Future Directions for the AIM and EDA Concepts - Good Cascade Impactor Practices, AIM and EDA for Orally Inhaled Products - page 394

5. The LPM / SPM ratio works like a charm when the cutoff (size boundary
delineating LPM from SPM ) is set at the MMAD . But if you are away from
the MMAD value, the expected ideal ratio is not 1.0, so what happens to
sensitivity? How far from MMAD could you go?
EDA works as long as the size boundary defi ning the LPM / SPM ratio is within
0.3-3.0. Also, the goal (specifi cation) is not to have the size boundary separat-
ing LPM from SPM set “close to 1.0” but close to whatever value you deter-
mined during development is typical for your product's APSD with the chosen
boundary size.
6. Does FDA have any data on the use of EDA or AIM?
[Mr. Bill Doub, FDA St, Louis, MO]: In my lab, we looked at four CFC MDIs
and fi ve HFA MDIs. (There was a fi fth CFC MDI but it turned out to be expired.)
We attempted to change their APSDs by introducing an inter-actuation delay,
actuator cleaning, and looking at different life stages. We used an Anderson
Cascade Impactor (ACI) and looked at the fi ne-particle dose as well. The exper-
iments were conducted because patients are complaining about the CFC-HFA
difference. Where we saw statistically signifi cant effects with the full-resolution
cascade impactor (FRCI), we applied EDA to see if that method showed similar
effects. Although the EDA data have not yet been internally validated, qualita-
tively we saw the same types of APSD changes as were observed using the
FRCI. Quantitatively, changes were greater with EDA although RSD was also
higher with EDA. For example, the effect of cleaning was 4-11 % from FRCI
data and 24 % for EDA. The effect of delay was 8 % with FRCI and 21 % with
EDA. All LPM / SPM ratios were 0.2-0.3, so we were probably pushing the
limits of sensitivity of EDA. MMAD s were 2.2-2.9

µ m, and the LPM -to- SPM

m.
7. If the ratio [ LPM/SPM ] is more sensitive than stage groupings, could it be used
to predict MMAD ?
LPM / SPM is more sensitive to change in MMAD than stage groupings (see
Chap. 8 ) . However, if you want to predict MMAD , it's better to calculate MMAD
from the cumulative mass-weighted APSD determined using a full-resolution
impactor. MMAD may be estimated using an appropriate method (e.g., probit
analysis). Incidentally, the USP method for MMAD determination assumes a
log-normal APSD, which is not always valid and which can lead to a biased
assessment of MMAD [ 4 ].
[In addition, it may be important to look at the impact of the entire collection
effi ciency curve for each stage rather than just the corresponding stage d 50 val-
ues for the most accurate MMAD assessment with some CIs, in particular the
ACI. The NGI appears to be less prone to such effects, most probably because
stage-to-stage overlap is largely absent and each collection effi ciency curve is
close to being symmetric about its d 50 value.]
8. Does it matter if the impactor is based on the viable versus the nonviable type?
Those differences can become an issue [if the chosen abbreviated CI has a
parent full-resolution CI that is different in internal design, e.g., compare the
FSA (which is based on the ACI and has fl at collection plates) and the ACVI

boundary was set at 2.1 µ