TECHNOLOGY EXPLAINER
Scale-up
— why cells that grow well in 2 L do not grow the same way in 2,000 L
Make the vessel bigger and you can still hold the same set points for temperature, pH and dissolved oxygen. But power per unit volume, impeller tip speed, mixing time and kLa cannot all be matched at once. A study of a 5,000 L production bioreactor found lower kLa, longer mixing time and poorer CO2 removal than in small vessels. This article sets out why, in the language of chemical engineering, and how it differs from “scale-out”, which simply adds more of the same equipment.
- What scale-up is (the short version)
- Our calculation: what changes, and by how much, when a vessel is scaled up geometrically
- Scale-up criteria — P/V, tip speed, mixing time and kLa
- Our calculation: at the same tip speed, the bigger the vessel, the lower the P/V
- Supplying oxygen and removing CO2 — in a big vessel the bubbles fill up with CO2
- A materials engineer's view (1): scale-up is a fight against area divided by volume
- Scale-down models — reproducing a big vessel in a small one
- Scale-out — lining up equipment of the same size
- A materials engineer's view (2): scale-out means “the same parts, in volume, without variation”
- What we could not confirm, and remaining problems
- Glossary / References / Claim-to-source audit
Sourced = stated in published material or peer-reviewed papers (link given)
Our calculation = a figure this article derived, with the assumptions spelled out
Not yet confirmed = a concept or proposal with no confirmed track record
Explanations of general chemical-engineering relations, and our readings of them, are marked separately as Commentary.
1. What scale-up is (the short version)
Scale-up is reproducing, in a large production bioreactor, the culture conditions settled on in a small one. The difficulty is that some quantities can be matched by set point and some cannot.
- Quantities you can match: an original paper published in 2026 (Lemire et al.) treats temperature, pH, dissolved oxygen and feeding strategy as scale-independent variablesSourced
- Quantities you cannot: impeller tip speed, kLa, gas residence time and mixing time are scale-dependent variables, and because they are interdependent it is “impossible to maintain all of them constant during scale-up”Sourced
- No guidelines: a 1999 review says no specific guidelines had been published for large-scale animal cell culture of 10,000 L and above, and that designs were often based on those for microbial culture, which may be far from optimal because microbial and animal cell cultures differ greatly in the energy put into themSourced
Xing et al. (2009), who studied mixing in a 5,000 L production bioreactor without cells, report in their abstract that the 5,000 L bioreactor in its configuration at the time had lower kLa, longer mixing time and a lower CO2 removal rate than 5 L and 20 L bioreactors, and that the maximum viable cell density it could support, limited by oxygen transfer, was estimated at 7 × 106 cells/mLSourced. Same cells, same medium — but change the vessel and the ceiling changes. That is the scale-up problem (our commentary).
2. Our calculation: what changes, and by how much, when a vessel is scaled up geometrically
Take a vessel whose shape is kept the same (geometric similarity) while its volume grows 1,000-fold — for example, from 2 L to 2,000 L. Every length grows by the cube root of 1,000, which is 10 times.
- Agitation power: P = Np · ρ · N³ · d5 (in the turbulent regime the power number Np is roughly constant; N is impeller speed, d is impeller diameter, ρ is liquid density)
- Power per unit volume: since V ∝ d³, P/V ∝ N³ · d²
- Impeller tip speed: u = π · N · d, so u ∝ N · d
- Mixing time: in the turbulent regime N × mixing time can be treated as roughly constant, so mixing time ∝ 1/N
Lemire et al., too, treat P/V as a quantity estimated from agitation speed, and tip speed as a quantity calculated from a formulaSourced.
Lemire et al. put the content of this table into words. Keeping mixing time constant raises tip speed and can damage cells; keeping tip speed constant reduces agitation and can lead to poor mixingSourced. They add that as vessels grow, the surface-area-to-volume ratio falls and gas transfer at the surface drops, so more gas has to be sparged, which can itself damage cells through bubble shearSourced.
