Food

Elemental Analysis Manual: Section 4.3B Appendix B - Supplemental Information on Interlaboratory Trials

<< Elemental Analysis Manual (EAM) for Food and Related Products 

Version 1 (June 2008)
Authors: William R. Mindak

Table of Contents

4.3B.1 ANALYTICAL LIMITS

4.3B.2 REFERENCE MATERIAL RESULTS

4.3B.3 FOOD RESULTS

4.3B.4 OTHER SUPPORTING RESULTS

4.3B.5 CONCLUSION

GLOSSARY

An FDA interlaboratory method trial was undertaken to evaluate proposed EAM Method 4.3 (draft E, January 2002). The method describes procedures for determination of cadmium and lead in food by graphite furnace atomic absorption spectrometry (GF-AAS) after microwave assisted nitric acid decomposition and includes specific analytical quality controls. An important quality control measure is the recovery of the analyte from a fortified analytical solution (FAS) to determine if matrix interference is present. If matrix interference is present then additional dilution of the analytical solution or standardization by method of standard additions was required.

Four FDA laboratories participated in the study administered by CFSAN/Elemental Research Branch: Northeast Regional Laboratory, Southeast Regional Laboratory, Atlanta Center for Nutrient Analysis and the Center for Food Safety and Applied Nutrition (CFSAN). All participating laboratories used the same brand of instrument (Perkin Elmer) equipped with a transverse heated graphite furnace. Each laboratory was given 3 reference materials and 3 foods for analysis by EAM Method 4.3 (draft E) for cadmium and lead. In addition, each laboratory was also given high-purity nitric acid, cadmium and lead stock solution, independent check solution, a copy of the method, and reporting forms. CFSAN also determined cadmium and lead in its GF-AAS analytical solutions by inductively coupled plasma-mass spectrometry (ICP-MS) using an Agilent 7500C instrument.

4.3B.1 ANALYTICAL LIMITS

4.3B Tables 1 and 2 list GF-AAS instrument conditions and standardization information for cadmium and lead, respectively, used by the participating laboratories. The limit of quantification (LOQ) for each element is calculated using the analytical solution quantification limit (ASQL), analytical portion and dilution of the analytical portion. Only results above the laboratory's LOQ are used for assessing accuracy and precision except results of reference materials, which are assessed only if the certified value is above the LOQ. The characteristic mass (m0), ASDL, and ASQL reported by each laboratory are listed in 4.3B Table 3.

Note: When the interlaboratory trial was conducted, the protocol at the time required ASQL and LOQ be calculated based on 10 times the standard deviation of the blank. The values reported in this appendix thus reflect "10×σ" values rather than the current protocol of "30×σ".

Instrument sensitivity for cadmium was similar among the laboratories as indicated by the small variation in m0 values. Except for Laboratory 1, ASDL and ASQL values for cadmium varied by a factor of 5, which is not an unreasonable variation between laboratories. Laboratory 1 reported a value of 0 for cadmium's ASDL and ASQL that is not a real representation of these analytical limits. As directed by the method, a value of 0 is avoided by obtaining an appropriate number of significant digits for the data used to calculate ASDL and ASQL.

Instrument sensitivity for lead was similar for Laboratories 2, 3 and 4 as indicated by the small variation in m0 values. The difference in m0 between Laboratory 1 and the other laboratories is probably due to the use of end-capped tubes by Laboratory 1 verses regular tubes by the other laboratories. ASDL and ASQL values only varied by about a factor of 2 (excluding Laboratory 1's values).

