The complexity of biopharma processes requires innovative solutions.
Dec 01, 2014
By Cynthia Challener, PhD
A key component of the quality-by-design (QbD) approach to pharmaceutical manufacturing is the implementation of process analytical technology (PAT) for ongoing monitoring and adjustment of production processes to ensure the production of biologic APIs with consistent quality that meet specifications every time. Implementing PAT for bioprocesses presents unique challenges given the greater complexity of cell-culture and fermentation processes, the variability of raw materials (particularly living organisms), and the move from stainless-steel to single-use equipment. The recent emphasis on establishing a thorough understanding of the manufacturing process from the earliest process design stages through the use of process characterization studies is, however, a key enabler of PAT approaches, according to Kumar Dhanasekharan, director of process development at Cook Pharmica. Advances in high-throughput systems, Raman and near-infrared (NIR) spectroscopy, high-performance single-use sensors, and chemometric methods are also helping manufacturers of biopharmaceuticals overcome these hurdles.Unique challenges
The key to successfully implementing PAT for bioprocesses is the availability of robust, reliable, low-cost, easy-to-use online/atline analyzers; a complete understanding of the variability inherent in the process and the variability introduced by raw materials; and sufficient time to develop a process that is amenable to the application of PAT during manufacturing, according to Rajesh G. Beri, head of mammalian research and technology for Lonza Custom Development and Manufacturing. In addition, PAT is much more than just one sensor integrated into a process. “Effective PAT implementation implicates the use of various measurement and control devices and hence requires a commitment to sophisticated automation solutions,” says Mario Becker, director of marketing and product management for PAT and automation at Sartorius Stedim Biotech.
Because bioprocesses are complex and dependent on multi-factor interactions between the process variables, using PAT for biotherapeutics can be more challenging than for small-molecule drug manufacturing. “Real-time in-process measurements of critical quality and performance attributes of raw and in-process materials and processes are not always possible for biologics due to the complex nature of most analytical methods and the reaction systems,” says Dhanasekharan. Using spectroscopy for the analysis of a complex broth of cells, where the active ingredient often remains in the cell bodies, is difficult, adds Ivo Backx, manager of business and project development for the pharmaceutical industry at Siemens. “Often, secondary metabolites and related parameters must be detected as critical process parameters and monitored to track the batch status in real time, which requires a lot more process understanding,” he explains.
According to Becker, as biotechnology has advanced, the number of factors and setpoint accuracy and precision of control have become more demanding. At the same time, regulatory agencies have continued to press for tighter controls, greater understanding, and higher consistency. “Therefore, each step of the process requires a level of PAT that is balanced between financial investment, legal implications, and profitability. In the end, it is all about risk mitigation,” Becker asserts.
Further complicating the use of PAT for biologics is the complexity introduced by the variety of raw materials and the variability in raw material lots/quality used in product manufacturing. “It is possible with extensive experience in biologics manufacturing to be able to identify critical sources of variability; however, much work remains to be completed in order to gain control over the variability posed by raw materials,” says Beri.
The shift from classic stainless-steel equipment toward smaller production skids, most often based on disposables, is also presenting challenges for measurement methodologies, according to Backx. Finally, implementation of PAT for bioprocesses implies the extensive use of the design-of-experiment (DoE) approach. “The cost for DoEs in the biopharmaceutical industry can be quite high, and DoEs can also be quite time consuming, making the use of PAT difficult. One solution may be to work with small volumes and run several experiments in parallel,” Backx says.
Upstream and downstream
Although upstream processes tend to be more challenging and the sources of variability (raw materials and process) are reduced in downstream processes (e.g., after Protein A purification for monoclonal antibodies), PAT has been more widely used in the production fermenter or bioreactor. The increased use of PAT for upstream processes is driven by the fact that most of the critical quality attributes of a product, the product yield, and the quantities of process-related impurities are determined.
