Microchemical Journal 79 (2005) 357 – 365

www.elsevier.com/locate/microc

ICP-MS determination of Pt in biological fluids of patients treated

with antitumor agents: evaluation of analytical uncertainty
M. Bettinelli*
Laboratorio di misure ambientali e tossicologiche, Fondazione S. Maugeri, Via Ferrata 8, 29100 Pavia, Italy

Accepted 17 August 2004

Available online 14 November 2004

Abstract

The estimation of the uncertainty associated to the analytical methods is necessary in order to establish the comparability of results.
Methods of Pt determination in biological fluids lack very often of information about uncertainty of results, with likely implications when
results are used to interpret the mechanism of action of platinum compound or when they are considered to optimise the clinical therapies.
An inductively coupled plasma-mass spectrometer (ICP-MS) method for the determination of Pt in biological fluids (plasma, ultrafiltrate
and urine) of patients treated with antitumor agents has been developed and validated. The limits of quantification (LOQ) in the three
matrices were 1.0, 0.1, and 2.0 Ag/l, respectively.
Intraday and interday precisions and accuracies were in good agreement with the FDA criteria for the validation of analytical methods.
The validation study was implemented by assessing the uncertainty evaluation for Pt determination in the different matrices according to
EURACHEM/CITAC Guide.
D 2004 Elsevier B.V. All rights reserved.

