Controlling Integrated, Continuous Processes: Real-Time Monitoring with Feed-Back and Feed-Forward Controls Enables Synchronization and Enhances Robustness

Integration of continuous unit operations has the potential to reduce footprints, costs, and time associated with bioprocessing while also boosting yields and quality. Successful linking of upstream and downstream steps, however, cannot be achieved without applying enhanced process monitoring and control technologies. Complexity increases with the integration of multiple processes that have their own monitoring and control requirements. Thus, both biopharmaceutical manufacturers and developers of process analytical technologies (PATs) and control solutions are leveraging learnings from previous implementations to enable successful connection of continuous bioprocesses. Robust and user-friendly real-time analytics for monitoring and executing control strategies are under development to help users maximize productivity on integrated operations while maintaining safety and efficacy of biopharmaceutical products.

Up- and Downstream Integration

Many continuous-bioprocessing solutions are available for cell culture, purification, and fill–finish of biologic drug substances and drug products. Those individual continuous operations can be linked together to increase further the benefits of continuous manufacturing.

“The field of continuous, integrated biomanufacturing is slowly but surely coming into its own,” contends Carrie Mason (associate director of biologics research and development at Lonza Biologics). Upstream, increasing numbers of companies use perfusion N – 1 processes that couple cell growth with continual removal of waste products to achieve high inoculation densities in production reactors. For instance, continuous cell-culture growth can be integrated with alternating tangential-flow (ATF) filtration, says Adriano Leuzzi (head of process development at ReiThera).

Depending on the mode of operation (seed-train intensification, intensified fed-batch, concentrated fed-batch, or continuous perfusion), multiple upstream systems and auxiliary equipment can be connected, observes Boris Aleynik (director of strategy and product marketing, 908 Devices). Those can include automated media-preparation systems; seed and production bioreactors; and cell-retention devices, pumps, scales, and surge tanks to facilitate perfusion media exchange, cell bleeding, and/or continuous harvesting of biotherapeutic drug substances.

Downstream, integrated continuous chromatography systems for protein A affinity capture and ion-exchange polishing are linked with continuous viral inactivation and in-line concentration/diafiltration (DF) systems, according to Shunsuke Shiina (R&D scientist, AGC Biologics). Continuous multicolumn chromatography (CMCC) and periodic countercurrent chromatography (PCC) are good examples, Aleynik notes. “Continuous low-pH viral inactivation using plug-flow reactors or dual tanks, and single-pass tangential-flow filtration (SPTFF) enable the transition to continuous and semibatch downstream processing,” he adds. Continuous formulation and fill–finish systems, meanwhile, ensure consistent product blending to help optimize production efficiency while maintaining high product quality, Leuzzi says.

As an example, WuXi Biologics uses integrated, intensified perfusion cell culture, capture chromatography, and low-pH hold steps for monoclonal antibody (mAb) manufacturing — significantly enhancing productivity while reducing costs associated with protein A resin use. In a recent case, Mingyue Fang (director of the non-GMP pilot plant at WuXi Biologics) notes that a productivity of 105 g/L was realized in combination with a 92% reduction in resin use. Overall, 3 kg of mAb were produced using one 50-L single-use (SU) bioreactor.

The ultimate goal is to achieve full end-to-end integration, with upstream operations linked to downstream purification and fill–finish steps in one continuous process. However, continual product flow from upstream to downstream would require enhanced automation solutions, multilayered process orchestration, sophisticated model-based control systems, and advanced process analytical technology (PAT) for real-time monitoring and quality control (QC).

In the near term, integration of perfused production reactors with CMCC systems can allow for continuous capture of product in permeate streams, according to Mason. “In the long term, processes developed with an in-line concentration of different streams should enable reduced hold-up volumes between integrated unit operations, reducing the footprint required to operate continuous processes,” she says.

In the meantime, Harvey Branton (senior consultant for eXmoor Pharma Concepts) observes that it often makes sense to intensify a unit operation by running it continuously before switching to a batch approach to overcome challenges associated with smaller volumes. “The continuous process is run until the collection tank is full. The product is processed while another tank is filled,” he explains. “This approach may not meet the purist definition of end-to-end continuous [manufacturing], but it leverages technology when and where it adds value, and ultimately product is produced in a continuous manner.”

