Determination of Parabens and Triclosan in Indoor Dust Using Matrix

Anal. Chem. 2007, 79, 1675-1681

Determination of Parabens and Triclosan in Indoor Dust Using Matrix Solid-Phase Dispersion and Gas Chromatography with Tandem Mass Spectrometry P. Canosa, I. Rodrı´guez,* E. Rubı´, and R. Cela

Departamento de Quı´mica Analı´tica, Nutricio´ n y Bromatologı´a, Instituto de Investigacio´ n y Ana´ lisis Alimentario, Universidad de Santiago de Compostela, Santiago de Compostela 15782, Spain

A simple sample preparation method for the determination of four parabens and triclosan in indoor dust is presented. Analytes were extracted from the sample and isolated from interfering species using the matrix solidphase dispersion technique. After that, they were silylated and determined by gas chromatography combined to tandem mass spectrometry (GC/MS/MS). The influence of several factors on the yield and selectivity of the extraction was evaluated in detail. Under final working conditions, samples (0.5 g) were mixed with the same amount of anhydrous sodium sulfate and dispersed on 1.25 g of C18. This blend was transferred to the top of a polypropylene cartridge containing 2 g of Florisil. After removing less polar species with 10 mL of dichloromethane, analytes were recovered using 10 mL of acetonitrile. This extract was concentrated to 1 mL, derivatized, and injected in the GC/MS/MS system. Derivatization was carried out at 45 °C in 5 min using 100 µL of N-methyl-N-(tert-butyldimethylsilyl) trifluoroacetamide. Quantification limits from 0.6 to 2.6 ng/g and absolute recoveries between 80 and 114% were achieved. Analysis of dust samples demonstrated the presence of the target species in indoor dust from private houses. The highest average concentration (702 ng/g) corresponded to triclosan. Esters of p-hydroxybenzoic acid (parabens) and 2-(2,4-dichlorophenoxy)-5-chlorophenol (triclosan) are compounds with bactericide and antimicrobial properties employed mainly in the formulation of personal care products (PCPs) such as tooth pastes, deodorants, beauty creams, solar filters, and bath gels. In addition, parabens are added to canned foods and beverages as preservatives and, less often, to textiles. Triclosan is also incorporated as a biocide in sportive clothes, footwear, carpets, plastic toys, and kitchenware.1-3 Although the acute toxicity of parabens and triclosan is low, a growing concern has arisen in relation to their potential long* To whom correspondence should be addressed. E-mail: [email protected]. Fax: 00 34 981 595012. Tf: 00 34 981 563100 ext. 14387. (1) Lokhnauth, J. K.; Snow, N. H. Anal. Chem. 2005, 77, 5938-5946. (2) Zhang, Q.; Lian, M.; Liu, L.; Cui, H. Anal. Chim. Acta 2005, 53, 31-39. (3) Tixier, C.; Singer, H. P.; Canonica, S.; Mu ¨ller, S. Environ. Sci. Technol. 2002, 36, 3482-3489. 10.1021/ac061896e CCC: $37.00 Published on Web 01/06/2007

