Biomedical Engineering Reference
In-Depth Information
since there is no way to compensate. In a dynamic process, information on variability of
inputs or outputs can be used to tune the process (Fig. 2.5b). This information can be real
time or based on off-line testing. Variable input parameters can then be compensated for.
In addition to process monitoring, a dynamic manufacturing system needs flexibility in
setting process parameters or a design space. To allow for process adjustments, the design
space also needs to predict how movement within the space will impact product
attributes. A simple example of compensating for variability, using the design space
shown in Fig. 2.4c, is as follows. The product yield from an upstream process was higher
than expected, and material would need to be discarded based on protein load limits.
However, adjusting the pH could support a higher protein load in the upper left corner of
the design space. This would facilitate taking advantage of variability in the upstream
process.
A design space can be generated for one unit operation, as done for the chromatog-
raphy example above, or for an entire process [11]. In an entire process, such as shown in
Fig. 2.3, final product quality attributes can drive the design of the final manufacturing
step and process inputs for the final step can drive the outputs of the preceding step. This
approach to unit operations can continue upstream until it defines acceptable inputs and
outputs from the initial thawing of a cell bank vial.
2.5 DEVELOPING A DESIGN SPACE
In the previous section, a hypothetical design space for chromatography was presented.
Figure 2.6 describes some of the initial steps for developing such a design space.
It is important to first define the outputs that the manufacturing step will need to
achieve. These will be the responses evaluated in studying the chromatography step.
This requires establishing the CQAs that the final product must meet. After that, the
chromatography output performance measures must be set so that the complete process
will deliver the final product CQAs. The general requirements of the manufacturing
step should also have been considered in the initial process design (e.g., choice of the
methodology).
In Fig. 2.6a, the step requirements are removal of subpotent charge variants and
impurities. A flow-through ion-exchange column is then chosen to achieve these
requirements (Fig. 2.6b). The potential factors that could impact ion-exchange perfor-
mance are listed in a cause and effect diagram, such as the fishbone or Ishikawa diagram
(Fig. 2.6c). The list of factors should be extensive as not to miss any variables. Since
not all the factors can be studied, an assessment of relative risk is performed. A
failure modes and effects analysis (FMEA) is one such risk assessment and management
tool (Fig. 2.6d) that considers the probability, severity, and detectability of an event
[38, 39]. Appropriate cross-discipline expertise [40] and other prior knowledge are
important in generating a meaningful risk assessment.
After the risk assessment, a more limited set of variables can be studied using DOE.
In addition to defining the variables to be studied, appropriate ranges need to be set.
For screening, the ranges should be set above experimental noise and wide enough to
detect variables that matter. Screening approaches, such as fractional factorial or
