Direct solid sampling ETAAS determination of cadmium in equine muscle† Ernst Lu�cker Institute of Veterinary Food Science, Justus-Liebig University, Frankfurter Strasse 92, D-35392 Giessen, Germany. E-mail: ernst.h.luecker@vetmed.uni-giessen.de Received 10th November 1998, Accepted 22nd December 1998 Direct solid sampling by means of electrothermal atomic absorption spectrometry (ETAAS) was evaluated as a rapid procedure for the determination of Cd in equine muscle tissue with respect to the introduction of a legal limit for Cd in equine muscle tissue in Germany.Analysis of the distribution of Cd within individual muscles of horses showed the variance component ‘heterogeneity’ (relative standard deviation, RSD: 16%)to be clearly dominated by residual analytical variance (RSD 24%). Error of mean value estimation can be reduced by increasing the number of samples—each taken from diVerent sampling sites—from approximately 30% (n=1) to 11% (n=6) . In a comparison between solid sampling ETAAS and a conventional compound procedure as reference (regression analysis, n=394), the Cd content ranged from 0.01 to 0.9 mg g-1 fresh substance.Results were found to be closely correlated (r=0.98) and proved to be analytically and statistically (P>0.10) of no relevant diVerence. Ranges of erroneous classification were calculated, which allow the identification of transgressing samples. For the legal limit of 0.2 mg g-1 fresh substance, erroneous classification (false positive and false negative) ranged from 0.14 to 0.27 mg g-1 fresh substance.Limits for the 1% range of erroneous classification were calculated to range from 0.13 to 0.32 mg g-1 fresh substance. Based on these results, a model strategy is proposed for the application of direct solid sampling analysis as a rapid in-process screening procedure within the scope of meat hygiene regulations. Musculus longissimus dorsi is given in the literature as ranging Introduction from 0.001 to 0.01 mg g-1 fresh substance.13–17 However, for Residues of heavy metals in food are oYcially controlled equine muscle tissue a distinctly higher Cd content is reported almost exclusively by means of electrothermal atomic absorp- as shown in Table 1.tion spectrometry (ETAAS) as part of compound procedures. In addition, the meat hygiene legislation in Germany has These procedures are nationally standardized1 or prescribed2 recently introduced limits for heavy metals for all slaughtered and internationally recommended as reference.3 Compound animals.23 These limits are based on the recommended control procedures include homogenization and transformation of the values for specified tissues of pigs and cattle.24 Thus, the solid sample into solution prior to instrumental analysis.higher Cd content of equine muscle tissue will lead to a However, compound procedures are not only expensive and distinctly increased probability for the transgression of the time consuming but also markedly prone to errors.The main legal limit (0.2 mg g-1 fresh substance). Consequently, the Cd source of analytical error is known to be caused by pre- content of the muscle tissue of all of the horses which are analytical contamination due to the omnipresence of heavy slaughtered in Germany should be determined. However, there metals at relevant concentrations.4 As an alternative, solid is no analytical procedure at hand which allows a rapid and sampling ETAAS was introduced5 and later characterized as inexpensive determination of Cd within the process of a rapid and low-cost analytical procedure in numerous studies.6 slaughtering. Even so, a serious drawback of solid sampling ETAAS was and still is seen in the very low sample masses (microgram to milligram range) which are needed for analysis.Thus, solid Table 1 Cd (mg g-1 fresh substance) in equine muscle tissue as reported sampling ETAAS has been predominantly applied to homoin the literature genized samples.Some workers, however, have studied the matrix-dependent