Analytical and Bioanalytical Chemistry

, Volume 391, Issue 4, pp 1397–1408

Results of a European inter-laboratory comparison study on the determination of EU priority polycyclic aromatic hydrocarbons (PAHs) in edible vegetable oils

Authors

    • European Commission, Directorate General Joint Research CentreInstitute for Reference Materials and Measurements
  • José Angel Gomez Ruiz
    • European Commission, Directorate General Joint Research CentreInstitute for Reference Materials and Measurements
  • Christoph von Holst
    • European Commission, Directorate General Joint Research CentreInstitute for Reference Materials and Measurements
  • Thomas Wenzl
    • European Commission, Directorate General Joint Research CentreInstitute for Reference Materials and Measurements
  • Elke Anklam
    • European Commission, Directorate General Joint Research CentreInstitute for Health and Consumer Protection
Original Paper

DOI: 10.1007/s00216-007-1771-4

Cite this article as:
Simon, R., Gomez Ruiz, J.A., von Holst, C. et al. Anal Bioanal Chem (2008) 391: 1397. doi:10.1007/s00216-007-1771-4

Abstract

A collaborative study on the analysis for 15 + 1 EU priority PAHs in edible oils was organised to investigate the state-of-the-art of respective analytical methods. Three spiked vegetable oils, one contaminated native sunflower oil, and one standard solution were investigated in this study. The results of 52 laboratories using either high-performance liquid chromatography with fluorescence detection or gas chromatography with mass-selective detectors were evaluated by application of robust statistics. About 95% of the laboratories were able to quantify benzo[a]pyrene together with five other PAHs included in the commonly known list of 16 US-EPA PAHs. About 80% of the participants also quantified seven additional PAHs in most samples, two of which were benzo[b]fluoranthene and benzo[k]fluoranthene, which were also known from the EPA list. Only about 50% of the participants quantified cyclopenta[cd]pyrene, benzo[j]fluoranthene, and benzo[c]fluorene. The robust relative standard deviations of the submitted results without discrimination between the methods applied ranged between 100% for 5-methylchrysene in spiked olive oil and 11% for the same analyte in spiked sunflower oil. The results clearly showed that for these analytes the methods of analysis are not yet well established in European laboratories, and more collaborative trials are needed to promote further development and to improve the performances of the respective methods.

Keywords

Polycyclic aromatic hydrocarbons (PAHs)European priority PAHsEdible vegetable oilsInter-laboratory comparison study

Introduction

Polycyclic aromatic hydrocarbons (PAHs) constitute a large class of organic substances containing two or more fused aromatic rings made up of carbon and hydrogen atoms. Hundreds of individual PAHs may be formed and released during the incomplete combustion of organic carbon-containing materials such as wood-, coal-, and mineral-oil-derived fuels. In the 1970s, the US Environmental Protection Agency created a list of PAHs most frequently found in environmental samples, on which research should focus. The PAHs on this list are commonly known as the 16 EPA-PAHs (EPA-16). Eight of these PAHs (EPA-8) are known to be mutagenic and/or carcinogenic and thus give rise to serious health concerns [1]. Benzo[a]pyrene (BaP) was the first PAH to be identified as a carcinogen and has therefore been studied most often. It has often been used as a marker for PAH contamination in general, although it contributes only 1–20% of the total carcinogenicity found in samples from the environment [2].

In 2002 the European Commission’s Scientific Committee on Food (SCF) assessed the toxicity of 33 PAHs, confirmed the subset of EPA-8 from the EPA-16, and identified seven additional PAHs as being of major concern to human health. These 15 EU priority PAHs should be monitored in food to enable long-term exposure assessments and to verify the validity of the use of the concentrations of benzo[a]pyrene (BaP) as a marker for a “content of carcinogenic PAHs in food” [3]. The toxicological importance of these compounds was confirmed in October 2005 by the International Agency for Research on Cancer (IARC), which assessed the potential carcinogenicity of 60 PAHs and (re)classified BaP as a carcinogen to human beings (IARC group 1), cyclopenta[cd]pyrene, dibenz[a,h]anthracene, and dibenzo[a,l]pyrene as probably carcinogenic to human beings (group 2a), and additionally nine of the 15 EU priority PAHs as possibly carcinogenic to human beings [4]. The European Commission (EC) also reacted in 2005 by issuing Recommendation 2005/108/EC for the analysis of the 15 EU priority PAHs in food and by relating data for specific food processing practices [5]. As a consequence of this development, the European Food Safety Authority (EFSA) asked the EU Member States for further investigations into the 15 EU priority PAHs together with one additional PAH (benzo[c]fluorene), as the analysis of it in food “may help to inform future evaluations” according to the outcome of an evaluation undertaken by the Joint FAO/WHO Expert Committee on Food Additives (JECFA) in 2005 (15 + 1 EU priority PAHs) [6]. The areas for investigation should especially cover the food categories for which a maximum permitted BaP concentration was given in Commission Regulation (EC) No 208/2005, which is now being replaced by Commission Regulation (EC) No 1881/2006 [7, 8]. A European workshop on analytical methodologies held in 2005 showed that many EU Member States’ laboratories were not fully prepared for the analysis of the recommended group of 15 + 1 EU priority PAHs in food products, and that there was an urgent need to solve this problem [9]. There is still a lack of available validated analytical methods covering all of the PAHs as required (Table 1).
Table 1

