Biomedical Engineering Reference
In-Depth Information
TABL E 5.3. Fermentation SSM Qualification Against Manufacturing Performance Criteria
Total Elapsed Time
to Induction (h)
Product Yield (g)
Biomass (OD)
SSM Mean
SD
100
3
26.3
0.3
100
2
SSM interval
(corrected for
sample frequency)
94.4-106
(94.4-98.0)
24.8-27.0
(27.76-28.24)
89.1-109
(90.2-97.6)
Manufacturing data
tolerance interval
(
a¼
0.95, p
¼
0.95)
86.5-135
26.6-29.9
81.3-113.8
Product yield and biomass data have been normalized against the SSMmean performance. SD, standard deviation.
A small-scale model was developed and qualified on the basis of manufacturing-
scale data for this case study. The small-scale model is deemed qualified if the
performance parameters fall within the tolerance interval bounds determined from
manufacturing data. Table 5.1 shows the key operational parameters of the small-scale
model. The growth performance in the small-scale model, expressed here as the total
elapsed time to induction, initially did not meet the acceptance criteria based on
manufacturing scale performance although all other qualification criteria were achieved
(Table 5.3). As seen in Fig. 5.6, biomass growth and the corresponding dissolved oxygen
profile at the manufacturing scale lag behind that in the small-scale model from the early
stages of the fed-batch phase onward. The root cause of this discrepancy was the relative
high sampling frequency at bench scale. When the sample frequency was accounted for
the small-scale model met all the acceptance criteria of the manufacturing scale
production (Table 5.3). Therefore, the small-scale model was deemed as qualified.
5.5 DESIGN OF EXPERIMENT STUDIES
These studies are performed on the basis of OPs identified and prioritized by FMEA and
using qualified small scale models. The characterization ranges (CRs) were set at three
times the normal operating range (NOR). A variety of DOE studies can be planned
depending on the number of parameters that need to be examined and the resolution
required for the study [4, 8, 10]. The approaches vary in their resolution and the number of
experiments required for a given number of parameters to be studied. For our case,
fractional factorial and full factorial studies conducted were deemed appropriate. The
fractional factorial DOE study was designed such that the main effects could be distin-
guished from all two-factor interactions for those parameters expected to have the most
effect on process performance. Each experiment was performed at least in duplicate, and
theDOE designs containedmultiple center points used for independent estimation of pure
error as well as identification of potential quadratic curvature in the data. All experiments
were carefully monitored to ensure control of the operation parameter(s) within the CR.
Datafromfractional andfull factorialDOEdesignstudieswereanalyzedbyanalysisof
variance (ANOVA) to understand main effects of operational parameters on process
