Report Chlorinated Paraffins Ils Quasimeme 2011

IVM Institute for Environmental Studies

Accredited under No. L476 (RvA)

7

Interlaboratory Study on the Analysis of Chlorinated Paraffins in Environmental Matrices Phase 1

Ike van der Veen IVM Institute for Environmental Studies, VU University, Amsterdam, the Netherlands

Wim Cofino Wageningen University and Research Centre, Alterra, QUASIMEME Laboratory Performance Studies, The Netherlands

Steven Crum Wageningen University and Research Centre, Alterra, QUASIMEME Laboratory Performance Studies, The Netherlands

Jacob de Boer IVM Institute for Environmental Studies, VU University, Amsterdam, the Netherlands

Report W-12/11, version 1 11 September 2012


This report is released by: Prof. Dr. J. de Boer Head of the department C&B

It was internally reviewed by: Prof. Dr. J. de Boer

IVM Institute for Environmental Studies VU University Amsterdam De Boelelaan 1087 1081 HV AMSTERDAM The Netherlands T +31-20-598 9555 F +31-20-598 9553 E [email protected]

Wageningen University and Research Centre, Alterra, QUASIMEME Laboratory Performance Studies, The Netherlands P.O. Box 47 6700 AA Wageningen The Netherlands T +31 (0)317 486546 F +31 (0)317 419000 E [email protected]

Copyright 2012, Institute for Environmental Studies All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photo-copying, recording or otherwise without the prior written permission of the copyright holder


Interlaboratory Study on the Analysis of Chlorinated Paraffins in Environmental Matrices

Contents

Summary 5

1 Introduction 7

2 Materials and methods 9

2.1 Study design 9 2.2 Material preparation 9 2.3 Methods used by participants 9 2.4 Data Assessment 11

3 Results 17

4 Discussion 19

4.1 Laboratory performance 19 4.2 Comparison with other ILSs 22

5 Conclusions 25

Acknowledgements 27

References 29

Appendices 31

Appendix 1 List of Participants 32

Appendix 2 Results and graphical presentation 33

Appendix 3 Numerical z-score values per matrix 38

Appendix 4 Consistency of data 39

Appendix 5 Graphical output of the Cofino Model statistics 40

Appendix 6 Additional method information 44


Interlaboratory Study on the Analysis of Chlorinated Paraffins in Environmental Matrices 5

Summary

The first round of an international interlaboratory study (ILS) on chlorinated paraffins (CPs) was organised by the Institute for Environmental Studies (IVM) in cooperation with the proficiency testing scheme of QUASIMEME. This study was initiated because it was generally agreed at a workshop on the analysis of CPs in Ostend, Belgium, organised by QUASIMEME in March 2010, that there was a clear need for an interlaboratory study, preferably designed in a step-wise way. The first objective of the study was to assess the intercomparability of the data produced and secondly to initiate improvements.

In total 15 laboratories participated in the present study on the analysis of CPs. The participants were requested to quantify the total concentration of CPs and the concentration of the three individual CPs 1,1,1,3,9,10-Hexachlorodecane (1,1,1,3,9,10-HCD) , 1,1,1,3,11,12-Hexachlorododecane (1,1,1,3,11,12-HCDD), and 1,1,1,3,11,13,13,13-Octachlorotridecane (1,1,1,3,11,13,13,13-OCTD) in an iso-octane solution of CPs. Two solutions containing CPs in known concentrations were provided to the participants for calibration purposes. Participants used their in-house quantification methods and techniques. The results were collected and statistically evaluated using the Cofino statistics. Between lab coefficients of variation (CVs) and z-scores were appointed to the laboratorys results as an expression of accuracy.

The majority of the laboratories (57-71%) obtained satisfactory z-scores for the analysis of the three individual CPs. For individual CP analysis the model between lab CVs varied between the compounds from 22 to 46%. For total CP a between lab CV of 56% was found. Since no other intercomparison studies were known for the analysis of individual CPs, no comparison could be made with previous intercomparison studies.

Even though a variety of quantification techniques were used, the difference between the median value and the target value of the total CP concentration has decreased compared to a study of Tomy et al. (1999), in which all participants used the same quantification technique.

The present study also showed that the overall performance of participants in the analysis of total CPs has improved compared to the interlaboratory study of Pellizzato et al. (2009) although the present study only dealt with a solution of CPs, while the study of Pellizzato et al. (2009) dealt with an oil extract.

The overall experience in CP analysis in the present study varied between the participants, but no significant difference could be calculated between the performance of participants with different levels of experience.