Keywords: ICP-MS; Plasma; Ultrafiltrate; Urine; Platinum

1. Introduction absorption and to decrease the metabolic activation through

Chrysoteraphy with Pt-based drug is gaining increasing
importance, especially in the treatment of solid tumors:
testicular and ovarian tumors; head, neck and bladder
carcinomas; brain malignancies [1,2].
A variety of oral analogues of known antitumor agents
have been developed in the past few years for economic,
pharmacological and practical reasons. To merit further
clinical development, the oral analogue should have a
preclinical antitumor activity and toxicity comparable to
those of parental compound, no or only limited gastro-
intestinal toxicity, acceptable bioavailability, and safe and
reproducible pharmacokinetic profile.
A series of ammine/amine platinum (IV) dicarboxilates
of higher lipophilicity and stability than cisplatin and
carboplatin have been developed in an effort to improve
* Present address: C.M.B. Central Laboratory, 29100 Piacenza, Italy.
E-mail address: maurizio.bettinelli@libero.it (M. Bettinelli).
0026-265X/$ - see front matter D 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.microc.2004.08.003
processes in the gastrointestinal track [3].
From this point of view, it is essential to correctly
interpret the mechanism of action of platinum compound in
order to optimise the therapies which have been proposed to
date and to develop adequate analytical methods for
monitoring the Pt levels in biological fluids.
Method validation is the process of proving that an
analytical method is acceptable for its intended purpose. For
pharmaceutical methods, guidelines from United States
Pharmacopeia (USP) [4], International Conference on
Harmonization (ICH) [5], and the Food and Drug Admin-
istration (FDA) [6] provide a framework for performing
such validations. In general, methods for regular submission
must include studies on specificity, linearity, accuracy,
precision, range, detection limit, quantitation limit, and
robustness.
Today, it is generally acknowledged that the fitness for
the purpose of an analytical result cannot be assessed
without some estimates of the measurement uncertainty to
358 M. Bettinelli / Microchemical Journal 79 (2005) 357–365
compare with the confidence required. The Guide to the
Expression of Uncertainty in Measurement (GUM) pub-
lished by ISO [7] establishes general rules for evaluating
and expressing uncertainty for a wide range of measure-
ments. The guide was interpreted for analytical chemistry
by EURACHEM in 2000 [8]. The approach described in
the GUM requires the identification of all possible sources
of uncertainty associated with the procedure, the estimation
of their magnitude from either experimental or published
data, and the combination of these individual uncertainties
to give standard and expanded uncertainties for the
procedure as a whole. However, the GUM principles are
significantly different from the methods currently used in
analytical chemistry for estimating uncertainty which
generally make use of bwhole methodQ performance
parameters, such as precision, recovery, and ruggedness
test. The purpose of this paper is to underline that, if
validation studies are properly planned and executed, the
data produced can be easily handled in the uncertainty
estimation process.
2. Experimental
2.1. Reagents and standard solutions
– Platinum (PtCl2, 10% HCl) and iridium (IrCl3d 3H2O,
10%HCl) were provided by BDH Laboratory Supplies,
England.
– Nitric acid 65% Suprapur was provided by Merck,
Darmstadt, Germany.
– Water used for the inductively coupled plasma-mass
spectrometer (ICP-MS) analysis was prepared in-house
using a Milli-RO 10 PLUS, Milli-Q 185 PLUS purifier
(Millipore, Bedford, USA).
Platinum chloride stock solution of 1 g/l was diluted
(1:10 v/v) in ultrapure water (Milli-RO 10 PLUS, Milli-Q
185 PLUS, Millipore) to obtain working solutions for the
preparation of calibration curves and quality control
samples.
In order to reduce the matrix effect, Ir was used as
internal standard. The iridium chloride stock solution of 1 g/
l was diluted (1:10 v/v) in ultrapure water to obtain the
working solution.
2.2. Preparation of calibration curves and quality control
samples
Aliquots of the working solutions were spiked into blank
human plasma (diluted 1:30 v/v with 1% HNO3 v/v) to
reproduce concentrations of 1, 10, 50, 100, and 200 Ag/l and
into plasma ultrafiltrate samples (diluted 1:30 v/v with 1%
HNO3 v/v) to reproduce concentrations of 0.01, 0.1, 1, 10,
and 20 Ag/l. These samples were used as calibration
standards. The same working solutions were used for the
preparation of the quality control samples at three different
concentrations: 2, 41, and 152 Ag/l for blank human plasma
and 0.02, 6, and 18 Ag/l for blank plasma ultrafiltrate.
Aliquots of the working solutions were spiked into blank
human urine (diluted 1:50 v/v with 1% HNO3 v/v) to
reproduce concentrations of 2, 10, 50, 100, and 200 Ag/l.
These samples were used as calibration standards. The same
working solutions were used for the preparation of the
quality control samples at three different concentrations: 5,
40, and 160 Ag/l. Iridium was used as internal standard in all
the samples employed for the preparation of calibration
curves and quality control samples (1 Ag/l for plasma and
ultrafiltrate, 3 Ag/l for urine sample).
2.3. Sampling procedure
Five-milliliter blood samples were collected in tube
containing 3 ml of EDTA and immediately centrifuged at
2000
g for 15 min at 4 8C, and the plasma was removed.
Plasma was separated and divided in two portions, one
for the determination of total Pt (TPt) and the other one for
the preparation of free Pt species (UPt). Separation of UPt
species was performed by centrifugal ultrafiltration using a
modified version of the method of Bannister et al. [9]. A
portion of 2 ml from each plasma sample was placed in a
microconcentrator Centricon-10 (Amicon Dan, MA) having
a cut-off value of 10,000 daltons and centrifuged at 2000

g for 180 min at 4 8C.
Two urine samples (fraction 0–8 and 8–24 h) were
collected during the treatment on days 1 and 14, and the
volume was measured and stored at 20 8C until analysis.
2.4. Sample handling
Plasma, ultrafiltrate, and urine samples were thawed in a
warm bath immediately before analysis.
Plasma and ultrafiltrate samples were diluted (1:30 v/v)
in polypropylene tubes with 1% HNO3 (v/v). After addition
of 1-Ag/l Ir as internal standard, the samples were analysed
by ICP-MS. The sample uptake rate was of 1.0 ml/min.
Table 1
Instrumental specifications and analytical conditions
195
Element/mass Pt
193
Internal standard Ir
Replicate time (ms) 900
Dwell time (ms) 300
Scanning mode Peak hopping
Sweeps/reading 3
Number of replicates 10
Points/spectral peak 1
Resolution Normal
Power 1100 W
Plasma argon 12 l/min
Auxiliary argon 0.80 l/min
Nebulizer argon 0.99 l/min
 M. Bettinelli / Microchemical Journal 79 (2005) 357–365 359

