Product Life Cycle Approach to Cascade Impaction Measurements - Good Cascade Impactor Practices, AIM and EDA for Orally Inhaled Products - page 136

of the product, for example from different units, different batches, different life
stages through individual inhaler content testing (as a minimum from beginning
and end of unit) and at various times during stability testing, in numbers suffi -
cient to obtain adequate statistical power.
4. Choose LPM , SPM values and correlate AIM-QC CI-based measurements of
both metrics to their equivalents determined by full-resolution CI measurements
after: (1) selecting an optimum particle size boundary between LPM and SPM ;
and (2) demonstrating preferably that a linear relationship exists between
LPM / SPM and MMAD.
Note the following considerations:
(a) The appropriate boundary between LPM and SPM must be determined by
full-resolution CI.
(b) The traditional coeffi cient of determination ( R 2 ) may not be appropriate for
all cases, when establishing the correlation between AIM- and full resolution-
based metrics. For instance, when the range of MMAD values for a given
product is narrow, the value of R 2 may appear low relative to other products
possessing higher variability in MMAD , even though their correlation is just
as good. This coeffi cient is therefore more appropriate for comparisons of
distributions with similar ranges of MMAD , and not as an absolute indicator
of goodness of fi t. The root mean square error ( RMSE ) divided by the slope
of the linear regression ( b ) is an alternative goodness-of-fi t statistic that may
be more robust in terms of predictive power [ 4 ].
(c) Release batches against specifi cations based on MMAD , together with EDA
metrics, LPM/SPM and ISM , after correlation between EDA metrics obtained
from the full resolution and AIM systems has been established, and the tar-
get profi le has been created. Establishing the correlation could occur either
in development or after approval (depending on when a suffi cient number of
batches is available to justify the proposed approach—a sponsor company
may make that decision based on its own risk assessment).
(d) Based on their regulatory strategy, the sponsor company will also have to
determine if they are going to include the AIM method as the primary APSD
method in the NDA submission (or a similar appropriate application for
product registration in Europe, Canada or other countries), to support the
registration stability program, or whether, the switch to the AIM method will
be a post-approval submission. It is recommended to obtain prior approval
from regulatory agencies if a matrix approach is to be utilized in the NDA
registration stability studies.
(e) For the near term, determination of appropriate acceptance limits for LPM/
SPM ratio and ISM , could be accomplished by developing limits that produce
operating characteristic (OC) curves that match existing approaches (i.e.,
groupings) with respect to type II error (false acceptance) consistent with lim-
its for approved products, to achieve the same minimum acceptable quality
standard. Longer term, QbD is likely to drive the desire for limits driven by
some relationship to product performance, i.e., OIP effi cacy and/or safety.