Online HPLC Bioprocess Monitoring: A Complete Guide to Real-Time CQA Measurement in Upstream Biopharmaceutical Production

A fed-batch mAb production run lasts 10–14 days. Every off-line HPLC sample you pull takes 2–6 hours to return a result. In that window, glucose depletes, aggregates accumulate, and the culture drifts in ways that a titer number from six hours ago cannot tell you. The batch either survives that blind spot or it doesn't — and you find out at harvest.

Online HPLC bioprocess monitoring closes that feedback gap by connecting chromatographic analysis directly to the production timeline.^[7] It qualifies under FDA's 2004 PAT Framework Guidance as a real-time measurement tool, provides the molecular specificity that spectroscopic methods lack for protein isoforms and sequence variants, and is now deployable at bench scale without the engineering burden it once required.

This guide covers the three monitoring modes (at-line, online, inline) and when each applies; which critical quality attributes HPLC can and cannot measure; how automated sampling integration works step by step; what miniaturized HPLC systems change about deployment economics; and how HPLC fits into a complete PAT toolkit for monoclonal antibody production.

What Is Online HPLC and How Is It Used for Real-Time Bioprocess Monitoring?

Online HPLC, in the chromatographic sense, means the LC system is connected to and fed by the process stream — not that it operates over a network. An automated aseptic sampling interface withdraws process fluid from the bioreactor at programmed intervals, conditions the sample if needed (filtration, dilution), and injects it directly into the LC without manual handling. Results return in minutes. The instrument then cycles again.

The enabling hardware is the sampling interface, not the HPLC itself. A highly capable analytical HPLC sitting on a bench two rooms away from the reactor is always off-line regardless of its performance specifications — because someone has to carry the sample to it. Without automated sample withdrawal and transfer, you have off-line analysis with a short walk. This distinction matters when scoping a PAT implementation: instrument performance is rarely the limiting factor; sampling architecture almost always is. Budget and timeline accordingly.

FDA's PAT Framework Guidance (2004) defines PAT as "a system for designing, analyzing, and controlling manufacturing through timely measurements of critical quality and performance attributes.^[1]" HPLC qualifies because it provides quantitative, chemically specific measurements in a timeframe that can inform in-process control decisions. It sits alongside Raman spectroscopy, NIR, and mass spectrometry as one of the primary PAT analytical tools — but unlike spectroscopic methods, it physically separates analytes before detection, which is the key advantage for structurally similar species.^F In practical terms, this means HPLC is not interchangeable with Raman or NIR for protein quality attributes: for charge variants or aggregate fractions, it is typically the only tool that will give you a defensible number.

How an Online HPLC System Connects to a Bioreactor

The connection requires an aseptic sampling port on the bioreactor equipped for automated actuation — typically a steam-in-place (SIP) valve, single-use diaphragm valve, or a dedicated sampling probe. Sampled fluid travels through temperature-controlled transfer tubing (minimizing dead volume and preventing protein precipitation) to a sample conditioning step, then to the HPLC injection port. The conditioning step is where most implementations hit their first technical decision: inline filtration adds dwell time but preserves sample integrity; dilution is simpler but reduces sensitivity for low-abundance analytes. For impurity monitoring at sub-percent levels — aggregates, fragments, charge variants — inline filtration is generally the safer choice; the sensitivity cost of dilution becomes significant precisely where your most critical measurements occur.

What Are the Differences Between At-Line, Online, and Inline HPLC Monitoring in Bioprocessing?

This comparison gets handled poorly in most vendor literature — "online" and "at-line" are frequently used interchangeably, which obscures meaningfully different automation levels and engineering requirements.

Inline HPLC is not currently practical for bioreactor applications. Flow cell geometry, pressure requirements, and sterility validation create technical barriers that have not been solved commercially. Some vendor materials imply inline LC is on the near-term horizon for bioprocessing; it is not.

Online HPLC delivers the highest monitoring resolution but requires the most engineering investment: validated sampling hardware, aseptic transfer lines, and often DCS or historian data integration. At-line HPLC — where the instrument sits adjacent to the reactor and samples are drawn discretely, either manually or via a semi-automated module — is the most common practical implementation at development scale. Axcend's InFocus module is designed specifically for this at-line discrete sampling configuration, sitting directly adjacent to a reaction vessel without requiring DCS integration.

Which Monitoring Mode Is Right for Your Upstream Process?

