Transfusion and Apheresis Science 51 (2014) 126–131

Contents lists available at ScienceDirect

Transfusion and Apheresis Science

j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / t r a n s c i

Review

Analytical model for calculating indeterminate results interval

of screening tests, the effect on seroconversion window
period: A brief evaluation of the impact of uncertain results on
the blood establishment budget
Paulo Pereira a,*, James O. Westgard b, Pedro Encarnação c, Jerard Seghatchian d
a
Department of Quality Assurance, Portuguese Institute of Blood and Transplant, Avenida Miguel Bombarda 6, 1000-208 Lisboa, Portugal
b Department of Pathology and Laboratory Medicine, University of Wisconsin Medical School, Madison, WI, USA
c
Católica Lisbon School of Business and Economics Research Unit, Catholic University of Portugal, Lisbon, Portugal
d
International Consultancy In Blood Components Quality/Safety Improvement and DDR Strategy, London, UK

A R T I C L E I N F O A B S T R A C T

Keywords:
The evaluation of measurement uncertainty is not required by the European Union regu-
Bias
lation for blood establishments’ laboratory tests. However, it is required for tests accredited
Delta-value
Precision
by ISO 15189. Also, the forthcoming ISO 9001 edition requires “risk based thinking” with
Seroconversion window period risk described as “the effect of uncertainty on an expected result”. ISO recommends GUM
Total analytical error models for determination of measurement uncertainty, but their application is not intend-
ed for ordinal value measurements, such as what happens with screening test binary results.
This article reviews, discusses and proposes concepts intended for measurement uncer-
tainty of screening test results. The precision model focuses on cutoff level allowing the
evaluation of the indeterminate interval using analytical sources of variance. The interval
is considered in the estimation of the seroconversion window period. The delta-value of
patients and healthy subjects’ samples allows ranking two tests according to the proba-
bility of the two classes of indeterminate results: chance of false negative results and chance
of false positive results (waste on budget).
© 2014 Elsevier Ltd. All rights reserved.

Contents

1. Introduction ......................................................................................................................................................................................................................... 127

2. Methods and materials .................................................................................................................................................................................................... 127
2.1. The indeterminate interval concept and the potential impact of uncertain results in blood establishment
clinical decision .................................................................................................................................................................................................................. 127
2.2. Indeterminate interval measurement using analytical bias and precision ...................................................................................... 128
2.3. Impact of indeterminate interval on seroconversion window period ............................................................................................... 129
2.4. The evaluation of the impact of uncertain results on the blood establishment budget ............................................................. 129
3. Results and discussion ...................................................................................................................................................................................................... 129
4. Conclusions .......................................................................................................................................................................................................................... 130
Appendix: Supplementary material ............................................................................................................................................................................ 131
References ............................................................................................................................................................................................................................. 131

* Corresponding author. Department of Quality Assurance, Portuguese Institute of Blood and Transplant, Avenida Miguel Bombarda 6, 1000-208 Lisboa,
Portugal. Tel.: +351 210063047; fax: +351 210063070.
E-mail address: paulo.pereira@ipst.min-saude.pt (P. Pereira).