1 Introduction

Chlorinated paraffins (CPs), also known as Polychlorinated alkanes (PCAs), are complex mixtures of chlorinated n-alkanes with carbon chain lengths of 10 to 30 and a chlorination degree between 30% and 70% by mass. Characterization of CPs is performed by their alkane chain lengths. They are divided into three groups: short-chain (C10-13) (SCCP), medium-chain (C14-17) (MCCP) and long chain (C18-32) (LCCP) CPs.

CPs are used in several industrial applications like flame-retardants in the rubber industry, as high temperature lubricants and cutting fluids in the metalworking industry and as additives in liquids, in paints and textile. The analysis of CPs is highly complicated. There tens of thousands of congeners which make separation by gas chromatography (GC) and even by GCxGC difficult. Alternative methods are scarce and may suffer from false positive results. Therefore, the data reported on CPs up to the present include very high uncertainties (100% or more).

Chlorinated paraffins are currently of utmost importance to the environmental analyst. They are being produced in high volumes, they are under discussion in the United Nations Environmental Programme (UNEP) for possible definition as a persistent organic pollutant (POP), and they are a mandatory determinand in the European Water Framework Directive. However, their analysis is subject to very large variation, as the CP patterns are extremely complex. In March 2010, QUASIMEME organized a workshop on the analysis of CPs in Ostend, Belgium. A number of critical issues in the analysis of these CPs were discussed. It was generally agreed that there was a clear need for an interlaboratory study, preferably designed in a step-wise way and accompanied by expert comments.

The present study focuses on the analysis of CPs in a solution of undisclosed concentration, to assess the possibility of producing precise and accurate data for a range of CPs in a mixture.

In total 15 laboratories participated in the present ILS, of which 11 submitted data. The ILS study focussed on the CPs listed in table 1.

Table 1 Full names, abbreviations and chemical formulas of individual CPs covered in this study and provided.

Full name Abbreviation Chemical

formula

Chlorination

degree (% Cl)

1,1,1,3,9,10-Hexachlorodecane 1,1,1,3,9,10-HCD C10H16Cl6 61.0

1,1,1,3,11,12-Hexachlorododecane 1,1,1,3,11,12-HCDD C12H20Cl6 56.4

1,1,1,3,11,13,13,13-Octachlorotridecane 1,1,1,3,11,13,13,13-OCTD C13H20Cl8 61.7

This study was carried out by the Institute for Environmental Studies (IVM) in collaboration with QUASIMEME (www.QUASIMEME.org).



2 Materials and methods

2.1 Study design

The aim of the sample preparation was to select individual CPs as representative for CPs in the environment. Unfortunately, only commercial standards with chlorine atoms at the end carbon atoms of the CP backbone were available. Due to this lack of standards, three congeners with CPs on the ends of the backbone were selected.

Three Ampoules were prepared and sent to the participants with the aim of quantifying in triplicate the three individual CPs, mentioned in table 1, and the total amount of CPs in ampoule C by using ampoule A and B (see chapter 2.2).

The first invitation for participation in the study was sent out in November 2010 and the samples were distributed in August 2011.

2.2 Material preparation

Ampoule A

Ampoule A contained an iso-octane solution of three individual CPs of a known identity and in a known concentration (see Table 2).

Table 2 Content of Ampoule A

Abbreviation Concentration (ng/mL)

1,1,1,3,9,10-HCD 1366

1,1,1,3,11,12-HCDD 1389

1,1,1,3,11,13,13,13-OCTD 1009

Ampoule B

Ampoule B contained an iso-octane solution of a technical mixture of SCCPs (C10-C13, 51,5% Cl) in a disclosed concentration (53 g/mL).

Ampoule C

The iso-octane solution in ampoule C consisted of a technical mixture of SCCPs (C10-C13, 55.5% Cl) spiked with the three individual CPs, which are mentioned in table 1. Details on the concentrations of the CPs can be found in Table 3 and in Appendix 2.

2.3 Methods used by participants

A short description of the methods reported by each individual participant is provided in Appendix 6. One participant (CPP15) handed in two data sets, obtained with two different analysis methods (marked with m1 and m2). Although, participants used different detection methods for the analysis of CPs, all participants used a GC method for the separation (see Figure 1). The preferred method was GC-mass spectrometry (GC-MS) (69%), and a few laboratories used GS-MS/MS (23%). Only one participant did not use an MS for the detection. A GCxGC- electron capture detector (ECD) system was used instead. The ion sources used by the participants who did use an MS technique



varied. Almost half of the participants (46%) used chemical ionization (CI), while 31% used electron impact (EI) and 23% used electron capture chemical ionization (ECNI) (see Figure 2). Whenever EI was used the polarity used was positive, while the polarity used with CI and ECNI was negative.

