Leveraging Data for Better Biopharmaceutical Process Control

RGtimeline/Shutterstock.com

Upstream biopharmaceutical manufacturing has always depended on process measurements, because precise levels of nutrients and oxygen must be monitored closely to ensure cell culture viability and the health of the process. For many years, though, extending this approach into true process analytical technology (PAT) and using it to track and control processes was done mainly in the research lab.

Recently, there has been a significant mindset change in the industry. Biopharmaceutical companies are starting to use PAT in more advanced development and even some manufacturing work, while methods are also moving downstream. Francisca F. Gouveia, a biopharmaceutical process specialist who is now working for a biopharmaceutical company, evaluated the industry’s progress in a PhD thesis written while she worked for 4Tune Engineering, a consulting company and software vendor based in Portugal, which she defended at the University of Copenhagen (1) in March 2018. 

As she sees it, biopharmaceutical companies are now developing the mindset that will be needed to take PAT and advanced process control to the manufacturing area and to different areas of manufacturing. As she says, “PAT is not just sticking a probe into a bioreactor. A great deal of work is needed to make advanced process monitoring and control approaches feasible.” At this point, due to higher complexity of unit operations and production infrastructures, biopharma remains a step behind small-molecule manufacturing (where the methods have routinely been used in GMP facilities).

One major challenge is developing a lifecycle approach to using PAT and advanced control that would allow these approaches to be used, from R&D through biopharmaceutical tech transfer, and manufacturing, she says. Ms. Gouveia and Jose C. Menezes, CEO and founder of 4Tune Engineering, shared trends they are seeing in biopharmaceutical process monitoring and control with BioPharm International. 

Greater use of PAT 

BioPharm:At the International Forum for Process Analytics and Control (IFPAC) conference in 2018, a record number of presentations addressed biopharma PAT and process control. Are more biomanufacturers adopting PAT and advanced control in biopharma?

Menezes:The initial use of PAT in bioprocessing was focused on monitoring specific culture attributes in upstream processing, with mid- or near-infrared (MIR or NIR) spectroscopies, or some transition analysis and performance studies with ultraviolet (UV) and NIR spectroscopies used in downstream processing. Those applications followed the experience and successes in pharma solid-dosage forms manufacturing. However, the low concentrations and/or aqueous matrices involved created several challenges, particularly for NIR spectroscopy, due to a lack of sensitivity and specificity.

Monitoring calibration models could be built and used across lifecycle and process scales only if very detailed care was taken in designed experiments during calibration development (2) to avoid ‘indirect calibrations’ that can result when vibrational spectroscopies are used in bioprocessing.

A second moment for PAT in biopharma involved moving from parameter monitoring to process state estimation and exploring control opportunities. Instead of putting effort into unspecific and insensitive methods pushed to their analytical limits, the multi-parametric character of all PAT methods was explored with multivariate analysis to provide a complete estimation of the bioprocess state over time (e.g., a process trajectory during an upstream cultivation), creating the opportunity in pharma to define guided sampling and end-point determination strategies. This involved most notably NIR and MIR spectroscopies and, more recently, Raman spectroscopy as well. There are accounts of in-situand at-line under GMP conditions use of different spectroscopies (2,3).

Currently, we are observing a trend toward connecting process condition-monitoring to accurate product quality-control during processing, by very specific techniques such as in-situmass spectrometry (MS). The capability to use MS as a PAT, and aggregate multiple attributes that MS can very specifically measure, is a real breakthrough. IFPAC 2018 already showed this year several talks on MS-based multiple attribute methods (MAM), a new acronym being used by FDA’s Emerging Technology Team (ETT) to name the PAT use of these methods.

The aims of optimization and control were present at earlier phases for PAT in biopharma, but today it is possible to monitor product quality attributes end-to-end (E2E) in upstream and downstream processing. Use of approaches such as MAM could open a completely new field for PAT used earlier in development for building. This approach would result in very fast process/product understanding, but more importantly, would support the process performance qualification (PPQ, Stage 2) and the entire commercial lifecycle (Stage 3).

Organizational obstacles

BioPharm: What obstacles remain to greater use of advanced monitoring and control in biopharma?

Gouveia: Some of the obstacles are not technical at all, but organizational. Companies need to align different groups (e.g., the process engineers, analytical experts, regulatory departments) in efforts to leverage PAT and advanced process control. Some companies still have a disconnect between production  experts and those who work on statistics and data analysis. This approach does not facilitate data-driven thinking. 

As I found while researching my thesis, people working in PAT at biopharmaceutical companies often have difficulty trying to take Raman or NIR spectroscopy out of the lab into development and manufacturing. Extending the lifecycle concept to PAT method development will be essential to ensure its fit for purpose during routine production.