© 2007 American Chemical Society

term effects on human health and wildlife. In the case of parabens, environmental effects are supposed to be practically negligible since they are effectively removed during the treatment of urban wastewater;4,5 however, a potential link has been proposed between some types of cancer and the prolonged dermal exposure to paraben-containing products, e.g., underarm deodorants.6 Triclosan is a relatively lipophilic compound, not removed completely during conventional treatments of domestic wastewater, presenting a certain trend to be accumulated in sludge, sediments, and biota.7-9 Moreover, it can undergo different transformation reactions rendering more toxic and persistent pollutants such as chlorophenols, dioxins, tetra- and pentachlorinated phenoxyphenols, and methyl triclosan. These byproducts represent a higher risk for the environment and human health than the parent bactericide.10-13 Levels and fate of parabens and triclosan in different environmental compartments have been investigated in detail; however, less information is available regarding their distribution in indoor atmospheres.14 The presence of an increasing number of chemicals in air, airborne particulate matter, and dust from working centers and private houses has been related to some respiratory diseases, e.g., asthma and allergies.15 Analysis of settled dust is useful to assess chronic exposures (oral, dermal, and respiratory uptake) to semivolatile organic compounds (SVOCs) in indoor areas. Most of the sample preparation procedures developed for (4) Lee, H. B.; Peart, T. E.; Svoboda, M. L. J. Chromatogr., A 2005, 1094, 122129. (5) Canosa, P.; Rodrı´guez, I.; Rubı´, E.; Bollaı´n, M. H.; Cela, R. J. Chromatogr., A 2006, 1124, 3-10. (6) Darbre, P. D.; Aljarrah, A.; Miller, W. R.; Coldham, N. G.; Sauer, M. J.; Pope, G. S. J. Appl. Toxicol. 2004, 24, 5-13. (7) Bester, K. Water Res. 2003, 37, 3891-3896. (8) Morales, S.; Canosa, P.; Rodrı´guez, I.; Rubı´, E.; Cela, R. J. Chromatogr., A 2005, 1082, 128-135. (9) Alaee, M.; D’Sa, I.; Bennett, E.; Letcher, R. Organohalogen Compd. 2003, 62, 136-138. (10) Canosa, P.; Morales, S.; Rodrı´guez, I.; Rubı´, E.; Go´mez, M.; Cela, R. Anal. Bioanal. Chem. 2005, 383, 1119-1126. (11) Rule, K. L.; Ebbett, V. R.; Vikesland, P. J. Environ. Sci. Technol. 2005, 39, 3176-3185. (12) Sa´nchez-Prado, L.; Llompart, M.; Lores, M.; Ferna´ndez-Alva´rez, M.; Garcı´aJares, C.; Cela, R. Anal. Bioanal. Chem. 2006, 384, 1548-1557. (13) Balmer, M. E.; Poiger, T.; Droz, C.; Romanin, K.; Bergqvist, P.; Mu ¨ ller, M. D.;Buser, H. R. Environ. Sci. Technol. 2004, 38, 390-395. (14) Rudel, R. A.; Camann, D. E.; Spengler, J. D.; Korn, L. R.; Brody, J. G. Environ. Sci. Technol. 2003, 37, 4543-4553. (15) Bro ¨ms, K.; Sva¨rdsudd, K.; Sundelin, C.; Norba¨ck, D. Indoor Air 2006, 16, 227-235.

Analytical Chemistry, Vol. 79, No. 4, February 15, 2007 1675

Table 1. Summary of MS/MS Detection Conditions for Parabens and Triclosan, as tert-Butyldimethylsilyl Derivatives

a

comp

mol wt

parent ions (m/z)a

MeP EtP PrP BuP TCS

266 280 294 308 404.5

209 223 237 251 347

product ions (m/z)

quantification ions (m/z)b

excitation amplitude (V)

storage level (m/z)

195, 177, 149 195, 177, 163, 151 195, 163, 151 195, 151 310, 200, 219

177 151 + 177 + 195 151 + 195 151 + 195 200 + 310

0.5 0.5 0.5 0.5 1.5

92 98 104 110 140

Employed also as quantification ions in the MS detection mode. b Using MS/MS detection.

the determination of SVOCs in dust use exhaustive extraction approaches, e.g., soaking with organic solvents, Soxhlet, accelerated solvent extraction, and supercritical fluid extraction.14,16-18 All these techniques present a poor selectivity; consequently, primary sample extracts usually contain a high level of coextracted organic species, which have to be removed in further timeconsuming cleanup steps. Matrix solid-phase dispersion (MSPD) is a low-cost sample preparation technique that presents a reduced consumption of organic solvents, provides high extraction yields, and offers a considerable degree of selectivity. It involves blending the sample with a solid sorbent, in a mortar with a pestle, to reduce the particle size of the original matrix and to achieve its homogeneous distribution around dispersant particles. After that, the blend is transferred to an empty cartridge, or a syringe barrel, and analytes are extracted using a small volume of an appropriate solvent. Involved equilibriums are somehow similar to those occurring in a chromatographic separation, with the dispersed sample serving as the stationary phase. Extracts can be analyzed as recovered from the MSPD cartridge or subjected to a purification step. The latter can be performed simultaneously to extraction, just by placing a cosorbent at the bottom of the cartridge or syringe barrel.19,20 In the earlier applications of MSPD, only functionalized silica sorbents, mainly C8 and C18, were considered as dispersants. Organic compounds contained in the sample are dissolved and dispersed into the silica-bonded apolar phases.19 In further works, normal-phase sorbents and relatively inert materials, such as sand and diatomaceous earth, have been employed also as dispersants.20-22 Obviously, in this case, interactions between sample components and dispersant take place only through adsorption processes. Regarding its application field, MSPD has been used mainly in the analysis of food and biota samples;23-25 however, some successful methods have been developed also for the determina(16) Clausen, P. A.; Lindeberg, Bille, R. L.; Rikke, L.; Nilsson, T.; Hansen, V.; Svensmark, B.; Bowadt, S. J. Chromatogr., A 2003, 986, 179-190. (17) Reighard, T. S.; Olesik, S. V. Anal. Chem. 1997, 69, 566-574. (18) Ingerowski, G.; Friedle, A.; Thumulla, J. Indoor Air 2001, 11, 145-149. (19) Barker, S. A. J. Chromatogr., A 2000, 885, 115-127. (20) Kristenson, E. M.; Ramos, L.; Brinkman, U. A. Th. Trends Anal. Chem. 2006, 25, 96-111. (21) Bogialli, S.; Curini, R.; Di Corcia, A.; Nazzari, M.; Samperi, R. Anal. Chem. 2003, 75, 1798-1804. (22) Furusawa, N. Anal. Bioanal. Chem. 2004, 378, 2004-2007. (23) Ramil, M.; Hernandez, D.; Rodriguez, I.; Cela, R. J. Chromatogr., A 2004, 1056, 187-194. (24) Pensado, L.; Casais, M. C.; Mejuto, M. C.; Cela, R. J. Chromatogr., A 2005, 1077, 103-109. (25) Valsamaki, V. I.; Boti, V. I.; Sakkas, V. A.; Albanis, T. A. Anal. Chim. Acta 2006, 573+574, 195-201.

