VALIDATIEDOSSIER DOSSIER DE VALIDATION SOP/TRA/ANA/0

Transcription

1 Leuvensesteenweg 17 - B 38 Tervuren PRO/5.4/2/DOC1/V3 VALIDATIEDOSSIER DOSSIER DE VALIDATION SOP/TRA/ANA/ TRA/ANA/4 11/9/213 Page 1 of 15

2 PRO/5.4/2/DOC2/V1 HISTORIEK VAN DE VALIDATIEDOSSIER / HISTORIQUE DU DOSSIER DE VALIDATION SOP/TRA/ANA/4 DATUM / DATE FASE WIJZIGING / PHASE - MODIFICATION May 211 October 212 Method development November 212-September 213 November 213 Method validation, and creation of first version of the document of validation After remarks audit BELAC (25/9/213) : adjusted calculation of measurement uncertainty for carrot and mushroom 11/9/213 Page 2 of 15

3 PRO/5.4/2/DOC3/V1 TOEPASSINGSGEBIED VAN DE SOP / DOMAINE D APPLICATION DE LA SOP: Deze methode wordt toegepast op plantaardige matrices (landplanten). Cette procédure s applique aux matrices végétales (plantes terrestres). Définitions As Arsenic As species Specific chemical form of As defined according to e.g. its oxidation state or its molecular structure As III Arsenite As v Arsenate As i Inorganic arsenic (=As III +AS V ) Certified reference material (CRM) A certified reference material (CRM) is a material in which a specific analyte content has been specified. In the present study the CRM NMIJ 753a (rice flour) is used. The material is certified for total As and for inorganic arsenic. Limite of quantification (LOQ) MultiQC Procedure blanc Speciation analysis Supplemented material The limit of quantification (LOQ) is the minimal concentration of an analyte which can be measured in a routine analysis. The limit of quantification for As species is calculated as 1 times the standard deviation of the background signal of the chromatogram control chart of the type Shewart test, or value of a test, corresponding to a complete analytical cycle (preparation+measurement), realized in conditions identical to the sample analysis, but in the absence of the sample analytical activity which identifies and/or quantifies chemical species sample enriched with a known quantity of the analyte of interest VALIDATIEPARAMETERS / PARAMETRES DE VALIDATION The method concerns a quantitative method of confirmation including the following parameters: -calibration -limit of quantification -specificity -trueness -repeatability, reproducibility and measurement uncertainty -peak resolution 11/9/213 Page 3 of 15

4 1. Analytical specificity The measurement of 2 blank samples, as described in PRO/5.4/2, to demonstrate analytical specificity is not for our method. Arsenic is a natural constituent of the earth crust and is ubiquitous in all biological matrices. This means that it is impossible to find blank samples. In trace element analysis by means of ICP-MS, analytical specificity can refer to the absence of significant interferences during the measurements. In case of speciation analysis of arsenic, the most important potential interference is caused by chloride, because the 4 Ar 35 Cl + and 38 Ar- 37 Cl polyatomic ions can interfere with the detection of As species on mass 75. On VARIAN ICP-MS this interference can be counteracted by the use of H 2 as a reaction gas. During the As speciation measurements a H 2 flow of 9mL/min is activated (cfr. SOP/TRA/ANA/4). In order to verify a potential disturbance of the As speciation chromatograms by a Cl - interference signal, HCl solutions of various concentrations.1%,.2%,.5%, 1% were injected and analysed by HPLC-ICP-MS. On these chromatograms, at all injected HCl concentrations, a small peak at the retention time of As V is observed. However, even at the highest Cl concentration this peak is more than ten times below the LOQ of the method (for a detail of results see file Asi_interference Cl in 71\ACCREDITATIE\ I ARSENIC INORGANIC \VALIDATIEDOSSIER inorganic arsenic). Most likely the peak originates from the presence of a very low concentration of inorganic As in the acid, and not from a chloride interference on the 75 As signal. Moreover, such high Cl - concentrations are not expected to be present in the extracts. The potential presence of matrix specific effects or matrix specific interferences is investigated and discussed in paragraph 1 for various food samples of plant origin (rice, carrot, champignons). 2. Robustness The parameters that have to be considered are discussed on p 4/4 3. Precision (repeatability reproducibility) Currently no performance criteria have yet been laid down for inorganic arsenic by the Commission of the European Communities. The performance criteria laid down in Commission Regulation 333/27 (for amongst others cadmium and lead) were used as a guide in the present document. According to the latter Commission Regulation the performance criterion for precision is HORRAT R < 2. The data used to calculate HORRAT R are the data from CRM NMIJ753A (rice flour; certified for total As at 98±7 µg/kg, As III at 71.1 ± 2.9 µg/kg and As V at 13. ±.9 µg/kg; total inorganic As can be calculated as 84.1±3. µg/kg) introduced in the MultiQC between 31/1/212 and 21/2/213 as these data represent reproducibility conditions. The HORRAT R values for inorganic arsenic (cfr. Table 1). are below 2 HORRAT R = Observed RSD R /RSD R Horwitz Table 1. HORRAT R values based on the CRM NMIJ 753 a (rice flour) Asi µg/kg Mean 9.1 SD 3.7 RSD R 4.1 RSD Horwitz 22 HORRAT R.19 11/9/213 Page 4 of 15