http://dx.doi.org/10.1016/j.transci.2014.10.004
1473-0502/© 2014 Elsevier Ltd. All rights reserved.
 P. Pereira et al./Transfusion and Apheresis Science 51 (2014) 126–131
1. Introduction
ISO 15189 guideline (2012) requires the estimation of
measurement uncertainty (MU) as well as the specifica-
tion of the performance requirements (quality goal) for MU
and the regular review of the estimated MU [1]. MU deter-
mination is optional on blood establishment laboratories
field, since it is not mandatory by the European Union di-
rectives [2–5] or by US Clinical Laboratory Improvement
Amendments (CLIA) requirements integrated in the Amer-
ican Association of Blood accreditation program [6]. ISO 9001
has been widely applied in European blood establish-
ments [7] and the introduction of “risk based thinking” on
its forthcoming edition could be supported in blood estab-
lishment laboratories by determining MU since it defines
risk as “the effect of uncertainty” (entry 3.09 of [8]).
ISO recommends determination of MU by following the
Guide to the Expression of Uncertainty in Measurement
(GUM) principles [9]. Dimech et al. successfully applied a
GUM modeling approach [10] and an interlaboratory em-
pirical approach [11] to a screening test. However, the GUM
concept is focused on numerical quantity values (entry 1.20
of [12]), rather than ordinal quantity (entry 1.26 of [12]) tests
providing binary results (i.e., positive/negative). For an in
depth discussion of GUM measurement uncertainty ap-
proach, further information can be found elsewhere [9].
Ordinal quantity is defined in the International Vocabu-
lary of Metrology (VIM) as the “quantity, defined by a
conventional measurement procedure, for which a total or-
dering relation can be established, according to magnitude,
with other quantities of the same kind, but for which no
algebraic operations among those quantities exist”.
Pulido et al. proposed for general chemistry the use of
statistical intervals for the estimation of uncertainty of binary
results to evaluate compliance with supervisory limits [13]
and the performance curve to determine the cutoff con-
centration with an uncertainty interval [14]. Several authors
proposed additional alternative (to GUM) concepts for di-
agnostic sensitivity and specificity, such as the Bayesian
theorem on conditional probability [13–16].
When a sample with concentration equal to cutoff is tested
for viral agents, theoretically the chances of yielding a pos-
itive or a negative result are 50/50. Thus, according to the
normal distribution principles there is a 50% chance of getting
a false result (false negative, if the sample is contaminated,
false positive if it is uninfected). During seroconversion
window period (WP) there is a moment when this phe-
nomena occurs, representing a high chance of untrue result.
The effect of uncertainty on the clinical decision value or
cutoff (i.e., indeterminate interval due to analytical error)
constrains the statistical significance of diagnostic results;
consequently its determination is relevant to the screening
test risk evaluation in blood establishments.
Therefore, the model for MU determination should be
chosen according to the measurement purpose, for example,
GUM modular models in reagents manufacturer’s labora-
tories, GUM empirical, probability and other (alternative to
GUM) models in medical laboratories.
Although the international metrology organizations rec-
ommend VIM terminology, it is seldom used for screening
immunoassays since it does not contain the appropriate
127
vocabulary for the immunoassays specific statistical pro-
cedures. Fuentes-Arderiu [17] and Dybkaer [18] have
proposed terminologies suitable for ordinal tests, but they
are not used in common practice. In this paper, the termi-
nology particular to the presented models is adopted.
This article reviews, discusses and proposes a random
and systematic error concept focused on screening immu-
noassays cutoff, as well as its impact on seroconversion
window period and the role of false positive results in the
waste of the blood establishment budget.
The data calculations for these models have been per-
formed using standard spreadsheet software (Microsoft®
Excel® 2013). The spreadsheets validation was done ac-
cording to the U.S. Food and Drug Administration (FDA) The
Office of Regulatory Affairs (ORA) Laboratory Manual prin-
ciples [19]. The spreadsheets supporting the results of this
article are included as supplementary material. They were
intended to perform modeling for this paper to allow easy
handling and automation of calculation, but they are not
suited for the use in laboratory field.
2. Methods and materials
2.1. The indeterminate interval concept and the potential
impact of uncertain results in blood establishment clinical
decision
The indeterminate (or uncertainty) interval (or “gray-
zone”) term is used in this article to describe the uncertainty
interval around the cutoff point where the binary result has
a significant probability to be untrue due to analytical error.
Hypothetically (analytical) MU should be evaluated for a
