Solution crystallization is a solvent-intensive separation and purification technique that is commonly utilized in the manufacturing processes of high-value-added chemicals such as active pharmaceutical ingredients (APIs). APIs can either be crystallized in their pure forms or with other pharmaceutically acceptable compounds. Model-based optimization tools are becoming increasingly common as alternatives for trial-and-error-based experimentation in the crystallization process and product design. However, important challenges limiting their full adoption still exist. Firstly, the use of model-based optimization tools may not have reached the full potential in the industry due to the complexity of their implementation. Secondly, it is often desirable to identify solvents and operating con...[ Read more ]

Solution crystallization is a solvent-intensive separation and purification technique that is commonly utilized in the manufacturing processes of high-value-added chemicals such as active pharmaceutical ingredients (APIs). APIs can either be crystallized in their pure forms or with other pharmaceutically acceptable compounds. Model-based optimization tools are becoming increasingly common as alternatives for trial-and-error-based experimentation in the crystallization process and product design. However, important challenges limiting their full adoption still exist. Firstly, the use of model-based optimization tools may not have reached the full potential in the industry due to the complexity of their implementation. Secondly, it is often desirable to identify solvents and operating conditions that work well for a whole process, particularly in view of the industry trend toward continuous manufacturing. However, such approaches suitable for many commonly occurring process types are not currently available. Thirdly, the existing approaches commonly treat process and product design as separate problems even though process and product performance are often closely connected. Fourthly, dynamic models to describe cocrystal dissolution in the presence of API precipitation are lacking, which makes the predictive simulation of their product performance unattainable. The work presented in this thesis develops novel modeling and optimization tools to address those challenges.

In Chapter 2 of this thesis, a generalized workflow is developed for equilibrium-based, simultaneous solvent selection and process optimization for stand-alone solution crystallization processes. This workflow is based on the perturbed-chain statistical associating fluid theory (PC-SAFT) equation-of-state (EoS), which is a commonly utilized thermodynamic model for crystallization solvent selection and process optimization. This workflow provides a step-by-step guide that can be executed with minimal expertise to identify optimal solvent candidates and operating conditions for the common solution crystallization methods. Each step of the workflow is presented with readily executable computational tools so that the user does not have to commit substantial resources in programming. The use of the workflow is demonstrated through an industrially relevant case study, which shows that the workflow allows rapid screening of different crystallization methods, solvents, and operating conditions for an application.

The reaction-extraction-crystallization is a commonly occurring solvent-intensive unit operation sequence in pharmaceutical processes. Simultaneous solvent selection and process optimization considering the whole process rather than individual unit operations is essential to account for various trade-offs arising among unit operations accurately, which have not been developed yet. In Chapter 3, a new PC-SAFT-based optimization framework is developed for the equilibrium-based, simultaneous solvent selection and process optimization for the reaction-extraction-crystallization sequence with possible solvent recycling. The application of the developed optimization framework is illustrated using a case study involving a continuous manufacturing process of an API. The optimization problem can be solved successfully with a mixed-integer nonlinear programming (MINLP) relaxation strategy, followed by either continuous mapping or a branch-and-bound approach for solvent identification. The computational tractability of the proposed computational framework indicates the good potential for applications to industrially relevant cases featuring similar thermodynamic equilibria and complexity.

In pharmaceutical cocrystallizing systems, where an API is crystallized simultaneously with a coformer to form a cocrystal, process and product performance are highly interrelated. The selected coformer, solvents, and operating conditions should facilitate the selective crystallization of the pure cocrystal, while the coformer type may need to be tailored for enhanced bioavailability of the final product by improving its dissolution behavior. Currently, computational approaches that can capture the trade-offs arising due to these competing process and product performance criteria do not exist. Therefore, in Chapter 4, a new model-based optimization framework that allows the identification of coformers, solvent types, and operating conditions for cocrystallization is developed, which simultaneously optimizes the process efficiency and product dissolution behavior. The crystallization process is described by an equilibrium model based on the PC-SAFT, and the API dissolution behavior is evaluated using a dynamic model that describes cocrystal dissolution in a standard dissolution testing apparatus. The developed optimization framework is illustrated with a case study involving a model API whose bioavailability can be enhanced by a cocrystal-based formulation. The proposed solution strategy, which is a combination of branch-and-bound and continuous mapping, can successfully solve the optimization problem to identify coformers, solvents, and operating conditions when considering process and product performance simultaneously.

Cocrystal dissolution often provokes API precipitation due to supersaturation generation, either in the bulk solution or at the cocrystal surface, which is generally undesirable. The design of cocrystal-based drug delivery systems would benefit from improved predictive dissolution models that can account for such surface and bulk precipitation, which are currently lacking. Therefore, in Chapter 5, a predictive in vitro model for cocrystal dissolution considering simultaneous surface and bulk precipitation is developed and validated with experimental data. The dissolution medium is compartmentalized based on the API concentration. Three crystal populations are considered to exist, i.e., the cocrystals, API crystals at the cocrystal surface, and API crystals in the bulk medium. The developed model demonstrates a good capability to capture the experimental trends of cocrystal dissolution. The model also makes predictions in good agreement with the experimental data. Finally, integration of the validated model with numerical optimization tools is achieved to identify optimal formulation conditions in terms of dose, particle size, and pre-dissolved coformer concentration, which could benefit the pharmaceutical industry through reduced trial-and-error experimentation for cocrystal formulation development.