Analysis and Quality Control of WHO- and UNEP-Coordinated Human Milk Studies 2000–2019: Polybrominated Diphenyl Ethers, Hexabromocyclododecanes, Chlorinated Paraffins and Polychlorinated Naphthalenes

Abstract

Four different analytical methods were used for the determination of (1) polybrominated diphenyl ethers (PBDE), (2) hexabromocyclododecanes (HBCDD), (3) chlorinated paraffins (CP) and (4) polychlorinated naphthalenes (PCN) in human milk samples of the WHO/UNEP-coordinated exposure studies. As a laboratory accredited according to EN ISO/IEC 17025, a comprehensive quality control program was applied to assure the reliability of results. This included procedural blanks, the use of numerous quality control samples as in-house reference materials and the participation in proficiency tests (PTs). Trueness was estimated from the PT samples using the assigned values.

The mean absolute deviation of the sum parameters ∑PBDE₆ and ∑PBDE₇ from the assigned values of 53 PT samples analysed between 2006 and 2021 was 12% and 14%, respectively.

For α-HBCDD as the most abundant diastereomer and the sum of α-, β- and γ-HBCDD, deviations of the reported value from the assigned value of the proficiency tests (31 samples, analysed between 2007 and 2021) were in most cases below 40% over a large concentration range, e.g., for α-HBCDD, between 0.0084 and 19 ng/g fw. For concentrations above 0.5 ng/g lipid, the deviation was in the range of approximately 0–30%.

For short-chain and medium-chain CP (SCCP and MCCP) all z-scores achieved in interlaboratory comparisons during 2017–2020 were within ±2 z and therefore satisfactory (13 PT samples were analysed for ΣCP, ΣSCCP and ΣMCCP using the GC-ECNI-Orbitrap-HRMS method, eight results achieved for ΣCP using the GC-EI-MS/MS method).

Due to the lack of available proficiency tests for PCN at the time of measuring the human milk samples of the 2016–2019 period, an external validation for control of the trueness was performed through an interlaboratory comparison with an independent laboratory. The deviation of the ΣPCN₁₃ in five test samples between the external laboratory and CVUA Freiburg was in the range from 3 to 20%. At a later stage (in 2021), the laboratory participated successfully in the first interlaboratory comparison study on PCN congeners in cod liver oil. The z-scores for seven congeners and two sum parameters were within ±2 z and therefore satisfactory. Also, the results for other of the altogether 26 PCN congeners were in accordance with the median values reported by all participants.

As a result, the determination of PBDE, HBCDD, CP and PCN in human milk samples of the WHO/UNEP-coordinated exposure studies followed the strict rules of the accreditation system and the general criteria for the operation of testing laboratories as laid down in EN ISO/IEC 17025.

1 Introduction

This compendium “Persistent Organic Pollutants in Human Milk” comprises a series of articles on human milk surveys structured in five parts. Part I, the introduction, provides an overview of the World Health Organization (WHO) and the United Nations Environment Programme (UNEP)-coordinated exposure studies on persistent organic pollutants (POPs) in human milk and their link to the Stockholm Convention, including protocols for the collection of samples and an overview on the participating countries with respect to regional distribution and temporal differentiation (Malisch et al. 2023a). It also includes a review on human milk surveys on POPs (Fürst 2023) and a review on the Stockholm Convention and its implementation by regional and global monitoring reports (Šebková 2023). Part II presents analytical methods and quality control. In Part III, the findings between 2000 and 2019 are presented in various publications, some of which are relevant to this article: (1) polybrominated substances (Schächtele et al. 2023), (2) chlorinated paraffins (CP) (Krätschmer et al. 2023) and (3) polychlorinated naphthalenes (Tschiggfrei et al. 2023). Part IV comprises assessments of time trends and of possible health risks for the breastfed infant that arise from dioxin-like compounds, and Part V presents conclusions and key messages.

This chapter describes the analytical methods and quality control measures used for the determination of polybrominated diphenyl ethers (PBDE), hexabromocyclododecanes (HBCDD), chlorinated paraffins (CP; comprising short-chain chlorinated paraffins [SCCP] and medium-chain chlorinated paraffins [MCCP]) and polychlorinated naphthalenes (PCN) in human milk samples obtained from the WHO/UNEP-coordinated exposure studies performed between 2000 and 2019.

2 Analytical Criteria

Accuracy depends on systematic errors and random components. “Trueness” (closeness of the agreement between the expectation of a test result or a measurement result and a true value (ISO 3534-2:2006, /24/)) and “Precision” (closeness of the agreement between independent test/measurement results obtained under stipulated conditions (ISO 3534-2:2006, /24/)) are used to describe accuracy and are therefore important criteria for the assessment of reliability of analytical methods (Eppe et al. 2017).

To provide reliable monitoring information for the Parties to the Stockholm Convention, the guidance document for the Global Monitoring Plan (GMP) proposed that a quantified objective for temporal studies should be stated, e.g. “to detect a 50% decrease in the levels of POPs within a 10-year period” (UNEP 2013, 2019). The statistical model used in the “Bi-ennial Global Interlaboratory Assessments on Persistent Organic Pollutants” is based on a target error of 25% to assess the performance of each laboratory for each analyte in each matrix. The analyte groups in the third round (2016/2017) and fourth round (2018/2019) of these assessments included organochlorine pesticides, polychlorinated biphenyls, dioxin-like POPs, PBDE, hexabromobiphenyl (only in the 2016/2017 round; however, without an assigned value for human milk), toxaphene, HBCDD (without assigned values for human milk in the 2016/2017 and 2018/2019 rounds) and perfluorinated alkyl substances, but not CP and PCN (UNEP 2017, 2021).

As of 2022, there were no analytical criteria defined for PBDE, HBCDD, PCN and short-, medium- or long-chain chlorinated paraffins (SCCP, MCCP, LCCP, respectively) in the European Union. The European Union Reference Laboratory (EURL) for halogenated POPs in feed and food published a general guidance document on SCCP and MCCP analysis in food samples (EURL for Halogenated POPs in Feed and Food 2021, with the annex describing analytical criteria under review). Interlaboratory studies and proficiency tests in that field commonly use a variation of ±25% as target standard deviation, leading to results within a variation of ± 50% being acceptable (Krätschmer and Schächtele 2019). Additionally, the EURL POPs published a guidance document on the determination of brominated contaminants, initially specifying analytical criteria for PBDE and HBCDD (EURL for Halogenated POPs in Feed and Food 2022a; Fernandes et al. 2022a).