sample with cutoff concentration; in practice the cutoff con-
centration is affected by between-batch, within-batch,
metrological conditions, etc. So, the cutoff measured in a
screening immunoassay has an associated analytical im-
precision, where there is always an indeterminate interval
at this decision point, which means that results inside (i.e.,
positive/negative or above/below point, respectively) have
a statistically significant probability of being untrue (false
result) [20]. CLSI EP12-A2 guideline describes an indeter-
minate interval model using the term “C5-C95 interval”
defined as “the range of analyte concentrations around the
cutoff such that observed results at concentrations outside
this interval are consistently negative (concentrations < C5)
or consistently positive (concentrations > C95)”. It states that
“observed results at concentrations inside this interval are
not consistent due to imprecision” (entry 5.3 of [20]). C5 rep-
resents the concentration sample with up to 5% false positive
results, and C95 represents the concentration sample with
more than 95% true positive results. This concept was de-
signed for In Vitro Diagnostic medical device manufacturers
with tests under research and development (R&D). Other
definitions of indeterminate interval include the interval
where the results cannot be assured to be true positive or
true negative, or the interval where the binary results are
uncertain [21–24] due to imprecision (entry 2.15 of [12])
and bias (entry 2.17 of [12]). The indeterminate interval can
be taken to be the MU at the cutoff concentration. It is
usually modeled by a confidence interval centered at the
cutoff, whose width is proportional to the MU [25].
128 P. Pereira et al./Transfusion and Apheresis Science 51 (2014) 126–131
Higher uncertainties lead to wider confidence intervals
which lead to a higher risk in the binary decision taken (e.g.,
a blood establishment virology screening immunoassay has
a higher probability (higher risk) of yielding a false nega-
tive result for higher MU).
To recognize the impact of a clinical decision based on
the laboratory outcomes, the intended use of the test result
should be taken into consideration. On a blood establish-
ment virology field it is the assurance of post-transfusion
safety [26]. Therefore, false positive results (α-errors) are a
minor concern, since they do not represent post-transfusion
risk representing increase of cost per reported result. On the
other side, false negative results (β-errors) represent a major
component of post-transfusion residual risk. Test results are
typically expressed as the ratio between the sample’s value
and the cutoff value (thus yielding one for a sample with
cutoff concentration) [27].
Considering the purpose of screening test results, if u is
the (combined standard) measurement uncertainty syn-
onymous of the indeterminate interval width, a blood
establishment should classify as positive, negative or inde-
terminate according to: positive if greater or equal to one,
negative if lower than 1-u, and indeterminate if setting in
the interval [1 − u, 1] [20]. Samples with indeterminate
results should be tested on complementary tests (e.g., con-
firmatory tests, nucleic acid techniques) and/or a subsequent
sample should be collected and tested as a result of
seroconversion WP.
2.2. Indeterminate interval measurement using analytical
bias and precision
Analytical random and systematic error compo-
nents at the cutoff concentration result are major causes
of lack of stability of measurement that influence the
trueness of binary results. The described MU models are
focused on the intended use of blood establishment test
results.
The analytical random error could be estimated through
precision in reproducibility condition of measurement
defined in VIM as the “condition of measurement, out of a
set of conditions that includes different locations, opera-
tors, measuring systems, and replicate measurements on the
same or similar objects” (entry 2.24 of [12]). CLSI EP5-A2
(2004) describes an evaluation protocol for the precision of
quantitative tests also referred as within-laboratory repro-
ducibility, within-device precision or total precision. It can
also be used to evaluate the precision of an ordinal quan-
tity test since the ordinal scale of measurement (including
binary result) is numerical. According to GUM principles,
it is classified as an intralaboratory empirical model [28].
It requires that data to be collected for a sample with con-
centration equal to cutoff concentration during a series of
20 days twice in 2 daily runs to assure focusing on clinical
decision point. This replication experiment fulfills the US
Center for Medicare and Medicaid Services’ guidance for CLIA
recommending a minimum of 20 control samples to esti-
mate precision.
Precision under repeatability conditions r (within-run
precision) (entry 2.20 of [12]) and precision under repro-
ducibility conditions R (within-laboratory precision) are
computed leading to an estimation of (total) precision. Refer
