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

net/publication/324733001

Measurement uncertainty and metrological traceability of whole blood

cyclosporin A mass concentration results obtained by UHPLC-MS/MS

Article  in  Clinical Chemistry and Laboratory Medicine · April 2018

DOI: 10.1515/cclm-2018-0120

CITATIONS READS

0 231

3 authors, including:

Pedro Alía Francesca Canalias

Hospital Universitari de Bellvitge Autonomous University of Barcelona
35 PUBLICATIONS   387 CITATIONS    69 PUBLICATIONS   426 CITATIONS   

SEE PROFILE SEE PROFILE

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

metrological traceability of enzyme measurements View project

All content following this page was uploaded by Francesca Canalias on 22 June 2018.

The user has requested enhancement of the downloaded file.


Raül Rigo-Bonnin*, Pedro Alía and Francesca Canalias
Measurement uncertainty and metrological
traceability of whole blood cyclosporin A mass
concentration results obtained by UHPLC-MS/MS
https://doi.org/10.1515/cclm-2018-0120
Received December 21, 2017; accepted February 21, 2018
Abstract
Background: Traceable and accurate results of cyclo-
sporine A (CsA) mass concentrations in whole blood are
required to ensure the monitoring of immunosuppressive
therapy in transplant recipients. Metrological traceability
and measurement uncertainty can allow ensuring reliabil-
ity and comparability of these results over time and space.
In this study, we provide a practical and detailed example
of how the traceability and uncertainty of mass con-
centration of CsA results, obtained using an ultra-high-­
performance liquid chromatography-tandem mass spec-
trometry (UHPLC-MS/MS) procedure, can be described
and estimated.
Methods: Traceability was described mainly according to
ISO 17511 and information obtained from certificates facil-
itated with the manufacturer’s calibrators. Uncertainty
estimation was performed using the bottom-up and top-
down approaches. For the bottom-up approach, the most
relevant sources of uncertainty were identified and later
used to estimate the standard, combined and expanded
uncertainties. For the top-down approach, expanded
uncertainty was estimated directly using intralab quality
control data mainly.
Results: Mass concentration of CsA results was trace-
able to the manufacturer’s product calibrators used to
calibrate the UHPLC-MS/MS procedure. The expanded
uncertainties estimated by the bottom-up and top-down
approaches were 7.4% and 7.2%, respectively.
Conclusions: After performing the bottom-up and top-
down approaches, we observed that their results were
*Corresponding author: Raül Rigo-Bonnin, Laboratori Clínic,
IDIBELL, Hospital Universitari de Bellvitge, L’Hospitalet de Llobregat,
Barcelona, Spain, Phone: +34932607543, Fax: +34932607546,
E-mail: raulr@bellvitgehospital.cat
Pedro Alía: Laboratori Clínic, IDIBELL, Hospital Universitari de
Bellvitge, L’Hospitalet de Llobregat, Barcelona, Spain
Francesca Canalias: Laboratori de Referència d’Enzimologia Clínica,
Departament de Bioquímica i Biologia Molecular, Universitat
Autònoma de Barcelona, Bellaterra, Spain
Clin Chem Lab Med 2018; aop
quite similar. This fact would confirm that the top-down
approach could be sufficient for estimating uncertainty of
CsA mass concentrations in whole blood results in clini-
cal laboratories. Finally, we hope that this study can help
and motivate clinical laboratories to describe metrologi-
cal traceability and to perform measurement uncertainty
studies based on the simpler top-down approach.
Keywords: bottom-up; cyclosporine A; top-down; trace-
ability; ultra-high-performance liquid chromatography-
tandem mass spectrometry (UHPLC-MS/MS); uncertainty.
Introduction
Cyclosporine A (CsA) is an immunosuppressive drug that is
widely administered to recipients of solid organ transplants
[1–3]. Therapeutic drug monitoring of CsA in these patients
is required because of the narrow therapeutic intervals
and significant interindividual variability in whole blood
concentrations. Trough mass concentration of CsA in
whole blood (cCsA) is commonly monitored and generally
regarded as a good surrogate for CsA exposure [1–3]. Thus,
dose adjustments, critical to regulate the appropriate level
of immunosuppression, are made in part based on labora-
tory results. Thereby, traceable and accurate results of cCsA
should be provided by laboratories to ensure the monitor-
ing of immunosuppressive therapy. Application of concepts
like metrological traceability and measurement uncertainty
can help ensure reliability and comparability of these
results over time and space. Additionally, the knowledge of
metrological traceability of the results and the estimation
of measurement uncertainty of measured values are man-
datory for clinical laboratories aiming to achieve ISO 15189
compliance and accreditation [4].
Metrological traceability is defined as “property of a
measurement result whereby the result can be related to a
reference through a documented unbroken chain of calibra-
tions, each contributing to the measurement uncertainty”
[5]. Ideally, traceability should be materialized by relat-
ing a measurement result to a stated reference through
an unbroken chain of calibrations. Stated references may
range from a standard to a certified primary reference
Brought to you by | Universitat Autònoma de Barcelona
Authenticated
Download Date | 4/25/18 1:06 PM
2      Rigo-Bonnin et al.: Traceability and uncertainty of cyclosporine A results obtained by UHPLC-MS/MS
material that embodies a unit of the International System
of Units (SI). These reference materials must have certain
well-defined characteristics, such as homogeneity and
stability, as described in the ISO 17511 [6]. Reference meas-
urement procedures must be characterized, validated and
documented according to specifications outlined in ISO
15193 [7].
The measurement uncertainty can provide a quanti-
tative indication of the quality and accuracy of measure-
ment values. Measurement uncertainty is a non-negative