Figure 1 Methods used for analysis of CPs, for details see Appendix 6.

Figure 2 Ionization techniques of the MS methods used.

For the GC separation often a DB-5MS column was used (46%), but also other columns were utilised, such as Rtx-5MS and Rtx-1614 (see Figure 3). Splitless injection was used (69%) as well as on-column injections (23%) and cooled injection with solvent venting (8%).

Figure 3 Columns used for analysis of CPs on a GC,

2.3.1 Quantification

For the quantification of the concentration of total CPs in ampoule C participants were requested to use ampoule B. Since the mixture in ampoule B contained CPs with 51.5% Cl, and the mixture in ampoule C contained 55.5% Cl some participants reported that an in-house mixture was a better match to quantify CPs in ampoule C. Participant CPP4 and CPP8 handed in two data sets. The first data set, marked with m1, was obtained by quantification of total CP with ampoule B and the second data set, marked with m2,

GC-MS69%

GC-MS/MS23%

GCxGC-ECD8%

Analysis method

CI46%

EI31%

ECNI23%

Ionization

DB-5MS46%

Rtx-5MS15%

Restek

15%

Rtx-16148%

Rtx-2008%

DB-1MS x innowax

8%

Column



was obtained with quantification with in-house mixtures. In-house mixtures were also used by participants CPP3 for the quantification of total CPs in ampoule C.

The majority of the participants (64%) used in addition an internal standard, like 13C Chlordane, 13C12-138-HxBDE, d10 anthracene, and 13C HCB, for the quantification of CPs.

2.4 Data Assessment

The data assessment was carried out according to the principles employed in the data assessment of the QUASIMEME laboratory performance studies. All data received from the participants were entered into a database and assessed using a standard procedure enabling direct comparison between participants. The approach to the assessment is based on standard ISO 13528 (Thompson et al., 2006). Additions or differences in the assessment from these standards are given or referred to in this report. However, the assigned value and the laboratory assessment using z-scores is calculated using the model developed by Cofino et al. (2000).

The standard, (ISO 13528), includes statistics for proficiency testing schemes, and uses robust statistics as a basis for the assessment. However, it is generally acknowledged that robust statistics cannot cope with more than 10% extreme values, particularly with a skewed distribution. The Cofino model is able to routinely handle these types of distribution and provide the best estimate of the consensus value, which may be used as the assigned value.

The Cofino model is used routinely for QUASIMEME assessments. In this study the Normal Distribution Approximation (NDA) has been employed (Cofino et al., 2000). This approach includes all data in the evaluation and no subjective truncation or trimming is made. This model has been further developed to include Left Censored Values (LCV, which is the correct nomenclature for less than values) The development of these models has been fully documented and published (Cofino et al., 2000, Cofino, et al., 2005, Wells, et al., 2004) and accepted by the Dutch Accreditation Body. An overview of the assessment with explanation and examples is given in the Assessment Rules for the Evaluation of the QUASIMEME LP Studies Data (Wells and Scurfield, 2004).

The details of the Cofino Model were provided elsewhere (Wells et al., 2004, Wells and Scurfield, 2004), but in summary the approach is as follows:

All data included in the assessment; No data trimmed or down weighted; Assigned values (AV) based on Cofino NDA model; All LCV are also included, provided certain criteria are met (see chapter 2.4.2).

2.4.1 Plots

The performance of the laboratories in this study is illustrated in the z-score histograms in Appendix 5, and an example of the plots is shown in Figure 4. Where the assigned value for a determinand is indicative, the values are plotted as their original reported concentrations. The rules for confirming whether the consensus value should be an assigned value or an indicative value are given in the Assessment Rules for the Evaluation of the QUASIMEME LP Studies Data (Wells and Scurfield, 2004) with appropriate examples.

Normally, four plots are given for each determinand (see Figure 4). The upper left plot provides an impression of the probability density function (PDF) model for all data (black) and for the first mode (PMF1) model of the data (blue dotted). Superimposed on



these PDFs is a histogram of the individual measurements (grey bars). This plot shows the distribution of the data as a whole, and of the data in the main mode (PMF1) model on which the assigned value is based.