Table 2 3. Results and discussion

Natural abundance and potential interferences of Pt isotopes
Pt mass Abundance (%) Potential interferences 3.1. Specificity
189 0.013 Os, YbO, HfO
191 0.780 Os, YbO, HfO, LuO The two major sources of analytical inaccuracy during
193 32.900 YbO, HfO
ICP-MS analyses are matrix-induced signal suppression and
194 33.800 HfO
195 25.300 Hg, HfO, WO, TaO spectral interferences.
197 7.210 Hg, HfO, WO, TaO In the case of matrix-induced signal suppression, the ion
intensity of an analyte element is somewhat dependent upon
Urine samples were diluted (1:50 v/v) in polypropylene the total composition of the sample; for a matrix of
tubes with 1% HNO3 (v/v). After addition of 3-Ag/l Ir as unknown composition, the method of addition calibration,
internal standard, the samples were analysed by ICP-MS. in which known concentration of standards are added
The sample uptake rate was of 1.0 ml/min. directly to a sample solution, allows to compensate for the
difference in sensitivity between samples and standards.
2.5. ICP-MS analysis Spectral interference may occur from several sources:
isobaric overlaps, plasma-induced polyatomic ions, matrix
The Perkin-Elmer SCIEX ELAN 5000 inductively solvent-induced polyatomic, ions and matrix-induced pol-
coupled plasma-mass spectrometer (Perkin-Elmer, Norwalk, yatomic ions [10]. Considering the natural abundance and
CT, USA) was designed for routine and rapid multielement potential interferences of Pt isotopes reported in Table 2 and
quantitative determinations of trace and ultratrace elements preliminary studies not reported in the present paper, the
195
and isotopes [6]. An IBM PS/2 Model 70 386 with an Pt was identified as the isotope with the lowest number of
operating system Xenix 386 (version 2.3.4) and an Intel interferences.
80386 microprocessor controls the ELAN. Under the described experimental conditions, no signifi-
The instrument is equipped with Cool Flow CFT-75 cant interference between Pt and human urine and plasma
Neslab and autosampler PE AS90. and ultrafiltrate matrix was observed in the drug-free
To optimise the ICP-MS signal, a standard solution (PE samples.
Number N812-2014) containing 10 Ag/l of three elements ICP-MS tracing of real sample obtained from a subject
that covered the entire mass range (e.g., 24Mg, 103Rh, and treated with an anticancer agent is reported in Fig. 1.
208
Pb) was used. A fourth isotope, 197Au, was used to
monitor background noise in the system. Typically, the 3.2. Calibration curves
intensity values obtained for 10 Ag/l were the following:
24
Mgz5,000,; 103Rhz30,000, 208Pbz5,000, and 197AuV20 The standard addition method was used for plasma,
ions/s. ultrafiltrate, and urine matrices.
The instrumental specifications and the analytical con- The calibration curves were established by plotting the
ditions have been reported in Table 1. current intensity (ions/s) ratio between 195Pt and the internal

Fig. 1. Tracing of real human urine sample obtained from a subject treated with Pt anticancer agent.
360 M. Bettinelli / Microchemical Journal 79 (2005) 357–365

