Ultrafiltration-diafiltration (UF-DF) sits at a critical junction in biopharmaceutical manufacturing, and real-time process analytical technology (PAT) in UF-DF is becoming essential to ensure control at this stage. UF-DF defines the final protein concentration, formulation composition, and, ultimately, product quality in the drug substance. However, many processes still rely on delayed, offline measurements, and that gap risks compromising the development of a robust purification process.
Expert Perspective: A Shift Toward Real-Time Monitoring
Apurva Godbole, a PAT specialist at Merck & Co., Inc., Rahway, NJ, USA, argues that the industry must shift toward real-time monitoring to unlock tighter control, stronger process understanding, and more efficient manufacturing. Her experience reflects a broader industry shift: PAT is no longer experimental. It is becoming central to the design and control of biological processes.
Real-Time PAT in UF-DF vs. Endpoint Testing Limitations
Traditional UF-DF control strategies rely on endpoint or offline testing. That approach leaves long periods where operators lack visibility into process performance. "When key process attributes are tested only at the end of the run, the process effectively operates without feedback for extended periods," Godbole explains.
Working in analytical science?
Register for a FREE Separation Science account to subscribe to the Separation Science Newsletter.
UF-DF steps often run for several hours. During that time, undetected deviations—such as membrane fouling, concentration drift, protein shear, or even protein breakthrough into the permeate—could occur. Fouling can reduce flux and distort concentration profiles, while breakthrough directly impacts yield and product recovery. "Instrumental or operational issues could also lead to breakthrough of protein into the permeate, directly impacting the process yield and revenue," she notes. When samples are only tested at the end, the opportunity to intervene has passed.
Offline testing does not fully solve the problem. Sample transport and lab turnaround times can introduce delays lasting hours. Additionally, piecing together data from individual measurements adds to turnaround time on the shop floor. "Traditional offline/end-point tests lead to a reactive process control with missed opportunities to intervene as opposed to a proactive control that could be enabled by real-time process data," asserts Godbole.
The takeaway is direct: without real-time data, UF-DF operates blind during its most critical phase.
Where Is Product Quality Most at Risk?
UF-DF simultaneously concentrates protein and exchanges buffer. That combination places several product quality attributes (PQAs) under stress.
Godbole highlights three key vulnerabilities:
- Protein concentration: Rapid concentration—often doubling within short timeframes—increases the risk of overshooting or undershooting targets and reduces the window for corrective action.
- Aggregation: Higher concentrations and process conditions can promote aggregate formation.
- Formulation integrity: Buffer exchange can alter buffer component levels due to phenomena such as the Donnan effect, potentially compromising protein stability.
"Protein concentration is the most vulnerable," she clarifies, because it defines the final deliverable of the process.
At the same time, charge-induced shifts in buffer composition during diafiltration can compromise stability. "The Donnan effect causes undesired loss of ions from the DF buffer into the permeate," explains Godbole. This shift can push buffer concentrations below target levels, directly impacting protein stability and formulation performance.
These risks do not act independently. They interact with process parameters such as flux, pressure, and buffer composition. That complexity reinforces the need for continuous monitoring rather than intermittent sampling.
Real-Time PAT in UF-DF: Tools and Technologies
PAT adoption in UF-DF continues to expand, with several technologies proving practical on the manufacturing floor.
Godbole points to the most established tools:
- UV spectroscopy: Widely used for inline and at-line protein concentration measurement.
- Refractive index (RI) and viscometry: Support process characterization.
- Online size-exclusion chromatography (SEC): Tracks aggregation trends.
- Raman and IR spectroscopy: Enable inline monitoring of buffer composition and progress of buffer exchange.
"UV spectroscopy has proven to be the most effective PAT tool to monitor protein concentration," she notes, citing its simplicity and ease of deployment.
However, challenges remain. "Accurately measuring low levels of excipients in the presence of an overwhelming protein signal remains a key challenge," shares Godbole, noting that this requires robust chemometric models to separate the two. That limitation keeps method development front and center.
The direction of travel is clear: more integrated, multi-attribute monitoring systems that provide a fuller picture of process state.
Real-Time PAT in UF-DF Enables Data-Driven Decision-Making
Real-time PAT in UF-DF changes how operators run UF-DF. Instead of waiting for results, they act on live data. As Godbole explains, "Real-time PQA data enables decision-making during the run."
This capability supports several operational improvements:
- Determining precise transition points between UF and DF based on live concentration data.
- Accurately monitoring process endpoints to avoid overshooting or undershooting concentration targets.
- Detecting abnormal trends (for example, rapid aggregation or flux decline) as they develop and intervening before irreversible damage to product quality or batch failure.
