Evaluation of analytical methods for determining pesticides in baby food

Analytica Chimica Acta 399 (1999) 135–142

Evaluation of analytical methods for determining pesticides in baby food Jane C. Chuang a,∗ , Mary A. Pollard a , Mark Misita a , Jeanette M. Van Emon b b

a Battelle, Columbus, OH, USA National Exposure Research Laboratory, US Environmental Protection Agency, Las Vegas, NV, USA

Received 5 November 1998; received in revised form 26 April 1999; accepted 29 April 1999

Abstract Three extraction methods and two detection techniques for determining pesticides in baby food were evaluated. The extraction techniques examined were supercritical fluid extraction (SFE), enhanced solvent extraction (ESE), and solid phase extraction (SPE). The detection techniques used were enzyme-linked immunosorbent assay (ELISA) and gas chromatography/mass spectrometry (GC/MS). Different SFE and ESE conditions were considered, and the resulting extracts were analyzed by either ELISA or GC/MS. The use of C18 SPE cartridges to extract pesticides from baby food was also evaluated. Using SFE–ELISA, recoveries of most spiked pesticides were less than 50% in both non-fat and fatty baby food. Using SFE–GC/MS, recoveries of target pesticides were greater than 80% in dried baby food, but 10–60% of the spiked pesticides were lost during the freeze-drying process. Off-line coupling of SPE–GC/MS provided a quantitative measure (>80%) of the pesticides in non-fat baby food (fruits and vegetables). Direct ELISA applied to diluted non-fat food also gave a quantitative measure of the target pesticides. The ESE–ELISA method provided a quantitative determination of atrazine in both non-fat and fatty baby food. ©1999 Elsevier Science B.V. All rights reserved. Keywords: SFE–GC/MS; ESE–ELISA; Baby food; Atrazine; Carbofuran; Chlorpyrifos; Metolachlor

1. Introduction Children can be exposed to pesticides by inhaling contaminated air, ingesting tainted food or non-dietary substances (i.e., dust or soil), or absorbing them through the skin from contaminated media. Pilot-scale human exposure studies [1–3] suggest that dietary ingestion may be an important pathway for children to become exposed to persistent organic pollutants including pesticides. Determination of pesticides in food is often complicated by the presence of fats, and therefore requires multiple cleanup steps before ∗ Corresponding author. Tel.: +1-614-424-6424; fax: +1-614424-5263.

analysis. Cost-effective analytical methods are needed for conducting large-scale studies that produce large numbers of samples. Off-line coupling of supercritical fluid extraction (SFE) and enzyme-linked immunosorbent assay (ELISA), a method developed by the US EPA, National Exposure Research Laboratory-Las Vegas (NERL-LV), promises a streamlined approach for food analysis in support of large-scale human exposure assessment studies [4,5]. Additional evaluation/confirmation of this method is required to completely document performance characteristics for real-world samples. The goal of this study was to evaluate SFE–ELISA, as well as other analytical methods for determining pesticides in baby food. The extraction methods

0003-2670/99/$ – see front matter ©1999 Elsevier Science B.V. All rights reserved. PII: S 0 0 0 3 - 2 6 7 0 ( 9 9 ) 0 0 5 8 4 - X

136

J.C. Chuang et al. / Analytica Chimica Acta 399 (1999) 135–142

examined were SFE, enhanced solvent extraction (ESE), and solid phase extraction (SPE). The two detection techniques used in this study were ELISA and gas chromatography/mass spectrometry (GC/MS).