For biologics that are extracellular and present in relatively large concentrations, there is a good potential for direct detection through the use of spectroscopic measurement techniques. This type of biologic is better suited for PAT applications, according to Kjell François, manager of the PAT data management team at Siemens. He adds that nutrient uptake can often also be measured using NIR as a soft sensor. If the cell metabolism is known, the detection of changes as secondary indicators can be adopted as PAT techniques for process monitoring, even if the active compound to be produced cannot be detected explicitly. “Using PAT as a tool to control the process based on secondary or derived parameters is more difficult, although there are some cases where feed rates are controlled using PAT in order to maintain optimal reaction conditions,” François notes. For example, Dhanasekharan points out that during cell culture, adaptive feed strategies for glucose and other nutrients are becoming more routine due to technologies for fast turnaround on metabolite analysis, such as advanced real-time glucose analyzers and, in some cases, Raman spectroscopy.
With respect to the use of PAT in downstream bioprocessing, François suggests that a lack of good real-time online measurement techniques may be a limiting factor. There has, however, been significant progress in the application of rapid methods for microbial testing and control, as well as the determination of protein concentration, according to Dhanasekharan. “At-line A280 measurements of protein concentrations are now routinely used in downstream operations for real-time control. There have been recent advances in UV-Vis spectroscopy that do not require dilution, allowing for the direct determination of protein concentration without sample preparation, thus increasing speed and accuracy,” he explains.
High throughput methods
High-throughput process development systems and rapid analytical methods have the potential to reduce both the labor and time required to study both the many different variables associated with bioprocesses and the variability introduced by raw materials, according to Beri. “By working closely with manufacturers of high-throughput bench scale systems, we are currently making them more suitable for achieving increased process knowledge while simultaneously reducing the labor and time required to gain this knowledge,” he adds.
One challenge with using high-throughput techniques, however, is that they generate more data than can be handled by commonly used Windows-based software programs. As a consequence, real-time PAT Q-data management systems and multivariate analysis programs are increasingly being used.
The continuous real-time quality control and assurance that is highly desired in biopharmaceutical manufacturing can be achieved using sophisticated process control strategies based on multivariate monitoring techniques, according to Becker. “Multivariate monitoring is based on models representing the relationship between parameters, whereas traditional process monitoring is based on univariate rules incapable of revealing correlation patterns, and sometimes results in misleading conclusions,” he comments.
Software systems for multivariate data analysis (MVDA) in real-time are highly efficient tools for process monitoring and control of previously established design spaces based on current process parameters and analytical data, according to Becker. The software permits early detection of process deviations and provides user guidance for identifying potential root causes of these deviations by displaying easy-to-understand graphics, resulting in not only improved health, safety, and environmental (HSE) performance, but also enhanced control and assurance of the overall process and product quality. Systematic cost savings also can be achieved, and these systems can be fully integrated into the automation architecture of single-use unit operations.
“MVDA techniques are being increasingly used for scale- and batch-to-batch comparison investigations to support or derive process understanding and to ultimately improve the quality, safety, and efficacy of a drug product,” says Becker. DoE, meanwhile, is a standard technique for accelerating process development tasks and gaining additional process understanding with the fewest number of experiments. “Nowadays,” Becker adds, “chemometric software can even be fully integrated into supervisory control and data acquisition (SCADA) systems and therefore can easily be integrated into standard working procedures in a process-development environment.”
“Lonza and many other companies have been very successful in prospectively applying MVDA to increase process performance and control product quality. Some biologics manufacturers have also implemented online MVDA to enable more real-time process monitoring and control, which Lonza is also actively implementing,” comments Beri.
