Proficiency tests for contaminants in food and herbal medicine in the Asia Pacific region

Trends

Trends in Analytical Chemistry, Vol. 29, No. 6, 2010

Proficiency tests for contaminants in food and herbal medicine in the Asia Pacific region Kimmy M. Chan, Samuel T.C. Cheung, Yee-Lok Wong, Amos L.S. Cheng, Chuen-shing Mok, Wang-wah Wong, Dan W. Tholen, Yiu-chung Wong Proficiency-test (PT) programs organized by Asia Pacific Laboratory Accreditation Cooperation have been recognized as supporting mutual recognition arrangement amongst member laboratories for more than 15 years. Responding to the needs of laboratories in the region, several recent programs have had specific focus on food and herbal medicine testing. This article describes the overall performance of participating laboratories and the operation of three related PTs for trace elements, organochlorine pesticides and veterinary drugs. We also discuss the effectiveness of the PT programs and the assessment trends in PTs. Crown Copyright ª 2010 Published by Elsevier Ltd. All rights reserved. Keywords: Asia Pacific; Contaminant; Food; Herbal medicine; Laboratory performance; Organochlorine pesticide; Performance assessment; Proficiency test; Trace element; Veterinary drug

1. Introduction Kimmy M. Chan, Samuel T.C. Cheung, Yee-Lok Wong, Amos L.S. Cheng, Chuen-shing Mok, Yiu-chung Wong* Analytical and Advisory Services Division, Government Laboratory, Hong Kong Wang-wah Wong Hong Kong Accreditation Service, Innovation and Technology Commission, Hong Kong Dan W. Tholen Statistical Consulting, Traverse City, Michigan, MI 49686, USA

*

Corresponding author. Tel.: +852 2762 4042; E-mail: [email protected]

562

Asia Pacific Laboratory Accreditation Cooperation (APLAC) is a cooperation of accreditation bodies in the Asia Pacific region that accredits laboratories, inspection bodies and reference-material producers [1]. It is one of five specialist regional bodies under the framework of Asia Pacific Economic Cooperation (APEC) and also one of the three recognized regional cooperation bodies [together with European Accreditation (EA) and Inter-American Accreditation Cooperation (IAAC)] of International Laboratory Accreditation Cooperation (ILAC) [2]. ILAC itself is the worldÕs principal international organization for the development of laboratory-accreditation practices and procedures, the promotion of laboratory accreditation as a trade-facilitation tool and the recognition of competent test facilities around the globe. One of APLACÕs primary roles is to organize proficiency tests (PTs) in the region so as to strengthen technical

competence for member laboratories and to support development of mutual recognition arrangement (MRA). To achieve and to maintain such an important role, the APLAC PT Committee was founded in early 1994. The responsibilities of the PT Committee are to oversee all work in relation to the conduct of PT programs and measurement audits via the arrangement of workshops, seminars and training programs. Since their first meeting convened in October 1994, regular meetings have been held in conjunction with APLAC General Assemblies. To initiate APLAC PT programs, accreditation bodies are required to submit proposals to the PT Committee for review and approval. The proposed program should comply with management and technical criteria stipulated in APLACÕs and relevant international standards for PT. It should contain the objective and adequate scientific data to support the design and the operation of the program. Priority will be given to those that have not been organized by other scheme providers but have potential demands from users. Upon receiving a proposal from an accreditation body, the PT CommitteeÕs Chairman will circulate the document to members for comment and endorsement. The organizing accreditation body will then revise the protocol, if necessary, to address the comments and the feedback received. The cost for material preparation, transport of samples and relevant analytical work involved in the program is borne by the organizing accreditation

0165-9936/$ - see front matter Crown Copyright ª 2010 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.trac.2010.02.014

Trends in Analytical Chemistry, Vol. 29, No. 6, 2010

body, though partial funding is normally provided. Since APLACÕs programs are organized by different accreditation bodies, the majority are on ad hoc basis. Since its first PT program held in May 1994, APLAC has organized about 100 programs on calibration and measurement. More than 70 of the completed programs were in the chemical-measurement field, covering different test parameters over a wide variety of matrices (e.g., childrenÕs products, commodities, construction materials, environmental samples, food, herbal materials and pharmaceuticals). The programs share the ultimate target of establishing mutual agreement on the equivalence of the operation of member laboratories in the region and hence facilitating the removal of technical trade barriers related to testing activities. The programs also provide an invaluable tool for the comparability of testing and the consolidation of laboratory accreditation under ISO/IEC 17025 [3]. Several food problems in the region (e.g., Sudan dyes in duck eggs [4], malachite green (MG) in fish [5] and melamine in milk products [6]) have had a severe impact on food-safety and public-health issues. However, the rapid growth of herbal-medicine trade [7], especially in the Asia Pacific region, has resulted in the enactment of contemporary regulations for import and export of medicinal commodities (e.g., control of herbs arising from the contamination of residual pesticides and toxic elements [8–11] during cultivation). As a consequence, quantitative measurement of contaminants in food and herbal materials has become routine monitoring work in regulatory authorities and testing laboratories. In view of the current trend, APLAC has been organizing a number of PT programs on contaminants in food and herbs in order to reveal the competence of participating laboratories involved in testing. This article presents the general operation procedures, the associated quality requirements and performance assessment of APLAC PT programs. As illustrative case studies, we report on three recently completed PTs on contaminants in food and herbal-medicine samples. These include residual organochlorine pesticides in ginseng root, heavy metals in a herbal plant and MG in eel muscle. Finally, we discuss the use of assigned values and deviation from target performance in assessment in PT programs.

2. Infrastructure of APLAC PT programs The PT programs provide analytical forums for the comparability of calibration and measurement, and build up participantsÕ ability to take appropriate corrective actions where technical deficiencies are found. They also provide a flow of know-how between their accreditation bodies and establish a mechanism for achieving high levels of technical performance. The

Trends

APLAC PT Committee regularly evaluates the effectiveness of programs so that they fully meet the objectives of supporting removal of technical barriers to trade and underpinning testing and measurement activities. All chemical-testing programs are conducted in accordance with the requirements in APLAC PT002 [12], ISO Guide 43-1 [13] or ILAC Guide 13 [14]. (Note: ISO Guide 43 is being replaced by the new ISO/IEC 17043 [15] standard in 2010). As shown in Table 1, a number of programs encompassing a wide range of test parameters in various food and herbal matrices were organized by different accreditation bodies over the past few years. The duration of the programs (from invitation to issuance of final report) varied from four months to over a year, depending on the analysis time given and the complexity involved in the coordinators preparing materials. The number of participating laboratories in the programs showed that the involvement was consistently high.

2.1. General operation procedures The organizing accreditation body invites national accreditation bodies within APLAC and the invitation usually extends to the two regional accreditation bodies (EA and IAAC) as well as ILACÕs unaffiliated accreditation bodies. Invited accreditation bodies receive an information protocol that details the sample nature and analytes to be tested, the maximum number of participating laboratories, and statistical methods of assessing participantsÕ performance. For participantsÕ reference, it also contains a confirmed schedule (timeline for homogeneity test, stability test, sample dispatch, resultsubmission deadline, and issuance of interim and final reports). In addition, methods for the determination of assigned values and performance-standard deviation should be clearly defined before the program commences. Accreditation bodies are requested to pass the information to their interested accredited laboratories and to notify the organizer of the particulars of their nominees. After receiving analytical results from participants, the organizing accreditation body prepares and issues an interim report. Participants should check the correctness of their submitted data in the interim so as to avoid erroneous statistical analysis in the final report. A draft report (e.g., containing the assessment results and technical commentary on the sources of error, method effects, and overall performance) is then submitted to the PT Committee for review. The organizing accreditation body should respond with sound explanations on the comments and questions raised by the PT Committee and, if necessary, incorporate them into the revised report. Upon PT Committee approval, a final report is distributed to all parties concerned and posted on the APLAC website [1] for

