Theoretical Basis for the EDA Concept - Good Cascade Impactor Practices, AIM and EDA for Orally Inhaled Products - page 153

in order for the product to be considered “fit for purpose.” In contrast, the expectation
in the European Union and Canada is that acceptance criteria are established for
“fine particle dose/mass” where FPM is defined as the mass of active ingredient in
the collected size fraction below about 5 μm aerodynamic diameter [ 22 , 23 ].
EDA was developed primarily as a quality control tool with the intended purpose
of detecting aberrant aerodynamic particle size distributions (APSD) [ 11 ]. In other
words, EDA was designed primarily to address the QC decision: accept or reject a
batch (with respect to APSD). In order to characterize the performance of EDA, the
Cascade Impactor Working Group (CI-WG) has utilized CI data gathered by
IPAC-RS and used three fundamental statistical approaches: measurement system
analysis (MSA), operating characteristic curves (OCCs), and principal components
analysis (PCA). Basic background information on measurement processes and
these two statistical approaches is given in the following section to aid the reader in
assessing the subsequent material that compares the performance of EDA to current
approaches.
Making good quality decisions relies on the following principles:

1. Measuring the right quality attributes
2. Understanding how well one can measure those attributes and apply the
measurements to quality decisions

The work of the CI-WG has paralleled these concepts. To understand what to
measure, the group has explored from a risk assessment perspective what impacts the
aerosol properties of OIPs. To characterize the measurement capabilities, the group
has evaluated the accuracy and precision of proposed measurements and applied the
findings from these investigations and other related information to undertake formal-
ized statistical analyses evaluating the ability to make the correct quality decision
concerning the release, rejection, or recall of a batch of pharmaceutical product. This
approach quantifies the fundamental uncertainties in the CI measurement system that
give rise to type I and II errors, cast in terms of QC decisions as the probability of
rejecting acceptable product and releasing unacceptable product, respectively. Thus,
the capability of a QC test of metrics derived from CI-generated APSD data can be
judged based on a characterization of these error rates.

7.5 Defining the EDA Metrics and Their Background

The EDA metrics were developed out of a need to have a practical set of metrics for
OIP QC purposes. Mathematically, APSD is a multivariate measurement (i.e., it
requires an array of numbers to describe it). For QC purposes, it is desirable to have
univariate metrics (i.e., single-numbered) that sufficiently describe the distribution
and that are sensitive to variation in the original APSD (Fig. 7.6 ).
These ideas have driven the selection of the two EDA metrics, with the underly-
ing intent that both metrics can be easily obtained from CI-generated data.