2. Experimental 2.1. Off-line coupling of SFE–ELISA and SFE–GC/MS The four target pesticides evaluated by SFE–ELISA were atrazine, carbofuran, chlorpyrifos, and metolachlor. Initially, SFE–ELISA was performed based on the method described in two related papers [4,5]. The baby food samples used in this study were commercially available (Heinz, Pittsburgh, PA). The baby food was stored at room temperature before use and kept at approximately 0◦ C after the jar was opened. An aliquot (nominal 4 g) of the baby food was removed after thoroughly mixing the contents of the glass jar (baby food container) with a glass rod. The spiked samples were prepared prior to extraction. Known amounts of the four target pesticides (nominal spike level was 2 ng/g of food sample) in a composite solution were spiked into the baby food, and the spiked baby food was then mixed with 1.5 g of Hydromatrix, a pelletized diatomaceous earth (Varian, Harbor City, CA), added to disperse the food sample. The dispersed food sample was placed in the extraction vessel between two 1 g plugs of basic alumina, which retained any fatty material that may have dissolved in the extraction fluid. The spiked and non-spiked baby food samples were extracted with CO2 using an ISCO SFX 200 SFE system. Initially, the extractions were performed at 2205 psi, 70◦ C for 15 min in static mode and 60 min in dynamic mode. The flow rate of CO2 was approximately 1 ml/min (as liquid CO2 ). The CO2 was of SFE grade (Praxair, Danbury, CT). The extracted material was collected in 5 ml of reagent water. An aliquot of 300 ␮l of acetonitrile (ACN) was used as a matrix modifier. Additional SFE tests were carried out in an attempt to improve the recoveries of the spiked pesticides from baby food. For these additional tests, the extraction pressure was kept at 2500 psi, and the

extraction temperature was kept at 70◦ C. The SFE extraction fluid evaluated were 100% CO2 , 10–20% dichloromethane (DCM) in CO2 , and 10–15% ACN in CO2 . The amount of matrix modifier, ACN, used ranged from 0 to 600 ␮l. As chlorpyrifos cannot be recovered from water, dichloromethane (DCM) was used as the collection solvent for most tests. When DCM was used as the collection solvent for SFE, the DCM extract (approximately 15 ml) was concentrated to 1 ml by Kuderna–Danish evaporation. A known amount of internal standard was added to the concentrated DCM extract for GC/MS analysis. The sample extract and standard solutions were analyzed by 70 eV electron impact (EI) GC/MS. Four ELISA test kits – for atrazine, carbofuran, chlorpyrifos, and metolachlor were purchased from Strategic Diagnostics, Inc. (SDI) (Newark, DE). All immunological reagents used were obtained from SDI, including paramagnetic particles coated with anti-pesticide antibody (suspended in a buffered saline with preservatives and stabilizers), pesticide enzyme conjugate, phosphate buffer, hydrogen peroxide solution, chromogenic solution (3,30 ,5,50 -tetramethylbenzidine), stopping solution (2 M sulfuric acid) and washing solution (preserved deionized water). An aliquot of 200–250 ␮l of the aqueous sample extract (depending upon the ELISA test kit), 250 ␮l of pesticide enzyme conjugate, and 500 ␮l of anti-pesticide antibody-coated paramagnetic particle solution were combined in a test tube. After vortexing for 1–2 s, the test tube was incubated at room temperature for 15–30 min (depending upon the test kits). The mixture was separated using a magnetic separation rack. An aliquot of 1 ml washing solution was added to each test tube, and left to stand for a minute; the washing solution was gently decanted. The washing process was repeated twice. The rack containing the tubes was removed from the magnetic separation rack and a 500 ␮l aliquot of a freshly prepared chromogenic solution was added to each test tube. The reaction was stopped after 20 min incubation by adding 500 ␮l stopping solution. The color intensity in each test tube was measured at 450 nm using a PRA-1 photometric analyzer (formerly Ohmicron now SDI). Concentrations of the four target pesticides were determined based on the calibration curve generated from the ELISA of the particulate pesticide at four concentrations – 0, 0.1, 1.0, and 5.0 ng/ml.

J.C. Chuang et al. / Analytica Chimica Acta 399 (1999) 135–142

137

A Hewlett-Packard (HP) GC/MS was operated in the selected ion monitoring mode. Data acquisition and processing were performed with a ChemStation data system. The GC column was a DB-5 fused silica capillary column (60 m × 0.32 mm; 0.25 ␮m film thickness). Helium was used as the GC carrier gas. Following injection, the GC column was held at 70◦ C for 2 min and then was temperature-programmed to 150◦ C at 15◦ C/min and then to 290◦ C at 6◦ C/min. The MS was operated in the SIM mode. Peaks monitored were the molecular ion peaks and their associated characteristic fragment ion peaks. Identification of the target analytes was based on their GC retention times relative to those of corresponding internal standards and relative abundance of the monitored ions. Quantification of target analytes was based on comparisons of the integrated ion current responses of the target ions to those of the corresponding internal standards using average response factors of the target analytes generated from standard calibrations.