http://www.elsevier.com/locate/trac

563

Trends

Trends in Analytical Chemistry, Vol. 29, No. 6, 2010

Table 1. APLAC PT programs for food and herbal medicine Program T004: Benzoic acid in oyster sauce T007: Fat protein, moisture and ash in whole milk powder T009: Arsenic, cadmium, lead, mercury, zinc and tin in fish flesh T021: Methanol and sulfur dioxide in alcoholic beverage T025: Ash, moisture, protein, copper and preservatives in wheat flour T028: BHC in egg powder T029: Protein, carbohydrate, fat, moisture, ash and energy in meat paste T030: Coliforms and aerobic counts in food T032: Moisture, ash, fat and protein in milk powder T034: Genetically-modified organisms in soybean powder T035: Benzoic acid, sorbic acid and sodium saccharin in beverage T036: Sulfonamides in degreased dry sheep liver powder T037: Protein, ash, total fat, dietary fiber, energy and carbohydrate in rice flour T041: E. coli, Enterobacter aerogenes, and Enterococcus faecalis in reference sample T043: Cadmium and lead in herbal medicine T046: Aerobic plate count, Listeria monocytogenes and Vibrio parahaemolyticus T047: Animal materials in feedstuff T048: Beef veterinary drug residues T049: Organochlorine pesticides in ginseng root T050: Nitrofuran metabolites in prawn T056: Pesticide residues in rice T057: Total arsenic, cadmium and lead in seawater shrimp T058: Malachite green and leucomalachite green in swamp eel T059: Organochlorine pesticides in ginseng root T065: Cadmium and lead in herbal medicine T069: Melamine in fish feed T071: Melamine in milk

Date started/completed

Organizers

No. of participants

Nov 96/Jul 97 Jun 97/Dec 97

GLHK & HKAS, Hong Kong IANZ, New Zealand

143 85

Dec 97/Jul 98

NATA, Australia

104

May 05/Feb 06

TAF, Taiwan

49

Dec 00/Dec 01

SAC, Singapore

97

Aug 01/Mar 03 Jul 01/Nov 01

NABL, India NATA, Australia

72 101

Jun 02/Jan 03 Jun 02/Sep 02 Sep 03/May 04

CNAL, China NATA, Australia CNAL, China

117 126 55

May 03/Aug 04

CNAL, China

83

Mar 03/Nov 04

CNAL, China

40

May 03/Nov 03

NATA, Australia

89

May 04/Oct 04

NATA, Australia

38

Jun 05/Jan 06 Apr 05/Dec 05

GLHK & HKAS, Hong Kong CNAS, China

38 25

Nov 05/Jul 06 Jun 05/Nov 05 Feb 06/Jun 06 Jun 07/Nov 08 Aug 07/Dec 08 Nov 06/Jul 07

CNAS, China CNAS, China GLHK & HKAS, Hong Kong CNAS, China CNAS, China GLHK & HKAS, Hong Kong

21 26 70 37 45 103

May 07/Feb 08

GLHK & HKAS, Hong Kong

48

Dec 07/Jun 08 Jul 08/Oct 08 May 09/Sep 09 Jul 09/Nov 09

GLHK GLHK GLHK GLHK

& & & &

HKAS, HKAS, HKAS, HKAS,

Hong Hong Hong Hong

Kong Kong Kong Kong

55 109 52 76

CNAL, China National Accreditation Board for Laboratories; CNAS, China National Accreditation Service; HKAS, Hong Kong Accreditation Service; GLHK, Government Laboratory of Hong Kong; IANZ, International Accreditation New Zealand; NABL, Accreditation Board for Testing and Calibration Laboratories; NATA, National Association of Testing Authorities; SAC, Singapore Accreditation Council; TAF, Taiwan Accreditation Foundation.

membersÕ perusal. During the course of any programs, organizing accreditation bodies are responsible for keeping participantsÕ identities confidential. 2.2. Quality-assurance requirements and performance assessment To ensure the prepared material is suitable for use in a PT program, the homogeneity and the stability conditions of analytes must be confirmed. In accordance with ISO Guide 43-1 [13] and ILAC Guide 13 [14], at least 10 samples are selected at random for the homogeneity test. Analysis is performed in duplicate using repeatable conditions (e.g., same instrument, same method and same 564

http://www.elsevier.com/locate/trac

operator) and should be completed within the shortest possible time. Analytical results are subject to valid mathematical treatments (e.g., analysis of variance (ANOVA) or other methods recommended in ISO13528 [16], a document that complements ISO G43-1 [13]). The results of statistical analysis should indicate no significant effect attributable to variability amongst the PT samples. Stability of test materials at storage temperature or other conditions should be shown to be acceptable over the time that elapses from dispatch of samples to result submission by participants. Unless otherwise approved by the PT Committee, participantsÕ performance is assessed using a z-score

Trends in Analytical Chemistry, Vol. 29, No. 6, 2010

Trends

approach, which is a convenient, internationally accepted way to express participantsÕ performance in PT programs, and calculated by the following Equation (1): z ¼ ðxi  xÞ=r

ð1Þ

where xi is the arithmetic mean for individual participant, x is the assigned value and r is an estimate of the spread of results (or target standard deviation). Interpretation of z-score in ISO Guide 43-1 [13] and ILAC Guide 13 [14] is as follows: jzj 6 2 Satisfactory 2 < jzj < 3 Questionable jzj P 3

Unsatisfactory

For those who achieve |z| scores in the range 2–3, organizers will recommend them to conduct a thorough review (e.g., analytical procedures and data treatment) in order to find out the cause of discrepancy from other participants. As a general rule, it is mandatory for participants having |z| P 3 for any test to take corrective actions. The corrective action, a responsibility between the participating laboratory and its accreditation body, should be undertaken soon after the release of the final report – and it might vary from a discussion with the laboratory to a withdrawal of accreditation for the tests involved.

3. PT programs for contaminants in food and herbal medicine Amongst the recently completed APLAC PT programs, we selected the programs on trace metals in a herbal plant (T043), organochlorine pesticides in ginseng (T059) and MG in swamp-eel muscle (T058) as working examples to illustrate the overall performance of participantsÕ laboratories on determining contaminants in food and herbal medicine. The three programs were organized in the period 2005–08 and the number of participants enrolled in the programs was 38–55 from a total of 35 economies within the Asia Pacific region and other regions (Table 2). 3.1. Heavy metals in herb (APLAC T043) 3.1.1. Overview of the program. The program was the first APLAC program for a herbal matrix that required the determination of cadmium and lead in Herba Demodii Styracifolii, which is a popular Chinese medicine with indications of promoting diuresis and relieving stranguria. It is useful to treat infectious hepatitis, cholelithiasis and cholecystitis with dampness-heat syndrome. The program aimed at addressing the contamination problems of heavy metals in herbal plants [17,18] and provided a means to evaluate the relevant testing capabilities of participants.