3. Scale-up criteria — P/V, tip speed, mixing time and kLa
Lemire et al. list the most commonly used scale-up criteria as constant kLa, constant P/V, constant impeller tip speed and constant gas flow rateSourced. Each is trying to protect something different.
| Criterion | What it protects | Guide values in primary sources |
|---|---|---|
| P/V (power per unit volume, W/m³) | Energy for mixing, keeping cells in suspension and dispersing bubbles | 10 to 80 W/m³ is typical in small equipment; in large single-use bioreactors physical constraints limit it to about 20 to 30 W/m³ at most [Ref. 2]. The design range for one single-use bioreactor family is 10 to 250 W/m³ (laboratory to production scale) [Ref. 3] |
| Impeller tip speed (m/s) | Shear near the impeller | Typically kept below 1 m/s in small equipment to avoid cell damage [Ref. 2]. Below 2.0 m/s as a design range [Ref. 3] |
| Mixing time (s) | Uniformity of pH, nutrients and dissolved gases | Below 60 s as a design range; that bioreactor family achieved below 30 s at all scales [Ref. 3] |
| kLa (h−1) | Capacity to supply oxygen | Above 7 h−1 with pure oxygen, assuming processes of up to 27 to 28 × 106 cells/mL [Ref. 3] |
| Gas flow rate (vvm: litres of gas per litre of liquid per minute) | CO2 removal and oxygen transfer | Lemire et al. cite a report that holding both P/V and vvm constant gave comparable viable cell density, viability and titre at 3 L, 500 L and 2,000 L [Ref. 2] |
All Sourced (Lemire et al. [Ref. 2], Dreher et al. [Ref. 3]). The values apply to the equipment and cells each paper dealt with and are not universal benchmarks for every process.
Lemire et al. themselves scaled up from 1 L to 10 L with constant P/V as the primary criterion, also holding gas flow rate (vvm) constant, and obtained comparable cell growth and productivity. They note that constant P/V has been shown to be the most reliable approach when the difference in vessel volume is smallSourced.
4. Our calculation: at the same tip speed, the bigger the vessel, the lower the P/V
Dreher et al. report, for their company's single-use stirred bioreactor family (50 to 2,000 L), that the larger the vessel, the lower the P/V at the same tip speedSourced. Using figures from the same paper (height/diameter = 2, impeller/vessel diameter = 0.38, power number 1.3), we check this by calculation.
- At a tip speed of 1.0 m/s, 2 L gives about 36 W/m³ and 2,000 L about 3.6 W/m³Our calculation. Since P/V ∝ u³/d, a tenfold increase in length means one tenth
- Even raising the tip speed at 2,000 L to 1.8 m/s gives a P/V of only about 21 W/m³Our calculation
- That matches, in order of magnitude, Lemire et al.'s statement that the maximum P/V in large single-use bioreactors is about 20 to 30 W/m³Sourced
- The experimental condition in Lemire et al.'s 1 L vessel (P/V of about 35 W/m³)Sourced is also the same order as our 2 L, 1.0 m/s value
Reading: when a process run at 35 W/m³ in a small vessel moves to 2,000 L, matching P/V means raising the tip speed, and matching tip speed drops P/V by an order of magnitude — the “every row has red” of Fig. 1 happens in real numbers (our commentary). Assumptions and limits: the geometric ratios and power number are those of one bioreactor family. The power number changes with impeller type, number of impellers and the presence of baffles.
