JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 199 1 VOL. 6 559 Accuracy of Multi-element Analysis of Human Tissue Obtained at Autopsy Using Inductively Coupled Plasma -Mass Spectrometry* Thomas D. B. Lyon and Gordon S. Fell Trace Element Unit Institute of Biochemistry Royal Infirmary Glasgow G4 OSF UK Keith McKay and Roger D. Scott Scottish Universities Research and Reactor Centre National Engineering Laboratory East Kilbride UK The accuracy of inductively coupled plasma mass spectrometry (ICP-MS) for the multi-element analysis of human tissue has been assessed by analysis of National Institute of Standards and Technology (NIST) Standard Reference Material (SRM) 1577a Bovine Liver International Atomic Energy (IAEA) Certified Reference Material (CRM) H 8 Kidney and IAEA CRM H 4 Animal Muscle for All Mg Mn Fe Co Zn Cu Mo Rb Sr and Cd.Twenty three out of 24 determinations agreed with certified values. An intercomparison was also made between ICP-MS and the well established technique of atomic absorption spectrometry (AAS) for Al Mg Mn Fe Cu Zn and Cd in human tissue samples which covered a much wider range of analyte concentrations than the reference materials. Drift in the ICP-MS instrument was studied for 3 h and all but the lighter elements Al and Mg were found to be acceptably controlled by an In internal standard. The equivalence of the ICP-MS and AAS data was compared using the usual statistics of correlation and regression. In addition comparisons were made using the statistical analysis advocated by Bland and Altman.In spite of the problems found ICP-MS is capable of accurate and rapid multi-element analysis of human autopsy tissue. Keywords Inductively coupled plasma mass spectrometry; multi-element analysis; accuracy; autopsy tissue; reference materials Several have appeared recently which discuss the determination of trace elements in biological samples by inductively coupled plasma mass spectrometry (ICP-MS). The purpose of this paper was to investigate the suitability of ICP-MS as a routine method for multi-element analysis of human autopsy material. The accuracy of this technique was examined by the analysis of National Institute of Standards and Technology (NIST) Standard Reference Material (SRM) 1577a Bovine Liver International Atomic Energy Agency (IAEA) Certified Reference Material (CRM) H 8 Kidney and IAEA CRM H 4 Animal Muscle and by an intercomparison with standard methods of atomic absorption spectrometry (AAS) e.g.flame AAS (FAAS) and electrothermal AAS (ETAAS) using a selection of tissue samples takenat autopsy covering the wide concentration range found in both healthy and diseased tissues. Experimental Instrumentation The ICP-MS results were obtained using a PlasmaQuad (V.G. Elemental Winsford Cheshire UK). Details of the instrumental operating conditions used are given in Table 1. The applied dead time was set at the unusually low value of 3 ns. Choice of applied dead time will be discussed in a future publication. The flame AAS results for Mg (285.2 nm) Zn (213.9 nm) Fe (248.3 nm) and Cd (228.8 nm) were obtained using a Perkin-Elmer Model 3030 spectrometer while a Perkin-Elmer Model 2280 spectrometer with an HGA-500 graphite furnace and an AS-1 autosampler was used for the Al Mn and Cu study.Duplicate injections of 20 ,ul were used in the furnace work. The temperature programmes were as follows A1 (309.3 nm) ambientkamp 10 s to 1400 "Champ 2 s to atomize at 2700 "C hold 10 s; Mn (279.5 nm) ambienthamp 1 s to 140 "C and 7 s/ramp 2 s to 1100 "C hold 30 s/ramp 2 s to 2700 "C hold 5 s; and Cu (325 nm) ambienthamp 1 s to 140 "C hold 7 s/ramp 7 s to 900 "C hold 23 s/ramp 2 s to 2700 "C hold 10 s. * Presented at the Second International Conference on Plasma Source Mass Spectrometry University of Durham UK September 24th-28th 1990. Table 1 Plasma operating details Plasma- R.f.power Forward Reflected Gas controls Auxiliary Coolant Nebulizer Nebulizer Spray chamber Ion sampling- Sampling cone Skimmer cone Sampling distance Lens settings- 1.37 kW t 1 0 w 0.6 1 min-I 13 1 min-I 0.7 1 min-' Babington V-groove type high solids Scott-type double bypass water cooled pumped at 0.7 ml min-' (ambient) Nickel sampler (Nicone) with 1.0 mm Nickel (001 Type) with 0.75 mm orifice 10 mm from load coil orifice L1 -36V L2 -52V L3 -1.8 V L4 -62V Extraction lens -200 V Optimization- The lenses were adjusted to maximize the llsIn signal Vacuum- Expansion stage Intermediate Analyser Data acquisition- Peak jumping Sweevs 2.4 mbar < l x mbar 3 x 10+-4 x 1 0-6 mbar 10 Channels per peak 3 Dwell time 50 000 pus Sample Preparation Five replicate samples of the reference materials and single freeze-dried autopsy samples of heart@) muscle(m) liver(1)560 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 199 1 VOL.6 and kidney(k) were weighed (0.500 g) into perfluoroalkoxy resin ‘bombs’ (60 ml Savillex Minnesota USA) and 1 ml of