distribution of analytes by means of solid Muscle n x: x� Min. Max. Ref. sampling and the eVect on analytical precision in fresh and Ba 68 0.13 —b 0 1.5 18 untreated samples. They reported a relatively low degree of Ba 40 0.15 —b 0.02 0.69 18 heterogeneity for elements such as Cd, Zn, Pb and Hg in some —b 142 0.17 0.13 0.01 1.7 19 animal tissues—such as porcine and bovine liver7–10 and renal Diaphragm 109 0.13 0.12 0.01 0.47 20 cortices11 as well as avian kidney.12 Thus, they have shown M.rectus abdominis 100 0.06 0.0 0.007 0.26 20 that a true ‘direct’ solid sampling is not only possible but also Diaphragm 17 —b 0.11c 0.01c —b 21 is totally error-free with regard to sample preparation. M. rectus abdominis 17 —b 0.05c 0.005c —b 21 M. longissimus 17 —b 0.05c 0.005c —b 21 Within the scope of oYcial heavy metal control in food, Musclesd 17 0.11 0.06 0.005c 1.5 21 muscle tissue is usually of minor concern, as its heavy metal M.longissimus 209 0.08 0.05 0.03 0.62 22 burden is very low. The Cd content of porcine and bovine M. semitendinosus 209 0.07 0.05 0.03 0.54 22 aB: Muscle samples from bacteriological examination. bNo data given. †Presented at the 8th Solid Sampling Spectrometry Colloquium, cValues taken from graphs. d12 diVerent muscles. Budapest, Hungary, September 1–4, 1998. J. Anal.At. Spectrom., 1999, 14, 583–587 583Therefore, the present study was performed in order to Analytical quality assurance evaluate the applicability of direct solid sampling as a rapid The certified and in-house standardized reference materials in-process procedure for the determination of Cd in equine used were: bovine liver (NIST SRM 1577a, LIS-G01, LISG10), muscle tissue within the scope of meat hygiene regulations. porcine kidney (BCR CRM 186), and bovine muscle (BCR Furthermore, the determination of Cd in equine muscle tissue CRM 184).presents a suitable model for the general characterization of Compound as well as solid sampling procedures as used in representativity of both direct solid sampling and conventional this study were evaluated in several national and international compound procedures. For this purpose, (i) information about round-robin exercises.26,29–31 the distribution of Cd between and within individual muscles is obtained by applying a multifactorial hierarchic analysis of Analytical models variance, (ii) a suitable sampling strategy is developed, (iii) the correctness of direct solid sampling ETAAS results is Analysis of Cd distribution in equine muscle tissue was evaluated with regard to the oYcially prescribed reference performed using three muscles (MLC: M.longus capitis, MG: method, (iv) limits of erroneous classification are established M. gastrocnemius and MD: M. diaphragmaticus) taken from and (v) the performance of direct solid sampling analysis is two horses.evaluated with respect to analytical time and cost. Within each muscle, six sampling sites were chosen in a totally randomized way. From each sampling site six microsamples were taken from an area—which was circumscribed Experimental as closely as possible—and analyzed. Additionally, the material taken from a sampling site was homogenized during Material microsampling. Thus, multifactorial analysis of variance of Muscle tissue samples (459) were taken during regular the mixed model (ANOVA) gave estimations of betweenslaughtering of horses.The main sampling sites were: M. sampling site variance (heterogeneity, sh) and within-sampling longissimus thoracis (n=154), M. semitendinosus (n=153) and site variance (residual analytical variance, sa). Prior to M. diaphragmaticus (n=99). Additionally, muscle samples ANOVA, data were logarithmically transformed in order to (n=53) of the distal extremities taken for bacteriological achieve homogeneity of variance.The resulting dispersion investigation were used. If necessary, samples were stored in factor (df ) around the geometric mean is dimensionless and close-fitting polyethylene foil at -21 °C. can be