(600f) Development of a Hybrid Quality-By-Control Framework for Optimal Crystallization Process Design

Strict regulatory standards are dictated by the FDA, EMEA and WHO for the active pharmaceutical ingredients (API) to meet high level quality criteria such as purity, size and shape of the API crystals or polymorphic structure. One of the main criteria is to achieve size uniformity as well as desired yield in industry. Despite of the efforts put in pharmaceutical industry, intensive fines formation and agglomeration of crystals are still common problems, often having a negative impact on the product crystal size distribution (CSD). A general practice to eliminate fines and agglomerates is to apply temperature cycles for internal fines dissolution and de-agglomeration.¹ For this purpose, quality-by-design (QbD) approaches can be applied for the design of the suitable temperature profile, but this requires intensive experimental investigation to explore the whole design space. Quality-by-Control (QbC) has been introduced more recently as a new novel alternative to QbD, which can be applied to control critical quality attributes of a certain process, whereby the desired operating curve like temperature profile is automatically determined by the feedback control algorithm rather than by excessive open-loop experimentation. In this way QbC can significantly reduce required process development time and material used.²

The work proposes a novel hybrid framework combining the model-free and model-based quality-by-control (mfQbC and mbQbC) paradigms for the optimal crystallization design with the aim of minimization of fines and agglomerates in the product. This developed framework describes the mfQbC based rapid design used in conjunction with direct nucleation control (DNC) and supersaturation control (SSC), along with (2) the mbQbC framework using a novel population balance model-based platform for the robust optimization of the operating trajectory of the crystallizer to achieve crystals with desired quality attributes.³ This framework provides an improved alternative to the current industrial practice, by the combined use of feedback control and mathematical models for rapid design of robust crystallization processes. The implementation of the proposed framework enhanced the prediction of crystallization events and was able to achieve better control of the crystallization processes by providing controlled temperature cycles, for de-agglomeration and fines destruction. The information for parameter estimation and optimization calculations was collected from the experiments and the system diagnostics was done in mfQbC part, which also provided a rapid design of robust operating conditions, that were improved to achieve optimal operation subsequently in the mbQbC approach.

The advantage of the mfQbC is that it requires minimal preliminary system information, hence, it enables quick and efficient crystallization process design. On the other hand, the mf-QbC operation policy might be sub-optimal, which can be overcome by exploiting the predictive capabilities of the process models used by the mbQbC. In the mfQbC part, DNC experiments were run for two purposes: to determine the primary and secondary nucleation metastable zone widths (MSZW) and to obtain the temperature cycle profile that is required for producing the crystals with desired properties. DNC is an efficient approach to eliminate fines due to the application of temperature cycles. SSC was applied after DNC generates the MSZW information to obtain the temperature profile for only growth mechanism. The resulting temperature profile from SSC can also be easily implemented in industrial distributed control systems (DSCs). After mfQbC, using the mbQbC framework, a population balance model (PBM) was developed to simulate primary and secondary nucleation, growth, agglomeration and de-agglomeration mechanisms as well as dissolution. Model parameters were identified using experimental data for a model API crystallization with high agglomeration and secondary nucleation tendency. Using a model-based robust optimization, in which parametric uncertainty obtained from the model-identification step was also included, different temperature profiles were determined considering the dual objective of fine minimization and maximal de-agglomeration. The optimal temperature profiles obtained was validated experimentally to show the higher quality product that can be obtained using the this QbC framework.

References:

1. Wu, Z., Yang, S. & Wu, W. Application of temperature cycling for crystal quality control during crystallization. CrystEngComm 18, 2222–2238 (2016).
2. Yang, Y., Song, L. & Nagy, Z. K. Automated Direct Nucleation Control in Continuous Mixed Suspension Mixed Product Removal Cooling Crystallization. Cryst. Growth Des. 15, 5839–5848 (2015).
3. Szilágyi, B., Borsos, Á., Pal, K. & Nagy, Z. K. Experimental implementation of a Quality-by-Control (QbC) framework using a mechanistic PBM-based nonlinear model predictive control involving chord length distribution measurement for the batch cooling crystallization of L-ascorbic acid. Chem. Eng. Sci. 195, 335–346 (2019).