696 Biotechnol. Prog.

2006, 22, 696−703

A Systematic Approach for Scale-Down Model Development and Characterization

of Commercial Cell Culture Processes
Feng Li,*.† Yasunori Hashimura,† Robert Pendleton,‡ Jean Harms,† Erin Collins,‡ and Brian Lee†
Process Engineering and Cellular Science and Technology, Amgen Inc., Thousand Oaks, California 91320

The objective of process characterization is to demonstrate robustness of manufacturing processes

by understanding the relationship between key operating parameters and final performance.
Technical information from the characterization study is important for subsequent process
validation, and this has become a regulatory expectation in recent years. Since performing the
study at the manufacturing scale is not practically feasible, development of scale-down models
that represent the performance of the commercial process is essential to achieve reliable process
characterization. In this study, we describe a systematic approach to develop a bioreactor scale-
down model and to characterize a cell culture process for recombinant protein production in
CHO cells. First, a scale-down model using 2-L bioreactors was developed on the basis of the
2000-L commercial scale process. Profiles of cell growth, productivity, product quality, culture
environments (pH, DO, pCO2), and level of metabolites (glucose, glutamine, lactate, ammonia)
were compared between the two scales to qualify the scale-down model. The key operating
parameters were then characterized in single-parameter ranging studies and an interaction study
using this scale-down model. Appropriate operation ranges and acceptance criteria for certain
key parameters were determined to ensure the success of process validation and the process
performance consistency. The process worst-case condition was also identified through the
interaction study.