Choose online if you are running commercial or late-phase production, CQA deviations carry batch failure risk, and you have engineering resources to validate the sampling interface. Choose at-line if you are in process development, monitoring frequency of 30–60 minutes per sample is sufficient, and you need results without a six-month integration project. Off-line remains appropriate for reference testing and low-criticality CQAs. If you are early in development with no existing PAT infrastructure, at-line is the right starting point: it generates the process understanding dataset that justifies online investment at scale, while remaining achievable with a single instrument and a semi-automated sampling module.

What Critical Quality Attributes Can HPLC Monitor in Real Time During Upstream Bioprocessing?

HPLC covers more ground than most process engineers expect — but it also has genuine limits worth stating honestly.

CQAs measurable by HPLC in upstream production:

1. Monoclonal antibody titer — Protein A affinity HPLC or reverse-phase C4/C8 methods; most commonly targeted CQA for fed-batch mAb production. Enables data-driven harvest timing.

2. Protein aggregates — SEC-HPLC; aggregate accumulation during production can signal cell stress before yield is affected. Catching aggregate accumulation mid-run rather than at harvest gives you the option to adjust feeding strategy, temperature, or dissolved oxygen before quality is irreparably compromised.

3. Charge variants — IEX or CEX-HPLC; monitors acidic and basic species critical to mAb potency and stability.

4. Antibody fragments — Reverse-phase HPLC; detects fragmentation products that indicate proteolytic activity or process instability.

5. Amino acids and selected metabolites — Derivatization HPLC methods; complements inline Raman for more specific metabolite profiling.

6. Peptide and oligonucleotide intermediates — For RNA therapeutics, gene therapy vectors, and fermentation-derived peptide APIs; ion-pair reverse-phase HPLC methods are well-established.

7. Small molecule APIs in microbial fermentation — UV or MS detection; applicable where the product is a defined small molecule.

HPLC vs. Spectroscopic PAT Tools: When Separation Matters

Raman and NIR can measure glucose, lactate, and osmolality continuously and inline without touching the culture. HPLC cannot match that on speed or convenience for bulk metabolites. Where HPLC is irreplaceable is molecular specificity for species that co-elute spectroscopically — charge variants differing by a single amino acid modification, aggregate fractions at 1–2% of total protein, fragmented species at sub-percent levels. If your CQA can be estimated by Raman, use Raman. If your CQA requires structural resolution, HPLC is the only routine option. In practice, this means the two approaches are complementary rather than competitive: Raman handles the continuous metabolite baseline, and HPLC handles the periodic structural quality checkpoints that Raman cannot provide.

How Is Real-Time Titer Measurement Performed During Upstream Biopharmaceutical Production?

Real-time titer measurement bioproduction workflows most commonly use Protein A affinity columns (e.g., MabSelect, POROS A series) for mAb quantitation. The antibody binds selectively to the Protein A ligand, is eluted under acidic conditions, and is quantified by UV absorbance at 280 nm. Run times are typically 5–15 minutes per injection, allowing four to twelve titer measurements per hour during continuous operation.

The automated cycle: aseptic valve opens → defined sample volume withdrawn → optional dilution or inline filtration → injection onto HPLC column → UV detection → peak area integrated → concentration calculated against calibration curve → time-stamped result written to CDS and process historian.

The output is a titer-versus-time profile across the entire production run — not isolated data points from manual sampling. That profile reveals peak expression timing, productivity trends, and deviations that a scatter plot of six-hour-delayed off-line results cannot show. Specifically, it gives you the ability to make harvest timing decisions based on where you are on the titer curve rather than where you were six hours ago — a meaningful difference when in some optimized or short duration cell lines, the productivity plateau can be as short as 12–24 hours.

One frequently overlooked variable: sample bleed volume. Each automated sampling cycle withdraws a defined volume from the bioreactor. For commercial 2,000 L bioreactors, this is inconsequential. For 500 mL to 2 L development reactors, it is not. Miniaturized capillary HPLC systems operating at nanoliter injection volumes — such as Axcend's Focus LC®, which supports 4–40 nL fixed-loop injections — reduce the volume withdrawn per cycle by one to two orders of magnitude compared to standard analytical HPLC, which matters when the reactor itself holds less than a liter. At analytical HPLC injection volumes, cumulative sample removal across a 14-day monitoring run from a 1 L bioreactor can represent a meaningful fraction of the culture volume; at capillary scale, it does not.

When HPLC data is captured under a CDS with audit trail capability and 21 CFR Part 11 compliance, real-time titer measurements can contribute directly to the batch record, supporting RTRT pathways and reducing post-production testing burden.