1676 Analytical Chemistry, Vol. 79, No. 4, February 15, 2007

tion of concerning organic pollutants in other matrixes such as soils, sediments, and freeze-dried sludge from sewage water treatment plants.26,27 Aims of this work are as follows: (1) to assess the suitability of the MSPD technique for the one-step extraction of parabens and triclosan from settled dust samples, achieving a certain selectivity in this process, and (2) to determine the levels of these bactericides in a representative number of dust samples from private homes. After extraction, target species were transformed in the corresponding tert-butyldimethylsilyl derivatives and determined by gas chromatography using tandem MS detection. EXPERIMENTAL SECTION Standards and Material. Methanol, n-hexane, dichloromethane, ethyl acetate, acetonitrile, all trace analysis grade solvents, and anhydrous sodium sulfate were supplied by Merck (Darmstadt, Germany). Triclosan (TCS), methyl (MeP), ethyl (EtP), propyl (PrP), and butyl paraben (BuP), as well as the derivatization reagent N-methyl-N-(tert-butyldimethylsilyl) trifluoroacetamide (MTBSTFA), were purchased from Aldrich (Milwaukee, WI). Individual solutions of each analyte were prepared in methanol. Further dilutions and mixtures of them were made in methanol, ethyl acetate, and acetonitrile. tert-Butyldimethylsilyl derivatives of the five bactericides were obtained by addition of a given volume of MTBSTFA to solutions (0.5-mL volume) of the compounds in ethyl acetate or acetonitrile. Under final working conditions, derivatization was accomplished in acetonitrile. Optimal conditions were established using 0.5-mL aliquots of a pooled extract obtained from several dust samples. Influence of time, temperature, and volume of MTBSTFA on the yield of the silylation reaction was evaluated simultaneously using an experimental factorial design. The factor temperature was considered at three different levels (20, 45, and 70 °C). Time and volume of MTBSTFA were evaluated at two: 5-60 min and 20-100 µL, respectively. Florisil (60-100 mesh), alumina (150 mesh), silica (230-400 mesh), C18 (70-230 mesh), and sand (50-70 mesh) were acquired from Aldrich and Merck. The normal-phase materials were activated at 130 °C for 48 h and then allowed to cool down in a desiccator before being used as cleanup cosorbents in the MSPD cartridge. C18 and sand were used as received. Polypropylene solid-phase extraction cartridges (15-mL capacity) and 20µm polyethylene frits were purchased from International Sorbent (26) Li, Z. Y.; Zhang, Z. C.; Zhou, Q. L.; Gao, R. Y.; Wang, Q. S. J. Chromatogr., A 2002, 977, 17-25. (27) Blanco, E.; Casais, M. C.; Mejuto, M. C.; Cela, R. Anal. Chem. 2006, 78, 2772-2778.