parameter characterizing the dispersion of the quantity
values being attributed to the measurand based on the
information used [5]. There are two different approaches
for the estimation of uncertainty: the bottom-up and the
top-down [8–10]. In the bottom-up approach, all conceiva-
ble sources of uncertainty are fundamentally considered,
and standard uncertainty is estimated either by a direct
experiment (type A) or from other sources of informa-
tion (type B), or a combination of both to obtain then a
combined uncertainty for, finally, estimate an expanded
uncertainty of a measured value. The top-down approach
considers the uncertainty as a whole. Uncertainties are
evaluated directly using validation data or intralab quality
control data produced by a measurement system.
Unfortunately, no high-order reference material or
reference measurement procedure exists for cCsA in the
Joint Committee for Traceability in Laboratory Medicine
database [11]. This fact compromises the reliability and
comparability of cCsA results and, consequently, hinders
their interpretation. The aim of this study was to describe
the measurement traceability of cCsA results obtained by
a measurement system based on ultra-high-performance
liquid chromatography (UHPLC)-tandem mass spectro-
metry (MS/MS). We also aimed to provide to clinical labo-
ratory specialists a detailed estimation of the uncertainty
of cCsA values using the bottom-up and the top-down
approaches.
Materials and methods
Reagents and materials
LC-MS-grade ammonium acetate (purity ≥ 98%), zinc sulfate
(purity ≥ 99.5%) and formic acid (purity ≥ 98%) were purchased
from Sigma-Aldrich (St. Louis, MO, USA). Liquid chromatography
coupled to mass spectrometry (LC-MS)-grade acetonitrile (quality of
solvent ≥ 99.9%), LC-MS methanol (quality of solvent ≥ 99.9%) and
LC-MS water (quality of solvent ≥ 99.99%) were supplied by Merck
KGaA (Darmstadt, Germany).
A commercially available lyophilized calibrator set ClinCal®
Whole Blood Calibrators, for Immunosuppressants Levels 0–6 (lot: 405)
was purchased from Recipe (Darmstadt, Munich, Germany). To carry
out the experimental part of this study, the Liquicheck™ Whole Blood
Immunosuppressant Control Level 2 (lot: 26242) from Bio-Rad Labora-
tories (Hercules, CA, USA) was used, as well as blood samples from
patients treated with cCsA having values near to those of the quality
control. Calibrator and quality control materials were reconstituted
with LC-water and stored as recommended by manufacturers.
Different sets of ClinMass® internal standard (IS), lyophilized, for
immunosuppressants (lot: 1312) were supplied by Recipe, containing
CsA-D12 (purity and isotopically purity ≥ 98%). The working IS solu-
tion (CsA-D12 25 μg/L) was prepared in acetonitrile and stored as indi-
cated by Rigo-Bonnin et al. [12].
Measurement procedure and equipment
cCsA was measured using a previously described procedure [12]
based on an UHPLC-MS/MS. The measurement system used was
Acquity® UPLC®-TQD® (Waters, Milford, MA, USA). Furthermore, the
following equipment was used: ADA-120/L analytical balance (Adam
Equipment, Bletchley, UK), Biofuge 13 centrifuge (Heraeus Holding
GmbH, Hanau, Germany), Acura® 825 adjustable 100–1000 μL vol-
ume micropipette (Socorex Isba, Ecublens, Switzerland), Nichipet®
EX II adjustable 20–100 μL volume micropipette, Nichimate® ­Stepper
repetitive dispenser (Nichiryo Co. Ltd., Koshigaya-shi, Saitama,
Japan) and 1000-mL and 250-mL BLAUBRAND® volumetric flasks
(BRAND GmbH + Co. KG, Wertheim, Germany).
Methodology
Metrological traceability
Description of metrological traceability was based on the
ISO 17511 [6] and IUPAC [13]. Information related to the
selection of metrological references, calibration hierarchy
and suitably validated measurement procedures, acqui-
sition and verification of end user’s calibrators and end
user’s measurement on system or sample to obtain meas-
urement results was considered. The information was
obtained from certificates and statements facilitated by
the calibrator’s manufacturer (Recipe).
Measurement uncertainty
Bottom-up approach
The steps followed to estimate the uncertainty were as
follows [8, 9]:
Specification of the measurand
The measurand is the quantity intended to be measured
[5]. According to this definition, the measurand should
Brought to you by | Universitat Autònoma de Barcelona
Authenticated
Download Date | 4/25/18 1:06 PM
 Rigo-Bonnin et al.: Traceability and uncertainty of cyclosporine A results obtained by UHPLC-MS/MS      3
be well defined and sufficient information of the meas-
urement should be detailed. Thus, we have described the
quantity according to the IUPAC and IFCC recommenda-
tions [14].
Identification of uncertainty sources
The bottom-up approach requires the development of
a measurement function in which the relation between
measurand values (dependent variable [y]) and rel-
evant influencing factors (independent variables [x]) is
described. We used the cause and effect diagram [9–11]
to identify the different sources of uncertainty, as shown
in Figure 1. Considering that the influencing factors may
themselves be viewed as measurands and depend on
other quantities, we proposed only an approximation of
measurement function:
 AS  by choosing an appropriate distribution of their values, by
 A ⋅ cIS − aw 
cCsA (µg / L) =  IS 
 ⋅ fsol ⋅ fsp ⋅ fMS ⋅ fUHPLC ⋅ fdp ⋅ fδ
 bw
Figure 1: Cause and effect diagram of the most relevant measurement uncertainty sources of cyclosporin A mass concentration in whole
blood using the bottom-up approach.
where As is the peak area obtained for a patient sample;
AIS, the peak area obtained for the IS in the sample; cIS, the
mass concentration of IS in the sample preparation solu-
tion; aw, the intercept of the weighted regression line of
the calibration curve; bw, the slope of the weighted regres-
sion line of the calibration curve; fsol, the factor for the
solutions preparation; fsp, the factor for the sample prepa-