How Can HPLC Be Integrated with Automated Bioreactor Sampling Systems?

Bioreactor sampling automation for chromatography integration follows a consistent sequence regardless of HPLC platform. The complexity lives in the interfaces between steps, not in the HPLC itself.

Step 1: Aseptic sampling valve selection. The bioreactor requires an aseptic sampling port compatible with automated actuation — SIP valves, single-use diaphragm valves, or dedicated sampling probes (e.g., SecureCell Numera). This is typically the longest lead-time engineering item and should be specified before instrument selection begins. Valve selection also determines whether you can transition from at-line to online monitoring later without bioreactor modification — worth confirming upfront if online deployment is on the roadmap.

Step 2: Sample conditioning. Bioreactor fluid contains cells and debris incompatible with most HPLC columns. Options: inline hollow-fiber or flat-membrane filtration (adds 2–5 minutes dwell time; acceptable for most methods), or mobile phase dilution (simpler, but reduces sensitivity for low-abundance analytes). Inline centrifugation provides cleaner samples but adds significant capital and engineering cost. For most development-scale implementations, inline hollow-fiber filtration is the practical default: it adds manageable dwell time and protects column lifetime without the complexity of centrifugation.

Step 3: Transfer line design. Minimize tubing length and internal volume between the sampling valve and HPLC injection port. Use PEEK or PTFE; temperature-control the line if protein precipitation is a risk. Transfer line dead volume directly delays the measurement feedback loop. Every additional minute of transfer delay is a minute subtracted from your window to intervene when a CQA deviates — keep lines as short as physically achievable.

Step 4: HPLC injection and method configuration. The sample delivery is configured based on your sampling setup. For discrete samples that can be delivered into standard vials, an autosampler like the AutoFocus can be configured to handle unattended sequences. For continuous at-line monitoring directly adjacent to a reactor, a sampling device like the InFocus interface can take pre-filtered samples and introduce them into the Focus LC at pre-determined intervals. The Focus LC's low flow rates make it well-suited for coupling directly to a mass spectrometer positioned at the point of need, providing confirmatory identification without returning samples to a central lab.

Step 5: Data capture and historian integration. HPLC output must connect to a process data historian for time-stamped CQA trending. Integration paths include OPC-UA output from the CDS, CSV export to a PI historian, or LIMS middleware. This step is where most implementation projects run late — it is an IT and automation engineering task, not an analytical chemistry task. Scope it as such. Engaging your IT and automation groups at Step 1 rather than Step 5 is the single most reliable way to keep a PAT implementation on schedule.

Step 6: System qualification. GMP environments require IQ, OQ, and PQ before HPLC data can be used for batch disposition. IQ/OQ packages vary significantly by vendor; confirm availability before committing to a platform for regulated production use. For development-scale or IND-stage implementations where GMP disposition is not yet required, this step can be deferred — but documenting instrument performance from the start will reduce the qualification burden when GMP deployment follows.

What Is the Minimum Engineering Requirement for At-Line HPLC Monitoring?

At-line HPLC monitoring is achievable without DCS integration, SIP valve installation, or historian connectivity. A semi-automated sampling module, a miniaturized HPLC system on the bench adjacent to the reactor, and a standard CDS with manual data export constitutes a functional at-line PAT implementation suitable for process development and IND-stage work. Full automation is the destination, not the starting point. Starting with this minimal configuration also gives you real process data to justify the engineering investment in full online integration — making the business case with evidence rather than projections.

What Advantages Does Miniaturized HPLC Offer for Upstream Bioprocessing Analytics?

Miniaturized HPLC biopharmaceutical manufacturing applications have grown as capillary-scale instruments have become more robust and operationally straightforward. The advantages relative to standard analytical HPLC are concrete and specific:

1. Reduced sample bleed volume. Capillary HPLC injection volumes of 4–1,000 nL versus 1–50 µL for analytical HPLC reduces per-cycle sample withdrawal by one to two orders of magnitude — critical for development-scale bioreactors where cumulative sample removal affects culture viability.

2. Dramatically lower solvent consumption. Capillary flow rates of 0.4–10 µL/min versus 0.5–2 mL/min analytical — 100–1,000x less solvent per run. A monitoring workflow running 80 injections over a 14-day production run consumes milliliters of mobile phase rather than liters, reducing procurement, disposal, and safety compliance costs materially. For labs running continuous bioprocess monitoring campaigns across multiple bioreactors simultaneously, this reduction in solvent volume also meaningfully lowers fire risk and chemical exposure, which matters in process development environments where biosafety cabinets and fume hood space are already constrained.

