(401a) Process Understanding Tools in PAT Implementations

For the past years the pharmaceutical industry has been trying to increase its process understanding in order to keep pace with regulatory demands and also to increase its competitiveness by being able to find improvement opportunities and increase efficiency. This is also a requirement from the new paradigm of Quality by Design.

For this increase in process understanding demand, the industry has started to collect increasingly amounts of new data from the process and the data itself is now of different types, particularly with the introduction of PAT tools. These tools that are able to collect complex data (NIRS, 2D-Fluorescence, dielectric spectroscopy, etc.) pose challenges for the data organization, accessibility, prioritization and knowledge extraction. In fact, without these challenges solved, the amount and type of data collected can reveal to be a considerable investment and irrelevant for practical use.

If data is not systematically and automatically organized (presuming that it is being automatically gathered), making sense of it can become an unreasonably time-consuming activity. Diagnostic and troubleshooting efforts are lost in the data matrix complexity and energy is spent in the organization of the data in the context of the production process. Also, this lack of organization also poses barriers to the implementation of prediction and MSPC models across the entire cascade of process steps (from raw-materials to end product). To perform true whole-process analysis across the entire process and process life-cycle in the lights of the new guidelines and regulatory demands, this issue needs to be solve with new tools that enable the shift of energy from data organization to the more useful process understanding and process knowledge sphere with a true system engineering environment.

Therefore a new tool that provides a solution to the industry to tackle these problems is needed. This tool should allow the organization and storage of different types of data: Process data (temperatures, pressures, flow rates, etc.), PAT fingerprint data (NIR. MIR, 2D-Fluorescence, etc.), LIMS data and in-line monitoring data. These data have to be added in real-time or be a part of an historical database of the process enabling fast monitoring and diagnosis of the process or historical analysis of previous runs. The necessary graphical instruments are also required in this tool to enable quick and intuitive visualization of the data according to its type (single plots, landscapes, etc.).

Comparison of data should also be possible in that graphical interface that takes the previous automatic organization of data and allows easy evaluation of different data types, from different process steps (ex: unit operations) at different time steps or from different runs. This should prove very useful when performing root-cause analysis and assessing cause & effect on the process.

Many of the data being gathered from pharmaceutical production process is used (or could be used) to predict production quality, control the process, diagnose performance, etc. This is especially true since the introduction of PAT tools that can extract increasingly more informative data from the process. By applying modeling techniques to these data (PAT data, process data, etc.) more process understanding and control can be implemented. For these reasons, this tool should support an easy integration of inline prediction models, calibrations, MSPC, etc. from third-party software. In fact, by having this feature, the user has the chance to improve its models in accuracy and robustness by taking advantage of the organization and accessibility of all the data that ultimately leads to a greater process understanding. 

The importance of fast sharing of information and findings is an increasing necessity in the industry when knowledge-based decisions often have to be taken with haste but also truthful. In this context, a new tool has to support mechanisms to share the information found by using it so it can be spread across the company (through e-mail, dropbox or saved in the company’s database).

In order to exemplify the above-mentioned needs in the pharmaceutical industry and the advantages of such tool, it is intended to present it in the context of a case study with data from a bio-production pharmaceutical process.