15 + 1 EU priority PAHs under investigation (with acronyms)

https://static-content.springer.com/image/art%3A10.1007%2Fs00216-007-1771-4/MediaObjects/216_2007_1771_Tab1_HTML.gif

The main principle used for the determination of PAHs is based on gas chromatography with mass spectrometric detection (GC–MS) and high-performance liquid chromatography with fluorescence detection (HPLC–FL). Most established methods were developed and validated for the analysis of BaP [9]. As the lack of official methods has been noted by the scientific community, ISO methods are currently in preparation for the determination of an enlarged set of PAHs in edible oils and fats, and in one case for drinking, ground, and surface water, even including all 15 EU priority PAHs [10, 11].

Only the results of one collaborative study on the validation of the performance characteristics of a method for the determination of the 15 EU priority PAHs in smoke flavourings have been published so far [12]. Proficiency testing schemes such as FAPAS® have only recently focussed on the determination of the 15 + 1 EU priority PAHs [13].

A collaborative study on the analysis for PAHs in edible oils was organised in order to investigate the actual state-of-the-art of analytical methods favoured for use in official food control laboratories, whereas a grading of the individual laboratories' performance was considered less important and no proficiency scores were calculated. Nevertheless, the study was mostly designed along the lines of the harmonised protocol for proficiency testing [14].

Participating laboratories were asked to report their results uncorrected for recovery and to also deliver the recovery values. The set of samples was composed of three spiked vegetable oils (materials 1–3) and one unspiked crude sunflower oil (later referred to as native sunflower oil, material 4). A standard solution containing all PAHs under investigation in 2-propanol (material 5) was also included in the study. This article summarises the methods used for sample preparation, clean-up, and quantitation, presents the overall outcome of the study, and explains the presentation of the raw results by using the examples of BaP and DlP. The entire dataset can be found as Electronic Supplementary Material, sorted by material and analyte.

Experimental

Reagents

The activated charcoal, n-hexane, 2-propanol, acetonitrile, and ethyl acetate used were all of analytical grade or higher (VWR, Darmstadt, Germany). The analytes benz[a]anthracene (BaA) CAS# 56-55-3, chrysene (CHR) CAS# 218-01-9, 5-methylchrysene (5MC) CAS# 3697-24-3, benzo[b]fluoranthene (BbF) CAS# 205-99-2, benzo[j]fluoranthene (BjF) CAS# 205-82-3, benzo[k]fluoranthene (BkF) CAS# 207-08-9, benzo[a]pyrene (BaP) CAS# 50-32-8, indeno[1,2,3-cd]pyrene (IcP) CAS# 193-39-5, dibenz[a,h]anthracene (DhA) CAS# 53-70-3, benzo[ghi]perylene (BgP) CAS# 191-24-2, dibenzo[a,l]pyrene (DlP) CAS# 191-30-0, dibenzo[a,e]pyrene (DeP) CAS# 192-65-4, dibenzo[a,i]pyrene (DiP) CAS# 189-55-9 and dibenzo[a,h]pyrene (DhP) CAS# 189-64-0 were commercially available BCR reference materials (IRMM, Geel, Belgium). Cyclopenta[cd]pyrene (CPP) CAS# 27208-37-3, purity >99.0 % by GC, was manufactured on request (Biochemical Institute for Environmental Carcinogens, Prof. Dr. Gernot Grimmer-Foundation, Großhansdorf, Germany). Benzo[c]fluorene (BcL) CAS# 205-12-9 was purchased from Dr. Ehrenstorfer GmbH (Augsburg, Germany).

