Monitoring And Control Of Reproducibility In Quasi-Continuous Integrated Production Processes Of Active Pharmaceutical Ingredients
The development of integrated production processes include the combination and transformation of current batch oriented unit operations into linked sequential/parallel production strategies.
The presented process starts with a two-stage upstream consisting of cell cultivation and subsequent protein production, which in turn results in a five step downstream process, consisting of cell clarification via a separator, retention of cellular debris using microfiltration, concentration of the secreted product by ultrafiltration with subsequent buffer exchange through diafiltration, followed by a final purification using column chromatography.
The three main operations cell breeding, protein production and the complete downstream line are running in series, but also in parallel with a one-day-offset each. Such strategies were developed at HAW Hamburg [Luttmann et al., 2015].
To achieve reproducible process conditions, the process development was done in accordance with the known industry guidelines from FDA and ICH regarding QbD and PAT. In this context the identification of optimal Design Spaces and Control Spaces was in the foreground.
The in-line measurement of important media components and cell physiological parameters as well as the on-line evaluation of process reproducibility, are remaining unsolved problems in industrial biotechnology - irrespective of whether a process is operated batch wise or continuously. A way out of this dilemma can be found by on-line MVDA data processing. This paper describes a comprehensive application of MVDA in process monitoring and control using the example of an integrated production of potential Malaria vaccines with Pichia pastoris.
Cell mass, glycerol and secreted target proteins as well as cell internal AOX content are measured with NIR-, Raman- and 2D-Fluorescence-spectrometry. Here, intensive off-line analysis of the concerned process variables form the foundation for the training of spectral observations as well as for the evaluation of cell specific reaction rates from routinely measured on-line variables with MVDA-investigations.
The main approach of MVDA was an on-line monitoring of reproducibility of involved unit operations. This was achieved by off-line modeling of Golden Batch tunnels and on-line evaluation of the process trajectories using SIPAT® and SIMCA® software tools.
On top of this, on-line process prediction and on-line Golden Batch control were implemented. The prediction is based on IBR-Imputation by Regression (SIMCA® Q) and the control of processes evolving outside their Golden Batch limitations is based on BOBYQA-Bound Optimization by Quadratic Approximation (SIMCA® online). Such methods for process monitoring and control of quasi continuous pharmaceutical production pave the way for Real Time Release of APIs.
All approaches have been approved and tested in real industrial-like production processes which have been performed over several weeks.