4.3B Table 1. GF-AAS Instrument Conditions for Cadmium

Laboratory:1234
GF-AAS ManufacturerPEPEPEPE
Instrument ModelPE 4100ZLPE 4100ZLPE 5100ZLPE 5100ZL
Lamp Type (Cd)EDLEDLEDLEDL
Wavelength (nm)228.8228.8228.8228.8
Char Temp. °C600600600650
Atomization Temp. °C1600160016501700
Standardization
  Standard 1 (µg/L)0.30.510.5
  Standard 2 (µg/L)0.5121
  Standard 3 (µg/L)1232
  Standard 4 (µg/L)2344
  Standard 5 (µg/L)34--
AlgorithmLinearLinearLinearLinear

4.3B Table 2. GF-AAS Instrument Conditions for Lead

Laboratory:1234
GF-AAS ManufacturerPEPEPEPE
Instrument ModelPE 4100ZLPE 4100ZLPE 5100ZLPE 5100ZL
Lamp Type (Pb)HCLHCLEDLHCL
Wavelength (nm)283.3283.3283.8283.3
Char Temp. °C750750820800
Atomization Temp. °C1700170017001600
Standardization
Standard 1 (µg/L)3552.5
  Standard 2 (µg/L)510105
  Standard 3 (µg/L)10201510
  Standard 4 (µg/L)20302020
  Standard 5 (µg/L)3040--
AlgorithmLinearLinearLinearLinear

4.3B Table 3. Analytical Sensitivity and Limits

Laboratory:1234
Cadmium
mo (pg)11.51.61
ASDL (µg/L)00.0270.0470.010
ASQL (µg/L)00.0640.110.030
Lead
mo (pg)23363133
ASDL (µg/L)0.650.900.870.39
ASQL (µg/L)1.52.12.20.93

back to top

4.3B.2 REFERENCE MATERIAL RESULTS

Three reference materials were included in the study: Cocoa Powder (CFSAN in-house reference material), Bovine Liver (NIST SRM 1577b) and Trace Elements in Spinach (NIST SRM 1570). These materials were chosen because the reference value lead levels are near the estimated limit of quantification (LOQ) or because of the analytical challenge presented by the matrix. Reference materials were analyzed, in duplicate, without drying. Reference material results were corrected for moisture as determined at CFSAN by freeze-drying portions of the reference materials. The moisture values used for conversion are listed in 4.3B Table 4. 4.3B Tables 5 to 10 list the reference material results. Additional dilution of the analytical solution was necessary in some cases to overcome matrix interference. The HORRAT value1 was calculated and used to assess results. The HORRAT value is the ratio of the reproducibility relative standard deviation, expressed as a percent (RSDR, %) to the predicted reproducibility relative standard deviation, expressed as a percent (PRSDR, %) Results with HORRAT values greater than 2 were investigated for outliers.

4.3B Table 4. Moisture in RMs

Reference MaterialMoisture (%)
Cocoa Powder (CPFDA)4.90
Bovine Liver (NIST SRM 1577b)4.40
Spinach (NIST SRM 1570)5.70

Cadmium

Cadmium results were generally good (4.3B Tables 5 to 7) and all levels were well above LOQ in each reference material. Mean RM recoveries for the 3 reference materials were excellent (87% to 102%). Results from Laboratory 1 were consistently lower than the other laboratories but no outliers were detected using the Dixon outlier statistical test. Results from Laboratory 4 were consistently higher than the reference value, although by an acceptable amount. Unfortunately, Laboratory 1 reported an ASDL and ASQL of 0 so no estimate of LOQ was possible. Assuming Laboratory 1's LOQ was similar to those of the other laboratories then all cadmium values were well above LOQ. All analytical solutions required additional dilution to obtain a cadmium concentration within the calibration curve or to diminish matrix interference. The dilution factor (DF) required for a given sample varied approximately by a factor of 2. The HORRAT value was ≤1.5 for each reference material indicating between lab variability was in the expected range. ICP-MS results were in agreement with GF-AAS results and reference values.