Test materials

Three different edible oils, in particular sunflower, maize, and olive oil, were purchased at a local retailer store. To eliminate possible traces of PAHs, five litres of each oil were treated with 200 g of activated charcoal overnight and were vacuum-filtered carefully at 40 °C through 0.45-μm cellulose nitrate membrane filters (Whatman International Ltd, Maidstone, England) to remove remaining traces of the adsorbent. Subsequently a weighed amount of approximately 20 ml of a solution with the 15 + 1 EU priority PAHs in n-hexane was added to each of the treated oils (materials 1–3), vigorously mixed overnight, and aliquots were filled into vials to be sent to the attending laboratories. In addition, native sunflower oil kindly provided by the European Oil and Proteinmeal Industry (FEDIOL) was used (material 4); the concentrations of the PAHs (if present at all) in this oil were unknown. To investigate the influence of instrumental calibration on the results, a standard solution of the 15 + 1 PAHs in 2-propanol was prepared (material 5). The concentrations of the analytes were about ten times higher in the standard solution than in the oils to account for the usual enrichment of analyte during sample preparation and clean-up. An overview of the spiked concentrations in the respective materials is given in Table 2.
Table 2

Analyte concentrations in the three spiked vegetable oils and the solution distributed as test materials

Matrix

Vegetable oil

Solvent

Analyte

Material 1 (ng/g)

Material 2 (ng/g)

Material 3 (ng/g)

Material 5 (ng/ml)

BaA

0.8

2.4

8.4

9.5

BaP

0.8

2.4

8.4

4.9

BbF

7.4

0.8

2.0

11.9

BcL

1.8

7.2

0.7

19.3

BgP

11.0

1.2

3.0

24.0

BjF

3.0

12.0

1.1

33.4

BkF

0.9

2.6

9.3

11.2

CHR

0.9

9.7

2.5

7.7

CPP

9.7

1.0

2.6

23.1

DeP

2.0

0.8

7.6

16.0

DhA

1.1

3.2

11.4

11.1

DhP

2.4

1.0

9.3

27.9

DiP

0.8

8.3

2.1

16.0

DlP

8.3

2.4

0.8

11.0

IcP

2.1

8.4

0.8

12.9

5MC

1.1

0.4

4.2

6.6

Homogeneity testing

The test materials were all liquids and as such were considered to be real solutions, which, by definition, are homogeneous. The materials were therefore not subjected to testing for homogeneity.

Design of the inter-laboratory comparison study

Participating laboratories

Laboratories were invited for participation either directly or via announcement of the study on the website of the organiser. As a result, 68 laboratories from 22 different countries confirmed their interest in participation. They were asked to use their own in-house established methods and to analyse as many of the EU 15 + 1 priority PAHs as possible in one sample of each material in duplicate. Laboratories were from official and private food control, industry and research organisations (academia). At the time of the study most laboratories were located in the EU: 65 laboratories in 19 countries of the 27 EU Member States, one in Turkey, one in Canada, and one in Malaysia.

Statistical evaluation of results

After calculating the average of the values from the duplicate analysis, the mean values and the standard deviations of the submitted results were estimated by the application of robust statistics. This statistical technique has the advantage that the detection and rejection of outliers is not necessary, because the impact of extreme values on the average and the standard deviation is down-weighted. Kernel density plotting was used to identify modes of the distribution. This technique replaces every data point with a normal distribution, the kernel. The density plot is then generated by adding all kernels [15]. The robust mean values, robust standard deviations, and the kernel density plots were computed using MS Excel® macros written by the Analytical Methods Committee (AMC) of The Royal Society of Chemistry (RSC) [16].

Results and discussion

Of the 68 registered participants 52 reported results back. Two types of methods were used for analysis: GC–MS and HPLC–FL. The latter method was slightly dominant: 28 laboratories used HPLC-FL and 22 applied GC/MS (Table 3). The preference for HPLC was even more pronounced in official food control laboratories (24 reporting out of 29): nine laboratories used GC–MS and 15 laboratories preferred HPLC. For sample preparation simple dissolution/dilution (28) of the oil was preferred over saponification (13) and liquid–liquid partitioning (7). Saponification was mostly combined with (subsequent) liquid–liquid partitioning (8). The favourite clean-up method was solid-phase extraction (23) followed by gel-permeation chromatography (11) and donor-acceptor column chromatography (6). In two cases gel-permeation chromatography and solid-phase extraction were combined. Most of the laboratories reported recoveries (33). Several participants used internal standards to correct directly for recovery (14). Ten out of 22 laboratories using GC–MS applied isotopically labelled internal standards. Using HPLC–FL isotopically labelled compounds cannot be distinguished by the detector. Therefore laboratories with HPLC–FL and internal standardisation used benzo[b]chrysene instead (although deuterated compounds might be separated chromatographically from native ones) (4). The majority of the laboratories were accredited according to ISO 17025 (36), while only 18 laboratories used an accredited method for the determination of PAHs. However, it was unclear which PAHs the methods were accredited for. It was found to be reasonable to assume that most laboratories used an accredited method for only a couple of selected PAHs or just one PAH, most probably BaP. Three laboratories made a specific statement in this respect, confirming the latter. One laboratory was using an accredited method for EPA-16 and another one for EPA-8 (Table 3).
Table 3

Overview of the methods used for sample preparation, clean-up, and final measurement

Lab No.