5 Repeatability and reproducibility (intermediate precision) were calculated based on different samples of plant origin being: rice, carrots and champignons, supplemented with stock solutions containing inorganic arsenic (spiked as As III ), at two different levels and measured on three different days. The spiking levels were: For carrot and champignon: Natural level of inorganic arsenic (2.4 µg/kg for carrot, 1.6 µg/kg for champignon) 96 µg/kg 192µg/kg For rice: Natural level of inorganic arsenic (136 µg/kg) 272 µg/kg 476 µg/kg The calculations can be found in ValMethTerv 1.4 inorganic As (71\ACCREDITATIE\ARSENIC INORGANIC\VALIDATIEDOSSIER inorganic arsenic). The theoretical natural level was determined as the average of the data for the unspiked samples as determined in a preliminary experiment. The long term reproducibility (intermediate precision) will be calculated when sufficient data will be available. Addenda: -template ValMethTerv 1.4 inorganic arsenic in rice -template ValMethTerv 1.4 inorganic arsenic in carrot -template ValMethTerv 1.4 inorganic arsenic in champignon 4. Accuracy/ recovery Accuracy indicates the deviation between the mean value found and the true value. Currently only a rice certified reference material is available in which the concentrations of As III and As V are certified. For rice the determination of accuracy is based on the difference between the sum of the certified values (As i =As III + As V ) and the measured value of As i. For the other matrices accuracy of the method is illustrated by calculating the recovery of spiked samples. This can be determined by applying the inorganic arsenic analysis method to samples to which known amounts of inorganic arsenic species have been added. The accuracy is then calculated from the test results as a percentage of the analyte recovered. The data that have been used to determine the accuracy in the vegetable samples, are the measurement results of spiked samples used for precision calculation. The calculations can be found in Accuracy calculations inorganic As.xlsx (71\ACCREDITATIE\ARSENIC INORGANIC\VALIDATIEDOSSIER inorganic arsenic). For the certified rice sample the bias of the measured value compared to the certified value is statistically significant. Due to the fact that the uncertainty on the certified value and the uncertainty on the measurement of that material are both very small (1.8% and.7% respectively), even a small bias easily becomes significant. Given the fact that the bias is included in the calculation of the measurement uncertainty (see 7. Measurement uncertainty ), we consider a bias of less than 1% acceptable. 11/9/213 Page 5 of 15