Introduction
As development of a commercial cell culture process for
production of a biological product is completed at the laboratory
and pilot scales, commercialization process begins with process
characterization, scale-up, technology transfer, and validation
for a manufacturing site. Process characterization is a systematic
investigation to understand the commercial scale cell culture
process in detail, including the relationship between key
operating parameters and the final cell culture performance for
culture growth and productivity, as well as the final product
quality attributes such as potency, post-transcriptional modifica-
tions, and impurity profiles (1). Objectives of process charac-
terization include identification of key operational and perfor-
mance parameters, establishment of acceptable ranges for key Figure 1. Process characterization flowchart.
parameters, and demonstration of process robustness (2, 3). Risk Assessment. Before starting a characterization study,
Technical information from the characterization study has the potential risk of failure during operation of the entire cell
become a regulatory expectation in recent years as prerequisite culture process including medium preparation, seed train, and
for manufacturing process validation as well as for the long- production stage needs to be evaluated in order to prioritize
term commercial manufacturing support (4, 5). Since performing the risks. As a systematic method to prioritize operating
a characterization study at the manufacturing scale is not parameters and to evaluate their potential impact to the process,
practically feasible due to cost of operation and limited Failure Mode and Effects Analysis (FMEA) is a useful risk-
availability of large-scale bioreactors, scale-down models that analysis tool to guide the process evaluation and to determine
represent the performance of manufacturing scale process are which operating parameters require further characterization (2).
usually employed at the laboratory scale. It is also important to FMEA is based on weighing of three aspects: severity of the
use qualified analytical methods throughout the characterization impact of an operating parameter excursion, frequency of
study for consistency and accuracy of assays. The overall occurrence, and detectability of the excursion. On the basis of
procedure of cell culture process characterization is shown in historical data from their previous development work, process
Figure 1 and described in detail below. development scientists are able to assess the severity of operating
parameters. Manufacturing personnel familiar with the full-scale
* To whom correspondence should be addressed. Ph: 805-313-4481.
Fax: 805-499-6819. Email: fengl@amgen.com.
manufacturing equipment controllability should provide the
† Process Engineering. occurrence and detection information. Therefore, FMEA often
‡ Cellular Science and Technology. requires a joint effort from various groups including process
10.1021/bp0504041 CCC: $33.50 © 2006 American Chemical Society and American Institute of Chemical Engineers
Published on Web 04/20/2006
Biotechnol. Prog., 2006, Vol. 22, No. 3
Figure 2. Scale-down model development and qualification.
development, process engineering, manufacturing, and quality
control. Through this risk assessment, key operating parameters,
their control and characterization ranges, and key performance
parameters should be identified.
Scale-Down Model Development and Qualification. During
development of a scale-down model, it is important to demon-
strate equivalency of the key performance parameters between
the bench and commercial scales (6). Ideally, cell culture
performances at the commercial scale should be used as a
baseline to guide scale-down model development. If the full-
scale data is not available, pilot-scale performances can be used
to provide preliminary scale-up information. Under these
circumstances, process characterization has to proceed at risk,
and the models should be re-evaluated once the commercial
scale data becomes available.
A typical process of scale-down model development and
qualification is shown in Figure 2. Bench-scale bioreactors are
usually used as a primary scale-down model for the production
stage, whereas other culture systems such as shaker flasks and
spinner flasks are used for the seed train stage and to screen
raw materials. A good scale-down model needs to match not
only the large-scale performance at the operating set point but
also the response of the key operating parameter changes. Since
it is not practical to compare the two scales beyond the normal
operating conditions, a proper scale-down strategy is necessary
to ensure intrinsic consistency between scales. Then it can be
used in the process characterization study to test the impacts of
operating parameter variations on process performances and
product quality attributes. The cell culture operating parameters
can be categorized into volume-dependent parameters (e.g.,
Figure 3. Ranges of process parameters.
697
working volume, feed volume, agitation, aeration) and volume-
independent parameters (e.g., pH, dissolved oxygen, tempera-
ture). A general strategy for scale-down model development is
to proportionally scale down the volume-dependent parameters
while maintaining the volume-independent parameters at the
same set points used in the large-scale process. However, some
volume-dependent parameters, such as bioreactor agitation speed
and aeration rate, are difficult to be scaled down linearly because
of differences in bioreactor geometry, liquid surface-to-volume
ratio, gassing regime, and control capability. Thus, different
operating conditions for the volume-dependent parameters may
be employed for the different scales as long as these variations
do not significantly change the final process performance and
product quality.
Once a scale-down model is established, the model needs to
be qualified by demonstrating equivalent performance with the
large-scale process. Key performance parameters of the cell
culture process such as cell growth and product quality attributes
and their acceptable ranges need to be defined as the scale-
down model qualification criteria. Ideally, the scale comparison
study is performed in parallel runs with the same operating
conditions, including the raw materials, culture media, and
inoculum, at the center point of each parameter except the
volume-dependent parameters.
Single Parameter Ranging Study. The main purpose of this
study is to identify the key operating parameters and their
acceptable ranges. The ranging study will vary one operating
parameter at a time while keeping the others at their set point.
Figure 3 shows the different ranges of an operating parameter.
Center point of the operating range (OR) is usually used as a
process control set point except for the operational parameter
that is under evaluation. Effects of an operational parameter on
the final process performance are tested with a wider range up
to three times the OR. The acceptable range (AR) of a parameter
may reside either inside or outside of the characterization range
(CR) depending on the result of final cell culture performance
under the various operating conditions characterized. The CR
of each potential key operational parameter should be deter-
mined through FMEA based on the risk of process failure by
the individual parameter. Initial screening of key operational
parameters for a process with multiunit operations that involve
many operational parameters such as downstream purification
can be performed by design of experiments (DOE). However,
since the number of potential key operational parameters for a
cell culture process is relatively small and the effects of
operating parameters on process performances are highly
698 Biotechnol. Prog., 2006, Vol. 22, No. 3