5. Supplying oxygen and removing CO2 — in a big vessel the bubbles fill up with CO2
(1) What happened at 5,000 L
The key points from the abstract of Xing et al.'s (2009) study of a 5,000 L bioreactorSourced:
- The 5,000 L bioreactor had lower kLa, longer mixing time and a lower CO2 removal rate than 5 L and 20 L bioreactors
- Measurements with two probes showed that gradients in pH and dissolved oxygen (differences by location) could exist
- Empirical correlations indicated that increasing air sparged from the bottom was more efficient than increasing agitation power for improving oxygen transfer and CO2 removal
- As liquid volume rises during fed-batch culture, mixing across the whole vessel becomes a challenge
(2) Why CO2 is harder to remove from a big vessel
Lemire et al. explain that pCO2 accumulation is mainly an issue at larger scales, “due to longer gas residence times leading to bubbles becoming saturated with CO2 before reaching the liquid surface”Sourced. Dreher et al., meanwhile, attribute the higher kLa in their larger vessels to “probably” the longer residence time of the bubblesSourced.
The same “longer residence time” works in favour of oxygen and against CO2. Oxygen can keep moving from bubble to liquid, but once a bubble approaches equilibrium with the liquid, no more CO2 goes into it. The capacity to remove CO2 is therefore set by how full of CO2 the bubbles get (the CO2 concentration in the outlet gas) and by how much gas is blown in (our commentary).
- Assumptions: cell density 20 × 106 cells/mL; oxygen uptake 5.5 pmol/cell/day (within the 5.14 to 5.77 reported by Goudar et al.); respiratory quotient = 1 (0.98 to 1.14 in the same paper)
- CO2 production = 5.5 × 10−9 mmol × 2 × 1010 cells/L ÷ 24 h = 4.58 mmol/L/h
- At 2,000 L: 4.58 × 2,000 = 9.2 mol/h (about 403 g/h, about 9.7 kg a day)
Gas flow needed to carry this out in the exhaust (by CO2 concentration in the outlet gas)
- At 1% CO2 in the outlet: 9.2 ÷ 0.01 × 22.4 L ÷ 60 = about 342 L/min = 0.17 vvm
- 0.086 vvm at 2%, 0.034 vvm at 5%, 0.017 vvm at 10%
Reading: the higher the CO2 concentration in the outlet gas, the less gas is needed; but if the outlet gas is close to equilibrium with the liquid, a high outlet CO2 concentration means a high pCO2 in the liquid. In other words, if you want to keep pCO2 in the liquid low, the only way is to blow in a lot of gas (our commentary). Assumptions and limits: the bicarbonate buffer system, release from the liquid surface and CO2 retained in the liquid through base addition are not considered. 22.4 L/mol is an approximation at standard conditions.
Nienow's (2006) review, too, says that as fed-batch cultures have reached higher densities and oxygen or oxygen-enriched air has come into use, the rate of CO2 evolution has been affecting pH control, pCO2 and osmolality, and that if oxygen transfer can be achieved with more intense aeration and agitation, most of the problems disappearSourced. Hu et al. suggest, as one possible solution, separating the two jobs: pure oxygen in fine bubbles, and air in large bubbles to strip CO2Not yet confirmed.
6. A materials engineer's view (1): scale-up is a fight against area divided by volume
Line up the problems so far and almost all of them come down to area divided by volume.
- Liquid surface: Lemire et al. say that as vessels grow, the surface-area-to-volume ratio falls and gas transfer at the surface dropsSourced. For a cylinder with H/D = 2, liquid surface area divided by liquid volume is about 4.6 m−1 at 2 L and about 0.46 m−1 at 2,000 L — one tenthOur calculation
- Liquid depth: with the same shape, depth goes from about 0.22 m at 2 L to about 2.2 m at 2,000 L, ten timesOur calculation. That lengthens bubble residence time and makes CO2 saturation more likely (our commentary)
- Impeller: each unit of impeller area has more liquid to move, and at the same tip speed P/V falls (Fig. 2)
Engineers who have scaled up polymerisation reactors, crystallisers or coating-liquid mixing tanks will find this a familiar picture. What differs is how much headroom there is. As Varley and Birch point out, microbial and animal cell cultures differ greatly in the energy put into themSourced, so in animal cell culture it is hard to solve things by raising P/V a lot. That is why effort has gathered not on forcing it through with power, but on bubble size, sparger placement and gas flow (our commentary). Xing et al.'s finding that more air from the bottom was more efficient than more agitation powerSourced can be read as one example.
