What is measured
The introduction of QbD has challenged traditional thinking on pharmaceutical development and manufacturing. There has been
a greater focus on developing science and risk-based control strategies based on product and process understanding over the
more traditional approach of check-box compliance with reliance often on conventional end-product testing.
During process development, analysts are being challenged to provide more information on what process and material attributes
are truly crucial to process performance. Greater emphasis on understanding the role the physical attributes of input materials
play in the process has seen increasing adoption of advanced physical-properties measurement technologies. In-line measurement
technologies coupled with multivariate statistical analysis techniques are being used to provide greater understanding of
what is actually happening during each of the process steps.
Risk assessments of processes are increasingly being used to evaluate the need or value of a particular measurement within
the control strategy to reduce the risk to product quality. A measurement, for example, may be applied to in-process quality
control, or perhaps as an approach to improve detectability of a failure mode examined in a failure mode and effects analysis
The effectiveness of in-process control in assuring product quality has been recognized (8). A number of pharmaceutical companies
have developed manufacturing processes where the controls are in process and a control strategy similar to Figure 1 is used.
Figure 1: Manufacturing process where the contols are in-process.
A PAT-enabled process allows increased sampling of the process material to be performed, thereby increasing the information
about the process. The increased sampling can, however, increase the risk of failing a zero-tolerance specification found
in a uniformity of dosage units test, for example. Pharmacopoeial authorities have recognized this disincentive to the adoption
of large sample sizes and responded by proposing alternative approaches (9).
PAT-enabled processes may also present challenges to the traditional paradigm of batch release based on end-product testing
and specifications. The PAT-enabled process may evaluate different attributes or end-points, use multivariate statistical
analysis of process data, and also require an understanding of the distribution of data, all of which may affect the determination
of appropriate acceptance criteria.
The implications of these changes include tools that support process understanding (e.g., granule porosity), new technologies
and skill sets (e.g., near infrared, imaging, statistics), and new behaviors (e.g., risk assessment vs. check box compliance).