Performance Characterization of EDA and Its Potential to Improve Decision Making in Product Batch Release - Good Cascade Impactor Practices, AIM and EDA for Orally Inhaled Products - page 203

behavior of stage groupings used for OIP w9j601 as a function of shifting the MMAD
of the target APSD and superimposes the limits derived from the limiting cumula-
tive APSDs. One-sided limits were established for the groupings related to the upper
and lower tail of the APSD (groups 2 and 4), as is the common current practice,
while a two-sided limit was established for the central grouping (group 3).
Some immediate observations arise from closer inspection of Fig. 8.27 . Firstly, this
presentation supports the concept of using one-sided limits for groups 2 and 4 to
detect shifts in particle size through the mass-weighted median of the APSD ( MMAD ).
In principle, an increase in mass of API on stage group 2 signals an increase in MMAD ,
and conversely, an increase in mass of API collected by stage group 4 signals a
decrease in MMAD . Secondly, the predicted behavior of stage group 3 suggests that it
is unlikely to provide a practical control for changes in MMAD . This plot predicts only
a 4.25% label claim difference corresponding to a 1 μm shift in the value of MMAD .
Further, the relationship between stage group 3 mass content and MMAD is not mono-
tonic, having its maximum value at MMAD = ca. 3 μm.
The final steps to generate OCCs were to compute the probability of acceptance
for each metric based on the estimated precision of each metric ( s = 0.05 for
LPM / SPM and s = 1.0 for stage groupings) with the assumed limits as a function of
the MMAD of the displaced APSDs.
In the case of the stage grouping strategy, the overall requirement includes the
condition that the limits for all groupings are met. The final probability of acceptance
could thus be computed by considering the probability that the assumed limits are
simultaneously met for all stage groupings. As discussed above, the inclusion of a
group 3 requirement is predicted to be problematic. This fact is borne out by the
shapes of the actual OCCs generated by this scheme (Fig. 8.28 ). This illustration
presents the contributions of the individual stage groupings to the overall OCC for
one particular case and illustrates the issue of including a stage group 3 requirement
with respect to monitoring movement in the value of MMAD . As a result of this
finding, an additional set of OCCs were produced that dropped the requirement for
stage group 3 and instead considered the probability of acceptance for meeting
simultaneously only the stage group 2 and 4 requirements.
All resulting OCCs for LPM / SPM and stage groupings are compared and pre-
sented in Figs. 8.29 , 8.30 , 8.31 , 8.32 , 8.33 , 8.34 , 8.35 , 8.36 , and 8.37 . Figure 8.29
illustrates OCCs for LPM / SPM , “stage groupings,” including all three groupings,
and “stage groupings,” considering only groups 2 and 4, all superimposed on the
ideal OCC for comparison. The reader will recall that this particular case illustrates
the results for CFC suspension MDI (product w9j601 ) where the limits were derived
from setting R 2 to equal 0.9.
The issue of including group 3 is apparent in this figure where the probability of
acceptance ( p accept ) never reached 1.0 over the entire range of MMADs that were
considered. This feature was consistently observed over all limits and OIPs that
were considered. In this particular case, the same issue, though to a lesser extent,
was observed for stage grouping when only groups 2 and 4 were considered. At the
same time, the LPM / SPM metric demonstrated superior performance with a corre-
sponding OCC that exhibited appropriate acceptance and rejection characteristics,