These actions reduce variability and improve batch success rates. They also shorten investigation timelines by providing continuous process context.
In practice, real-time data turns UF-DF into a closely monitored, data-driven control system.
PAT Versus Offline Assays: Defining the Right Role
Inline PAT does not replace offline assays in every scenario. Its performance depends on the application and the technology's maturity. "Accuracy and robustness vary with scale and intended use. In some cases, a PAT method may be the only way to monitor certain process attributes," Godbole points out.
She frames PAT roles across a spectrum:
- Monitoring: Lower accuracy requirements; supports trending and understanding.
- Decision-making: Higher expectations for reliability.
- Control: Requires strong validation and tight comparability to reference methods.
Another distinct advantage of PAT, compared to offline testing, is its utility for potent processing. Potent samples are inherently difficult to extract and analyze due to the rigorous safety precautions required to handle these molecules. In these processes, real-time PAT testing options may quickly shift from a convenience to a critical, "need-to-have" tool.
Offline testing still plays a role in specific cases:
- Root-cause investigations when deviations occur.
- Verification of PAT-identified trends or anomalies via validated reference methods when the primary sensor is unvalidated or utilized for monitoring only.
- Situations where PAT lacks the required accuracy or robustness for release decisions.
The industry goal remains clear—reduce reliance on offline testing—but full replacement depends on validation and regulatory acceptance.
Real-Time PAT in UF-DF: GMP Validation Challenges
Moving PAT into GMP manufacturing introduces technical and operational hurdles.
Godbole outlines the main barriers to implementing these real-time analytical technologies within a GMP workflow:
- Ensuring site readiness and instrument compliance.
- Validating method performance (linearity, specificity, accuracy, robustness).
- Aligning acceptance criteria with manufacturing sites and process teams.
- Training personnel for operation and maintenance.
- Modifying process skids to integrate inline sensors.
"PAT transfer to GMP manufacturing is a resource and time-intensive effort. However, it is well worth the effort when there are clear benefits," she asserts.
These implementations demand early planning. Delaying integration increases cost and complexity, especially when retrofitting existing systems.
Real-Time PAT in UF-DF Supports Enhanced In-Process Control and Process Understanding
The long-term value of PAT extends beyond immediate control. Continuous data unlocks deeper insight into process behavior.
Godbole highlights two major outcomes:
- Enhanced process understanding: "Real-time monitoring enables linking operating conditions to outcomes such as concentration, aggregation, or buffer composition," she advises. Especially with accelerated timelines for bringing drugs to market, PAT tools are becoming invaluable to rapidly generate process understanding at all scales.
- Real-time release potential: "Validated PAT tools could replace the offline testing to release the concentrated and buffer-exchanged protein to the next formulation step," she explains.
This knowledge supports better process design and faster development timelines. It also opens the door to automated in-process control and, eventually, the transition to a real-time release testing framework.
Practical Implementation Considerations for Real-Time PAT in UF-DF
Implementing real-time PAT in UF-DF requires more than selecting the right analytical tool. It demands alignment across process design, infrastructure, and operations.
Several factors shape successful deployment:
- Integration into existing systems: Inline PAT tools must be incorporated into the UF-DF flow path. This often requires skid modifications, which can trigger revalidation and impact timelines.
- Data strategy and use case definition: Teams must define whether PAT will support monitoring, decision-making, or active control. This determines required accuracy, validation depth, and regulatory expectations.
- Method development and modeling: Techniques such as Raman or IR require robust chemometric models to separate protein and excipient signals, especially at low concentrations.
- Operational readiness: Site personnel must be trained to calibrate, maintain, and troubleshoot PAT systems to ensure consistent performance.
- Workflow alignment: Real-time data must connect to clear decision points—such as when to transition from UF to DF or when to stop a run due to aggregation or flux decline.
"PAT implementation requires alignment between development and manufacturing teams to ensure the technology delivers reliable and actionable data," she notes.
Teams that address these factors early reduce implementation risk and accelerate adoption. More importantly, they position real-time PAT in UF-DF as a practical control tool rather than just an analytical add-on.
The Bottom Line
UF-DF remains a high-impact step in biologics manufacturing, making real-time PAT in UF-DF a critical capability for modern process control. Traditional testing approaches cannot keep pace with the need for control.
Real-time PAT offers a clear path forward by:
- Reducing blind spots during long processing times
- Enabling immediate intervention
- Strengthening process understanding
- Supporting robust in-process control and future real-time release strategies
The transition requires investment, validation, and cross-functional alignment. But the payoff is measurable: tighter control, fewer failures, and more efficient operations.
The industry now faces a simple choice—continue reacting to results, or start controlling processes as they happen.