nonlinear, a single parameter ranging study is more suitable for

both screening of key parameters and identification of the
acceptable ranges.
DOE Interaction Study and the Worst-Case Study. Once
the key operating parameters are identified by single parameter
ranging study, interaction effects among the parameters are to
be evaluated. Statistically designed experiments using a frac-
tional factorial design at two levels representing the upper and
lower limits of each operating parameter are most commonly
used (7, 8). However, it should be noted that the fractional
factorial experiment does not resolve all interactions since the
interactions among the key parameters may be confounded by
the interactions of the non-key parameters. In the DOE design,
it is important to ensure the interactions expected to be important
are not confounded with interaction expected to be unimportant.
Results of the interaction study are used to confirm robustness
of the commercial process. DOE experiments can also identify
the worst-case condition, which is a combination of operating
parameters at their limits, which generates the worst performance
during manufacturing. The key performance parameters are used
to identify the worst-case condition, such as cell viability,
product titer, and product quality attributes. The materials
generated from the worst-case condition should be further
processed to purification in order to evaluate any potential
impact on downstream operations and product quality.
The characterization study can be broken down into several
parts based on different unit operations. A typical cell culture
manufacturing process includes vial thaw, seed train scale-up,
and production stages, which usually need to be characterized
separately. In addition, the media hold test and cell line stability
should also be addressed. Cell culture characterization will need Figure 4. Cell growth (a) and viability (b) at different agitation speed
in the 2-L scale-down model.
tremendous support from the analytical group to monitor product
quality. Cooperation from downstream harvest and purification Results and Discussion
groups is also necessary. Some cell culture materials from
characterization studies can be used as feed stream for down- Scale-Down Model Development and Qualification. The
stream operations, especially for the worst-case study. physical and chemical parameters that are often considered
critical during development of a scale-down model for cell
Methods and Materials culture process are dissolved oxygen, temperature, pH, inocula-
tion density, nutrient concentration, and shear force by agitation
Cell Line and Medium. A CHO cell line expressing a and gas sparging, as these parameters can significantly affect
recombinant glycosylated protein was used in this study. The the cell growth and protein production. The 2-L Applikon
cells were grown in suspension using proprietary serum-free bioreactors were used for scale-down model development. All
cell culture media and feed solutions; 0.5 g/L Pluronic F-68 volume-independent parameters were controlled at the same set
(Sigma, St. Louis, MO) was added into the culture media as a points in both large-scale and small-scale bioreactors. Among
shear-protection additive. volume-dependent parameters, the bioreactor working volume
Cell Culture Operation. The cell culture process employs and the nutrient feed volume were scaled down linearly based
a 13-day fed-batch process with bolus feeds at defined time on the volume difference, whereas the impeller agitation speed
points. The small-scale cell culture experiments were carried and the aeration rate needed to be adjusted independently from
out in 2-L Applikon bioreactors (Applikon, Netherlands) with the volume ratio. As mammalian cells are considered to be
a starting volume of 1.4 L and the final working volume of 2 shear-sensitive (9, 10), high agitation speed can cause shear
L. For the large-scale cell culture study, a 2000-L bioreactor damage to the cells, but low agitation speed can result in poor
(B.Braun, Allentown, PA) with a starting volume of 1400 L mixing and sometimes cell aggregation. Several criteria such
and the final working volume of 2000 L was used. Four 2000-L as power per volume (P/V), impeller tip speed. or energy
runs were performed to establish the baseline performance for dissipation have been proposed as a target to be matched during