Table 2. Geographic distribution of registered participants for APLAC T043, T058 and T059 Country

T043

T058

T059

Argentina Australia Austria Belgium Brunei Canada Chile China Cyprus Czech Republic Ecuador Estonia Germany Hong Kong Indonesia Ireland Israel Italy Japan Korea Latvia Malaysia New Zealand Norway Philippines Poland Romania Russia Singapore Switzerland Taiwan Thailand Trinidad & Tobago United States of America Vietnam Total number Number of returned results Return rate (%)

2 — — 1 — — 5 4 — 1 — 1 — 7 4 1 — 2 — — — — 3 — 1 — — — 1 — 1 2 1 — 1 38 38 100

— 4 2 2 1 4 2 4 1 — — — 2 5 3 — — — 2 — 1 — 1 1 — 2 2 — 3 3 2 — — 1 — 48 45 94

2 4 — — — 2 4 4 — — 2 — — 6 4 — 2 — 4 3 — 1 3 — 5 — — 4 1 — 4 — — — — 55 50 91

Several batches, each containing about 10 kg of dried Herba Demodii Styracifolii samples (authenticated by a Chinese medicine expert) that had been confirmed to contain trace quantities of incurred cadmium (0.1–0.5 mg/kg) and lead (1–5 mg/kg) were purchased from a herbal shop. Samples were immersed in distilled water overnight to remove dirt and foreign matter. Pre-rinsed samples were air-dried in a cleanroom (Class 1000), then subjected to freeze drying, grinding, sieving (through 100-lm sieve) and mixing. Aliquots of about 25 g of the homogenized fine powder were packed into pre-cleaned and nitrogen-flush plastic bottles. More than 100 bottles of test samples were prepared and stored at room temperature before distribution. Three test portions (1 g) were drawn from each of the 10 randomly selected bottles in the homogeneity test. The 30 sub-samples were analyzed for cadmium and lead contents in a http://www.elsevier.com/locate/trac

565

Trends

Trends in Analytical Chemistry, Vol. 29, No. 6, 2010

randomized order using an accredited inductively coupled plasma mass spectrometry (ICP-MS) method. Homogeneity was determined from sampling variance (S2sv ) and analytical variance (S2av ) using a one-way Table 3. Results of F-test for the homogeneity test of cadmium and lead in APLAC T043 Element

Mean (mg/kg)

Cd Pb

0.296 1.41

S2sa

S2av

F

Fcritical

0.000023 0.001437

0.000017 0.002077

1.319 0.092

2.393 2.393

Fcritical at n = 30 and at the confidence interval of 95%.

ANOVA and an F-test. As shown in Table 3, homogeneity of test material was confirmed, as the F values obtained for cadmium and lead were smaller than those of the critical F-test at the confidence interval of 95%. Stability of analytes in the samples was monitored by duplicate analysis of random samples over a 10-month period. Absolute percent deviations of mean values were <3% of those of the mean value in the homogeneity test, indicating that matrix cadmium and lead were stable over the study period. Consensus means from participantsÕ data using robust test and r estimated from Horwitz function [19] were used to calculate participantsÕ z-scores [Equation (2)]:

Table 4. ParticipantsÕ methodologies and reported mean values for cadmium and lead in Herba Demodii Styracifolii Lab. code

Methodology Instrumentation

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38

ICP-OES ICP-MS ICP-MS ICP-MS ICP-MS ICP-MS & AAS ICP-MS AAS NAA ICP-OES ICP-MS ICP-OES AAS AAS AAS AAS ICP-MS ICP-MS ICP-MS AAS AAS ICP-MS AAS AAS AAS AAS AAS AAS AAS AAS ICP-MS ICP-OES ICP-MS AAS ICP-MS AAS ICP-MS ICP-MS

Digestion method Microwave Microwave Microwave Microwave Microwave Microwave Microwave — Electrical plate Pressure digester Microwave Microwave Dry ashing Muffle furnace Microwave Muffle furnace Microwave Water bath Hot block Wet ashing Dry ashing Microwave Microwave Microwave Wet ashing High-pressure Graphite block Microwave Microwave Muffle furnace Microwave Microwave Microwave Microwave Microwave Microwave High-pressure asher Microwave

Mean concentration (mg/kg) Digestion medium HNO3, H2O2 HNO3, H2O2 HNO3, HCl HNO3 HNO3 HNO3 HNO3 — H2SO4, H2O2 HNO3 HNO3, H2O2 HNO3, H2O2, HF NA HNO3, H2SO4 HNO3, H2O2 HNO3 HNO3 HNO3 HNO3, HCl HNO3, HClO4 — HNO3 HNO3, H2O2 HNO3 HNO3 HNO3 HNO3, HCl HNO3 HNO3 — HNO3, H2O2 HNO3, H2O2 HNO3, H2O2 HNO3 HNO3 HNO3, H2O2 HNO3 HNO3, H2O2 n Robust mean RSD (%)

Cadmium 0.235 0.247 0.301 <0.400 0.241 0.287 0.274 0.284 0.310 0.261 0.251 0.192 0.297 0.159 0.326 0.158 0.259 0.280 0.312 — 0.371 0.186 0.334 0.253 0.294 0.342 0.267 0.23 0.273 0.224 0.231 0.211 0.332 0.294 — 0.323 0.274 0.274 35 0.270 18.2

Lead 1.28 1.39 1.46 1.47 1.553 1.41 1.46 — — 1.43 1.415 — 3.358 1.507 1.380 1.504 1.32 1.40 1.34 5.74 1.145 1.123 1.42 1.436 1.37 1.50 1.62 1.5 1.88 0.953 0.946 1.253 1.457 1.436 1.35 1.58 1.54 1.249 35 1.43 9.5

—, No data or information provided; AAS, Atomic absorption spectrometry; ICP-MS, Inductively coupled plasma mass spectrometry; ICP-OES, Inductively coupled plasma optical emission spectrometry; NAA, Neutron activation analysis.

566

http://www.elsevier.com/locate/trac

Trends in Analytical Chemistry, Vol. 29, No. 6, 2010

Horwitz function ¼ 0:02C0:8495

Trends

ð2Þ

where C was the mean mass fraction of analytes obtained from homogeneity test. 3.1.2. Discussion on participantsÕ performance. All participating laboratories submitted results for cadmium or lead (Table 4); only two did not report for cadmium and three did not report for lead. Atomic absorption spectrometry (AAS) and ICP-MS were used most by participants, followed by inductively coupled plasma optical emission spectrometry (ICP-OES). Apart from these common methods for trace-element analysis by field laboratories, one laboratory employed a nondestructive neutron-activation analysis (NAA) method. Pre-treatment of samples is normally not required in NAA, so it is referred as an accurate primary method for inorganic analysis by national metrology institutions [20]. More than 60% participants used microwave-assisted digestion due to convenient operation, rapid and complete digestion for food and plant matrices [21–23]. Other common digestion methods (e.g., heating and wet ashing) were also employed. Nitric acid or a mixture of nitric acid and hydrogen peroxide were known to be effective in digesting food samples and hence used by the majority of participants. Although these digestion media could not completely decompose siliceous materials in plant tissue, a recovery of 90–115% for cadmium and lead was reported [24] when applied to plant-reference materials, including pine needles, tomato leaves, apple leaves and peach leaves. A recent study showed that siliceous materials in plant was readily dissolved at high temperature using a combination of nitric acid and hydrofluoric acid [25], but only one participant used it in this program. Further, the use of nitric acid and perchloric acid was not recommended (the latter is potentially hazardous during digestion and has relatively low recovery for heavy metals [26]); one participant who used it gave the highest result for lead. Consensus mean values of cadmium and lead were, respectively, 0.270 mg/kg (0.158–0.371 mg/kg) and 1.43 mg/kg (0.946–0.371 mg/kg). Between-laboratory variations for cadmium (18.2%) and lead (9.5%) were comparable to the r estimated from the Horwitz function. Three participants gave questionable z-score results for cadmium, whereas it had one questionable and five unsatisfactory z-score results for lead (Fig. 1). Analysis of the same herbal material was later carried out using an isotope-dilution ICP-MS (ID-ICP-MS) technique that had been verified to have high accuracy and precision through a Comite´ Consultatif pour la Quantite´ de Matie`re pilot-study program (CCQM-P97) [27]. Reference values (0.2797 mg/kg for cadmium and 1.5065 mg/kg for lead) obtained by ID-ICP-MS were found to agree (3–5% bias) with the consensus mean and satis-

