(146d) Development of an Immersion Mill Integrated Crystallization Process Model As Digital Twin for in-Silico Process Optimization

Control of crystal size distribution (CSD) and shape is critical in the pharmaceutical industry for meeting tight critical quality attribute (CQA) requirements in the manufacturing of active pharmaceutical ingredients (APIs). In general, to increase the efficiency of downstream operations such as filtration and drying, and increase flowability and manufacturability of powders, large crystals with low aspect ratio (AR) are preferred.¹ A general method to enhance simultaneous control of CSD and AR is to use wet-milling in conjunction with the crystallization process. To find the process variables that will give the desired product with CQAs might require extensive experimentation and long time, especially with the increasing number of decision variables by introducing a new component such as wet-mill. To overcome this issue and have a methodological process design based on in-silico experiments and model-based optimization, along with the current adoption of the emerging Industry 4.0 practices, digital twins are increasingly used in the pharmaceutical industry.

In this work, we present the development of a digital twin for the immersion-mill integrated crystallization process of a pharmaceutical API (Compound A), from Takeda Pharmaceuticals International Co. Compound A forms high AR crystals and has size dependent growth. The objectives of the crystallization design were to produce product crystals with low AR (< 3.5) and with bulk density values higher than 0.25 g/ml. Implementing only temperature cycles was shown to be insufficient to achieve these design objectives.² For this reason, the integration of milling was necessary, which considerably increased the number of critical process parameters. A digital twin of the process was developed for the integrated process and used to optimize the process parameters such as milling time or number of temperature cycles.³ Several model training experiments were done with different numbers of temperature cycles and different milling start times to have a broad spectrum of data and product properties. For milling a T25 digital ULTRA-TURRAX from IKA Works, Inc. immersion mill was used. For data acquisition and system monitoring during the experiments, process analytical technology tools (PAT) were used, including Mettler Toledo’s ParticleTrack G400 for in-line chord length distribution (CLD), for detecting the crystallization events such as nucleation and dissolution, as well as for the crystal count measurement, the Zeiss MCS621 ATR-UV/Vis spectrophotometer for concentration measurement, and the Mettler Toledo’s ParticleView v19 for in-situ microscopy images. Additionally, Malvern’s Mastersizer 3000 Hydro MV was used as an off-line characterization tool to measure crystal size distribution (CSD) and the data from the PAT tools were directly used in the parameter estimation step.

The process model was developed by including secondary nucleation, size dependent growth (SDG), size dependent dissolution (SDD), and size dependent breakage mechanisms. A novel SDG rate expression was used in the model to capture the CSD dynamics considerably better than the standard SDG rate models. Size dependency of the breakage was demonstrated by conducting several breakage-only experiments by using the same set-up also to find the smallest attainable size for the breakage by immersion-mill to be used in the breakage formulation. The kinetic parameter estimation of the digital twin was done by minimizing the difference between the experimental and simulated concentration profiles and maximizing the correlation between the simulated crystal number density and measured FBRM counts, while also minimizing the difference between the predicted CSDs and measured CSDs of the samples taken at the sampling times.

The prediction accuracy of the developed digital twin is tested by validation experiments which are designed separately. After validating the model, it is used for in-silico optimization to find the optimum milling time, periods when the mill has to be on or off, and the number of temperature cycles and features of the temperature profile, with the aim of maximizing the product mean size while minimizing the AR. The optimized operating conditions are validated experimentally to demonstrate the benefits of the digital design approach of the immersion-mill integrated crystallization process.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported.

References:

1. Horio, T., Yasuda, M. & Matsusaka, S. Effect of particle shape on powder flowability of microcrystalline cellulose as determined using the vibration shear tube method. Int. J. Pharm. 473, 572–578 (2014).
2. Eisenschmidt, H., Bajcinca, N. & Sundmacher, K. Optimal Control of Crystal Shapes in Batch Crystallization Experiments by Growth-Dissolution Cycles. Cryst. Growth Des. 16, 3297–3306 (2016).
3. Szilagyi, B. & Nagy, Z. K. Model-based analysis and quality-by-design framework for high aspect ratio crystals in crystallizer-wet mill systems using GPU acceleration enabled optimization. Comput. Chem. Eng. 126, 421–433 (2019).