scale-down model development. scale-up of agitation speed (11, 12). In this study, scale-up
Analytical Methods. Cell culture samples were taken daily strategies based on equivalent power per volume (P/V) and
for viable cell density and viability measurements using a Cedex impeller tip speed were investigated. In the 2-L bioreactors, five
cell counter (Innovatis AG, Germany). Cell culture metabolites agitation speeds at 150, 250, 350, 450, and 550 rpm were
such as glucose, glutamine, lactate, and ammonia were moni- screened, where 350 and 550 rpm represents equivalent P/V
tored using a NOVA Bioprofile 400 (NOVA Biomedical, and impeller tip speed, respectively, based on the agitation set
Worcester, MA). Off-line pH and dissolved CO2 were analyzed point at the 2000-L commercial scale. Figure 4 shows the
using a blood gas analyzer (Bayer Healthcare). The protein profiles of growth and viability of the cells at the different
product concentration was measured using a HPLC system. The agitation speeds. Whereas the cell growth and viability in the
glycosylation isoform percentage was measured using isoelectric 2-L bioreactors with 150 250, and 350 rpm were comparable
focusing gel (IEF). with those of the commercial scale, lower cell density and
Biotechnol. Prog., 2006, Vol. 22, No. 3
Figure 5. The comparison of viable cell density (a), viability (b), dissolved CO2 (pCO2) (c), normalized titer (d), and normalized glycosylation
isoform percentage (e), between 2-L (n ) 6) and 2000-L (n ) 4) scales. Mean ( one standard deviation.
viability were observed with high agitation speeds at 450 and
550 rpm. Both agitation and bubble-sparge aeration can cause
shear-induced cell damage. Since the high agitation speed should
decrease oxygen sparge flow rate to maintain the same oxygen
transfer rate, the observed viability drop is most likely due to
the agitation-induced shear damage. This result indicates that
shear stress resulting from the high agitation speeds in the 2-L
bioreactor becomes detrimental to the cells. Since the slightly
higher final cell density at 350 rpm matched more closely with
the growth profile of the 2000-L scale, the agitation speed, 350
rpm, based on equivalent P/V was selected for the scale-down
model to provide mixing and shear conditions comparable with
those of the large-scale bioreactor.
The main function of aeration during cell culture process is
to supply oxygen and remove CO2 from the culture medium.
Oxygen is supplied to the bioreactor in two different modes;
one is overlaying air to the headspace and the other is sparging
oxygen through a sparger. The aeration rate through a sparger
changes during the process in order to maintain a constant level
of dissolved oxygen (DO) in response to the oxygen demand
by the cells. Gas bubble residence time in the small-scale
bioreactor is much shorter than in the large-scale bioreactor as
a result of the shorter liquid height. Since the dissolved oxygen
699
levels in both scale bioreactors are controlled at the same set
point, the air flow rate through the sparger per volume (vvm)
in the small-scale bioreactor is usually higher than that in the
large-scale bioreactor to meet the DO demand of the culture,
which also resulted in lower dissolved carbon dioxide (pCO2)
in the medium as a result of more efficient CO2 removal (13).
Moreover, in small-scale bioreactors, a significant portion of
CO2 stripping occurs from the surface of the liquid by headspace
aeration. In contrast, CO2 removal becomes less effective by
headspace aeration in large-scale reactors because of the reduced
liquid surface-to-volume ratio (14). To avoid high pCO2
accumulation at the 2000-L scale, a mixed gas of air and oxygen
was used to increase the overall sparge rate (vvm), which can
strip CO2 more efficiently. The percentages of air and oxygen
in the gas mixture were fixed throughout the entire run while
the oxygen and air sparge flow rates varied and were determined
by the DO control. This stripping approach using the air and
oxygen mixed gas was not used in the 2-L bioreactor because
of the highly efficient CO2 removal at the small scale. The pCO2
level during the run in the scale-down model was about 20
mmHg lower than that in the large scale (Figure 5). To
understand the potential impact of the different pCO2 level, a
pCO2 ranging study was performed by sparging CO2 through
700 Biotechnol. Prog., 2006, Vol. 22, No. 3