Figure 1. Comparison of extraction efficiencies achieved as a function of the cosorbent material (2 g), using acetonitrile or ethyl acetate (20 mL) as extraction solvents. Data obtained for 0.5-g fractions of a pooled dust sample dispersed on 2 g of C18. Average responses of duplicate extractions. Error bars represent the values for individual extractions.

Technology (Mid Glamorgan, UK). Syringe filters (Millex GV, 13 mm, 0.22 µm) were obtained from Millipore (Billerica, MA). Samples and Sample Preparation. Indoor dust was obtained from private houses using conventional vacuum cleaners equipped with paper dust bags. The content of bags was sieved and the fraction with a particle size under 60 µm considered in this study. Sieved samples were stored at 4 °C using light-protected glass vessels. Spiked samples were prepared by adding a standard solution in methanol, containing all or some of the considered compounds, to accurately weighed fractions of sieved dust samples, ∼1 mL of standard solution was used/g of dust. The resulting slurry was homogenized thoroughly and left at room temperature until complete evaporation of the solvent. Then, it was stored in amber glass vessels at 4 °C and aged for at least two weeks before being extracted. Optimization of extraction conditions was performed using two different pooled and spiked dust samples. The first was fortified with all compounds at 100 ng/g. The second was spiked only with EtP and BuP (300 ng/g) since it already contained important levels of MeP, PrP, and triclosan. Their TOC were 20.0 and 25.6%, respectively. Influence of different experimental factors on the performance of the MSPD method was evaluated using both an univariant approach and the experimental factorial design methodology. In the second case, as for optimization of silylation conditions, the Statgraphics statistical package was used to generate the experimental matrix and to calculate the standardized main effects of the considered factors.28 In all experiments, 0.5 g of dust was first dried with 0.5 g of anhydrous sodium sulfate to improve the interaction between the analytes and the extraction solvent. Then, they were dispersed on C18 in a glass mortar with a pestle until obtaining a visually homogeneous blend, ∼5 min. This blend was transferred to the top of a SPE cartridge containing an appropriate cosorbent and a polyethylene frit at the bottom. Another frit was placed on top of the cartridge. Under final working conditions, 1.25 g of C18 and 2 g of Florisil were used as dispersant and cosorbent, respectively. MSPD cartridges were first rinsed with 10 mL of dichloromethane to remove organic compounds with a lower polarity than analytes. Then, parabens and triclosan were quantitatively recovered using 10 mL of acetonitrile. Both solvents were passed through the (28) Statgraphics Plus Manual, Experimental Design. Manugistics, Rockville, MD, 2005.

Figure 2. GC/MS/MS chromatograms (total ionic current signal) for a silylated standard (100 ng/mL of each compound) (A), acetonitrile extracts obtained from fractions of a pooled, spiked dust sample without considering a washing step (B), and after rinsing the MSPD cartridge with dichloromethane previously to analyte extraction (C).

dispersed sample by gravity. The acetonitrile extract was evaporated, using a gentle stream of nitrogen, until an approximate volume of 0.5-0.6 mL and then made up to 1 mL in a volumetric glass flask. After filtration, 0.5 mL of this extract were derivatized under optimized conditions. Determination. Silylated species were determined by GC/ MS/MS. The employed system was a Varian CP 3900 gas chromatograph (Walnut Creek, CA) connected to an ion trap mass spectrometer (Varian Saturn 2100). Separations were carried out using a HP-5 MS capillary column (30 m × 0.25 mm i.d.; df, 0.25 µm) supplied by Agilent (Wilmintong, DE). Helium (99.999%) was used as carrier gas at a constant flow of 1 mL/min. The GC oven was programmed as follows: 50 °C (held for 1 min), at 10 °C/ min to 270 °C (held for 10 min). The GC/MS interface and the ion trap temperatures were set at 270 and 220 °C, respectively. Injections (1-µL volume) were made in the splitless mode (splitless time 1 min), with the injector port at 260 °C. The mass spectrometer was operated in the electron impact ionization mode (70 eV). MS spectra were recorded in the range from 90 to 550 m/z units. The parent ion for each silylated species, which appeared at a [M - 57]+ m/z units, was isolated from other ions and subjected to collision-induced dissociation. Optimization of Analytical Chemistry, Vol. 79, No. 4, February 15, 2007

