A2LA
P103b – Annex: Policy on Estimating Document Revised:
Measurement Uncertainty for Life 3/5/18
Sciences Testing Labs Page 1 of 10
A2LA has compiled information for classifying some common types of test methods to meet the
A2LA Policy on Measurement Uncertainty for Testing Laboratories. The A2LA Policy is
intended to facilitate compliance with ISO/IEC 17025, and is subject to change as additional
guidance is made available internationally.
The annex to this policy, developed and reviewed by the A2LA Life Science Advisory
Committee, provides guidance for categorizing methods when determining measurement
uncertainty. Laboratories must comply with 5.4.6.2 and 5.4.6.3 of ISO/IEC 17025:2005 or 7.6.1,
7.6.2 and 7.6.3 of ISO/IEC 17025:2017 regardless of whether a method is listed as Category I, II,
or III. Thus, there must be a procedure that describes how the laboratory intends to address
measurement uncertainty for all test methods. Appended to this document the reader will find a
Guideline where test methods are grouped by discipline along with their suggested category
designation.
LABORATORY PROCEDURE
The laboratory is required to identify and document the applicable measurement uncertainty
category (I-III below) for each of the test methods identified on the laboratory's proposed scope
of accreditation. This requirement is in addition to the requirement that the laboratory list all
significant components of uncertainty (including sub-sampling where applicable) and make
reasonable estimates as to their magnitude for each accredited test method. Quantitative estimates
of measurement uncertainty are not required for Category I methods and uncertainty can be
estimated by other means for Category II methods.
I. Test Methods that are reported on a qualitative basis, or on a categorical or nominal
scale. These are methods where test items (samples) are classified using visual observation
or other similar methods to determine, detect, or identify the target. The requirement to
calculate measurement uncertainty does not apply to test methods or studies where the end
point is an opinion or diagnosis.
II. Well-recognized test methods are those methods that specify limits to the values of the
major sources of uncertainty of measurement and specify the form of presentation of
calculated results. This category includes:
1) Rapid method kits that specify limits to the values of the major sources (contributors)
of uncertainty, as well as well-recognized rapid methods where kits are used to
determine qualitative results, (for example, a semi-quantitative kit assay that reports
qualitative results such as “presence” or “absence” based on a numeric value).
2) Semi-quantitative test methods where the determination is based on a continuous-
scale measurement.
III. All other test methods, these include chemical, environmental, or biological test
methods based on published regulatory or consensus methods (examples: FDA, EPA,
OECD, AOAC, ASTM, ISO) as well as those test methods needing major (or all)
components of uncertainty identified. In such cases measurement uncertainty estimates
L:\REQUIREMENTS\P103b - Annex-Policy on Estimating Measurement Uncertainty for Life Sciences Testing Labs.doc
A2LA
P103b – Annex: Policy on Estimating Document Revised:
Measurement Uncertainty for Life 3/5/18
Sciences Testing Labs Page 2 of 10
are to be generated, based on appropriate techniques specified below. Laboratory –
developed methods require validation per ISO/IEC 17025:2005, section 5.4.5.2 or
ISO/IEC 17025:2017, section 7.2.2. As part of this validation, the significance of
measurement components or the significance of the modifications of the measurement
components from the standard test method must be considered so that the measurement
uncertainty for the method can be estimated.
NOTE 1: To determine whether the uncertainty affects compliance to a specification limit
(ISO/IEC 17025:2005, 5.10.3.1 or ISO/IEC 17025:2017, 7.8.6, Annex A.2), uncertainty must be
estimated and reported (see EUROCHEM-CITAC Guide Use of Uncertainty Information in
Compliance Assessment 1st Edition 2007).
Estimating Measurement Uncertainty:
The laboratory must have and apply a procedure for identifying the sources of uncertainty
associated with testing technologies and/or test methods. This procedure must identify the
mechanism used for documenting and identifying the major components contributing to the
uncertainty and where applicable, present the calculations used for quantifying the measurement
uncertainty for the test method. The components of uncertainty are to be identified for all test
methods or studies, accompanied by reasonable estimates of their magnitude. Then the estimate
of the measurement uncertainty may be determined from either reference or control samples,
from method validation data, or from combining the individual components. Quantitative
estimates of measurement uncertainty are not required for Category I and may not be required for
Category II.