Table 3 calibration curve). The same samples analysed for the intra-
Accuracy and precision of the determination of platinum in human plasma and interday precision were used to calculate the accuracy
samples evaluated in three different days
of the method. Accuracy was calculated as the percent ratio
Nominal concentration (Ag/l) QC1 QC2 QC3
of the concentration found to the nominal concentration
(2 Ag/l) (41 Ag/l) (152 Ag/l)
value.
1st day
The results of this study are reported in Tables 3–5,
Found 1st 2.043 40.917 152.632
concentrations measurement respectively for QC1, QC2, and QC3 samples and sum-
(Ag/l) 2nd 2.072 41.462 152.292 marized in Tables 6–8 for a concentration of analyte close to
measurement estimated limits of quantification (LOQ).
3rd 2.072 41.037 151.554
measurement
3.4. Limit of detection (LOD) and limit of quantification
Intraday mean (Ag/l) 2.063 41.138 152.159
Intraday S.D. (Ag/l) 0.018 0.286 0.551 (LOQ)
Intraday CV (%) 0.87 0.70 0.36
Intraday accuracy (%) 103.2 100.3 100.1 For ultrafiltrate plasma, the instrumental detection limit
(LOD) and the limit of quantification (LOQ) were
2nd day
Found 1st 2.042 40.825 152.175
concentrations measurement Table 4
(Ag/l) 2nd 2.024 40.879 151.808 Accuracy and precision of the determination of platinum in human plasma
measurement ultrafiltrate samples evaluated in three different days
3rd 1.965 40.940 152.442
measurement Nominal concentration (Ag/l) QC1 QC2 QC3
Intraday mean (Ag/l) 2.010 40.881 152.142 (0.02 Ag/l) (6 Ag/l) (18 Ag/l)
Intraday S.D. (Ag/l) 0.040 0.057 0.318 1st day
Intraday CV (%) 2.01 0.14 0.21 Found 1st 0.0196 6.0680 17.7872
Intraday accuracy (%) 100.5 99.7 100.1 concentrations measurement
(Ag/l) 2nd 0.0188 6.1388 17.7792
3rd day measurement
Found 1st 2.006 40.390 152.508 3rd 0.0201 5.8652 17.7765
concentrations measurement measurement
(Ag/l) 2nd 1.958 41.736 151.724 Intraday mean (Ag/l) 0.0195 6.0240 17.7810
measurement Intraday S.D. (Ag/l) 0.0007 0.1420 0.0056
3rd 2.017 41.399 151.571 Intraday CV (%) 3.36 2.36 0.03
measurement Intraday accuracy (%) 97.50 100.40 98.78
Intraday mean (Ag/l) 1.994 41.175 151.935
Intraday S.D. (Ag/l) 0.031 0.700 0.503 2nd day
Intraday CV (%) 1.57 1.70 0.33 Found 1st 0.0203 6.0564 18.0200
Intraday accuracy (%) 99.7 100.4 99.9 Concentrations measurement
(Ag/l) 2nd 0.0223 5.9348 18.1639
Interday accuracy and precision measurement
Mean (Ag/l) 2.023 41.065 152.079 3rd 0.0222 5.9344 17.8857
S.D. (Ag/l) 0.041 0.404 0.420 measurement
CV (%) 2.05 0.98 0.28 Intraday mean (Ag/l) 0.0216 5.9752 18.0232
Accuracy (%) 101.1 100.2 100.1 Intraday S.D. (Ag/l) 0.0011 0.0703 0.1391
Intraday CV (%) 5.22 1.18 0.77
standard (193Ir) against the concentration of Pt. Slopes and Intraday accuracy (%) 108.00 99.59 100.13
intercepts were determined by linear regression analysis. 3rd day
Found 1st 0.0210 6.0493 17.9209
3.3. Precision and accuracy concentrations measurement
(Ag/l) 2nd 0.0204 5.9637 18.3330
measurement
Intraday precision was assessed by replicate determina-
3rd 0.0209 5.9306 17.9139
tions (n=3) of Pt concentration added as PtCl2 at three measurement
concentrations to the blank human plasma (2, 41, and 152 Intraday mean (Ag/l) 0.0208 5.9812 18.0559
Ag/l), human plasma ultrafiltrate (0.02, 6, and 18 Ag/l), and Intraday S.D. (Ag/l) 0.0003 0.0613 0.2400
to the urine (5, 40, and 160 Ag/l). Intraday CV (%) 1.55 1.02 1.33
The same determinations were repeated in three different Intraday accuracy (%) 103.83 99.69 100.31
working days in order to evaluate the interday precision and Interday accuracy and precision
accuracy. The precision of the method was calculated as the Mean (Ag/l) 0.0206 5.9935 17.9534
coefficient of variation of the mean value found at each S.D. (Ag/l) 0.0011 0.0880 0.1902
concentration level (one near the lower limit of quantitation, CV (%) 5.51 1.47 1.06
one near the middle, and one near the upper boundary of the Accuracy (%) 103.11 99.89 99.74
 M. Bettinelli / Microchemical Journal 79 (2005) 357–365 361