factory performance of participants was evidenced from the good correlation results. Measurement uncertainty (MU) is one of the crucial requirements of ISO/IEC 17025 [3] and it provides a means for accreditation bodies to assess the competence of the respective analysis. In this program, participants were requested to submit MU data to the organizer, though it was not taken into account in the z-score assessment. Four participants did not report MU and the others gave a significant variation of relative expanded uncertainty (1.3–51% for cadmium; 0.33–43% for lead) at the coverage factor of 2. Although international guidelines are already well established (e.g., Eurachem/ CITAC guide [28]), a number of participating laboratories still have difficulties in estimating MU. Bearing in mind that the absence of reliable associated MU for any analytical measurement would introduce the risk of results being incorrectly interpreted [29], accreditation bodies need to provide training and practical guidance to those laboratories. 3.2. Organochlorine pesticides (OCPs) in ginseng root (APLAC T059) 3.2.1. Overview of the program. The program was a continuation of the first PT (APLAC T049) on five OCPs in ginseng root (Table 1). APLAC T059 targeted the measurement of seven OCPs – hexachlorobenzene (HCB), a-, b-, c- and d-hexachlorocyclohexane (a-, b-, cand d-BHC), pentachloronitrobenzene (PNCB) and o,p 0 -dichlorodiphenyldichloroethylene (o,p 0 -DDE) – in ginseng root. 20 kg of Panax ginseng samples containing detectable quantities of incurred OCPs were purchased from a local market. Various batches of the root samples were rinsed with distilled water to remove dirt and other foreign matter, then freeze dried and ground. The sample was sieved through 100-lm sieves and the fine powder collected was transferred to a container for mixing. Aliquots of 30 g were finally dispensed into nitrogen-purged amber-glass bottles, capped and disinfected with 137Cs at a dose of about 1 kGy. The 150 bottles prepared were vacuum sealed inside polypropylene bags and stored in electronic desiccators at 25C prior to dispatch. The homogeneity test involved duplicate analysis of 12 random samples using an accredited test method of Soxhlet extraction (2 x 4 h) and gas chromatography with mass spectrometric detection (GC-MSD) determination. Relative standard deviations (RSDs) of withinbottles and between-bottles for a-BHC, b-BHC, c-BHC, dBHC, HCB and PNCB were 0.85–3.5% and 3.1–6.1%, respectively. Samples were found containing no o,p 0 -DDE (below the limit of quantification of 20 lg/kg) and this parameter was purposely designed as a qualitative test in this program. Referring to ISO13528 [16], sample homogeneity was accepted if the between-sample variation (Ss) was 6 0.3r. Ss was calculated from the between-sample standard deviation (Sx) and the absolute http://www.elsevier.com/locate/trac

567

Trends

Trends in Analytical Chemistry, Vol. 29, No. 6, 2010

Figure 1. Distribution of participantsÕ z-scores for (a) cadmium and (b) lead in Herba Demodii Styracifolii.

difference of duplicate analysis (wt) in the following relationship [Equation (3)]: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi X w2t =nÞ ð3Þ Ss ¼ S2x  ð where n was the number of replicate results, and r was estimated from the Horwitz function [19]. A 5-month stability test for residual OCPs was conducted using two sets of randomly selected samples at room temperature of about 25C and at an elevated 568

http://www.elsevier.com/locate/trac

temperature of 37C. Concentrations of OCPs in each sample set were analyzed in triplicate on a monthly basis after dispatch. Samples were stable if the difference of the mean in homogeneity test and the mean in stability test was 60.3r. RSDs of the six OCPs during the five-month period were 0.45% to 4.0% and the mean values of triplicate analysis were <0.3r. Similar to the stability test in APLAC T049, results showed that those OCPs under study in ginseng root were stable at both temperatures.

Trends in Analytical Chemistry, Vol. 29, No. 6, 2010

Trends

Table 5. ParticipantsÕ analytical methods and reported mean values for the seven OCPs in ginseng root Lab. code

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50

Instrumentation

GC-ECD GC-MS GC-MS GC-ECD GC-MS GC-MS GC-ECD GC-MS2 GC-MS GC-ECD GC-MS GC-ECD GC-ECD GC-ECD GC-ECD GC-MS GC-ECD GC-ECD GC-MS2 GC-ECD GC-ECD GC-MSD GC-ECD, -MS GC-MS GC-ECD GC-ECD GC-ECD GC-ECD GC-ECD GC-ECD GC-ECD GC-ECD GC-MS GC-ECD GC-ECD, -MS GC-ECD, -MS GC-ECD GC-ECD GC-ECD GC-ECD GC-ECD GC-ECD GC-ECD GC-MS GC-MS GC-ECD GC-ECD GC-ECD GC-ECD, -MS GC-MS n Robust mean RSD (%)

Mean concentration (lg/kg) a-BHC

b-BHC

c-BHC

d-BHC

HCB

PNCB

— 86.8 67.3 106.80 99 90.0 101 287 88.6 83.5 239 202 166 80.0 187 200 39 193 211 95.808 187 202 49 217.498 9.480 52.3 143 139 49.9 64.5 189 90.5 237.5 71.3 250 242.47 198 126 189 152.722 115 135.507 72 231 91.3 79.8 — 188 210.950 211 48 141.8 53.6

149 63.4 46.6 60.16 109 90.5 62.8 99.1 49.0 42.6 100 102 55.7 61.0 91.3 123 38 103 725 — 96.1 216 30 345.686 21.253 26.5 213 86.1 19.6 44.3 95.0 48.6 107.4 64.3 99.4 93.16 101 125 98.7 1407.505 129 14.820 53 111 BDL 35.7 — 174 104.259 106 48 87.4 54.7

108 443.5 61.3 564.83 171 84.9 77.8 208 71.5 59.1 150 149 118 81.0 146 173 110 166 167 967.208 149 199 221 193.537 5.667 43.3 677 113 59.0 47.7 155 67.3 161.5 58.2 180 166.91 155 88 154 170.117 64 117.737 51 189 BDL 56.6 — — 165.714 160 48 133.1 51.0

— 83.4 203 348.87 82 — 365 1086 304 266 — 705 — 377 — — — — 725 — 693 3140 44 788.053 — 467 127 484 227 150 62.2 289.5 733.7 235 — 6723.35 738 186 754 946.139 349 7.014 291 587 BDL 279 — 171 658.304 836 38 442.9 76.3

87 55.7 47.7 68.38 — 42.4 65.3 188 59.5 54.7 127 162 133 65.0 197 230 26 173 154 — 161 163 32 277.889 7.980 39.8 112 147 6.74 — 131 59.6 171.3 69.7 — — 183 — 184 — 775 113.657 77 186 41.5 62.5 — 172 — 199 41 114.7 65.7

762 — 330 91.08 461 — 560 1811 587 448 1536 1290 — 377 232 247 300 1391 — — 1320 1485 421 1786.553 64.123 286 1160 868 1860 — 105 564.7 1240.2 519 — — 1300 — — — 652 1015.788 — 1329 BDL 497 — — 114.768 1530 36 807.2 76.7

o,p 0 -DDE 7 <0.5 BDL 0 — <5 — BDL — — BDL 0 — — 9.6 9 — BDL BDL — — 6.2 — — <7.000 <2.0 BDL BDL — — <10 — 0 — — — BDL — — — <15 4.587 — <10 BDL BDL 38.329 BDL BDL BDL 29 NA NA

—, No data or information provided; BDL, Below detection limit; GC-ECD, Gas chromatography with electron-capture detection; GC-MS, Gas chromatography mass spectrometry; GC-MS2, Gas chromatography tandem mass spectrometry; n, Number of numeric data; NA, Not applicable.

Consensus mean and r estimated from Horwitz function were used to calculate z-scores in the performance assessment.

