploring a large design space, substantial time and resources.2,3 This work presents an integrated methodology combining solvent selection, high-throughput screening, and AI-driven analysis to optimize the crystallization of meloxicam (MLX).
A multicriteria solvent screening approach based on Hansen solubility parameters and Environmental, Health, and Safety criteria was applied to identify suitable solvents for MLX crystallization. This analysis revealed dimethyl sulfoxide (DMSO) as the most promising solvent. Experimental validation was performed using a high-throughput platform equipped with turbidity measurements and high-resolution imaging. AI-based image analysis enabled real-time monitoring of morphological changes during crystallization. These results confirmed DMSO as the most suitable solvent, providing a favorable balance between solubility, controllable supersaturation, induction time, and acceptable EHS performance.
Cooling crystallization of MLX in DMSO, however, produced elongated agglomerated needles with broad size distributions and fouling, while the relatively low slope of the solubility curve limited yield. To overcome these limitations, the methodology was extended to antisolvent crystallization. Water was identified as the most suitable antisolvent, and a DMSO:H2O system was investigated experimentally. Temperature cycling improved crystal habit, transforming needle shaped crystals into more compact quasi-equant forms, although some agglomeration remained. The addition of polyvinylpyrrolidone further reduced agglomeration and produced well-defined quasi-equant crystals with a smaller particle size.
Overall, this integrated methodology lays a strong foundation for systematic optimization of crystallization procedures toward safer and more sustainable pharmaceutical manufacturing and provides a scalable framework for robust pharmaceutical process development.
Acknowledgement:
This project was funded by the UKRI (10038378) and ERC (HORIZON-HLTH-2021-IND-07, 101057430).
References
- J. Liu et al., Computer Aided Chemical Engineering, 2021, 50, 1221-1227.
- W. Li et al., Chemical Engineering Research and Design, 2023, 202, 126-146.
- X. Yuan et al., Chemical Engineering Journal Advances. 2025, 23, 100823.