Category I Test Methods: No calculated estimates of uncertainty are required for test methods
that are qualitative, categorical or nominal scale test methods.
Category II Test Methods: No additional estimates of measurement uncertainty are required if
the laboratory can demonstrate their ability to meet the measurement uncertainty specified in the
test method and its associated reporting requirements. Similarly, no additional estimates of
measurement uncertainty are required for well recognized rapid method kits that produce a
qualitative response.
Qualitative and semi-quantitative tests that are based on continuous or quantitative
responses and have pre-determined cutoff points are influenced by measurement uncertainty.
The effect of the uncertainty can be an incorrect qualitative response. To account for this,
many methods have an allowance for an “indeterminate” response. Therefore, samples where
results are close to the decision point (if available) are those most at risk, and should be the
basis for investigative studies on measurement uncertainty (using, for example, conventional
models for detection limits). In these situations, measurement uncertainty can be expressed as
either:
1. A traditional MU statement for samples at levels near the decision point(s).
2. A statement about false classification rates for results near the decision point(s).
3. Overall rates of correctness for different known classes of samples (e.g., true positives
and true negatives; sensitivity and specificity; etc).
L:\REQUIREMENTS\P103b - Annex-Policy on Estimating Measurement Uncertainty for Life Sciences Testing Labs.doc
A2LA
P103b – Annex: Policy on Estimating Document Revised:
Measurement Uncertainty for Life 3/5/18
Sciences Testing Labs Page 3 of 10
Category III Test Methods: For these methods, MU shall be estimated using available data,
published information, and/or designed experiments, as described in the latest version of the
“A2LA G104 Guide for the Estimation of Measurement Uncertainty in Testing” or other similar
guidance documents. Uncertainty can be estimated using laboratory control samples, method
validation studies, or by an appropriate model for the propagation of error components.
Laboratory control sample (LCS) results may be used to estimate MU, provided the
samples are an appropriate matrix and concentration. Laboratories should follow the
procedures in ASTM E 2554-07: Standard Practice for Estimating and Monitoring the
Uncertainty of Test Results of a Test Method in a Single Laboratory Using a Control Sample
Program. Alternatively, they may estimate uncertainty using the following guidance:
1. When the LCS has been through all method steps, then the laboratory can use the
standard deviation (SDP) from the LCS intermediate precision data as an estimate of
combined standard uncertainty. A relative SD (or CV) may also be used.
2. When the LCS have not been run through all method steps, then the laboratory should
incorporate any appropriate additional components or considerations in the uncertainty
calculations, for example, those uncertainty components from sub-sampling, aliquoting
or sample preparation. The additional components should be combined with SDP using
the root sum square (RSS) method.
3. When a method has a known consistent bias that is inherent to the method (e.g. low
recovery on difficult analytes) the bias must not be added to the uncertainty
calculations. The bias shall, however, be clearly stated and recorded along with the
uncertainty estimate. If a bias adjustment is made prior to reporting a result (e.g.,
adjusting for recovery on a sample that is spiked with a known amount of substance),
then an additional source of uncertainty is introduced and must be included in the
uncertainty estimate. However, if LCS data routinely include adjustments for recovery,
then the error from the adjustment is already included in SDP and does not need to be
added again.
It is recommended that 20 or more individual LCS data points be obtained to estimate
SDP. The estimate of combined uncertainty is then expanded using the formula:
Measurement Uncertainty for a Defined Matrix (LCS) = k x SDP,
where k (the coverage factor) equals 2 (for 95% confidence)
If fewer than 20 LCS results are available, the coverage factor should be the
appropriate t statistic for 95% confidence for the associated number of degrees of
freedom (10=2.228, 20=2.086, 30=2.042, 40=2.021, 60=2.000, 120=1.980 & ∞=1.960,
NIST SP260-100: 1993 Table B.3.4).