Advances in spectroscopy
Near-infrared (NIR) and Raman spectroscopy are attracting a lot of attention as PAT techniques for bioprocesses. Spectroscopy techniques specifically tailored for bioprocess applications are important solutions for non-destructive process analysis, according to Becker. “We are excited about the use of Raman and NIR probes for biopharmaceutical manufacturing because these probes have the ability to provide real-time information about the process and product in the bioreactor,” observes Beri. “As a result, companies can use this knowledge to help control the process so that it delivers product that meets predefined specifications while also maintaining a high product yield,” he adds.
When combined with chemometrics tools, Becker explains that NIR enables the acquisition of data that otherwise would be unavailable with conventional wet chemistry. “In addition to cell parameters (density and viability), nutrients, and metabolites, NIR is capable of predicting process trajectories, reflecting the overall state of an entire process as a sum parameter,” Becker says. “This functionality allows a real-time comparison of the current batch with the ideal process state, or the so-called ‘golden batch,’” says Becker.
He continues, “With standard adaptation via Ingold ports, compact and robust designs, and an attractive total cost of ownership, spectroscopy will become a standard tool for in-process monitoring and control during commercial manufacturing,” Becker concludes. Process adaptations for single-use unit operations are also under development.
Sensors for single-use and smaller-scale production
The increased use of single-use technologies in biopharmaceutical production processes has resulted in a strong demand for robust and reliable single-use sensors that enable the application of PAT as a basis for effective automation and optimization, according to Becker. “In addition to established multi-use offline sensor technologies, such as those for temperature, pressure, pH, dissolved oxygen, conductivity, flow rate, and basic molecular percentages, PAT has shifted to inline/online methods including the determination of cell, nutrient, and metabolite concentrations and cell viability using spectroscopic measurement techniques. These sensors, which have been developed for and utilized in stainless-steel multi-use systems, now need to be fully integrated and function equivalently in single-use unit operations,” he says.
Most importantly, interface issues between single-use components, sensors, and local control and SCADA systems must be addressed to ease the implementation, qualification, and performance of PAT-enabled unit operations. “Stable and robust processes require sophisticated monitoring and control of critical process parameters based on reliable data acquisition, storage, and evaluation capabilities from inline/online sensors and process analytics,” Becker asserts.
Sartorius recently extended its bioreactor portfolio with automated glucose-feed-control and integrated real-time measurement of viable cell density for automated induction, infection, and harvesting processing routines based on fully integrated single-use sensor technologies. This functionality significantly reduces sampling frequency and, therefore, operator influence, according to Becker. In addition, it increases the level of process automation, thus reducing batch-to-batch variability and out-of-specification production.
Along with the shift to single-use technologies, there is a move in the biopharmaceutical industry to smaller-scale production operations as interest in the local production of drug products designed for local markets and personalized medicine increases. This shift is driving the use of fast and reliable atline analyzers for the determination of numerous biologically relevant parameters using a small amount of sample, according to François. “These techniques are needed to support the smaller scale of batch fermentations expected as the concept of personalized medicines filters down to biologics production operations. Combining such atline analyzers with real-time spectroscopic techniques should allow the effective monitoring of bioreactor farms with small-scale reactors,” he notes. Siemens is also exploring synergies with the analytical techniques applied for the healthcare sector, such as blood gas analyzers, to biologics production. “Sharing technologies looks quite promising and provides new perspectives,” says François.
At the same time, there is a clear trend toward smaller, more dynamic production skids that enable continuous production using PAT, integrating upstream processes with downstream processes, in a feed-forward, feed-backward manner for overall process control, according to Backx. “This shift can fundamentally change the production landscape in the (near) future, and PAT will be a key element in the success of this approach,” he states. “Importantly, because there is no intermediate offline, time-consuming QA control between the different steps, every operation should be monitored and controlled in an optimal way driven by PAT.”
ALL FIGURES ARE COURTESY OF THE AUTHORS
Vol. 27 No. 12
Citation: When referring to this article, please cite it as C. Challener, "Improving PAT for Biologics with Online Spectroscopy and Multivariate Data Analysis," BioPharm International 27 (12) 2014.