Accredited (PAH in oil)

Sample preparation

Clean-up

Internal standard

Quantification method

LAB 01

No

Dissolution in cyclohexane/ethyl acetate

GPC+SPE (silica gel)

Yes

GC–HRMS

LAB 02

Yes

Dissolution in cyclohexane/ethyl acetate

GPC+SPE (aluminium oxide)

Yes

GC–HRMS

LAB 03

Yes

Dissolution in dichloromethane

GPC

No

HPLC–FL

LAB 04

No

Liquid–liquid partitioning with DMSO and cyclohexane

TLC

No

GC–MS

LAB 06

Yes

Dissolution in 2-propanol

DACC

No

HPLC–FL

LAB 07

Yes

Dissolution in 2-propanol

DACC

No

HPLC–FL

LAB 08

Yes

Dissolution in cyclohexane/ethyl acetate

GPC

No

GC–MS

LAB 09

Yes

Saponification and liquid/liquid extraction with cyclohexane

SPE (aluminium oxide)

No

GC–MS

LAB 10

Yes

Liquid–liquid extraction with acetonitrile/acetone

SPE (C18 + silica gel)

No

HPLC–FL

LAB 12

No

Saponification and liquid–liquid extraction with cyclohexane

SPE (silica gel)

No

HPLC–FL

LAB 13

Yes

Dissolution in cyclohexane

SPE (silica gel)

Yes

GC–MS

LAB 14

Yes

Saponification and liquid–liquid extraction with cyclohexane

SPE (silica gel)

No

GC–MS

LAB 15

No

Liquid–liquid extraction with acetonitrile/acetone

SPE (C18 + silica gel)

No

HPLC–-FL

LAB 16

Yes

Dissolution in acetonitrile/acetone

SPE (silica gel & Florisil)

No

HPLC–FL

LAB 17

Yes

Saponification and dissolution in cyclohexane

Liquid–liquid partitioning

Yes

GC–MS

LAB 18

Yes

Saponification and liquid/liquid partitioning with cyclohexane and DMF

SPE (silica gel)

No

HPLC–FL

LAB 20

No

Saponification and liquid/liquid partitioning with cyclohexane and DMF

SPE (silica gel)

No

GC–MS

LAB 21

Yes

Dissolution in cyclohexane and liquid–liquid partitioning with methanol/water

Liquid–liquid partitioning & SPE (silica gel)

Yes

GC–MS

LAB 22

Yes

Saponification and liquid–liquid extraction with cyclohexane

SPE (silica gel)

No

GC–MS

LAB 23

Yes

Dissolution in isohexane/butyl methyl ether

SPE (polystyrene)

Yes

HPLC–FL

LAB 24

Yes

Saponification and liquid–liquid partitioning with methanol/water

SPE (silica gel)

No

GC––MS

LAB 25

Yes

None

GPC

No

HPLC–FL

LAB 26

Yes

Saponification and liquid–liquid extraction with cyclohexane

SPE (silica gel)

No

HPLC–FL

LAB 27

Yes

Dissolution in cyclohexane/ethyl acetate

GPC+SPE (silica gel)

Yes

GC–MS

LAB 28

Yes

Dissolution in cyclohexane

SPE

Yes

GC–MS and GC–HRMS

LAB 29

No

Dissolution in n-hexane

SPE (silica gel & NH2)

No

HPLC–FL

LAB 31

No

Dissolution in cyclohexane/isooctane

SPE (SAX & C18)

No

HPLC–FL

LAB 32

No

Dissolution in n-hexane

SPE (silica gel)

No

HPLC–FL

Lab No.

Accredited

Sample preparation

Clean-up

Internal standard

Quantification method

LAB 33

Yes

Dissolution in cyclohexane/ethyl acetate

GPC

No

HPLC–FL

LAB 34

Yes

Liquid–liquid extraction with acetonitrile/acetone

SPE (C18 + silica gel)

Yes

HPLC–FL

LAB 35

Yes

Dissolution in ethyl acetate/cyclohexane

GPC

No

GC–MS

LAB 36

Yes

Dissolution in dichloromethane

GPC

No

HPLC–FL

LAB 39

Yes

Liquid/liquid extraction with acetonitrile/acetone

SPE (C18 + silica gel)

No

HPLC–FL

LAB 43

Yes

Dissolution in ethyl acetate/cyclohexane

GPC and SPE (silica gel)

No

GC–MS

LAB 44

Yes

Saponification and liquid/liquid extraction with cyclohexane

Multilayer LC and SPE (silica gel)

Yes

GC–HRMS

LAB 45

Yes

Saponification and liquid/liquid extraction with cyclohexane

SPE

No

GC–MS

LAB 47

Yes

Saponification and liquid/liquid extraction with n-hexane and washing with ethanol/water