Florisil SPE cleanup. The Florisil SPE column was conditioned with 6 ml of DCM, 6 ml 50% DCM in hexane, and 6 ml hexane. The sample extract (in 1 ml of hexane) was applied to the conditioned Florisil column and eluted with 18 ml of 15% ethylether (EE) in hexane. The collected fraction was concentrated to 1 ml for GC/MS analysis. Therefore, two fractions were generated for each fatty baby food for GC/MS analysis.

2.2. Off-line coupling of solid phase extraction (SPE) and GC/MS

An aliquot (nominal 4 g) of baby food was removed after thoroughly mixing the contents of the glass jar with a glass rod. The baby food was mixed with 4 g of sand. Known amounts of target pesticides were spiked into the baby food and sand mixture and transferred to an extraction cell for ESE. The baby food and sand mixture was sandwiched between two plugs of silanized glasswool. The ESE was performed using an ISCO SFX 200 SFE system; deionized water was used as the extraction solvent. The ESE conditions evaluated were extraction pressure, 2000–2500 psi; extraction temperature, 70–150◦ C; and extraction time, 35–50 min. An aliquot (200–250 ␮l) of the collected aqueous extract was then analyzed by ELISA. The ELISA procedure used is the same as described in the SFE–ELISA section.

The C18 SPE column (Bakerbond) was conditioned with 5 ml of DCM, 5 ml of methanol, 5 ml of 1 : 1 methanol/reagent water, and 5 ml of reagent water. For spiked baby food samples, an aliquot of the baby food sample was placed in a clean beaker and spiked with known amounts of target pesticides. Both the spiked and non-spiked baby food samples were diluted with reagent water (2.5–5 ml). The diluted samples were passed through the conditioned C18 SPE column through 60 ml reservoirs packed with silanized glasswool. After all the samples were introduced to the C18 SPE columns, the columns were air-dried for 1 h. Then the columns were eluted with 18 ml of 50% DCM in hexane to remove the target pesticides from the columns. The collected fractions were concentrated to 1 ml for GC/MS analysis. The GC/MS procedures used are similar to those described before. For baby food containing fat, additional preparation steps were applied to the baby food residue. The residue was sonicated with 2 × 10 ml of DCM. The DCM extracts were combined, dried with sodium sulfate anhydrides, and concentrated to 1 ml. The DCM extract was then solvent exchanged into hexane for

2.3. Direct ELISA for non-fat baby foods An aliquot (1 g) of baby food was diluted with 5 ml of reagent water. The diluted baby food was filtered through silanized glasswool in a 60 ml reservoir attached to a SPE manifold. The aqueous extract was ready for ELISA. The ELISA procedure used is the same as described in the SFE–ELISA section. 2.4. Off-line coupling of ESE and ELISA

3. Results and discussion 3.1. Evaluation of SFE–ELISA and SFE–GC/MS The recovery data for the off-line coupling of SFE–ELISA, as summarized in Table 1, are the average and standard deviation values from triplicate assays. Initially, the SFE conditions used were based

138

J.C. Chuang et al. / Analytica Chimica Acta 399 (1999) 135–142

Table 1 Summary of recovery data for SFE–ELISA Baby food (spiked level)

SFEa condition code

Recovery (%) Carbofuran

Atrazine

Metolachlor

Chlorpyrifos

Chicken noodle (2 ppb)

A

35 ± 7.2

40 ± 8.1

11 ± 2.3

NDb

Chicken noodle (2 ppb)

B

51 ± 9.3 28 ± 5.5

55 ± 9.7 36 ± 7.5

22 ± 4.7 11 ± 2.1

ND ND

Chicken noodle (4 ppb)

B

57 ± 4.2 49 ± 3.5

54 ± 1.5 48 ± 4.4

39 ± 4.0 20 ± 1.5

ND ND

Banana/tapioca (2 ppb)