The right questions need to be asked: what is our problem statement, what do we want the PAT method to do, in what range, in the presence of what other substances (e.g., metabolites in cell culture), what technique will work best for a specific application, and what performance requirements [established in the International Council on Harmonization’s (ICH) Q2, the European Medicines Agency (EMA) Near-Infrared Spectroscopy NIRS guideline, and guidance from the US and EU pharmacopoeias], will be required for control purposes?

Efforts taken early in development will pay off, and the PAT method will be robust even if there are changes in raw materials or in a specific process step. In general, more thought [to use of process monitoring and control] is needed earlier on. There’s a need to take knowledge of the process and use it to build a method, then to align the lifecycle of the method with that of the process through teamwork and collaboration.It’s much more than understanding chemometrics and spectroscopy. You need to understand the organization, its quality culture and the regulatory framework.

Leveraging soft sensors

BioPharm: Has the industry grown more comfortable with the concept of soft sensors?

Menezes: The use of soft sensors in bioprocessing is seeing a rebound from its promising inception in the mid-1990s. Today, they are being used across operations, making true the promises of quality by design (QbD).  In time, they may even permit the use of real-time release (RTR) in bioprocessing. As data management becomes more sophisticated, the possibility of retrieving and aggregating data from critical process parameters (CPPs) and critical quality attributes (CQAs) in real-time will allow CQAs to be estimated during processing for subsequent operations. That opens many possibilities for E2E optimization and feedforward control and will make QbD a reality in bioprocessing. We have developed a soft sensor applications for GMP biomanufacturing in which media components quality attributes are related to upstream yields or bulk product CQAs (e.g., stability) (4,5).

BioPharm: Can you give an example of how soft sensors are being used to leverage data?

Gouveia: In the usual commercial biopharma setup, you have a complex package of specs and very complex upstream and downstream unit operations. It can often be difficult to pinpoint which step(s) is/are responsible for quality results. Soft sensors can be effectively used to link processing steps using different data sources (e.g., online and offline process measurements, in-process quality measurements) to shifts in product quality. For example, chromatography data can provide a fingerprint of the product at a specific point in the process so that changes in peak shape can be used to better understand the underlying buffer preparation variability or to anticipate packing problems that might have been overlooked.

BioPharm: In general, what progress are you seeing in the use of advanced process control and the use of predictive models?

Menezes: We see a significant change in how advanced control practices from other processing and systems engineering areas are today applied in biopharma. The realization that in bioprocessing there is one more control agent present (i.e., the biologic component used, such as bacteria or cells) took some time to accept. Feedback control has always been carried out based on extracellular measurements, specific sensors, and multiparametric ones (e.g., PAT).

Effective advanced control must start early in the lifecycle by considering the process E2E: What are the CQAs, the CPPs available for establishing a control strategy?  This is a risk-management exercise that will make the knowledge about causality between CPPs and CQAs explicit. From this point on it will be clear where, when, which, and by how much each CPP should be set or changed to achieve a specific set of CQAs.

Gouveia: Companies are becoming more open to the idea of predictive models, as they start to address data gathering and management. This issue often centers around the manufacturing execution system (MES). Commercial data-aggregation software packages can help companies develop the necessary models. This is true upstream, but we’re also seeing more advanced control move downstream to help better control yields and purification steps, and even to improve the management of plant and equipment.

BioPharm: What is preventing a more unified approach to integrated continuous bioprocessing?

Gouveia: Several companies are  working on end-to-end continuous biopharmaceutical processes, but the industry’s existing plants have a totally different facility design. A strong business case will be needed for taking this approach before much more can be done.

Menezes: Integrated bioprocessing presents specific challenges, compared to continuous small-molecule processes, in terms of PAT, controllability, and RTR. Over the next decade, plant modularization and new facility designs promise to move biopharm and small-molecule manufacturing closer to Industry 4.0 models.

References

1. F. F. Gouveia, “Chemometrics and PAT Applications in the Pharma and Biopharma Industries,” PhD Thesis, University of Copenhagen, March 2018, SL grafik, Frederiksberg, Denmark, March 2018. 
2. JC Menezes, “Process Analytical Technology and Quality by Design in Bioprocess Development and Manufacturing,” Industrial Biotechnology, in Comprehensive Biotechnology, Vol.3, 2nd edition, Ed. A Moreira, Moo-Young Ed., (Elsevier, 2011).
3. “Near-Infrared Spectroscopy in Laboratory and Process Analysis,” Encyclopedia of Analytical Chemistry (Wiley, 2012) DOI:10.1002/9780470027318.a9361
4. “Spectroscopic finger-printing of raw materials,” Hoffmann-La Roche AG, WO 2012/059520 A1. 
5. “Method and system for preparing synthetic multicomponent biotechnological and chemical process samples,” Hoffmann-La Roche AG, WO/2015/097217. 

Article Details

BioPharm International
Vol. 31, No. 5
May 2018
Pages: 42–45

Citation

When referring to this article, please cite it as A. Shanley, "Leveraging Data for Better Biopharmaceutical Process Control" BioPharm International  31 (5) 2018.