expressed as relative standard deviation (e.g. df= x·1.23±1lusmn;23%)32 in a simplified albeit more illustrative manner. Direct solid sampling ETAAS Comparison of methods consisted of the analysis of 394 The analyses were carried out using solid sampling samples: first, by means of solid sampling ETAAS and second, spectrometers (SM1 and SM20) with Zeeman-eVect back- by means of the compound procedure after sample homogenizground correction (Gruen AMS, Ehringshausen, Germany).ation and decomposition. Note that virtually the same sample In addition, a conventional high energy deuterium background was thus analyzed following solid sampling analysis, as the corrected atomic absorption spectrometer (PU9100x, TJA mass of the six microsamples taken was about 2–4 orders of Unicam, OVenbach, Germany) was used with a modified magnitude lower than that of the field sample.graphite system for solid sampling.25 Calibration was eVected Results were analyzed by means of regression analysis by means of working standard solutions (5, 10, 20 and ( least-squares method), t-test and Wilcoxon test. Logarithmic 30 mg l-1) derived from a 1000 mg l-1 stock standard solution transformation of data was used in order to normalize and ( Titrisol, Merck, Darmstadt, Germany) and checked by at standardize the standard deviation of the y-values for given least one certified and one in-house standardized reference x-values.The resulting standard deviation sy.x facilitated the material. The reference material was also used to optimize the estimation of limits for erroneous classification with respect basic optical and thermal conditions (Table 2) prior to the to legal limits according to Lu� cker.33,34 Calculations were analyses of field samples.Field samples of equine muscle tissue performed with the help of the following software packages: (5–100 g) were analyzed without any pre-treatment. During MS-Excel 97 (Microsoft, Redmond, WA, USA), Limit33 and microsampling, these field samples were kept at +2 °C and BMDP.35 99% relative humidity. The actual analytical sample (‘microsample’) was plugged from the field sample with the help of Results two Inox micro-tweezers (Kretschmer, Giessen, Germany) and rapidly transferred to the tared platform on a M500P or Distribution analyses 4503MP6 microbalance (Sartorius, Go� ttingen, Germany).The distribution of Cd within and between sampling sites of Typical sample masses were 0.1–5 mg. The instrumental analythe analyzed equine muscles is depicted in Fig. 2. The Cd sis was started after the platform had been introduced into content ranges around 0.02 mg g-1 in muscles of horse A and the graphite cuvette of the spectrometer (Fig. 1). Instrumental around 0.2 mg g-1 fresh substance in horse B. The mean parameters are listed in Table 2. Quantification was achieved relative standard deviations (RSDs) of the sampling sites are using peak height measurement with the SM1 spectrometer 22 and 15%, respectively. Reference material and working and using peak area integration with the SM20 and PU9100x standard solutions were analyzed alternately after each analysis spectrometers.Further details are given elsewhere.9–12,26–28 of three field samples. The respective mean RSD of the reference material ranges from 3.2 to 13.2% (4–8 micros- Reference procedure amples) and that of the working standard solutions from 2.4 to 8.9% (n=4, manual pipetting). Thus, in direct solid sam- Field samples were homogenized by means of a noncontaminating kitchen mixer (Moulinette, Moulinex, Solingen, pling analysis the observed variance of non-homogenized equine muscle tissue is somewhat higher than in homogenized Germany).Then, 5 g of the homogenized material were decomposed with purified HNO3 in an open system (Tecator, reference material and distinctly higher than in liquid standards. Frankfurt, Germany).26 Instrumental analysis was performed using a PU9100x atomic absorption spectrometer with high Results of the four-factorial hierarchic analysis of variance of the mixed model ( logarithmically transformed data, 2 energy deuterium background correction (D2-ETAAS). 