3 Polybrominated Diphenyl Ethers (PBDE)

3.1 Analytes

For PBDE analysis, six prevalent analytes (comprising seven congeners) were recommended by the GMP Guidance Document for the groups from tetra-BDE to hepta-BDE (BDE-47, BDE-99, BDE-100, BDE-153, BDE-154, BDE-175/183 [co-eluting]) (UNEP 2013, 2019). The sum of these six PBDE analytes was used as the most important summarizing parameter until 2017, when BDE-209 was added (UNEP 2019). However, questions were raised whether BDE-175 is an important enough component of commercial octabromodiphenyl ether mixtures to be listed in Annex A of the Stockholm Convention. Since BDE-175 and BDE-183 co-elute on common HRGC columns, the presence of BDE-175 as an important component in technical octa-BDE mixtures has not been illustrated. The successful HRGC/LRMS separation of a 1:1 mixture of BDE-175 and BDE-183, as well as ¹H NMR analysis of technical material, has allowed to confirm that this congener is not present in technical products (e.g. Great Lakes DE-79™) in quantifiable amounts (Konstantinov et al. 2011). Therefore, the “Guidance document on the determination of organobromine contaminants for analytical parameters in food and feed” recommends 9 PBDE congeners as analytes of interest: seven congeners covered by the GMP Guidance Document (BDE-47, BDE-99, BDE-100, BDE-153, BDE-154, BDE-183 and BDE-209) and in addition BDE-28 and BDE-49 (EURL for Halogenated POPs in Feed and Food 2022a).

For the human milk samples of the period 2016–2019, in addition to the congeners recommended by the GMP guidance document (except BDE-175), 18 congeners were included to cover a broader range (BDE-15, BDE-17, BDE-28, BDE-49, BDE-66, BDE-75; BDE-77; BDE-85; BDE-119; BDE-126; BDE-138; BDE-190; BDE-196; BDE-197; BDE-203; BDE-206; BDE-207; BDE-208). Whereas BDE-28 and BDE-49 were included in EURL proficiency test, the other additional congeners and BDE-175 were not included in any proficiency tests, and therefore are currently not validated with the same degree of quality control as the nine analytes recommended by the EURL Guidance Document. However, results can give a first provisional indication whether other congeners might be of interest in addition to the congeners recommended by the GMP guidance document. Table 19 (in the appendix) lists these 25 PBDE congeners and the 13 ¹³C₁₂-labelled internal standards.

3.2 Extraction

The lipid portion of the milk is extracted as a first step, and the two procedures used for this are described below. ¹³C₁₂-labelled surrogates (or other ¹³C₁₂-labelled congeners) of many of the analytes are added to an aliquot of the extracted lipids and serve as internal standards, for more accurate quantitation and also for the calculation of recoveries.

3.2.1 Separation of the Cream (Lipid) Layer by Centrifugation

For extraction of the lipids, approx. 200 g of human milk sample was centrifuged in stainless steel centrifuge tubes for 10 min at approx. 3000 rpm at approx. 4 °C. The supernatant cream layer was transferred into a glass beaker. Sodium sulphate was added to the cream until the ground material was powdery when stirred. This powder was extracted 3–4 times with n-hexane by stirring well with a glass rod and filtering the extract. The solvent was rotary evaporated to give the extracted lipids. The lipids were dried using a nitrogen stream on a sand bath at approx. 80 °C for 30 min and in a drying oven at 103 °C for 30 min.

3.2.2 Twisselmann Hot Extraction

Before extraction, the sample was freeze-dried or mixed with a drying agent (polyacrylate or sodium sulphate). A mixture of ethanol and toluene (7/3, v/v) was used for the Twisselmann extraction (permanent “hot extraction” at the boiling point of the solvent mixtures for six hours). After evaporation of the solvent, the raw extract was dissolved in tert-butyl methyl ether (t-BME) for separation of polar co-extractives and for lipid determination after evaporation of t-BME.

3.2.3 Addition of Internal Standards

To an aliquot of the lipid aliquot, the ¹³C₁₂-labelled standards listed in Table 19 were added.

3.3 Clean-up

Principle

The extracts obtained from the above procedures require further purification. Manual or automated purification procedures are based on similar principles that essentially separate PBDE from the lipids and other co-extracted interferences using adsorbents such as alumina, silica, Florisil (activated magnesium silicate) or techniques such as gel permeation chromatography (GPC).

For the analysis of human milk, two different procedures were applied. A partly automated method including gel permeation chromatography, a multi-layer silica column and a Florisil column was used for the determination of BDE-47, BDE-99, BDE-100, BDE-153, BDE-154, BDE-183, whereas a fully automated method applying DEXTech Plus/Pure (LCTech, Obertaufkirchen, Germany) was used for these analytes including BDE-209 and the other specified BDE. Both methods include elements of the analytical procedures that are used for the determination of PCDD/PCDF and PCB and allow the simultaneous clean-up in different fractions, where required.

3.3.1 Partly Automated Clean-up Procedure

Gel permeation chromatography using Bio-Beads S-X3 (Bio-Rad, Hercules, CA, USA) was used to remove the lipid (in four runs with a maximum of 0.75 g fat each; 50 g Bio-Beads S-X3; eluent ethyl acetate/cyclohexane [1/1, v/v]). Small amounts of remaining lipid and oxidizable substances were removed using a mixed column loaded with layers of 1 g sulphuric acid (96%) impregnated silica gel and 1 g NaOH-impregnated silica gel (eluent: 20 mL heptane). A Florisil column (deactivated with 3% water) was used for further clean-up (elution of PBDE by heptane containing 0.2% of toluene; PBDE collected as the first fraction). (If PCB and PCDD/PCDF are co-analysed, PCB co-elute with PBDE in this first fraction; PCDD/PCDF are eluted by toluene in a second fraction).

3.3.2 Fully Automated Clean-up Procedure

The fat extract was dissolved in 1 mL of acetone and 5 mL of cyclohexane and injected into the purification system of the fully automated clean-up system (DEXTech Plus/Pure) with a 15 mL sample loop). The method for determination of PBDE uses the same method columns and eluents as used for the determination of PCDD/PCDF and PCB, thus, the clean-up for these three groups of analytes can be performed simultaneously. The method combines three columns: (1) a multi-layer silica sulphuric acid column (column 1); (2) an alumina column (column 2) and (3) a carbon column (column 3). Pre-packed columns were provided ready-to-use by the supplier. The sample extract was loaded with hexane first onto the multi-layer silica column, then transferred with hexane onto the alumina column. PBDE were eluted from the alumina column by a mixture of n-hexane and dichloromethane (1/1, v/v 50:50%) onto the carbon column. PBDE were not adsorbed by the carbon column (in contrast to PCDD/PCDF and non-ortho PCB) and collected as Fraction 1 (Table 1).

Table 1 Method parameters for the DexTech Plus/Pure system collecting PBDE in fraction 1 (if required, ndl-PCB and mono-ortho-PCB may also be collected in fraction 1; non-ortho PCB and PCDD/PCDF, in fraction 2)

This fraction was concentrated to 3 mL followed by the addition of 10 μL of dodecane, then carefully evaporated to dryness under a gentle stream of nitrogen and finally reconstituted with 90 μL toluene containing the recovery (syringe) standards ¹³C₁₂-BDE-77 and ¹³C₁₂-BDE-206.

3.4 GC-HRMS Measurement

The measurement of the PBDE was carried out using a GC-HRMS system (MAT95XP and DFS, Thermo Fisher Scientific, Waltham, MA, USA) monitoring M⁺ or M-2Br⁺ clusters at a resolution of 10,000 (5% valley). 5 μL of the final extract was injected into a PTV inlet using the solvent vent mode. Analytes of interest were separated on a 15 m or 30 m Rtx-1614 column (Restek, Bellefonte, PA, USA) or a 30 m DB-5 MS column (Agilent Technologies, Santa Clara, CA, USA) (Fig. 1).