Table 5 case of plasma and urine analysis, considering that the

Accuracy and precision of the determination of platinum in urine samples concentrations of interest were one order of magnitude
evaluated in three different days
higher than those in ultrafiltrate, the LOQ was set,
Nominal concentration (Ag/l) QC1 QC2 QC3
arbitrarily, at 1 and 2 Ag/l, respectively.
(5 Ag/l) (40 Ag/l) (160 Ag/l)
For the LOQ, we also verified the criterion reported by
1st day
FDA [6] bas the lowest concentration that can be
Found 1st 5.048 40.161 160.116
concentrations measurement determined with a satisfactory degree of accuracy (devia-
(Ag/l) 2nd 5.028 39.971 159.707 tion from nominal value V20%) and precision [coefficient
measurement of variation (CV)V20%] to provide pharmacokinetically
3rd 5.036 40.069 160.543 useful results.Q
measurement
Intraday mean (Ag/l) 5.037 40.068 160.122
Intraday S.D. (Ag/l) 0.010 0.093 0.418 3.5. Plasma and ultrafiltrate
Intraday CV (%) 0.19 0.23 0.26
Intraday accuracy 100.7 100.2 100.1 Matrix-matched calibration was achieved by adding
(mean found/ dilute Pt standard solution to pooled plasma or ultrafiltrate
added %)
samples and kept at 25 8C until use.
2nd day For plasma, the mean slope and intercept values of the
Found 1st 5.006 40.049 159.13 calibration curves (0–200 Ag/l) obtained over the three-day
concentrations measurement validation period were 0.199F0.008 (CV=3.86%) and
(Ag/l) 2nd 5.014 40.007 159.77 0.055F0.043, with a correlation coefficient higher than
measurement
0.9996.
3rd 5.009 40.099 160.99
measurement For ultrafiltrate plasma, the statistical parameters for the
Intraday mean (Ag/l) 5.009 40.052 159.96 calibration curves (0–20 Ag/l) obtained over the three-day
Intraday S.D. (Ag/l) 0.004 0.046 0.94 validation period were 0.376F0.042 (CV=11.1%) and
Intraday CV (%) 0.08 0.11 0.59 0.005F0.006, with a correlation coefficient higher than
Intraday accuracy 100.2 100.1 99.9
0.9989.
(mean found/added %)
Intraday and interday accuracies and precisions of the
3rd day method for the determination of platinum in human plasma
Found 1st 5.069 39.93 159.69 and ultrafiltrate samples have been reported in Tables 3 and
concentrations measurement 4, respectively.
(Ag/l) 2nd 5.070 39.92 160.31
Intraday precision and accuracy for Pt determination in
measurement
3rd 4.925 40.07 159.81 human plasma samples ranged, respectively, between 0.14
measurement and 2.01 and between 99.7 and 103.2 (CV %), whereas, in
Intraday mean (Ag/l) 5.021 39.98 159.94 the ultrafiltrate, they ranged between 0.03 and 5.22 and
Intraday S.D. (Ag/l) 0.083 0.082 0.33 between 97.5 and 108.0 (CV %). Slight better ranges of
Intraday CV (%) 1.66 0.21 0.21
interday precision and accuracy were achieved for plasma
Intraday accuracy 100.4 99.9 100.0
(mean found/added %) with respect to ultrafiltrate probably due to higher Pt
concentrations investigated.
Interday accuracy and precision The instrumental detection limit (LOD) was 0.002 Ag/l,
Mean (Ag/l) 5.0228 40.0325 160.0077 with a minimum quantifiable platinum concentration (LOQ)
S.D. (Ag/l) 0.0436 0.0784 0.5479
for ultrafiltrate equal to 0.01 Ag/l. In the case of plasma
Precision (CV %) 0.87 0.20 0.34
Accuracy (mean found/added %) 100.46 100.08 100.00 analysis, the concentrations of interest were one order of
magnitude higher than those in ultrafiltrate samples, there-
fore, the LOQ was set at 1 Ag/l.
evaluated on the basis of the 3j and 10j criterion, The mean slope value of the calibration curves obtained
respectively, by analyzing five bblanksQ of plasma, ultra- over the three-day validation period was 0.1277F
filtrate, and urine on three nonconsecutive days. In the 0.0038(CV=2.94%). The correlation coefficients were