4.3B Table 5. Cadmium in Cocoa Powder (CPFDA)

Laboratory:1234
Rep 1 Anal Portion (g)0.50430.48710.50710.5626
Rep 2 Anal Portion (g)0.50220.49610.49080.5512
ASQL (µg/L)-0.0640.110.030
Rep 1 Result (µg/L)5.257.227.608.96
Rep 2 Result (µg/L)5.156.927.228.84
DF54.324
LOQ (mg/kg)-0.020.020.006
Rep 1 Result (mg/kg)0.2730.3840.3940.419
Rep 2 Result (mg/kg)0.2680.3700.3870.422
Mean Result (mg/kg)0.2710.3770.3910.421
Interlaboratory Mean Result ± SD = 0.365 ± 0.0653 mg/kg
RSDR = 17.9%; PRSDR = 18.6%; HORRAT = 1.0
Reference Value ± SD = 0.388 ± 0.059 mg/kg
RM Recovery (%)7097101109
Interlaboratory Mean RM Recovery = 94%
ICP-MS Result (mg/kg)   0.387

4.3B Table 6. Cadmium in Bovine Liver (NIST SRM 1577b)

Laboratory:1234
Rep 1 Anal Portion (g)0.52240.49980.50890.5386
Rep 2 Anal Portion (g)0.51450.50170.51480.5284
ASQL (µg/L)-0.0640.110.030
Rep 1 Result (µg/L)8.659.5411.411.3
Rep 2 Result (µg/L)8.509.3011.111.2
DF56410
LOQ (mg/kg)-0.020.030.02
Rep 1 Result (mg/kg)0.4330.4980.5860.548
Rep 2 Result (mg/kg)0.4320.4840.5630.552
Mean Result (mg/kg)0.4330.4910.5740.550
Interlaboratory Mean Result ± SD = 0.512 ± 0.0634 mg/kg
RSDR = 12.4%; PRSDR = 17.7%; HORRAT = 0.7
Reference Value ± SD = 0.50 ± 0.03 mg/kg
RM Recovery (%)8798115110
Interlaboratory Mean RM Recovery = 102%
ICP-MS Result (mg/kg)   0.507

4.3B Table 7. Cadmium in Spinach (NIST SRM 1570)

Laboratory:1234
Rep 1 Anal Portion (g)0.52950.49170.48920.5270
Rep 2 Anal Portion (g)0.52850.49220.48800.5553
ASQL (µg/L)-0.0640.110.030
Rep 1 Result (µg/L)22.925.121.930.9
Rep 2 Result (µg/L)22.425.623.533.3
DF10161010
LOQ (mg/kg)-0.060.060.02
Rep 1 Result (mg/kg)1.151.361.251.56
Rep 2 Result (mg/kg)1.121.381.271.59
Mean Result (mg/kg)1.131.371.261.57
Interlaboratory Mean Result ± SD = 1.33 ± 0.186 mg/kg
RSDR = 13.9%; PRSDR = 15.3%; HORRAT = 0.9
Reference Value (non-certified) = 1.5 mg/kg
RM Recovery (%)769184105
Interlaboratory Mean RM Recovery = 87%
ICP-MS Result (mg/kg)   1.48

back to top

Lead

Lead results (4.3B Tables 8 to 10) did not agree as well as the cadmium results. This can be attributed, in part, to the low levels of lead (near or below LOQ) in reference materials Cocoa Powder and Bovine Liver. However, results for the reference material Spinach was quite variable even though lead was present at approximately 5 times LOQ. ICP-MS results were in agreement with reference values.

The reference value of lead in Cocoa Powder was near or below the LOQ for Laboratories 2-4 and the obtained results are consistent with the LOQ (4.3B Table 8). Laboratory 1's LOQ indicated the ability to measure the reference value but the RM recovery was very high (255%) and matrix interference was not observed. The reason for this excessively high result is unclear but one possible explanation is that LODs and LOQs based on measurements of a pristine matrix (1% nitric acid or method blanks) do not accurately estimate these limits for a sample matrix. The trace level found by Laboratory 4 was very close to the reference value (109%). This laboratory also determined that there was matrix interference (using FAS recovery) and diluted the analytical solution by a factor of 3. The trace levels of Laboratories 2 and 3 were 136% and 48%, respectively, of the reference value. Neither of these laboratories experienced matrix interference and thus did not dilute the analytical solution. Cocoa Powder was a challenging sample because of the low level of lead and the possibility of matrix interference. FAS recovery did not detect possible matrix interference for Laboratories 1-3. Inaccurate assessment of or inability to detect matrix interference could have contributed to the poor results for Laboratories 1-3 but the low lead concentration was probably a factor as well. The high result from Laboratory 1 (255%) cannot be explained from matrix interference because this type of interference usually causes signal suppression and thus low results.