GPC

No

GC–MS

LAB 48

No

Dissolution in 2-propanol

DACC

No

HPLC–FL

LAB 49

Yes

Dissolution in n-hexane (ISO 15302)

SPE (aluminium oxide)

No

HPLC–FL

LAB 50

Yes

Liquid–liquid partitioning with THF/water and cyclohexane, saponification

Liquid–liquid extraction with cyclohexane and SPE (silica gel)

Yes

GC–HRMS

LAB 52

Yes

Dissolution in cyclohexane/ethyl acetate

GPC

No

HPLC–FL

LAB 53

Yes

Dissolution in n-hexane

Silica gel

No

GC–MS

LAB 54

No

Saponification and liquid/liquid partitioning with iso-octane

No

No

HPLC–FL

LAB 56

Yes

Dissolution in 2-propanol or iso-butanol

DACC

No

HPLC–FL

LAB 58

No

Dissolution in 2-propanol

DACC

Yes

HPLC–FL

LAB 59

Yes

Dissolution in 2-propanol

DACC

No

HPLC–FL

LAB 60

No

Liquid/liquid partitioning with DMSO

SPE

Yes

GC–MS

LAB 61

Yes

Saponification and liquid/liquid extraction with n-hexane

SPE (silica gel) + GPC

No

GC–MS

LAB 62

No

Dissolution in iso-octane / cyclohexane

SPE (BOND ELUT-PPL)

Yes

HPLC–-FL

LAB 64

No

Dissolution in n-hexane

SPE (silica gel)

No

HPLC–FL

LAB 65

Yes

Dissolution in dichloromethane

SPE (aluminium oxide)

No

GC–MS

LAB 67

No

Dissolution in 2-propanol

DACC

No

HPLC–FL

Glossary:

DACC, donor acceptor column chromatography

DMF, dimethyl formamide

DMSO, dimethyl sulfoxide

GC–HRMS, gas chromatography–high resolution mass spectrometry

GC–MS, gas chromatography–mass spectrometry

GPC, gel permeation (or: size exclusion) chromatography

HPLC-FL, high––performance liquid chromatography with fluorescence detection

ISO, International Organisation for Standardization

PAH, polycyclic aromatic hydrocarbon(s)

SPE, solid–phase extraction

TLC, thin-layer chromatography

* = BaP only, x = EPA 16, + =EPA 8

The number of reported concentration values varied for the different materials and analytes. In the cases of indeno[1,2,3-cd]pyrene (IcP), benz[a]anthracene (BaA) and dibenz[a,h]anthracene (DhA), a trend for a decreasing number of reported values with decreasing concentrations of the respective analyte could be observed. This effect was probably linked to the performances of the methods applied in the respective laboratories, as described by the limits of detection and limits of quantification. Consequently, a given analyte was not detected in one material at the lower concentration level(s), while it could be quantified in materials with higher concentration levels. However, this trend was weak and was not observed for all analytes (Fig. 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs00216-007-1771-4/MediaObjects/216_2007_1771_Fig1_HTML.gif
Fig. 1

Number of reported values for concentration for all analyte–matrix combinations sorted by the median value (* marks EPA-8 analytes)

The numbers of reported concentration values for a specific analyte were mostly consistent across the materials. After sorting the analytes by decreasing numbers of values reported, a trend was found (Fig. 1). Most data were reported by the laboratories for BaP followed by the other seven analytes included in the group “EPA-8”. BaP was found by nearly all participants in all materials (49) and quantified in almost all analyses (>90%), even in the samples with a BaP content of 0.8 μg/kg, which was below the required limit of quantitation according to Commission Directive 2005/10/EC, laying down the methods of analysis for the official control of the levels of benzo[a]pyrene in foodstuffs [17]. Directly following in decreasing order were chrysene (CHR), benzo[ghi]perylene (BgP), IcP, BaA, and DhA, all five included in EPA-8 (48–47). This outcome was to be expected, as the analytes included in EPA-8, a subset of the EPA-16, were known to the analytical community for quite some time. Less often reported were the contents of benzo[b]fluoranthene (BbF) and benzo[k]fluoranthene (BkF), the two remaining members of EPA-8. This was not surprising either, as difficulties with the gas-chromatographic separation of the two benzofluoranthenes in presence of benzo[j]fluoranthene (BjF) were already documented in the literature [18]. Most laboratories using HPLC (20) were not able to determine cyclopenta[cd]pyrene (CPP) due to the fact that the compound does not give rise to fluorescence and can only be quantified with UV absorption detection. Concerning BjF, 22 laboratories did not report results. This could either be due to its weak fluorescence (HPLC–FL) or a lack of chromatographic separation (GC–MS) [19].

Several groups of analytes were generated to assess the state of the art in analytical laboratories with respect to the requirements of legislation [bearing in mind that all 16 analytes had been added to all materials except for the native sunflower oil (material 4)].