Fig. 1

Example chromatogram showing separation of PBDE congeners by GC-HRMS (30 m Rtx-1614; R = 10,000)

3.5 Quality Control

Human milk samples were received at the reference laboratory over 20 years in five rounds between 2000 and 2019, each covering approximately 4 years, and analysed for PBDE between 2005 and 2020, including determination of PBDE in left-over samples of the 2000–2004 period. In order to check the comparability of results, an internal quality control program was run covering procedural blank samples, various kinds of in-house reference samples and external quality control based on regular participation in interlaboratory studies and proficiency tests.

3.5.1 Procedural Blank Samples

For the six PBDE congeners BDE-47, BDE-99, BDE-100, BDE-153, BDE-154 and BDE-183, 116 procedural blank samples were analysed between 2005 and 2021. These procedural blanks were included in all steps of the analytical method. Calculations for the limits of quantification (LOQs) were based on the use of an aliquot equivalent to 3 g lipid. In contrast to BDE-153, BDE-154 and BDE-183, the lower brominated congeners BDE-47 and BDE-99 were above the respective LOQs in more than half of the samples. Median procedural blank levels for the individual congeners were in most cases below the lowest levels found in the analysed human milk samples. In some cases, the maximum levels for procedural blanks were in the range of the minimum levels found in human milk samples for these congeners. In these cases, a contribution from the method background (as seen in the procedural blanks) to very low levels found in human milk cannot be excluded. However, the median concentrations of the individual BDE congeners in human milk were about two orders of magnitude higher than the median of the blank samples (range 34–650 times higher), the sum parameter ∑PBDE₆ by a factor of 140 (in comparison with the maximum found in the blank samples, by a factor of 8) (Table 2). The sum parameter was calculated as “upper-bound” (ub) result (with calculation of the contribution of not quantified congeners to the sum parameter as the limit of quantification).

Table 2 Concentrations of six PBDE congeners and ∑PBDE₆ (ub) of reagent blank samples analysed together with human milk and fatty food samples between 2005 and 2021 and of BDE-209 and ∑PBDE₇ (ub) analysed between 2019 and 2021 together with the human milk samples of the 2016–2019 period; the concentration range found in human milk samples is included for comparison

BDE-209 was determined in human milk samples collected in the 2016–2019 period; these analyses were performed from 2019 to 2021. The BDE-209 results for procedural blanks analysed in the same batch with these human milk samples were in the range 0.04–0.07 ng/g lipid and therefore below the levels found in human milk. Results for the sum parameter ∑PBDE₇ (ub) for the procedural blank samples were in all cases below levels found in the human milk samples (median of ∑PBDE₇ in human milk by a factor of 30 higher than median of blank samples).

3.5.2 Quality Control Samples as In-House Reference Material

Six different quality control (QC) samples were used for monitoring of the precision of PBDE analysis between 2004 and 2021. Both naturally contaminated samples and samples spiked with native PBDE congeners were analysed. An overview of the different quality control samples, the number of replicate analyses and the results are given in Table 3. The concentration for the sum parameters ∑PBDE₆ and ∑PBDE₇ ranged between 0.1 and 9.2 ng/g lipid, and 0.2 and 9.3 ng/g lipid, respectively, and therefore covered most of the concentration range in human milk samples, except for very highly contaminated samples. The coefficients of variation (CVs) for the sum parameters ∑PBDE₆ and ∑PBDE₇ were in the acceptable range between 9 and 16%, and 6 and 34%, respectively.

Table 3 Quality control samples analysed between 2004 and 2021 for PBDE and CVs of the sum parameters ∑PBDE₆ and ∑PBDE₇

Figure 2 gives an overview of the CVs of the individual congeners for the six different QC samples compared with the mean content of these congeners in the QC samples. For most of the congeners, CVs below 30% were observed and only in some cases did CVs exceed 40%.

Fig. 2

Mean content of PBDE congeners in six quality control samples (ng/g lipid) plotted against coefficients of variation of these congeners

In conclusion, these QC samples analysed between 2004 and 2021 show a good comparability of the results for sum parameters and in most cases also for individual congeners over this period of time.

3.5.3 Participation in Proficiency Tests

Between 2006 and 2021 CVUA Freiburg participated in numerous interlaboratory studies and proficiency tests generating results for individual congeners and sum parameters for up to 53 different food test samples. The concentration ranges of these proficiency test covered a very wide concentration range from the low pg/g fresh weight (fw) range to the high ng/g fw range. Many of these PT samples were fat or oil samples and in these cases the concentration reported on a fat basis equalled the concentration on a fresh weight basis. An overview of the results is given in Table 4 showing the wide concentration range and the mean absolute deviation. Figure 3 illustrates the deviations of individual results from the assigned values, Fig. 4 the deviations of the sum parameters.

Table 4 Results of proficiency tests for PBDE congeners and sum parameters between 2006 and 2021

Fig. 3

Deviation of results reported for individual congeners of the assigned values for proficiency tests between 2006 and 2021 (Note that many of these PT samples were fat or oil samples and in these cases the concentration on fat basis equals the concentration on a fresh weight basis.)

Fig. 4

Deviation of results reported for PBDE sum parameters from assigned values for proficiency tests between 2006 and 2021 (Note that many of these PT samples were fat or oil samples and in these cases the concentration on fat basis equals the concentration on a fresh weight basis.)

The mean absolute deviation of the reported results of the individual congeners from the assigned values was in most cases below 20% with higher values above 30% for BDE-183 and BDE-209, the deviation of the sum parameters 12% for ∑PBDE₆ and 14% for ∑PBDE₇.

4 Hexabromocyclododecanes (HBCDD)

4.1 Analytical Procedure and Analytes

For determination of HBCDD, the extraction and clean-up steps of a method based on the European Norm 1528 for determination of residues of nonpolar organochlorine pesticides and contaminants were used, which is also part of the official collection of test methods in Germany (European Norm [EN] 1528; German Standard § 64 LFGB). It is based on a modular structure and covers a wide range of analyte-matrix combinations (suitable for food of animal origin and human milk).

After centrifugation at 3000 rpm for 10 min and separation of the cream, lipids were extracted with nonpolar solvents. The supernatant cream layer was transferred into a glass beaker. Sodium sulphate was added to the cream until the ground material was powdery when stirred. This powder was extracted 3–4 times with n-hexane by stirring well with a glass rod and filtering the extract. The solvent was evaporated to give the extracted lipids.

As internal standards, TCB (2,4,5′-trichlorobiphenyl), triphenylphosphate and PCB 209 (decachlorobiphenyl) were used. After addition of the internal standards, up to 0.5 g lipids was separated by gel chromatography using polystyrene gel (Bio-Beads S-X3; length 740 mm, 20 mm i.d., filling level 500–500 mm; cyclohexane/ethyl acetate mixture [1:1] as eluent). After concentration to about 2–3 mL volume, the eluate was further concentrated with isooctane added and evaporated to about 1 mL. Ethyl acetate had to be completely removed, otherwise isooctane had to be added again and concentrated.

Chromatography on a small column of partially deactivated silica gel was performed as the final clean-up step. The silica gel (70–230 mesh) was heated overnight at 130 °C and allowed to cool in a desiccator. After adding 1.5% of water, it was shaken for 30 min and then stored in a tightly sealed container. The chromatographic tube was packed with 1 g of deactivated silica gel. After pre-washing the column with 2 × 5 mL hexane, the isooctane solution was loaded onto the silica column and the analytes of interest were eluted by a total volume of 10 mL toluene.

