3. BLOOD AND URINE ANALYSIS
3.2. Calibration, Limits of Detection, and Method
In our laboratory, we have measured the urine and blood concentrations of BTEX by SPME–GC–MS to assess individual exposure to monoaromatic hydrocarbons polluting the urban air [43]. Unlike the use of urine, the use of human blood for method development raises ethical problems, and this strongly limits the number of possible preliminary experiments. Therefore, only few experiments were per- formed to optimize exposure time, temperature, and stirring rate. On the other hand, relevant data on SPME of BTEX are already available from the literature and it is always advisable to consider all the previous work, although in different matrices [34]. For headspace sampling of BTEX, we used a 75-àm Carboxen/
PDMS SPME fiber, which is known to show a very high affinity for aromatic compounds [34]. The GC–MS analyses were carried out on a HP-5MS column (30 m⫻0.25 mm, 0.25-àm film) using hydrogen as carrier gas. Analytes were desorbed at 280°C for 3 minutes in the GC injector. No carryover was observed.
The method was applied to the quantitative determination of blood and urine concentrations of BTEX, as biomarkers of exposure to environmental air pollut- ants, among a group of 24 volunteers cycling for 2 h along city routes. Just before and immediately after the runs, subjects had to provide blood and urine samples.
Each sample was analyzed in duplicate. Blood concentrations of 180, 285, 213, and 722 ng/L (median values) were found in prerun specimens for benzene, tolu- ene, ethylbenzene, and total xylenes, respectively. The corresponding urinary concentrations were 89, 280, 73, and 220 ng/L. Changes in blood benzene and toluene were observed after exercise, consistent with airborne concentrations.
Significant variations between prerun and postrun urine samples were not ob- served. A GC–MS chromatogram obtained in selective-ion monitoring mode of a blood sample from a nonsmoker subject obtained at the equilibrium conditions is shown in Figure 6.
Solid-Phase Microextraction 241
Figure 6 A GC–MS chromatogram in selective ion monitoring mode of a blood sample from a non-smoker subject. SPME fiber: 75-àm-Carboxen/PDMS. Extraction conditions:
37°C, 30 minutes, 500 rpm. Peak identification: 1, benzene⫹I.S.; 2, toluene; 3, ethylben- zene; and 4,m-⫹p-xylene; 5,o-xylene.
3.2.1. Calibration
Since blood and urine have a natural content of BTEX, the standard addition procedure was chosen to assess linearity of the SPME–GC–MS method. External calibration and the use of literatureKvalues are only possible with simple and clean matrices, i.e., air or clean water [37,42]. In the case of more complex sam- ples, such as soils or biological fluids, the use of isotopically labeled internal standards or the standard addition is recommended [1,12]. Blood (with heparin) and urine were spiked with BTEX mixtures in order to obtain concentration in- creases in the 0 to 50àg/L range. We prepared a calibration standard set in order to spike the calibration samples (2 ml) with the same small volume of methanolic standard (20àl). The addition of methanol or other organic solvents to aqueous matrices should always be lower than 1%. Moreover, methanol itself contains trace amounts of BTEX. We used13C6-benzene as internal standard to improve the reproducibility of the data. Although not separated from benzene,13C6-ben- zene has unique ions for quantitation. Since isotopically labeled compounds have chemical and physical properties very similar to those of the unlabeled analogs, they are considered as the best SPME–GC–MS internal standards. Valuable quantitative results at very low concentration levels were obtained using particu- lar care during sample preparation and storage. Samples for calibration and au- thentic samples were prepared and frozen in a similar way. In the case of blood, we could obtain linear response curves only when the samples were frozen 2 hr after the spiking, indicating the relevant role of the equilibration time of the standard spiked into the matrix. Without equilibration of the analytes in the ma-
242 Manini and Andreoli
trix, no reliable calibration could be obtained. This behavior is peculiar to blood and was not observed in the case of urine, probably due to the different lipophilic components of the two matrices. In general, it is always advisable that the spiked analytes equilibrate with the natural content already present in the matrix and with the matrix itself. Moreover, when extracting conditions are optimized, it should be kept in mind that the interaction between the matrix and native analytes should be stronger than in the case of spiked analytes [44].
With the standard addition method, experimental data fit a linear model y⫽a ⫹bx, wherexis the concentration increases andythe chromatographic peak areas to IS area ratio. The correct standard addition procedure requires the standard addition to every sample, to overcome the matrix effect, i.e. the differ- ence in the composition of human fluids due to the interindividual biological variance. As a drawback, standard addition leads to a high number of samples to be processed. In the case of urine, there is also the possibility of minimizing the matrix effects using a pool of urine for calibration or minimizing differences in salt composition by adding salt up to saturation. By contrast, blood samples from different subjects cannot be mixed because they are incompatible with each other. Moreover, in the case of blood, the requirement of a larger volume of sample to perform standard additions is not always acceptable.