1677

Figure 3. Distribution of parabens and triclosan in consecutive dichloromethane and acetonitrile fractions eluted from dust samples dispersed on 1.25 g of C18, considering 1 or 2 g of Florisil as cosorbent in the MSPD cartridge. Elution protocol: (A) 3 × 10 mL of CH2Cl2 followed by 1 × 15 mL of CH3CN, (B) 1 × 10 mL of CH2Cl2 followed by 4 × 5 mL of CH3CN. N ) 2 replicates. Table 2. Experimental Domain and Standardized Values for Main Effects of Factors Considered in the Experimental Design factor

range of values low

solvent volume (mL) C18 mass (g) a

ethyl acetate 5 0.5

standardized main effects

medium 15

high acetonitrile 25 2.0

MeP 2.70a 0.86 -0.02

EtP

PrP

BuP

TCS

5.00a

4.29a

6.52a

0.08 1.16

0.58 0.35

0.34 0.73

4.52a 0.79 -0.69

Statistically significant factors at the 95% confidence level

MS/MS fragmentation conditions has been described in detail elsewhere.5,8 A summary of MS/MS detection parameters is given in Table 1. Concentrations of analytes in spiked and nonspiked dust samples were calculated by external calibration, comparing the responses obtained for standards and sample extracts in the same solvent (acetonitrile under final optimized conditions), and derivatized using identical conditions. GC coupled to MS/MS was used as the quantification technique; moreover, in some cases, single MS detection was also considered to assess the level of coextracted interferences under different experimental conditions. RESULTS AND DISCUSSION Optimization of Extraction Conditions. One of the most attractive features of MSPD methods is that analytes can be extracted from the sample and separated from interfering species 1678

Analytical Chemistry, Vol. 79, No. 4, February 15, 2007

in a single step. Preliminary extraction experiments were carried out with the aim of obtaining a first estimation of the role of different cosorbents on the selectivity of the MSPD process. For this purpose, GC/MS was employed as determination technique. Fractions (0.5 g) of the pooled sample spiked with the analytes at 100 ng/g and 2 g of C18 were used in all cases. As cosorbents, 2 g of one of the following materials was considered: C18, silica, alumina, or Florisil. Ethyl acetate (20 mL) was chosen as the extraction solvent in this first set of experiments since it had been employed previously for the elution of TCS and parabens from solid-phase extraction cartridges.10,29 Sample extracts were evaporated, adjusted to 1 mL, and filtered. An aliquot of the filtrate (500µL volume) was mixed with 40 µL of MTBSTFA, allowed to react (29) Canosa, P.; Rodrı´guez, I.; Rubı´, E.; Negreira, N.; Cela, R. Anal. Chim. Acta 2006, 575, 106-113.

Table 3. Comparison of Extraction Efficiencies Using C18 or Sand as Dispersant Materialsa concentration (ng/g) ( standard deviation dispersant C18 sand

MeP 598 ( 18 537 ( 14

EtP 122 ( 7 107 ( 4

PrP 733 ( 54 678 ( 14

BuP 119 ( 5 127 ( 6

Table 4. Precision, Recoveries for Spiked Samples (n ) 4 Replicates), and Quantification Limits (S/N 10) of the MSPD Method precision, RSD (%)

TCS 1169 ( 86 1217 ( 93

a Data for a nonspiked composite dust sample. GC/MS/MS detection, n ) 4 replicates.