to CLSI EP5-A2 document for detailed calculations [29]. Using
VIM terminology, the estimated precision result is synon-
ymous of “combined standard measurement uncertainty”
u and the product of its result by a factor k larger than the
number one estimates the “expanded uncertainty”, k·u, U
[9]. In general metrology, when the conditions of the Central
Limit Theorem are met, a value for the coverage factor is
taken from a one or two-tailed t-Student distribution with
effective degrees of freedom veff for veff < 5 (please refer to
[30] for the computation of veff). When veff > 5, k is assumed
to be 2, corresponding roughly to the 95% confidence in-
terval. In laboratories statistics is also used k = 1.65 for a one-
sided estimate.
On screening immunoassays, within-laboratory preci-
sion could be intended to access immunoassay quality and
to set its quality goal. The “quality” is considered the inde-
terminate interval expressed by within-laboratory precision
and the “quality goal” the allowable indeterminate inter-
val expressed by the allowable precision (e.g., cutoff ± 30%)
[31]. The goal should be defined by laboratories taking into
consideration the test results’ clinical decision.
When an estimate of bias is available, the total analyt-
ical error (TAE) concept could be used. It is an extension to
the precision featuring both types of error in a single result.
ISO claims “the deviation from the true value is composed
of random and systematic errors. The two kinds of errors,
assumed to be always distinguishable, have to be treated
differently. No rule can be derived on how they combine to
form the total error of any given measurement result, usually
taken as the estimate.” Despite ISO’s statement, TAE is cur-
rently used in reagents’ manufacturer and also in the medical
laboratory field [32].
Westgard et al. (1974) introduced the TAE concept to clin-
ical chemistry [33], defining it as “the net or combined effect
of random and systematic errors” [34]. Another definition
comes from CLSI EP21-A (2003) which defines TAE as the
“result of a measurement minus a true value of the
measurand” containing random and systematic effects (entry
3 of [35]). The TAE measurement is performed combining
the estimate of bias from a screening immunoassay com-
parison of methods study and the estimate of precision from
a replication study, such as that occurs in the EP5-A2 model.
It can be computed through the simplified formula [36]
TAE = biastotal + k·sd where biastotal is the sum of bias com-
ponents and sd is the measurement standard deviation under
reproducibility conditions of measurement (precision) (entry
2.25 of [12]). Thus, it expresses the combined effect of all
the error components arising within the measurement pro-
cedure. CLSI EP21-A presents another estimation approach
requiring a minimum of 120 patient samples intended for
reagents’ manufacturers [37].
The types of bias in immunoassays are classified as (a)
the proportional bias modeled by biasp = (s + (b·s))/(co + (b·co))
where b is the constant slope with y intercept equal to zero,
e.g., proportional bias effect on samples and calibrators used
for cutoff determination (note: bias is zero since (s + (b·s))/
(co + (b·co)) = (1 + b)/(1 + b)·s/co = s/co; and (b) the fixed bias
modeled by biasf = (s + (a + b·s))/(co + (a + b·co)) where b is
the constant slope and a is the y intercept where (a + b·s)/
(a + b·co) ≠ 0, e.g., different bias effect on sample and
 P. Pereira et al./Transfusion and Apheresis Science 51 (2014) 126–131
calibrators such as sample bias due to nonconforming sample
centrifuge and cutoff bias due to nonconforming reagent
storage during transport from manufacturer to laboratory
[38]. The statistical process for estimating the relation-
ships among fixed bias variables is a very complex task to
the laboratory field. The fixed bias components are mea-
surable mainly in R&D of reagents on the manufacturer level.
To exemplify the model, a single screening anti-hepatitis
C virus (HCV) immunoassay was used to provide a common
base for discussion of results (chemiluminescence method).
Test results are expressed by the ratio of the sample’s result
and the cutoff result s/co using the mathematical model
[14] snc ( xNCnc + (0.55 ⋅ xPCnc )), where snc is the sample net
counts expressed by the number of photons in the sample,
xNCnc is the negative calibrator net counts expressed by the
average result from the two lowest replicates out of three,
and xPCnc is the positive calibrator net counts expressed by
the average result from the two highest replicates out of three
[39]. The method is not metrologically traceable. For an in
depth discussion of traceability in medical laboratories,
further information can be found elsewhere [40].
The approaches and results will be discussed in “results
and discussion” section.
2.3. Impact of indeterminate interval on seroconversion
window period
According to [41], “the window period for a test de-
signed to detect a specific disease (particularly an infectious
disease) is the time between first infection and when the