7. Scale-down models — reproducing a big vessel in a small one
You cannot experiment by trial and error in a big vessel. So you build a scale-down model that reproduces the conditions of the big vessel in a small one. The review by Hu et al. describes a study that built a scale-down model of a 2,000 L commercial processSourced.
- In a 2 L vessel, 350 rpm corresponded to the same P/V as that organisation's particular 2,000 L vessel and gave comparable performance
- Raising the 2 L vessel to 450 to 550 rpm began to harm cell growth and viability from day 5 onwards
Li et al.'s (2010) review likewise says that shake flasks and small bioreactors are used to screen many clones and conditions, but are unsuited to process optimisation because environmental conditions cannot be monitored or controlled and fed-batch culture is hard to run routinelySourced. What a scale-down model needs is not smallness but the ability to bring in the constraints of the big vessel correctly (our commentary).
8. Scale-out — lining up equipment of the same size
Making the vessel bigger is not the only way to raise output. ICH Q13, the guideline on continuous manufacturing (adopted at Step 4 on 16 November 2022), lists the following ways of changing production outputSourced.
- Change in run time: “Issues not observed over shorter run times may become visible as run time increases”Sourced
- Increase in mass flow rate: the risks relate to changes in process dynamics and to the system's capability to handle increased mass flow ratesSourced
- Duplication of equipment (scale-out): either replicating a production line with the same equipment and setup, or replicating only some unit operations in parallel on the same line. For the latter, the points to consider include uniform flow distribution among the parallel operations, synchronisation and re-integration of parallel flow streams, and material traceabilitySourced
- Increasing equipment size (scale-up): “General principles of equipment scale-up apply, as in the case of batch manufacturing.” Because residence time distribution and process dynamics may change, the risks and the control strategy should be assessed at the new scaleSourced
In cell therapy, scale-out is also used in a different sense. A review by researchers at the US NIH Clinical Center (Song et al. 2022) says that autologous therapies, which use the patient's own cells, particularly require large scale-out capacitySourced. As examples of closed, automated culture systems, the review describes one with a culture chamber fixed at 180 mL, one with a 250 mL chamber, and a rocking-motion system that can be expanded by changing bag size, and mentions a concept for the first of these in which dozens of units are connected to use vertical space and reduce the floor area of a manufacturing facilityNot yet confirmed.
Training material presented in 2022 by a reviewer at the PMDA (Japan's Pharmaceuticals and Medical Devices Agency), explicitly marked as the presenter's personal views, contrasts batch manufacturing, where “verification work is needed at each scale during development and validation”, with continuous manufacturing, where “in some cases the work can be reduced by matching development equipment to the commercial production scale” (our translation)Sourced.
| Aspect | Scale-up (make it bigger) | Scale-out (line them up) |
|---|---|---|
| What changes | P/V, tip speed, mixing time, kLa, liquid depth, area divided by volume (Sections 2 to 6) | Conditions inside each unit do not change. The number of units, and variation between them, become the issue |
| Verification | Verification needed at each scale (for batch manufacturing) [Ref. 8] | Replicating the same equipment and control strategy makes it easier to carry the verification approach over [Ref. 7] |
| Suited to | Products where one batch makes many doses (such as antibody drugs) | Autologous cell therapies made one batch per patient [Ref. 9]; line replication in continuous manufacturing [Ref. 7] |
| Weaknesses | “Everything constant” is impossible [Ref. 2] | Equipment and consumables are needed for every unit. With parallelisation, flow distribution and synchronisation are challenges [Ref. 7] |
Items with a reference are Sourced. The construction of the table and the “suited to” grouping are this article's commentary.
9. A materials engineer's view (2): scale-out means “the same parts, in volume, without variation”
With scale-out, the chemical-engineering problems inside each unit (mixing, oxygen, CO2) stay small. In their place, consumables for every unit, and variation between units, become the issue.