3. Compact footprint for at-line deployment. A miniaturized HPLC system measuring ~9x15x17 in (~23x38x43cm) can fit on a crowded process development bench or inside a biosafety cabinet directly adjacent to the bioreactor. Standard analytical HPLC systems occupy between 24-39 in (60-100cm) in width x 20-27 in (50-70cm) in depth and 24-39 in (60-100cm) in height of dedicated bench space plus the separate PC and detector hardware. In practice, this means the instrument can live where the process is rather than requiring the sample to travel to a dedicated analytical lab — which is the difference between true at-line monitoring and off-line analysis with a short walk.

4. Lower operating cost with documented payback. Solvent reduction exceeding 99% translates to measurable consumable savings per run. Capillary HPLC systems with these operational characteristics have been documented to achieve payback within approximately two to three years from operational savings alone — well within the typical 5–10 year equipment lifecycle.

5. Native ESI-MS compatibility. Capillary flow rates (0.4–10 µL/min) sit in the optimal range for electrospray ionization without post-column splitting — simplifying LC-MS method development for titer and impurity monitoring applications where mass confirmation is needed. This eliminates the post-column splitter hardware, reduces ionization suppression from excess solvent, and lowers limits of detection compared to split-flow configurations from analytical HPLC — a meaningful advantage for low-abundance impurity monitoring.

Method Transfer Considerations: What Changes When Moving from Analytical to Capillary HPLC?

This is the non-obvious limitation that matters most operationally: capillary HPLC is not a drop-in replacement for a method developed at 4.6 mm column format. Four variables require specific attention during method transfer — dwell volume (the Focus LC has ~1.9 µL dwell volume, affecting gradient timing), gradient delay, injection volume (from µL to nL scale), and column equilibration time. For labs without capillary LC experience, this represents real method development work. In Axcend’s experience adapting validated USP monographs to capillary scale under <621> guidelines, method conversions typically complete in one to four weeks depending on method complexity.^[9] Factor it into project timelines. The investment is a one-time cost per method; once transferred and validated at capillary scale, the method runs with the same specificity and quantitative performance as the analytical format — with all the operational advantages described above.

How Does Process Analytical Technology (PAT) Enable Real-Time Control in Upstream Bioprocessing?

The regulatory framework matters here because it determines what online HPLC data can actually do for your process — not just what it measures.

FDA issued its PAT guidance framework in 2004: *"Guidance for Industry: PAT — A Framework for Innovative Pharmaceutical Development, Manufacturing, and Controls."* ICH Q8 (Pharmaceutical Development) and Q10 (Pharmaceutical Quality System) build on these principles and apply directly to biopharmaceutical upstream processes.^[3] The core premise: understand and control the process rather than testing quality into the product after the fact.

What HPLC process analytical technology specifically enables under this framework:

• Real-time release testing (RTRT): Validated PAT data collected during manufacturing can substitute for end-product testing, reducing batch release time from weeks to days.^[8] This is one of the most concrete regulatory returns on PAT investment — if your HPLC monitoring data is validated and audit-trailed under a CDS with 21 CFR Part 11 compliance, it can directly replace post-production release testing for the CQAs it covers, compressing the time between production completion and product release.
• In-process deviation detection: Real-time CQA monitoring catches aggregate accumulation, titer plateau, or impurity accumulation during the run — enabling intervention rather than batch failure post-mortem.
• Design space definition: PAT data from development runs populates the multivariable design space described in ICH Q8, directly supporting process validation documentation. The more HPLC data points you collect during development, the more tightly you can define acceptable operating ranges — which gives you more flexibility to justify process changes post-approval without triggering major variation submissions.
• Continuous process verification (CPV): FDA's 2011 Process Validation Guidance requires Stage 3 ongoing data collection; real-time HPLC monitoring contributes directly to CPV programs.^[4]
• Regulatory filing support: PAT data strengthens CTD Module 3 content for BLAs and supports post-approval change management with a documented process understanding foundation.

One honest reframe worth stating: full PAT implementation including online HPLC is not required by FDA. It is encouraged, and rewarded with regulatory flexibility — reduced in-process testing requirements, faster release pathways, stronger change control justifications. Many companies deploy at-line HPLC as a development PAT tool without it appearing in the filed control strategy. That is a legitimate and common approach, not a compromise. The practical implication: deploying at-line HPLC during development costs less and requires less integration than online production monitoring, but the process understanding it generates directly supports the filed control strategy and can be cited in the BLA even if the PAT tool itself is not part of the commercial control system.

