(471f) Combining Numerical Modelling (CFD), Statistical Modelling (MVDA) and Laboratory Experiments for the Optimization of Large Scale UF/DF Operations

Plant scale experiments are both costly and time consuming to run. Typically, in pharmaceutical manufacturing, these experiments involve large quantities of high value product. APC combined multiple models; numerical, statistical and laboratory-scale, to optimise large scale UF/DF process runs. Computational Fluid Dynamics (CFD) used direct numerical simulations to model the developed flow and associated velocity and shear profiles. A representative laboratory scale unit operation was built, used to test the outcomes from CFD and develop a predictive MVDA model in a time and cost-effective manner.

CFD modelling was deployed to gain an understanding of two UF/DF manufacturing unit operations at two manufacturing sites. One site was exceeding the specified protein aggregation level while the other met the criteria. CFD determined the hydrodynamic environment generated within each site and potential factors (pipe network, elbow joints, control valves, etc.) contributing to the protein aggregation levels. CFD showed that shear stress within the systems was related to flowrate, pipe diameters, elbow joints in the pipe network and the actuator valve.

A numerical lab-scale model that could be constructed in a laboratory and replicated the shear effects present in the failing manufacturing UF/DF process was developed. The lab-scale unit operation accurately represented the process by including steel tubing, elbow joints and a T-valve (to mimic the disruption to flow caused by an actuator valve). A 0.29 % increase in aggregation, similar to levels observed at the failing site, was achieved during runs of the CFD developed lab-scale model. MVDA models were built to correlate the Raman spectral data to off-line SE-HPLC protein aggregation data, enabling the prediction of protein aggregation.

This integrated approach between mathematical models and experiments allowed for the rapid assessment of potential practical solutions on a laboratory scale unit operation before large scale pharmaceutical manufacturing runs. While the MVDA model allowed for real time assessment of critical quality attributes.