584 J.Anal. At. Spectrom., 1999, 14, 583–587Table 2 Instrumental parameters for the determination of Cd in fresh and untreated equine muscles by means of direct solid sampling ETAAS SM1 SM20 PU9100x Optical parameters: Emission source Electrodeless Electrodeless Hollow cathode low frequency low frequency Resonance line l/nm 228.8 228.8 228.8 Bandpass/mm 0.25 0.25 0.20 Background correction Direct Zeeman Direct Zeeman High energy deuterium Graphite system: Cuvette Pyrolytically coated Uncoated Pyrolytically coated Carrier (platform) Boat-like Boat-like Drawer-like pyrographite graphite pyrographite Chemical modifier NH4H2PO4 — NH4 H2PO4 Cut-oV value (%) 25 25 — Argon/l min-1 0.4 0.4 2.8 Program: Temperature/time: °C s °C s °C °C/s s Phase 1 (Dry 1) 100 15–20 200 25 120 5 30 Phase 2 (Dry 1) — — — — 300 10 40 Phase 3 (Ash) 200 15–100 600 15–100 500 50 50 Phase 4 (Atomize) 2200 3–7 2400 3–7 1550 >2000 5 Phase 4 (Clean)a 3000 0–5 3000 0–5 2400 >2000 5 Phase 6 (Cool ) 5 0–5 5 0–5 5 — 25 aOptional for SM1 and SM20.muscles of each horse (P>0.1). Within all muscles, no positive eVects are obtained for the ‘sampling site’ (P>0.6), or for the possible interactions. In this model of direct solid sampling a total error of mean value estimation of about 30% (dft=1.30±1) is obtained for the analysis of one microsample. The estimate of the variance components shows that residual analytical variance (dfa= 1.24±1) clearly dominates heterogeneity induced variance (dfh=1.16±1).Comparison of methods The Cd contents of the 394 equine muscle samples range from Fig. 1 Flow scheme of analytical steps in direct solid sampling ETAAS 0.01 to 0.9 mg g-1 fresh substance (Fig. 3). The control value analysis of fresh and untreated equine muscle tissue. of 0.1 mg g-1 fresh substance (as recommended by the German Ministry of Health) is exceeded by 133 samples (33.8%) as analyzed by direct solid sampling ETAAS, whereas only 111 samples (28.2%) exceed the control value when applying the compound procedure. For the legal limit of 0.2 mg g-1 fresh substance, the respective figures are 41 (10.4%) for solid Fig. 2 Distribution of Cd (mg g-1 fresh substance, FS) in muscles of two horses (A and B): total mean values of muscles (dotted lines), mean values and standard deviation of sampling sites within the muscles (sampling scheme: 2×3×6×6, MLC: M.longus colli, MG: M. gastrocnemius, MD: M. diaphragmaticus). animals ×3 muscles ×6 sampling sites ×6 microsamples, n= 216) indicate an influence on total variance for the error sources ‘animals’ (P<0.001) and ‘muscles’ (P<0.05). When Fig. 3 Histogram of the analytical results of direct solid sampling regarding the muscles as a whole, the increased (P<0.01) Cd ETAAS and compound ETAAS with sample decomposition and content of the diaphragmatic muscle (MD) becomes apparent frequency of transgression of Cd contents in equine muscles with respect to the German legal limit of 0.2 mg g-1 fresh substance (FS).(Fig. 2), whereas no positive eVects are found for the other J. Anal. At. Spectrom., 1999, 14, 583–587 585the significant increase in Cd in the diaphragmatic muscle— was shown.20,21 This demonstrates that representativity36 of sampling is not easily achieved, when only one muscle is sampled. Homogenization of a sample usually leads to decreased analytical variances.However, the observed variance does not correspond with analytical imprecision as (i) information about native variance is lost, (ii) no additional information is obtained regarding the Cd content of the total muscle mass of the respective animal, (iii) the probability of secondary contamination is increased and (iv) information about such a contamination may be lost after homogenization. All of this may lead to a false kind of security as regards analytical quality.Still, a wealth of information is needed rder to