The reference value of lead in Bovine Liver was near or below the LOQ for Laboratories 2-4 and the obtained results are consistent with the LOQ (4.3B Table 9). Laboratory 1's LOQ indicated the ability to measure the reference value but the RM recovery was very high (232%) and matrix interference was not observed. Laboratories 3 and 4 had acceptable recoveries of 111% and 99%, respectively even though Laboratory 4's value is a trace level. Laboratory 1's lead result of 0.299 mg/kg (232% recovery) failed the Dixon outlier test (at 5% possible false rejection) and was labeled as an outlier. Although no reason for this outlier is clear, this laboratory's LOQ may be an under estimated value. As with Cocoa Powder, only Laboratory 4 found it necessary to dilute the analytical solution due to matrix interference.

The reference value of lead in Spinach was above the LOQ for all laboratories (4.3B Table 10). Recovery of lead in Spinach ranged from 12% to 101% of the reference value with a mean of 65%. This reference material was meant to provide a challenging matrix with a lead concentration well above LOQ. The relatively high lead content was supposed to allow quantification even after extensive dilution of the analytical solution due to matrix interference. Laboratory 4 found extensive matrix interference that required a DF of 4 in order to achieve an accurate result (101% recovery). In contrast, Laboratory 2 was able to achieve a good result (99% recovery) without the need for further dilution. The reason behind this difference in interference assessment needs further investigation. Laboratories 1 and 3 had poor recoveries of 12% and 49%, respectively. A possible explanation for these low recoveries is an improper assessment of interference. A DF of 3 or more should have been required.

4.3B Table 8. Lead in Cocoa Powder (CPFDA)a

Laboratory:1234
Rep 1 Anal Portion (g)0.50430.48710.50710.4811
Rep 2 Anal Portion (g)0.50220.49610.49080.5512
ASQL (µg/L)1.62.12.20.93
Rep 1 Result (µg/L)6.62.340.862.16
Rep 2 Result (µg/L)5.93.321.122.56
DF1113
LOQ (mg/kg)0.080.20.10.2
Rep 1 Result (mg/kg)0.2990.1240.0420.118
Rep 2 Result (mg/kg)0.2640.1760.0630.122
Mean Result (mg/kg)0.2810.150 (TR)0.053 (TR)0.120 (TR)
Reference Value ± SD = 0.110 ± 0.022 mg/kg
RM Value Recovery (%)25513648109
ICP-MS Result (mg/kg)   0.117

a Interlaboratory statistics were not calculated since two or more laboratories reported values less than LOQ.

4.3B Table 9. Lead in Bovine Liver (NIST SRM 1577b)a

Laboratory:1234
Rep 1 Anal Portion (g)0.52240.49980.50890.5386
Rep 2 Anal Portion (g)0.51450.50170.51480.5284
ASQL (µg/L)1.62.12.20.93
Rep 1 Result (µg/L)6.702.053.802.36
Rep 2 Result (µg/L)6.901.742.402.82
DF1113
LOQ (mg/kg)0.070.10.10.2
Rep 1 Result (mg/kg)0.2920.1080.1630.115
Rep 2 Result (mg/kg)0.3060.0910.1220.140
Mean Result (mg/kg)0.2990.099 (TR)0.1430.128 (TR)
Reference Value ± SD = 0.129 ± 0.004 mg/kg
RM Value Recovery (%)2327711199
ICP-MS Result (mg/kg)   0.114
Dixon Outlier Test Value: 0.783 Laboratory 1
Dixon Tabulated Value (5%): 0.765 Laboratory 1 Outlier

a Interlaboratory statistics were not calculated since two or more laboratories reported values less than LOQ.

