Comparison of inductively coupled plasma mass spectrometry techniques in the determination of platinum in urine: quadrupole vs. sector field



In recent years the increasing use of platinum (Pt) both in medical and in industrial applications has caused its growing anthropogenic emission and spread in the environment. Pt is released into the atmosphere by exhaust catalytic converters, and Pt compounds are often used in antitumour therapies. As a consequence, significant amounts of Pt can be detected in hospital wastewaters. This can lead to an increase in the exposure levels to Pt, especially in urban areas. It is therefore necessary to determine Pt reference values in the general population, by using suitable procedures able to achieve adequate analytical performances. Several measurements of Pt in biological fluids have been reported, but the analytical methods used for the determination of Pt often lack information about the uncertainty of the results, especially for low concentrations of urinary Pt in non-occupationally exposed subjects. The present paper considers the measurement of urinary Pt levels in a general population group from central Italy, by both quadrupole (Q) and sector field (SF) inductively coupled plasma mass spectrometry (ICP-MS). The two procedures were validated and their expanded uncertainties were evaluated. The limits of detection (LODs), calculated taking into account dilution factors, were 0.18 and 0.05 ng L−1 of Pt for the Q and SF procedures, respectively. The median value observed was 4.13 ng L−1 of Pt in urine, while the relative combined uncertainty at 5 ng L−1 was below 20% with both ICP-MS techniques. These data are in good agreement with those reported in the literature for similar studies. Copyright © 2005 John Wiley & Sons, Ltd.

The use and diffusion of platinum in the environment have increased rapidly in recent years, both in the medical and in the industrial field. Platinum is widely used in the antineoplastic chemotherapy of several tumours, such as lung, cervical, testicular, head, neck ovarian and bladder carcinomas.1, 2 The resulting excretion of Pt compounds by treated patients leads to not negligible concentrations of Pt in wastewaters, and then in the environment.3 In addition, modern three-way automotive catalysts, which abate the emission of NOx, CO and aromatic hydrocarbons, typically contain high concentrations of Pt group elements (PGEs) and, thus, can be considered as the main sources of Pt dispersion.4–6 Many studies have reported high Pt concentration in airborne particulate,7, 8 in sewage sludges,9 and in road dust,6 thus raising concern about potential harmful consequences on humans.

The most efficient way to monitor and evaluate the real incidence of this exposure is to determine the reference value (RV) of Pt in the urine of the general population.10 The assessment of the RV of Pt in biological fluids is a difficult task for analytical laboratories, due to the low levels at which the metal is present, requiring adequate detection limits, careful and minimal manipulation of the samples, and effective control of pre-analytical factors in order to assure low levels of blanks.11

The determination of Pt in biological fluids is performed by means of different inductively coupled plasma mass spectrometry (ICP-MS) techniques, i.e., quadrupole (Q), sector field (SF) and instruments with dynamic cell reaction (Q-DRC), and related hyphenated techniques,12–15 although some alternative procedures like voltammetry,16 ETAAS with on-line pre-concentration,17 and ICP-OES18 have also been investigated.

There is considerable experimental data on the Pt concentrations in biological fluids available in the literature, but very few studies have involved a robust evaluation of the analytical performance, in terms of assessment of the measurement uncertainty. It is now, however, generally acknowledged that the fitness of analytical data cannot be assessed without some estimation of the measurement uncertainty to be compared with the required confidence levels. It is thus necessary to identify all the possible sources of uncertainty arising from the adopted procedure, to define their magnitude, and to combine their estimations to give standard and expanded uncertainties.

This is particularly important when trace or ultra-trace elements have to be assessed as RVs. At low level it is important to establish if a procedure is able to give satisfactory results, i.e., which is the lower difference between two samples that the procedure is able to distinguish. From this point of view, overall uncertainties higher than 50% seem to be unacceptable.

In a previous study19 the assessment of the measurement uncertainty was applied to validate the method of Pt quantification in fluids of patients treated with antitumour agents. The same approach has been adopted here for comparing the performances of two analytical procedures involving different ICP-MS techniques, namely the Q and SF systems, in the determination of Pt in the urine of selected healthy subjects. The tentative RVs for urinary Pt of a general population group from a central region of Italy were also calculated.



To quantify Pt in urine two ICP-MS approaches were adopted. The first used a quadrupole system (Q-ICP-MS), an ELAN DRC II instrument (Perkin Elmer SCIEX Instruments, Concorde, Ontario, Canada), equipped with a cyclonic spray chamber and a Meinhard type concentric nebulizer. The second used a sector field (SF-ICP-MS), an ELEMENT I spectrometer (Thermo-Finnigan, Bremen, Germany), equipped with a guard electrode device and operating at a mass resolution of mm = 300. For both instruments the radio-frequency power and the gas flows were optimized so as to obtain good sensitivity and, at the same time, low production of oxides (<2%) and doubly charged ions (<1%). The instrumental specifications and analytical conditions are reported in Table 1.

