Biomedical Engineering Reference
In-Depth Information
Figure
2.6. Generation of a design space starts with defining appropriate responses, the
process methodology, and important process variables. (a) Charge variants and impurities are
defined as important product attributes based on a risk assessment, such as the decision tree
described in Fig. 2.2. (b) Based on prior knowledge, a flow-through ion-exchange purification
step is felt to be an appropriate strategy for controlling these attributes in the product. (c) A
fishbone or Ishikawa diagram can be used to describe a broad variety of variables and the
relationship of the variables toproduct quality. (d) A risk assessment such as FMEA can be used to
rank the importance of these factors based on probability, severity, and detectability. High risk is
indicated in red, moderate risk in yellow, and low risk in green. (See the insert for color
representation of this figure.)
Plackett-Burman designs, may be used to reduce the number of variables to be studied in
detail [36, 37, 41]). These screening designs may confound different variable interactions
or even confound main effects with interactions. After the further reduction of important
variables by screening, a more extensive DOEmay be used to refine estimates of variable
effects and interactions. In Fig. 2.7, one such design, a central composite circumferential
design (CCC), is shown. This design assumes that the risk assessment and screening
studies have defined pH and protein load as the important factors in the flow-through
ion-exchange step. In addition to assessing all the interactions at high and low levels (full
factorial), the design includes multiple center points to assess experimental variability
and extended axial points for each parameter (the other parameter is held at a midpoint
value). The interpretation of DOE depends on modeling the relationship between
variables and responses. Amore extensive design, such as CCC, may be used to generate
contour plots linking variables and responses. Such plots can be the basis of a design
space. In Fig. 2.8a, three hypothetical response contour plots for basic variants, acidic
variants, and an impurity are shown. On the basis of the need to achieve product CQAs,
the acceptable limits for all three responses are indicated above the plots. Overlaying the
acceptable areas for all three responses can result in a two-factor design space (Fig. 2.8b).
The use of DOE for chromatography of biotechnology products is not a new
approach[42].DOEhasbeenusedtodemonstrate robustness within the empirical
chromatography parameters. DOE can also be used to optimize chromatography
parameters. For QbD, DOE may be used to establish an initial design space that
extends beyond the empirical ranges used for pilot and full-scale lots. To support a
