Biomedical Engineering Reference
In-Depth Information
avoid confounding variables. Just as for DOE in development, organizing variables so
that raw material lot changes are not confounded will increase the ability to detect a
change in a manufacturing setting and to understand the impact.
11.6 QbD FOR ORGANIZATIONS
Multivariate analysis not only applies to data processing and experimental design but
also applies multiple ways to view a biopharmaceutical manufacturing process. This
chapter so far has used two views, process flow and development timing. A third view
deals with the dimension of corporate organization. In biopharmaceutical companies,
there is a group of people who develop the manufacturing process while another group
carry out the manufacturing. Not surprisingly, groups within a company have specific
requirements to meet their differing goals. Development thrives on more information
about the process. Manufacturing depends on operational simplicity and robustness.
In some ways, these are competing needs: how can simplicity, cost containment, and
increased understanding of raw materials coexist? Of course, these must coexist to
achieve the highest quality products in a consistent manner. The nonobvious point is that
by building in both aspects, the total process achievesmaximal efficiency. The paradox of
DOE holds for raw materials—more measurements really can mean less complexity.
11.6.1 Type I and II Risk
The organizational dimension described above boils down to a difference in information
needs that can be quantitated. The result, leads to an understanding of how to organize
analytical testing schemes applied to raw materials over the life of a product—a
time-dependent QbD approach. The development scientist's view would be one willing
to accept more type I error as defined in statistics. Specifically, a batch of raw material
that would give acceptable outcomes in manufacturing process gets erroneously
identified as unacceptable. The tolerance for type I error is higher because it would
also come with more information about the raw material composition. A manufacturing
scientist's perspective, however, might be the one willing to tolerate more type II errors
and accept batches that differ from each other at the expense of having less information
about a material. The tolerance for each error type will change over the development
timescale for a product and will remain strongly influenced by the outcome of the
manufacturing process. If a batch failure is observed in a manufacturing plant, for
example, the desire for richer information content will increase and so also the
probability of failure. Designing quality into the process for raw materials involves
balancing the risks from type I and type II errors in a way that adapts with experience.
11.7 TESTS AVAILABLE
Small-scale use tests for critical, complex raw materials are widely implemented in the
biopharmaceutical industry as a way to give definitive information about the probability
