May 201 6 DLA 49/201 5 - Contact Material II: Nickel Release DLA Evaluation Report proficiency test Dienstleistung Lebensmittel DLA 49/2015 Analytik GbR Contact Material II: Nickel Release of Metal Contact Material (Costume Jewelry and Coin) Dienstleistung Lebensmittel Analytik GbR Waldemar-Bonsels-Weg 170 22926 Ahrensburg, Germany [email protected] www.dla-lvu.de Coordinator of this PT: Dr. Matthias Besler Reprint, also in part, only with written permission from DLA-Ahrensburg Page 1 of 22 May 201 6 DLA 49/201 5 - Contact Material II: Nickel Release Inhalt / Content 1. Introduction.........................................................3 2. Realisation..........................................................3 2.1 Test material...................................................3 2.1.1 Homogeneity....................................................4 2.2 Shipment of test items and information for analysis..............5 2.3 Submission of results............................................5 3. Evaluation...........................................................6 3.1 Assigned value...................................................6 3.2 Standard deviation...............................................6 3.3 Outliers.........................................................6 3.4 Target standard deviation........................................6 3.4.1 General model (Horwitz)........................................7 3.4.2 Value by precision experiment..................................7 3.4.3 Value by perception............................................8 3.5 z-Score..........................................................9 3.6 z'-Score.........................................................9 3.7 Quotient ........................................................9 3.8 Standard uncertainty............................................10 4. Results.............................................................11 4.1 Test Item I: Nickel Release in µg/cm2/week......................12 4.2 Test Item II: Nickel Release in µg/cm2/week.....................14 5. Documentation.......................................................16 5.1 Primary data....................................................16 5.2 Homogeneity.....................................................18 5.3 Analytical Methods..............................................19 6. Index of participant laboratories...................................21 7. Index of references.................................................22 Reprint, also in part, only with written permission from DLA-Ahrensburg Page 2 of 22 May 201 6 DLA 49/201 5 - Contact Material II: Nickel Release 1. Introduction The participation in proficiency testing schemes is an essential element of the quality-management-system of every laboratory testing food and feed, cosmetics and food contact materials. The implementation of proficiency tests enables the participating laboratories to prove their own analytical competence under realistic conditions. At the same time they receive valuable data regarding the validity of the particular testing method. The purpose of DLA is to offer proficiency tests for selected parameters in concentrations with practical relevance. Realisation and evaluation of the present proficiency test follows the technical requirements of DIN EN ISO/IEC 17043 (2010) and DIN ISO 13528:2009 (6). 2. Realisation 2.1 Test material Two different test materials were offered to be analysed. According to pre-tests the nickel release from the first material was below the limit of migration of 0,5 µg/cm2/week (0,88 µg/cm2/week respectively) for articles intended to come into direct and prolonged contact with the skin according to EU-regulation 1907/2006 and EN 1811 (ASU §64 82.02-6) while the nickel release of the second material was above the migration limit (13). Test item I (jewelry): The test material is a bracelet (costume jewelry) made of metal (diameter appr. 6,6 cm, hight appr. 4 mm, see fig. 1a). The material was purchased in the trade by DLA as specimen from one production unit. Test item II (coin): The test material is a 2 Euro commemorative coin from the Federal Republic of Germany: material ring copper/nickel alloy CuNi25 and kernel brass (diameter 25,75 mm, hight 2,20 mm, see fig. 1b). The material was purchased in the trade by DLA as specimen from one production unit (original rolls, 25 x 2 Euro commemorative coins 2010 “City Hall and Roland (Bremen)”, coining site A). Fig. 1: a) left bracelet test item I and b) right 2 Euro coins Reprint, also in part, only with written permission from DLA-Ahrensburg Page 3 of 22 May 201 6 DLA 49/201 5 - Contact Material II: Nickel Release The samples were packed in transparent plastic bags and labeled. 