B

69 ± 3.2 69 ± 2.3 63 ± 4.7

77 ± 4.2 74 ± 7.2 71 ± 4.9

37 ± 3.6 45 ± 4.0 32 ± 6.6

ND ND ND

Banana/tapioca (4 ppb)

B

57 ± 3.5 51 ± 2.5 52 ± 3.2

68 ± 8.1 55 ± 2.5 72 ± 5.9

48 ± 3.8 41 ± 2.9 15 ± 26

ND ND ND

a A: extraction pressure: 2205 psi; extraction time: 15 min (static) and 60 min (dynamic) for 100% CO ; B: extraction pressure: 2500 psi; 2 extraction time: 10 min (static) and 60 min (dynamic) for 100% CO2 . b ND: not detected.

on the published protocol [4,5]. We were unable to achieve quantitative recoveries (>80%) of the spiked pesticides from baby food using the protocol. We then modified the SFE condition in an attempt to improve the recoveries. The modifications included increasing pressure from 2205 to 2500 psi and reducing the restrictor temperature from 100 to 70◦ C. Recoveries of the spiked pesticides in the chicken noodle sample ranged from 0 to 40% using the published protocol, and from 0 to 57% using the modified SFE condition. Better recoveries (0–77%) of the spiked pesticides were obtained from the banana/tapioca samples. Note that the spiked chlorpyrifos were not recovered from either the chicken noodle or the banana/tapioca samples under the SFE conditions tested. Therefore, collection efficiency tests were performed to investigate the loss of chlorpyrifos through SFE–ELISA. The target pesticides were spiked in 5 ml of distilled water and supercritical CO2 was bubbled into the collection vessel to mimic the extraction at 2500 psi/70◦ C for 15 min in the static mode, and followed by 10, 30, or 60 min in dynamic mode. The results showed that quantitative recoveries (>80%) were obtained for carbofuran, atrazine, and metolachlor but not chlorpyrifos. Recoveries of chlorpyrifos decreased from 10 to 2% as the dynamic extraction time increased from 10 to 60 min. We repeated the collection efficiency tests

by using DCM as the collection solvent and analyzing the DCM extracts by GC/MS. Quantitative recoveries (98–114%) of chlorpyrifos were achieved when DCM was used as the collection solvent. The loss of chlorpyrifos was mainly from the depressurizing step when water was used as the solvent. This is partly because chlorpyrifos is not stable in water and may degrade through this step. In addition, chlorpyrifos is more soluble in DCM and CO2 than in water. Thus, significant amounts of chlorpyrifos may be soluble in CO2 and lost at the depressurizing step. Since the other three pesticides are at least 20 times more soluble in water than chlorpyrifos, the loss through the depressurizing step is negligible. Additional tests to evaluate SFE conditions were carried out using DCM as the collection solvent. The recoveries of the spiked pesticides from most baby food samples were less than 50% and reproducible recovery data could not be obtained. We repeated the SFE condition and used sand instead of Hydromatrix to disperse the food, but the results were also unacceptable. In summary, we were unable to achieve quantitative (>80%) and reproducible data of spiked pesticides from wet baby food using the SFE conditions tested. Baby food is typically greater than 80%. The presence of water in baby food may interfere with the removal of pesticides from the baby food under SFE.

J.C. Chuang et al. / Analytica Chimica Acta 399 (1999) 135–142

139

Table 2 Summary of recovery data in dried chicken noodle for SFE–GC/MS SFEa condition

10% 15% 10% 15%

ACN ACN ACN ACN

in in in in

CO2 CO2 CO2 CO2

Recovery (%)

(post-spiked) (post-spiked) (pre-spiked) (pre-spiked)

Carbonfuran

Atrazine

Metolachlor

Chlorpyrifos

85 ± 11 80 ± 14 22 ± 2.1 30 ± 3.5

106 ± 14 96 ± 6.4 64 ± 1.4 65 ± 2.8

115 ± 7.3 100 ± 7.8 74 ± 2.1 71 ± 1.4

103 ± 11 94 ± 8.5 84 ± 5.7 82 ± 2.8

a Extraction pressure: 2500 psi; extraction temperature: 70◦ C; extraction time: 15 min (static) and 60 min (dynamic) for 10% ACN in CO2 and 5 min (static) and 20 min (dynamic) for 15% ACN in CO2 .