Platform and Modular Approaches

Integrated continuous bioprocesses can be set up using a platform or modular approach. Alex Sargent (director of process development at Charles River Laboratories) says that a single platform for integrated unit operations reduces difficulties with software and hardware standardization and integration because everything is controlled and captured by one system. The approach also improves process control, uniformity, and robustness. Full automation also reduces human variability in manufacturing.

Using a single platform, however, can limit flexibility, Sargent cautions. It can be more difficult to make products at larger scales, to collect more samples, and to integrate different sensors or analytics into platform workflows. “Using a more modular approach to continuous bioprocessing can be beneficial if such needs are anticipated,” he says. “Chaining together a wide array of different systems and technologies provides flexibility to support different processes and/or process changes.” That leads to a challenge of integrating hardware and software from different systems to ensure process control and uniformity.

Standardization: Controlling linked continuous bioprocesses presents challenges spanning regulatory compliance, software integration, hardware compatibility, and standardization, notes Giuseppina Miselli (process development senior scientist at ReiThera). In addition to a lack of standardization in consumables and the plethora of open and proprietary communication protocols that exist today, Aleynik at 908 Devices also highlights the limited availability of scalable SU continuous-processing systems and robust PATs, the sparse regulatory guidance, and lack of industry experience.

Sterility Concerns: Maintaining sterility also can be challenging in integrated continuous bioprocessing. Multiple connections are involved when different unit operations are connected. Each must be managed carefully to ensure sterility throughout each run, which can take several days or longer. Soyeon Ahn (principal scientist and senior director of downstream manufacturing at Samsung Biologics) says that in perfusion systems, for instance, connections are needed for addition of fresh medium to the bioreactor, transfer of cells into the perfusion system, transfer of process fluid to the ATF/TFF system, return of cells to the bioreactor, and so on.

It is thus important for fluid transfers to be functionally closed, states Dalip Sethi (commercial leader for cell-therapy technologies in North America at Terumo Blood and Cell Technologies). He adds that sterile welding technology is key to achieving that important goal.

Timing Challenges: Linking continuous processes together creates numerous challenges related to timing of activities. That includes input and output flow rates as well as analyses that enable process adjustments when needed. Maintaining a steady state between the output flow of a preceding unit operation and the input flow of the subsequent unit operation is primary, according to Fang at WuXi Biologics.

In a downstream example, if a continuous polishing step is linked with viral filtration and the chromatography operation is performed in bind–elute mode, then product elutes periodically with fluctuating protein concentration, pH, and conductivity values. During continuous viral filtration, however, consistent flow is needed to ensure that the filtration pressure is maintained within a set operating range. Periodic and uneven outflow from the chromatography step therefore would be problematic. WuXi Biologics has addressed that issue by introducing a surge tank(s) to mitigate differences in chromatography column output and input flow for viral filtration.

On the upstream side, Ahn at Samsung notes that keeping each process operating within optimal ranges also requires tuning of inlet/outlet parameters. “This goal can be achieved only by linking real-time values for key parameters within the bioreactor and media vessels,” she says.

Lonza’s Mason agrees that one of the largest technical challenges in controlling continuous bioprocesses with linked unit operations (regardless of the specific operation) is combining necessary automation to trigger process adjustments with timely measurements of critical process parameters (CPPs). “Replacement of off-line assays with like-for-like at-line and in-line PAT solutions that operators can use is critical to overcoming this gap,” she states.

Need for Adaptable, Real-Time Monitoring Technologies

In fact, effective integration of continuous processes hinges mainly on real-time monitoring capabilities. PAT tools must provide “instantaneous feedback on CPPs without disrupting operations, ensuring high accuracy, precision, and reliability over extended periods,” observes Miselli at ReiThera. PAT systems also should have comprehensive analytical capabilities, enabling multivariate analysis to monitor multiple parameters simultaneously and support seamless automation and integration with existing systems.