6 Table 3: Mean accuracy for inorganic arsenic in different matrices. As i Rice 17% Carrot 18% Champignons 19% 5. Beslissingsgrens (CCα) en detectievermogen (CCβ) As there are no legal limits for As species in rice, grains or grain products, CCα and CCβ are not. 6. Calibration (Linearity of the method) The calibration is an external calibration of the linear type. Calibration curves for As V (representing inorganic As in the sample) are presented in the file calibration Asi (71\ACCREDITATIE\ARSENIC INORGANIC\VALIDATIEDOSSIER inorganic arsenic). The concentrations used are the following:,.5, 1, 2, 5, 1 en 25 pbb. Based on these concentrations 2 different calibration curves were made: one including all concentrations, and a second one including only the 5 lowest concentrations levels. As these low concentrations correspond to the range of expected sample concentrations (taking into account the dilution factor), and as no difference is observed between the 2 calibration lines, calibration range during routine measurements will be limited to the five lowest concentrations. Calibration details are shown in figure 1 below. Observed R 2 value is >.99, and residuals are <1%. The random variation of the residuals (Res%) with increasing standard concentrations supports the trueness of the linear model. When higher Asi concentrations are present/expected in the samples, there is no need to extent the calibration range (not up to 25 ppb), because -as illustrated in Figure 1- the calibration curve remains linear also at these higher concentration levels. A. Area 1,5 1, 9,5 9, 8,5 8, 7,5 7, 6,5 6, 5,5 5, 4,5 4, 3,5 3, 2,5 2, 1,5 1, ppb 4 5 Regression equation: y= x R 2 =.9998 Name ppb area area(recalc) Res % Point Point Point Point Point /9/213 Page 6 of 15

7 B. 45, 4, 35, 3, Area 25, 2, 15, 1, 5, ppb Regression equation: y= x R 2 =1. Name ppb area area(recalc) Res % Point Point Point Point Point Point Point Figure 1: Examples of As V calibration curves in case of A. calibration range -5 ppb, B. calibration range -25 ppb. 7. Measurement uncertainty The measurement uncertainty was calculated via the method imposed by the Federal Agency for the Safety of the Food Chain 1. The combined measurement uncertainty is calculated from the intra-reproducibility and the bias. %u = (%u Rw ² + %u bias ²) The intra-reproducibility is calculated from the control chart of a certified rice sample (NMIJ 753 a): %u Rw ² = RSD cc ² 1 Procedure Bepaling van de meetonzekerheid voor kwantitatieve chemische analyses. LAB P 58 Meetonzekerheid-v.1-nl. Goedgekeurd 3/11/28 2 ISO/BIPM GUM. 28. Evaluation of measurement data Guide to the expression of uncertainty in measurement. JCGM 1:28 GUM 1995 with minor corrections. 11/9/213 Page 7 of 15

8 for rice matrices, or from the relative standard deviation obtained from the data from section 4 (accuracy) for carrot and mushroom matrices: %u Rw, vegatable ² = RSD vegetable ² The data from section 4 (accuracy) were also used to calculate the bias for rice (based on NMIJ 753a) and for carrot and mushroom (based on recovery experiments). rice: %u bias = %b² + RSD bias ² + %u CRM ² carrot, mushroom : %u bias = (Σ%b i ²)/n with %b i = (x i -C spike, i )/C spike, I and n the number of recovery experiments As only one control material was used, measurement uncertainties were not calculated at different concentration levels with this method. The expanded measurement uncertainty is the product of the combined uncertainty and the coverage factor k = 2, which corresponds to a confidence level of approximately 95%. The calculations of the measurement uncertainties can be found in Measurement uncertainty inorganic arsenic (71\ACCREDITATIE\ ARSENIC INORGANIC\VALIDATIEDOSSIER inorganic arsenic). The combined standard uncertainty u c is 9% for rice, 14% for carrot and 17.5% for mushroom. The extended uncertainty U is 18% for rice, 28% for carrot and 35% for mushroom. As an alternative, the measurement uncertainty can also be calculated via the ISO/BIP GUM method 2 from the uncertainties related to repeatability (r), intermediate precision (ip) and trueness (t): u c = Uncertainties related to repeatability and intermediate precision are taken from section 3 (see ValMethTerv 1.4 inorganic As ). The uncertainty related to trueness is taken from the accuracy determination (section 4). The uncertainty calculations via the ISO/BIP GUM method are included as an illustration, but will not be used for real samples. The extended uncertainty U based on the ISO/BIP GUM method vary from 1 to 16% at the low level, 13-21% at the median level, and 1-23% at the high level. Taking into account the presence of a small bias (7%) as shown in paragraph 4, the extended measurement uncertainties that will be reported, correspond to the ones calculated via the method imposed by the Federal Agency for the Safety of the Food Chain. 8. Quantification limits The limit of quantification (LOQ) is the minimal concentration of an analyte which can be measured in a routine analysis. The limit of quantification for inorganic arsenic corresponds to 1 times the standard deviation of the background signal of the chromatogram. Standard deviations of the background signal of NMIJ 753a chromatograms were calculated for a zone of.5 minute after As V elution. Average calculated LOQ values (µg/kg) obtained on different measurement days are shown in table 4 (values rounded to.5 unity). Table 4. Calculated LOQ values in solution (µg/l) As V Average signal (c/s) 461 Average standard deviation (c/s) 27.5 Relative standard deviation (RSD) 7.7% 11/9/213 Page 8 of 15

