(576e) Utilizing Mechanistic Modelling to Support Crystallization Seeding Design

Digital design of manufacturing processes using mechanistic models is fast becoming an essential tool during Active Pharmaceutical Ingredient (API) early process development activities. It enables rapid and effective exploration of the decision space for Critical Process Parameters (CPP), helping to reduce risk and product time-to-market, and aiding in the effective and safe production of high quality pharmaceutical products. In this work, we demonstrate how mechanistic modelling can be utilised to support the early stage design of seeding strategy for a batch cooling crystallization process. We outline a step-wise workflow consisting of model validation from which primary nucleation and growth kinetics are estimated, as a minimum list of crystallization mechanisms. The data used during the bespoken parameter estimation was experimentally measured in an unseeded small-scale crystallization process.

As part of this work, the model predictions for the seeded crystallization process was subsequently verified experimentally. This model-based approach along with rich experimental data yielded a more efficient workflow for seeding design, with reduced experimental effort compared to a more traditional purely empirical approach. A range of process data was utilized for the purposes of qualitative mechanism discrimination and quantitative model validation as follows:

• Qualitative mechanism & model discrimination
• FBRM data to inform active crystallization mechanisms to consider
• SEM and/or optical images of product crystals to inform crystallization mechanisms to consider within the mechanistic model
• Quantitative model validation
• FBRM data to provide indication of onset of nucleation and access to the rate of nucleation in the unseeded crystallization runs.
• Solute concentration data over time both online (IR/Raman) &/or offline (HPLC)
• Particle size measurements of the product crystals from laser diffraction

The validated mechanistic model was subsequently used for model-based design of seeding policy, to assist in the seeding design (seed mass, seed addition point, and/or seed particle size distribution) required to achieve the target PSD specifications (defined by the volume-based quantiles d10, d50 and d90 of the product PSD). Uncertainty and sensitivity analysis were performed to understand the attainable region for the product PSD in terms of quantiles for the crystallization process and to identify the relative importance of process parameters, including seed mass and seed PSD.

Conclusions

The main conclusions of the work include the following:

• It has been shown that a model-based approach can be utilised to support the early stage development of crystallization processes to achieve the desired quality attributes of the product particles.
• A model-based approach has been shown to help to reduce the degree of experimental effort required compared to a purely empirical approach.