Ideally, sensors will not need to come in contact with the product stream, AGC’s Shiina cautions, or they should be able at least to withstand sterilization/sanitation. In-line PAT tools also need to be stable over a long period of use (>25 days) and be suitable for use a current good manufacturing practice (CGMP) environment, notes Mason. Sensors and instruments should resist fouling and ideally have built-in redundancy, Aleynik adds. Direct measurements are preferred over indirect ones with soft sensors. Where in-line PATs cannot be implemented readily, Mason advises companies to consider at-line solutions that are significantly less complex than traditional QC assays.

In some cases, speed is not a critical attribute. More important is the ability to track and predict the effects of process changes over time. Branton of eXmoor pharma points to perfusion systems running at a dilution rate of two vessel volumes per day (VVD). The effects of vessel perturbations will be spread over the following 12-hour period until a vessel returns to a steady state. Hourly measurements thus could be sufficient to indicate the need to increase the load volume on a chromatography column for maintaining a constant product amount. “Even more predictive capabilities can be achieved by running models in the background,” he says, “for example, to look at how different process parameters impact productivity. Yield then could be inferred, then confirmed using in-line/at-line or even off-line measurements.”

Robust data-handling capabilities are also crucial, Miselli notes. They facilitate efficient storage, retrieval, and analysis of large data volumes through advanced analytics and ensuring operational stability under diverse conditions.

Finally, given the many diverse biologic modalities in the pipeline today and the different unit operations involved in continuous bioprocessing, it is essential that monitoring and control technologies be highly adaptable. “The more adaptable the technology [is] to different systems,” Sargent observes, “the more powerful it becomes when integrated into different manufacturing workflows.”

Implementation Considerations

Start with a Good, Scalable Process: Although continuous processing offers cost and quality advantages, and integration of multiple continuous unit operations can magnify those benefits, a manufacturing process must be both robust and reliable before it can run in continuous mode. “In fact,” Branton says, “because continuous processes constantly generate product, the process often must be even more robust than its batch counterpart so that it can continue to function effectively in the face of many different possible problems.”

If integrated continuous processes are to be implemented at large scale, then scale-up must be considered during early development to ensure that the right materials, equipment, and PAT solutions will be available, adds Ahn. “If the right equipment and process monitoring and control technologies are not used, a process might run outside the design space, which can affect smooth transfer from one unit operation to another. Clogged filters and other problems ultimately can affect product quality — and, in the worst-case scenario, lead to stoppage of the entire process.”

Thoughtful Technology Selection: The optimum monitoring and control strategies for a given continuous-manufacturing process, whether integrated or not, depends on the nature of the drug substance and the bioprocess itself. “Manufacturers should use a risk-based approach to evaluate the suitability of a PAT tool for a particular process and its potential impact on product quality and process performance,” Aleynik suggests.

That starts with understanding a product’s requirements. Aleynik points to a number of factors to consider beyond standard analytical performance metrics (e.g., specificity, precision, accuracy, linearity), including cost and ease of implementation, calibration and maintenance requirements, process and automation system integrations, scalability, and ease of method validation. “In addition to providing real-time or near-real-time measurements, PAT solutions should offer robust, long-term, validated performance; support standard industrial communication protocols; be scalable from lab to commercial production; and be CGMP-compliance ready.”

Enabling Synchronization: Control strategies for integrated, continuous processes must ensure that upstream operations provide a consistent output to downstream operations. Therefore, as Mason advises, control solutions must be able to identify and adjust process feeds in response to detrimental conditions that arise during production. Feedback from downstream unit operations to upstream process engineers is equally important. In both cases, Sargent stresses, timely access to information needed to make decisions is paramount.

Alessia Noto (upstream supervisor at ReiThera) agrees. To prevent bottlenecks, she says, synchronization is vital to ensuring that the output from upstream processes matches the input needs downstream. “Maintaining process stability is crucial. Quality assurance through continuous monitoring and defect minimization ensures high-quality transitions between processes. Demand prediction helps optimize feed rates and resource planning to align with downstream requirements. Finally, effective feedback integration allows real-time adjustments and predictive control models to harmonize upstream and downstream operations.”

