Manufacturing Vaccines: An Illustration of Using PAT Tools for Controlling the Cultivation ofBordetella pertussis

Abstract

An illustration of the operational consistency of the upstream part of a biopharmaceutical process is given. For this purpose four batch cultivations of Bordetella pertussis have been executed under identical conditions. The batches have been monitored by means of two fundamentally different process sensors. First, common single channel measurements such as temperature, pH, dissolved oxygen (DO), and flow rates are used and second, the multichannel measurements from the NIR (Near Infrared) analyzer. Because of the fundamental differences between the two types of measurements, two models have been developed to evaluate the operational consistency. The last sensor studied is a typical representative of process analyzers which are described in the PAT (Process Analytical Technology) guidance document issued in 2004 by the American Food and Drug Administration (FDA). Data from both sensors have been evaluated by a multivariate tool for data acquisition. This resulted in two different performance models. Again this approach is characteristic for the implementation of PAT for the manufacture of biopharmaceuticals.With both performance models, we were able to explore the operational consistency of the batches. In addition, the performance models were also able to detect a deviating batch. Further, it was shown that both sensor types gave partly overlapping information since a deviation in the batch profiles of the logged process variables was accompanied by a deviation in the spectral batch profiles.The performance models are valuable tools in developing advanced monitoring and control systems for biopharmaceutical processes. Using such models, advanced knowledge based systems can be developed to detect abnormal situations in an early stage and remove the cause.The procedure of data processing described in this article is relatively new in the biopharmaceutical industry. The NIR analyzer and both performance models presented in this article are clear ingredients for better process understanding and process control, as intended in the FDA's PAT Initiative. This initiative is part of the FDA's strategy of cGMP (current good manufacturing practice) for the 21st century and aims at introducing innovations Address correspondence to Eric N. M. van Sprang, in both the manufacturing of biopharmaceuticals and the development of new biopharmaceuticals.This study shows the feasibility of two typical PAT tools for controlling the manufacturing of biopharmaceuticals. To the best of our knowledge such feasibility study is not documented up to now in the scientific literature.

Gene‐expression‐based quality scores indicate optimal harvest point in Bordetella pertussis cultivation for vaccine production

“…Implementation of typical PAT (FDA 2004) sensors such as NIRS during batch fermentation, would generate a continuous, multivariate process fingerprint of the inoculation-to-harvest trajectory. On-line monitoring of nutrient concentrations for automated detection of the harvest point is also possible (Arnold et al, 2002;Ferreira et al, 2005;van Sprang et al, 2007). Such a science-based PAT approach, in combination with the presented model to measure product quality, could result in a process that yields whole-cell vaccines of such intrinsic quality and consistency (Gnoth et al, 2007) that it would change the perspective for the use of the Kendrick potency test (Irwin and Standfast, 1957;Kendrick et al, 1947).…”

Hybrid modeling framework for process analytical technology: Application toBordetella pertussiscultures

“…Setting (b) contains the online measured data of pH, temperature, percentage of DO, and a selection of wavelength of the spectra (1,111–1,397 and 1,587–1,852 nm). This wavelength selection originates from van Sprang et al7 was applied in Soons et al4 and is chosen in this study because the excluded wavelengths correspond to saturated intensities due to water. Both settings, (a) and (b), are augmented by the initial component concentration values of every batch to compare the PLS model to the hybrid model providing the exact same data.…”

“…Multivariate regression, (nonlinear) partial least square (NPLS), evolving factor analysis, support vector machines, or principal component analysis (PCA) are probably the ones that are most commonly applied and most successful 6, 7. These methods are data driven, and in most cases, they are applied on their own disregarding other valuable process knowledge.…”

Future Trends for PAT for Increased Process Understanding and Growing Applications in Biomanufacturing

“…55 The measurements were used for a consistency analysis of the batch process, providing better process understanding which includes the understanding of the variations in the data. MSPC analysis of four batches of data was performed to monitor the batch trajectories, and indicated that one batch had a deviation in the pH.…”