Table 6
Accuracy and precision data for plasma ultrafiltrate at LOQ concentration
Nominal concentration 0.01 (Ag/l) Intraday Interday
Mean (Ag/l)FS.D. CV (%) Accuracy (%) Mean (Ag/l)FS.D. CV (%) Accuracy (%)
1st day 0.0098F0.0008 8.10 98.0
2nd day 0.0108F0.0007 6.52 107.7 0.011F0.001 7.49 104.6
3rd day 0.0108F0.0006 5.56 108.0
362 M. Bettinelli / Microchemical Journal 79 (2005) 357–365

Table 7
Accuracy and precision data for plasma at LOQ concentration
Nominal concentration 1.0 (Ag/l) Intraday Interday
Mean (Ag/l)FS.D. CV (%) Accuracy (%) Mean (Ag/l)FS.D. CV (%) Accuracy (%)
1st day 1.032 (0.005) 0.46 103.2
2nd day 1.032 (0.028) 2.69 103.2 1.034 (0.017) 1.61 104.6
3rd day 1.037F0.017 1.64 103.7

0.999 in the entire period of the validation assessment.
Accuracy and precision for plasma and ultrafiltrate samples
at the LOQ concentration are reported in Tables 6 and 7.
3.6. Urine
Matrix-matched calibration was achieved by adding
dilute Pt standard solution to pooled urine samples, which
were previously centrifuged to remove the sediment and
kept at 25 8C until analysed.
The mean slope and intercept values of the calibration
curves (0–200 Ag/l) obtained over the three-day validation
period were 0.128F0.04 (CV=2.97%) and 0.033F0.041,
with a correlation coefficient higher than 0.9999.
The instrumental detection limit (LOD) was 0.002 Ag/l,
with a minimum quantifiable platinum concentration (LOQ)
set at 2 Ag/l.
Intraday and interday accuracies and precisions of the
method for the determination of platinum in human urine
samples have been reported in Table 5.
Intraday precision, expressed as CV %, ranged between
0.08 and 0.59 (with only one value higher than 1%), and
intraday accuracy, calculated as the percentage of the found/
added amounts, ranged on average between 99.9% and
100.7%. Precision lower than 1% and accuracy very close to
100% were achieved during the interday evaluation.
Accuracy and precision for urine samples at the LOQ
concentration are reported in Table 8.
The results obtained in the present study meet the FDA
criteria for the validation of analytical methods [6]. For each
concentration level, the mean value is within F15% of the
nominal value (at the LOQ, within F20%), and the
coefficient of variation (CV) around the mean value is
lower than F15% (at the LOQ, F20% for the CV is
accepted).
These results prove that, in keeping with international
standards, the method is adequate for the assay of platinum
in human plasma and urine samples.
3.7. Uncertainty evaluation from validation data
The validation study of the present assay was imple-
mented by assessing the uncertainty evaluation for Pt in
the different matrices at the four levels of concentration.
According to EURACHEM/CITAC Guide [8], the com-
bined uncertainty (u combined) has been calculated estimat-
ing the components associated with the repeatability
(u repeatability), the instrumental calibration (u calibration), and
the sampling (u sampling). The uncertainty component due to
the recovery (u recovery) was not considered because the
calibration curve was performed by standard addition
method (in the same matrix as the sample) determining the
analytes according to the procedure described in the
experimental section.
Consequently, the equation that best represents the
mathematical model upon which the relative uncertainty is
evaluated is taken to be
uðComb:Þ
ðComb:Þ
s ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
    
