I N S I D E L A B O R A T O R Y M A N A G E M E N T
■
© A O A C I N T E R N A T I O N A L
■
N O V E M B E R / D E C E M B E R 2 0 1 4
23
Thompson also showed that the robust
standard deviation for PT data can usu-
ally be directly compared to the RSD
(R)
determined in a collaborative study (10).
Ellison et al. identified some condi-
tions that must be met in order for PT
data to be used to assess reproducibil-
ity: 1. collection of reproducibility data
is designed into the PT scheme before
the PT scheme is initiated; 2. the can-
didate method should be characterized
for precision and bias in a single-labo-
ratory type of evaluation prior to being
included in a PT validation project;
3. there must be a formal set of method
instructions; 4. some minimum number
of the laboratories participating in the
PT program must use the candidate
method under review; and 5. the PT
scheme must include a range of materi-
als covering the scope of the method.
Using Intermediate Reproducibility to
Determine Measurement Uncertainty
Estimation of measurement uncer-
tainty is an integral part of the modern
accreditation process. ISO 17025
states that measurement uncertainty
must be estimated and made avail-
able if requested by the customer. The
Codex Alimentarius Commission has
guidelines that require laboratories
involved in the import/export of foods to
be accredited and report measurement
uncertainty (11). Historically, AOAC
has relied on the RSD
(R)
as an adequate
estimate of the measurement uncer-
tainty, and therefore AOAC does not
require method developers to calculate
or report measurement uncertainty as a
part of the method evaluation process.
However, measurement uncer-
tainty can be estimated using many
procedures which are described in the
literature (12–14). In principle, two
approaches may be used when calcu-
lating the measurement uncertainty of
a test result: the ‘Top-down’ or Type A
approach which is based upon a statis-
tical evaluation of the test results from
samples that have undergone the entire
analytical process; and the ‘Bottom-up’
or Type B approach in which all pos-
sible sources of variation of the result
are listed separately and the contribu-
tion of each source to the measurement
uncertainty is estimated. The Bottom-up
approach to estimate the uncertainty
of analytical results seems to be rather
impractical for methods of analysis
(15). In practice most laboratories have
used the Top-down or Type A approach,
estimating the measurement uncer-
tainty using the data available from
quality control, sample duplicates, and
method validation, especially intermedi-
ate reproducibility.
There are three kinds of data that
may be used to calculate the expanded
uncertainty (U) using the Top-down
approach:
1. Data from the original validation of
the method
2. Data obtained from collaborative
studies
3. Data obtained within a laboratory
using the method (16)
ISO Technical Standard 21748
“Guide to the Use of Repeatability,
Reproducibility and Trueness Estimates
in Measurements Uncertainty
Estimation” provides several procedures
for the estimation of the measurement
uncertainty using repeatability and true-
ness data. This would make it possible
to determine measurement uncertainty
using only in-house or single-laboratory
data. This may be an attractive option
in lieu of the difficulties in organizing
collaborative studies.
In a recent article entitled The
Estimation and Use of Measurement
Uncertainty for a Drug Test Procedure
Validated According to USP <1225>,
Weitzel illustrated with examples the
procedures to determine measurement
uncertainty from single-laboratory vali-
dation (SLV) data (17). Weitzel used
accuracy, bias, precision, ruggedness,
and intermediate reproducibility data to
calculate the measurement uncertainty.
In a separate communication,
Weitzel also pointed out that that some
AOAC method manuscripts already
include a measurement uncertainty
calculation (18). For example, in AOAC
Official Method
2011.07,
a method for
the determination of vitamins A and E
by UPLC-UV or FLD, the authors use
a simplified approach described by
Barwick and Ellison (19) to calculate
measurement uncertainty using preci-
sion and trueness study data. AOAC
Official Method
2011.12,
a method for
the determination of vitamins D
2
and D
3
in food by UPLC/MS/MS, also includes an
estimate of the measurement uncertainty
calculated using a combination of preci-
sion and analytical competence data.
These method evaluations demon-
strate that calculating measurement
uncertainty from a variety of in-house
or SLV data is a relatively trivial task
if the evaluation studies are properly
planned to consider the necessary data
required to calculate measurement
uncertainty.
On-Site Verification
The main purpose for method vali-
dation is to ensure that an analytical
method designed and developed for a
specific purpose can actually achieve
an acceptable accuracy and precision.
The main purpose for investigating the
reproducibility of a method is to assess
T
he main purpose for method validation
is to ensure that an analytical method
designed and developed for a specific
purpose can actually achieve an acceptable
accuracy and precision.
(Continued on page 24)
83