The eluate of this fraction was evaporated to a small volume (no complete dryness) and immediately taken up with 0.1 mL methanol in the intended final volume for determination by HPLS-MS/MS (Agilent TQ 6410, Agilent Technologies, Santa Clara, CA, USA) using a C18 column (50 × 2.1 mm, 1.8 μm ID, Agilent Technologies, Santa Clara, CA, USA).

4.2 Quality Control

4.2.1 Procedural Blank Samples

No α-, β- and γ-HBCDD concentrations above the limit of quantification were found in procedural blank samples (LOQ for 90% of the procedural blank samples: <0.1 ng/g lipid; max 0.5 ng/g lipid).

4.2.2 Quality Control Samples as In-House Reference Material

A fat sample spiked with α-, β- and γ-HBCDD at 5 ng/g lipid was used as an in-house reference material for quality control. Coefficients of variations for individual diastereomers and for the sum parameter were in an acceptable range between 15 and 17% (Table 5).

Table 5 HBCDD concentrations and precision (CV) determined in 106 quality control samples (spiked fat, 5 ng/g lipid) used as in-house reference materials

4.2.3 Participation in Proficiency Tests

The samples analysed for proficiency tests between 2007 and 2021 covered a wide concentration range between below 0.01 ng/g and above 10 ng/g fresh weight. Many of these PT samples were fat or oils and in these cases the concentration on fat basis equals the concentration on fresh weight basis. This range covers the concentrations found in human milk samples: The α-HBCDD levels of 102 pooled samples from 72 countries collected between 2006 and 2019 ranged between <0.1 ng/g lipid and 15 ng/g lipid (median: 0.5 ng/g lipid; 90% of all results were below 2 ng/g lipid). In nearly all samples, β-HBCDD and γ-HBCDD occurred below or around the limit of quantification (Schächtele et al. 2023).

The deviations of the reported value from the assigned value of the proficiency tests for α-HBCDD as most abundant diastereomer and the sum of α-, β- and γ-HBCDD were in the range of approximately 0–30% for concentrations above 0.5 ng/g fw respectively lipid. The mean value of the absolute deviation was below 40% for all diastereomers and the sum parameter over the whole concentration range, e.g. for α-HBCDD the value was between 0.0084 ng/g fw and 19 ng/g fw (in this case, equal to 19 ng/g lipid) (Table 6, Fig. 5).

Table 6 Results of proficiency tests performed between 2007 and 2021 for HBCDD congeners and the sum parameter between 2007 and 2021

Fig. 5

Deviation of results reported for HBCDD from assigned values for proficiency tests between 2007 and 2021 (Note that many of these PT samples were fat or oil samples and in these cases the concentration on a fat basis equals the concentration on a fresh weight basis.)

5 Chlorinated Paraffins (CP)

The method for routine analysis of chlorinated paraffins in food and human milk was developed and validated as part of a doctoral thesis (Krätschmer 2022).

5.1 Analytical Procedure and Analytes

5.1.1 Sample Preparation

Sample preparation of the frozen human milk sample was performed as described elsewhere (Krätschmer et al. 2018, 2019, 2021). In brief, 50 g of warmed, homogenized samples were filled into baked-out glass centrifuge tubes. Cooled centrifugation (4 °C, 3000 rpm, 10 min) was used to separate the cream which was then removed from the hydrogenous phase. The cream was fortified with the recovery standard (¹³C₁₀-1,5,5,6,6,10-hexachlorodecane, Cambridge Isotope Laboratories, Tewksbury, Ma, USA) and dried by grinding with sodium sulphate until a powdery consistency was reached. Manual cold extraction with dichloromethane/n-hexane (1:1, v/v) was performed in triplicate and the decanted, filtered solvent evaporated to dryness using a rotary evaporator.

Further sample clean-up was performed using open column chromatography (Fig. 6). In a first step, the lipids were hydrolysed on an acidified silica column. The resulting extract was fractionated on a Florisil^® column (magnesium silicate primed with 1.5% water) eluted with 75 mL n-hexane followed by 60 mL dichloromethane. The latter fraction contained the CP and was concentrated, initially using a rotary evaporator followed by a gentle nitrogen stream before the addition of the syringe standard ε-hexachlorocyclohexane (ε-HCH, Dr. Ehrenstorfer, Augsburg, Germany).

Fig. 6

Description and elution protocol of the acidified silica column (left) and Florisil column (right) used for sample clean-up

Each sample batch additionally included a procedural blank (sodium sulphate prepared like a sample) as well as different quality control (QC) samples which consisted of raw cow’s milk with and without fortification with SCCP and MCCP standards at different levels. For other sample series, QC samples originating from interlaboratory studies (e.g. fortified coconut fat, fortified lard) or prepared from market food (e.g. extra virgin olive oil) were used.

5.1.2 Analysis

Unlike all the other POPs analysed in these samples, CP are not separable by chromatographic methods (Fig. 7). Accordingly, standard quantification procedures using response factors of quantification standards of known concentration are not applicable here. Over the last decade, a multitude of quantification methods both differing in instrumentation used (e.g. liquid or gas chromatography, tandem gas chromatography, direct injection combined with low- or high-resolution mass spectrometry or tandem mass spectrometry) and data treatment strategies (e.g. peak deconvolution, different forms of calibration, direct application of pre-modelled response factors) have been published in the literature (van Mourik et al. 2015; Yuan et al. 2019; Fernandes et al. 2022b).

Fig. 7

Total ion chromatograms on a GC-ECNI-Orbitrap-HRMS instrument of an SCCP, MCCP and LCCP commercial quantification standard. LCCP are almost impossible to quantify using gas chromatography due to low volatility

Although many of these approaches could be proven as comparable during international interlaboratory studies (van Mourik et al. 2018; Krätschmer and Schächtele 2019), the true concentration of CP in any given sample remains unknown. Another major difficulty is the poor availability of suitable quantification standards (Schinkel et al. 2018; Fernandes et al. 2022c), a situation that has shown some improvement in the last year or so. Given the major developments in this field of analysis in recent years, the method applied to human milk samples at CVUA Freiburg has also changed over the years. Therefore, both the method applied until 2016 and the current method will be briefly described in the following sections.

5.1.2.1 Semi-quantitative Analysis (GC-EI-MS/MS)

Until 2016, human milk samples were analysed by GC-EI-MS/MS (Reth et al. 2005). This allowed for a reliable and comparable estimation of the total CP amount in the sample, but not for a distinction between SCCP and MCCP. The following parameters, as also shown in the EURL POPs Guidance Document on the Analysis of CP in Food, were applied (EURL for Halogenated POPs in Feed and Food 2021).

Electron ionization (EI) is a hard ionization method, i.e. it tends to create many small fragment ions. As CP have similar structures, this fragmentation is non-specific and reduces information of the parent molecules. Tandem mass spectrometry at least allows for the identification of three mass transitions that are specific for all CP and allows for separation from other organochlorine contaminants.