Figure 6. Cell culture metabolite profile scale comparison of glucose (a), glutamine (b), lactate (c), and ammonia (d) between 2-L (n ) 6) and
2000-L (n ) 4) scales. Mean ( one standard deviation. Metabolites in cell culture samples were measured by Nova Bioprofile. Data shown here
are normalized.

Table 1. Operating Parameters Scale Down from 2000-L to 2-L Scale

parameter type operating parameter scale-down strategy
volume-dependent batch medium volume, inoculum volume, volumetrically scale down 1000 times
feed volume, working volume
volume-independent inoculation density, ph, dissolved oxygen, similar set points at different scales
temperature, feed regime
nonlinear parameter impeller agitation P/V equivalencea
oxygen/air aeration no air sparge to reduce CO2, only pure oxygen
used in the small scale.
a P/V is determined from the equation P/V ) (N FN3D5)/V, where P is the total power input rate (g cm2 s-3), N is the impeller power number, F is the
p p
fluid density (g cm-3), N is the impeller speed (rev s-1), D is the impeller diameter (cm), and V is the total culture volume.

Table 2. TOST Analysis of Key Performance Parameters for Scale Comparison

parameter mean value 3σ max p-value result
peak viable cell density (cells/mL) 26.30 × 106 3.46 × 106 0.0198 equivalent
integrated viable cell density (cells‚day/mL) 170.54 × 106 27.14 × 106 0.0139 equivalent
normalized titer 81.95 10.22 0.0095 equivalent
normalized isoform % 30.80 6.67 0.0006 equivalent

the sparger at various flow rates. The result indicated that the
difference in pCO2 within the range did not affect the final cell
culture performance (data not shown). Therefore, pCO2 was not
considered as a key performance parameter for scale-down
model qualification. Table 1 summarizes the different types
operating parameters and their scale-down strategies.
Six 2-L scale runs using the defined scale-down conditions
were compared with four 2000-L scale runs. The performance
parameters including cell growth, viability, product titer, and
product glycosylation isoform were comparable between the two
scales (Figure 5). The metabolite profiles of glucose, glutamine,
lactate, and ammonia were also very similar between the scales
except for the pCO2 profiles, which were notably different
(Figure 6). Lactate concentration increased with cell growth and
then was consumed by the cells at the late stage of the run. It
was noticed that the lactate concentration varied a lot during
its consumption stage. Since these big variations were observed
from both scales, it was from intrinsic process variability and
not from scale difference.
Statistical Analysis of Process Comparability for Scale-
Down Model. Peak viable cell density, integrated viable cell
density (IVC), product titer, and glycosylation isoform percent-
age were considered as key performance parameters for scale-
down model qualification. Statistical approaches were employed
to demonstrate equivalency of the key performance parameters
between the bench and commercial scales.
The conventional Student’s t test is not testing for equiva-
lency, but only testing whether sets are statistically different.
With the t test, the null hypothesis (hypothesis that data sets
are not statistically different within a defined confidence level)
may be rejected or may not be rejected. Rejection of the
hypothesis demonstrates statistical difference, but not rejecting
Biotechnol. Prog., 2006, Vol. 22, No. 3
Figure 7. Viable cell density (a), viability (b), normalized product titer (c), and normalized glycosylation isoform percentage (d) at five different
dissolved oxygen levels.
the hypothesis only means that not enough data were available
to demonstrate difference and does not mean equivalency.
Moreover, the conventional t test is only able to demonstrate
the statistical difference, which may be different from practical
difference. For example, a statistical difference that is smaller
than the assay or equipment variability cannot demonstrate the
practical differences caused by the different scale processes.
Therefore, a t test may declare nonequivalency of process
performance between the different scales even if the statistical
difference is negligible from a practical standpoint. Here, a new
equivalency test approach, Two-One-Side Test (TOST) is used
to determine the equivalency between the scale-down model
and the large-scale process performance. The scale equivalence
is defined as -θ < µS - µL < θ, where µS and µL are
performance parameter mean values at the small and large scale,
respectively. If the difference in the performance parameter
means of two different scales falls within an acceptable range
of [-θ, θ], the process performances at these scales are
considered to be equivalent. Setting the acceptable ranges should
be approached from a practical standpoint, accounting for the
intrinsic process and assay variability. For instance, historical
data at large-scale were evaluated to gauge how much variability
in performance occurred under the identical process conditions,
and the information was used to set acceptable ranges. Depend-
ing on the availability and quality of historical data, setting
acceptance criteria would be difficult. Small number of repeat
runs usually increases leniency for acceptance.
In this study, the data from four 2000-L runs were used to
determine how much variability in performance can be expected
in the process and the range of variability within which
equivalency can be demonstrated. Variations within (3 standard
deviations of the mean (3σ) were considered acceptable, and
the range was set as [-3σ, 3σ] based on the 2000-L process
data. Using JMP software, TOST analysis was performed on
four performance parameters: final cell density, integrated viable
701
cell density (IVC), product titer, and isoform percentage of the
scale-down model and the large-scale processes. The results that
are summarized in Table 2 indicate that the key process
performance between the different scales is equivalent; therefore
the scale-down model is qualified to be used for process
characterization study.
Single-Parameter Ranging Study. The production stage
characterization started with a single parameter ranging study.
Five potential key operating parameters were identified during
FMEA for characterization, including dissolved oxygen (DO),
inoculation density, culture pH, temperature, and dissolved
carbon dioxide (pCO2). Each parameter was tested at three to
six different levels with duplicate bioreactor runs for each
condition. These ranging studies examined the impact on the
process performance by various set points of each parameter.
The acceptable range for each operating parameter was deter-
mined on the basis of these study results. The studies of DO
and pH are shown here as two examples.
Figure 7 shows the effects of dissolved oxygen on cell growth
and product quality related attributes. The cell growth and cell
viability were reduced at the higher set points of DO than at
the center point. While the slower cell growth yielded lower
product titer at the high DO, the glycosylation isoform level
(%) of product does not seem to be affected by high DO. The
lower DO levels did not affect the cell culture performance.
Figure 8 shows the effect of different culture pH on process
performance. Within the range of pH levels tested, cell growth
and product titer were not significantly affected compared to
the control. However, the final glycosylation isoform level (%)
in product increased with higher culture pH. These results
indicate that a certain cell culture condition can have significant
impact on one performance parameter while the other perfor-
mance parameters are unaffected. Also, understanding the effect
of individual operational parameter on the final cell culture
702 Biotechnol. Prog., 2006, Vol. 22, No. 3