3.2.2. Discussion on participantsÕ performance. Participants used GC with electron-capture detection (GC-ECD), GC-MSD or both as their analytical methods for http://www.elsevier.com/locate/trac

569

Trends

Trends in Analytical Chemistry, Vol. 29, No. 6, 2010

quantifying OCPs in ginseng root (Table 5). Despite liquid chromatography with tandem MS (LC-MS2) gaining popularity, in terms of selectivity and sensitivity, in pesticide analysis [30] in recent years, GC remains the preferred technique used by food laboratories worldwide. Chromatographic separations normally relied on 30–60-m DB-5 or DB-1701 capillary columns. Extraction by agitation and sonication was used by the majority because of the ease of operation and simple instrumentation. Only seven participants used Soxhlet, microwave-assisted and matrix solid-phase dispersion extraction. Acetonitrile, acetone alone or mixed with other common organic solvents (e.g., petroleum ether, dichloromethane, and hexane) were reported as the choices for extraction solvent. Removal of sample interference was conducted by solid-phase extraction (SPE) and/or gel permeation chromatography (GPC) (60%) or adding concentrated sulfuric acid (20%). About 20% of the participants did not use any clean-up treatment. While the OCP analytical methods used by participants were relatively standard and routine [31], the spread of data was found to be very wide (between-laboratory variation was in the range 51.0–76.7%). Consensus mean values (±SD) were 114.7 lg/kg (±75.4 lg/ kg) for HCB, 141.8 lg/kg (±76.0 lg/kg) for a-BHC, 87.4 lg/kg (±47.9 lg/kg) for b-BHC, 133.1 lg/kg (±67.8) for c-BHC, 442.9 (±338.0) lg/kg for d-BHC and 807.2 lg/kg (±618.8) for PNCB, respectively. Due to the substantial between-laboratory variation and deviation from those values obtained in the homogeneity studies for d-BHC and PNCB, the consensus mean might not correctly represent the actual incurred values. Hence, participants were informed that no performance assessment was provided for these two test parameters. A similarly large variability of OCP results was also reported in other PT programs (e.g., an RSD of 78% in mussel tissue [32]). The causes were said to be noncompliance of QA/QC requirements (e.g., loss during clean-up and preparation procedures, inadequate internal/external standards, and erroneous calibration, dilution or calculation). Furthermore, the effects of matrix interference from plant extracts on the response of pesticides might be even more variable than from extracts of animal origins [33]. *enyuva and Gilbert [34] also commented that performance of pesticide analysis in difficult matrices (e.g., tea and hops) was relatively poor. Ginseng root is considered to be a ‘‘difficult matrix’’ and the presence of high oil content and intrinsic interferences should require extensive clean up (e.g., SPE and GPC after extraction [35]). Another explanation for the variation in this program was that it was possibly due to the difficult extraction of the incurred OCPs. During the development of an accurate GC-ID-MS technique [36], it was found that extractable incurred OCPs in ginseng root using sonication were only about 50–60% of that with Soxhlet extraction, and the recovery also varied 570

http://www.elsevier.com/locate/trac

Table 6. z-score performance of participants for four of the incurred OCPs in ginseng root Performance achieved

|z| 6 2 2 < |z| < 3 |z| P 3 Total

Number of Participants (%) a-BHC

b-BHC

c-BHC

HCB

28 (58%) 9 (19%) 11 (23%) 48 (100%)

30 (64%) 6 (13%) 11 (23%) 47 (100%)

27 (57%) 12 (26%) 8 (17%) 47 (100%)

16 (39%) 15 (37%) 10 (24%) 41 (100%)

with the duration of sonication extraction. An independent comparative study on the effect of various extraction techniques on OCPs in animal feed [37] supported these findings. Of the 29 results for o,p 0 -DDE, 23 reported ‘‘0 lg/kg’’ or ‘‘below detection limit’’, which were considered satisfactory qualitative results. Six participants had falsepositive results and they later explained that it was caused by positive interferences, insufficient clean up or sample contamination. Out of 50 laboratories, 33 claimed their test methods were accredited, but they did not demonstrate performance superior to that of their non-accredited peers. In fact, some studies showed that participantsÕ results depended on the performance of the analyst, instead of sophisticated instrumental methods or accreditation status [38]. Results of z-score for HCB, a-BHC, b-BHC and c-BHC are summarized in Table 6. With the use of generous target r (estimated from the Horwitz function) in the range 25–30%, the overall performance was not very good. Only 39–64% of the participants achieved satisfactory z-score results. In brief, 27 of the 50 participants (54%) were identified as having reported one or more unsatisfactory results; and 40 of 183 submitted data sets (21.8%) were identified as unsatisfactory. Compared with the performance of the first OCP program [39], there was no significant improvement made in the analysis of incurred OCPs in ginseng root. Participants in this program were also found to repeat the same problems in MU estimation as in APLAC T043. Apart from a number of overestimated or underestimated expanded uncertainties, about 10% of the participants showed hesitation in presenting MU data. 3.3. Malachite green and leucomalachite green in swamp eel (APLAC T058) 3.3.1. Overview of the program. MG had been used as a veterinary drug to treat various fish diseases. MG and its metabolite, leucomalachite green (LMG), exhibit carcinogenic properties, and both chemicals are now banned in many countries. However, illegal use of MG in fish

Trends in Analytical Chemistry, Vol. 29, No. 6, 2010

farming in some countries led to a serious food incident in fish and fishery products in 2005. At that time, only a very limited number of PT programs were available to assess the testing capability of MG and LMG, so a program was organized upon requests from local and overseas laboratories. About 5 kg of live swamp eels (Monopterus albus) were purchased from a local fish market. Eels were divided into two groups:  one group (fed group) comprised 1 kg of eels and was placed in a 40-L fish tank containing 0.5 mg/kg of MG; and,  the second group (blank group) of 4 kg of eels was placed in a separate tank containing water only. After 24 hours, eels in both groups were taken out, washed with water, then slaughtered, boned, freezedried, ground to powder and sieved through 250-lm sieves. Concentrations of MG and LMG in the dried powder of the fed group were initially determined using a validated LC-MS2 method. Appropriate quantities of the powder from both groups were thoroughly mixed in order to produce a desirable concentration range (lg/kg level) of MG and LMG in the final product. An aliquot of 5 g of the mixed powder was independently dispensed into a clean nitrogen-flushed amber bottle. The bottle was capped, disinfected by c-irradiation at a dose of about 10 kGy and vacuum sealed inside a polyethylene bag. More than 150 bottled samples were eventually prepared. The homogeneity and the stability of samples were tested, as described in APLAC T059 (para 3.2.1.). Duplicate portions of 0.5 g from the 10 random bottles were analyzed using a validated LC-MS2 method in the homogeneity test. Mean concentrations and 0.3r of MG and LMG were 30.75 lg/kg and 2.47 lg/kg, and

Trends

2.05 lg/kg and 0.25 lg/kg, respectively. The corresponding Ss values were smaller than those of 0.3r, indicating that the homogeneity of the prepared samples was satisfactory. A 9-month stability study for MG and LMG at 25C and 37C was monitored and it covered the entire period of the program. One sample from each temperature set was analyzed in duplicate or triplicate under the same sample-treatment and operational conditions as in the homogeneity test. Results in the stability test were in the range 27.08–31.70 lg/kg for MG (RSD = 1.4% at 25C and 6.1% at 37C) and 1.86– 2.11 lg/kg for LMG (RSD = 3.6% at 25C and 4.5% at 37C), respectively, indicating that both analytes were stable throughout the program. Assigned reference values were determined by an LCID-MS method. The technique was also used to determine MG and LMG in salmon and to compare results with other national metrology institutes in a CCQM intercomparison program (CCQM-P88). Results of the program (Fig. 2) verified that the LC-ID-MS (Participant B) method offered good accuracy and precision. Reference values from the LC-ID-MS used for assessment were 28.2 ± 0.25 lg/kg for MG and 2.21 ± 0.15 lg/kg for LMG. r for z-score assessment was estimated from the Horwitz function. 3.3.2. ParticipantsÕ performance. ParticipantsÕ results and analytical methodologies (LC-MS2 and LC-UV) have been reported elsewhere [40]. In brief, most participants used acidic buffer solution [e.g., McIlvaine buffer (citric acid and sodium hydrogen phosphate) or ammonium acetate] as scavengers to reduce the conversion of MG to LMG during extraction with acetonitrile or a mixture of acetonitrile with dichloromethane, then followed with