4.3B Table 10. Lead in Spinach (NIST SRM 1570)

Laboratory:1234
Rep 1 Anal Portion (g)0.52950.49170.48920.5270
Rep 2 Anal Portion (g)0.52850.49220.48800.5553
ASQL (µg/L)1.62.12.20.93
Rep 1 Result (µg/L)3.320.911.523.7
Rep 2 Result (µg/L)3.823.010.125.6
DF1114
LOQ (mg/kg)0.070.20.10.2
Rep 1 Result (mg/kg)0.1371.120.6251.19
Rep 2 Result (mg/kg)0.1471.240.5461.22
Mean Result (mg/kg)0.1421.180.5851.21
Interlaboratory Mean Result ± SD = 0.779 ± 0.5130 mg/kg
RSDR = 65.8%; PRSDR = 16.6%; HORRAT = 4.0
Reference Value ± SD = 1.2 ± 0.2 mg/kg
RM Value Recovery (%)129949101
Interlaboratory Mean RM Recovery = 65%
ICP-MS Result (mg/kg)   1.14

back to top

4.3B.3 FOOD RESULTS

Three food materials were included in the study: ketchup, sweet potato (puréed baby food) and yellow mustard. These foods were chosen because of the expected low levels of lead or cadmium or because of the analytical challenge presented by the matrix. Two replicate analytical portions of each food were analyzed. A third fortified analytical portion (FAP) was prepared and analyzed by the laboratory to estimate analyte recovery in the food matrix. 4.3B Tables 11 to 16 list the food results. Additional dilution of the analytical solution was necessary in some cases to overcome matrix interference.

Cadmium

Cadmium results between laboratories were generally in good agreement (4.3B Tables 11 to 13) and RSDR ranged from 6.6% to 13.4%. The only outlier was the result for ketchup from Laboratory 1. The level of cadmium in the foods was well above LOQs. Unfortunately, Laboratory 1 reported an ASQL of 0 so no estimate of LOQ was possible. Assuming Laboratory 1's LOQ was similar to those of the other laboratories then all cadmium values were well above LOQ. Two replicate analytical portions analyzed by each laboratory agreed well with each other. FAP recovery was good with a range of 90% to 105% and a mean of 99%. Results from Laboratory 4 were consistently higher than the others while results from Laboratory 1 were generally lower. The HORRAT value was ≤1.5 for each food and indicates between lab variability was in the expected range. Some analytical solutions required additional dilution to obtain a cadmium concentration within the calibration curve or to diminish matrix interference. The DF required for a given sample varied approximately by a factor of 2. However, the analytical portion mass varied between laboratories for a given sample and a smaller mass would lessen the potential for interference. Laboratory 4's ICP-MS results were in good agreement with GF-AAS results with a maximum relative percent difference (RPD) between the two techniques of 15% for yellow mustard.

4.3B Table 11. Cadmium in Sweet Potato Baby Food

Laboratory:1234
Rep 1 Anal Portion (g)2.51452.42048.55263.1087
Rep 2 Anal Portion (g)2.54222.42268.67003.0389
ASQL (µg/L)-0.0640.1100.030
Rep 1 Result (µg/L)0.4360.3931.340.620
Rep 2 Result (µg/L)0.430.4281.450.609
DF1111
LOQ (µg/kg)00.70.30.3
Rep 1 Result (µg/kg)4.24.14.05.0
Rep 2 Result (µg/kg)4.14.44.05.0
Mean Result (µg/kg)4.24.34.05.0
Interlaboratory Mean Result ± SD = 4.4 ± 0.45 µg/kg
RSDR = 10.2%; PRSDR = 36.2%; HORRAT = 0.3
Fortification (µg/kg)12.420.024.3322
FAP Recovery (%)9310298105
Interlaboratory Mean FAP Recovery = 99%
ICP-MS Result (µg/kg)   3.8