ration; fMS, the factor for mass spectrometer parameters;
fUHPLC, the factor for chromatograph parameters; fdata, the
factor for data processing; and fδ, correction factors for
biases.
Estimation of standard uncertainties
Standard uncertainty is expressed as a standard deviation
or as a relative standard deviation. Therefore, the standard
uncertainties for each uncertainty source were estimated
using either a type A or a type B evaluation. A type A eval-
uation is based on statistical analysis of values obtained
from direct experiments – e.g. values from measurements
Brought to you by | Universitat Autònoma de Barcelona
Authenticated
Download Date | 4/25/18 1:06 PM
4      Rigo-Bonnin et al.: Traceability and uncertainty of cyclosporine A results obtained by UHPLC-MS/MS
repeated under defined conditions. However, it was not
always possible to obtain experimental data for individ-
ual uncertainty sources, so a type B evaluation was used.
Therefore, helpful information needed for a type B evalua-
tion was obtained from calibration certificates and manu-
facturers’ data.
According to the cause and effect diagram (Figure 1),
standard uncertainties were estimated for calibration pro-
cedure, solutions preparation, samples preparation, mass
spectrometer parameters, UHPLC parameters, data pro-
cessing and bias.
Calibration procedure: The standard uncertainty associ-
ated with the calibration procedure is generally close to
the individual uncertainties of the commercial calibra-
tors, which are mainly due to their assigned values, the
pipetted volumes used for their reconstitution and the sta-
bility of reconstituted calibrators, and with the calibration
curve fitting.
According to the manufacturer’s data, the cCsA
assigned values and associated expanded uncertainties
(k = 2) were 25.8 ± 0.64, 49.0 ± 1.1, 95.7 ± 2.5, 181 ± 4, 439 ± 11
and 1243 ± 23 μg/L.
The measuring volume from pipetting was i­ndicated
by the certificate of external calibration as 1000 ± 2.6  μL
(k = 2).
The relative standard uncertainty related to the stabil-
ity of reconstituted calibrators was estimated from experi-
mental data, ≤5.9% for at least 6 months [12].
Furthermore, the standard uncertainty of the indi-
vidual measurements from weighted linear calibration
curves (ufitt) was calculated according to the following
equation [15, 16]:
sw 1 1 ( y0 − yw )2
ufitt = ⋅ + +
bw w0 n b2 ⋅ n
w ∑
( wi ⋅ xi2 − n ⋅ xw2 )
i =1
n
sw =
∑ i =1
wi ⋅ ( yi − (aw − bw ⋅ xi ))2
n−2
n n
xw =
∑ i =1
( w i ⋅ xi )
                yw
n n
where n is the number of experimental points used to
perform the calibration curve; xi, the calibrator values;
yi, the response values for calibrators; wi, the weighting
factor (equal to 1/xi); aw, the y-weighting intercept of the
linear regression equation; bw, the weighting slope of
the linear regression equation; x0, the determined cCsA
of unknown sample calculated through the calibration
curve; y0, the response of x0; w0, the weighting factor for x0;
sw, the estimated standard deviation for x0; x̅w, the weight-
ing mean values of xi; and y̅w, the weighting mean values
of yi.
To estimate ufitt, a blood sample from a patient having
cCsA value near to the internal quality control was pro-
cessed using three different calibration curves.
Solutions preparation: Solutions needed to assess the
measurement procedure were mobile phases A and B,
working IS solution and precipitation solution.
Uncertainties associated with solutions preparation
were mainly due to the purity of reagents, the quality of
solvents, the mass weighted into the balance, the volu-
metric flasks and the pipetted volumes and the molar
mass of reagents. Standard uncertainties were obtained
from manufacturers’ data (purity of reagents and quality
of solvents) and calibration certificates (balance, pipettes
and volumetric flasks). Furthermore, keeping in mind
that the evaporation of organic solvents used to prepare
the working IS solution may lead to changing concentra-
tion of IS over the time and affect to the cCsA values, an
experiment was performed. Different aliquots from the
same patient’s blood sample with a cCsA value near to the
internal quality control – preserved at −80 °C to minimize
the possible loss of stability, were processed in quintupli-
cate using fresh IS working solution and the same solution
after 5 days (maximum time of conservation) at (2–8) °C.
Their uncertainty was estimated from the cCsA percentage
deviation obtained.
Samples preparation: Standard uncertainties associ-
ated to samples preparation were related to the sample,
the working IS and the precipitant solution pipetted
volumes, and the centrifugation was performed during the
protein precipitation process. Uncertainties were obtained
from calibration certificates of dispensers, micropipettes
and centrifuges.
Mass spectrometer and UHPLC parameters: Several
mass spectrometer and UHPLC parameters were consid-
ered as possible sources of uncertainty (Figure 1). Uncer-
=
∑ i =1
( w i ⋅ yi )
tainties associated to these parameters were obtained
from manufacturer’s data.
Data processing: The CsA and CsA-D12 peak areas inte-
gration, peak fitting and peak areas variability, as well as
the retention times peaks variability, were identified as the
main sources of standard uncertainty related to data pro-
cessing. To estimate them, a blood sample from a patient
with cCsA value near to the internal quality control was
processed in quintuplicate. The standard uncertainty
Brought to you by | Universitat Autònoma de Barcelona
Authenticated
Download Date | 4/25/18 1:06 PM
 Rigo-Bonnin et al.: Traceability and uncertainty of cyclosporine A results obtained by UHPLC-MS/MS      5
related to peak integration (uinteg(y)) was calculated as the
relative standard error of the mean values using pooled
relative standard deviations (sinteg(x)), of inter-replicate
data, given by the equation facilitated by the manufac-
turer of mass spectrometer [17]:
m νi ⋅ sinteg
2
(x)
∑ i =1 νi
uinteg ( y ) =
m slope of the function represents the sensitivity coef-
  n
 