Figure 2. Comparison of the LC-ID-MS method (Participant B) with other LC-ID-MS methods for the determination of total MG and LMG in salmon in the CCQM-P88 pilot study (Adapted with permission from the CCQM-P88 Draft A Report).

http://www.elsevier.com/locate/trac

571

Trends

Trends in Analytical Chemistry, Vol. 29, No. 6, 2010

Table 7. ParticipantsÕ methodologies and reported mean values for MG and LMG in eel muscle Lab. code

Methodology Instrumentation

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45

2

LC-MS LC-UV LC-MS2 LC-VIS LC-MS2 LC-UV LC-MS2 LC-MS2 LC-MS2 LC-MS2 LC-UV LC-DAD LC-MS2 LC-VIS LC-MS2 LC-MS2 LC-MS2 LC-MS2 LC-UV LC-MS2 LC-MS2 LC-MS2 LC-MS2 LC-MS2 LC-MS2 LC-MS2 LC-MS2 LC-UV LC-UV LC-MS2 LC-MS2 LC-VIS LC-MS2 LC-UV LC-MS2 LC-UV LC-MS2 LC-UV LC-MS2 LC-MS2 LC-MS2 LC-UV LC-MS2 LC-MS2 LC-MS2

Internal standard d5-MG, d6-LMG No No No d5-MG, d5-LMG Brilliant green No Brilliant green No d5-LMG No No d5-MG, d6-LMG No No d6-LMG d5-MG, d6-LMG d5-MG No d5-MG, d6-LMG d5-MG, d6-LMG d5-MG, d5-LMG d5-MG, d6-LMG d6-LMG No Brilliant green d5-MG, d5-LMG No No Crystal violet d5-MG, d5-LMG No d5-MG, d6-LMG No d5-MG, d6-LMG No d5-MG, d6-LMG No No No d5-MG, d6-LMG No No d5-MG, d6-LMG d5-MG, d5-LMG

Mean concentration (lg/kg) Clean up SCX SPE Hexane wash C18 SPE No SCX SPE SCX SPE MCX SPE SPE Hexane wash MCX SPE Hexane wash SCX SPE No C18 SPE SCX SPE SCX SPE — SCX SPE Filtration Alumina SPE SCX SPE No SCX SPE PRS SPE Alumina SPE No SCX SPE Sep-Pak C2t NO Hexane wash SCX SPE Alumina SPE MCX SPE SCX SPE MCX SPE Alumina, PRS SPE DCM wash SCX SPE — C18 SPE SCX SPE Alumina, PRS SPE C18 SPE SCX SPE SPE n Median RSD (%)

MG

LMG

16 4.3 12.6 <15 20.250 11.4 11.51 0.88 24.6 3.06 — 1.97 4.50 3.2324 4.01 25.5 26.5 33.0 ND 6.20 9.61 22.89 11 45.4 32.6 20.32 35 21.43 96 9.63 32 0.3716 27.0 39.2 22 712 66.8 10.7 3.15 7.0 25.0 16.2 78.3 20 18.07 41 19.03 78.3

1.9 — 0.58 59 ND <10 3.89 6 0.5 3.1 1.42 <5 <1.0 ND — 0.293 2.56 11.2 4.1 76 1.34 1.82 1.90 <5 2.4 92.1 ND 2.1 41.94 20 <2 1.2 8.545 2.1 — 1.9 ND 2.0 73.5 0.23 ND 1.63 — 2.11 <10 3.65 29 2.11 109

—, No data or information provided; ND, Not determined; DCM, Dichloromethane; LC-MS2, Liquid chromatography with tandem mass spectrometry; LC-UV, Liquid chromatography with ultraviolet detector; MCX, Mixed-mode cationic exchange; PRS, Propylsulfonic acid; SCX, Strong cationic exchange; SPE, Solid-phase extraction (Adapted with permission from [40]).

SPE (either cationic exchange or C18 columns) for removal of matrix interferences. Dispersions of participantsÕ data for both analytes were extremely large. Mean concentrations of MG and LMG were 0.4–712 lg/kg and 572

http://www.elsevier.com/locate/trac

0.23–92.1 lg/kg, respectively. The variability could be associated with lengthy experimental procedures, tedious sample preparations, instability of MG, matrix interference and trace levels in fish tissues [41]. Median

Trends in Analytical Chemistry, Vol. 29, No. 6, 2010

Trends

Figure 3. ParticipantsÕ results reported for (a) MG and (b) LMG in eel muscle. Dotted lines represent the reference values and error bars represent measurement uncertainty. The inset shows the lower reported concentration data for LMG.

values deviated from the reference assigned values by 32.7% (MG) and 4.5% (LMG) (Table 7). Fig. 3 illustrates the distribution of participantsÕ data. With the 42 MG and 29 LMG numeric data sets used for z-score

assessment, 20 (48%) and 18 (62%) participants achieved satisfactory z-scores; 9 (21%) and 3 (10%) achieved questionable z-scores; and, 13 (31%) and 8 (28%) achieved unsatisfactory z-scores for MG and LMG, http://www.elsevier.com/locate/trac

573

Trends

Trends in Analytical Chemistry, Vol. 29, No. 6, 2010

respectively. The overall performance was not promising as 21 (30%) of 71 data sets submitted were identified as unsatisfactory. Contrary to a study [42] that concluded that determination of MG and LMG in fish and shrimp was consistent between LC-MS and LC-UV methods, the results in this program showed different experimental findings. Analysis of participantsÕ data indicated precision and accuracy depended upon the choice of analytical method used. The RSD of MG and the accuracy of LMG (relative to the assigned value) were found to be better when using LC-MS than the LC-UV/Vis technique. For those who employed isotope internal standards (e.g., d5-MG, d6-LMG or 13C6-LMG) in LC-MS2 analysis, the precision for both analyses was further improved. Reliability of the technique employing isotope standards for MG and LMG in fish has been demonstrated by other research teams [43,44].

4. Performance-assessment trends in proficiency tests PT or interlaboratory comparison programs offer unique external QA/QC for qualitative and quantitative analysis [45,46] as well as a means of evaluation in assessing the competence of participating laboratories [47]. International harmonization of practices in PT could improve the comparability of analytical measurements [48], enhance universal acceptance of analytical data and is always beneficial to accreditation bodies, participating laboratories and their customers. One of the essential achievements in harmonization is the use of adequate statistical tools, as it establishes a scientific foundation for accurate, fair performance assessment. ISO 13528 [16] has outlined common statistical methods for homogeneity testing, stability testing and data treatment in PT programs. Five assessment approaches (z-score, En-score, zeta-score, z 0 -score and Ez score) or scoring systems on the basis of different mathematical models are listed for scheme providers to evaluate data. The advantages and the limitations of individual scoring systems on performance assessment are thoroughly discussed in two technical papers [49,50]. Unlike z-score, which utilizes target r, the other four include expanded uncertainty (U) or standard uncertainties (u) from both participants and the assigned values. The 1993 International Harmonization Protocol for PT [51] recommended z-score and it has been used frequently in many programs since then. It is simple and has demonstrated excellent applicability and acceptance by scheme providers and participating laboratories worldwide [52]. En-score, a complementary approach to z-score, is also commonly used in current practice by scheme providers [53]. 574