Therefore, dried baby food (chicken noodle) was used for the evaluation of SFE conditions. The recovery data of post-spiked and pre-spiked dried chicken noodle sample are summarized in Table 2. Quantitative recoveries (>80%) were obtained in post-spiked dried chicken noodle under these two SFE conditions. The overall method precision was within ±18%. In one condition, 10% ACN in CO2 was used as the extraction fluid, and the extraction time was 15 min in static mode and 60 min in dynamic mode. In condition two, we increased the polarity of the extraction fluid to 15% ACN in CO2 and shortened the extraction time to 5 min in static mode and 20 min in dynamic mode. The addition of matrix modifier, ACN, to the dried chicken noodle sample did not improve the recoveries when 10% ACN in CO2 was used as the extraction fluid. In order to ensure that there was no significant loss of pesticides through freeze-drying, the chicken noodle sample was spiked with known amounts of pesticides prior to the freeze-drying step and tested under these SFE conditions. The results showed that approximately 50% of carbofuran and approximately 10–20% of the other three pesticides were lost through the freeze-drying step. In summary, off-line coupling SFE–GC/MS can be used to determine non-volatile pesticides including atrazine, metolachlor, and chlorpyrifos in dried chicken noodle samples. 3.2. Evaluation of SPE–GC/MS Off-line coupling SPE–GC/MS was evaluated to determine the presence of pesticides in baby food. The recovery data of the spiked pesticides, as summarized in Table 3, are average and standard deviation values of triplicate baby food samples. Quantitative recoveries (>90%) were obtained for all spiked pesticides in both

the applesauce and the banana/tapioca samples. The overall precision for SPE–GC/MS was within ±20% in these two types of baby food. Acceptable recoveries were also observed in the pear sample, which ranged from 69% (carbofuran) to 109% (metolachlor). Quantitative recoveries of atrazine (94%) and metolachlor (78%) and lower recoveries of carbofuran (59%) and chlorpyrifos (32%) were observed in the green bean sample. Both the water-soluble and the residue fractions of fatty baby food (chicken noodle sample) were analyzed for pesticides. The data as reported in Table 3 are the sums of the two fractions. The recoveries of the spiked pesticides in the chicken noodle sample were lower than those in the non-fat baby food (fruits and vegetables). Note that most of the spiked pesticides in the chicken noodle sample were found in the water-soluble fraction, and approximately less than 10% of the overall recovered pesticides were in the residue fraction. 3.3. Evaluation of direct ELISA A direct ELISA on diluted, filtered nonfat baby food was also evaluated. The recovery data of the spiked pesticides, as summarized in Table 4, are the average and standard deviation values of triplicate baby food samples. Duplicate assays were performed on selected baby food samples, and the averages of the duplicate assay results were used for these samples. Four different ELISA test kits were used, one for each spiked pesticide. In general, the overall method precision was within ± 20%. Using the atrazine test kits, quantitative recoveries (>90%) were obtained in most of the tested baby food. Recoveries (>70%) of spiked carbofuran were obtained in the pear, applesauce, and car-

140

J.C. Chuang et al. / Analytica Chimica Acta 399 (1999) 135–142

Table 3 Summary of recovery data for off-line coupling SPE-GC/MS Baby food

Recovery (%)

Pear Apple sauce Carrot Banana/tapioca Green bean Chicken noodlea

Carbofuran

Atrazine

Metolachlor

Chlorpyrifos

69 ± 4.9 94 ± 2.3 53 ± 4.2 100 ± 6.4 59 ± 6.5 28 ± 2.5

92 ± 9.5 117 ± 20 82 ± 16 100 ± 2.8 94 ± 8.1 56 ± 3.2

109 ± 3.2 109 ± 3.2 56 ± 2.0 99 ± 0.7 78 ± 4.7 28 ± 0.6

72 ± 7.6 101 ± 11 58 ± 5.7 138 ± 17 32 ± 17 48 ± 2.5

a Data are from the sums of two fractions (water-soluble and residue) of each sample. The spiked levels were 20 ppb for each spiked pesticide in all the tests.