Table 1. Equipment and settings for the determination of Pt in urine
SpectrometerELAN DRC II, Sciex Perkin ElmerELEMENT I, Thermo-Finnigan
Resolutionmm = 300mm = 300
NebulizerMeinhard glass type with cyclonic spray chamberMeinhard glass type with water-cooled spray chamber, Scott type
InterfaceSampler and skimmer cones in NiSampler and skimmer cones in Ni
Data acquisitionPeak hopping; 5 sweeps; 3 readings; 5 replicatesElectric scan; 5 runs; 4 passes
RF power (kW)1.251.20
Argon flows (L min−1)Plasma, 15; auxiliary, 1.3; sample, 1.0Plasma, 14; auxiliary, 0.9; sample, 1.0
Analytical mass (amu)195Pt195Pt

Reagents and standard solutions

Standard solutions of Pt for calibration curves and of iridium (Ir) as internal standard (IS) were supplied by CPI (Amsterdam, The Netherlands) and SPEX (Edison, NJ, USA). Ultra-pure 65% HNO3 was provided by Carlo Erba (Milan, Italy), and the high-purity water used throughout was prepared by a Milli-Ro, Milli-Q system (Millipore, Bedford, MA, USA).

Population study

Fifty subjects aged between 20 and 68 years, living in Novafeltria (Pesaro, Italy), were randomly selected from the county register. Twenty-five males and 25 females, with mean ages of 48.2 and 36.8 years, respectively, were included in the study. Subjects were interviewed to obtain detailed information on family, dietary habits, lifestyle and potential exposures. Urine samples were collected in polypropylene tubes, which had previously been soaked in 1% HNO3, and then stored at −20°C until analysis.

Sample preparation and calibration

To maintain the integrity of the urine samples and to minimize contamination or loss of samples, manipulation of the specimens was performed in a Class-100 clean room (Tamco, Rome, Italy). After being thawed at room temperature and then shaken, 1 mL aliquots were sub-sampled from the original urine samples and diluted with a 1% HNO3 solution (3 or 9 mL of diluting solution for final analysis for the Q or SF instruments, respectively). In order to control possible matrix-induced variations, which could heavily affect the robustness of the methods, the standards addition approach for calibration was adopted and an IS was used. Initial experiments showed a very low repeatability (relative standard deviation (RSD) = 4.27%) of 193Ir signal added as IS (concentration: 1.0 μg L−1) when the same urine sample was determined (no. of replicates = 30), but a high variation in the Ir response (RSD = 18.34%) in the case of analysis of different samples (no. of samples = 30; Fig. 1). This evidence—obtained by Q-ICP-MS and confirmed by SF-ICP-MS—corroborated the very good instrumental performance (low RSD for the same urine analyzed 30 times) but, in contrast, showed the significant influence of matrix composition on the analytical signals. Platinum, as expected, displayed a behaviour similar to that of Ir (Fig. 1(b)), so the Pt/Ir ratio warranted adequate independence from matrix effects. Data coming from the SF experiments (on more diluted samples) showed a less pronounced, but still significant, influence of the matrix composition (RSD = 9.5%). On the basis of these preliminary tests the use of Ir as IS at a concentration of 1 or 0.5 μg L−1 for Q or SF determinations became mandatory. The calibration was performed according to the standards addition method (six spike levels at 1, 2, 6, 8, 10 and 20 ng L−1). Calibrants were prepared daily by diluting single element stock solutions of 1000 mg L−1 Ir and Pt with high-purity water.

Figure 1.

Signal intensity (counts s−1) in the Q-ICP-MS analysis: (a) intensity for 15 replicates of the same urine sample (Ir 1 μg L−1) and (b) intensity for 15 different urine samples (Ir 1 μg L−1) spiked with 0.2 μg L−1 of Pt.

Study of the measurements uncertainty

The uncertainty of measurements was evaluated as suggested by international bodies such as ISO20 and EURACHEM/CITAC.21 For both techniques the samples were directly analyzed after simple dilution; therefore, the measurement uncertainties can be calculated by considering the contributions of only three factors: (i) the uncertainty related to the calibration curve; (ii) that linked to the repeatability; and (iii) that associated with the dilution step.