9 LOQ in solution (µg/l).2 The calculations are detailed in the file LOD_LOQ calculations (71\ACCREDITATIE\ ARSENIC INORGANIC \Validatiedossier inorganic As\LOD_LOQ)) These calculated LOQ values were verified by measuring As V solutions with similar concentration levels (.1 µg/l,.2 µg/l,.3 µg/l and.4 µg/l) in triplicate. The results are shown in de file Val Asi_LOQ injecties (cfr. 71\ACCREDITATIE\ ARSENIC INORGANIC\VALIDATIEDOSSIER inorganic arsenic\lod_loq. Concentration of.2 µg/l were measured accurately, which confirms the suitability of the calculated LOQ-value. A summary of LOQ values, at different dilution factors of the matrix is presented in table 5. Table 5. Final LOQ values after recalculation to concentration in matrix (µg/kg) taking into account the dilution factor (DF) As V LOQ in solution LOQ in matrix for DF=5 LOQ in matrix for DF =2 LOQ in matrix for DF= Decisions related to the repeatability of results Because of the potential heterogeneity of biological samples, a homogenisation step is included before subsamples are taken for analysis. Indeed, heterogeneity of the sample influences the repeatability of results (variation among subsamples). Each analysis will therefore be performed in 2 replicates, and the following criteria will be taken into account: -Values of the reference sample need to fall within the limits specified in the file Accept CRM. These values take into account the extended measurement uncertainty of the analysis. -To be accepted, measurement results have to be higher than the LOQ. Results below this value will be reported as <LOQ. - Analyses will be performed in 2 replicates. The coefficient of variation needs to be <25%. The value of 25% is determined based on the extended measurement uncertainty of inorganic arsenic in the various matrices, which varies from 18% to 35%. -When the coefficient of variation is higher than 25%, the analysis will be repeated in 2 replicates. 1. Specificity of the method in various matrices For the determination of specificity the results of the spiking experiments were used in which known amounts of the inorganic arsenic were added to the matrices under investigation (rice, carrot, mushroom) and to some additional matrices (wine and babyfood-vegetable puree). Calculation of the slopes between the nominal and measured values of inorganic arsenic (Table 7), shows that these slopes are generally fluctuating around 1, which confirms the specificity of the method in the matrices concerned. The calculations are detailed in the file Specificity of Asi in different matrices.xls (71\ACCREDITATIE\ ARSENIC INORGANIC\VALIDATIEDOSSIER inorganic arsenic). 11/9/213 Page 9 of 15

10 Table 7: Slopes of As i spiked in various matrices. Significant differences in slopes are indicated by different superscript letters. Arsenic species Matrix Slope As i Rice 1.16 b As i Carrot.92 a As i Mushroom.95 a As i Wine.99 a As i Babyfood (vegetable puree).92 a 11. Peak resolution Peak resolution (Rs) can be calculated according to the formula Rs= 2(Tr2-Tr1)/(W1+W2), with Tr2 en Tr1 retention times of two consecutive peaks, and W2 en W1 baseline width of the corresponding peaks. Baseline peak width (W) corresponds to the distance between the baseline intercepts of tangent lines to the front slope and the back slope of a peak (Figure 6). This baseline peak width, W, is considered equivalent to the peak width at 13.4 % peak height, or for Gaussian peaks can be approached by taking 1.7* peak width at 5% peak height. Figure 6: Illustration of the baseline peak width, W, which corresponds to the distance between the baseline intercepts of tangent lines to the front slope and the back slope of the peak. A typical chromatogram of a 5 ppb standard solution containing a mix of 4 As species (AB, DMA, MMA, As V ) is presented in Figure 7. 11/9/213 Page 1 of 15