for 5 min at room temperature, and injected in the GC/MS system.10 Independently of the cosorbent material, similar responses (peak areas) were obtained for all compounds, data not given; however, the complexity of GC/MS chromatograms was lower for alumina or Florisil than in the case of silica and C18; figure not shown. These results indicated that interferences extracted from the dispersed sample, together with parabens and triclosan, were more strongly retained in the cosorbent layer by the first two materials. In a second series of experiments, the extraction capabilities of four solvents with increasing polarities, n-hexane, dichloromethane, ethyl acetate, and acetonitrile, using C18 as dispersant (2 g) and alumina or Florisil (2 g) as cosorbents, were compared. In all experiments, 20 mL of solvent and the same spiked sample as in the above paragraph was used. In order to avoid apparent changes in the yield of the extraction, due to variations in the performance of the silylation reaction as a function of the organic solvent, all extracts were evaporated to dryness and reconstituted with 1 mL of ethyl acetate before being derivatized. Compounds were determined by GC/MS/MS using conditions given in Table 1. Independently of the cosorbent material, the less polar solvents, dichloromethane and n-hexane, failed to recover triclosan and parabens from the MSPD cartridge. Probably, their interaction with the medium polar analytes (log Kow from 1.9, for MeP, to 4.8, for TCS) was too weak to remove them either from the dispersed sample itself or from the polar cosorbents placed at the bottom of the cartridge. Responses obtained using acetonitrile and ethyl acetate are shown in Figure 1. In general, both solvents provided higher extraction yields employing Florisil than using alumina. Thus, the first was chosen as the cosorbent, whereas both solvents were considered for further optimization experiments. Figure 2 compares the GC/MS/MS chromatograms obtained for a standard (Figure 2A) and an extract from the pooled dust sample using acetonitrile as elution solvent and Florisil as cosorbent (Figure 2B). In spite of the high selectivity of the MS/ MS detection mode, a relatively complex chromatogram, particularly in the detection window of BuP, was obtained (similar chromatographic profiles were obtained using ethyl acetate). The number of peaks in this region could be reduced by rinsing the MSPD cartridge with 10 mL of dichloromethane, previously to extract the analytes using acetonitrile or ethyl acetate, (Figure 2C). It seems evident that the washing step allowed one to remove many low polar organic compounds, which otherwise will appear in the same fraction as the analytes. The influence of the mass of dispersant (C18), volume, and type of elution solvent (acetonitrile or ethyl acetate) on the yield of the extraction was evaluated simultaneously using a mixed mode 31 × 22 experimental factorial design with two central

recovery ( RSD (%)

repeatability reproducibility spiked spiked (n ) 4 (n ) 12 conc conc QL compd replicates) replicates) 50 ng/ga 300 ng/gb (ng/g) MeP EtP PrP BuP TCS a

4.7 4.5 3.4 5.3 6.0

8.6 4.4 2.8 4.4 13.0

114 ( 9 87 ( 10 93 ( 7 80 ( 5 108 ( 6

102 ( 3 90 ( 5 98 ( 5 84 ( 4 102 ( 5

1.3 2.6 0.6 0.7 1.5

300 ng/g in the case of triclosan. b 700 ng/g in the case of triclosan.

Table 5. Concentrations of Parabens and Triclosan in Nonspiked Dust Samples, n ) 4 Replicates concentration (ng/g) ( standard deviation sample code

MeP

EtP

PrP

BuP

TCS

1 2 3 4 5 6 7 8 9 10 average

53 ( 7 89 ( 9 190 ( 27 970 ( 20 440 ( 90 580 ( 17 780 ( 80 610 ( 60 470 ( 8 190 ( 38 468

8(2 290 ( 31 48 ( 4 56 ( 3 60 ( 10 123 ( 6 370 (12 84 ( 6 23 ( 2 22 ( 4 108

16 ( 2 160 ( 19 250 ( 25 1050 ( 33 400 ( 70 540 ( 40 530 ( 30 470 ( 21 430 ( 15 220 ( 31 406

4(1 28 ( 5 15 ( 3 101 ( 3 51 ( 8 210 ( 10 190 ( 11 99 ( 1 35 ( 1 29 ( 4 76

700 ( 117 240 ( 16 580 ( 64 444 ( 9 280 ( 40 960 ( 70 2200 ( 170 850 ( 40 470 ( 23 300 ( 13 702

points.28 Extractions were carried out using 0.5-g fractions of the second pooled, spiked sample described in Samples and Sample Preparation. The volume of elution solvent was tested at three levels: 5, 15, and 25 mL, and the type of solvent and mass of C18 at two, Table 2. In the 14 experiments comprised in this design, dispersed samples were first rinsed with 10 mL of dichloromethane. After that, analytes were eluted with the corresponding volume of ethyl acetate or acetonitrile according to conditions defined in each experiment of the design. All extracts, including dichloromethane fractions, were evaporated to dryness, reconstituted with 1 mL of ethyl acetate, and mixed with a fresh aliquot of MTBSTFA. None of the analytes was detected in the washing dichloromethane fractions. Concentrations in extracts corresponding to the 14 experiments of the experimental factorial design were determined by external calibration. Standardized values for the main effects of type, volume of extraction solvent, and mass of C18 in the response achieved for each analyte are given in Table 2. The absolute value of a main effect is proportional to the variation on the yield of the extraction, for a given compound, when the considered factor changes from the low to the high level defined in the domain of the design. Its sign indicates whether the extraction yield increases (positive sign) or decreases (negative sign). Confirming the trend depicted in Figure 1, statistically significant (95% confidence level) higher extraction efficiencies were achieved with acetonitrile than with ethyl acetate, Table 2. The volume of solvent showed a positive but minor effect on the extraction process; however, the quadratic term associated with this factor was relatively important, particularly in the case of MeP, Analytical Chemistry, Vol. 79, No. 4, February 15, 2007