(i) Calibration uncertainty

In the equation of the calibration curve, Eqn. (1), xi is the concentration in ng L−1 of the single i-point of the curve and equation image the corresponding intensity.

equation image(1)

The residuals were obtained by difference of the theoretical signal intensity (equation image, from the calibration curve) and the real one (equation image, obtained from the instrumental analysis); the related standard deviation of residuals (sy/x) was also calculated. Real samples were then analyzed with spikes of known amounts of analyte, at six concentration levels (qi) (Q = 1, 2, 6, 8, 10 and 20 ng L−1). Then, for each replicate, the variances of the calibration curve (equation image, see Eqn. (2), Table 2), the mean of variances (equation image, Eqn. (3), Table 2) and the mean of replicates (equation image, Eqn. (4), Table 2) were calculated and used to determine the standard calibration uncertainty (equation image) and the relative standard calibration uncertainty (equation image) for every level of concentration (Q) (Eqns. (5) and (6) in Table 2).

Table 2. Equations used to estimate the calibration uncertainty
  1. n = no. of points of the calibration curve; m = no. of replicates for each i-point; equation image = mean of intensities of all the calibration points; yq =  signal intensity of the single replicate of the Q point of the repeatability test; xi = concentration (ng L−1) of the i-point of the calibration curve; equation image = mean of the six values of xi used for the calibration curve.

equation image(2)
equation image(3,4)
equation image(5,6)

(ii) Repeatability uncertainty

The standard deviation of the repeatability data mentioned above represented the repeatability standard uncertainty, i.e., [s(Q) = urep(Q)], which was used to obtain the repeatability standard uncertainty of the mean (Eqn. (7)).

equation image(7)

Finally, the repeatability standard uncertainty of the mean was divided by the mean of replicates to achieve the corresponding relative uncertainty (Eqn. (8)).

equation image(8)

(iii) Dilution step uncertainty

The dilution uncertainty equation image depends on the variability connected to the devices—volumetric flasks, tubes and pipettes—used for the dilution of the samples. In this case the uncertainty declared by the manufacturer was used, assuming that it had a rectangular distribution. As a consequence, the corresponding standard uncertainty was simply calculated by dividing the manufacturer's uncertainty by a factor of equation image. It is easy to verify that the effect of this component is in general negligible compared with the effect of the others.

Combining the data of the partial uncertainties previously described, the relative combined standard uncertainty was calculated according to the following formula:

equation image(9)

The expanded uncertainty (U) was obtained by multiplying the combined standard uncertainty by a coverage factor, k, typically in the range 2–3, gathered from the Student's t-values at the desired confidence level (95%) and defined degrees of freedom. So, the relative expanded uncertainty was:

equation image(10)


General performances of the methods

The linearity of calibration curves based on 1, 2, 6, 8, 10 and 20 ng L−1 of analyte resulted in correlation coefficients equal to 0.999 for Q-ICP-MS and 0.997 for SF-ICP-MS. Accuracy, expressed as precision and trueness, was evaluated on spiked samples at three different concentrations, i.e., 1.5, 5.0 and 15.0 ng L−1, as well as through participation in an interlaboratory study (34th RRT G-EQAS) organized by the University of Erlangen (Germany). The findings of the in-house recovery study, summarized in Table 3, were very satisfactory for both techniques. Also, the interlaboratory comparison gave good recoveries, by Q-ICP-MS, on two spike levels, i.e., 73.0 ± 4.90 ng L−1 vs. the reference value of 88.7 (tolerance range of 49.4–128.0 ng L−1) and 165.0 ± 1.40 ng L−1 vs. the reference value of 179.2 (124.3–234.2 ng L−1).

Table 3. Accuracy of ICP-MS analyses (no. of replicates = 10)
Spikes (ng L−1)Precision (%)Trueness (%)Precision (%)Trueness (%)

Moreover, the variability of the determinations was assessed in conditions of intermediate repeatability, i.e., two series of 20 real samples were analyzed on two different days, with different calibrations and by different operators. From these two sets of data the resulting repeatability was equal to 5.4% and 4.0% in Q- and SF-ICP-MS experiments, respectively. Method limits of detection (LODs), in their turn, were estimated by multiplying by 3.3 the standard deviation obtained from the analyses of 10 replicates of a sample of urine with low natural concentration of Pt. The LODs, calculated taking into account the dilution factors, were 0.18 and 0.05 ng L−1 for the Q and SF procedures, respectively. On the basis of the resultant performance, the two techniques had a similar degree of accuracy. Conversely, the SF-ICP-MS method was noticeably better than the other technique in terms of LOD, precision and intermediate repeatability, thus indicating intrinsic improved stability and sensitivity. On the other hand, the SF instrumentation has as its main drawbacks the higher purchase and maintenance costs and the need of operators with considerable expertise.