test can reliably detect that infection”. Typically, the WP is
described as the number of days between the day of infec-
tion and the first positive result [42]. The term is also
synonymous of “seroconversion sensitivity” [43]. For the
screening immunoassays used in blood establishment vi-
rology laboratories, the seroconversion WP is the expression
of the time taken for seroconversion. That is the reason for
the examination of blood donors before the donation as part
of minimum eligibility criterion order to suspend or to
exclude candidates with high risk to be on seroconversion
period [2,44].
The performance requirement should be the highest clas-
sification in tests’ rank for seroconversion, commonly
available from commercial panels’ inserts. A minimum
seroconversion WP is also related to a test generation. The-
oretically, the WP could consider the day of the first
indeterminate result sample instead of the day of the first
positive result sample, since the blood donations of inde-
terminate and positive results samples donors shall be
rejected according to a screening algorithm. Depending on
the screening panel, this may result in a shorter interval or
not. For an in depth discussion of seroconversion WP, further
information can be found elsewhere [26].
2.4. The evaluation of the impact of uncertain results on the
blood establishment budget
The delta-value is appropriate to compare test methods
and rank them according to their capabilities to generate
numerical results close to the cutoff, i.e., with a significant
probability to be on the indeterminate interval. It is defined
129
“as the distance of the mean of the distribution from the
cutoff in standard deviation units” [36]. The test with the
least chance of producing indeterminate results should be
preferred, since this choice reduces the chance of lack of
sustainability of blood components production, i.e., de-
creases the probability of blood establishment budget waste.
Consider a sample of patients and another of healthy sub-
jects. Let sdd and xd be the standard deviation and the mean
of log10(s/co), i.e., of the test results for the patients ex-
pressed as the logarithm to base 10 of the ratio s/co. sdnd
and xnd refer to the same quantities for the healthy sub-
jects sample. The delta measurements are given by delta-
plus δ + = xd (log10 s co) sdd (log10 s co) where result is ≥1
and delta-minus δ − = xnd (log10 s co) sdnd (log10 s co) where
the output is <1. The test from a series with delta-plus closest
to one has the highest probability of indeterminate results.
On the other hand, the test from a series with delta-minus
closest to one has also the highest probability of indeter-
minate results. The performance goal is to select the test
with the lowest chance to generate results close to the cutoff
value. The most uncertain test from a series of tests using
the same data for evaluation will have the highest chance
to produce indeterminate results. For an in depth discus-
sion of delta-value in blood establishments, further
information can be found elsewhere [36].
3. Results and discussion
Despite the TAE should be the first choice when com-
pared to the within-laboratory precision, its model is
equivalent to the precision model, since the fixed bias is
unknown. For its determination, the cutoff standard devi-
ation sdco was computed from the ratio between the cutoff
photons count obtained in a series of 17 runs yielding from
May 5 to July 2, 2011 and the average from 370 cutoff
photons count from January 3, 2010 to September 11, 2011
(i.e., closest to true cutoff) where sdco = 0.07. The within-
laboratory reproducibility standard-deviation sdRw was
determined from a human plasma sample ID 957 diluted
1:3 for near cutoff concentration (average result equal to
0.94) from April 3 to April 22, 2011 (20 runs) yielding
sdRw = 0.030. The random error component was estimated
by combining the variances: sd = (sdco2 + sdRw2)½ = 0.08. The
95.5% confidence interval is [0.84, 1]. The quality goals for
indeterminate interval must be defined by each laborato-
ry since they are unavailable in published tables [32]. The
claimed quality goal was [0.70, 1], same as cutoff ± 30%. The
measurement uncertainty is accepted, since the 95% con-
fidence interval is within the goal. The estimation can be
periodically reviewed to verify compliance such as when it
is considered another claimed indeterminate interval or
when critical test error components changed.
It was tested an HCV seroconversion panel to measure
the WP using not only the cutoff but also the validated al-
lowable indeterminate level of 30%. A period of 97 days was
necessary until the test was able to reliably detect the in-
fection of panel patient (see Fig. 1). The panel insert did not
feature the test under evaluation but 11 other tests featur-
ing nine with first positive results from the 97 th day,
one from the 104th day, and another from the 131st day.
Considering the cutoff or the indeterminate interval, the
130 P. Pereira et al./Transfusion and Apheresis Science 51 (2014) 126–131