Song et al. list among the advantages of closed, automated technology maintaining sterility through functionally closed single-use components and better consistency from reduced operator-to-operator variationSourced. If autologous therapies make one batch per patient, single-use culture chambers, tubing and bags are consumed in proportion to the number of patients (our commentary).
What is asked of materials here is a different set of properties from scale-up.
- Lot-to-lot consistency: for 100 units to give the same result, 100 consumables need the same surface and the same extractables profile. The control of antioxidants and irradiation conditions discussed in our explainer on single-use systems applies directly
- Contact area at small volumes: the smaller the container, the more plastic surface per mL of liquid (in that explainer's calculation, 1 L has about 12.6 times as much as 2,000 L). The more small units you line up, the relatively larger the influence of the plastic on the liquid
- Reliability of connections: ICH Q13 says single-use connections (tube welds, connectors) that experience long durations or frequent change-outs should be evaluated as potential contamination risksSourced. More units means more connections
Scale-up is a chemical-engineering problem; scale-out is a materials and quality-control problem — broadly, that is how it can be framed (our commentary).
10. What we could not confirm, and remaining problems
(1) Details of the 5,000 L study
We confirmed Xing et al. (2009) from its PubMed abstract. The coefficients of the empirical correlations in the full text, and the cell density supported by the improved configuration, were not checked and are not stated.
(2) Acceptable pCO2 values
Several primary sources state that CO2 accumulation affects cell growth and qualitySourced, but a generally applicable upper limit could not be confirmed in the primary sources within the scope of this article's research, so none is given.
(3) The concept of lining up many automated culture units
The concept of connecting dozens of units to reduce floor area is at the stage of being described in a reviewNot yet confirmed; this article did not confirm any operating record in a real commercial facility.
(4) The factors and values in this article are idealised calculations
The values in Figs. 1 to 3 are calculations assuming geometric similarity, a constant power number and so onOur calculation. In real equipment they change with the number of impellers, aeration and changing liquid volume.
- Temperature, pH and dissolved oxygen can be matched, but P/V, tip speed, mixing time and kLa cannot all be matched at once, as the literature statesSourced
- For a 1,000-fold volume increase at constant P/V, tip speed rises about 2.2 times and mixing time about 4.6 timesOur calculation
- At the same tip speed, P/V at 2,000 L is one tenth of that at 2 LOur calculation
- The 5,000 L bioreactor had lower kLa, slower mixing and poorer CO2 removal; more air was more efficient than more powerSourced
- In large vessels bubbles become saturated with CO2, so keeping pCO2 low takes gas flowSourced
- Scale-out means lining up the same equipment, and ICH Q13 lists flow distribution, synchronisation and more as points to watchSourced
11. Glossary
- Scale-up
- Reproducing, in a large vessel, conditions settled on in a small one.
- Scale-out
- Raising output by replicating or paralleling equipment of the same size.
- Geometric similarity
- Changing only size while keeping shape (dimension ratios) the same.
- P/V
- Agitation power per unit volume (W/m³).
- Power number (Np, Ne)
- A dimensionless form of agitation power. In the turbulent regime it is set largely by impeller shape.
- Impeller tip speed
- The speed of the impeller's outer edge (π × speed × impeller diameter). A guide to shear.
- Mixing time
- The time for an added substance to spread uniformly through the vessel.
- kLa
- Volumetric oxygen transfer coefficient. A measure of how fast oxygen moves from gas to liquid (h−1).
- vvm
- Litres of gas blown in per litre of liquid per minute. A measure of gas flow.
- pCO2
- Partial pressure of CO2 dissolved in the liquid. Affects cells when it builds up.
- Respiratory quotient
- CO2 production rate divided by oxygen uptake rate. Reported to be about 1 for CHO cells.
- Scale-down model
- An experimental system that reproduces the conditions (constraints) of a large vessel in a small one.