Estimate of the uncertainties

As pointed out by Thompson,22 the simple assessment of the LODs for an analytical procedure does not guarantee that all measurements above those limits are satisfactory. It is important, in fact, to critically discuss the produced data on the basis of their overall uncertainty, in order to evaluate the interval of values reasonably attributable to the parameter to be measured. In this context, the evaluation of the different uncertainties relating the two ICP-MS procedures under comparison gave a valuable picture of their performances. Table 4 and Fig. 2 summarize the calibration, repeatability, dilution, combined and expanded relative standard uncertainties as obtained following the above schemes. The worse coefficient correlation of the SF-ICP-MS was responsible for its worse uncertainty data even if the difference was not particularly significant. In general, the uncertainty findings depicted no significant differences between the two procedures under investigation, with a clear improvement of the expanded uncertainty at concentration higher than 1.5 ng L−1. Among the factors affecting the uncertainty it should be stressed that the contribution of the calibration is the more important component at lower levels (confirmed by the fact that the worse SF calibration data gave worse expanded uncertainty evaluation, even in the presence of a better repeatability) and this diminished, as expected, at higher concentration levels, up to one order of magnitude, for both techniques.

Table 4. Relative standard uncertainties of the ICP-MS techniques
Spikes (ng L−1)1.55.0151.55.015
equation image0.0860.0220.0080.1290.0360.013
equation image0.0380.0290.0220.0090.0110.008
equation image0.0060.0060.0060.0060.0060.006
equation image0.0940.0370.0250.1300.0380.016
equation image0.2310.0850.0630.3340.0980.036
Figure 2.

Combined relative uncertainties of Q-ICP-MS and SF-ICP-MS and their related components.

Analysis of real samples

The real samples were analyzed using the conditions discussed above. The data obtained by the two techniques were compared according to the maximum likelihood functional relationship (MLFR) calculation. Following this approach, results were plotted assuming that, beyond the uncertainty of the values of the Y, the variance of the errors of the X was also not negligible: the comparison was, therefore, more rigorous than in the case of the normal linear regression approach. The resulting curve was then evaluated in order to verify if the intercept and slope were statistically different from the theoretical values of 0 and 1, respectively. Comparing the Q- and SF-ICP-MS findings, this hypothesis was verified, again confirming the good agreement between the data from the two procedures, as can be seen by the Fig. 3, in which the Q and SF results were plotted together and the MLFR line was drawn. The equation of the comparison curve was:y = [0.86 ± 0.19]x + [−0.48 ± 1.03].

Figure 3.

Comparison of real samples analyses by means of Q-ICP-MS and SF-ICP-MS. (MLFR regression curve).

Platinum urine concentrations relating to the population group under study are displayed in Table 5, in terms of basic statistics averaged from data obtained by both techniques. The size of the population sample was statistically sufficient to consider these findings as tentative RVs. Moreover, the RVs showed good agreement with other literature results for similar studies. As examples, Wilhelm et al.10 recently obtained a RV in urine of 10 ng L−1; Schramel and Wendler23 gave an urinary range of 0.9–6.6 ng L−1 for adults without gold fillings in their teeth, and values up to 151 ng L−1 for subjects with gold fillings; Herr et al.16 showed similar results, i.e., Pt urine levels higher than 4.5 ng L−1 in subjects without assessable medical or dental devices; Iavicoli et al.6 reported a range of 0.55–10.1 ng L−1 in policemen involved in urban traffic control; and Bocca et al.24 found significantly different (p < 0.001) median concentrations for two urban groups living in Rome (1.70 ng L−1) and in Foligno (0.52 ng L−1), a small town with low traffic density.

Table 5. Reference values of Pt in urine (no. of subjects = 50)
 Pt (ng L−1)
TotalMale (n = 25)Female (n = 25)
Geometric mean4.094.513.70
10th Percentile1.732.441.61
90th Percentile9.258.429.57

Finally, if the real data are evaluated on the basis of the uncertainty calculations, it can be highlighted that, at values close to the median RV found in the population group, a relative expanded uncertainty below 20% is obtained.

From these data it is demonstrated that the described procedures are suitable for a correct evaluation of the RV of urinary Pt in the general population. The investigation will therefore continue by enlarging the statistical basis giving particular attention to the possible correlations between urinary Pt levels and subjects' lifestyle and dietary habits, and the presence of dental devices.


The analytical performances of two ICP-MS procedures for the quantification of Pt in urine showed a good agreement, in terms of accuracy and precision, between the two sets of data, and confirmed that ICP-MS techniques are suitable for monitoring the general population. Moreover, both techniques gave adequate LODs and overall uncertainties associated with the measurements.

The procedures, when applied to the evaluation of Pt in urine of 50 subjects from a central region of Italy, gave the median value of 4.13 ng L−1, with minimum and maximum values equal to 1.20 and 27.3 ng L−1, respectively.

This study will continue, as a correlation between the information collated on the study subjects' lifestyle and dietary habits and their urinary Pt levels was outside of the scope of this study.