Figure 4. Examples of the graphical output of the Cofino Model statistics for

1,1,1,3,11,13,13,13-OCTD in ampoule C. The Kilt Plot (Overlap Matrix) (upper right plot) provides an overview of the degree of overlap of each pair of data. It gives a clear indication of the degree of homogeneity of the data. As a key, the white areas indicate maximum overlap of the PDFs and, therefore, highest agreement (an overlap of one implies that the two laboratories of the pair report exactly the same results), while the black area show the pairs in poor agreement.

The lower left plot is a ranked overview of all data with an error bar of 2 SD. The numerical values are given in blue and the LCVs are given in red.

The ranked z-score plot (lower right) is based on the mean of the data, which is normally also the assigned value. However, if there is any adjustment required to the assigned value as a result of the assessment, e.g. use of the nominal concentration or a trimmed value, then the final z-score given in the z-score histograms will reflect these changes.

2.4.2 The assigned values and indicative values

The assigned value (AV) is obtained from the main mode model of the data using the Cofino Model (bleu dotted line in upper left panel in Figure 4), and is centred around the highest density of values. Unless otherwise stated, the assigned value is based on this consensus value of all data. Although all data are included in the assessment, those values that lie some distance from AV contribute less to the mean than values which occur at or near the mean.



In some instances it is not possible to set an AV, and an indicative value is given. No assessment of laboratory performance is given where an indicative value is set. An overview of the assessment, with explanation, decision flowcharts and examples, is given in the paper Assessment Rules for the evaluation of the QUASIMEME Laboratory Performance Studies Data, available on the QUASIMEME website (www.quasimeme.org). A summary of the categories is given below:

Category 1

For data with the number of numerical observations 7

An assigned value is based on the mean when 33% of values have a z-score of |Z| < 2. Where < 33% of the data has |Z| < 2 the value is indicative. i.e. at least 33% must be in good agreement.

Category 2

For data with the number of numerical observations > 3 and < 7

An assigned value is based on the mean when 70% of values have a z-score of |Z| < 3 and a minimum of 4 observations have |Z| < 2. Otherwise the value is indicative. i.e. for small datasets, n > 3 and n < 7, there need to be very good agreement and a maximum of one extreme value before an assigned value can be given.

Category 3

For data with the number of numerical observations < 4

No assigned value is given. Normally the median value is given as an indicative value.

Category 4

For data with the high Total Error% >100% in combination with bad performance

No assigned value is given.

2.4.3 The Z-score Assessment

A z-score (Thompson and Wood, 1993) is calculated for each participants data for each matrix / determinand combination which is given an assigned value. The z-score is calculated as follows:

z - score = Mean from Laboratory - Assigned Value

Total Error

It is emphasized that in many interlaboratory studies the between-laboratory standard deviation obtained from the statistical evaluation of the study is used as total error in the formula above. In the QUASIMEME data assessment, the total error is estimated independently taking the needs of present-day international monitoring programs as starting point. For each determinand in a particular matrix, a proportional error (PE) and a constant error (CE) have been defined. The total error depends on the magnitudes of these errors and on the assigned value:

Total Error = Assigned Value x Proportional Error (%)

100 + 0.5 x Constant Error

The values for the PE and CE are set by the QUASIMEME Scientific Assessment Group and are monitored annually. The values are based on the following criteria:

- Consistency of the required standard of performance to enable participating laboratories to monitor their assessment over time.



- Achievable targets in relation to the current state of the art and the level of performance needed for national and international monitoring programmes.

The assessment is based on the standard ISO 17043 and z-scores. The QUASIMEME model is designed to provide a consistent interpretation over the whole range of concentration of analytes provided, including an assessment where LCVs are reported.

The proportional error is set at 12.5% for all matrices. This applies to all determinands. The constant error (CE) has been set for each determinand or determinand group. This value was initially set to reflect the limit of determination, but is at present more closely related to the overall laboratory performance. The magnitude of the CE is set to provide a constant assessment in terms of z-score regardless of concentration. The CE is set at 0.025. Therefore at low concentrations the level of accuracy required to obtain a satisfactory z-score is less stringent than at high concentrations.

Following usual practices e.g. ISO 17043, the z-scores can be interpreted as follows for laboratories which take part in QUASIMEME to assure the quality of their data for use in international marine monitoring programmes:

|Z| < 2 Satisfactory performance

2 3 Unsatisfactory performance

|Z| > 6 frequently points to gross errors (mistakes with units during reporting, calculation or dilution errors, etc.).