Local control loops for individual unit operations must be combined with global control strategies for interdependencies, says Aleynik. The control logic can range from empirically determined setpoints to advanced digital-twin–based models. PAT solutions should support multiple connectivity options with local and global automation systems. For example, connecting directly to a unit controller (e.g., programmable logic) over a serial or analog connection provides rapid signal transmission without much latency, qualities that are required for some types of operations. Open platform communications (OPC) unified architecture or other transmission control protocol/Internet protocol (TCP/IP)–based communication methods are slower but provide advanced security and diagnostics.

Ahn at Samsung highlights the importance of proper automation designs in establishing control solutions for integrated, continuous processes. “If the automation system is not perfectly set up, errors can occur, such as time delays or malfunctioning equipment. Such problems an disrupt the entire bioprocess operation.”

Process Analytical Technologies

Use of in-line PAT tools is increasing for process monitoring and control across many different integrated, continuous processes. Siyuan Tang (director of manufacturing, science, and technology, WuXi Biologics) says that basic parameters such as product concentration and volumetric flow rates of input and output streams can be monitored using ultraviolet and weight determinations (e.g., of surge tanks), respectively. CPPs such as temperature, pH, and dissolved oxygen (DO) also can be tracked using sensors to support process control. The utility of Raman spectroscopy is increasing for in-line analysis of many different product quality attributes (PQAs such as aggregates, fragments, charge variants, N-glycans, and posttranslational modifications) as well as for tracking metabolite production. “The integration of PAT and process-control strategies will enhance disturbance management in large-scale manufacturing, thereby increasing overall process robustness,” Tang says.

In upstream applications, advances in physical and chemical sensor technologies such as pH and optical DO probes have reduced the amount of time required for upkeep, Mason notes. In-line capacitance is applied widely for monitoring perfusion feed rates. Biosensors for monitoring glucose and lactate and in-line spectroscopy methods such as near infrared (NIR) are used as well. “These technologies offer real-time monitoring, high specificity, and enhanced process understanding,” says Noto. And Mason observes that they also reduce the need for operators to make process adjustments, which mitigates risk.

However, Noto cautions, many of those technologies often bring concerns such as high costs, complexity, frequent calibration, and maintenance needs. Applying in-line Raman, NIR, and other traditional spectroscopy-based analyzers requires upfront development of empirical calibration models, which necessitates extensive expertise and infrastructure. Aleynik says that 908 Devices has tackled that issue by offering an in-line Raman solution that uses a first-principles model, enabling out-of-the-box measurement of multiple process parameters (currently glucose, lactate, and total biomass) following a simple two-point calibration protocol

In addition to in-line PAT solutions, at-line versions (and off-line options, too) of more complex analytical technologies such as chromatography and mass spectrometry (MS) also are expanding for use in process monitoring of integrated continuous processes. They typically are used in combination with sterile autosamplers and generally are too slow for process control. But Aleynik notes that they can provide information on multiple process parameters and quality attributes simultaneously. One example from 908 Devices is an at-line capillary electrophoresis–MS (CE-MS) analyzer for monitoring concentration of proteins and media components including amino acids, vitamins, and biogenic amines. That ability can be leveraged to determine the optimal composition of perfusion media for high cell densities and sustained productivity in prolonged cultures, which is essential for implementation of implementation of cell-specific perfusion-rate (CSPR) control strategies.

Tang says that balances, flow-meters, and in-line sensors (for UV, pH, and conductivity) are used extensively in integrated, continuous downstream purification processes. Using real-time and near-real-time analytics, meanwhile, aligns with quality-by-design (QbD) principles, points out Hang Zhou (senior vice president and head of bioprocess research and development at WuXi Biologics): “The information gathered from these devices is collected through a distributed control system (DCS), facilitating communication and parameter adjustment across each unit operation. Interfacing with an automatic control system enables feedback control and facilitates predictive control. That in turn can enhance the robustness of integrated continuous processes. Indeed, fully continuous processes cannot be achieved without use of control solutions fed with data collected using multiple PAT tools.”