Table 4 Summary of recovery data of direct ELISA on diluted baby food Baby food

Spike level in food (ppb)

Recovery (%) Atrazine

Carbofuran

Metolachlor

Chlorpyrifos

Pear

1 2 5 10 20

98 ± 5.8 96 ± 8.6 96 ± 7.6 98 ± 7.8 104 ± 9.2

–a 96 ± 7.3 93 ± 4.5 95 ± 7.5 –

– 129 ± 7.0 139 ± 6.0 –

– 77 ± 3.5 86 ± 33 42 ± 10

Apple sauce

2 5 10

143 ± 9.9 118 ± 0.7 –

– 105 ± 22 112 ± 9.4

115 ± 3.5 121 ± 4.2 –

130 ± 14 49 ± 13 93 ± 13

Carrot

2 5 10

101 ± 9.9 121 ± 5.7 –

– 70 ± 12 105 ± 2.5

97 ± 1.4 125 ± 6.4 –

260 ± 11 102 ± 25 54 ± 18

Banana/tapioca

2 5 10

120 ± 5.7 141 ± 7.8 –

100 ± 15 175 ± 35 93 ± 11

101 ± 24 148 ± 11 –

61 ± 11 59 ± 0 65 ± 6.6

Green bean

2 5 10

106 ± 20 90 ± 2.1 102 ± 12

50 ± 39 96 ± 41 112 ± 18

225 ± 7.1 165 ± 21 174 ± 9.7

47 ± 0 25 ± 7.0 27 ± 2.1

a

–Denotes the assay was not performed.

rot samples. Satisfactory recoveries of carbofuran in the banana/tapioca sample were obtained at the 2 ppb and 10 ppb spike levels, but not at 5 ppb. Satisfactory recovery (112 ± 18%) of carbofuran in the green bean samples was only observed at a spike level of 2 and 10 ppb, but not at the 5 ppb. Recoveries of metolachlor in the green bean sample were greater than 100% at all three spike levels (2, 5, and 10 ppb). This is probably due to a matrix interference in the ELISA. With few exceptions (over-recovery), quantitative recoveries (>90%) of metolachlor were obtained in other

baby food. In general, the recovery of chlorpyrifos in fruits and vegetables using direct ELISA was lower than 70% in most tests. Note that ELISA performed on the laboratory-prepared chlorpyrifos standard solution using the calibration curve generated from the standard solutions purchased from SDI gave less than half of the expected value. The recovery data were based on the calculated laboratory-prepared standard solutions, since they were used for the spiking. The same direct ELISA approach was tested on the chicken noodle sample, but the recoveries of the spiked pesti-

J.C. Chuang et al. / Analytica Chimica Acta 399 (1999) 135–142

Fig. 1. Recovery data of atrazine in baby food by ESE–ELISA. The baby food, Bro/Car/Che, denotes broccoli/carrot/cheese.

cides were less than 50%. This finding agreed with the SPE–GC/MS results showing low recoveries of spiked pesticides from the chicken noodle sample. 3.4. Evaluation of ESE–ELISA Off-line coupling ESE–ELISA was also evaluated to determine the presence of pesticides in baby food. Fig. 1 summarizes the recovery data obtained in the evaluation tests using ESE–ELISA. The recovery data were obtained form the average of the triplicate assays. The results showed that the extraction temperature had a significant effect on the recoveries of pesticides from baby food. None of the spiked atrazine were recovered at extraction temperatures of 70 and 100◦ C. The recoveries of atrazine in broccoli/carrots/cheese increased from 12 to 105% as the extraction temperature increased from 120 to 150◦ C. The average of all atrazine recoveries in the spiked pear samples were greater than 80% at extraction temperatures of 120, 140, and 150◦ C. ESE–ELISA performed at a pressure of 2000 psi and extraction temperature of 150◦ C provided the best overall atrazine recoveries (>90%) in all tested baby food, whether non-fat or fatty. In summary, off-line coupling ESE–ELISA can be used to determine atrazine in non-fat and fatty baby food. 4. Conclusions The evaluation of the SFE–ELISA protocol revealed that there was a significant loss of chlorpyrifos from