uðRepeat:Þ 2 uðCalib:Þ 2 uðSamp:Þ 2
¼ þ þ
ðRepeat:Þ ðCalib:Þ ðSamp:Þ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2  2  2
u̇u ðComb:Þ ¼ u̇u ðRepeat:Þ þ u̇u ðCalib:Þ þ u̇u ðSamp:Þ
where: u̇ (Comb.)=combined relative uncertainty, u̇ (Repeat.)=re-
lative uncertainty of repeatability, u̇ (Calib.) = relative uncer-
tainty of calibration, u̇ (Samp.) = relative uncertainty of
sampling
In order to investigate the contributions of all the
components to the combined relative uncertainty as a
function of Pt concentration (in plasma, ultrafiltrate, and
urine), these uncertainties were evaluated and plotted in
Figs. 2, 3 and 4, respectively.

Table 8
Accuracy and precision data for urine at LOQ concentration
Nominal concentration 2.0 (Ag/l) Intraday Interday
Mean (Ag/l)FS.D. CV (%) Accuracy (%) Mean (Ag/l)FS.D. CV (%) Accuracy (%)
1st day 2.091F0.012 0.58 104.6
2nd day 2.068F0.017 0.81 103.4 2.093F0.025 1.20 104.6
3rd day 2.119F0.012 0.57 105.9
 M. Bettinelli / Microchemical Journal 79 (2005) 357–365 363

Fig. 2. Relative uncertainty values (%) obtained at different concentration levels for Pt in plasma.

With respect to Pt in plasma and urine the graphs show
the relative uncertainty ranges from 4 to 7% at 1–2 Ag/L
and go down to 2–3% for the higher concentrations (40–
150 Ag/L).
While the uncertainty component associated to the
repeatability is constant (0.5–2%) overall the different
concentration levels and the different days, the component
due to the instrumental calibration increases significantly (7–
8%) in correspondence of the lowest concentrations of
analyte.
With respect to Pt in plasma ultrafiltrate, the graph
reported in Fig. 4 shows that the relative uncertainty
ranges from 4 to 5% at 8–16 Ag/l and increases up
to 20 to 30% for the higher concentrations (at 0.01–
0.02 Ag/l).
This occurrence means that the quantitative determi-
nation of low concentrations of Pt in ultrafiltrate is
greatly affected by the uncertainty assignable to the
instrumental calibration as done in the present study (0–
20 Ag/l). Further tests performed reducing the calibration
range but keeping the same number of concentration
levels (n=6), given better values of combined uncertainty
as consequence of a significant improvement of the
uncertainty component due to calibration. The mathemat-