http://www.elsevier.com/locate/trac

As stated in Equation (1), z-score is derived from assigned value and r. Assigned value acts as the reference point for an assessment and r estimates the allowable deviation range under specified circumstances from the reference point. These two parameters could be varied from one program to another depending upon providersÕ preference as well as the suitability of particular measurements. While fitness for purpose of a scheme is the predominant drive for the selection, assigned value and r must be well defined by the providers and their use should be transparent to participants prior to commencement of the programs. 4.1. Assigned values The latest ISO/IEC 17043 [15] defines assigned value as the value attributed to a particular property of a PT item. It should be the best practicable estimate of the ‘‘true’’ value of analytes in a test material, and therefore any inappropriate estimate of the assigned value will undermine the validity of performance assessment [54]. Five different types of assigned value, with descending order of confidence interval in measurement uncertainty, are described in ISO 13528 [16], namely formulation, certified reference value, derived reference value, and consensus values from expert laboratories and from participants. Consensus mean values from participants are often used by providers [55] because these values are easily determined with no additional experimental work, so they do not add to the cost of a program. The main problem is that performance of participants will not be properly interpreted if the consensus values deviate considerably from the true values. A simulation study using the Monte Carlo model predicted that the deviation, mainly contributed by the number of participating laboratories and the distribution of laboratory bias, could be as large as 40% [56]. In fact, discrepancy has often been found in a number of PTs and a study showed that the consensus values of four OCPs in a PT were underestimated by 2.7–14.1% compared with the reference values [57]. De Bie`vre [58] commented that a PT should offer a traceable quantity instead of using the participantsÕ data to calculate consensus value. Further, Kuselman et al. [59] strongly proposed the use of reference values with high metrological property (e.g., reference materials) for PTs. It was reported in the USA that a number of certified reference materials in food have been used successfully in the qualifying process for analytical contracts, and in the routine sample stream as part of the national program [60]. However, the cost of the program increases with the number of participants, so the usefulness of this approach is confined to small-scale PTs only. Scheme providers for food testing in Europe [61] decided to determine the reference values using accurate measurement methods (e.g., ID-MS) in an attempt to reduce possible bias of the assigned values. Similarly, a

Trends in Analytical Chemistry, Vol. 29, No. 6, 2010

French team relied upon technical assistance from their national metrology institute in order to produce more reliable reference values [62]. Although the concept of implementing accurately referenced assigned values is ideal for PTs, the costs of skilled manpower, sophisticated instrumentation and isotope standards might be major considerations, particularly for commercial scheme providers. Alternatively, gravimetric preparation values were thought to be excellent estimates of true values and have been reported as giving reliable assessments in an interlaboratory comparison for PAHs and pesticides in organic solutions [63]. Most recently, a PT on melamine in milk [64] using gravimetric preparation values as the reference values showed that consensus values at medium–high concentration (1–4.5 mg/kg) agreed well with reference values, but there was an obvious discrepancy at low concentration (0.05 mg/kg) due to the variability of participantsÕ capabilities at trace-level determination. Knowing that the quality of performance assessment improves with the use of a reference value in most circumstances, scheme providers should cautiously plan whether assigned values are the preferred choice on the basis of the nature of the program as well as acceptance by participants and their accreditation bodies. 4.2. Target-performance deviation The criteria for setting r should be considered on the basis that it is realistically achievable or provides sound performance assessment. There are several ways to determine r, as thoroughly described elsewhere [16,49]. For food testing, r derived from the Horwitz function [19] is commonly used by PT-scheme providers (e.g., FAPAS PT programs on vitamins in liquid supplement and breakfast cereal [65]). The Horwitz function generalizes the reproducibility of large PT data sets to a mathematical expression [Equation (2)]. In general, the value of r increases with decreasing analyte concentration. However, the actual standard deviation of food analytes in a PT is usually lower than that of the Horwitz function at low concentration but higher at high concentration, so modifications of the expression at <120 ppb and >13.8 ppm were suggested by Thompson [66]. While some workers have reservations about the application of the Horwitz function [67], the function is nonetheless useful in prescribing data uncertainty and defining in advance an appropriate standard uncertainty for participants in PT [68]. Other scheme providers tend to collate the data generated from their previous PTs and to establish r from adequate statistical analysis (e.g., the Confino model by QUASIMEME [69] for environmental samples and a fitness-for-purpose RSD of 25% for European Commission PTs [70] on pesticides in fruits and vegetables). Furthermore, a recent simple function comprising a two-

Trends

parameter model with concentration of the analyte as a single predictor variable has also been reported to work well for a PT in food analysis [71]. 4.3. Future development for the APLAC PT program The APLAC PT facilitates mutual confidence in the technical competence of members and their accredited laboratories. The very limited number of scheme providers in the PT network in the Asia Pacific region emphasizes the importance of APLAC PT programs to participating laboratories. Since many of the economies in the region are developing nations, continuity in organization of PT programs in food and herbs can assist the development of technical skills, new analytical techniques and method validation. After serving the communities for more than 15 years, the programs offered were reviewed and were found to have high quality and to suit the needs of participating laboratories. In order to ensure that the PT program operates at the highest standard, APLAC recognizes that there is still scope for improvement. Discussions are proceeding on how to enhance the capability of accreditation bodies in organizing PT programs for laboratories in the developing economies. Other issues are also being explored (e.g., the feasibility of programs for inspection bodies, diversity of analytes and matrices, use of more reliable assigned values, and collaboration with national metrology institutes and other scheme providers). Finally, the ultimate goal of ‘‘once tested and accepted everywhere’’ requires an integrated measurement system combining the activities of metrology, standards and accreditation. We are confident that APLAC PT programs will have a unique role to play in progress to meet that goal.

Acknowledgements The authors are indebted to T.L. Ting, Government Chemist, and C.M. Lau, Assistant Government Chemist, Government Laboratory of Hong Kong, for their support for the above PT programs and this manuscript.

References [1] Asia Pacific Laboratory Accreditation Cooperation (http://www. aplac.org). [2] The International Laboratory Accreditation Cooperation (http:// www.ilac.org). [3] International Standards Organization, ISO/IEC 17025, Gerneral requirements for the competence of testing and calibration laboratories, ISO, Geneva, Switzerland, 2005. [4] D. Patton, 11 November 2006 (http://www.ap-foodtechnology.com/Processing/Sudan-Red-found-in-new-regions-of-China). [5] Wikipedia, Malachite green (http://en.wikipedia.org/wiki/Malachite_green). [6] J.R. Ingelfinger, New Engl. J. Med. 359 (2008) 2745.

http://www.elsevier.com/locate/trac

575

Trends [7] [8] [9] [10] [11] [12]

[13]

[14]

[15]

[16]

[17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27]

[28]

[29] [30] [31] [32] [33] [34] [35] [36] [37] [38]

