You are on page 1of 2

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/266700238

Recommendation of measurement uncertainty models complying GUM and


alternative for blood bank screening tests results

Poster · October 2014


DOI: 10.13140/2.1.2027.9043

CITATIONS READS

0 171

4 authors:

Paulo Pereira James O. Westgard


Instituto Português do Sangue e da Transplantação, IP University of Wisconsin–Madison
81 PUBLICATIONS   111 CITATIONS    182 PUBLICATIONS   4,637 CITATIONS   

SEE PROFILE SEE PROFILE

Pedro Encarnação Gracinda de Sousa


Universidade Católica Portuguesa Instituto Português do Sangue e da Transplantação, IP
99 PUBLICATIONS   1,033 CITATIONS    39 PUBLICATIONS   288 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

P7-Statistical process control methods to the production of blood components View project

P6-Sampling methods to the statistical control of the production of blood components View project

All content following this page was uploaded by Paulo Pereira on 10 October 2014.

The user has requested enhancement of the downloaded file.


Recommendation of measurement uncertainty models
complying GUM and alternative for blood bank screening tests results
Pereira P, Westgard J, Encarnação P, De Sousa G
Instituto Português do Sangue e da Transplantação, Portugal

Introduction Results
The blood bank virology laboratory with screening tests Expanded uncertainty 95% confidence interval, indeterminate results interval (s/co)
Modeling models Empirical models
accredited by ISO 15189 must determine the
Monte Carlo
measurement uncertainty of the results. The model Partial derivative
simulation
Intralaboratory Interlaboratory EQA/PT
should be selected by the blood bank laboratory staff.
[0.14, 1.86] [0.74, 1.19] [0.74, 1.26] [0.88, 1.12] [0.52, 1.48]
Apart from ISO, the measurement uncertainty is
recognized as critical to the laboratory results' risk Unrealistic due to Unrealistic due
Realistic (i.e., expanded
undetectable to group
evaluation. The presented research recommends a set co-variance
uncertainty∩manufacturer uncertainty)
heterogeneity
of measurement uncertainty models complying the
Guide to the expression of uncertainty in measurement 95% Confidence interval (percentage)
(GUM) and alternative (to GUM), fulfilling ISO 15189 Window Overall Positive Negative Area under
Sensitivity Specificity
period agreement agreement agreement curve
principles, presenting the mathematical models as well
as the pros and cons of each one, considering the 97 days 88.3-100 98.5-100 98-99.9 87.9-100 97.8-99.9 99-100
impact of clinical decision on post-transfusion safety.
The modeling models are focused on measurement
uncertainty components and are complex to implement
Materials and Methods in blood bank; they are important for reagent research
& development at the manufacturer level. The empirical
It was determined the measurement uncertainty results models allow the determination of cutoff uncertainty
fulfilling GUM principles when the goal was the and are easy to use in blood bank. The alternative to
uncertainty on cutoff value, as well as alternative to GUM models outputs show the probability of results in
GUM concepts when the goal was the probability of blood donors population (infected and healthy
results in blood bank population. The determinations of persons), as well the estimation of trueness as the
measurement uncertainty were performed for a single seroconversion window period. The EQA/PT model
screening immunoassay (chemiluminescence method) should only be used when the laboratories’ group is
for the detection of anti-HCV antibodies allowing an homogeneous.
easier comparison of results between models. Measurement uncertainty
requirement
Performed GUM models:
a) Modeling models: partial derivative, and Monte
Cutoff Cutoff uncertainty or Seroconversion
Carlo simulation and; uncertainty probability of results?
Probability
window period

b) Empirical models: intralaboratory, interlaboratory,


and external quality assessment (EQA)/proficiency Reagent
manufacturer or
Are patients and
Blood healthy subjects
testing (PT)). Manufacturer blood bank
field?
bank
Yes
sample available?
No

Performed alternative models: Modeling models:


partial derivative or
Empirical models:
intralaboratory,
Diagnostic accuracy
models when the Agreement of results
a) Seroconversion window period Monte Carlo
simulation
interlaboratory &
EQA/PT
diagnosis is known

b) Diagnostic accuracy models when the diagnosis is


known Diagnostic accuracy
when cutoff varies?
c) Agreement when comparator is other than Overall agreement,
Expanded uncertainty Expanded uncertaitny No positive agreement &
diagnostic accuracy criteria Yes Sensitivty & negative agreement
specificity
d) Receiver operating characteristic (ROC) curve and Receiver operating
characteristic curve
area under ROC curve (AUC) Sensitivity, specificity & area under curve
The GUM modeling spreadsheets required advanced Reported evaluation of measurement uncertainty

mathematical skills since standard uncertainties must


Table 1. Recommended scheme for the measurement
be combined in a mathematical model assuring the uncertainty in blood bank screening laboratory
stoichiometry of reaction according to the rules for
products of variance components. The other models Conclusions
request simpler skills and are well known in blood
Despite all models are practicable in blood bank, they
banks laboratories.
should be chosen according to their relevance,
considering principally the roles of indeterminate
References results interval, seroconversion window period,
BIPM (2008) JCGM 100 Evaluation of measurement data - Guide to the sensitivity and specificity on post-transfusion
expression of uncertainty in measurement. JCGM, Sèvres risk/safety. The agreement of results should only be
S L R Ellison and A Williams (2012) Eurachem/CITAC guide: Quantifying considered when the diagnostic accuracy models are
uncertainty in analytical measurement. 3rd ed.
CLSI. EP12-A2 User protocol for evaluation of qualitative test performance not practicable, which is unexpected in blood bank. The
2nd ed. Wayne, PA: CLSI; 2008. area under ROC curve is a supplemental
CLSI. EP24-A2 Assessment of the clinical accuracy of laboratory tests using determination, since the mathematical model of
receiver operating characteristic curves 2nd ed. Wayne, PA: CLSI; 2011. commercial tests cutoff is unchangeable in blood bank.

View publication stats

You might also like