Figure 8. Viable cell density (a), viability (b), normalized product titer (c), and normalized glycosylation isoform percentage (d) at six different
pH set points.

Table 3. Interaction Study DOE Design and Test Results

run DO pH temp inoculum titer isoform %
1 - - - - 22 36
2 - - - - 22 37
3 - - + + 88 56
4 - - + + 91 57
5 - + - + 64 57
6 - + - + 72 62
7 - + + - 72 72
8 - + + - 74 82
9 0 0 0 0 84 56
10 0 0 0 0 82 59
11 + - - + 37 28
12 + - - + 43 32
13 + - + - 51 38
14 + - + - 53 38
15 + + - - 49 59
16 + + - - 40 36
17 + + + + 38 37
18 + + + + 67 56

performance using a wide range is important for selection of

acceptable ranges.
Interaction Study. After the single parameter ranging study,
four key operating parameters (inoculum density, pH, temper-
ature, and dissolved oxygen) were identified for further char-
acterization for interaction effects among the parameters. The
interactions between four parameters were evaluated by using
a two-level half-fractional factorial design with the upper and
lower limits of the acceptable ranges. Per design of the
experiment, all parameters were simultaneously varied at either
high or low level in each run, with the exception of the center-
point condition where all operating parameters were maintained
at the set points. This design allows examination of all main
effects and some two-factor interactions. The objectives are to
demonstrate process robustness and product quality consistency.
Table 3 lists the DOE design pattern and experimental results.
The worst-case condition was identified on the basis of the result
Figure 9. JMP analysis of the main effects and parameter interactions
in the DOE study. (a) Operating parameter effects for product titer (b)
effects for glycosylation isoform %.
of the lowest product titer and isoform level that was achieved
with the combination of high DO, low pH, low temperature,
and high inoculation density. This condition represents the
worst-case scenario of the manufacturing process within the
acceptable range for individual operational parameters.
Figure 9 shows the main effects and interactions on product
titer and isoform level. The results indicate that DO has a
significant negative effect on both product titer and isoform
Biotechnol. Prog., 2006, Vol. 22, No. 3
level. While pH has a significant impact on isoform level, it
has no significant effect on titer. Inoculum density only affects
the product titer and not isoform level, and temperature had
significant positive effects on both performance parameters.
Aside from confirming the main effects that were found in
the single parameter ranging studies, the JMP analysis also
resolved interactions between some of these operating param-
eters. For the product titer, two interactions were found to be
significant, DO with temperature and DO with inoculum density,
whereas for the isoform level, the only significant interaction
was between DO and pH.
Conclusions
In this study, we demonstrate a systematic approach to
characterize a commercial cell culture process for therapeutic
recombinant protein production. A scale-down model that
represents the large-scale cell culture process was developed
and qualified by comparing key process performance parameters
between the scales using a statistical analysis, the Two-One-
Side Test (TOST). For process characterization, single-parameter