NOTE 1: MU estimates from LCS samples should only include data from analysis runs that
were determined to be “in control”, and should exclude data from runs that were determined to
be “out of control” and where reasons for the problem were identified and corrected. When
there was no explanation for the “out of control” signal, it might reflect actual uncertainty and
should be retained in the MU estimate. However, this depends upon what the result of a root
L:\REQUIREMENTS\P103b - Annex-Policy on Estimating Measurement Uncertainty for Life Sciences Testing Labs.doc
A2LA
P103b – Annex: Policy on Estimating Document Revised:
Measurement Uncertainty for Life 3/5/18
Sciences Testing Labs Page 4 of 10
cause investigation revealed, for example if the investigation revealed that the out of control
event was not due to an assignable cause.
NOTE 2: If single LCS results are used in MU calculations but the average of multiple results
is reported to the client, then SDP has to be divided by the square root of the number of
measurements used in creating the average.
NOTE 3: Stated uncertainties for reference materials are usually quite small and are generally
considered to be included in the uncertainty calculations for an LCS that is run through all
method steps. If reference material uncertainties are significant they should be combined with
SDP using the root sum square (RSS) method.
Method validation data may be used to estimate measurement uncertainty if the validation
data were determined by studies that are consistent with ISO 5725 and/or the equivalent
AOAC Harmonized Collaboratively Validated Methods. Use of these data also requires that
the laboratory has demonstrated its competence with the method, as determined by criteria
below.
The laboratory may use a published SD for reproducibility (SDR) as an estimate of combined
standard uncertainty under the following conditions:
1. The validation study included all sources of uncertainty (including sample preparation
and different analysts)
2. The laboratory has acceptable bias
3. The laboratory has acceptable repeatability, or the estimate is modified appropriately.
To demonstrate competence with a method, the laboratory must calculate the SD for
laboratories (SDL), as the quadratic difference between reproducibility and repeatability
(SDr) from the validation study (SDL = √(SDR2-SDr2)). Then the laboratory must estimate
their bias using reference materials or other procedures, and estimate their repeatability using
a replicates study at the appropriate level.
The laboratory must demonstrate competence with the method by showing that:
1. Their Bias < 2SDL
2. Their Repeatability < √Fx(SDr), with F taken from a statistical F table using
appropriate degrees of freedom and 95% confidence. The laboratory has an option to
use 1.5 as a low limit for √F (and therefore a tight criterion).
NOTE: F tables are found in all introductory statistical textbooks and in many computer
packages and calculators. Unfortunately the format varies in different presentations regarding
the numerator and denominator degrees of freedom and significance level ( or /2). For the
purposes of comparing SDr with a lab’s repeatability, use the number of observations used to
estimate the SDr as the numerator degrees of freedom and the number of replicates used to
estimate the laboratory’s repeatability as the denominator degrees of freedom. Look only for
significance at the low end (repeatability much larger than SDr), so use a one-sided F, with
=0.05. As a rough rule, if the repeatability is less than 1.5 times SDr, it is acceptable.
L:\REQUIREMENTS\P103b - Annex-Policy on Estimating Measurement Uncertainty for Life Sciences Testing Labs.doc
A2LA
P103b – Annex: Policy on Estimating Document Revised:
Measurement Uncertainty for Life 3/5/18
Sciences Testing Labs Page 5 of 10
If a laboratory has much lower repeatability than SDr, then this lower estimate should be
combined with the SDL using RSS to obtain a lower estimate of combined uncertainty.
Similarly, if a laboratory has acceptable bias, but their repeatability is larger than the
criterion, then the laboratory may combine their repeatability with the validation study SDL
to obtain a larger combined uncertainty estimate.
If the validation study did not include all steps in the method, then standard uncertainties
from these steps may be added to the SDR with the RSS method.
The estimate of combined uncertainty (usually SDR) is then expanded using the following
formula:
Measurement Uncertainty for a Defined Matrix = k x SDR,
where k (the coverage factor) equals 2 (for 95% confidence)
For test methods that need identification of all components of uncertainty and detailed
measurement uncertainty budgets, these estimates are to be calculated in accordance with
published methods that are consistent with those described in the ISO "Guide to the
Expression of Uncertainty in Measurement" and subsequent guidance.