141

the depressurizing step when water was used as the collection solvent. The collection efficiency for chlorpyrifos was improved to greater than 90% when DCM was used as the collection solvent. In general, recoveries of the four spiked pesticides (atrazine, carbofuran, metolachlor, and chlorpyrifos) from baby food were less than 50% using SFE–ELISA. Additional SFE conditions were evaluated using DCM as the collection solvent. The DCM extracts were concentrated and analyzed by GC/MS instead of ELISA. We can achieve quantitative and reproducible recoveries of the spiked pesticides from dried baby food, but not from baby food as purchased using SFE–GC/MS. Two SFE conditions were developed to provide a quantitative measure of most target pesticides in dried baby food. The baby food was freeze-dried prior to SFE to remove water. The dried baby food was extracted with either 10% ACN in CO2 or 15% ACN in CO2 at 70◦ C and 2500 psi. However, the freeze-drying step caused an approximately 50% loss of carbofuran, a volatile pesticide. Two alternative approaches, SPE–GC/MS and direct ELISA, were developed for determining the presence of target pesticides in non-fat baby food. A C18 SPE column was used to extract pesticides from non-fat baby food, and the eluent of the C18 SPE column was concentrated and analyzed by GC/MS. With few exceptions, acceptable recoveries (>70%) of the spiked pesticides were obtained from non-fat baby food (pears, applesauce, carrots, bananas, tapioca, and green beans), but not from fatty baby food (chicken noodle). The overall method precision for the SPE–GC/MS was within ±20% for most pesticides. Acceptable recoveries (>70%) of some spiked pesticides were obtained by direct ELISA applied to diluted and filtered non-fat baby food. The ELISA–atrazine test kit provided the best result, with overall method accuracy greater than 90% and method precision within ±20%. Acceptable recoveries were obtained for pears, apple sauce, and carrots using the ELISA–carbofuran and ELISA–metolachlor test kits. Of the four ELISA test kits, the ELISA–chlorpyrifos test kit performed the worst. An ESE–ELISA method was also developed to measure target pesticides in both non-fat and fatty baby food. The results revealed that extraction temperature had a significant effect on removing the pesticides from the baby food. The ASE–ELISA

142

J.C. Chuang et al. / Analytica Chimica Acta 399 (1999) 135–142

consisted of extracting the food at 150◦ C and 2000 psi with water and conducting ELISA on the aqueous extract. This provided quantitative determination of atrazine in pear, broccoli/carrot/cheese, and chicken noodle samples. This method offered a potential for a streamlined approach for determining pesticides in both non-fat and fatty food; more evaluation tests are needed to establish a routine analytical method for processing large-scale samples.

Acknowledgements This work was supported by US EPA Contract No. 68-D4-0023 to Battelle Memorial Institute. It has been subjected to the Agency’s peer and administrative review and has been approved for publication as an EPA document. Mention of trade names or commercial products does not constitute endorsement or recommendation for use.

References [1] N.K. Wilson, J.C. Chuang, C. Lyu, Measurements of persistent organic chemicals in several day care centers. Presented at the 1997 annual meeting of the International Society of Exposure Analysis, Research Triangle Park, NC, November 1997. [2] J.C. Chuang, P.J. Callahan, C.W. Lyu, N.K. Wilson, Polycyclic aromatic hydrocarbon exposures of children in low-income families, J. Expos. Anal. Environ. Epidem., in press. [3] J.C. Chuang, C. Lyu, Y.-L. Chou, P.J. Callahan, M. Nishioka, K. Andrews, M.A. Pollard, L. Brackney, C. Hines, D.B. Davis, R. Menton, Evaluation and application of methods for estimating children’s exposure to persistent organic pollutants in multiple media. Final Report, Contract 68-D4-0023, WA 3-06, 1998. [4] V. Lopez-Avila, C. Charan, J.M. Van Emon, Quick determination of pesticides in foods by SFE–ELISA, Food Testing and Analysis 2(5) (1996) 28–37. [5] V. Lopez-Avila, C. Charan, J. M. Van Emon, SFE–ELISA applications for determination of pesticides in various matrices, in: R.C. Bier, L. Stanker (Eds.), Immunochemical Detection of Residues in Foods, ACS Symposium Series No. 621, Washington, DC, 1996b, pp. 439–449.