2 ficient value, which is expressed as the percentage
n  ∑ i i 
( I ⋅ t )
∑ i=1  Ii ⋅  ti − i=1 n  
 
 ∑ i=1Ii  
sinteg ( x ) =
n
∑ i =1 i
I
where m is the total number of replicates; ti, the ­retention
time value for the observation i; and Ii, the intensity
(signal: number of ions per second) associated to ti.
The standard uncertainty related to peak fitting
(ufitt (y)) was calculated as follows [17]:
uinteg ( y )
ufitt ( y ) =
((2 ⋅WinHalfsize ) + 1)0.5 ⋅ Iterations0.25
where Iterations is the number of smooths applied to the
raw data, and WinHalfsize is the half-size of the smooth-
ing window.
The standard uncertainties related to peak areas and
retention times variabilities were estimated as the relative
standard error of the mean values using their respective
standard deviations obtained in the replication study.
Estimation of combined uncertainty
Once the contribution of uncertainty sources to the
overall uncertainty was quantified as relative standard
deviations, we combined them to give relative combined
standard uncertainty (uc (y)) according to the following
equation:
n
uc ( y ) = ∑(cs ⋅ u ( x ))
i s
i =1
where us(xi) is the relative standard uncertainty and csi,
the sensitivity coefficient, which denotes uncertainty in
(y) arising from uncertainty in (x) [8, 9]. Sensitivity coef-
ficient describes how the measurand value varies with
changes in the value of each source of uncertainty [18].
Sensitivity coefficients were experimentally obtained in
the laboratory following the steps below [19]:
–– Define different values to each uncertainty source
(or measurement condition), in which the value
described in the measurement procedure should be
included.
–– Perform replicates of the cCsA for each condition
defined above.
–– Represent values graphically. The x-axis describes
the values of the uncertainty source, and the y-axis
stands for cCsA in percentages.
–– Calculate the linear regression function of values. The
of measurand value variation per uncertainty source
unit (e.g. X% of cCsA variation per °C on column
chamber).
A critical step in the sample preparation is associated
to the shaking mix time variability of the precipitated
solution. For this reason, the uncertainty source related
to this process was included in the uncertainty estima-
tion, and an experimental study to obtain the sensitivity
coefficient was performed. A blood sample from a patient
with cCsA value near to the internal quality control was
processed in triplicate at shaking mix times of 10, 20, 30,
40 and 50 s.
Furthermore, given the high degree of variability that
different mass spectrometer parameters (cone voltage
and collision energy) and UHPLC parameters (column
chamber and autosampler temperatures) could contribute
in the uncertainty estimation, sensitivity coefficients for
these parameters were obtained experimentally. To esti-
mate them, different blood samples from patients with
cCsA values near to the internal quality control were pro-
cessed in triplicate at cone voltages of 15, 18, 22, 25 and 28
V; at collision energies of 11, 15, 19, 23 and 27; at column
chamber temperatures of 40, 45, 50, 55 and 60 °C; and at
autosampler temperatures of 10, 15, 20 and 25 °C.
Estimation of expanded uncertainty
Relative expanded uncertainty (U) was obtained by mul-
tiplying the relative combined standard uncertainty by an
i
2
appropriate coverage factor (k):
U = k ⋅ uc ( y )
According to the GUM [8], the coverage factor was
obtained from t distribution table with a level of confi-
dence of 95%, according to the effective degree of freedom
(veff ) based on the Welch-Satterthwaite formula:
uc4 ( y )
νeff =
n us4 ( xi )
∑ i =1 νi
Brought to you by | Universitat Autònoma de Barcelona
Authenticated
Download Date | 4/25/18 1:06 PM
6      Rigo-Bonnin et al.: Traceability and uncertainty of cyclosporine A results obtained by UHPLC-MS/MS

where us (xi) is each individual relative standard uncer- relative bias (δr) and its uncertainty (uδ (y)) were calcu-
tainty, and vi is the degree of freedom of us (xi). lated as follows:
x−µ
δr = ⋅ 100
µ
Top-down approach
uδ ( y ) = δ2r + ucal
2
( y ) + up2 ( y ) + uµ2 ( y )

The steps followed to estimate the measurement uncer-

tainty were as follows [9, 10]: (1) specification of the
measurand, (2)  estimation of the uncertainty associated
to the intermediate precision of the measurement system,
(3) identification of any sources of uncertainty that are
not adequately covered by the imprecision data, (4) esti-
mation of combined uncertainty and (5) estimation of
expanded uncertainty. Steps 1, 3, 4 and 5 for the top-down
approach were performed in the same manner as the bot-
tom-up approach described above.
Estimation of the precision of the measurement
system was performed using Liquicheck™ Whole Blood
Immunosuppressant Control Level 2 quality control data.
Specifically, 425 internal quality control values were col-
lected over 11  months from August 1, 2016, to June 31,
2017. The relative uncertainty associated to the measure-
ment system imprecision (up (y)) was calculated as the
relative standard of values using pooled relative standard
deviations, of intermonth data, given by the following
equation:
n
νi ⋅ si2 ( x )
up ( y ) = ∑
i =1
νi
where n is the number of values for the month i; si (x), the
relative standard deviation obtained in the month i; and
vi, the degree of freedom of si.
where x̅ is the mean obtained in our laboratory after col-
lecting 425 internal quality control values; μ, the con-
ventional value calculated as the mean of the means of
all laboratories participating in the UNITY™ program that
use HPLC-MS/MS measurement procedures; ucal (y), the
relative uncertainty associated with the calibration pro-
cedure; up (y), the relative uncertainty associated to the
­measurement system intermediate imprecision; and uμ (y),
the r­ elative uncertainty associated with the assigned value
of the reference material calculated as [20]:
s/ µ
uµ ( y ) = 1.25 ⋅
⋅ 100
p
where s is the robust standard deviation obtained from
all laboratories participating in the UNITY™ program, and
p is the number of laboratories.
A correction factor for bias was applied and included
as source of uncertainty if
δr > 2 ⋅ uδ ( y )
To estimate the bias related to the REC, the ME, the
C-O and the SEL, data from a previously validated meas-
urement procedure [12] were used. Biases were estimated
using the following equations:
δREC = 1 − REC
δME = 1 − ME
δC-O = C-O − 0

