In the last few years, many pharmaceutical companies have started investing in continuous production, and some have already succeeded in filing new pharmaceuticals using a continuous flow manufacturing process. This article summarizes a study at GlaxoSmithKline, GSK, where real-time multivariate monitoring added value to the development of a continuous production process of an active pharmaceutical ingredient (API).
Multivariate data analytics (MVDA) tools support real-time monitoring for a continuous flow production process in pharmaceutical manufacturing.
Continuous production can provide several advantages for the pharmaceutical industry. Spatial footprint can be decreased, consumption of solvent and energy can be reduced, and a well-functioning continuous flow production can increase both quality and yield. Furthermore, production volumes can be tailored to demand rather than to the size of the batch vessels. Continuous production also allows for the integration of sensors and PAT tools for real-time monitoring and advanced process control.
Using MVDA for process control of continuous flow manufacturing
In the study, Melanie Dumarey, PhD in Pharmaceutical Sciences, and her colleagues at GSK developed and implemented multivariate data analytics (MVDA) tools for real-time monitoring of a continuous flow production process. The study was performed at the R&D site as well as at the commercial manufacturing plant.
The process that was studied was an API synthesis where the material was processed in four reactors and reagents were added at several stages. A total of 69 sensors were used as part of the control strategy to measure conductivity, temperature, flow rate, pressure, and pump speed, with temperature and flow rate constituting critical parameters.
Monitoring continuous production by summarizing 69 sensors in one trend
While it is impossible for an operator to continuously monitor 69 different sensors, multivariate monitoring tools can be used to summarize process trends in one or a few multivariate trends, based on covariance patterns (principal component analysis, PCA). Another advantage of using multivariate trends for process control is that the model is more sensitive to small process drifts.
From multiple sensor information to one multivariate trend.
In the early development stage of the continuous flow production process, the aim was to gain process understanding. The team used retrospective data analysis to better understand the variability patterns in the sensor data. One insight was that there were step changes in pump pressure when feed vessels were changed. In order to have a robust model, a full feed cycle had to be included to avoid false alarms.
Building the MVDA model for real-time monitoring
In the second development stage, the team wanted to test different process conditions to learn more about the process and efficient manufacturing. This time, real-time monitoring architecture was used. Two hours of data from the 69 sensors were extracted from the process control system to build a model offline in SIMCA-P from the Umetrics® Suite of Data Analytics Solutions. Once the model was finalized, it was uploaded to SIMCA-online to make real-time predictions.
By using real-time multivariate tools in this stage, the team identified two main failure modes – pump blockage and gas reagent supply blockage – and the root cause of the blockages could be identified.
Gaining insights from real-time multivariate monitoring.
Adapting the model for commercial production
The final stage of the study was to transfer the continuous production process to the commercial manufacturing plant. The colleagues at the manufacturing plant wanted to be able to increase or decrease the manufacturing rate when needed, and not run at fixed set points, and the model was successfully changed to support that type of monitoring.
The study was a first step to develop a continuous flow production process using multivariate methods for advanced process control. By customizing the multivariate tools, already from the early stage process development to commercial manufacturing, the team could add value to the process and gained insights they probably would not have attained without the MVDA tools. Accordingly, the approach could be useful for Quality by Design, QbD. Furthermore, the MVDA tools made it easier for operators to monitor the process and encouraged discussions at the manufacturing plant.
Melanie Dumarey currently works at the Centre of Excellence Continuous Processing, Product Development, AstraZeneca, Sweden.
The results of the study are published in the Journal of Pharmaceutical Innovation:
M. Dumarey, M. Hermanto, C. Airiau, P. Shapland, H. Robinson, P. Hamilton, M. Berry. Advances in continuous active pharmaceutical ingredient manufacturing: real-time monitoring using multivariate tools. J. of Pharm. Innov. (2018) in press.
Read more about lean manufacturing and six sigma.
Want to know more?
In this webinar, Melanie Dumarey gives a detailed presentation of the study.
Register to watch the recording.