According to our study design, a large number of urine and blood samples to be analyzed in a time limited by analyte stability was expected (n ⫽ 192).
For these reasons, we decided to evaluate preliminarily the variability due to matrix effects. Using blood and urine from different subjects (n⫽5), we repeated the entire standard addition calibration procedure several times. What we found was only a slight matrix effect, comparable with the precision of the method:
different blood (or urine) samples used for calibration gave rise to equations with different intercept values (a, the native content), but with similar (CV ⬍10%) slope values (b). We decided to use the equations obtained with standard addition to a limited number of samples as external calibrations to calculate blood and urine concentration of BTEX in all the samples, aftery-axis shift (y⫽bX, where X⫽x⫹a/b). On the other hand, our purpose was to measure BTEX at the ppt level and also to appreciate differences between prerun and postrun blood and urine samples from the same subject. This approach cannot be generalized with- out the evaluation of the matrix effect case by case. The affinity of the matrix for the analytes strongly varies depending on the chemical and physical properties of both the matrix and the analytes. In some cases, the variability is too large to be acceptable [17]. In any case, when the number of biological samples is limited and a sufficient amount of sample is available, it is always advisable to spike each sample.
Figure 7 shows the calibration graphs obtained in the case of blood BTEX.
Linearity was established for concentrations up to 50àg/L. A very wide linear
Solid-Phase Microextraction 243
Figure 7 Calibration graphs obtained in the case of blood BTEX.
dynamic range is recognized for SPME–GC using flame ionisation or quadrupole MS detection. The use of ion trap MS can limit the linearity of SPME [45].
3.2.2 Limits of Detection
Since measurable concentrations of BTEX are present in the blank matrix, detec- tion limits were determined using the corresponding deuterated compounds. Iso- topically labeled compounds are also useful to improve specificity and to ensure precision in peak assignment. Limits of detection were 5 ng/L for benzene and toluene, and 10 ng/L for ethylbenzene and xylenes. Detection limits of 10 to 20 ng/L were also obtained for halogenated anaesthetics. The sensitivity of SPME depends onKfsvalues and is not enhanced by larger sample volumes, especially for compounds withKfs⬍500 [12].
3.2.3 Precision
Precision of the method is controlled by several factors, first of all, the status of the SPME fiber. The mean fiber lifetime is 100 runs when desorbing into an injector heated to 220°C or immersing into water saturated with salt and at pH 2. Direct sampling from dirty matrices may cause faster degradation of the fiber, due to the adsorption of high-molecular-weight species such as proteins, salt crys- tals, or humic materials. If the process is reversible, the fiber can be soaked with a proper solution [8]. When a fiber used for quantitative analysis starts to degrade,
244 Manini and Andreoli
it should be substituted immediately. Also the aging of the injector septum causes deterioration of precision. Our approach was to change the fiber and the septum after 100 sampling-injections, or earlier when necessary. Before use, new fibers have to be conditioned in the GC injector at the temperature and for the time suggested by the manufacturer, 280°C for 30 minutes in the case of Carboxen/
PDMS. Besides the fiber status, another important factor for the precision of the method is the sample volume, which should be measured accurately, especially in the case of small volumes. In headspace mode, the headspace volume also should be kept constant from vial to vial. A delay between sample collection and extraction or between extraction and analysis may affect the precision. The most volatile compounds are reported to be stable on the fiber at room temperature for about 2 minutes [15]. The use of unsilanized glassware may cause analyte losses up to 70% in water samples stored for 48 hr [12]. For these reasons, we always analyzed samples immediately after sampling and used only silanized glass vials. In our BTEX study, we evaluated the repeatability of the method from six consecutive extraction-injections of the same blood or urine sample nonspiked. Repeatability was in the range of 6.5–9.2% (expressed as % RSD) for aromatic compounds. In the case of very volatile analytes, permanent gases or compounds with very low boiling temperatures, such as anaesthetics, the use of gaseous standard solutions instead of liquid solutions was found to improve the reproducibility in standard preparation. We also used a gaseous IS, dichloro- methane, to avoid interference arising from other solvents and peaks overlapping with the peak of nitrous oxide. The calculated intraday and interday precisions for all the anesthetics were 3.0 to 7.2% and 6.5 to 12.9%, respectively. Our find- ings are in agreement with data reported by other authors [14,22,27].
3.2.4 Validation
Validation of a SPME method for target analytes should be performed using standard reference materials with similar matrix, when available. Another possi- ble and frequently used way is validation of a SPME method against well- accepted extraction techniques, such as purge-and-trap [13,25,46] or static head- space [46]. Several interlaboratory studies demonstrated that SPME is a reliable technique for the quantitative analysis of volatile organic compounds [46] and pesticides in water samples [47–48]. We have validated our SPME–GC–MS method for the determination of nitrous oxide in urine by means of the compari- son with static headspace [33].