Estimation of uncertainty associated to the bias δSEL = SEL − 0

The uncertainty associated to the REC (uREC(y)), the

The different possible sources of bias related to the cali-
ME (uME(y)), C-O (uC-O(y)) and the SEL (uSEL(y)) were calcu-
bration procedure, the recovery of extracted samples
lated as follows:
(REC), the matrix effect (ME), the carryover (C-O) and the
selectivity (SEL) were evaluated using data of a previously uREC ( y ) = δREC
2
+ us2−REC ( y )
validated measurement procedure [12]. Studies of compat-
uME ( y ) = δ2ME + us2−ME ( y )
ibility of measurement results were performed to known if
the biases were significant. uC-O ( y ) = δC-O
2
+ us2−C-O ( y )
Because there is no high-order reference material or
uSEL ( y ) = δSEL
2
+ us2−SEL ( y )
material with a value assigned by a reference measure-
ment procedure to be used, we utilized the Liquicheck™ where us−REC(y) and us−ME(y) were the standard error of the
Whole Blood Immunosuppressant Control Level 2 and the mean values obtained in the recovery and ME of the vali-
UNITY™ Interlaboratory Program (BioRad) data to evalu- dated measurement procedure [12]; us-C-O(y) and u ­ s-SEL(y)
ate the bias associated to the calibration procedure. The were estimated using a right-angled triangle distribution

Brought to you by | Universitat Autònoma de Barcelona

Authenticated
Download Date | 4/25/18 1:06 PM
 Rigo-Bonnin et al.: Traceability and uncertainty of cyclosporine A results obtained by UHPLC-MS/MS      7

(type B approach) from data of the measurement proce- Manufacturer’s standing measurement system.
dure [12]. LC-MS/MS measurement system was used to assign the
Correction factors for biases were applied and final mass concentration of each CsA working calibrator.
included as source of uncertainty if each absolute bias
value was higher than k-times its respective uncertainty. Manufacturer’s product calibrators. ClinCal® Whole
The k values were obtained using the Welch-Satterthwaite Blood Calibrators for Immunosuppressants were the
formula in the same way described above. working calibrators whose CsA values were assigned
using the standing measurement procedure.

End-user’s routine measurement procedure. Routine

Results human samples were processed using the end-user’s
measuring system (Acquity® UPLC®-TQD®) accord-
Metrological traceability ing to the procedure described in Rigo-Bonnin et  al.
[12], and using the product calibrators with associated
The metrological traceability that applies to cCsA in
calibration.
patients’ results was as follows:
Bearing in mind the calibration hierarchy described,
cCsA in patients’ results were metrologically traceable to
Manufacturer’s selected measurement proce-
Recipe’s ClinCal® Whole Blood Calibrators for Immunosup-
dure. Recipe’s selected measurement procedure was
pressants. The metrological traceability chain and calibra-
based on gravimetry. This procedure is neither an inter-
tion hierarchy are shown in Figure 2.
national conventional reference procedure nor an interna-
tional conventional calibrator, and there is no metrological
traceability to the SI. Measurement uncertainty
Manufacturer’s working calibrators. Six standard solu- Bottom-up approach
tions of CsA in a hemolyzed human whole blood. Each
standard solution was prepared from the material C-093 The measurand was defined as the mass concentration
(1.000 ± 0.005 g/L of CsA in acetonitrile) from Cerilliant (μg/L) of the cyclosporine A in human whole blood meas-
Corp. (Round Rock, TX, USA) and a hemolyzed human ured according to an in-house measurement procedure
whole blood solution and, then, subjected to a freeze- using an Acquity® UPLC®-TQD® measurement system. Fur-
drying process. thermore, the quantity can be described as follows:

Manufacturer’s selected Gravimetry

measurement procedure
n
sig
As
Materials prepared from the
material C-093 (Cerilliant Manufacturer’s working
Corp.) and a hemolyzed calibrators Ca
human whole blood solution lib
ra
te

Manufacturer’s standing
Metrological traceability chain

measurement procedure LC-MS/MS

n
sig
As

ClinCal Calibrator for

Relative uncertainty

Manufacturer’s product
immunosuppressants calibrators Ca
(Recipe) lib
ra
te

Clinical laboratory routine

Acquity UPLC-TQD
measurement procedure
n
sig
As

Routine patient
samples

Measurement result

Figure 2: Scheme of metrological traceability chain and calibration hierarchy for the cyclosporin A mass concentration in whole blood results.

Brought to you by | Universitat Autònoma de Barcelona

Authenticated
Download Date | 4/25/18 1:06 PM
8      Rigo-Bonnin et al.: Traceability and uncertainty of cyclosporine A results obtained by UHPLC-MS/MS

Blood–Cyclosporine A; mass concentration (ClinCal®
Whole Blood Calibrators for Immunosuppressants Ref.
9933; UHPLC-MS/MS).
The cause and effect diagram in Figure 1 indicates
the most relevant sources of uncertainty associated to the
cCsA values. In this study, the sources of uncertainty were
identified and classified into seven main contributions
as follows: calibration procedure, solutions preparation,
samples preparation, mass spectrometer parameters,
UHPLC parameters, data processing and bias.
Relative standard uncertainty values obtained and
considered to calculate the relative combined uncertainty
are summarized in Table 1.
Sensitivity coefficients of shaking mix time vari-
ability of the precipitated solution, cone voltage and
collision energy for the mass spectrometer, and column
chamber and autosampler temperatures for the UHPLC
were 0.315% · s−1, 0.152% · V−1, 0.162% · eV−1, 0.177% · °C−1
and 0.196% · °C−1, respectively. Figure 3 shows, as an
example, the estimation of shaking mix time vari-
ability of the precipitated solution of sample sensitivity
coefficient.
The combined uncertainty obtained considering only
the sources with the greatest contribution was of 3.5%
(Table 1). Figure 4 shows a block diagram of the relative
standard uncertainties contributing to the combined
uncertainty of cCsA.
Given that the calculated effective degree of freedom
was 16.7, the coverage factor used to calculate the expanded
uncertainty, from relative combined uncertainty, was
2.110. The relative expanded uncertainty obtained using
the bottom-up approach was 7.4% (Table 1).