Each of the three mass transitions shown in Table 7 may be used to provide a semi-quantitative estimate of the total CP amount in a sample. As partial characterization, the results of all three transitions, quantified via (a) SCCP and (b) MCCP calibration solutions, may be reported as intermediate results. However, the end result is the mean value between those two results and represents the total CP amount in the sample, as the sum of both intermediate results would be a gross overestimation (Fig. 8).

Table 7 Method parameters for CP determination using GC-EI-MS/MS

Fig. 8

Schematic quantification workflow for CP analysis using GC-EI-MS/MS

5.1.2.2 Homologue Group Specific Analysis (GC-ECNI-Orbitrap-HRMS)

In 2016, a new quantification method using the GC-ECNI-Orbitrap-HRMS technology was established for CP at the CVUA Freiburg. The very high mass resolution of this instrument allowed for a differentiation between SCCP and MCCP while operation in full scan mode as well as reduced analysis time compared to other GC-ECNI-MS methods. The choice of the comparatively softer ionization method additionally allowed for detection of the molecular or pseudo-molecular ions and a variety of other fragment ions, offering further insight into the homologue group distribution and patterns of each sample. The specific parameters applied for CP analysis have been described in several publications (Krätschmer et al. 2018; Mézière et al. 2020; EURL for Halogenated POPs in Feed and Food 2021) and are summarized in Table 8.

Table 8 Method parameters for CP quantification using GC-ECNI-Orbitrap-HRMS.

While it is possible to determine homologue group specific peak areas for [M-Cl]^– and [M-HCl]^– fragment ions with this method, currently no suitable standards with defined homologue group concentrations are available.

Therefore, the quantification strategy applied to the GC-ECNI-Orbitrap-HRMS data works in three interdependent steps (Fig. 9):

• 1. Use of alkyl chain length specific CP standards of different chlorination degrees to model homologue group specific response factors according to Yuan et al. 2017

• 2. Use of these response factors to characterize an SCCP and MCCP mixture of alkyl chain length specific standards designed to resemble the homologue group pattern of the target sample matrix (e.g. fish)

• 3. Use of the characterized standard mixture as linear or exponential calibration curve for sample quantification

Fig. 9

Schematic quantification workflow for CP analysis using GC-ECNI-Orbitrap-HRMS

This approach allows for maximum flexibility to incorporate day-to-day changes in the analytical system while allowing for first indications even of homologue group specific concentrations and reliable CP alkyl chain length specific results. For easier comparison to other studies, results reported for the human milk samples were summarized as ΣSCCP and ΣMCCP.

5.2 Quality Control

A comprehensive quality control programme was applied to prove the long-time reliability of results between 2012 and 2021.This included spiked samples at different levels and different kinds of quality control samples. Possible systematic errors were checked by analysis of reference material or participation in several interlaboratory studies. This validation is part of the general quality control programme applied in the daily routine for analysis of all kinds of samples. Analyses were performed by different operators using different chemicals over a long time and data collected in separate runs—therefore, these validation data are much more robust than that obtained from a single initial validation when one technician performs repeated analyses under the same conditions using the same chemicals in one sequence. An overview spanning results of the quality control for the whole period 2000–2019 is summarized in the following section.

5.2.1 Initial Method Validation

5.2.1.1 Method Validation: GC-EI-MS/MS

For initial method validation, raw cow’s milk was fortified with a CP standard at two different levels: 2 ng/g milk (=1 ng/μL injected sample) and 10 ng/g milk (=5 ng/μL injected sample). Additionally, chemical blanks and milk blanks were analysed with each sample batch. The fortified samples were prepared and analysed five times each. Table 9 describes the method criteria derived from this validation study.

Table 9 Validation results and resulting parameter for GC-EI-MS/MS

As the sample preparation method cannot eliminate the blank levels completely, they are used for determining a limit of detection (LOD) and limit of quantification (LOQ). For calculation purposes, the blank levels of sample batches spanning at least 3 months were taken into account and LOD and LOQ values were calculated according to the following formula:

$$ \mathrm{LOD}={x}_{\mathrm{blank}}+3\times {s}_{\mathrm{blank}} $$

(1)

$$ \mathrm{LOQ}={x}_{\mathrm{blank}}+10\times {s}_{\mathrm{blank}} $$

(2)

where x_blank is the mean blank level [ng/μL] and s_blank the standard deviation [ng/μL] of the blank levels taken into account for the calculation.

During initial method validation, a mean blank level of 0.12 ng/μL with a standard deviation of 0.05 ng/μL resulted in the following parameters:

• LOD = 0.12 ng/μL + 3 × 0.05 ng/μL = 0.26 ng/μL injected sample (= 0.5 ng/g milk)

• LOQ = 0.12 ng/μL + 10 × 0.05 ng/μL = 0.62 ng/μL injected sample (= 1.2 ng/g milk)

The dilution factor used to convert ng/μL injected sample into ng/g milk includes the initial sample weight (50 g) and final sample volume before injection (100 μL). Using similar dilution factors, the LOD and LOQ can also be applied to other matrices for a first indication of the working range.

5.2.1.2 Method Validation: GC-ECNI-Orbitrap-HRMS

For initial method validation, coconut fat and lard were fortified with SCCP and MCCP standards at different levels (Table 10).

Table 10 Fortification levels for initial method validation

These samples were also part of different international interlaboratory studies. For the purpose of this validation study, both samples were analysed in triplicate with accompanying chemical blank samples. Tables 11 and 12 show the results and performance with regard to recovery and RSD as important method criteria derived from this validation study.

Table 11 Results and validation parameters for the coconut fat samples

Table 12 Results and validation parameters for the lard samples

As the sample preparation method cannot eliminate the CP background (blank level) completely, they are used for determining a limit of detection (LOD) and limit of quantification (LOQ). For calculation purposes, the blank levels of sample batches spanning the last 2 years on a rolling scheme were taken into account and LOD and LOQ for SCCP, MCCP and the sum of CP were calculated according to Eqs. (1) and (2). Table 13 gives an overview of the mean blank levels and resulting LODs and LOQs during the initial validation period.

Table 13 Overview of limit of detection (LOD) and limit of quantification (LOQ) determined during initial method validation based on 14 blank levels analysed over the duration of 3 months

5.2.2 Quality Control Samples

Between 2017 and 2020, several different matrices were prepared as quality control samples in order to provide a good fit with the sample matrix. In the case of human milk sample, raw cow’s milk was analysed as the procedural blank matrix and this was fortified with SCCP and MCCP standards. Beside such matrix-specific QC samples, fortified coconut fat or lard samples from the 2017 and 2018 interlaboratory studies on SCCP and MCCP in food were routinely added to each sample batch and can therefore give a more accurate view of long-term stability and repeatability of the method (Fig. 10).

Fig. 10

Recoveries of several different QC samples analysed 2016–2020 (left) and fortified raw cow’s milk samples analysed in tandem with each human milk batch 2017–2020 (right). The yellow line indicates warning levels for daily quality control, red lines equal warning limits of current CP interlaboratory studies

5.2.3 Participation in Interlaboratory Studies

Interlaboratory studies and proficiency tests on chlorinated paraffins in biota are to this day very rare. The performance of the GC-EI-MS/MS method was compared to other laboratories in the interlaboratory testing scheme organized by QUASIMEME 2011–2017 (van Mourik et al. 2018), which focused on SCCP in environmental matrices and standard solutions. Additionally, EURL POPs organized yearly interlaboratory studies on SCCP and MCCP in food matrices starting 2017, including CP later on in their POPs proficiency tests as optional analytes (EURL for Dioxins and PCBs in Feed and Food 2018; Krätschmer and Schächtele 2019; EURL for Halogenated POPs in Feed and Food 2019, 2020a, b).

