PAT Plus Integrated Data Analytics Enables Bioprocess Automation

Automation

Apr 19, 2022 | 6 min read

Process Analytical Technology (PAT) can help biopharma manufacturers automate and optimize bioprocesses effectively. Sartorius offers PAT-enabled instruments and data analytics tools specifically designed to support process optimization and automation using a QBD-approach favored by regulators.

^{This article is posted on our Science Snippets Blog}

As bioprocess development becomes more complex and regulatory oversight continues to lean toward documentation of processes using statistical models, biopharma manufacturers are faced with increasing pressure to improve product quality and reduce manufacturing costs. Biopharma manufacturers are responding by adopting Process Analytical Technology (PAT) to enable automation and ongoing process optimization.

What Is PAT?

PAT is method for measuring and controlling a process based on product quality attributes. Controlling product quality is a key differentiator from “traditional” manufacturing, which uses empirical process experience rather than statistical models to show that a process is yielding the desired result.

Traditional process control can be dependent upon individual operators and fixed process limits. Experience may have shown that a specific recipe normally delivers an acceptable product. However, this often doesn’t take into account variables such as raw material or intermediate product variation, environmental changes, unexpected process input deviations, process or sensor drift, or other factors that an affect the results. The use of newer and more advanced sensors is limited or not used to its full potential.

PAT, on the other hand, uses a well-documented approach that can deliver huge commercial benefits. Biopharma companies have reported up to three-fold increase in productivity. In some companies, batch process production times have been reduced from 30 days to 30 hours. Other companies converting from batch to continuous manufacturing have reduced manufacturing time from 30 days to 90 minutes.

PAT also helps improve sustainability and margin by allowing companies to source raw materials more widely (with less risk). It helps improve the quality and consistency of the product, while reducing waste and rejects. PAT is a key element of a Quality by Design (QbD) approach to biopharma process development and manufacturing favored by regulatory agencies.

Making use of large amounts of data gathered by sensors opens the door to analytics-based optimization. Replacing traditional time- and resource-demanding options with multivariate soft sensors lays the groundwork. With an increased understanding of the interactions of very complex production processes, you gain the ability to create closed control loops and reduce costs related to CO2, energy, resources and time.

The scale of potential benefits from PAT1 include:

Increase productivity three-fold
Reduce production times from 30 days to 30 hours
Reduce manufacturing time from 30 days to 90 minutes

PAT Automation Goals for Bioprocesses

A key reason that biopharma companies move toward PAT implementation is to support automation and continuous process manufacturing.
Two main goals when utilizing DOE and MVDA for PAT automation are: repeatability and optimization. You want to keep your process in control, and you want to be able to ensure the process can be repeated with same output (yield and quality). In addition, you want to be able to optimize your process to allow for factors such as changes in environmental influences or raw materials.

Repeatability

Repeatability is all about monitoring and controlling the process to detect drifts, shifts or outliers. With the right tools (such SIMCA^®-online), you can set up warnings and detect drifting processes before they lead to costly mistakes. In addition, you can make corrections if, for example, a batch goes out of spec. By applying what we call “golden batch models,” which are created based on your MVDA models, you can keep your process under control.

Data analytics can also add a predictive layer, allowing you to forecast quality attributes, yield, and other key batch-related indicators based on the trajectory of the data. Using a control strategy based on data analytics models also helps support FDA requirements towards continuous process verification (CPV).

Optimization

Optimizing your process is the next part of ensuring a well-controlled and repeatable process. It’s important to look at the causes of any batch-to-batch variation and understand the cause. With the proper modeling tools, you can quickly assess all variables at the same time to help pinpoint root causes or deviations that might occur during the process. You can also simplify and enhance process development to improve cycle times, yield and product quality.

In addition, PAT allows you to fine-tune your models to make sure that they remain effective. If any changes are made to the process, you will have the process insights needed to understand the impact the changes have on your model and will be able to adjust for them. Model verification and optimization helps ensure your models continue to work for various applications.

What Does a PAT Journey Look Like?

In a previously recorded Sartorius webinar focusing on PAT Automation, Paul Gilham, Innovations Director, Optimal Industrial Technologies, outlined the high level steps of a PAT Journey. They can be summarized as follows:

Identify the product Critical Quality Attributes (CQAs). These may change over time as more is learned about the product. Therefore, you should plan to collect and store all the data.
Identify instrumentation suitable for measuring CQAs. If no correlation can be found with an instrument, then another technology has to be tried. If no instrument works, then project is a non-starter.
Combine data sources. Sometimes ‘data fusion” works best – either by combining more than one spectral instrument, and/or: combining spectra data with univariate data.
Install the instruments, MVDA software and PAT knowledge manager software (e.g. synTQ, SIPAT, DeltaV, EasyPAT). Installing the instrument in the right position is critical to getting the correct result. It’s important that all the instruments and software are communicating with each other.
Account for operational considerations. Be sure the instrument is installed in such a way that production is not impeded, and you receive optimized measurements.
Run a Design of Experiments (DOE) and develop the MVDA models. Use the collected data from chosen instruments to develop your models. Normally an analytical laboratory is required for the calibration exercise.
Use data to calibrate your model. MVDA model-building is effectively a calibration exercise for the instruments, and you use it to prove the model.
Test and refine the models for operational use. This verifies the precision of the calibration. Use should use a different data set for verification from that used for calibration.
Run the process models in real time. Run the process in real time (whether batch or continuous) to build a picture of the process to develop a quality-based control algorithm.
Build product quality-centric control models based on process understanding. A key difference between a PAT system and traditional system is that the control is based on real-time quality, not just empirically based recipe data.
Test the system, and when performance is assured, move to manufacturing. Huge benefits are possible in R&D, pilot and full-scale manufacture.
Implement continuous improvement. Ideally over weeks, months and years, further refine the models to maximize benefits.

What Tools Are Needed?

Data analytics solutions play an essential role in implementing PAT. Selecting instrumentation that lets you incorporate PAT methodologies and utilize data analytics is important. Today in biopharma that often includes using spectral instruments, such as RAMAN, IR, or NIR, so you need data management and analysis designed to support large volumes of multivariate data.

In addition, you need multivariate data analysis (MVDA) software such as SIMCA^® for model building and understanding, a control system (PLC/DCS), laboratory equipment (such as bioreactors), PAT knowledge manager (such as synTQ, SIPAT, DeltaV, or EasyPAT) with the SIMCA-Q execution engine to enable use of the multivariate models in real-time, and a Design of Experiments (DOE) tool (MODDE^®.).

^{The Umetrics® Suite supports the process development and manufacturing stages of the biopharmaceutical product lifecycle.}

Biprocess instruments and devices fit into a framework that is known as a process control hierarchy. Data analytics falls into a category known as SCADA (Supervisory Control and Data Acquisition), which controls equipment automation and recipe execution. (See illustration below). It is connected to PLCs (controllers) and devices, including sensors, actuators, pumps and other analytical instrumentation.

In addition, in biopharma, analytical instruments connected into the system may include spectroscopy, chromatography, mass spec and HPLC, which provide advanced insights into the process. Data from these can be coupled with “soft sensors” and use modeling software to provide additional insight into what’s happening within your processes.

^{In the process control hierarchy, data analytics falls under the SCADA (Supervisory Control and Data Acquisition) layer.}

Moving Toward Continuous Processes in Biopharma

The FDA is encouraging pharma and biopharma manufacturers to move toward continuous processe using PAT and QbD. Recently, the director of the FDA’s Office of Pharmaceutical Manufacturing Assessment said:

PAT and [quality by design] initiatives have contributed to substantial progress; however… as an industry we have not reached the intended vison that we set in the early 2000s. We are still focused on monitoring rather than controlling the processes to ensure product quality.”

- Stelios Tsinontides, director of the FDA’s Office of Pharmaceutical Manufacturing Assessment

The FDA has asserted that using QbD and PAT can reduce product defects, lower recalls, and produce higher quality products, as well as reduce the risk of drug shortages. ICH guidelines outline how important a control strategy that takes into account critical process parameters and critical quality attributes is to your ability to produce a good Certificate of Analysis (COA) for your product.

Using a QbD Approach

The ultimate vision behind a QBD approach is to achieve a state of control in which you know what the quality of the product will be at the end. To make that happen, it’s necessary to incorporate sensor data as well as equipment control and have analytics integrated into the equipment.

Using data-based models, you can create control strategies built around critical process parameters, which control critical quality attributes. By keeping critical parameters in control, you can predict what the quality of the product will be during release.

Instruments and Tools to Support PAT

Sartorius offers a suite of PAT-enabled instruments and embedded data analytics solutions that support automation and a QbD approach to process development.

These include the Umetrics Suite of Data Analytics Solutions (including MODDE^® DOE, SIMCA^® MVDA and SIMCA^®-online real time data analytics, and SIMCA-Q^®, the embedded solution for OEM manufacturers) as well as the Sartorius BIOPAT^®, Unit Operation, PAT Technology and MFCS software that controls a bioreactor.

^{Sartorius products support data aggregation, monitoring and control, data evaluation, as well as process optimization.}

All of these things will help you enhance process control and the robustness of your process, ultimately leading to improved yield and product quality, as well as enabling predictive control and prescriptive analytics.

^References:

^{1. Paul Gilham, Innovations Director, Optimal Industrial Technologies, PAT Automation Focus Days, Webinar Presentation.}