Table 1: Uncertainty budget for the measurement of cyclosporin A mass concentration in whole blood using bottom-up approach.

Uncertainty   Category   Variable/   Distribution   Origin of data   Cs  us, %  (Cs · u)2, %

source factor type

Calibration   Weighted linear regression   cCsAcal   A   Experimental   1  1.40  1.96

procedure   aw, bw          
  Calibrators assigned values   cCsAcal   B   Manufacturer   1  1.11  1.23
Calibrators reconstitution   cCsAcal   B   Calibration certificate   1  0.09  0.01
  Calibrators stability   cCsAcal   A   Experimental [12]   1  1.39  1.93
Solutions   Mobile phases   fsol   B   Manufacturer   1  1.19  1.42
preparation         Calibration certificates      
  IS working solution   fsol   B   Manufacturer   1  0.83  0.68
        Calibration certificates      
  Precipitant solution   fsol   B   Manufacturer   1  0.19  0.04
        Calibration certificates      
  Organic solvent evaporation in   cIS   A   Experimental   1  0.28  0.08
IS working solution
Samples   Solutions pipetting and   fsp   B   Manufacturer   1  0.50  0.25
preparation centrifuge calibration       Calibration certificates      
  Shaking mix time variability of   fsp   A   Experimental   0.315  1.05  0.11
the precipitated solution
Mass   Cone voltage   fMS   B/A   Manufacturer/experimental  0.152  1.86  0.08
spectrometer   Collision energy   fMS   B/A   Manufacturer/experimental  0.162  2.15  0.12
parameters   Others   fMS   B   Manufacturer’s data   1  1.62  2.63
UHPLC   Column chamber temperature   fUHPLC   B/A   Manufacturer/experimental  0.177  0.74  0.02
parameters   Autosampler temperature   fUHPLC   B/A   Manufacturer/Experimental  0.196  2.72  0.28
  Others   fUHPLC   B   Manufacturer   1  1.15  1.31
Data   CsA peaks   fdp   A   Experimental   1  0.33  0.11
processing   CsA-D12 peaks   fdp   A   Experimental   1  0.31  0.10

Combined uncertainty, %   3.5

Expanded uncertainty, %, k = 2.110   7.4

Cs, sensitivity coefficient; us (%), relative standard uncertainty; cCsAcal, mass concentration of cyclosporin A calibrators values; aw, inter-
cept of the weighted regression line of the calibration curve; bw, slope of the weighted regression line of the calibration curve; CsA,
­cyclosporin A; fsol, factor for the solutions preparation; IS, internal standard (cyclosporin A-D12); cIS, mass concentration of internal standard
in working IS; fsp, factor for the sample preparation; fMS, factor for mass spectrometer parameters; fUHPLC, factor for chromatograph para-
meters; fdata, factor for data processing.

Brought to you by | Universitat Autònoma de Barcelona

Authenticated
Download Date | 4/25/18 1:06 PM
 Rigo-Bonnin et al.: Traceability and uncertainty of cyclosporine A results obtained by UHPLC-MS/MS      9

115.0 other conditions involving changes, such as different oper-

Cyclosporine A mass concentration, %

ators, solutions, samples and calibrators preparation, lot-

110.0 y = 0.315x + 94.34
R2 = 0.966
to-lot reagents (except for calibrators and internal quality
control), different calibrations, possible variabilities on
105.0 the measurement system parameters and data processing.
The relative uncertainty associated to the measurement
100.0
system intermediate imprecision was 3.50%.
Of all possible sources of uncertainty not adequately
95.0
covered by the imprecision data, only the uncertainty
associated with the assigned values of calibrators was
90.0
0 10 20 30 40 50 60 considered (1.11%).
Time, s As indicated below, biases were negligible, and no
Figure 3: Sensitivity coefficient of shaking mix time variability of source of uncertainty related to bias was considered.
the precipitated solution. Relative uncertainties were combined to obtain a rela-
The linear regression and the determination coefficient (R2) are shown. tive combined uncertainty of 3.7%. Furthermore, given
Symbols are the mean values of replicated cCsA measurements for that the calculated effective degree of freedom was 631, the
each uncertainty source condition. For a shaking mix of 20 s, the 100%
coverage factor used was 1.972 and the relative expanded
cCsA mean value corresponds to a cCsA mean value of 165 μg/L.
uncertainty obtained using the top-down approach was
7.2% (Table 2).
Top-down approach

The sources of uncertainty considered in the top-down Uncertainty associated to the bias
approach were associated with the assigned values of
commercial calibrators, the intermediate precision and Absolute relative bias associated with the calibration
biases of the measurement system. procedure and its expanded uncertainty were 3.1% and
Internal quality control replicates were performed 13.7%, respectively. Furthermore, biases related to the
using the same measurement procedure, which includes REC, the ME, the C-O and the SEL were 0.037, 0.010, 2.20

Figure 4: A block diagram about the relative standard uncertainties contributing to the combined uncertainty of cyclosporin A mass concen-
tration in whole blood for the bottom-up approach.