11 6, 55, 5, 45, 4, c/s AsB DMA solb As b_5ppb_13416_29.data 35, 3, 25, MMA As5 2, 15, 1, 5, SPW.7 STH STH # Name Time [Min] Quantity [ppb] Height [c/s] Area [c/s.min] Area % [%] Width 13.4% [Min] 1 AsB DMA MMA As Figure 7: Chromatogram of a 5 ppb standard solution containing a mix of 4 As species (AB, DMA, MMA, As V ). Values for retention times and peak width at 13.4% are presented in Figure 7. For the current method, describing the determination of Asi in food of plant origin, only the resolution of the peak-pair MMA-As V is important. Using the values presented in Figure 7 the resolution, Rs, of this peak-pair corresponds to a value Rs= 2.3, indicating an adequate separation. Rs values >1.3 are characteristic for well separated peaks. The latter value will be used as separation criterium during routine measurements RT [min] 11/9/213 Page 11 of 15

12 PRO/5.4/2/DOC4/V1 FACTOREN DIE DE JUISTHEID EN BETROUWBAARHEID VAN DE BEPROEVINGEN BEÏNVLOEDEN FACTEURS QUI INFLUENCENT L EXACTITUDE ET LA FIABILITE DES ESSAIS SOP/TRA/ANA/3 Parameter Composants character Control 1.Equipement and laboratorium material 2. Products and samples 3.Method principles 4.Room temperature 5.Staff mixer balances pipettes glassware for standards microwave teflon tubes for microwave ICP-MS HPLC pumps HPLC autosampler ph-meter milliq water Standards for As species Ammonium carbonate mobile phase 1 routine samples 2 ringtest samples Ion exchange colonne Temperature of HPLC-chain Room temperature of LC-ICP-MS laboratory Analists, technical responsible non if clean and working non if clean non if clean non if working and injection volume constant non if produced daily fresh (physic-chemical characteristiques of the column) non because of thermostatisation by oven non mixer is cleaned with bi-distilled water after use maintenance by external firm 1x/year and daily verification control by external firm 2x/year and maintenance by external firm 1x/year flasks stored in bi-distilled water control by external firm 3x/year and maintenance by external firm 1x/year. cleaning of tubes with HNO3 25% after use (wash cycle) verification of tube state at each use; if cracks are observed in a tube, it is not used anymore 1 maintenance by external firm 1x/year or interventions mentioned in logbook 2 daily control of sensitivity of the apparatus by use of a tuning solution (cfr SOP/TRA/ANA/2). intervention by firm when a pump related problem is observed (see logbook ICP-MS) check injection volume 1x/year (RSD<.5%) ph meter is calibrated daily and T dependency is taken into account via maintenance mentioned in logbook and daily verification daily fresh preparations of calibration dilutions constant use of same mark daily fresh preparation 1 verification of extraction efficiency in sample (total As in extract versus total As after mineralisation) 2 contrôle of Z-score of PT-Test constant use of same column mark cleaning or replacement of column if problems are observed (high pressure, bad peak separation, tailing of peaks, ) control of oven temperature 1x/year daily tuning of ICP-MS signal with a tuning solution optimises signal in function of daily conditions (cfr SOP/TRA/ANA/2) first, second, third line controls formation of people 11/9/213 Page 12 of 15