576

Trends in Analytical Chemistry, Vol. 29, No. 6, 2010

Z.-G. Wang, J. Ren, Trends Pharmacol. Sci. 23 (2002) 347. K. Chan, Chemosphere 52 (2003) 1361. J. Xue, L. Hao, F. Peng, Chemosphere 71 (2008) 1051. E. Ernst, J.T. Coon, Clin. Pharmacol. Ther. 70 (2001) 497. R.B. Saper, S.N. Kales, J. Paquin, M.J. Burns, D.M. Eisenberg, R.B. Davis, R.S. Phillips, J. Am. Med. Assoc. 292 (2004) 2868. Asia Pacific Laboratory Accreditation Cooperation, APLAC PT002: Testing Interlaboratory Comparisons, APLAC, 2008 (http://www.aplac.org). International Standards Organization, ISO Guide 43, Part I: Selection and Use of Proficiency Testing Scheme by Laboratory Accreditation Bodies, ISO, Geneva, Switzerland, 1997. International Laboratory Accreditation Cooperation, ILAC Guide 13: Guidelines for the Requirements for the Competence of Providers of Proficiency Testing Schemes, ILAC, 2007 (http:// www.ilac.org). International Standards Organization, ISO/IEC 17043, Conformity assessment - General requirements for proficiency testing, ISO, Geneva, Switzerland, 2010. International Standards Organization, ISO 13528: Statistical methods for use in proficiency testing by interlaboratory comparisons, ISO, Geneva, Switzerland, 2005. E.D. Caldas, L.L. Machado, Food Chem. Toxicol. 42 (2004) 59. U. Divrikli, N. Horzum, M. Soylak, L. Elci, Int. J. Food Sci. Technol. 14 (2006) 712. W. Horwitz, Anal. Chem. 54 (1982) 67A. Y.C. Yip, J.C.W. Lam, W.F. Tong, Trends Anal. Chem. 28 (2009) 214. S. Walas, M. Wo´jtowicz, H. Mrowiec, W. Zegar, Int. J. Environ. Anal. Chem. 84 (2004) 1023. S.P. Dolan, S.G. Capar, J. Food Comp. Anal. 15 (2002) 593. L. Huang, R.W. Bell, B. Dell, J. Woodward, Comm. Soil Sci. Plant Anal. 35 (2005) 427. S. Wu, X. Feng, A. Wittmeier, J. Anal. At. Spectrom. 12 (1997) 797. J.C.W. Lam, Y.C. Yip, W.F. Tong, Accred. Qual. Assur. 13 (2008) 311. Z.-Y. Hseu, Bioresource Technol. 95 (2004) 53. Y.C. Wong, D.W.M. Sin, Y.C. Yip, L. Valiente, A. Toervenyi, J. Wang, G. Labarraque, P. Gupta, D. Soni, Sumardi, E. Hwang, C. Yafa, O. Cankur, E. Uysalg, G. Turk, R. Huertas, Accred. Qual. Assur. 14 (2009) 151. Quantifying Uncertainty in Analytical Measurement, Eurachem/ CITAC Guide, 2nd Ed. 2000 (http://www.measurementuncertainty.org/mu/guide/index.html). A. Williams, Accred. Qual. Assur. 13 (2008) 633. O. Nu´n˜ez, E. Moyano, M.T. Galceran, Trends Anal. Chem. 24 (2005) 683. D. Derek, E. Sverko, Anal. Bioanal. Chem. 386 (2006) 769. J.-P. Villeneuve, F.P. Carvalho, M. Horvat, C. Cattini, Int. J. Environ. Studies 6 (2004) 437. A. Ambrus, Accred. Qual. Assur. 9 (2004) 288. H.Z. *enyuva, J. Gilbert, Trends Anal. Chem. 25 (2006) 554. F.E. Ahmed, Trends Anal. Chem. 20 (2001) 649. S. Chan, M.F. Kong, Y.C. Wong, S.K. Wong, D.W.M. Sin, J. Agric. Food Chem. 55 (2007) 3339. M. Gfrerer, S. Shuo, E. Lankmayr, X. Quan, F. Yang, Anal. Bioanal. Chem. 378 (2004) 1861. L. Van Nevel, P.D.P. Taylor, U. O¨rnemark, J.R. Moody, K.G. Heumann, P. De Bie`vre, Accred. Qual. Assur. 3 (1998) 56.

http://www.elsevier.com/locate/trac

[39] M.F. Kong, S. Chan, Y.C. Wong, S.K. Wong, D.W.M. Sin, J. AOAC Int. 90 (2007) 1133. [40] Y.C. Wong, T.C. Cheung, Food Addit. Contam. 26 (2009) 1472. [41] D.R. Doerge, M.I. Churchwell, T.A. Gehring, Y.M. Pu, S.M. Plakas, Rapid Commun. Mass Spectrom. 21 (1998) 1625. [42] W.C. Andersen, S.B. Turnipseed, J.E. Roybal, J. Agric. Food Chem. 54 (2006) 4517. [43] K. Halme, E. Lindfors, K. Peltonen, J. Chromatogr., B 845 (2007) 74. [44] Z. Hall, C. Hopley, G. OÕConnor, J. Chromatogr., B 874 (2008) 95. [45] I. Taverniers, M. De Loose, E. Van Bockstaele, Trends Anal. Chem. 23 (2004) 535. [46] B.M. Simonet, Trends Anal. Chem. 24 (2005) 525. [47] H. Emons, T.P.J. Linsinger, B.M. Gawlik, Trends Anal. Chem. 23 (2004) 442. [48] A.M.H. van der Veen, T.L. Hafkenscheid, Accred. Qual. Assur. 9 (2004) 657. [49] M. Thompson, S.L. Ellison, R. Wood, Pure Appl. Chem. 78 (2006) 146. [50] Y.C. Yip, W.F. Tong, Trends Anal. Chem. 28 (2009) 1276. [51] M. Thompson, R. Wood, Pure Appl. Chem. 65 (1993) 2123. [52] P. Armishaw, B. King, R.G. Millar, Accred. Qual. Assur. 8 (2003) 184. [53] N. Boley, A.M.H. van der Veen, P. Robouch, M. Golze, J. Van de Kreeke, U. O¨rnemark, B. Tylee, Accred. Qual. Assur. 11 (2006) 373. [54] R.E. Lawn, M. Thompson, R.F. Walker (Editors), Proficiency Testing in Analytical Chemistry, Royal Society of Chemistry, Cambridge, UK, 1997. [55] M. Thompson, Accred. Qual. Assur. 10 (2006) 574. [56] S.K. Wong, Accred. Qual. Assur. 10 (2005) 409. [57] Y.C. Wong, S.K. Wong, T.T. Kam, Anal. Bioanal. Chem. 392 (2008) 1507. [58] P. De Bie`vre, Accred. Qual. Assur. 11 (2006) 105. [59] I. Kuselman, M. Belli, S.L.R. Ellison, A. Fajgelj, U. Sansone, W. Wegscheider, Accred. Qual. Assur. 12 (2007) 563. [60] M.B. de la Calle Guntin˜as, I. Wysocka, C. Que´tel, E. Vassileva, P. Robouch, H. Emteborg, P. Taylor, Trends Anal. Chem. 28 (2009) 454. [61] K.M. Phillips, W.R. Wolf, K.Y. Patterson, K.E. Sharpless, J.M. Holden, Anal. Bioanal. Chem. 389 (2007) 219. [62] L. Noe¨l, C. Vastel, R. Chekri, C. Chafey, C. Testu, T. Gue´rin, Microchem. J. 92 (2009) 73. [63] M. Ricci, O. Bercaru, R. Morabito, C. Brunori, I. Ipolyi, C. Pellegrino, A. Sahuquillo, F. Ulberth, Trends Anal. Chem. 26 (2007) 818. [64] M. Chan, C.K. Lo, L.S. Cheng, T.C. Cheung, Y.C. Wong, Food Addit. Contam. 26 (2009) 1450. [65] A. Earnshaw, R.A. Smith, L. Owen, Food Chem. 113 (2009) 781. [66] M. Thompson, Analyst (Cambridge, UK) 125 (2000) 385. [67] T.P.J. Linsinger, R.D. Josephs, Trends Anal. Chem. 25 (2006) 1125. [68] T.P.J. Linsinger, R.D. Josephs, Trends Anal. Chem. 26 (2007) 659. [69] D.E. Wells, J. de Boer, Trends Anal. Chem. 25 (2006) 350. [70] P. Medina-Pastor, C. Rodriguez-Torreblanca, A. Andersson, A.R. Fernandez-Alba, Trends Anal. Chem. 29 (2010) 70. [71] M. Thompson, K. Mathieson, A.P. Damant, R. Wood, Accred. Qual. Assur. 13 (2008) 223.