The GC-ECNI-Orbitrap-HRMS method was used in this second study scheme as the other method was incapable of determining SCCP and MCCP separately. Due to the very complex and specialized field of analysis, the number of participants in each study was comparatively lower than for well-established analyte groups, leading sometimes to evaluations being provisional or in some cases completely impossible. Figure 11 shows all z-scores achieved in interlaboratory comparisons for the duration of the human milk studies discussed in the results for chlorinated paraffins. As conclusion, all z-scores (13 samples analysed for ΣCP, ΣSCCP and ΣMCCP using the GC-ECNI-Orbitrap-HRMS method, 8 results achieved for ΣCP using the GC-EI-MS/MS method) were within ±2 z and therefore deemed satisfactory.

Fig. 11

Results of interlaboratory studies and proficiency tests on the determination of CP, SCCP and MCCP 2017–2020. The z-scores were calculated using a fitness-for-purpose-based standard deviation for proficiency tests σ = 25%. Results within ±2 z are deemed satisfactory. (a) Results achieved using the GC-ECNI-Orbitrap-HRMS method, (b) results for ΣCP achieved using the GC-EI-MS/MS method

6 Polychlorinated Naphthalenes (PCN)

6.1 Analytical Procedure and Analytes

For extraction of the lipids, approx. 200 g of human milk sample at approx. 4 °C was centrifuged in stainless steel centrifuge tubes for 10 min. at approx. 3000 rpm. The supernatant cream layer was transferred into a glass beaker. Sodium sulphate was added to the cream with grinding until the material was powdery when stirred. This powder was extracted 3–4 times with n-hexane by stirring well with a glass rod and filtering the extract. The solvent was evaporated by rotary evaporation to give the extracted lipids. The lipids were dried in a nitrogen stream on a sand bath at approx. 80 °C for 30 min and in a drying oven at 103 °C for 30 min.

A 2 g aliquot of this extract was spiked with ¹³C₁₀-labelled standards and dissolved in 8 mL hexane. Table 20 (in the appendix) lists the 26 determined native congeners and the eight ¹³C₁₀-labelled standards that were used. Seven ¹³C₁₀-labelled standards listed in Table 20 in bold were added as internal standards to the lipid aliquot. ¹³C₁₀-labelled PCN 65 listed in Table 20 was used as the recovery standard and added after clean-up before the final determination step. The congeners were chosen based on the toxicological characteristics, reported levels of occurrence, congener patterns and the availability of analytical standards. Thus, as an example, congener 54 was not included in the target scope even if relative potency factors (REP) were reported (Fernandes et al. 2017) as the single PCN 54 standard was not available at the beginning of method development. As the availability of standards continues to improve, it is recommended and planned to extend the scope to other relevant congeners. Standards were obtained from LGC Standards (Wesel, Germany) and diluted to the appropriate levels in toluene.

Samples were purified by a fully automated clean-up system (DEXTech Plus, LCTech Obertaufkirchen, Germany) using three columns:

• Standard multi-layer silica sulphuric acid column

• Alumina column

• Carbon column

The method for determination of PCN uses the same columns and eluents as described above in the method for PBDE (Sect. 3.3.2) and for the determination of PCDD/PCDF and PCB. Pre-packed columns were provided ready-to-use by the supplier. The sample extract was loaded with hexane first onto the multi-layer silica column, then transferred with hexane onto the alumina column. PCN were eluted from the alumina column by a mixture of n-hexane and dichloromethane (1/1, v/v) onto the carbon column. The target PCN were eluted by toluene as fraction 2, spiked with the ¹³C₁₀-labelled recovery standard, evaporated to near dryness (~0.5 mL) and transferred into a vial. The final extract was gently blown off with nitrogen to a final volume of 50 μL. 25 μL of the final extract was transferred to a second vial and stored for a second, confirmation measurement.

The measurements were carried out using HRGC/HRMS (Trace 1310 GC coupled to DFS MS, Thermo Fisher Scientific, Waltham, MA, USA) at a resolution of 10,000 (at 5% peak height) and quantified against a 5-point calibration curve. PCN congeners (5 μL injection volume) were separated on a DB5-MS GC column (Agilent Technologies, Santa Clara, CA, USA). Confirming measurement was carried out using the same column in a GC-Orbitrap Q Exactive MS (Thermo Fisher Scientific, Waltham, MA, USA) at a resolution of 60,000 (FWHM @ m/z 200) and quantified against a 4-point calibration curve.

Five congener pairs (PCN 28/36, 52/60, 64/68, 66/67 and 71/72) could not be separated using the existing set-up, and although five other columns (Rtx-Dioxin2, Rtx-PCB, Rtx-2330 [Restek, Bellefonte, PA, USA] and DB-Dioxin, ZB-Dioxin [Phenomenex, Torrence, CA, USA]) were tested during the earlier method development phase, none of them were able to separate all the co-eluting congeners. In particular, PCN 66/67 could not be separated with any of the tested conventional chromatography columns. Columns used by other laboratories, i.e., ZB-1701 P fused silica column (Phenomenex, Torrence, CA, USA), were similarly not suitable to separate PCN 66/67 (Zacs et al. 2021). Only a special GC column (Rt-βDEXcst, Restek, Bellefonte, PA, USA) could separate PCN 66 and 67, however, with considerable limitations under routine conditions (long GC run times of about 100 min, short column lifetime, high column bleeding and therefore increased maintenance of the MS ion source) (Helm 2002). Furthermore, this special GC column was used for separation of closely eluting PCN congeners including PCN 66 and 67 in technical mixtures by two-dimensional GC/quadrupole mass spectrometric detection (GC × GC/qMS) on Rt-βDEXcst and DB-Wax phases. However, no quantitative data were given on the composition, and neither PCN 66 nor 67 was shown in the figures on relative abundance (Hanari et al. 2013); for other stationary phases, see Fernandes et al. (2017). As a conclusion, the determination of the concentration of the individual congeners PCN 66 and 67 in human milk is not possible under routine conditions and requires research for development of a practical and valid method with sufficient sensitivity.

6.1.1 Sum Parameter for Selected PCN

Concentrations of the sum of 26 congeners (Table 20) were calculated and are part of the report on PCN in human milk (Tschiggfrei et al. 2023).

It was observed that the number of congeners reported in the literature varies for different reasons (i.e. lack of standards depending on when determinations were carried out, knowledge on occurrence and toxicology, analytical feasibility). At least the following congeners (including congeners in co-eluting pairs) were reported frequently in literature (Fernandes et al. 2017): PCN 52/60, 53, 66/67, 64/68, 69, 71/72, 73, 74 and 75. These 13 congeners show toxicological relevance due to high REP factors up to 0.004 for PCN66/67 (Falandysz et al. 2019; Fernandes et al. 2010, 2011, 2017, 2022b, d; Zhihua et al. 2019; Zacs et al. 2021). In order to give an additional overview of the quality control with focus on these congeners, the sum of these 13 congeners was calculated in addition to the sum of 26 congeners and is used for certain quality control parameters.