correctly assess representativity when estimating the Cd burden of all muscles of an individual animal on the basis of only one sample. The carcass of a slaughtered animal is Fig. 4 Comparison of Cd contents of 394 equine muscle sample as composed of more than 200 individual muscles. In equines analyzed by direct solid sampling (x) and after homogenization and digestion of the respective samples ( y).The probability function and (and bovines) the mass of these muscles may exceed 400 kg. 5% range of erroneous classification is given for the German legal The analysis of one muscle sample being 3–4 orders of limit of 0.2 mg g-1 fresh substance (FS). magnitude lower in mass, however, is a standard procedure in oYcial meat hygiene residue control. Representativity has been tacitly taken as given. Interestingly, the same problem is sampling and 36 (9.1%) for the compound procedure.Comparison of the corresponding results obtained with direct inherent in solid sampling ETAAS where microsamples are usually several orders of magnitude lower than the total sample solid sampling ETAAS (x) and after sample homogenization and decomposition ( y) yields the equation of y=0.97x for the mass. Thus, much might still be learned from solid sampling distribution analyses. non-normalized data (r=0.983) and y=0.87x0.96 for the logarithmically transformed data (r=0.981).In the latter case, the Another result of the present distribution analysis is the low degree of heterogeneity within individual muscles. Previous standard deviation of the y-values for given x-values is about 20% (sy.x=1.21). On average, the Cd content as obtained by studies have demonstrated corresponding results for the distribution of Pb, Cd and Hg in fresh and untreated porcine,7 direct solid sampling ETAAS is expected to be about 4% above the corresponding Cd content as obtained by use of the bovine8 and equine liver,10 bovine renal cortices11 and avian kidneys.12 compound procedure.The noted diVerences are, however, not significant for both normalized and non-normalized data The total error of mean value estimation can be reduced by increasing the number of microsamples analyzed.37 As shown (P>0.01, t-test, Wilcoxon test). Erroneous classification of samples as analyzed by direct in Table 3, the maximum reduction of total variance is achieved when each microsample is taken from a diVerent sampling solid sampling can be characterized by means of a probability function using the legal limit, the standard deviation of the y- site.Corresponding results were found in a variety of solid sampling ETAAS distribution studies (e.g. Hg in avian kidney, values for given x-values (sy.x=1.21) and the t-distribution. An example is given in Fig. 4 for a significance level of 95%. Cd in equine liver) and in a study of the distribution of Fe in rabbit muscle tissue using a compound procedure.38 When For the legal limit of 0.2 mg g-1 fresh substance, erroneous classification (false positive and false negative) ranges from using this sampling strategy in the determination of Cd in equine muscles by means of solid sampling ETAAS, the 0.14 to 0.27 mg g-1 fresh substance.Limits for the 1% range of erroneous classification are calculated to range from 0.13 reduction of heterogeneity induced variance is impressive (Table 3), even though heterogeneity was found to be only a to 0.32 mg g-1 fresh substance.When applying the solid sampling spectrometer with direct minor component of total analytical variance. In solid sampling analyses, results (mean value, standard Zeeman-eVect background correction, the analysis of six microsamples takes about 10 min. Thus, in one hour 40–60 results deviation) are acquired continuously with the analysis of each microsample.Taking advantage of this analytical in-process can be obtained. This is equivalent to 4–7 samples, including analysis of reference material and working standards. The information, the number of microsamples to be analyzed can be