It was then necessary to count (Table 4) how many laboratories were able to
  • quantify all 15 + 1 EU priority PAHs

  • quantify the 15 SCF PAHs (no BcL)

  • had problems with CCP and/or the benzofluoranthenes

  • analyse the EPA-8 (amongst some others)

  • determine any remaining combination of analytes (rest).

Table 4

Number of reported groups of analytes

 

EU 15 + 1

15 SCF

No CPP, BbF, BjF, and/or BkF

EPA-8

The rest

Recovery

14

2

8

4

22

Material 1

7

3

8

3

29

Material 2

8

2

10

3

27

Material 3

8

3

6

3

29

Material 4

4

0

2

2

42

Material 5

22

6

6

0

16

It appeared that in 49 laboratory/sample combinations (∼20%), all 15 + 1 and in 14 further cases (6%), the 15 SCF priority analytes were found. In 32 cases (13%) up to four and in 11 cases (4%) up to eight priority analytes were missing, but the well-known set of the EPA-8 was still quantified. In 143 cases (57%) even the subset of the EPA-8 was not found completely (Table 4). These data show that only about 26% of the data from 13 laboratories were in line with Recommendation 2005/108/EC [5].

Despite the fact that all laboratories were requested to report the recovery values for each of the PAHs analysed, recoveries were only specified by 45 participants (Table 4). This could partially be attributed to the fact that the analyst automatically corrects for recovery by using internal standards and thus does not necessarily calculate the recovery. Thus several participants, who were correcting for recovery through the use of internal standards, did not report a value for recovery and were not documented for that. Those data were evaluated as “corrected for recovery”. The respective median values of the uncorrected and the recovery-corrected individual results for materials 1–4 were taken as the assigned values in the figures of the Electronic Supplementary Material, where the complete information is given and all results reported are presented. Two specific examples were selected and are shown below for illustration.

Detailed results for BaP and dibenzo[a,l]pyrene (DlP) are given in Figs. 2 and 3. Here the data were grouped by analyte/material combination corrected for recovery. The comparison of the two examples not only confirmed that more laboratories were able to analyse for BaP than for DlP but also demonstrated that the results were more scattered for DlP than for BaP. The kernel density plots showed that most data belonged to one distribution, whereas some of the data could be considered as outliers.
https://static-content.springer.com/image/art%3A10.1007%2Fs00216-007-1771-4/MediaObjects/216_2007_1771_Fig2_HTML.gif
Fig. 2

Analytical results for benzo[a]pyrene in material 1 (a) with kernel density plot (b)

https://static-content.springer.com/image/art%3A10.1007%2Fs00216-007-1771-4/MediaObjects/216_2007_1771_Fig3_HTML.gif
Fig. 3

Analytical results for dibenzo[a,l]pyrene in material 3 (a) and kernel density plot (b)

The robust relative standard deviations (RSDR) of the submitted results without discrimination between the methods applied varied between the extremes of 100% for dibenzo[a,l]pyrene in spiked olive oil and 13% for benzo[a]pyrene in spiked maize oil. Table 5 summarises the values found for the respective robust estimate of the mean (med), robust estimate of the standard deviation (median absolute deviation, MADe), and RSDR.
Table 5

Median (med)+, robust standard deviation (MADe)+, and robust relative standard deviation (RSDR as [%])

 