Brought to you by | Universitat Autònoma de Barcelona

Authenticated
Download Date | 4/25/18 1:06 PM
10      Rigo-Bonnin et al.: Traceability and uncertainty of cyclosporine A results obtained by UHPLC-MS/MS
Table 2: Uncertainty budget for the measurement of cyclosporin A mass concentration in whole blood using top-down approach.
Uncertainty source Category
Calibration procedure Calibrators assigned values
Intermediate precision Internal quality control data
Combined uncertainty, %
Expanded uncertainty, %, k = 1.972
Cs, sensitivity coefficient; us (%), relative standard uncertainty; cCsAcal, mass concentration of cyclosporin A calibrators values; aw,
intercept of the weighted regression line of the calibration curve; bw, slope of the weighted regression line of the calibration curve;
CsA, ­cyclosporin A; CV, coefficient of variation.
and 3.20, respectively; and their expanded uncertainties,
0.093, 0.039, 5.50 and 8.00, respectively.
All biases were lower than their respective expanded
uncertainty values. Thus, biases were negligible, no cor-
rection factor was necessary to be applied and, in con-
sequence, no source of uncertainty related to bias was
considered to estimate the uncertainty associated to cCsA
values.
Discussion
Values of cCsA are commonly used by the clinicians for
monitoring the status of a transplant patient and for check-
ing whether the administered dose of CsA is effective. Thus,
clinical laboratories must provide traceable cCsA results,
and as accurate as possible. To achieve it, several clinical
laboratories use measurement procedures mainly based
on HPLC-MS/MS or UHPLC-MS/MS. Unfortunately, few
of them use metrological traceability and measurement
uncertainty data in order to express the comparability and
accuracy of their reported results, although the importance
of these concepts is increasing [21–25]. Therefore, in this
study, we described the measurement traceability of cCsA
results obtained by a measurement system based on an
UHPLC-MS/MS procedure. Information from certificates
and statements, facilitated by manufacturers of calibra-
tors and reference materials, was sufficient to describe
the metrological traceability. Thus, patients’ cCsA results
were traceable to the Recipe’s product calibrators. We
also aimed to provide clinical laboratory specialists with a
detailed estimation of the uncertainty of cCsA values using
the bottom-up and the top-down approaches, accord-
ing to different guidelines [8–10]. It should be noted that
although the bottom-up is considered the best approach
to estimate the uncertainty, in the actual situation of clini-
cal laboratories this approach is labor intensive, time con-
suming and typically too complex and cumbersome to be
implemented. Although a certain level of statistical skill is
Variable Distribution type Origin of data Cs us, % (Cs · u)2, %
cCsAcal B Manufacturer 1 1.11 1.23
CV A Experimental 1 3.50 12.2
3.7
7.2
necessary to become familiar with the rationale, we have
tried to show it as simple as possible. For the bottom-up
approach, we found that the main contribution to the
uncertainty measurement was the UHPLC autosampler
temperature parameter, followed by all mass spectrometer
parameters, the weighted linear regression and calibrators
stability included into the calibration procedure and the
mobile phases preparation (Table 1 and Figure 4). For the
top-down approach, measurement uncertainty was evalu-
ated directly using intralaboratory quality control data
produced by the measurement system and data related
with the assigned values of calibrators, showing that
the major contribution was from intermediate precision
(Table 2). Our study showed that both approaches gave
similar uncertainty results (7.4% and 7.2%), as observed
by different experts [21, 22, 26, 27]. However, several con-
siderations must be taken into account. To simplify, the
described example of the estimation of uncertainty used
only one cCsA value within therapeutic interval. In a real
situation, considering the heterocedasticity of the meas-
urement systems based on UHPLC-MS/MS, an uncertainty
profile covering the measuring interval should be studied.
In addition, because no high-order reference material or
reference procedures exist for cCsA, a commercially availa-
ble internal quality control was used as reference material
to estimate the uncertainty associated to the intermedi-
ate precision and calibration bias. This material could be
non-commutable with human samples, thus affecting to
the results. Furthermore, to our knowledge, there is only
one published study [28] related to the estimation of meas-
urement uncertainty for cCsA results using a system based
on HPLC-MS/MS. Our expanded uncertainty values differ
substantially from those obtained in the study, perhaps
because of the use of a different measurement system, or
the way the performance characteristics were studied and
the sources of uncertainty considered.
In conclusion, we showed a detailed example to
describe metrological traceability and to estimate the
measurement uncertainty for the cCsA results. The results
Brought to you by | Universitat Autònoma de Barcelona
Authenticated
Download Date | 4/25/18 1:06 PM
 Rigo-Bonnin et al.: Traceability and uncertainty of cyclosporine A results obtained by UHPLC-MS/MS      11
were traceable to the manufacturer’s product calibra-
tors used to calibrate the UHPLC-MS/MS measurement
system. After performing the bottom-up and top-down
approaches, we observed that their results were quite
similar. This fact would confirm that the simpler top-
down approach should be sufficient for estimating uncer-
tainty of CsA mass concentration in whole blood results
by HPLC-MS/MS in clinical laboratories. Finally, we hope
that this study can help and motivate clinical laboratories
to describe metrological traceability and to perform meas-
urement uncertainty studies based on the simpler top-
down approach for this pharmacological quantity as well
as for other quantities that they may measure.