Fig. 3. Relative uncertainty values (%) obtained at different concentration levels for Pt in plasma ultrafiltrate.
364 M. Bettinelli / Microchemical Journal 79 (2005) 357–365
Fig. 4. Relative uncertainty values (%) obtained at different concentration levels for Pt in urine.
ical function [9] used for the calculation of uncertainty
of (x pred) due to calibration is the following:
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u  2
  u1 s2y=q 1 xpred  x̄Cal
u xpred ¼ t þ þ P
b m n ðxi  x̄Cal Þ2
where: S 2y/q is the residual variance of the regression
model, b is the slope of the regression curve, m is the
number of repetitions from which the predicted value is
derived, n is the number of calibration points on the
regression line, x pred is the predicted value, and x̄ cal is the
mean value of the concentration levels used in the
calibration curve.
Inspection of the previously reported equation confirms
that, as x pred approaches x̄ cal, the third term under the square
root approaches zero, and the u(x pred) thus approaches a
minimum value. In a practical analysis, therefore, a
calibration of this type will give the most precise results
when the measured instrument signal corresponds to a point
close to the centroid of the regression line (x̄ cal, ȳ cal). In our
study, with a calibration range of 0–20 Ag/l and a centroid
corresponding to (5.18; 1.84), the u̇ Calib. for a predicted
mean value of 0.36 Ag l1 Pt in ultrafiltrate (n=3 replicates)
was about 21.9%; the same results became lower (0.22%)
when the calibration range was set at 0–2 Ag/l with the same
number of calibration point and with a centroid correspond-
ing to (0.527; 0.163).
These results show the importance of calibration in
evaluating the measurement uncertainty of low Pt concen-
trations in biological fluids. Keeping constant the repeat-
ability of measurements, in the case of wider calibration
range, we should accept larger uncertainty of calibration,
which will reflect significantly on the combined uncertainty.
If we wished to improve (i.e., narrow) the confidence
limits of calibration, we could reduce the calibration range,
increase the number of calibration points on the regression
line or make more that one measurement, and use the mean
value in the calculation of the predicted value.
4. Conclusions
An ICP-MS method for the determination of Pt in
biological fluids (plasma, ultrafiltrate, and urine) of patients
treated with antitumor agents has been developed and
validated. The limits of quantification (LOQ) in the three
matrices were 1.0, 0.1, and 2.0 Ag/l, respectively.
Intraday and interday precisions and accuracies were in
good agreement with the FDA criteria for the validation of
analytical methods. The present study was implemented by
assessing the uncertainty evaluation for Pt determination in
the different matrices according to EURACHEM/CITAC
Guide. The combined uncertainty, calculated estimating the
components associated with the repeatability, the instru-
mental calibration, and the sampling resulted always lower
than 10% for Pt concentrations higher than 1 Ag/l.
While the uncertainty component associated to the
repeatability was constant overall the different concentration
levels, and the sampling uncertainty was negligible, the
component due to the instrumental calibration increased
significantly in correspondence of the lowest concentrations
of analyte.
References
[1] J.A. Gottlieb, B. Drewinko, Cancer Treat. Rep. 59 (1975) 621.
 M. Bettinelli / Microchemical Journal 79 (2005) 357–365
[2] A.B. Khan, B.J. D’Souza, M.D. Wharam, L.A.A. Champion, L.F.
Sinks, S.Y. Woo, D.C. McCullough, B.C. Leventhal, Cancer Treat.
Rep. 66 (1982) 2013.
[3] C. Sessa, C. Minoia, A. Ronchi, M. Zucchetti, J. Bauer, M. Corner, J.
de Joung, O. Pagani, J. Renard, C. Weil, M. D’incalci, Ann. Oncol. 9
(1998) 1315.
[4] U.S. Pharmacopeia 23, (1994) :1982–84, United States Pharmacopeial
Convention.
[5] International Conference on HarmonizationDraft Guidance on Vali-
dation of Analytical Procedures. Definitions and Terminology—
Federal Register, vol. 60, 1995 (March 1), pp. 1260.
365
[6] FDA, Guidance for Industries. Bioanalytical Method Validation, U.S.
Department of Health and Human Services Food and Drug Admin-
istration, 2001 (May).
[7] ISO, Guide to the Expression of Uncertainty in Measurement,
International Standards Organization, Geneva, 1993.
[8] EURACHEM/CITAC Guide, Quantifying Uncertainty in Analytical
Measurement, 2nd ed., 2000.
[9] S.J. Bannister, Y. Chang, L.A. Sterson, A.J. Repta, Clin. Chem. 24
(1978) 877.
[10] E.H. Evans, J.J. Giglio, J. Anal. At. Spectrom. 8 (1993) 1.