Additional guidance, including examples of methods that fit the categories (I-III) listed above and
guidance on the calculation of measurement uncertainty for testing laboratories has been
developed and posted on our website at www.A2LA.org. As of the date of this revision, the
A2LA-developed guidance currently on the website includes: “A2LA Guide for the Estimation of
Uncertainty for Testing” and “A2LA Guide for the Estimation of Uncertainty for Dimensional
Calibration and Testing Results.” Links to additional, external guidance documents are also
included. It is suggested that our website be checked frequently, since further guidance
documents will be made available as more information is collected. The most recent version of
this policy will also be posted on our website.
Reporting Measurement Uncertainty:
Measurement uncertainty is to be estimated for all methods in Category III, and is to be reported
when one or more of the following conditions occur:
1. When requested by the client.
2. When required by specification or regulation.
3. When the result is being used to determine conformance with a specification limit.
In these cases, the laboratory must report the expanded uncertainty in the same units as the
measurement result and with the same number of significant digits as the reported value. The
coverage factor must be included in the uncertainty statement. If the MU was estimated using
relative SDs or percentage relative SDs, the percentage must be transformed into the reported
units prior to reporting the uncertainty.
If the method has a known bias and this bias was not adjusted (for example, adjustment for
recovery), this bias should be reported in addition to the result and the uncertainty.
L:\REQUIREMENTS\P103b - Annex-Policy on Estimating Measurement Uncertainty for Life Sciences Testing Labs.doc
A2LA
P103b – Annex: Policy on Estimating Document Revised:
Measurement Uncertainty for Life 3/5/18
Sciences Testing Labs Page 6 of 10
For example, a measurement method has an average recovery of 89% of the target analyte, and
the expanded measurement uncertainty has been estimated as 2.3% at levels below 300ppm. A
test result is 210 ppm, and the result is used to prove conformance with a specification limit of
300ppm. The result could be reported as follows:
Sample result = 210 ppm. The expanded uncertainty of this result is +/- 5ppm, with a coverage
factor of 95%. This method has an average recovery of 92%, or at this level, a possible bias of
23ppm.
DOCUMENT REVISON HISTORY
Date Description
3/5/18 Updated refences to 2017 ISO/IEC 17025 version.
L:\REQUIREMENTS\P103b - Annex-Policy on Estimating Measurement Uncertainty for Life Sciences Testing Labs.doc
A2LA
P103b – Annex: Policy on Estimating Document Revised:
Measurement Uncertainty for Life 3/5/18
Sciences Testing Labs Page 7 of 10
Appendix A
Guideline for Category Determination
Discipline Types and Categories of Types of Test Methods:
Animal Drug Testing Program
Test Method Descriptive test method Category
ELISA Screening (semi-quantitative) II/III
TLC Screening (semi-quantitative) I/II
HPLC Screening (semi-quantitative) II/III
LCMS Definitive Determination (quantitative) III
GCMS Definitive Determination (quantitative) III
Food Chemistry Testing Program
Chromatography
GC Quantitative III
HPLC Quantitative III
TLC/TPC Semi-quantitative/quantitative I/II/III
Combustion Protein by LECO III
Filth-light Semi-quantitative II
Filth-macroanalysis Qualitative I
Kits Semi-quantitative II
Spectrophotometry
GC/MS Quantitative III
GC/MS Qualitative/semi-quantitative II/III
ICP Quantitative III
ICP Qualitative/semi-quantitative II/III
LC/MS Quantitative III
LC/MS Qualitative/semi-quantitative II/III
Wet Chemistry
Acid Digestion/Kjeldahl Quantitative III
Ash Quantitative III
Fat by Hydrolysis Quantitative III
Fat by Soxhlet Quantitative III
Gravimetric Quantitative III
Food Microbiology:
Aerobic/Anaerobic Plate Count Quantitative III
Membrane Filtration Quantitative III
MPN Quantitative III
ELISA Qualitative/Semi-quantitative II/III
PCR Qualitative/Semi-quantitative II/III
Cultures by pathogen/ID Qualitative I
L:\REQUIREMENTS\P103b - Annex-Policy on Estimating Measurement Uncertainty for Life Sciences Testing Labs.doc
A2LA
P103b – Annex: Policy on Estimating Document Revised:
Measurement Uncertainty for Life 3/5/18
Sciences Testing Labs Page 8 of 10
Rapid Check kits Semi-quantitative II
Kits (general) Semi-quantitative II
VIDAS/VITEK/ELFA Qualitative/semi-quantitative II
Discipline Types and Categories of Types of Test Methods:
Veterinary Program:
Pathology Qualitative I
GC/HPLC Quantitative III
Virus Isolation/Identification Qualitative I
Bacterial Isolation/Identification Qualitative I
AGAR Gel Immunodiffusion Qualitative/semi-quantitative I/II
Well defined ELISA kits Semi-quantitative II
ELISA kits lacking defined parameters/controls III
Standard,
Well-defined Serologic Assays with appropriate controls II
(for example, Virus Neutralization, Complement Fixation,
hemagglutination Inhibition, etc.)