Med

MADe

RSDR

Med

MADe

RSDR

Uncorrected

Recovery corrected

Material 1 (spiked sunflower oil)+

BaA

0.9

0.4

44

0.9

0.3

33

BaP

0.8

0.2

25

0.9

0.2

22

BbF

6.1

1.6

26

6.6

1.4

21

BcL

1.4

0.3

21

1.5

0.3

20

BgP

9.4

2.2

23

9.5

2.5

26

BjF

2.6

0.6

23

2.7

0.7

26

BkF

0.8

0.3

38

0.8

0.3

38

CHR

1.1

0.3

27

1.0

0.4

40

CPP

6.2

1.9

31

6.3

2.1

33

DeP

1.6

0.4

25

1.7

0.4

24

DhA

1.0

0.3

30

1.0

0.3

30

DhP

1.6

0.7

44

1.8

0.7

39

DiP

0.6

0.3

50

0.6

0.3

50

DlP

7.3

2.3

32

7.5

1.8

24

IcP

1.7

0.3

18

1.8

0.4

22

5MC

0.9

0.1

11

0.9

0.2

22

Material 2 (spiked olive oil)+

BaA

2.8

1.1

39

2.9

1.0

34

BaP

2.2

0.4

18

2.3

0.9

39

BbF

1.3

0.9

69

1.1

0.5

45

BcL

5.9

2.5

42

6.9

2.5

36

BgP

1.5

0.5

33

1.5

0.6

40

BjF

10

2.3

23

10

3.0

30

BkF

2.2

0.8

36

2.3

0.8

35

CHR

9.5

1.9

20

9.9

2.6

26

CPP

1.2

0.5

42

1.1

0.5

45

DeP

0.6

0.2

33

0.8

0.3

38

DhA

2.7

1.0

37

2.9

0.8

28

DhP

0.9

0.8

89

0.8

0.4

50

DiP

6

2.9

48

6.8

2.0

29

DlP

2.1

1.2

57

2.1

0.8

38

IcP

6.6

1.3

20

6.9

1.2

17

5MC

0.8

0.8

100

0.8

0.7

88

Material 3 (spiked maize oil)+

BaA

7.5

1.1

15

7.8

1.5

19

BaP

7.2

1.3

18

7.5

1

13

BbF

1.9

0.3

16

1.9

0.5

26

BcL

0.6

0.2

33

0.7

0.2

29

BgP

2.6

0.7

27

2.8

0.5

18

BjF

1.2

0.3

25

1.1

0.4

36

BkF

7.2

1.3

18

7.4

1.4

19

CHR

2.3

0.6

26

2.3

0.5

22

CPP

2.1

0.9

43

1.8

0.6

33

DeP

5.6

1.9

34

6.8

1.5

22

DhA

10

2.1

21

11

2.5

23

DhP

6.4

3.3

52

6.8

2.2

32

DiP

1.8

0.8

44

1.9

0.7

37

DlP

0.9

0.4

44

1

0.5

50

IcP

0.8

0.2

25

0.8

0.2

25

5MC

3.2

0.7

22

3.5

0.5

14

Material 4 (native sunflower oil)+

BaA

5.1

1.3

25

5.7

2.1

37

BaP

4.4

0.8

18

4.7

0.9

19

BbF

4.4

1.0

23

4.6

0.6

13

BcL

1.5

1.1

73

1.5

1.1

73

BgP

3.2

1.0

31

3.6

0.8

22

BjF

2.6

1.2

46

2.6

0.5

19

BkF

2.3

0.4

17

2.4

0.4

17

CHR

6.8

1.7

25

7.5

1.9

25

CPP

2.5

0.7

28

2.6

0.5

19

DeP

0.9

0.7

78

0.8

0.5

63

DhA

0.6

0.3

50

0.8

0.4

50

DhP

n.a.*

n.a.*

n.a.*

n.a.*

n.a.*

n.a.*

DiP

n.a.*

n.a.*

n.a.*

0.3

0.2

67

DlP

0.5

0.4

80

0.6

0.6

100

IcP

2.9

0.7

24

3.2

0.7

22

5MC

n.a.*

n.a.*

n.a.*

n.a.*

n.a.*

n.a.*

Material 5 (spiked solvent)+

BaA

9.5

1.0

11

   

CPP

23.4

4.9

21

   

BaP

5

0.6

12

   

DeP

14.8

2.5

17

   

BbF

11.7

1.3

11

   

DhA

11.7

2.1

18

   

BcL

19.3

3.6

19

   

DhP

24.4

7.0

29

   

BgP

21.5

4.7

22

   

DiP

18.2

3.1

17

   

BjF

32.5

4.0

12

   

DlP

10.5

1.4

13

   

BkF

11.2

1.6

14

   

IcP

12.7

2.1

17

   

CHR

8.1

1.3

16

   

5MC

6.7

0.8

12

   

* n.a. = not applicable, because no value was assigned

+ in ng/g for the materials 1–4 and ng/ml for material 5

When comparing the results for BaP delivered by the two different methods used, no significant trend can be observed (Fig. 4).
https://static-content.springer.com/image/art%3A10.1007%2Fs00216-007-1771-4/MediaObjects/216_2007_1771_Fig4_HTML.gif
Fig. 4a–e

Results for benzo[a]pyrene in edible oils corrected for recovery (ad) and solvent (e) by analytical method and measurement value (* = use of internal standard, MADe = median absolute deviation)

In order to analyse the distribution of the individual results, to search for tendencies and potential bias, the differences of all individual analytical values from the respective assigned value were plotted sorted by laboratory (Fig. 5). It appears that there is room for improvement at all of the laboratories. The comparison with the reduced dataset acquired from the analysis of the solvent solution shows that some of the data scattering can be attributed to the calibration of the instruments, but that the largest contribution still comes from sample preparation. Similar diagrams to those in Fig. 5 have been used to search for differences in results related to chromatographic methods, sample preparation techniques, sample clean-up, methods of calibration (internal standardisation vs. calibration with external standard), and accreditation status of the laboratories (data not shown). No correlations could be found.
https://static-content.springer.com/image/art%3A10.1007%2Fs00216-007-1771-4/MediaObjects/216_2007_1771_Fig5_HTML.gif
Fig. 5