The following schematic presentation illustrates the interpretation of the z-scores:

It is not possible to calculate a z-score for LCVs as LCVs represent a cut-off value rather than continuous data. However, Quasimeme provides a simple quality criterion:

LCV/2 < (concentration corresponding to |z|=3): LCV consistent with assigned value

Questionable

Assigned Value (AV)

AV - 3*TE

AV - 2*TE AV+2*TE

AV+3*TE

Satisfactory performance Unsatisfactory performance performance

performance

TE : total error

Assigned Value (AV)

AV - 3*TE

AV - 2*TE AV+2*TE

AV+3*TE

Unsatisfactory

TE : total error



LCV/2 > (concentration corresponding to |z|=3): LCV inconsistent with assigned value, i.e. LCV reported by laboratory much higher than numerical values reported by other laboratories.

Z-score key: S Satisfactory

Q Questionable

U Unsatisfactory

LCV key: C Consistent

I Inconsistent

No data: B - Blank



3 Results

The submitted results have been evaluated statistically and whenever the data met the requirements (as described in chapter 2), an assigned value was established. Z-scores were calculated based on the assigned value. Summary statistics are presented in Table 3. A summary of the assigned values and the percentage of satisfactory to unsatisfactory z-scores are presented in Table 4. Whenever less than values (LCV) were submitted, the percentage of consistent and inconsistent LCVs with the assigned value is given. The submitted data is presented in Appendix 2. Tables with individual z-scores are presented in Appendix 3, consistencies of the individual results are presented in Appendix 4 and z-score plots in Appendix 5.

Table 3 Summary of results of CPs in ampoule C (results in ng/mL).

Determinand Target value

Assigned value

Indicative Value

Model Mean Median Min* Max**

Model Between Lab CV%

Percentage in PMF1 n>LOQ

1,1,1,3,9,10-HCD 129 138 NA 138 148 51 810 22 57 7

1,1,1,3,11,12-HCDD 132 177 NA 177 197 122 2800 46 76 7

1,1,1,3,11,13,13,13-

OCTD 96 111 NA 111 120 90 1500 28 78 7

Total CP 12000 NA 18500 18180 18500 8100 49300 56 76 14

* Min: lowest value submitted > LOQ ** Max: highest value submitted > LOQ

Table 4 Summary of laboratory performance for CPs in ampoule C (results in ng/mL).

Determinand Assigned value Indicative Value

Total Error %

% data received

% |Z||Z|>2 (Questi-onable)

% 6>|Z|>3 (Unsatis-factory)

% |Z|>6 (Extreme)

1,1,1,3,9,10-HCD 138 NA 12.9 50 57 0 14 29

1,1,1,3,11,12-HCDD 177 NA 12.8 50 57 14 14 14

1,1,1,3,11,13,13,13-

OCTD 111 NA 13.0 50 71 14 0 14

Total CP NA 18500 0 100 0 0 0 0



4 Discussion

In total 15 laboratories from all over the world participated in the present study. Of those laboratories, 11 submitted data, of which three laboratories handed in two data sets obtained with either two different analyses methods(see chapter 2.3), or with different calibration solutions (see chapter 2.3.1). Seven participants were already working on CP analysis for over 3 years, while two laboratories were analysing CPs between 1 and 3 years. Only one laboratory had an experience in CP analysis of shorter than one year. The experience of one participant was unclear. No significant difference is observed between the reported concentrations of the less experienced participants and the more experienced participants (Figure 5).

Figure 5 Experience of participant vs total CP concentration reported.

4.1 Laboratory performance

As shown in Table 3, the model between lab CVs varied between the compounds from 22-56%. Satisfactory z-scores (i.e. Z|2|) were obtained from the majority (57-71%) (depending on the determinand) of the participants that submitted results (see Table 4).

Comparing the assigned values (or indicative value in case of total CP) with the target values (see Table 3 and Figure 6) shows that all assigned values (and indicative values) are 1.11.5 fold higher than the target values, which might be due to coeluting CPs, since the deviation is not observed by Participant CPP3 who used high resolution (HR) MS.

Due to the deviation between assigned value and target value some participants, who reported concentrations which agreed well with the target concentrations, received higher, though still satisfactory, z-scores. For example participants CPP3, CPP10 and CPP15m2 received z-scores of -1.9, -2.0 and -1.8 respectively for 1,1,1,3,11,12-HCDD,

0

5000

10000

15000

20000

25000

30000

35000

40000

45000

50000

(ng

/mL)

Participating laboratories

Total CP

Target value

Indicative value

experience> 3 years

experience 1-3 years

experienceunknown

experience < 1 year


20 Discussion

while the reported concentrations were 135, 131, and 136 ng/mL and the target value was 132 ng/mL.

