ISPE 博客分析生物制品配制中气液界面应作为工艺参数记录

iSpeak Blog

7 October 2026

A biologic drug product compounding step often transfers seamlessly on paper. The formulation is unchanged, the order of addition is the same, and the target concentration remains identical. The batch record at the receiving site calls for the qualified mixing speed and mixing time, and that is what the operator follows. Everything appears to match.

Then the subvisible particle counts at the receiving site consistently settle a little higher than they ever did in development. They remain within specification, but the shift is noticeable and persistent.

The investigation works through the usual suspects: the sterilizing filter, the tubing set, silicone from the container closure system, and the incoming drug substance. Sometimes one of them provides the answer. Often none of them does, and the investigation concludes with no assignable cause identifies, while the product conforms.

The air-liquid interface is doing the damage. It changes when you change scale. And it is the one condition in the compounding step that is not writen down.

1. The Interface, Not the Impeller, Does the Damage

Proteins are surface active. At an air-water interface a monoclonal antibody adsorbs and then partly unfolds, because its hydrophobic core is more comfortable facing air than facing water. Disturb that film and the unfolded molecules return to the bulk, where they act as nuclei for aggregation.1 Every time the surface is renewed a fresh interface is made and the cycle is run again: a vortex folding air down, a splash, a stream falling from an addition port into the bulk below.

It is worth being precise about how one-sided this is. Two concentrated IgG1 formulations were once held at shear rates between 20,000 and 250,000 per second, far above anything a compounding impeller generates, and shear alone did not cause aggregation.2 The same authors put the force available at an air-water interface at roughly 140 piconewtons, against the 20 to 150 piconewtons needed to unfold a protein, and concluded that air-bubble entrainment and adsorption to surfaces matter far more in production than shear does.

Polysorbate protects largely by getting to the surface first and holding the space, a competitive adsorption that has now been mapped directly at the air-water interface.3 That is why surfactant level and interfacial area are coupled, and why a formulation comfortably protected at one scale can sit closer to the edge at another with no change to the formulation at all.

2. No Scale-Up Rule Holds the Surface Constant

How far a free surface deviates from flat is governed by the Froude number, Fr = N²D/g, where N is impeller speed and D is impeller diameter.4 Under geometric similarity, each of the scale-up rules in common use moves it in a different direction.

Figure 1. What each scale-up rule does to the free surface. Holding revolutions per minute (RPM), power per volume, or tip speed constant each moves the Froude number, and with it vortexing, surface renewal, and air entrainment. Only scaling on Froude number itself conserves the interface, and almost nobody scales that way.

Take a 200 L development vessel and a 2,000 L commercial vessel: ten times the volume, about 2.15 times the diameter. Carry the same RPM across, and the Froude number is multiplied by 2.15. Match tip speed instead and it is divided by 2.15. Two entirely defensible engineering choices land a factor of 4.6 apart in how the surface behaves. Neither one is wrong. Neither one is recorded either.

The direction is often the opposite of the intuition. Scaling on constant power per volume or constant tip speed gives a calmer surface in the larger vessel, which means the development data may have been generated under more interfacial stress than the commercial process will ever see. The product then looks better at scale than it did in the laboratory. That is a happy accident, not a control strategy, and it works until the day it does not—a receiving site that carries the RPM across unchanged, or a large vessel run near the bottom of its fill range where impeller submergence drops and the vortex reaches down toward the blades.

3. The Batch Record Has Nowhere to Record It

Look at where surface condition actually lives in a compounding transfer package. Impeller speed is recorded. Mix time is recorded. Vessel volume is recorded. Fill level is usually a range, and often a wide one. Vortex depth is nowhere. Minimum impeller submergence is nowhere. Whether the polysorbate stock free-falls into the bulk or runs down the vessel wall is nowhere. Whether the operator watched air fold into the surface for twenty minutes is nowhere.

None of these are difficult measurements. Vortex depth is something an operator can see through a sight glass and describe in a sentence. ISPE and PDA both publish detailed guidance on what a technology transfer package should carry,5^, 6 and the gap here is not really in the guidance. It is that surface condition was never classified as a process parameter in the first place, so no field exists to hold it.