6.1.2 Toxic Equivalents (TEQ)

Toxic Equivalents (TEQ) were calculated as the sum of the products of the concentration of each compound (26 congeners) multiplied by the corresponding aryl hydrocarbon receptor-mediated (dioxin-like) relative potency factors (REP) and provided an estimate of the 2,3,7,8-TCDD-like activity.

REP values for PCN suggested by Falandysz et al. 2014 and REPs used in human exposure studies (Falandysz et al. 2019; Falandysz and Fernandes 2020; Fernandes et al. 2010, 2011, 2017, 2022b, d; Pratt et al. 2013; Zhihua et al. 2019; Zacs et al. 2021) were used for the estimation of PCN-TEQ. The applied REPs are compiled in the article on PCN in human milk (Tschiggfrei et al. 2023).

6.1.3 Limits of Quantification and Acceptable Differences Between Upper-Bound and Lower-Bound Results for PCN-TEQ

The limit of detection (LOD) and/or limit of quantification (LOQ) are important parameters for the evaluation of the reliability of analytical results. Due to similar analytical attributes between polychlorinated dibenzo-p-dioxins (PCDD), polychlorinated dibenzofurans (PCDF) and PCN, the requirements for the determination of the LOQ for PCDD/PCDF in food according to European legislation were taken as guidance. Thus, the following conditions were followed to determine the LOQ for all 26 PCN congeners (European Commission 2012):

The accepted specific LOQ of an individual congener is the concentration of an analyte in the extract of a sample which produces an instrumental response at two different ions, to be monitored with an S/N (signal/noise) ratio of 3:1 for the less sensitive signal and fulfilment of the basic requirements such as retention time, isotope ratio according to the determination procedures as described in EPA method 1613 revision B.

Following the concept for reporting of TEQ results as established for PCDD/PCDF and dioxin-like polychlorinated biphenyls (dl-PCB), the upper-bound (ND=LOQ) and the lower-bound (ND = 0) values should be given (Malisch et al. 2023b; UNEP 2019). The following two harmonized quality criteria were applied:

• (i) Calculation of the contribution of each non-detected congener to the TEQ as zero (lower-bound concentrations)

• (ii) Calculation of the contribution of each non-detected congener to the TEQ as the limit of detection (upper-bound concentrations)

The upper- and lower-bound values have important implications for the interpretation of the analytical results (Malisch and Schächtele 2023). As a performance criterion, the difference between these two should be less than 20% (UNEP 2019). In cases of co-eluting congeners and different REPs used in human biomonitoring or as suggested by Falandysz et al. 2014, both REP factors were applied and conclusions on the differences of the PCN-TEQ results drawn. The median of the differences between lower- and upper-bound PCN-TEQ in all 40 human milk samples of the 2016–2019 period was 0.3%. The range was between 0 and 2%, if REPs as used for human biomonitoring were applied, and between 0 and 3%, if other suggested REPs were used. Therefore, all samples analysed in this study fulfilled this QA/QC criterion. These differences were considered negligible (Tschiggfrei et al. 2023).

6.2 Quality Control

Before analysing the WHO-human milk study samples on PCN, the developed method was checked for precision and trueness by a small validation study in raw milk. Possible systematic errors were checked by applying a quality control programme including procedural blank samples, two different kinds of in-house reference material (spiked milk fat and butter) and confirmation of certain results by a different detection technique (GC-Orbitrap measurement). Quality control samples were included in the routine analysis for a broader picture of the PCN spectrum and to check the accuracy of the method. As the analysis of PCN is not widespread yet and no proficiency tests were available at the time of performance of the analyses, external validation was not possible at that time. As a substitute, contaminated fat samples with incurred PCN and a vegetable oil sample spiked with PCN were analysed by an independent laboratory to check the trueness. At a later stage, CVUA Freiburg took part in a first interlaboratory comparison study in cod liver oil for 26 PCN congeners conducted by the EURL POPs in the second half of 2021; the results are included in the following (see Sect. 6.2.5). All these steps illustrate the extent of validation, if methods have to be developed for new POPs.

6.2.1 Procedural Blank Samples

For PCN, the median of 22 procedural blank samples analysed between 2020 and 2021 is 0.61 pg/g fat (lower-bound [lb]) for ΣPCN₂₆ (Table 14). In most cases, congeners were below the LOQ. Therefore, the lower-bound value of the procedural blank is a better indication of the worst-case contamination than a reagent blank which could be considered for possible subtraction. The influence of procedural blank samples was negligible for human milk samples.

Table 14 ΣPCN₂₆ (pg/g lipid) levels and ΣPCN₁₃ (pg/g lipid) levels of reagent blank samples analysed together with human milk and fatty food samples between 2020 and 2021

6.2.2 Fortified Raw Milk and Butter Samples as In-House Reference Material

Fortification of fat or oil (i.e. sunflower oil) samples (which were tested to be free of PCN residues) with native PCN congeners at relevant levels is a well-established procedure to check the recovery of native analytes. The extracted fat of four raw milk samples was spiked at 15 pg/g lipid for each of the 26 PCN congeners. Evaluation of the recovery and precision was carried out by calculating the coefficient of variation (CV), before starting the analyses of the WHO/UNEP-human milk samples. Additionally, seven butter samples spiked at 15 pg/g lipid for each of the 26 PCN congeners were evaluated together with the performance of the analyses of the human milk samples. The mean recoveries for milk fat and butter for ΣPCN₁₃ and ΣPCN₂₆ were in the range 100–101% with a CV of the recovery in the range 1–2% (Table 15). For each of the congeners, the recoveries were between 86% (PCN 42) and 131% (PCN 31) with a CV between 0.7 and 11% (Fig. 12).

Table 15 Recovery and CV (%) of samples fortified at 195 pg/g lipid ΣPCN₁₃ and 390 pg/g lipid ΣPCN₂₆ respectively, analysed before (milk fat) and together (butter) with human milk samples

Fig. 12

Recovery and CV (%) of samples fortified at 15 pg/g lipid for each of the 26 PCN congeners analysed before (milk fat) and together (butter) with human milk samples

6.2.3 Quality Control Samples as In-House Reference Material and Precision

In order to control the analytical performance at different concentration levels of PCN congeners, a quality control sample (fish oil contaminated at different levels of single PCN congeners) was analysed together with human milk samples. In this way, the linearity of the response for PCN (range between 0.2 pg/g lipid for PCN 70 and 130 pg/g lipid for PCN 52/60) was checked. The fish oil, with a concentration of 324 pg ΣPCN₂₆/g lipid, had on average 2–3 times higher concentrations of ΣPCN₂₆ than human milk (26–170 pg ΣPCN₂₆/g lipid; ΣPCN₁₃ 20–134 pg/g lipid; Tschiggfrei et al. 2023), but covers the same order of magnitude as findings reported in the literature based on the determined congeners (483–3081 pg ΣTetraCN-OctaCN/g lipid (Lundén and Norén 1998), 59–168 pg ΣPCN₁₂/g lipid (Pratt et al. 2013), 211–2497 pg ΣMonoCN-OctaCN/g lipid (Li et al. 2020).