chosen according to the observed mean value: After analyz- average analytical cost with respect to argon, graphite and reference material in the analysis of one sample (six microsamples) is calculated to be approximately $0.4–0.7. Analytical Table 3 Error of mean value estimation in the determination of Cd expenditure is found to be increased when using the convenin fresh and untreated equine muscles by means of solid sampling tional atomic absorption spectrometer with high energy deu- ETAAS with respect to number of microsamples analyzed and terium background correction and modified graphite system.sampling strategy applied Respective factors for average analytical time and cost are Error of mean Number of value estimation 1.5–2.0.This is directly correlated with the need for a prolonged furnace program (Table 2). However, with both syssampling sites microsamples per tems as used in this study, analytical time and cost range far per field sample sampling site df a RSD (%)b below the limits as given by the circumstances of horse slaughtering. 1 1 1.299 30 1 6 1.188 19 3 1 1.163 16 Discussion 2 3 1.147 15 6 1 1.113 11 In this study the analysis of the distribution of Cd in equine 9 1 1.091 9 muscle tissue indicates analytically and legally relevant diVer- 20 1 1.060 6 ences between muscles within individual animals.This finding aDistribution factor around geometric mean. bApproximation for corresponds with previous studies where heterogeneity of Cd relative standard deviation for arithmetic mean. between diVerent muscles of individual animals—especially 586 J. Anal. At. Spectrom., 1999, 14, 583–5873 Commission of the European Communities, Commission ing, e.g.three microsamples, we can observe a mean Cd Decision 90/515/EEC, ABl, EC 1990, L 286, 33. content which does not exceed the lower 1% limit of erroneous 4 P. Tscho� pel, in Hazardous Elements in the Environment. classification (0.13 mg g-1), the field sample can then be classi- Techniques and Instrumentation in Analytical Chemistry, fied as non-suspect (with respect to the 5% limit) and analysis ed. M. Stoeppler, Elsevier, Amsterdam, vol. 12, 1992, pp. 73–95.is stopped. Samples exceeding the range of the lower 1% limit 5 U. Kurfu� rst, Nachr. Chem. Tech. Lab., 1981, 29, 854. 6 U. Kurfu� rst, in Solid Sample Analysis. Direct and Slurry Analysis have to be analyzed further. After increasing the number of Using GF-AAS and ETV-ICP, ed. U. Kurfu� rst, Springer-Verlag, microsamples analyzed to, e.g. n=6, samples can be finally Berlin, Heidelberg, 1998, pp. 1–127. classified as suspect or non-suspect with respect to the 5% 7 B.Klu� ßendorf, A.Rosopulo and W. Kreuzer, Fresenius’ Z. Anal. limit of erroneous classification. The number of microsamples Chem., 1985, 322, 721. analysed with regard to the 1 and 5% limits should be chosen 8 A. Besse, PhD Thesis, University of Gießen, 1987. according to the observed variances and with respect to 9 E.Lu� cker, C. Gerbig and W. Kreuzer, Fresenius’ Z. Anal. Chem., 1993, 346, 1062. variances obtained for working standard solutions, reference 10 E. Lu�cker, J. Meuthen and W.Kreuzer, Fresenius’ Z. Anal. Chem., material and previously analyzed samples. Maximum tolerable 1993, 346, 1068. observed variances were recorded in the literature.34 11 E. Lu� cker, A. Rosopulo and W. Kreuzer, Fresenius’ Z. Anal. Samples suspected of exceeding the legal limit can be verified Chem., 1987, 328, 370. either by using a compound reference procedure or by further 12 E. Lu�cker, Fresenius’ Z. Anal. Chem., 1997, 358, 848. increasing the number of microsamples analyzed.This latter 13 H. Hecht, Ber. Landwirtschaftswiss. 1978, 55, 828. 14 H. Hecht, Fleischwirtschaft, 1979, 59, 1621. step, however, depends on the legal acceptance of solid 15 L. Jorhem, S. Slorach, B. Sandstro�m and B. Ohlin, Food Addit. sampling ETAAS. Contam., 1991, 8, 201. As shown by the comparison of methods, results of direct 16 A. Niemi, E.