The cost of that shows up later. When the receiving site's particle counts run high, there is no record of what the surface looked like at either site. The comparison cannot be made, even in hindsight, and the investigation has no choice but to close unresolved.

4. What to Write Down Instead

Four additions, none of which need new equipment or new science:

• A free-surface acceptance state: "No visible vortex at the qualified speed across the full fill range," or "vortex depth not to exceed 10 percent of liquid depth." Verify it at both sites, on the actual vessel, at both fill extremes rather than at the setpoint alone.
• A minimum impeller submergence, tied to fill volume: A large vessel run at low fill is a different mixer from the same vessel run full. Bound the fill range and state the reason for the bound, so the receiving site knows it is a constraint and not a convenience.
• The addition method, not just the addition order: Sub-surface addition, addition down the wall, and free fall into the bulk create three different interfaces. Transfer packages are careful about sequence and silent about geometry.
• The scale-up rule you used, and what it did to the Froude number: One line—If tip speed was matched, say so, and say the Froude number fell by roughly half. It costs nothing to write and it tells the receiving site exactly what changed.

What to Take Away

• Interfacial stress, rather than bulk shear, drives most compounding-induced aggregation in protein drug products
• No commonly used scale-up approach preserves the free surface. Regardless of the parameter held constant, the Froude number changes
• At a ten-fold change in volume, maintaining the same RPM and constant tip speed differ by a factor of about 4.6 in Froude number
• Vortex depth, impeller submergence, and addition geometry are observable, inexpensive to record, and absent from most technology transfer packages
• Fill range is as critical as the operating setpoint. Process qualification should challenge the extremes, not just the nominal condition

Conclusion

Technology transfer packages are good at carrying settings and poor at carrying conditions. Mixing is where that gap is widest, because the setting is impeller speed, the condition is what the surface does, and the two are related by a vessel geometry that changes at precisely the moment of transfer.

Fixing it does not require new science or new instruments. It may be as simple as adding a single line to the batch record and documenting one observation through a sight glass, made at both sites and then comparing the results. The alternative is to uncover the discrepancy during process performance qualification, which is arguably the most expensive stage to learn anything.

About the Author

Sr. Manager, External R&D and Technical Services

Alovgen Inc.

Bhasker Sambar is a pharmaceutical scientist and Senior Manager of External Research and Development and Technical Services at Alvogen Inc., where he leads complex drug...

References

• 1

  van Haaren, C., B. Byrne, and S. G. Kazarian. "Study of Monoclonal Antibody Aggregation at the Air–Liquid Interface under Flow by ATR-FTIR Spectroscopic Imaging." Langmuir 40, no. 11 (2024): 5858–5868. https://doi.org/10.1021/acs.langmuir.3c03730

• 2

  Bee, J. S., J. L. Stevenson, B. Mehta, J. Svitel, J. Pollastrini, R. Platz, E. Freund, J. F. Carpenter, and T. W. Randolph. "Response of a Concentrated Monoclonal Antibody Formulation to High Shear." Biotechnology and Bioengineering 103, no. 5 (2009): 936–943. https://doi.org/10.1002/bit.22336

• 3

  Thompson, B. R., K. G. Pham, M. D. Phan, P. Sanchez-Puga, P. Gutfreund, T. Wang, K. K. Qian, F. Heinrich, Y. Liu, and N. J. Wagner. "Competitive Adsorption of Monoclonal Antibodies and Nonionic Surfactants at the Air–Water Interface." ACS Applied Materials & Interfaces 17, no. 28 (2025): 40116–40128. https://doi.org/10.1021/acsami.5c07162

• 4

  Paul, E. L., V. A. Atiemo-Obeng, and S. M. Kresta, eds. Handbook of Industrial Mixing: Science and Practice. Hoboken, NJ: Wiley, 2003. https://doi.org/10.1002/0471451452

• 5

  International Society for Pharmaceutical Engineering. ISPE Good Practice Guide: Technology Transfer, Third Edition. North Bethesda, MD: ISPE.

• 6

  Parenteral Drug Association. Technical Report No. 65 (Revised 2022): Technology Transfer. Bethesda, MD: PDA, 2022.