4.3B Table 12. Cadmium in Ketchup

Laboratory:1234
Rep 1 Anal Portion (g)2.70171.01802.02122.2000
Rep 2 Anal Portion (g)2.72681.03872.02082.0337
ASQL (µg/L)-0.0640.110.030
Rep 1 Result (µg/L)0.9540.9511.802.24
Rep 2 Result (µg/L)1.090.9921.822.10
DF1114
LOQ (µg/kg)-212
Rep 1 Result (µg/kg)9.023.422.025.4
Rep 2 Result (µg/kg)10.023.923.025.8
Mean Result (µg/kg)9.5242326
Interlaboratory Mean Result ± SD = 20 ± 7.3 µg/kg
RSDR = 36%; PRSDR = 28.8%; HORRAT = 1.2
Fortification (µg/kg)165096459
FAP Recovery (%)9610490102
Interlaboratory Mean FAP Recovery = 98%
ICP-MS Result (µg/kg)   21
Dixon Outlier Test Value: 0.808 Laboratory 1
Dixon Tabulated Value (5%): 0.765 Laboratory 1 Outlier
Results Excluding Laboratory 1
Interlaboratory Mean Result ± SD = 24 ± 1.6 µg/kg
Interlaboratory Mean FAP Recovery = 99%
RSDR = 6.6%; PRSDR = 28%; HORRAT = 0.2

4.3B Table 13. Cadmium in Yellow Mustard

Laboratory:1234
Rep 1 Anal Portion (g)2.54611.01352.61802.1585
Rep 2 Anal Portion (g)2.55030.98942.57102.3195
ASQL (µg/L)-0.0640.110.030
Rep 1 Result (µg/L)2.771.423.363.23
Rep 2 Result (µg/L)2.761.353.563.49
DF2144
LOQ (µg/kg)0242
Rep 1 Result (µg/kg)27.035.032.037.4
Rep 2 Result (µg/kg)26.934.035.037.6
Mean Result (µg/kg)27353438
Interlaboratory Mean Result ± SD = 33 ± 4.4 µg/kg
RSDR = 13.4%; PRSDR = 26.7%; HORRAT = 0.5
Fortification (µg/kg)405079496
FAP Recovery (%)9697100105
Interlaboratory Mean FAP Recovery = 100%
ICP-MS Result (µg/kg)   32

back to top

Lead

Unfortunately, the lead levels in the foods were near or below the LOQs (4.3B Tables 14 to 16). However, the laboratories correctly obtained results that were consistent with LOQs and the concentration obtained by Laboratory 4 using ICP-MS. Two replicate analytical portions analyzed by each laboratory agreed well with each other considering the low levels. FAP recovery was good and ranged from 89% to 121% with a mean of 98%. Statistics were not calculated because the measured levels in the foods were not above LOQ in 2 or more laboratories. Some analytical solutions required additional dilution because of matrix interference. The DF required for a given sample varied approximately by a factor of 3. The analytical portion variation between laboratories must be taken into consideration when comparing results. The ICP-MS results appear to be considerably lower than the GF-AAS results indicating that a positive bias in the GF-AAS results may be present. This discrepancy supports the qualification of results below LOQ. Analytical limits were estimated using a pristine matrix (method blanks) and probably do not accurately estimate these limits for a sample matrix especially a high-salt containing matrix (e.g., yellow mustard) which is known to interfere with measurements by GF-AAS.

Ketchup (4.3B Table 14) was a challenging sample because of low lead concentration and high salt content. Laboratory 1 found lead to be below LOD. The other laboratories determined lead to be at trace levels. There was inconsistency in the dilution required for interference free analysis (as determined by the FAS recovery). Laboratory 1 (2.7 g analytical portion) found that no additional dilution was necessary whereas Laboratory 4 (2.1 g analytical portion) determined that a DF of 3 was necessary. FAP recovery was very good but for one slightly high result (121%). The results from Laboratories 2-4 were similar even thought they were at trace levels. However, the ICP-MS lead result of <0.006 mg/kg indicates a positive bias may be present by GF-AAS.