Author contributions: All the authors have accepted
responsibility for the entire content of this submitted
manuscript and approved submission.
Research funding: None declared.
Employment or leadership: None declared.
Honorarium: None declared.
Competing interests: The funding organization(s) played
no role in the study design; in the collection, analysis, and
interpretation of data; in the writing of the report; or in the
decision to submit the report for publication.
References
1. Zhang Y, Zhang R. Recent advances in analytical methods for the
therapeutic drug monitoring of immunosuppressive drugs. Drug
Test Anal 2018;10:81–94.
2. McShane AJ, Bunch DR, Wang S. Therapeutic drug monitoring of
immunosuppressants by liquid chromatography-mass spectrom-
etry. Clin Chim Acta 2016;454:1–5.
3. Freudenberger K, Hilbig U, Gauglitz G. Recent advances in
therapeutic drug monitoring of immunosuppressive drugs. TrAC
2016;79:257–68.
4. International Organization for Standardization. Medical labo-
ratories. Requirements for quality and competence. ISO 15189.
Geneva, Switzerland: ISO, 2012.
5. Joint Committee for Guides in Metrology. International vocabulary
of metrology. Basic and general concepts and associated terms
(VIM), 3rd ed. JCGM 200:2012. http://www.bipm.org/en/publica-
tions/guides/vim.html. Accessed: Jan 11, 2018.
6. International Organization for Standardization. In vitro diag-
nostic medical devices. Measurement of quantities in biological
samples. Metrological traceability of values assigned to calibra-
tors and control materials. ISO 17511. Geneva, Switzerland: ISO,
2003.
7. International Organization for Standardization. In vitro diagnos-
tic medical devices. Measurement of quantities in samples of
biological origin. Requirements for content and presentation of
reference measurement procedures. ISO 15193. Geneva, Switzer-
land: ISO, 2009.
8. Joint Committee for Guides in Metrology. Evaluation of measure-
ment data. Guide to the expression of uncertainty in meas-
urement (GUM). JCGM 100:2008. http://www.bipm.org/en/
publications/guides/#gum. Accessed: Jan 11, 2018.
9. EURACHEM/CITAC. Guide CG4: Quantifying uncertainty in
analytical measurement, 3rd ed. Eurachem, 2012. https://
www.eurachem.org/index.php/publications/guides/quam.
Accessed: Jan 11, 2018.
10. Clinical and Laboratory Standards Institute. Expression of
measurement uncertainty in laboratory medicine; approved
guideline. CLSI EP29-A. Wayne, PA: CLSI, 2012.
11. Joint Committee for Guides in Metrology. https://www.bipm.org/
en/committees/jc/jcgm/. Accessed: Jan 11, 2018.
12. Rigo-Bonnin R, Arbiol Roca A, González de Aledo-Castillo
JM, Alía P. Simultaneous measurement of cyclosporine
A, ­everolimus, sirolimus and tacrolimus concentrations
in human blood by UPLC–MS/MS. Chromatographia
2015;78:1459–74.
13. De Bièvre P, Dybkaer R, Fajgelj A, Hibbert DB. Metrological
traceability of measurement results in chemistry: concepts
and implementation (IUPAC Technical Report). Pure Appl Chem
2011;83:1873–935.
14. International Union of Pure and Applied Chemistry, International
Federation of Clinical Chemistry. Properties and units in the
laboratory sciences. Part X. Properties and units in general clini-
cal chemistry. Pure Appl Chem 2000;72:747–972.
15. Kim JY, Kwon W, Kim HS, Suh S, In MK. Estimation of measure-
ment uncertainty for the quantification of 11-Nor-delta-9-tetrahy-
drocannabinol-9-carboxylic acid and its glucuronide in urine
using liquid chromatography–tandem mass spectrometry. J Anal
Toxicol 2014;38:164–70.
16. Mansilha C, Melo A, Rebelo H, Ferreira IM, Pinho O, Domingues
V, et al. Quantification of endocrine disruptors and pesticides
in water by gas chromatography-tandem mass spectrometry.
Method validation using weighted linear regression schemes. J
Chromatogr A 2010;1217:6681–91.
17. Waters Corporation. MassLynx 4.1: Peak Integration and quanti-
tation algorithm guide–71500122009/Revision A. Milford, MA,
2005.
18. Farrance I, Frenkel R. Uncertainty of measurement: a review of
rules for calculating uncertainty components through functional
relationships. Clin Biochem Rev 2012;33:49–75.
19. International Organization for Standardization. Guidance for
the use of repeatability, reproducibility and trueness estimates
in measurement uncertainty evaluation. ISO 21748. Geneva,
Switzerland: ISO, 2017.
20. International Organization for Standardization. Statistical meth-
ods for use in proficiency testing by interlaboratory compari-
sons. ISO 13528. Geneva, Switzerland: ISO, 2005.
21. Oosterhuis WP, Bayat H, Armbruster D, Coskun A, Freeman KP,
Kallner A, et al. The use of error and uncertainty methods
in the medical laboratory. Clin Chem Lab Med 2018;56:
209–19.
22. Kallner A. Estimation of uncertainty in measurements in the
clinical laboratory. Clin Chem Lab Med 2013;51:2249–51.
23. Comite Europeen de Normalisation. Health Informatics. Expres-
sion of results of measurements in health sciences. ENV 12435.
Brussels: CEN, 2006.
24. Kessler A. Mass spectrometry. A key technique for traceability in
clinical chemistry. Trends Analyt Chem 2016;84:74–9.
Brought to you by | Universitat Autònoma de Barcelona
Authenticated
Download Date | 4/25/18 1:06 PM
 12      Rigo-Bonnin et al.: Traceability and uncertainty of cyclosporine A results obtained by UHPLC-MS/MS
25. Vesper HW, Thienpont LM. Traceability in laboratory medicine.
Clin Chem 2009;55:1067–75.
26. Rami L, Canalias F. An approach to establish the uncertainty
budget of catalytic activity concentration measurements in a
reference laboratory. Clin Chem Lab Med 2015;53:743–51.
27. Lee JH, Choi JH, Youn JS, Cha YJ, Song W, Park AJ. Comparison
between bottom-up and top-down approaches in the
View publication stats
estimation of measurement uncertainty. Clin Chem Lab Med
2015;53:1025–32.
28. Grote-Koska D, Czajkowski S, Klauke R, Panten E, Brand K,
­Schumann G. A candidate reference measurement procedure
for cyclosporine A in whole blood. Accred Qual Assur
2014;19:147–57.
Brought to you by | Universitat Autònoma de Barcelona
Authenticated
Download Date | 4/25/18 1:06 PM