Reiner LUTTMANN, Hamburg University of Applied Sciences, Germany
Jens FRICKE, Boehringer Ingelheim, Austria
Sven-Oliver BORCHERT, Bayer AG, Germany
Jan-Patrick VOSS, Evonik, Germany
A development of continuously operated integrated pharmaceutical production processes needs a tremendous expenditure. Beside the installation of a full-scale production, scale-down concepts are essential to meet QbD-specifications of FDA.
In this presentation the surrounding PAT-field of such plants for production of potential Malaria vaccines is discussed in order to create model based QbD-compliant strategies. This includes fully automated processing and global process monitoring with additional classical and spectroscopic measurement procedures including enhanced data processing and MVDA.
A two-stage upstream line is also scaled-down in a sequential/parallel operated six-fold bioreactor-plant including flow analysis at-line procedures for substrates- and target protein-detection. The plant allows a fully automated simultaneous DoE-process optimization and identification of Critical Process Parameters. The DoE-model and Monte Carlo simulations create also the Design Space and the Control Space of QbD-production. ·
Similar methods are used in the down-stream area for optimization of cyclic protein purification. These methods are applied with an AEKTAT avant which is developed especially for DoE.
One main focus of the work lies on the development of a global MVDA-based monitoring system for variables like cell density, glycerol-, ammonium-, total secreted-, and target protein-concentrations but also on evaluation of process quality (reproducibility) of cyclic processing.
Applications of NIR-, Raman-, and 2D-Fluorescence-Spectroscopy and the appropriate PLSR-modeling leads to a partly success. This was improved by using the nonlinear SVR-Support Vector-machine Regression.
However, a MVDA with only classical bio-engineering variables creates also satisfying results in bio-monitoring up to an on-line detection of secreted target protein.
Qualities of running processes were evaluated with Golden Batch approaches. GB-tunnels were created with QbD-conformed process courses and then used for on-line observation and prediction of actual first principal components. MPC-Model Predictive Control was also implemented in order to avoid a leaving of the GB-tunnel by correction of process set-points. These methods open the way to an on-line release of pharmaceutical products.