- ICH Q13
- The international guideline on continuous manufacturing of drug substances and drug products (Step 4 in 2022).
- Autologous therapy
- A cell therapy that uses the patient's own cells, made as one batch per patient.
12. References
- Xing Z et al. (Bristol-Myers Squibb) “Scale-up analysis for a CHO cell culture process in large-scale bioreactors”, Biotechnology and Bioengineering 103(4):733–746 (2009) (PubMed abstract) https://pubmed.ncbi.nlm.nih.gov/19280669/
- Lemire L et al. (Polytechnique Montréal / National Research Council Canada) “Scale-up of a monoclonal antibody CHO fed-batch production in stirred tank bioreactors: Effect of hydrodynamic conditions and feeding regimen”, Biotechnology Progress (January–February 2026 issue, doi:10.1002/btpr.70073) https://pmc.ncbi.nlm.nih.gov/articles/PMC12908111/
- Dreher T et al. (Sartorius Stedim Biotech) “Design space definition for a stirred single-use bioreactor family from 50 to 2000 L scale”, BMC Proceedings 7(Suppl 6):P55 (2013) https://pmc.ncbi.nlm.nih.gov/articles/PMC3980816/
- Varley J, Birch J “Reactor design for large scale suspension animal cell culture”, Cytotechnology 29:177–205 (1999) https://pmc.ncbi.nlm.nih.gov/articles/PMC3463394/
- Goudar CT, Piret JM, Konstantinov KB “Estimating cell specific oxygen uptake and carbon dioxide production rates for mammalian cells in perfusion culture”, Biotechnology Progress 27(5):1347–1357 (2011) (PubMed abstract) https://pubmed.ncbi.nlm.nih.gov/21626724/
- Hu W, Berdugo C, Chalmers JJ “The potential of hydrodynamic damage to animal cells of industrial relevance: current understanding”, Cytotechnology 63:445–460 (2011) https://pmc.ncbi.nlm.nih.gov/articles/PMC3176934/
- ICH “Continuous Manufacturing of Drug Substances and Drug Products Q13”, final version (adopted 16 November 2022) https://database.ich.org/sites/default/files/ICH_Q13_Step4_Guideline_2022_1116.pdf
- PMDA (Kyoko Sakurai, Office of New Drug IV) “Regulatory points to note on continuous manufacturing of biopharmaceuticals”, Regulatory Science Expert Training material, 9 September 2022, stated to be the presenter's personal views (in Japanese) https://www.pmda.go.jp/files/000248673.pdf
- Song HW et al. (US NIH Clinical Center) “Scaling up and scaling out: Advances and challenges in manufacturing engineered T cell therapies”, International Reviews of Immunology 41(6):638–648 (2022) https://pmc.ncbi.nlm.nih.gov/articles/PMC9815724/
- Nienow AW “Reactor engineering in large scale animal cell culture”, Cytotechnology 50:9–33 (2006) https://pmc.ncbi.nlm.nih.gov/articles/PMC3476006/
- Li F et al. (Genentech) “Cell culture processes for monoclonal antibody production”, mAbs 2(5):466–479 (2010) https://pmc.ncbi.nlm.nih.gov/articles/PMC2958569/
13. Claim-to-source audit
| Claim in the text | Basis | Label |
|---|---|---|
| That the 5,000 L bioreactor had lower kLa, longer mixing time and a lower CO2 removal rate than 5 L and 20 L bioreactors. That the maximum viable cell density, limited by oxygen transfer, was estimated at 7 × 106 cells/mL. That pH and dissolved-oxygen gradients could exist. That increasing bottom air sparging was more efficient than increasing agitation power for oxygen transfer and CO2 removal. That whole-vessel mixing becomes a challenge as liquid volume rises in fed-batch culture | Xing et al. 2009 (PubMed abstract), Reference 1 https://pubmed.ncbi.nlm.nih.gov/19280669/ | Sourced |