4.3B Table 14. Lead in Ketchupa

Laboratory:1234
Rep 1 Anal Portion (g)2.70171.01802.02122.2000
Rep 2 Anal Portion (g)2.72681.03872.02082.0337
ASQL (µg/L)1.62.12.20.93
Rep 1 Anal Result (µg/L)01.300.842.92
Rep 2 Anal Result (µg/L)01.571.741.97
DF1123
LOQ (µg/kg)70505040
Rep 1 Result (µg/kg)0322033
Rep 2 Result (µg/kg)0384324
Mean Result (µg/kg)0 (ND)35 (TR)32 (TR)29 (TR)
Fortification (µg/kg)3250096459
FAP Recovery (%)1019612193
Interlaboratory Mean FAP Recovery = 103%
ICP-MS Result (µg/kg)   < 6

a Interlaboratory statistics were not calculated since two or more laboratories reported values less than LOQ.

Sweet potatoes (4.3B Table 15) were challenging because of the very low lead concentration. Three laboratories determined lead to be below LOQ. Results had a wide variation (41% RPD) as would be expected at such low concentrations. Laboratory 4 (3 g analytical portion) found it necessary to dilute the analytical solution because of matrix interference in contrast to Laboratory 3 that found no interference even with an 8.5 g analytical portion. FAP recovery was excellent. The very low levels by GF-AAS, considering some are trace levels, are in close agreement with the level found by ICP-MS. As with ketchup, the LOQ is probably under estimated by using method blanks.

4.3B Table 15. Lead in Sweet Potato Baby Fooda

Laboratory:1234
Rep 1 Anal Portion (g)2.51452.42048.55263.1087
Rep 2 Anal Portion (g)2.54222.42268.67003.0389
ASQL (µg/L)1.62.12.20.93
Rep 1 Result (µg/L)1.81.552.051.58
Rep 2 Result (µg/L)1.91.252.331
DF1113
LOQ (µg/kg)2020630
Rep 1 Result (µg/kg)17.916612.7
Rep 2 Result (µg/kg)18.71378.2
Mean Result (µg/kg)18 (TR)15 (TR)6.511 (TR)
Fortification (µg/kg)11520024322
FAP Recovery (%)107979697
Interlaboratory Mean FAP Recovery = 99%
ICP-MS Result (µg/kg)   8.5

a Interlaboratory statistics were not calculated since two or more laboratories reported values less than LOQ.

Yellow mustard (4.3B Table 16) presented a similar challenge as ketchup having a low lead and high salt concentration. Laboratory 1 reported that the level of lead was below LOD and the other laboratories reported trace levels. Laboratory 4 found interference severe enough to require a 3-fold dilution. In contrast, Laboratory 3 did not find dilution necessary even though they used the largest analytical portion. Laboratory 2 also did not find it necessary to dilute the analytical solution but they used the smallest (1 g) analytical portion that may have minimized any matrix interference. FAP recovery was very good. The ICP-MS lead result of 5.7 µg/kg indicates a positive bias may be present by GF-AAS.

4.3B Table 16. Lead in Yellow Mustarda

Laboratory:1234
Rep 1 Anal Portion (g)2.54611.01352.61802.1585
Rep 2 Anal Portion (g)2.55030.98942.57102.3195
ASQL (µg/L)1.62.12.20.93
Rep 1 Result (µg/L)01.532.102.59
Rep 2 Result (µg/L)01.162.142.91
DF2113
LOQ (µg/kg)70503040
Rep 1 Result (µg/kg)0382030
Rep 2 Result (µg/kg)0292131
Mean Result (µg/kg)0 (ND)34 (TR)21 (TR)31 (TR)
Fortification (µg/kg)5550079496
FAP Recovery (%)94938992
Interlaboratory Mean FAP Recovery = 92%
ICP-MS Result (µg/kg)   5.7

a Interlaboratory statistics were not calculated since two or more laboratories reported values less than LOQ.

back to top

4.3B.4 OTHER SUPPORTING RESULTS

Other supporting information includes results from the use of the method by FDA laboratories other than the trail participants. One laboratory (Laboratory A) reported lead results for reference materials analyzed in support of regulatory analyses that are summarized below.