DNA Sequencing Qualitative II
Conventional PCR Qualitative/semi-quantitative II
Real-time PCR Semi-quantitative II
Aerobic culture Qualitative II
Anerobic culture Qualitative II
Agar gel immunodiffusion Qualitative II
Agglutination Qualitative I
Competitive ELISA Semi-quantitative/quantitative II/III
Non-compeditive ELISA Semi-quantitative II
Complement fixation* Semi-quantitative II/III
Electron microscopy Qualitative I
Fluorescent antibody Qualitative I
Hemagglutination inhibition Qualitative I
Immunofluorescent antibody Qualitative I
Immunoglobulin Qualitative I
Indirect hemagglutination assay Qualitative I
Immunoperoxidase Qualitative I
Kinetics enzyme-linked assay Semi-quantitative/quantitative II/III
Microscopic agglutination Qualitative I
Minimal inhibitory concentration Qualitative I
Radioimmunoassay Quantitative III
Serum neutralization Semi-quantitative I/II
*The test can be made quantitative by setting up a series of dilutions of patient serum and determining the highest
dilution factor that will still yield a positive CF test. This dilution factor corresponds to the titer, measure of
concentration.
L:\REQUIREMENTS\P103b - Annex-Policy on Estimating Measurement Uncertainty for Life Sciences Testing Labs.doc
A2LA
P103b – Annex: Policy on Estimating Document Revised:
Measurement Uncertainty for Life 3/5/18
Sciences Testing Labs Page 9 of 10
Toxicology
Cytotoxicity (In Vitro)
Agarose Overlay I
MEM Elution I
Direct Contact I
Growth Inhibition I
Colony Assay III
Genotoxicity (In Vitro)
Mouse Lymphoma III
Chromosomal Aberration III
Bacterial Reverse Mutation (Ames) III
Genotoxicity (In Vivo)
Mouse Micronucleus II
Hemocompatability (In Vitro)
Hemolysis Test III
Partial Thromboplastin Time (PTT) III
Complement Activation II
Hemocompatability (In Vivo)
Thromboresistance I
Sensitization (In Vivo)
Guinea Pig Maximization I
Murine Local Lymph Node Assay (LLNA) III
Closed Patch (Buehler) I
Maximization Sensitization Test (MHLW) I
Systemic (Acute) Toxicity (In Vivo)
Acute Systemic Toxicity I
Pyrogen – Material Mediated III
Subacute/Subchronic/Chronic Toxicity (In Vivo)
Subchronic Toxicity I
Implantation (In Vivo)
Muscle Implantatation I
Subcutaneous Implant I
Irritation (In Vivo)
Intracutaneous I
Skin Irritation I
Ocular Irritation I
L:\REQUIREMENTS\P103b - Annex-Policy on Estimating Measurement Uncertainty for Life Sciences Testing Labs.doc
A2LA
P103b – Annex: Policy on Estimating Document Revised:
Measurement Uncertainty for Life 3/5/18
Sciences Testing Labs Page 10 of 10
Intraocular Irritation I
Mucosal Irritation I
Test Method Type Category
Histopathology
Necropsy I
Pathology I
Other Toxicology Studies
In Vivo Pyrogen III
In Vivo Safety Test I
In Vitro Bacterial Endotoxin Test (LAL) II
Technologies:
Clinical Chemistry II/III
Hematology II/III
HPLC/GC for measurement of test article III
Environmental Programs
To be developed
Biosafety Programs
L:\REQUIREMENTS\P103b - Annex-Policy on Estimating Measurement Uncertainty for Life Sciences Testing Labs.doc
