(410c) Moisture Content Monitoring in Continuous Drug Substance Isolation Manufacturing Platform | AIChE

(410c) Moisture Content Monitoring in Continuous Drug Substance Isolation Manufacturing Platform


Hur, I. - Presenter, Purdue University
Casas Orozco, D., Purdue University
Nagy, Z., Purdue

With the shift of paradigm from batch to continuous in pharmaceutical manufacturing, integrated filtration-drying has become an important emerging technology in continuous drug purification to produce high-purity small molecule oral dosage drugs [1]. Among several alternatives for the intensified filtration unit, the continuous filtration carousel (CFC) has become an efficient technology for handling various production scales following crystallization [2]. The CFC’s main body is a cylindrical carousel, equipped with stations, which are each allocated to one or more steps including filtration, washing, drying, and material transfer and discharge. The port rotates to the next station every predefined cycle time, allowing a sequence of small batch processes to handle continuous separation of solid forms of the active pharmaceutical ingredient (API). By adopting separated small-scale batch process within the continuous manufacturing (CM) horizon time, the extensive knowledge gained previously on batch filtration and drying can be easily adapted in to the integrated framework : 1) their lack of steady states and unknown model parameters in filtration-drying step, 2) dynamic critical material attributes (CMAs), namely feed properties of the slurry coming from the upstream crystallization step.

Quality-by-design (QbD) initiatives have been developed in the last years to overcome the challenges posed by input variability and model parametric uncertainty, and thus expedite the modernization of the pharmaceutical process development in the CM [3]. QbD defines the viable operating region of given sets of CMAs and critical process parameters (CPPs) to fulfill the target quality of critical quality attributes (CQAs). With the increasing availability of process analytical tools (PAT) and digital design tools, QbD has been used in various open-loop studies that create and use process digital twins to optimize process metrics (CQAs, costs, waste generation) with CPPs as decision variables [4, 5]. For the pharmaceutical isolation step, crystal properties and final product impurities of regulated materials are the main CQAs. Since impurities remaining in the solid at the end of the drying step will be directed to the downstream process, failure to reach the drug's targeted purity will lead to off-spec products under the stringent regulatory screenings on the CQAs [6]. Furthermore, the dynamics of crystal properties and final product purity from the simulation are not always in sufficient agreement with real process data. This plant-model mismatch may come from the presence of gross measurement noise, such as outliers and drifts, and the fact that not all model states can be measured, which makes model calibration harder. Improving the estimation and monitoring of the CQAs is critical to incorporate the decision-making framework developed from the design space (DS) into the closed-loop control of the product quality to a higher level of assurance.

We have developed a moving horizon estimation (MHE) framework that relies on state estimation of average moisture content of API filtration cakes, and applied it in the pharmaceutical crystallization-filtration-drying step of a real continuous drug substance isolation line. The MHE framework is based on a first-principles mathematical model describing solid-liquid separation of porous media consisting of crystal feed from an upstream crystallization unit. Since main CQAs in the analyzed case study, namely, cake average moisture and API productivity, cannot be measured in real time, state estimator was critical for their accurate prediction and further model recalibration by the MHE framework. Additionally, offline analysis of the final product CQAs was carried out to validate the proposed MHE workflow. Considering the non-negligible computational time required to solve both the model equations and the MHE problem, a sensitivity analysis was performed to rank uncertain model parameters from the system to moisture content of the final product. The analysis shows that from the steps carried out on the continuous drug isolation operations, filtration and drying model parameters contribute the most to changes in moisture content. Hence, the MHE problem was formulated focusing on the filtration and dryer steps in order to minimize computational time. To correct the model prediction due to the accumulation of the error within the operation timeline, the MHE estimates the moisture content from carousel using a digital twin of continuous crystallizer-CFC and measurements of the inlet and outlet temperature of dryer, pressure drop during the processes, and dry gas outlet composition. From this we estimate the minimum cycle time for filtration and drying to satisfy the target purity constraint, so that the productivity of the system can be improved. The crystallization step is accounted by a calibrated mathematical model, and it enables the inline calculation of the dynamic characteristic cake property capturing various range of filterability and drying dynamics. In this work, the OPC communication protocol was utilized to access to real-time sensor data, and the Python-based, object-oriented platform, PharmaPy [7], has been used to solve the CFC model equations and the MHE problem. The proposed inferential monitoring scheme was evaluated with open-loop and closed-loop experimental data and then integrated into a continuous crystallizer-CFC process. The performance of the proposed monitoring framework was successfully demonstrated by reconstructing the moisture content of the final product by the estimator in the presence of gross measurement errors.


[1] Price, C. J., Barton, A., & Coleman, S. J. (2020). CHAPTER 13. Continuous Isolation of Active Pharmaceutical Ingredients. In The Handbook of Continuous Crystallization. The Royal Society of Chemistry. https://doi.org/10.1039/9781788013581-00469

[2] Destro, F., Hur, I., Wang, V., Abdi, M., Feng, X., Wood, E., ... & Nagy, Z. K. (2021). Mathematical modeling and digital design of an intensified filtration-washing-drying unit for pharmaceutical continuous manufacturing. Chemical Engineering Science, 244, 116803.

[3] Lee, S. L., O’Connor, T. F., Yang, X., Cruz, C. N., Chatterjee, S., Madurawe, R. D., Moore, C. M. V., Yu, L. X., & Woodcock, J. (2015). Modernizing Pharmaceutical Manufacturing: from Batch to Continuous Production. Journal of Pharmaceutical Innovation, 10(3), 191–199. https://doi.org/10.1007/s12247-015-9215-8

[4] D. Casas-Orozco et al., “Digital Design of a Lomustine Manufacturing Process Using Pharmapy”, in AICHE Annual Meeting, 2021

[5] Su, Q., Bommireddy, Y., Shah, Y., Ganesh, S., Moreno, M., Liu, J., Gonzalez, M., Yazdanpanah, N., & Connor, T. O. (2019). Data reconciliation in the Quality-by-Design ( QbD ) implementation of pharmaceutical continuous tablet manufacturing. International Journal of Pharmaceutics, 563(April), 259–272. https://doi.org/10.1016/j.ijpharm.2019.04.003

[6] International Council for Harmonisation, 2016. Impurities: guideline for residual solvents Q3C(R6)

[7] Casas-Orozco, D., Laky, D., Wang, V., Abdi, M., Feng, X., Wood, E., Laird, C., Reklaitis, G. V., & Nagy, Z. K. (2021). PharmaPy: An object-oriented tool for the development of hybrid pharmaceutical flowsheets. Computers and Chemical Engineering, 153, 107408. https://doi.org/10.1016/j.compchemeng.2021.107408