Corn Bran, NIST Reference Material 8433

Lead Reference Value = 0.140 mg/kg

Mean RM Recovery = 92.8% (n =12)

Minimum 86.4%, Maximum 102.0%

Oyster Tissue, NIST Standard Reference Material 1566

Lead Reference Value = 0.480 mg/kg

Mean RM Recovery = 92.5% (n =7)

Minimum 80.4%, Maximum 106.3%

Cadmium Reference Value = 3.5 mg/kg

Mean RM Recovery = 92.9% (n =5)

Minimum 85.1%, Maximum 101.1%

The laboratory reporting the above results and another laboratory (Laboratory B) also reported FAP recovery in support of regulatory analyses using the method. These results are summarized below.

Lead FAP Recovery

Laboratory A

Mean = 99.6% (n = 33)
Minimum 82.0%, Maximum 127.0%

Laboratory B

Mean = 99.9% (n=55)
Minimum 80.0%, Maximum 122.0%

Cadmium FAP Recovery

Laboratory B

Mean = 98.2% (n =56)
Minimum 82.0%, Maximum 116.0%

back to top

4.3B.5 CONCLUSION

The results of this interlaboratory method trial show that accurate results can be obtained for cadmium and lead in food using GF-AAS. FAS recovery provided a means to assess the presence of matrix interference and, therefore, a need for further dilution of the analytical solution. However, the need for and degree of additional dilution varied between laboratories. Therefore, analysts should carefully evaluate the possibility of interference not only by the FAS recovery but also by observing the peak profile for any irregularities. In addition, when interference is suspected, the analyst should compare results obtained at different dilutions if analyte level is sufficient (i.e., if dilutions provide measurements above LOQ).

Good estimates of analytical limits are necessary for qualifying results that are below LOQ. Detection and quantification limits determined in a pristine matrix are probably very optimistic for "real" samples especially samples with a challenging matrix. Actual analytical limits in foods are almost certainly greater than analytical limits estimated using method blanks. Although analytical limits probably vary among foods, it is impractical to determine limits for each food matrix. Therefore rules for estimating analytical limits include very conservative rounding and the analyst should consider raising the estimate of the analytical limit when interference is suspected. In some cases, for example, an estimated LOQ of 37 µg/kg should be rounded up to 50 µg/kg.

The results of the interlaboratory trial have presented three issues that should be monitored with use of the method: analytical limits, utility of FAS recovery, and measurement of lead above the ASQL.

  • The trial assumed that the procedure for determining analytical limits was followed and did not require submission of the data used to determine analytical limits. Therefore, critical aspects of determining analytical limits such as using sufficient significant figures and adding analyte to method blanks to obtain a signal near ASDL were not documented. The procedures given for analytical limits should be followed and if they do not provide good estimates then alternative procedures should be investigated.
  • The FAS recovery did not consistently indicate the presence of interference. The trial did not require submission of FAS recovery data but assumed the analyst would perform this task and correctly interpret the results. The utility of the FAS to correctly indicate interferences should be monitored and if it fails to function properly then alternative procedures should be investigated.
  • Unfortunately, the level of lead in the foods and 2 of the reference materials was below the LOQ. Even thought FAP recoveries are very good for lead, trial data on native levels of lead above LOQ are needed to demonstrate interlaboratory reproducibility. A small follow-up to the trail should be undertaken to provide this information.

REFERENCES

  1. Official Methods of Analysis of AOAC INTERNATIONAL (2005) 18th Ed., AOAC INTERNATIONAL, Gaithersburg, MD, USA, Appendix D: Guidelines for Collaborative Study Procedures To Validate Characteristics of a Method of Analysis. AOAC website. 

back to top

Page Last Updated: 11/04/2014
Note: If you need help accessing information in different file formats, see Instructions for Downloading Viewers and Players.