| The distinction between scale-independent variables (temperature, pH, dissolved oxygen, feeding) and scale-dependent ones (tip speed, kLa, gas residence time, mixing time), and that the latter are interdependent so all cannot be held constant. That constant mixing time raises tip speed and constant tip speed can worsen mixing. That a lower surface-area-to-volume ratio reduces surface gas transfer, requiring more sparging that can cause bubble damage. The most commonly used criteria (constant kLa, P/V, tip speed and gas flow). Tip speed below 1 m/s and P/V of 10 to 80 W/m³ in small equipment; a maximum of 20 to 30 W/m³ in large single-use vessels. The 1 L experimental condition of about 35 W/m³. That 1 L to 10 L was done at constant P/V and constant vvm with comparable results, and that constant P/V is regarded as most reliable for small volume differences. The cited report of comparable results at 3 L, 500 L and 2,000 L at constant P/V and vvm. That pCO2 accumulation is mainly a large-scale issue, because longer gas residence times saturate bubbles with CO2 before they reach the surface | Lemire et al. 2026, Reference 2 https://pmc.ncbi.nlm.nih.gov/articles/PMC12908111/ | Sourced |
| The design range of the 50 to 2,000 L bioreactor family (tip speed below 2.0 m/s, kLa above 7 h−1 with pure oxygen, mixing time below 60 s, P/V 10 to 250 W/m³), and mixing time below 30 s at all scales. H/D 2:1, impeller/vessel diameter 0.38, power number 1.3. That P/V falls with vessel size at the same tip speed. That the higher kLa in larger vessels was attributed, probably, to longer bubble residence time | Dreher et al. 2013, Reference 3 https://pmc.ncbi.nlm.nih.gov/articles/PMC3980816/ | Sourced |
| That no specific guidelines had been published for large-scale animal cell culture above 10,000 L, and that designs often based on microbial systems may be far from optimal because energy inputs differ greatly | Varley & Birch 1999 (abstract), Reference 4 https://pmc.ncbi.nlm.nih.gov/articles/PMC3463394/ | Sourced |
| CHO cell oxygen uptake rate of 5.14 to 5.77 pmol/cell/day and respiratory quotient of 0.98 to 1.14 | Goudar et al. 2011 (PubMed abstract), Reference 5 https://pubmed.ncbi.nlm.nih.gov/21626724/ | Sourced |
| That in a scale-down model of a 2,000 L commercial process, 350 rpm in a 2 L vessel matched P/V and gave comparable performance, while 450 to 550 rpm began to cause harm from day 5 onwards | Hu et al. 2011 review, Reference 6 https://pmc.ncbi.nlm.nih.gov/articles/PMC3176934/ | Sourced |
| The idea of decoupling pure oxygen in fine bubbles from CO2 stripping with air in large bubbles | A proposal Hu et al. 2011 give as one possible solution; no large-scale track record was confirmed for this article, Reference 6 https://pmc.ncbi.nlm.nih.gov/articles/PMC3176934/ | Not yet confirmed |
| That Q13 was adopted at Step 4 on 16 November 2022. The ways of changing production output (run time, mass flow rate, duplication of equipment (scale-out: line replication and parallelisation), increasing equipment size) and the cautions for each. That general principles of equipment scale-up apply as in batch manufacturing. That single-use connections should be evaluated as contamination risks | ICH Q13 final version, Reference 7 https://database.ich.org/sites/default/files/ICH_Q13_Step4_Guideline_2022_1116.pdf | Sourced |
| The contrast that batch manufacturing needs verification at each scale while continuous manufacturing can in some cases reduce the work by matching development equipment to commercial scale (from material stated to be the presenter's personal views) | PMDA training material 2022, Reference 8 https://www.pmda.go.jp/files/000248673.pdf | Sourced |
| That autologous therapies particularly require large scale-out capacity. Examples of a fixed 180 mL chamber, a 250 mL chamber, and a rocking system expandable by bag size. The advantages of closed automation (sterility through functionally closed single-use components, reduced operator variation and so on) | Song et al. 2022 review, Reference 9 https://pmc.ncbi.nlm.nih.gov/articles/PMC9815724/ | Sourced |