13 Addendum 1: 1 Template ValMethTerv 1.4 inorganic arsenic in rice Notes for user : This page is for up to 5x1x3. Only blue cells are for inputs. N(repl) and N(days) calculated automatically (Check!). For CCa etc see instructions on line 46. CALCULATION of REPEATABILITY, WITHIN-LAB REPRODUCIBILITY, CCa, CCb and RECOVERY ACCORDING TO 22/657/EC, ISO and Van Loco & Beernaert (24) DATA : RESULTS AND GRAPHS : Analyte : Asi spiked in rice Slope intercept formulas : SSrepl = Ed Er (Y-Ymoy(level,day))2 Method : 121.8% Srep2 = Srepl2 = MSrepl = SSrepl/(n-ndays) MRL : 272 NB: no MRL available! SSdays = nrepl E(Ymoy(level,day)-Ymoy(level))2 unit : ug/kg t(26,.5) Yc CCa CCb MSdays = SSdays / (ndays-1) Nrepl : 3 (number of replicates) ug/kg % of MRL ug/kg % of MRL Sdays2 = (Msdays-Msrepl) / nrepl Ndays : 3 (number of days or analysts) % % Srw2 = Sdays2 + Srep2 Nlev : 3 (number of levels)) nom.ccn av.meas. uncertainty (st.dev.) ext. unc. (2 x st.dev.) Measured concentrations : ug/kg ug/kg ug/kg % ug/kg % (NOT corrected for recovery) : % % % % day or analyst : % % level 1 : level 2 : level 3 : replicate : replicate : replicate : SSrepl MSrepl Srepl average for level SSdays MSdays Sdays Sdays Srw SSrepl MSrepl Srepl average for level SSdays MSdays Sdays Sdays Srw SSrepl MSrepl Srepl average for level SSdays MSdays Sdays Sdays Srw Level Level Level % 6% 5% 4% 3% 2% 1% % linear regression y = 1.218x R² = Uncertainty Srepl Sdays Srw Uncertainty (%) reps days rw /9/213 Page 13 of 15

14 Addendum 2: Template ValMethTerv 1.4 inorganic arsenic in carrot Notes for user : This page is for up to 5x1x3. Only blue cells are for inputs. N(repl) and N(days) calculated automatically (Check!). For CCa etc see instructions on line 46. CALCULATION of REPEATABILITY, WITHIN-LAB REPRODUCIBILITY, CCa, CCb and RECOVERY ACCORDING TO 22/657/EC, ISO and Van Loco & Beernaert (24) DATA : RESULTS AND GRAPHS : Analyte : Asi spiked in carrots Slope intercept formulas : SSrepl = Ed Er (Y-Ymoy(level,day))2 Method : 14.1% 1.57 Srep2 = Srepl2 = MSrepl = SSrepl/(n-ndays) MRL : 96 NB: no MRL available! SSdays = nrepl E(Ymoy(level,day)-Ymoy(level))2 unit : ug/kg t(26,.5) Yc CCa CCb MSdays = SSdays / (ndays-1) Nrepl : 3 (number of replicates) ug/kg % of MRL ug/kg % of MRL Sdays2 = (Msdays-Msrepl) / nrepl Ndays : 3 (number of days or analysts) % % Srw2 = Sdays2 + Srep2 Nlev : 3 (number of levels)) nom.ccn av.meas. uncertainty (st.dev.) ext. unc. (2 x st.dev.) Measured concentrations : ug/kg ug/kg ug/kg % ug/kg % (NOT corrected for recovery) : % % % % day or analyst : % % level 1 : level 2 : level 3 : replicate : replicate : replicate : SSrepl MSrepl Srepl average for level 5.28 SSdays 6.14 MSdays 7.33 Sdays Sdays Srw SSrepl MSrepl Srepl average for level SSdays MSdays Sdays Sdays Srw SSrepl MSrepl Srepl average for level SSdays MSdays Sdays Sdays Srw Level Level Level % 8% 6% 4% 2% % linear regression y = 1.48x R² = Uncertainty Srepl Sdays Srw Uncertainty (%) reps days rw /9/213 Page 14 of 15

15 C O D AA-TERVUREN - C E R V AA-TERVUREN Addendum 3: 3: Template ValMethTerv 1.4 inorganic arsenic in mushroom 11/9/213 Page 15 of 15