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 188

Table 8.5 (continued)

95% Prediction
Interval

Precision/
Total
Variability
(%)

CI
Stage
d50(µm)

average
MMAD
(µm)

Lower
limit

Upper
limit

Discrimination
index(mm-1) Product Metric

w9k001 Ratio 3.3 3.54 3.504 3.575 13.96 14.8

Group 1 2.880 4.200 0.76 272.8

Group 2 3.339 3.741 2.48 83.3

Group 3 0.974 6.105 0.19 1060.7

Group 4 3.329 3.751 2.36 87.4

w9kw01 Ratio 3.3 2.86 2.807 2.910 7.43 19.6

Group 1 2.091 3.630 0.65 292.1

Group 2 2.654 3.066 2.42 78.4

Group 3 1.753 3.967 0.45 420.3

Group 4 2.718 3.001 3.53 53.7

Discrimination Index Color Code Range

>10
>5
>1
<1

An alternative approach commonly used in MSA is to consider the levels of
discrimination possible with a particular measurement within a critical range, typi-
cally the specification requirements. In the present case, MMAD did not have spe-
cific requirements, but a reasonable range can be assumed and applied to all metrics
(i.e., EDA and grouped stages), leading to a valid relative comparison of perfor-
mance. A range of 1 μm aerodynamic diameter was assumed, and figures of merit
were computed as 1.0 μm divided by the width of the transposed 95% prediction
interval at the mean value of MMAD from the inverse regression approach above.
The resulting values are tabulated in Table 8.5 . The discrimination index was much
higher for the ratio metric, LPM / SPM , compared to the grouped stages, again show-
ing the ability of the ratio metric to detect smaller shifts or changes of the MMAD ,
compared to grouped stages.
The precision to total variability figure of merit was also examined through a
graphical representation. In Fig. 8.12 , 95% prediction intervals for these metrics
were compared to intervals constructed to encompass the total variability for each