1679

Figure 4. GC/MS/MS chromatogram corresponding to a nonspiked dust sample (code 5, Table 4).

figure not show. Experimentally, for all compounds, the highest extraction yields were achieved considering the medium level (15 mL) defined for this factor in the domain of the experimental design. Finally, the main effect associated with the amount of dispersant never achieved the statistical significant bound and its sign was compound dependent, Table 2. On the basis of these data, acetonitrile was chosen as elution solvent and the mass of dispersant fixed at 1.25 g, an intermediate value within the domain of the design. At this point, no decision was taken regarding the optimal volume of acetonitrile. In order to avoid the need of introducing a solvent exchange step in the analytical procedure, derivatization conditions were optimized using acetonitrile, instead of ethyl acetate, as the organic solvent. The employed methodology is described above. As in the case of ethyl acetate, the reaction was rather fast; however, it required a slightly higher temperature and a larger volume of MTBSTFA. Optimal values were 45 °C, 100 µL of MTBSTFA, and a reaction time of 5 min. The stability of the derivatized analytes was evaluated by injecting them at different times after preparation. When being stored at 4 °C, they were stable for at least one week. As a further step in the development of the MSPD method, optimal volumes for washing and extraction solvents (dichloromethane and acetonitrile, respectively) as well as the mass of Florisil were established using an univariant approach. Two series of experiments were carried out, in duplicate, using 0.5-g fractions of dust dispersed on 1.25 g of C18 and considering 1 or 2 g of Florisil as cosorbent. In the first series of extractions, three 1680

Analytical Chemistry, Vol. 79, No. 4, February 15, 2007

consecutive fractions of dichloromethane (10 mL each) followed by one of acetonitrile (15 mL) were passed through the dispersed samples. In the second one, a fraction of 10 mL of dichloromethane was followed by four of acetonitrile (5 mL volume). Figure 3 depicts the average responses (normalized peak areas) for each analyte in the different fractions collected from MSPD cartridges for both series of experiments. Differences between duplicate experiments ranged from 1 to 10%, with the higher variations corresponding to the first series of extractions. Using 1 g of Florisil, significant amounts of TCS, the less polar of the analytes (log Kow 4.8), already appeared in the first 10 mL of dichloromethane collected from the MSPD cartridge, whereas employing 2 g of Florisil, neither triclosan nor parabens were detected in the first two dichloromethane fractions, Figure 3A. In the second series of experiments, independently of the amount of Florisil, the first fraction of acetonitrile (5 mL) extracted ∼95% of the analytes, whereas percentages higher than 98% were achieved with two fractions (10 mL) of this solvent, Figure 3B. Obviously, employing 1 g of Florisil, TCS was lost partially in the previous 10-mL dichloromethane fraction, Figure 3B. Dichloromethane fractions from the first series of experiments were also analyzed by GC/MS. Even when 2 g of Florisil was used as cosorbent, the bulk of interfering organic compounds with a lower polarity than target analytes was removed from the MSDP cartridge using just 10 mL of this solvent. Taking into account the above results, the mass of Florisil was fixed at 2 g, and the volumes of washing (dichloromethane) and elution solvent (acetonitrile) at 10 mL. In fact, the last was a rather conservative decision. In order to avoid