Fig. 1. Window period in a screening immunoassay for the PHV901 panel (SeraCare Life Sciences, Inc., Massachusetts) featuring 11 genotype 1a samples
from a blood or plasmapheresis donor who seroconverted over the course of their donation history.

patient’s seroconversion panel tested in the assay under eval-
uation, obtained a window period rated at the shorter period
level. The seroconversion WP aims at determining the most
critical measurable bias (biological bias) component in blood
establishment virology immunoassays, since it represents
the major component of the risk of post-transfusion infec-
tion [26].
The definition of the seroconversion WP makes its de-
termination dependent of the seroconversion period from
a single patient when only a panel is tested. The days
between first bleed and the first indeterminate result cannot
be inferred to the patients’ population. Window period is
a property of a test for a patient and its revision should only
happen when another patient panel with shorter period is
available.
The delta-value example used a healthy subject popu-
lation of 257 blood donors with negative results for anti-
HCV, ALT, and HCV RNA in the latest three blood donations
and a patient population containing 17 patients with con-
firmed HCV infection from a serum bank of a Portuguese
blood establishment with positive tests results of immu-
noassay, recombinant immunoblot, and polymerase chain
reaction, and twelve samples from commercial panels (6
samples from a qualification panel QHV711, and 6 samples
from a genotype performance panel PHV350). The samples
were measured by the same test, which is classified in this
model as “test 1”, and by another screening test for anti-
HCV designated as “test 2”. For the “test 1” δ+cand = 3.831 and
for the “test 2” δ+comp = 2.792. From the considerations above,
“test 1” has a lower probability of generating indetermi-
nate results from the infected candidates’ sample, i.e., higher
chance of false negative results. For the “test 1” δ−cand = 9.441
and for the “test 2” δ−comp = 9.404. Despite “test 1” having
a somewhat lower probability of generating indetermi-
nate results from the healthy subject sample, i.e., higher
chance of false positives results, these results do not show
a statistically significant difference. Equating false positive
results to budget waste, the tests are classified in the same
rank. The measurement should be reviewed as stated for
the precision model.
Figure 2 shows a scheme of the presented models at-
tending their role in the post-transfusion safety and blood
establishment budget.
4. Conclusions
The precision model is important to evaluate the screen-
ing immunoassays indeterminate interval due to analytical
error components satisfying ISO 15189 and GUM prin-
ciples. The total analytical error could be alternative;
however, it is impracticable in screening immunoassays in
blood establishments. Since the binary result is influ-
enced by the indeterminate interval, a tripartite reporting
should be used featuring not only positive and negative
results but also indeterminate results. The window period
should be measured using also the indeterminate interval,
since it allows the identification of a most accurate period
without increasing risk to post-transfusion safety. Even when
analytical bias is measurable, the WP traduces a major com-
ponent of biological bias and it is recommended to be
reported with precision/measurement uncertainty. The delta-
value is a useful model not only to measure the chance of
false negative results but also to measure the chance of
waste in blood establishment budget due to false positive
results.
Fig. 2. Models for assuring post-transfusion safety, and a model for the
management of waste in budget due to false positive results.
 P. Pereira et al./Transfusion and Apheresis Science 51 (2014) 126–131
Appendix: Supplementary material
Supplementary spreadsheets to this article can be found
online at doi:10.1016/j.transci.2014.10.004.
References
[1] International Organization for Standardization. ISO 15189 medical