0

50

100

150

200

250

300

(ng

/mL)


1,1,1,3,9,10-HCD (n=21)

Target value

Assigned value

0

50

100

150

200

250

300

(ng

/mL)


1,1,1,3,11,12-HCDD (n=21)

Target value

Assigned value



Figure 6 Comparison of assigned values (or indicative value) with the target values.

In Figure 7 a comparison is given between the different ionization techniques used and the reported concentration of total CP. Although only limited data is available per method and no hard conclusions can be made, it looks like EI with positive polarity gives results, which are closer to the target value. Only one participant reported results on the concentrations of total CPs obtained with ECD, which resulted in a deviation of only 4.7% of the target value. Two of the participants, marked with a asterisk in figure 7 used HR MS.

0

20

40

60

80

100

120

140

160

180

200(n

g/m

L)


1,1,1,3,11,13,13,13-OCTD (n=21)

Target value

Assigned value

0

5000

10000

15000

20000

25000

30000

35000

40000

45000

50000

(ng

/mL)


Total CP (n=42)

Target value

Indicative value


22 Discussion

Figure 7 Methods used by participant vs total CP concentration reported.

4.2 Comparison with other ILSs

No previous intercomparison studies are known on the analysis of individual PCAs. In an interlaboratory study of Tomy et al. (1999) two SCCP containing solutions (PCA-1 and PCA-70) were quantified for the total SCCP concentration by 7 participants who all used HRGC-ECNI-MS. Concentrations of the mixtures in that study were 6-10 fold higher than the concentration of the total CP concentration in ampoule C of the present study. Target CP concentrations of the mixtures were 74 and 118 ng/L respectively, while the means of the determined concentrations were 99.3 and 297 ng/L respectively, which is a difference of 34 and 152%. CVs were 20 and 44%. The model between lab CV for the total CP determination in the present study was 56%, which is higher than the CVs of the study of Tomy et al. (1999). The difference between the mean of the reported concentrations and the true value in the present study is 35 %, which is equal to the performance on PCA-1 but better than the performance on PCA-70 of the study of Tomy et al. (1999). In the study of Tomy et al. (1999) all participants used the same quantification technique, while in the present study a variety of techniques is used, which might explain the higher variance in the present study.

In another intercomparison study (Pellizzato et al., 2009) the total SCCP concentration was determined in an extract of industrial oil. Reported concentrations varied between 8.5 and 3200 mg/L, with 4 of the 6 participants reporting concentrations between 8.5 and 16.1 mg/L, which are in the same range as the concentrations of total CPs in ampoule C of the present study. In the study of Pellizzato et al. (2009) the CV was 209%. When the results of one participant were excluded, the CV decreased to 82%. The better performance in the present study compared to the study of Pellizzato et al. (2009) is most likely due to the fact that a standard solution is cleaner than an oil extract.

Possible source for data variance

Although participants were requested to use Ampoule A and B for the quantification of total CPs in ampoule C, some participants used a better fitting in-house mixture for

0

5000

10000

15000

20000

25000

30000

35000

40000

45000

50000

(ng

/mL)


Total CP

Target value

Indicative value

EI positiveCI negativequantified with in-house std.

CI negativequantified with ampoule B

ECNI ECD

Target value

Indicative value

EI positiveCI negativequantified with in-house std.

CI negativequantified with ampoule B

ECNI ECD

HR MS

HR MS



the quantification (see chapter 2.3.1), which resulted in a better performance of those participants and a lower model between lab CV (see Figure 8).

Figure 8 Quantification solution used by participant vs total CP concentration reported.

CPP4 M1

CPP8 M1CPP4 M2

CPP8 M2

0

5000

10000

15000

20000

25000

30000

35000

40000

45000

50000

(ng

/mL)


Total CP

Target value

Indicative value

Quantiification with ampoule B Quantiification with in-house mixture



5 Conclusions

For individual CP analysis the between lab CVs varied between 22 and 46%. Since no other intercomparison studies were known for the analysis of individual CPs, no comparison could be made with previous intercomparison studies. For total CP a between lab CV of 56% was found.

A larger number of laboratories participated in this international interlaboratory study than in previous intercomparison studies on total SCCP, showing that an increasing interest in quantitative analysis of CPs, as well as the intention to evaluate laboratory performance so as to improve the data produced in the field. This study showed that even though a variety of quantification techniques were used the difference between the median value and the target value has decreased compared to a study of Tomy et al. (1999), in which all participants used the same quantification technique.