Overview of all results (* = internal standard used)

Conclusions

The results of this collaborative trial showed that most participants have experiences with the analysis of PAHs in edible oils. However, several weaknesses were detected, demanding further development on three levels: the analytes BcL, CPP, and BjF were covered by only half of the laboratories and can be considered to be the most difficult ones to analyse. Major improvements in analytical capacities will be needed in this aspect. Seven of the analytes (BbF, DlP, BkF, DeP, 5MC, DiP, DhP) were analysed by about 80% of the laboratories with a robust estimate of RSDR of 13–100%, and with some effort this could be adapted to suit. For six of the analytes (BaP, CHR, BgP, IcP, BaA, DhA), the expertise present in the laboratories (>90% reported values, robust RSDR of 13–40%) can be considered to be at a high level and some refining work is all that is required to lower the uncertainty of the values reported. Overall, the robust relative standard deviations varied between 13 and 100%, which translates to 0.6–4.5 times the truncated Horwitz relative standard deviation of 22% for results from collaborative trials [20]. These findings underline the need for continued collaborative testing to promote the development and refining of analytical methods for the analysis of PAHs in edible oils.

Acknowledgements

We thank the following institutions for participation in this study:

Bundesforschungsanstalt für Ernährung und Lebensmittel, DE; RIKILT-Instituut voor Voedselveiligheid, NL; Agencia Española de Seguridad Alimentaria, ES; Instituto Superiore di Sanità, IT; Unilever Hellas, GR; Chemical Laboratory Dr. A. Verwey, NL; Food and Consumer Product Safety Authority, NL; Eurofins, DE; Environmental Research Center NV, BE; ITERG-Expertise corps gras, FR; Barry Calletsaut France, FR; Chemisches und Veterinäruntersuchungsamt Freiburg, Freiburg, DE; Biochemisches Institut für Umweltcarcinogene, DE; Central Science Laboratory, UK; Laboratoire interrégional de la Direction Générale de la Concurrence, de la Consommation et de la Répression des Fraudes, FR; Chemisches Landes-und Staatliches Veterinäruntersuchungsamt Münster, DE; Chemisches und Veterinäruntersuchungsamt Karlsruhe, DE; Landeslabor Brandenburg, DE; Karlshamns AB, SE; Agrifood Research, FI; National Food Administration, SE; Chemisches und Veterinäruntersuchungsamt Stuttgart, DE; Finnish Customs Laboratory, FI; State Veterinary and Food Institute, SK; The Dublin Public Analyst's Lab, IE; Danish Institute for Food and Veterinary Research, DK; LABERCA-ENVN, FR; Instituto de la Grassa, IT; Plant Protection and Soil Conservation Service of Somogy County, HU; Unilever Italia SPA Division Foods, IT; University of Udine, IT; Warsaw Agriculture University, PL; General Chemical State Laboratory, GR; Health Protection Inspectorate, EE; GSF-Institut für Ökologische Chemie, DE; LCBA-DISCAFF, IT; Plant Protection and Soil Conservation Service of Jasz-Nagykun-Szolnok County, HU; National Institute of Hygiene, PL; VITUKI Institute for Environment Protection and Water Management, HU; Institute of Public Health, BE; Central Laboratory of Veterinary and Food Control, Station of Borsod-Abaúj-Zemplén County, HU; Regional lab of Aarhus, DE; VITO, BE; Institute of National Public Health and Medical Offer Service Gyor-Moson-Sopron County, HU; NBC Defence and Ecology Scientific Research Center, RO; Laboratorio Arbitral Agroalimentario del M.A.P.A., ES; Bunge Europe R&D Centre, HU; National Public Health and Medical Officer Service, HU; Unilever Safety & Environmental Assurance Centre, UK; FAVV-AFSCA FLVVT, BE; Agency for Health and Food Safety GmbH, AT; Galab Laboratories, DE; National Center of Public Health Protection, BG; ADM Oelmühle Hamburg AG, DE; SGS Nederland B.V., NE; Cargill ROE, NL; ADM Noblee & Thoerl GmbH, DE; Food and Consumer Product Safety Authority, NL; Gobierno Vasco y Universidad del Pais Vasco, ES; Institut Dr. Wagner, AT; Stazione Sperimentale Oli Grassi, IT; Laboratorio de Salud Publica de Valencia, ES; EGE GIBA VE END ANALIZ LAB LTB STI (EGEANALIZ), TR; MicroPollutants Technologie, FR; University of Liege, BE; Malaysian Palm Oil Board, MY; FEDIOL, EU.

Supplementary material

216_2007_1771_MOESM1_ESM.pdf (1.4 mb)
ESM(PDF 1.44 MB)

Copyright information

© Springer-Verlag 2007