laboratories – requirements for quality and competence. 3rd ed.
Geneva: The Organization; 2012.
[2] The European Parliament and the Council of the European Union.
Commission Directive 2002/98/EC Setting standards of quality and
safety for the collection, testing, processing, storage and distribution
of human blood and blood components and amending Directive
2001/83/EC. OJ 2003;33:30–40. <http://eur-lex.europa.eu/
LexUriServ/LexUriServ.do?uri=OJ:L:2003:033:0030:0040:EN:PDF>;
[accessed 14.07.15].
[3] The Commission of the European Communities. Commission Directive
2004/33/EC Implementing Directive 2002/98/EC of the European
Parliament and of the Council as regards certain technical
requirements for blood and blood components. OJ 2004;91:25–39.
<http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2004:
091:0025:0039:EN:PDF>; [accessed 14.07.15].
[4] The Commission of the European Communities. Commission Directive
2005/61/EC Implementing Directive 2002/98/EC of the European
Parliament and of the Council as regards traceability requi-
rements and notification of serious adverse reactions and
events. OJ 2005;256:32–40. <http://eur-lex.europa.eu/LexUriServ/
LexUriServ.do?uri=OJ:L:2005:256:0032:0040:EN:PDF>; [accessed
14.07.15].
[5] The Commission of the European Communities. Commission Directive
2005/62/EC Implementing Directive 2002/98/EC of the European
Parliament and of the Council as regards Community standards
and specifications relating to a quality system for blood
establishments. OJ 2005;256:41–8. <http://eur-lex.europa.eu/
LexUriServ/LexUriServ.do?uri=OJ:L:2005:256:0041:0048:EN:PDF>;
[accessed 14.07.15].
[6] American Association of Blood Establishments [Internet].
Accreditation program. Bethesda, MD: The Association; 2014. <http://
www.aabb.org/sa/overview/Pages/program.aspx>; [accessed 14.07.15].
[7] European Blood Inspection System. Common European standards and
criteria for the inspection of blood establishments. Europe: The
Organization; 2010.
[8] International Organization for Standardization. ISO/Draft International
Standard 9001 quality management systems – requirements. Geneva:
The Organization; 2014.
[9] Bureau International des Poids et Mesures. JCGM 100 evaluation of
measurement data – guide to the expression of uncertainty in
measurement (GUM 1995 with minor corrections). Sèvres:
The Organization; 2008. <http://www.bipm.org/utils/common/
documents/jcgm/JCGM_100_2008_E.pdf>; [accessed 14.07.15].
[10] Serology Uncertainty of Measurement Working Party. Uncertainty of
measurement calculation for the Abbott Murex HCV EIA using the
guide to the expression of uncertainty in measurement (GUM)
approach. Version 2.0. Fitzroy, Vic: The Working Party; 2005.
[11] Dimech W, Francis B, Kox J, Roberts G. Calculating uncertainty of
measurement for serology assays by use of precision and bias
(Technical brief). Clin Chem 2006;52(3):526–9.
[12] Bureau International des Poids et Mesures. JCGM 200 international
vocabulary of metrology – basic and general concepts and associated
terms (2008 version with minor corrections). 3rd ed. Sèvres:
The Organization; 2012. <http://www.bipm.org/utils/common/
documents/jcgm/JCGM_200_2012.pdf>; [accessed 14.07.15].
[13] Pulido A, Ruisánchez I, Boqué R, Rius F. Estimating the uncertainty
of binary tests results to assess their compliance with regulatory
limits. Anal Chim Acta 2002;455:267–75.
[14] Pulido A, Ruisánchez I, Boqué R, Rius F. Uncertainty of results in
routine qualitative analysis. Trends Anal Chem 2003;22(10):647–54.
[15] Ellison S. Uncertainties in qualitative testing and analysis. Accredit
Qual Assur 2000;5:346–8.
[16] Mil’man B, Konopel’ko L. Uncertainty of qualitative chemical analysis:
general methodology and binary test methods. J Anal Chem
2004;59(12):1128–41.
[17] Fuentes-Arderiu X. Vocabulary of terms in protometrology. Accredit
Qual Assur 2006;11(12):640–3.
131
[18] Dybkaer R. Metrology and protometrology: the ordinal question.
Accredit Qual Assur 2007;12(10):553–7.
[19] U.S. Food and Drug Administration. ORA laboratory manual. Volume
III. 4.5 Development and validation of spreadsheets for calcu-
