Biomedical Engineering Reference
In-Depth Information
for biotechnology products may afford more immediate opportunities for generation
of large design spaces. To illustrate approaches to a downstream unit operation, a
hypothetical column purification step will be considered. The use of a design space
approach is compared to more traditional process limits in Fig. 2.4. Although
manufacturing has always had ranges for operating parameters, they have generally been
univariate. The pH and protein load of a chromatography column are used as example
parameters. The linkage of these parameters to quality has been empirical, often based on
the limited ranges used during manufacture of clinical trial material (Fig. 2.4a). Complex
biological products have been defined by their manufacturing processes [34]; the process
is the product. Changes in the manufacturing process often required a clinical trial to
maintain the empirical link between process characteristics and product quality. By the
mid-1990s [35], a specified subset of well-characterized biological products could utilize
biochemical comparability to allow some manufacturing changes in the absence of new
safety and efficacy studies. However, the parameter ranges used in manufacturing clinical
or to-be-marketedmaterial are often limited and do not explore the full extent of the ranges
leading to acceptable product quality. Small-scale studies can be used to support wider
parameter ranges (Fig. 2.4b). However, the validity of the scale-down models needs to be
demonstrated and the column performance measurements used should link to product
performance. Although small-scale studies can expand univariate ranges, they do not
Figure
2.4. Spaces or ranges that can be used in product manufacturing. (a) Biologic and
biotechnology products were historically defined by their manufacturing process. The process
and process ranges were based on the ranges used for the product used in clinical trials showing
safety and efficacy (S&E). Comparability studies allowed for some process changes. (b) These
ranges were often narrow and could be expanded in small-scale models. The use of these wider
ranges in manufacturing was dependent on the validity of the scale-down models and the
performance criteria. (c) Since many variables interact, a multivariate space is more reflective of
reality andmay be generated using designof experiments. Generally, these experiments are also
done at small scale and depend on the validity of the DOE models as well as the scale-down
model and performance criteria.