ranging study and DOE study were conducted using the qualified
scale-down model system to evaluate the main effects and
interactions of key operational parameters. The worst-case cell
culture condition was identified as a combination of the key
operational parameters within the operating range, and the
intermediate material generated under the condition needs be
further evaluated for a potential impact on the subsequent
downstream process.
Acknowledgment
The authors thank Earl Pineda, Matt Nelson, Ben Beneski,
and Ran Zheng for technical discussion; Tamer Eris and Bernice
Yeung for analytical support; and Andy Coop for statistical data
analysis.
References and Notes
(1) Chirino, A. J.; Mire-Sluis, A. Characterizing biological products
and assessing comparability following manufacturing changes. Nat.
Biotechnol. 2004, 22, 1383-1391.
(2) Seely, J. E.; Seely, R. A rational, stepwise approach to process
characterization. BioPharm Int. 2003, Aug, 24-34.
703
(3) Rathore, A. S.; Wang, A.; Menon, M.; Riske, F.; Campbell, J.;
Goodrich, E.; Martin, J. Optimization, scale-up and validation issues
in filtration of biopharmaceuticals-Part I. BioPharm Int. 2004, Aug.
(4) Rathore, A. S.; Noferi, J. F.; Arling, E. R.; Sofer, G.; Watler, P.;
O’Leary, R. Process validation how much to do and when to do it.
BioPharm Int. 2002, Oct, 18-28.
(5) FDA Guidance for Industry: Q5E Comparability of Biotechno-
logical/Biological Products Subject to Changes in Their Manufactur-
ing Process. http://www.fda.gov/cder/guidance/index.htm, June,
2005.
(6) Rathore, A.; Krishnan, R.; Tozer, S.; Smiley, D.; Rausch, S. Scaling
Down of Biopharmaceutical Unit Operations Part 1: Fermentation.
BioPharm Int. 2005, Mar.
(7) Moran, E. B.; McGowan, S. T.; McGuire, J. M.; Frankland, J. E.;
Oyebade, I. A.; Waller, W.; Archer, L. C.; Morris, L. O.; Pandya,
J.; Nathan, S. R.; Smith, L.; Cadette, M. L.; Michalowski, J. T. A
systematic approach to the validation of process control parameters
for monoclonal antibody production in fed-batch culture of a murine
myeloma. Biotechnol. Bioeng. 2000, 69 (3), 242-255.
(8) Kelley, B. Establishing process robustness using designed experi-
ments. In Pharmaceutical Process Validation; Sofer, G., Zabriskie,
D., Eds.; Marcel Dekker Press: New York, 1999.
(9) Chisti, Y. Animal-cell damage in sparged bioreactors. TIBTECH
2000, 18, 420-432.
(10) Michaels, J. D.; Mallik, A. K.; Papoutsakis, E. T. Sparging and
agitation-induced injury of cultured animal cells: Do cell-to-bubble
interactions in the bulk liquid injure cells? Biotechnol. Bioeng. 1996,
51, 399-409.
(11) Venkat, R. V.; Chalmers, J. J. Characterization of agitation
enviroments in 250 mL spinner vessel, 3 L, and 20 L reactor vessels
used for animal cell microcarrier culture. Cytotechnology 1996, 22,
95-102.
(12) Marks, D. M. Equipment design considerations for large scale
cell culture. Cytotechnology 2003, 42, 21-33.
(13) Xie, L.; Metallo, C.; Warren, J.; Pilbrough, W.; Peltier, J.; Zhong,
T.; Pikus, L.; Yancy, A.; Leung, J.; Aunins, J. G.; Zhou, W. Large-
scale propagation of a replication-defective adenovirus vector in
stirred-tank bioreactor PER. C6 cell culture under sparging condi-
tions. Biotechnol. Bioeng. 2003, 83 (1), 45-52.
(14) Mostafa, S. S.; Gu, X. Strategies for improved dCO2 removal in
large-scale fed-batch cultures. Biotechnol. Prog. 2003, 19 (1), 45-
51.
Accepted for publication March 27, 2006.
BP0504041