The present study also suggests that the overall performance of participants in the analysis of total CPs has improved compared to the interlaboratory study of Pellizzato et al. (2009), although the present study only dealt with a solution of CPs, while the study of Pellizzato et al. (2009) dealt with an oil extract.

Obviously, further developmental work in laboratories and more interlaboratory comparison exercises are needed to improve the analysis of CPs. Analytical standards of CP congeners that occur in environmental samples are badly needed.

A second step of this exercise including analysis of clean extracts will be organised shortly.



Acknowledgements

The Chlorinated Paraffins Sector Group of Eurochlor is gratefully acknowledged for their financial support of this interlaboratory study, and their technical input.



References

Cofino, W.P., Stokkum, I.H.M. van, Steenwijk, J. Van & Wells, D.E. (2005) A new model for the inference of population characteristics from experimental data using uncertainties. Part II. Application to censored datasets. Anal. Chim. Acta, 533, 31-39.

Cofino, W.P., Wells, D.E., Arise, F., Stokkum, I. van , Wegener, J.W. & Peerboom, R. (2000). A new model for the inference of population characteristics from experimental data using uncertainties. Application to interlaboratory studies. Chemometrics. Intell. Lab. Syst. 53, 37-55.

Pellizzato, F., Ricci, M., Held , A., Emons, H., Bhmer, W., Geiss, S., Iozza, S., Mais, S., Petersen, M. & Lepom, P. (2009). Laboratory intercomparison study on the analysis of short-chain chlorinated paraffins in an extract of industrial soil. Trends Anal. Chem. 28, 1029-1035.

Thompson, M., Ellison, S.L.R. & Wood, R. (2006). The international harmonised protocol for proficiency testing of analytical chemistry laboratories. IUPAC Technical report. Interdivisional working party for harmonization of quality assurance schemes. Pure Appl. Chem. 78 , 145-196.

Thompson, M.& Wood, R. (1993). International harmonised protocol for proficiency testing of (chemical) analytical laboratories. J. AOAC Int, 76, 926-940.

Tomy, G.T., Westmore, J.B., Stern, G.A., Muir, D.C.G. & Fisk, A.T. (1999). Interlaboratory study on quantitative methods of analysis of C10-C13 polychloro-n-alkanes. Anal. Chem., 71, 446-451.

Wells, D.E., Cofino, W.P. & Scurfield, J.A. (2004). The application of the Cofino model to evaluate laboratory performance study data using bandwidth estimator. UK. Report nr. 04/04. FRS Marine Laboratory, 375 Victoria Road, Aberdeen 388 AB11 9DB.

Wells, D.E. & Scurfield, J.A. (2004). Assessment rules for the evaluation of the QUASIMEME laboratory performance studies data. UK, version 2. FRS Marine Laboratory, 375 Victoria Road, Aberdeen AB11 9DB, pp.390 -391.



Appendices

1. List of participants 2. Results and graphical representation 3. Numerical z-score values per matrix 4. Consistency of data 5. Graphical output of the Cofino Model statistics 6. Additional method information


32

Appendix 1 List of Participants



Appendix 2 Results and graphical presentation

Participant code:

Date Received:

Date Analyzed:

Weight Received:

Weight Analyzed:

Concentration in ng/mL Target value

Assigned value

Indicative value

Model Mean Median MIN MAX

Model Between Lab CV%

Model percentage

in PMF1 n>LOQ

1,1,1,3,9,10-HCD 129 138 NA 138 148 51 810 22 57 7

1,1,1,3,11,12-HCDD 132 177 NA 177 197 122 2800 46 76 7

1,1,1,3,11,13,13,13-OCTD 96 111 NA 111 120 90 1500 28 78 7

Total CP 12000 NA 18500 18180 18500 8100 49300 56 76 14

ND: not detected

NA: not analyzed


34

Participant code: CPP1

1 CPP1

2 CPP1

3 CPP2

1 CPP2

2 CPP2

3 CPP3

1 CPP3

2 CPP3

3

CPP4m1 1

CPP4m1 2

CPP4m1

3

CPP4m2

1

CPP4m2

2

CPP4m2

3 Date Received: 08/08/2011

02/08/2011 04/08/2011 04/08/2011

Date Analyzed: 10-14/10/2011

12/08/2011 11/09/2011 11/09/2011

Weight Received: 7.0712g 7.3768g 7.1024g

7.2569g 7.2209g 7.3395g 7.2241g 7.2393g 7.0773g 7.2241g 7.2393g 7.0773g

Weight Analyzed: 7.0716g 6.8108g 7.1030g

7.2566g 7.2214g 7.3400g - 200 l 200 l - 200 l 200 l

Concentration in ng/mL 1,1,1,3,9,10-HCD NA NA NA NA NA NA 138 132 139 NA NA NA NA NA NA 1,1,1,3,11,12-HCDD NA NA NA NA NA NA 135 130 139 NA NA NA NA NA NA 1,1,1,3,11,13,13,13-OCTD NA NA NA NA NA NA 91 96 94 NA NA NA NA NA NA Total CP 29600 28000 29300 NA NA NA 11400 12200 12050 49300 48500 40300 21600 20300 17100

ND: not detected NA: not analyzed


1 CPP5

2 CPP5

3 CPP6

1 CPP6

2 CPP6

3 CPP7

1 CPP7

2 CPP7

3

CPP8m1 1

CPP8m1 2

CPP8m1

3

CPP8m2

1

CPP8m2

2

CPP8m2

3 Date Received:

04/08/2011 05/08/2011 05/08/2011

Date Analyzed:

01/11 27/10 28/10/2011

28/10/2011 12/09/2011

Weight Received:

1.2267g 1.4115g 1.2328g 7.2778g 7.3531g 7.1863g 7.2778g 7.3531g 7.1863g

Weight Analyzed:

0.3597g 0.5237g 0.3420g 2 l 2 l 2 l 2 l 2 l 2 l

Concentration in ng/mL 1,1,1,3,9,10-HCD NA NA NA NA NA NA 51 51 60 640 810 700 NA NA NA

1,1,1,3,11,12-HCDD NA NA NA NA NA NA 220 210 210 2600 2700 2800 NA NA NA

1,1,1,3,11,13,13,13-OCTD NA NA NA NA NA NA 110 130 120 1400 1500 1300 NA NA NA

Total CP NA NA NA NA NA NA 12000 12000 12000 18000 17000 17000 8700 8100 8200





1 CPP9

2 CPP9

3 CPP10

1 CPP10

2 CPP10

3 CPP11

1 CPP11

2 CPP11

3 CPP12

1 CPP12

2 CPP12


01/08/2011

Date Analyzed: 20/10/2011

22/09/2011 14/11/2011

Weight Received: 7.2042g 7.3329g 7.2475g

1 ampoule 1 ampoule

1 ampoule

Weight Analyzed: 7.2042g 7.3329g 7.2475g

1 ampoule 1 ampoule

1 ampoule

Concentration in ng/mL 1,1,1,3,9,10-HCD NA NA NA 126.6 123.5 126.6 NA NA NA NA NA NA

1,1,1,3,11,12-HCDD NA NA NA 129.2 131 133.3 NA NA NA NA NA NA

1,1,1,3,11,13,13,13-OCTD NA NA NA 92.20 97.02 97.9 NA NA NA NA NA NA

Total CP 31900 33500 33080 22600 23130 23030 10300 11000 11100 18000 26000 28000



1 CPP13

2 CPP13

3 CPP14

1 CPP14

2 CPP14

3

CPP15m1 1

CPP15m1 2

CPP15m1

3

CPP15m2

1

CPP15m2

2

CPP15m2


01/08/2011 01/08/2011

Date Analyzed: 15/11 11/11 12/11

13/10/2011 13/10/2011

Weight Received: 7.2471 g 7.3842 g 7.1574 g

Weight Analyzed: 1.3518 g 1.4382 g 1.3758 g

Concentration in ng/mL 1,1,1,3,9,10-HCD 145.2 147.0 152.95 NA NA NA 308 308 301 153 143 159 1,1,1,3,11,12-HCDD 292.4 270.8 254.2 NA NA NA 196 200 195 147 122 139

1,1,1,3,11,13,13,13-OCTD 153.0 133.8 148.3 NA NA NA 118 124 123 101 90 101

Total CP 11942 12399 11504 NA NA NA 11200 11200 11900 31200 30200 31600



36


38

Appendix 3 Numerical z-score values per matrix



Appendix 4 Consistency of data

S = Satisfactory; Q = Questionable; U = Unsatisfactory; B = Blank


40

Appendix 5 Graphical output of the Cofino Model statistics


42


44

Appendix 6 Additional method information


46

Documents

Report Chlorinated Paraffins Ils Quasimeme 2011