lation of data. Silver Spring, MD: FDA; 2013. <http://www.fda.gov/
ScienceResearch/FieldScience/LaboratoryManual/default.htm>;
[accessed 14.07.15].
[20] Clinical and Laboratory Standards Institute. EP12-A2 user protocol
for evaluation of qualitative test performance. 2nd ed. Philadelphia:
The Institute; 2008.
[21] Strike P. Test interpretation. In: Strike P, editor. Statistical methods
in laboratory medicine. Oxford: Butterworth-Heinemann; 1991.
[22] Harris EK, Boyd JC. Analytical goals for reference values. In: Harris
EK, Boyd JC, editors. Statistical bases of reference values in laboratory
medicine. New York, NY: Marcel Dekker; 1995.
[23] Ekins R. Immunoassay design and optimisation. In: Price C, Newman
D, editors. Principles and practices of immunoassay. 2nd ed. London:
MacMillan Reference; 1997.
[24] Crowther J. Methods in molecular biology. In: The ELISA guidebook,
vol. 516. 2nd ed. Totowa, NJ: Human Press; 2009.
[25] Kisner H. The gray zone. Clin Lab Manage Rev 1998;12:77–280.
[26] Fiebig E, Wright D, Rawal B, Garrett P, Schumacher R, Peddada L, et al.
Dynamics of HIV viremia and antibody seroconversion in plasma
donors: implications for diagnosis and staging of primary HIV
infection. AIDS 2003;17:1871–9.
[27] Kim S, Kim JH, Yoon S, Park YH, Kim HS. Clinical performance
evaluation of four automated chemiluminescence immunoassays
for hepatitis C virus antibody detection. J Clin Microbiol
2008;46(12):3919–23.
[28] European Federation of National Associations of Measurement, Testing
and Analytical Laboratories (EUROLAB). Technical report No.1 MU
revisited: alternative approaches to uncertainty evaluation. Paris: The
Federation; 2007.
[29] Clinical and Laboratory Standards Institute. EP05-A2 evaluation of
precision performance of quantitative measurement methods. In:
Approved guideline. 2nd ed. Philadelphia, PA: The Institute; 2004.
[30] European Co-operation for Accreditation. EA 4/02 expression of the
uncertainty of measurement in calibration. Europe: The Organization;
1999.
[31] Westgard J, Westgard S. Total analytical error. Clin Lab News
2013;39(9):8–10.
[32] Westgard QC [Internet]. Quality requirements (CLIA and others).
Madison, WI: The Company; 2014. <http://westgard.com/clia-
quality.htm>; [accessed 14.07.15].
[33] Westgard J, Neil Carey R, Wold S. Criteria for judging precision and
accuracy in method development and evaluation. Clin Chem
1974;20:825–33.
[34] Westgard J. Basic method validation. 3rd ed. Madison, WI: Westgard
QC; 2008.
[35] Clinical and Laboratory Standards Institute. EP21-A user estimation
of total analytical error for clinical laboratory methods, approved
guideline. Philadelphia, PA: The Institute; 2003.
[36] Crofts N, Maskill W, Gust I. Evaluation of enzyme-linked
immunosorbent assays: a method of data analysis. J Virol Methods
1988;22:51–9.
[37] Clinical and Laboratory Standards Institute. EP21-A estimation of total
analytical error for clinical laboratory methods, approved guideline.
Philadelphia, PA: The Institute; 2003.
[38] Mo MF. Total analytical error evaluation in quantitative and qualitative
assays. Alexandria, VA: American Statistical Association; 2006.
<http://www.amstat.org/meetings/fdaworkshop/presentations/
2006/PS9_MayMo_Total%20Error_FDAInd%202006.ppt>; [accessed
14.07.02].
[39] Abbott Diagnostics Division. Abbott Prism HCV. Code: 84-5342/R9.
Wiesbaden: Abbott Laboratories; 2008.
[40] Vesper H, Thienpont L. Traceability in laboratory medicine. Clin Chem
2009;55(6):1067–75.
[41] Le T, Bhushan V, Vasan N. First aid for the USMLE step 1. 20th ed.
New York, NY: McGraw-Hill Medical; 2009.
[42] SeraCare Life Sciences, Inc. Hepatitis C seroconversion panel (PHV901)
HCV genotype 1a. Manufacturer data sheet. Massachusetts: The
Company; 2006.
[43] Clinical and Laboratory Standards Institute. M53-A Criteria for
laboratory testing and diagnosis of human immunodeficiency virus
infection; approved guideline. Philadelphia, PA: The Institute; 2011.
[44] American Association of Blood Establishments. Standards for blood
banks and transfusion services. 29th ed. Bethesda, MD: American
Association of Blood Establishments; 2014.