2.1.1 Homogeneity The suitability of the test material was checked by multiple determinati- ons of nickel release according to ASU B 82.02-6 (corresponds to EN 1811-2012). The mean was 0,078 µg/cm2/week for test item I and 59,5 µg/cm2/week for test item II. With 19% and 8,1% respectively the repeata- bility standard deviations were considered acceptable in comparison to the combined measurement uncertainty of 46% (EN 1811, annex A). The re- sults are given in the documentation (13). The calculation of the repeatability standard deviation of the partici- pants was also used as an indicator of homogeneity. The repeatability standard deviation between samples was calculated from the 3 results of each participant. For test item II the relative repeatability standard deviation was 6,2%. For test item I the results of participants showed two groups of results for which the mean differed about 10 times. Therefore all subsequent cal- culations for test item I were exclusively done with results > 0,010 µg/cm2/week. For 2 samples each of 5 participants the repeatabi- lity standard deviation within the (remaining) samples was 28%. With respect to the respective mean of test items I and II the repeatabi- lity standard deviations of participants were in an usual range of the method (13). The repeatability standard deviations of the participants' results are given in the documentation. Reprint, also in part, only with written permission from DLA-Ahrensburg Page 4 of 22 May 201 6 DLA 49/201 5 - Contact Material II: Nickel Release 2.2 Shipment of test items and information for analysis Three samples A, B and C of the test item I were sent to every participating laboratory in the 48th week of 2015 and three samples A, B and C of test item II in the 5th week of 2016. The tests should be finished at 15th january 2016 and 11th march 2016, respectively. With the cover letter the following conditions according to the German official methods ASU B 82.02-6 (corresponding to EN 1811-2012) for testing of nickel release of articles intended for skin contact were set to by applied (13): 1. test solution (EN 1811): 0,5% (m/m) NaCl, 0,1% (m/m) lactic acid, 0,1% (m/m) urea, adjust to pH 6,5 2. time and temperature (30°C, 168h) 3. results given in µg/cm2/week. 2.3 Submission of results The participants submitted their results in provided standard forms by email. The concentration of the parameters as given by the participants in the column "final result" was used for each statistical evaluation, in case at least 7 results were submitted. In case of test item I due to a bimodal distribution of results (all single results estimated by kernel density plot) exclusively results > 0,010 µg/cm2/week were considered for statistical evaluation. Queried and documented were single results and the testing methods used. All participants submitted their results in time. Reprint, also in part, only with written permission from DLA-Ahrensburg Page 5 of 22 May 201 6 DLA 49/201 5 - Contact Material II: Nickel Release 3. Evaluation 3.1 Assigned value Because the analysed material was no certified reference material the robust mean of the submitted results was used as assigned value X (6). For test item I the kernel density estimation of all single results (res- ults given as “below” excluded) showed a distribution of results in two major groups for which the mean differed about 10 times. Therefore all subsequent calculations for test item I were exclusively done with res- ults > 0,010 µg/cm2/week. For test item II the distribution of submitted results showed no hint for bimodal distribution or other reasons for a higher variability. The statistical evaluation is performed when there are at least 7 results for a parameter. Single results giving values outside the measuring range of the participating laboratory or given as „0“ are not considered for statistical evaluation (e.g. results given as > 25 mg/kg and < 2,5 mg/kg, respectively) (6). 3.2 Standard deviation For comparison to the target standard deviation a robust standard devi- ation (Sx) was calculated (6). 3.3 Outliers Statistical outliers were determined by Mandel´s-H-Statistic for 95% significance niveau (5). Detected outliers were stated for information only, when z-score was < -2 or > 2. 3.4 Target standard deviation The target standard deviation of the assigned value is determined according to the following methods. In general the Horwitz target standard deviation is suitable for the statistical evaluation of interlaboratory tests where different analytical methods are applied. The standard deviation from precision experiments are derived from proficiency tests where a specific analytical method is mandatory. For information the z-scores from both models were given in the result tables, if available. For valuation of the present results neither the general model according to Horwitz (s. 3.4.1) nor the values by precision experiments (s. 3.4.2) were suitable. The valuation was done according to chapter 3.4.3 and followed the prin- Reprint, also in part, only with written permission from DLA-Ahrensburg Page 6 of 22 May 201 6 DLA 49/201 5 - Contact Material II: Nickel Release ciple of "fitness for purpose" in order to ensure the suitability for decisions with respect to allowed maximum migration levels. 3.4.1 General model (Horwitz) The relative target standard deviation in % of the assigned value is derived from following equation (Horwitz) σ = 2(1-0,5logX) (%) From the result the target standard deviation is calculated σ = X * σ / 100. (%) 3.4.2 Value by precision experiment Using the reproducibility standard deviation σ and the repeatability R standard deviation σ of a precision experiment the between-laboratories r standard deviation can be calculated σ : L  =2−2 . L R r And then, using the number of replicate measurements n, each participant is to perform, the target standard deviation for proficiency assessment is calculated : =22/n . L r The target standard deviations in table 1 were calculated from the precision data of the respective method. For the determination of nickel release the data of repeatability and reproducibility