(82c) Predictive CFD Mixing Model for Scale up/Down of Large Molecule Dug Product Formulation

Large molecule drug product formulation typically requires mixing of multiple materials to form a consistent and uniform drug product before further processing. One common example is the mixing of multiple preformulated liquid streams to homogeneity in a mixing vessel. High viscosity products with high protein concentration and/or high excipient concentrations such as stabilizers represent a mixing challenge. Experiments at full scale manufacturing to determine operating ranges are costly and time consuming. Vessel geometries are also not always kept similar upon scale up or technical transfer of a process, which complicates design of scale down studies to determine suitable mixing parameters. CFD models have shown promise in delivering predictive tools [1] to shorten process development workflow and eliminate at scale experimental runs. However, there are fewer published reports in verifying these models over a wide range of tank scales and shapes, while quantifying accuracy and uncertainty.

A CFD model was built in MStar CFD to predict mixing time to homogeneity for multiple liquid streams. CFD predictions were verified against surrogate data at multiple process scales and tank geometries from ~10-1000L. Multiple tank geometry types (cylindrical and cubical tank shapes) were considered and will be compared to existing literature reports. Sensitivity of model inputs and the uncertainty in the model compared to experimental data will be discussed. The results of a qualified CFD model across relevant scales during process development is important for future use to reduce experimental runs at manufacturing scale.

[1] Thomas, J., Sinha, K., Shivkumar, G. et al. A CFD Digital Twin to Understand Miscible Fluid Blending. AAPS PharmSciTech 22, 91 (2021).