What PAT Tools Are Used for Monitoring Monoclonal Antibody Production in Real Time?

Online HPLC monoclonal antibody process control is one part of a multi-tool monitoring strategy. Understanding where each tool fits prevents over-engineering and under-measurement.

1. Raman spectroscopy — Continuous inline measurement of glucose, lactate, glutamine, and osmolality. No sample removal. Does not resolve protein isoforms or variants. Widely deployed in commercial mAb manufacturing as the primary metabolite monitoring tool.

2. Near-infrared (NIR) spectroscopy — At-line or inline; metabolite and osmolality estimation. Requires robust calibration models sensitive to culture variability; recalibration across cell lines adds operational overhead. NIR calibration models trained on one cell line or media formulation often perform poorly on another — account for recalibration effort when deploying NIR across a multi-product facility.

3. HPLC (Protein A, SEC, IEX, CEX) — At-line or online; the only routine PAT tool capable of separating and quantifying charge variants, aggregates, and fragmented antibody species with chromatographic specificity. Run times 5–30 minutes per injection. Irreplaceable for structural isoform monitoring.

4. Capacitance/impedance probes — Inline continuous viable cell density measurement. Provides the cells/mL context essential for interpreting HPLC-derived specific productivity calculations. Without viable cell density data, a rising titer number tells you the culture is producing — but not whether it is doing so efficiently or masking a decline in cell health.

5. Dissolved oxygen and pH probes — Foundational inline sensors; not analyte-specific but universally required.

6. Mass spectrometry (LC-MS or direct infusion) — Highest specificity for PTM monitoring, sequence variants, and metabolomics. High cost and complexity currently limit use to development scale. Where LC-MS is deployed at development scale, capillary HPLC's native ESI-MS compatibility becomes a direct advantage — no post-column splitting means better signal, lower limits of detection, and simpler method setup compared to coupling a standard analytical HPLC to a mass spectrometer.

7. BLI/SPR biosensors — Biolayer interferometry or surface plasmon resonance for rapid titer measurement; complementary to or competitive with Protein A HPLC depending on throughput and specificity requirements.

Based on published industry surveys, most commercial mAb monitoring programs use two to four of these tools in combination. HPLC is most consistently the primary tool for aggregate and charge variant monitoring, where its separation specificity cannot be matched by any spectroscopic or biosensor alternative.^[10] Programs that underperform on CQA coverage typically relied on Raman alone and assumed spectroscopic metabolite tracking was equivalent to structural protein monitoring. It is not. The consequence is discovering aggregate accumulation or charge variant drift at harvest rather than mid-run — when intervention is no longer possible and batch failure is the only outcome.

Conclusion

At-line and online HPLC monitoring is technically achievable at bench and development scale today. The engineering barriers — aseptic sampling, sample conditioning, data integration — are real but manageable, particularly for at-line configurations where discrete semi-automated sampling replaces the need for continuous SIP valve cycling and DCS integration. Miniaturized capillary HPLC systems have lowered the footprint, solvent, and sample volume barriers enough that a development lab can deploy meaningful at-line monitoring without dedicated analytical infrastructure.

The decision to move from off-line to at-line HPLC monitoring in upstream bioprocessing is less a capital decision than a workflow design decision — and making it earlier in development generates the PAT dataset that strengthens late-phase submissions and commercial process validation. Labs that begin collecting real-time CQA data at bench scale have a concrete process understanding advantage when they reach the regulatory submissions that require it; labs that wait until late-phase or commercial scale pay for that delay in engineering time, compressed timelines, and thinner data packages.

If you are evaluating at-line HPLC integration for a cell culture or fermentation process, Axcend's application team can review your CQA targets and run requirements to determine whether capillary-scale HPLC is a fit — and what method transfer would involve. Schedule a technical conversation with the Axcend applications team.

Explore the products referenced in this guide:

• InFocus at-line process interface— discrete at-line sampling for reaction monitoring and PAT applications
• Focus LC capillary HPLC system— the core instrument for miniaturized bioprocessing analytics
• AutoFocus automated sample introduction — including the only commercially automated direct-infusion MS workflow on the market

Frequently Asked Questions

Q: What is online HPLC and how is it used for real-time bioprocess monitoring?