Over the whole period of its use from 2019 until 2021, 24 replicates of the fish oil were analysed under intermediate precision conditions, i.e., by different technicians under varying conditions, various batches of chemicals, instrumental conditions, etc. Table 16 presents the coefficient of variation (CV) obtained by repeated analysis of this quality control sample, indicating a high precision of the analytical method. In the 3 years of use of this quality control sample, a CV of 7% was observed for the mean level for ΣPCN₂₆ of 324 pg/g lipid, a CV of 6% for the mean level for ΣPCN₁₃ of 220 pg/g lipid.

Table 16 CV (%) of the quality control sample “fish oil” at 324 pg ΣPCN₂₆/g lipid and 220 pg ΣPCN₁₃/g lipid, respectively, used between 2019 and 2021

Figure 13 illustrates the quality control charts for ΣPCN₂₆ resulting from use of this fish oil over the whole period of its use from 2019 until 2021. Around the mean (M), warning levels are set at two sigma (lower warning level at M-2s, upper warning level at M+2s), control levels at three sigma (lower control level at M-3s, upper control level at M+3s). The upper warning level of 372 pg ΣPCN₂₆/g lipid was reached by one of the 24 replicates, the lower warning level of 276 pg ΣPCN₂₆/g lipid was not exceeded. None of the 24 replicates exceeded the lower (252 pg ΣPCN₂₆/g lipid) or upper (387 pg ΣPCN₂₆/g lipid) control level.

Fig. 13

Quality control chart for fish oil (324 pg ΣPCN₂₆/g lipid) used from 2019 to 2021 (UCL, upper control level; UWL, upper warning level; LWL, lower warning level; LCL, lower control level)

In addition to these conclusions from the general quality control of the PCN analyses performed between 2019 and 2021, it should be noted that the quality control fish oil samples that were analysed together with the human milk samples did not show exceedance of any warning level or control level of the individual 26 PCN congeners.

As a result, based on this quality control sample, the applied method achieved a long-term precision of below 10% over the 2019–2021 period for ΣPCN₁₃ and ΣPCN₂₆.

6.2.4 External Validation

Due to the lack of available proficiency tests at the time of measuring the human milk samples of the 2016–2019 period, an external validation for control of the trueness was performed through an interlaboratory comparison with an independent laboratory. Results of four highly polluted fat samples with incurred PCN and a fortified rape seed oil (ΣPCN₁₃ 195 pg/g lipid) were compared. At CVUA Freiburg, PCN were determined by GC-HRMS (high-resolution mass spectrometry; sector field instrument, R = 10,000) and confirmed by GC-Orbitrap MS (Orbitrap mass spectrometry; R = 60,000) (see Sect. 6.1). The external laboratory also used a sector field mass spectrometer (R = 10,000) for PCN determination. Both laboratories analysed at least 13 of the most frequently reported congeners (see Sect. 6.1). The deviation for the results between the external laboratory and CVUA Freiburg is summarized in Table 17. The deviation of the ΣPCN₁₃ between the external laboratory and CVUA Freiburg was in the range between 3 and 20%.

Table 17 Deviation (%) of the results of ΣPCN₁₃ between CVUA Freiburg and an external laboratory both using high-resolution mass spectrometry (resolution 10,000) for determination of PCN in four contaminated samples (two of them highly) and one fortified sample

The PCN pattern achieved by both labs were in agreement (Fig. 14). An overestimation using GC-Orbitrap for PCN 28/36 and PCN 48 in comparison with sector field detection was recognized and may have been caused by a possible co-elution of interfering compounds or other PCN congeners.

Fig. 14

Comparison of a PCN pattern of a highly polluted human milk sample by CVUA Freiburg using GC-Orbitrap (R = 60,000) and sector field (R = 10,000) and an external lab using sector field (R = 10,000). The congeners PCN 31, PCN 59 and PCN 63 were only determined by CVUA Freiburg

6.2.5 Participation in Interlaboratory Studies

At a later stage, CVUA Freiburg took part in the first interlaboratory comparison study conducted by the EURL POPs in the second half of 2021 on determination of PCN in cod liver oil for 26 PCN congeners (EURL for Halogenated POPs in Feed and Food 2022b). The number of participants in this first interlaboratory study for PCN was lower than for well-established analyte groups. Furthermore, the scope on targeted PCN congeners varied between the participating labs, limiting the possibility of assessment of the comparability for some congeners. Assigned values could be derived for 7 congeners/co-eluting pairs of congeners and two sum parameters ([1] “ΣPCN₁₂ plus PCN 64” comprising the 12 main congeners recommended by the PT provider as initial focus when starting the method development plus PCN 64; [2] ΣPCN₂₆ comprising all 26 congeners) (Table 18). Figure 15 illustrates the z-scores achieved in this interlaboratory comparison using the sector field method. As a conclusion, all z-scores were within ±2 z and therefore deemed satisfactory.

Table 18 Results of the interlaboratory study on the determination of PCN in cod liver oil

Fig. 15

Results of the interlaboratory study on the determination of PCN in cod liver oil. The plot shows z-scores, calculated for a fitness-for-purpose-based standard deviation for proficiency tests σ = 20%. Results within ±2 z-score are deemed satisfactory

Additionally, the deviation of the results to the median was calculated as useful information in particular for congeners/co-eluting pairs of congeners, if no assigned value and thus no z-score could be calculated. This information is seen as meaningful, if results from five or more laboratories for concentrations above the lowest background level (>0.3 pg/g lipid) were available. In general, the results for 26 PCN congeners from CVUA Freiburg were in accordance with the median of the results of the participants (Table 18).

7 Accreditation

In 1993, new quality standards were introduced for laboratories entrusted with the official control of foodstuffs by the Member States of the European Economic Community. Laboratories had to comply with the general criteria for the operation of testing laboratories laid down in European Standard EN 45001 supplemented by standard operating procedures and the random audit of their compliance by quality assurance personnel not later than November 1998 (Council Directive 93/99/EEC). In a revision of the regulations on official controls in 2004, it was stipulated that laboratories that were designated for official control should operate and be assessed and accredited in accordance with the European Standard EN ISO/IEC 17025—“General requirements for the competence of testing and calibration laboratories” (EU Regulation 882/2004). Therefore, the CVUA Freiburg was accredited in 1998 and has since been re-accredited continuously.

As a result, all analyses performed by CVUA Freiburg for determination of PBDE, HBCDD, CP and PCN in human milk of the WHO/UNEP-coordinated exposure studies followed the strict rules of the accreditation system and the general criteria for the operation of testing laboratories as laid down in EN ISO/IEC 17025 (European Standard EN ISO/IEC 17025).

Appendix

Table 19 25 Native and 13 ¹³C₁₂-labelled PBDE. Recommended analytes (BDE-47, BDE-99, BDE-100, BDE-153, BDE-154, BDE-183 and BDE-209) are marked in bold

Table 20 26 native and eight ¹³C₁₀-labelled PCNs. Seven ¹³C₁₀-labelled standards marked in bold were added as internal standards to the lipid aliquot; ¹³C₁₀-labelled PCN 65 was used as the recovery standard and added after clean-up before the final determination step