| The concept of connecting dozens of units to use vertical space and reduce floor area | A concept described by Song et al. 2022; no operating record in a commercial facility was confirmed for this article, Reference 9 https://pmc.ncbi.nlm.nih.gov/articles/PMC9815724/ | Not yet confirmed |
| That with higher-density fed-batch culture using oxygen or oxygen-enriched air, the CO2 evolution rate affects pH control, pCO2 and osmolality, and that most problems disappear with more intense aeration and agitation | Nienow 2006 review (abstract), Reference 10 https://pmc.ncbi.nlm.nih.gov/articles/PMC3476006/ | Sourced |
| That shake flasks and small bioreactors are used for screening but are unsuited to process optimisation because conditions cannot be monitored or controlled and fed-batch culture is hard to run routinely | Li et al. 2010 review, Reference 11 https://pmc.ncbi.nlm.nih.gov/articles/PMC2958569/ | Sourced |
| Factors for a 1,000-fold volume increase (constant P/V: N x0.22, tip speed x2.2, mixing time x4.6; constant tip speed: P/V x0.1, mixing time x10; constant mixing time: tip speed x10, P/V x100). The tip speed versus P/V curves (at 1.0 m/s: 2 L 35.6, 200 L 7.7, 2,000 L 3.6 W/m³; about 21 W/m³ at 2,000 L and 1.8 m/s). CO2 of 9.2 mol/h at 2,000 L (about 403 g/h, about 9.7 kg/day), and 0.17, 0.086, 0.034 and 0.017 vvm at outlet CO2 of 1, 2, 5 and 10%. Liquid surface area over volume from 4.6 to 0.46 m−1, liquid depth from 0.22 to 2.2 m | Our calculation. Geometric similarity, a constant power number in the turbulent regime, N × mixing time = constant, applying H/D = 2, impeller/vessel diameter 0.38 and power number 1.3 to all volumes, 20 × 106 cells/mL, oxygen uptake of 5.5 pmol/cell/day, respiratory quotient 1 and 22.4 L/mol are premises set by this article. 2 L is an extrapolation beyond Dreher et al.'s range | Our calculation |
| The coefficients of Xing et al.'s empirical correlations; a generally applicable upper limit for pCO2 | Only the abstract was checked, or none could be confirmed in primary sources within the scope of this article's research, so none is stated (commentary) | Commentary |
| The general chemical-engineering relations P = Np · ρ · N³ · d5, P/V ∝ N³d², u = πNd and N × mixing time ≈ constant. The “red = worse” framing. The framing that residence time favours oxygen and works against CO2. The reading that keeping pCO2 low needs gas flow. The area-over-volume framing, and the reading that since power cannot force it through, solutions come via bubbles and gas flow. The framing that a scale-down model must reproduce constraints. The construction of the scale-up versus scale-out table. The framing of scale-out as a materials and quality-control problem | Explanation of general chemical-engineering relations and this article's own structuring and commentary based on the published content. Not views expressed by the authors of the papers | Commentary |
| That Figs. 1 to 4 are explanatory drawings, not real equipment or measurements, and that the hero image and Fig. 5 are AI-generated images | Our note (commentary) | Commentary |
Last updated 23 September 2026. Sources are limited to primary material (peer-reviewed papers and conference papers, PubMed abstracts, ICH guidelines and published PMDA material). The empirical correlations in the full text of Xing et al. (2009), a generally applicable upper limit for pCO2, and any operating record for concepts that line up many automated culture units could not be confirmed in published primary sources and are not stated. The factors and values in Sections 2, 4, 5 and 6 are this article's calculations based on stated assumptions. All figures are explanatory concept graphics. Figs. 1 to 4 are vector drawings; the hero image and Fig. 5 are AI-generated images, and none of them shows real equipment, a real product or a real measurement.