Online HPLC connects a liquid chromatography system to a bioreactor via an automated aseptic sampling interface. Process fluid is withdrawn at programmed intervals, conditioned if needed, and injected automatically into the LC. Results — typically titer, impurity concentrations, or metabolite levels — return in minutes rather than the 2–6 hours required for off-line lab analysis, enabling data-driven process control during the run.

Q: What is the difference between at-line, online, and inline HPLC monitoring in bioprocessing?

Inline monitoring places a sensor inside the bioreactor — not currently practical for HPLC hardware. Online HPLC connects automated aseptic sampling directly to the instrument for near-continuous measurement with minimal operator intervention. At-line HPLC uses discrete semi-automated or manual sampling at an adjacent instrument, with 15–60 minute turnaround. Off-line requires transporting samples to a separate lab, returning results in 2–6 hours.

Q: What critical quality attributes can be monitored in real time by HPLC during upstream bioreactor operation?

HPLC can monitor mAb titer via Protein A affinity methods, protein aggregates by SEC-HPLC, charge variants by IEX or CEX-HPLC, antibody fragments by reverse-phase HPLC, amino acids and selected metabolites, and peptide or oligonucleotide intermediates.[6] Its advantage over spectroscopic PAT tools is molecular specificity — it separates and quantifies structurally similar species that NIR and Raman cannot resolve.

Q: How is real-time titer measurement performed during upstream biopharmaceutical production?

Automated sampling withdraws a small volume from the bioreactor at set intervals. The cell-free sample is injected onto a Protein A affinity HPLC column; the antibody binds and is eluted under acidic conditions, detected by UV at 280 nm. Run times are typically 5–15 minutes per injection. The resulting concentration value is time-stamped and logged to the process historian, building a continuous titer profile across the production run.

Q: What advantages does miniaturized HPLC offer for upstream bioprocessing analytics compared to traditional systems?

Miniaturized capillary HPLC operates at 0.4–10 µL/min flow rates and nanoliter injection volumes — 100–1,000x less solvent than standard analytical HPLC. For bioprocessing applications: reduced sample bleed per cycle (critical for small-volume development bioreactors), compact footprint suitable for placement adjacent to bench-scale reactors, lower consumable and waste costs, and native ESI-MS compatibility without post-column splitting. Method transfer from analytical HPLC format is required and should be planned for.

References

1. U.S. Food and Drug Administration. (2004). *Guidance for industry: PAT — A framework for innovative pharmaceutical development, manufacturing, and controls*. U.S. Department of Health and Human Services.

2. International Council for Harmonisation. (2009). *ICH Q8(R2): Pharmaceutical development*. ICH Harmonised Tripartite Guideline.

3. International Council for Harmonisation. (2008). *ICH Q10: Pharmaceutical quality system*. ICH Harmonised Tripartite Guideline.

4. U.S. Food and Drug Administration. (2011). *Guidance for industry: Process validation — General principles and practices*. U.S. Department of Health and Human Services.

5. Fonteyne, M., Vercruysse, J., De Leersnyder, F., Van Snick, B., Vervaet, C., Remon, J. P., & De Beer, T. (2021). Process analytical technology for continuous manufacturing of solid-dose pharmaceutical products: A review. *Journal of Pharmaceutical and Biomedical Analysis, 198*, 114049. https://doi.org/10.1016/j.jpba.2021.114049

6. Gilardoni, M., & Regazzoni, L. (2022). *Liquid phase separation techniques for the characterization of monoclonal antibodies and bioconjugates*. Journal of Chromatography Open 2 (2022) 100034

7. Gyorgypal, A., & Chundawat, S. P. S. (2021). Development and application of real-time process analytical technology for biopharmaceutical manufacturing: A review. *Current Opinion in Biotechnology, 71*, 145–154. https://doi.org/10.1016/j.copbio.2021.07.016

8. U.S. Food and Drug Administration. (2004). *Guidance for industry: PAT — A framework for innovative pharmaceutical development, manufacturing, and controls*. U.S. Department of Health and Human Services.

9. Axcend Application Note. "Adapting Pharmaceutical Monographs for Capillary Scale HPLC Using USP <621>" (hydrochlorothiazide), 2025.

10. Glassey, J., Gernaey, K. V., Clemens, C., Schulz, T. W., Oliveira, R., Striedner, G., & Mandenius, C. F. (2011). Process analytical technology (PAT) for biopharmaceuticals. *Biotechnology Journal, 6*(4), 369–377.