It is important to consider the value of automation in enabling control of integrated, continuous processes. “Implementing automation and robotics in bioprocessing can minimize human error, enhance process consistency, and provide better control over process parameters,” Zhou says. For instance, WuXi Biologics is developing automated analytical workflows for multiple PQA monitoring through multiple chromatographic, electrophoretic, and mass-spectrometric methods. “Automation will facilitate data collection and thereby expedite development of continuous bioprocesses.”

Several Emerging Technologies: Despite the numerous advances in process monitoring and control technologies, several unmet needs remain. “Technologies such as spectroscopy, flow cytometry, and MS for in-line and at-line monitoring are being investigated for their ability to provide rapid and non-destructive analysis of process samples,” Zhou says. “When these solutions are available, they will enable quicker decision-making and process adjustments.”

With its ability to provide detailed information, MS could improve the monitoring and control of integrated continuous processes significantly. Mason says that simplifying its use (e.g., by facilitating sample preparation and automated analysis) and technical advances for instrumentation that support using MS in a manufacturing environment would be important steps toward increasing its utility for bioprocess monitoring.

Zhou adds that multiattribute methods (MAMs), which couple MS with liquid chromatography for characterization and monitoring of multiple PQAs, are attracting much attention. Specifically, when compared with typical release assays, MAM approaches are the more informative, streamlined, and productive workflows for amino-acid sequence confirmation and product-variant identification and quantitation, with shorter turnaround times.

Shiina notes that Fourier-transform infrared (FTIR) spectroscopy is also being investigated for use in real-time monitoring of high-concentration products and buffer components in TFF/DF processes. “This technique can provide quantitative results very quickly,” he says. “The trick is to ensure measurements are made correctly.” And Miselli mentions that microfluidic devices in development could enable real-time polymerase chain reaction (PCR) assays for continuous process monitoring.

Branton says that the advent of software with sufficient processing power for image analysis is creating opportunities for real-time particle analysis. “These systems can process vast amounts of data, allowing the construction of models that can interpret even complex images.”

Recent technologies including artificial intelligence (AI) and machine learning (ML), digital twins, and next-generation sensors have been introduced for process monitoring and control. “These technologies are designed to enhance real-time monitoring, improve process optimization, enable comprehensive quality control, and facilitate predictive maintenance,” says Miselli. “We expect that they will ultimately see widespread adoption in the industry.”

Digital-twin technology, Zhou notes, can optimize process routing, identify potential issues in scaling up and other operations, and facilitate smooth and efficient process execution. Such applications hold great promise for optimizing and controlling integrated continuous bioprocesses.

Process control is advancing through systems such as model predictive control (MPC) and adaptive control. “They leverage mathematical models and algorithms to adjust key variables in real-time, enabling optimization of process performance, minimization of variability, and enhancement of process robustness,” observes Zhou.

Ultimately, Branton adds, such approaches move beyond just controlling bioprocesses by adding value through providing process engineers with improved understanding of integrated bioprocess systems — and thus helping them to understand the potential implications of changes made in one unit operation across the entire system. Such control strategies can be implemented without the need to redesign existing systems, he notes, by configuring current technologies in new ways.

Other emerging technologies with significant promise for improving the monitoring and control of integrated continuous bioprocesses, Leuzzi notes, include quantum sensors, lab-on-a-chip technologies, biosensor networks, bioluminescence/electrochemical sensor algorithms, and metabolic flux analysis tools. “These innovations aim to enhance sensitivity, provide real-time and distributed monitoring, enable adaptive control, and offer detailed insights into metabolic processes.” 

Cynthia A. Challener, PhD, is principal consultant at C&M Consulting in Morrisville, VT; 1-802-249-3862; [email protected]. She has been providing freelance writing services to the pharmaceutical industry for nearly 25 years, covering small-molecule, biologic, and advanced-therapy discovery, development, and manufacturing.