-R. Vena�lainen, T. Hirvi and E. Karppanen, Z. solid sampling do not deviate from the respective results as Lebensm. Unters.Forsch., 1991, 192, 4nventional compound procedure applied in 17 J. Falandysz, Sci. Total Environ., 1993, 136, 193. this study. The slight increase in Cd content as observed in 18 J. Holm, Fleischwirtschaft, 1979, 59, 737. 19 A. Salmi and J. Hirn, Fleischwirtschaft, 1981, 61, 1199. solid sampling can be related to a loss of mass due to 20 H. Hecht, Fleischwirtschaft, 1984, 64, 1113. evaporation during preparation and mass determination of 21 P. Geppert and B.Brunner, in Tagung des Arbeitsgebietes the microsamples.12,27,28 Accordingly, the number of samples Lebensmittelhygiene, ed. German Veterinary Society, Gießen, exceeding the legal limit was slightly increased. With respect 1995, vol. 36, pt. 2, pp. 243–250. to a modified sampling strategy, however, the frequency of 22 F. Weyermann and E. Lu� cker, Fleischwirtschaft, 1998, 78, 251. erroneous classification can be expected to be extremely low. 23 Ministry of Health, Germany, Fleischhygiene-Verordnung, Bundesgesetzblatt, 1996, pt. 1, p. 1678. With respect to the conditions of horse-slaughtering and the 24 Zentrale Erfassungs- und Bewertungsstelle fu� r Umweltche- given legal situation, both analytical time and cost of direct mikalien (ZEBS), Bundesgesundheitsblatt, 1996, 39, 193. solid sampling ETAAS are suYciently low to characterize it 25 A. Besse, A. Rosopulo, C. Busche and G. Ku� llmer, Labor Praxis, as a rapid procedure. In addition, a further 90% reduction of 1986, 1/2, 64.analytical time and cost can be expected when applying the 26 A. Rosopulo, Fresenius’ Z. Anal. Chem., 1985, 322, 669. sampling strategy outlined above. 27 A. Rosopulo and W. Kreuzer, in Fortschritte in der atomspektrometrischen Spurenanalytik, ed. B.Welz, Verlag Chemie,Weinheim, This study shows that direct solid sampling ETAAS can be vol. 2, 1986, pp. 455–463. applied as a rapid screening procedure in oYcial meat inspec- 28 E. Lu�cker and O. Schuierer, Spectrochim.Acta, Part B, 1996, tion ad hoc. The present situation in meat hygiene is rather 51, 201. exotic, initiated by certain—probably transient—legal pre- 29 E. Lu� cker, A. Rosopulo and W. Kreuzer, Fresenius’ Z. Anal. scriptions. However, direct solid sampling is the first analytical Chem., 1991, 340, 234. procedure suitable for in-process analysis within the scope of 30 E. Lu�cker, H. Ko� nig, W. Gabriel and A. Rosopulo, Fresenius’ Z. Anal. Chem., 1992. 342, 941. meat hygiene control of heavy metals. It will certainly serve 31 R. F. M. Herber and K.-H. Grobecker, Fresenius’ Z. Anal. Chem., to further our understanding of analytical imprecision and 1995, 351, 577. could give an impetus towards the development of new rapid 32 L. Sachs, Applied Statistics, Springer-Verlag, New York, 1984, procedures. Furthermore, there might be other interesting pp. 105–110. fields in meat hygiene—such as non-oYcial meat quality 33 E. Lu� cker, LIMIT—Program for the calculation of erroneous programs—for applying direct solid sampling ETAAS. classification, University of Gießen, 1997. 34 E. Lu� cker, in Tagung des Arbeitsgebietes Lebensmittelhygiene, ed. German Veterinary Society, Gießen, 1995, vol. 37, pp. 129–134. References 35 W. J. Dixon, BMDP Statistical Software Manual, University of California Press, Berkeley, 1992. 1 Amtliche Sammlung von Untersuchungsverfahren nach § 35 des 36 R. Klockenka�mpfer, Fresenius’ Z. Anal. Chem., 1977, 285, 345. Lebensmittel- und Bedarfsgegensta�ndegesetzes, Federal Institute of 37 E. Lu�cker, Appl. Spectrosc., 1997, 51, 1031. Consumer Health Protection and Veterinary Medicine, Beuth 38 E. Lu� cker, K. Failing, K. Lange, G. Walker and M. Bu� lte, Food Verlag, Berlin, 1995, p. 31. Sci. Technol., 1998, 31, 150. 2 Allgemeine Verwaltungsvorschrift zur Durchfu�hrung der amtlichen Untersuchungen nach dem Fleischhygienegesetz vom 11.12.1986 (VwVFlHG), Ministry of Health, Germany, BAnz Nr. 238 a. Paper 8/08804G J. Anal. At. Spectrom