standard deviations are not sufficiently given in ASU §64 B 82.02-6 and EN 1811, respectively (13). In a proficiency test of quality control material a nickel migration rate of 0,31±0,06 µg/cm2/week with a relative reproducibility precision of 33,3% was obtained (annex B). Reprint, also in part, only with written permission from DLA-Ahrensburg Page 7 of 22 May 201 6 DLA 49/201 5 - Contact Material II: Nickel Release 3.4.3 Value by perception The target standard deviation for proficiency assessment can be set at a value that corresponds to the level of performance that the coordinator would wish laboratories to be able to achieve (6). Because neither the general model (s. 3.4.1) nor the values of precision experiments (s. 3.4.2) were suitable for valuation of the results, we choose the following basis for evaluation of the results. According to DIN EN 1811 annex A (ASU B 82.02-6) the combined measurement uncertainty of the method is 46%. The expanded measurement uncertainty is applied in order to identify significant exceeding of the maximum migra- tion limit. For this purpose the combined uncertainty is multiplied with the coverage factor k (1,65) for significance niveau 0,05 (13). For the present evaluation of results a suitable target standard deviati- on was set based on the combined uncertainty of EN 1811 considering the value of the respective robust mean: 1) Target standard deviation for test item I The nickel release of the material is below the migration limit of 0,5 µg/cm2/week (and 0,88 µg/cm2/week respectively). Half of the value for the expanded measurement uncertainty was considered for the target standard deviation. Consequently the relative standard target deviation is: 1,65 x 46% / 2 = 38%. Hereby it is ensured, that the valuation of results by z-scores is com- parable to the criterium for exceeding of the maximum migration limit. The limit of the z-score ≥ -2 and ≤ 2 correspond to the decision limits of the expanded uncertainty. 2) Target standard deviation for test item II The nickel release of the material is clearly above the migration limit of 0,5 µg/cm2/week (and 0,88 µg/cm2/week respectively). Therefore a lower target standard deviation was considered: half of the value for the com- bined measurement uncertainty. Hereby the uncertainty is not expanded. Consequently the relative standard target deviation is: 1/2 x 46% = 23%. Reprint, also in part, only with written permission from DLA-Ahrensburg Page 8 of 22 May 201 6 DLA 49/201 5 - Contact Material II: Nickel Release 3.5 z-Score To assess the results of the participants the z-score is used. It indicates about which multiple of the target standard deviation ( σ ) the result (x) of the participant is deviating from the assigned value (X) (6). Participants’ z-scores were derived as: z = (x – X) / σ ; the requirements for the analytical performance are generally considered as fulfilled if -2 ≤ z ≤ 2 . 3.6 z'-Score The z'-score can be used for the valuation of the results of the participants, in cases the standard uncertainty has to be considered (s. 3.8). The z'-score represents the relation of the deviation of the result (x) of the participant from the respective assigned value (X) to the square root of quadrat sum of the target standard deviation ( σ̂ ) and the standard uncertainty (Ux) (6). Participants’ z'-scores are derived as: If applied, we define the denominator as the target standard σ deviation ' in following sections. The requirements for the analytical performance are generally considered as fulfilled if -2 ≤ z' ≤ 2 . Sx/ 3.7 Quotient Following the Horrat-value the results of a proficiency-test (PT) can be considered convincing, if the quotient of robust standard deviation and target standard deviation does not exceed the value of 2. A value > 2 means an insufficient precision, i.e. the analytical method is too variable, or the variation between the test participants is higher than estimated. Thus the comparability of the results is not given (11). In the present proficiency tests the quotients Sx/ σ̂ were In the present proficiency test the quotient Sx/ σ̂ for test item I was below 2,0 and for test item II above 2,0. Therefore for evaluation of Reprint, also in part, only with written permission from DLA-Ahrensburg Page 9 of 22 May 201 6 DLA 49/201 5 - Contact Material II: Nickel Release test item II the z'-Score (s. 3.6) considering the standard uncertainty of the assigned value was applied (s. 3.8). 3.8 Standard uncertainty The assigned value X has a standard uncertainty u that depends on the X analytical method, differences between the analytical methods used, the test material, the number of participant laboratories and perhaps on other factors. The standard uncertainty u for this PT is calculated X as follows (6). u =1,25∗Sx/ p X If u ≤ 0,3∗ the standard uncertainty of the assigned value needs X not to be included in the interpretation of the results of the PT (6). Values exceeding 0,3 imply, that the target standard deviation could be too low with respect to the standard uncertainty of the assigned value. In the present proficiency tests the quotients U /σ̂ were 0,68 and 0,67, x respectively. For test item II the quotients U /σ̂ and Sx/ σ̂ were > 0,3 x and > 2,0 at the same time, therefore the z'-Score considering the standard uncertainty of the assigned value was applied (s. 3.6). Reprint, also in part, only with written permission from DLA-Ahrensburg Page 10 of 22