an excessive dilution of the extract, the volume of acetonitrile could be limited to only 5 mL with a very little diminution in the yield of the extraction. Recently, some authors have described the possibility of using inert materials, e.g., sand and diatomaceous earth, instead of functionalized silica sorbents as sample dispersants.20-22 As a result, the cost of MSPD extractions is reduced considerably. In order to explore the suitability of this approach for the determination of parabens and triclosan in dust, a pool of several nonspiked samples was prepared. Fractions of this composite sample were dispersed in a mortar using 1.25 g of C18 or sand. In both cases, 2 g of Florisil, 10 mL of dichloromethane, and 10 mL of acetonitrile were employed as cosorbent, washing, and elution solvents, respectively. The yield of the extraction was equivalent for C18 and sand, Table 3; however, cleaner chromatograms were obtained for the first, showing its capability to retain strongly some of the organic compounds contained in the sample; figure not shown. Considering the higher selectivity achieved in the extraction process, C18 was kept as dispersant. Performance of the Method. The linearity in the response of the GC/MS/MS system, was evaluated using silylated standards in acetonitrile at seven concentration levels in the range from 2 to 1500 ng/mL. Correlation coefficients from 0.996 to 0.999 were achieved. Instrumental quantification limits, defined for a signal-to-noise (S/N) ratio of 10, ranged from 0.3 ng/mL for PrP and BuP to 1.3 ng/mL for EtP. Precision of the whole method was evaluated with the relative standard deviation data (RSD) corresponding to the extraction of 0.5-g fractions of the second pooled dust sample described in the Experimental Section. Under repeatability conditions (4 extractions in the same day), RSD values remained below 6% for all species. Reproducibility (12 extractions performed in four different days) ranged from 3 to 13%, Table 4. Recovery studies were accomplished using a discrete dust sample containing low levels of parabens (code 1, Table 5) and a TOC of 17.9%. This sample was divided in three portions, and two of them were spiked with the analytes at two different concentration levels. After being aged, ∼10 days, they were extracted in triplicate. Recoveries were calculated by dividing the difference between concentrations measured for spiked and nonspiked portions of the same sample by the added amount. Data, expressed as percentage, ranged from 80% for BuP to 114% for MeP. Their associated RSD remained under 10%, Table 4. Procedural blanks did not reveal contamination problems for any of the considered analytes. Quantification limits of the whole method ranged from 0.6 to 2.6 ng/g, Table 4. For parabens, these values are ∼100-fold lower than the quantification limit reported for MeP in indoor dust using Soxhlet extraction and GC/MS analysis.14 In the case of triclosan, this work constitutes the first application to its determination in dust samples. Achieved recoveries and quantification limits are similar to those reported for other complex matrixes, such as sludge.7,8 Application to Real Samples. The optimized method was employed to assess the levels of parabens and triclosan in indoor dust from 10 private houses located in the northwest of Spain. The five analytes were presented at quantifiable levels in all samples, Table 5. In fact, in most cases, the developed method was sensitive enough to allow the quantification of target species

without concentration of acetonitrile extracts (10 mL) to a final volume of 1 mL. A typical GC/MS/MS chromatogram is shown in Figure 4. With the only exception of sample code 2, MeP and PrP were always found at higher concentrations than EtP and BuP. In fact, average concentrations for the first two compounds were ∼4 times higher than those corresponding to the latter ones. The same distribution pattern has been reported in raw wastewater samples,4,5 which receive the direct discharge of some PCPs (e.g., toothpastes and bath gels). Spillage of liquid PCPs and use of paraben-containing spray deodorants might explain the presence of these compounds in indoor dust. Paraben concentrations reported in Table 5 are lower than those measured for the same matrix in the United States;14 however, the percentage of positive samples is significantly higher in this study. Triclosan was also detected in all dust samples. The average value (702 ng/g) was not far away from the microgram per gram range, which is the typical level reported for this compound in sludge.7,8 In addition to PCPs products, direct leaching from plastic materials and textiles, particularly carpets treated with this biocide, might contribute significantly to the presence of triclosan in indoor dust. From the best of our knowledge, previous studies focused on the presence and distribution of triclosan in indoor environments are not available. CONCLUSIONS The suitability of the MSPD technique for the extraction of four parabens and triclosan from indoor dust has been demonstrated for first time. The developed sample preparation method does not require the use of expensive instrumentation and employs moderate volumes of organic solvents; moreover, it allows the inclusion of a cleanup step in the sample preparation scheme without requiring any additional manipulation of sample extracts. As an additional advantage of this method, it must be pointed that the extraction solvent was compatible with the further silylation of the analytes, advisable to improve the performance of their gas chromatography determination. On the whole, the proposed method provides acceptable recoveries and enough sensitivity for the analysis of real-life polluted samples. The analysis of a limited number of dust samples has proved the ubiquitous presence of parabens and triclosan in indoor environments from private homes. Particularly, important concentrations of MeP, PrP, and triclosan were found. Thus, indoor dust represents another source of chronic exposure to these compounds. ACKNOWLEDGMENT Financial support from the Spanish Government and FEDER funds (project DGICT CTQ2006-03334) are acknowledged. P.C. acknowledges a FPU grant from the Spanish Ministry of Education.

Received for review October 9, 2006. Accepted December 5, 2006. AC061896E Analytical Chemistry, Vol. 79, No. 4, February 15, 2007

1681