Biomedical Engineering Reference
In-Depth Information
primary effects and interactions between variables during process characterization
studies. Moreover, the large numbers of variables that require testing are best managed
through the simultaneous testing of multiple variables, which DOE provides. These
experiments should cover wide ranges for process variables and material attributes and
be designed to determine failure limits and acceptable ranges for the manufacturing
process. An in-depth discussion of DOE can be found elsewhere [6, 7].
Given the number of variables associated with downstream processing of biophar-
maceutical products, establishing an expanded design space for all operational para-
meters for all unit operations can be a daunting and ultimately unnecessary task. A risk
assessment will help identify the subset of process steps most likely to benefit from an
expanded design space. The selection criteria for such an analysis would include
considerations such as
(1) impact of a purification step on critical product attributes, (2)
potential factors that impact process consistency (e.g., yield), and (3) opportunities for
manufacturing operational flexibility.
In line with standard risk assessment practices
[8], these elements would be evaluated in terms of probability of occurrence and severity
of the consequences and prioritized appropriately. The timing of such an analysis would
depend on the availability of sufficient process and product understanding, but would
best be performed with an established purification process prior to extensive process
characterization.
Once process steps are selected, the early identification and focus on critical and key
operational parameters for each process step can further reduce the workload to a
manageable number of parameters that can be investigated with a reasonable amount of
time and resources. “Key” and “critical” operating parameters are terms adopted from
PDATechnical Report 42 [9]. Critical operating parameters affect critical product quality
attributes when varied outside of a narrow (or difficult to control) operating range. Key
operational parameters also have a narrow (or difficult to control) range. Key operating
parameters, however, affect process performance (e.g., yield, duration), but not product
quality. The remaining (nonkey) parameters can affect process or product but are easily
controlled within wide acceptable limits. Much of this knowledge may be available from
earlier process development and fromprevious experiencewith similar processes. At this
point, however, if there is large number of process variables to consider, additional
parameter screening with low-resolution DOE (resolution III or IV) [6, 7] may be
required to identify the main effects. Parameters showing minor effects on process or
product across wide ranges (i.e., nonkey parameters) are less pertinent to process control
and can be excluded from further process characterization studies.
Screening studies are followed by higher resolution designs (e.g., resolution V or
response surface) powered to address interactions between variables. In addition, if
nonlinear responses to input parameters were detected during screening studies, response
surface DOE should be used to map curvature of the operating space [6]. The goal of
these higher resolution studies is to develop a predictive model, determine failure points
and establish design space limits for critical and key parameters.
Given the number of experiments required to establish design space, the majority of
these studies will take place at laboratory scale. Thus, the successful implementation of
the design space aspect of quality by design is contingent upon carefully designed
scale-down models that are representative of the manufacturing-scale operation. The