HN03 [Aristar grade Merck (formerly BDH)] was added together with 1 ml of In as internal standard (1.25 mg 1-I). The In spike was added at this point ratherthan immediately before mass spectrometric analysis in order to compensate for possible aerosol losses on opening the bomb or for incomplete transfer of the digest. The bombs were placed in a domestic microwave oven and heated for 2 min at full power (600 W). After cooling the bombs were opened and the contents washed into an acid-washed plasticvial and made up to volume (25 ml). This digestion procedure does not oxidize all the fat contained in some tissue samples and any undigested globules of fat were removed with a disposable plastic pipette.The incomplete destruction of any fat haslittle or no effect on the accuracy of the results. This can be seen from the good agreement of the values determined for the reference materials. Earlier attempts to dissolve the fat by including hydrogen peroxide in the digestion were successful but were abandoned after the failure of several bombs during the digestion process. Blanks carried through the digestion procedure gave a negligible response and no correction was made to the results. Inductively coupled plasma mass spectrometry The digests were nebulized directly into the ICP mass spectrometer without further dilution.Quantification of all reference materials and samples was made using external multi-element standard solutions made up in 4% HN03 and containing In as the internal standard. The top standard contained 5000 pg 1-1 of Mg Fe Zn and Cd 1000 pg 1-’ of Cu and 100 pg 1-l of Al Mn Co Mo Rb and Sr. A low standard was prepared by making a 10-fold dilution of the top standard. The isotopes used for the determinations were selected on the basis of freedom from polyatomic isobaric species. Interference from polyatomic species in the biological matrix has been discussed elsewhere.’ The isotopes chosen are shown in Table 2. Atomic absorption spectrometry The digests were manually diluted with water as required for each element of interest and analysed sequentially against simple single element standards.All digests for the magnesium assay were diluted 50-fold with 0.1% m/v lanthanum chloride solution and all digests for the determi- nation of iron containing more than 8 pg 1-* were diluted 1 0-fold with water. Statistical Interpretation The results of the intercomparisons are plotted in Figs. 2-8 (coincident data points are represented by a number). The coefficient of correlation r and the equation oftheregression line aregiven. Uncertainties are quotedat the 95% probability level. Agreement between two methods is usually considered satisfactory ifrand the slope ofthe regression equation(rn) are both close to unity and the intercept of the regression equation is zero.However it has been suggested that neither r nor the regression equation measure agreement between methods. Bland and Altman5 have described a statistical analysis which is specifically designed to assess agreement between measure- ment techniques. In this test the difference (A) between the two methods AAS and ICP-MS for each sample is plotted against its mean value. Ideally the data should lie within a normal distribution about the horizontal line through the zero difference. The 95% normal range for the differences defines the ‘limit of agreement’ while the mean difference can be thought of as the ‘bias’. Results and Discussion Reference Materials The results for the three reference materials are shown in Table 2 and are satisfactory for most practical purposes.~ Table 2 Results obtained for reference materials using ICP-MS. Errors are given at the 95% confidence interval (n=5) IAEA H 4 IAEA H 8 NIST SRM 1577a Animal Musclelpg g-I Kidneylpg g-’ Bovine Liver/pg g-l Element Isotope Found Certified Found Certified Found Certified 25 1081* 1050k59 782 f 112 818 f 7 5 589 f 43 600+ 15 A1 27 0.45k0.7 -7 0.98 * 0.6 - 0.95 f 0.17 - Mg Mn 55 0.52 + 0.04 0.52 20.04 6.1 k 0.5 5.7 f 0.3 10k0.6 9.9 k 0.8 Fe 57 46k8 4 9 f 2 297 f 20 265 k 15 187+ 14 194220 co 59 - (0.004) 0.1 3 f 0.0 1 (0.13) 0.24 f 0.02 c u 65 4.1 k0.5 4.0k0.3 33.5 f 1.6 31.3 k 1.8 149+ 15 158k7 Zn 66 90k6 86 k 3.5 205 f 23 193 2 6 1 2 4 t 14 19f1.5 21.4k3 22.2 f 0.8 11.3f0.6 12.5fO.l Rb 85 21.8f2 Sr 88 0.06f0.02 - Mo 98 0.041 f0.008 (0.05) 2.4 f 0.1 2.2 k 0.3 3.8f0.12 3.5 f 0.5 Cd 111 Cd 106 - - l8Ok 13 189f5 - * Single determination. j- Not determined or not available.0.21 f 0.05 123k8 1.2f0.06 (1.1) 0.135 f 0.005 0.138 f 0.003 0.40 f 0.03 0.44 f 0.06 - - - - - Table 3 Change in concentration of the top standard measured over 150 min period. The results after In compensation are expressed as % change of the actual concentration Elapsed In/ 1 O3 counts time/min S-’ Mg A1 Mn Fe Co Cu Zn Rb Sr Mo Cd Ba 0 279 0 0 0 0 0 0 0 0 0 0 0 0 36 219 + 5.4 + 1.9 + 1.2 -1.4 -5.3 -6.9 0.0 -1.6 0.0 -4.3 -0.8 -0.4 66 178 +14.6 + 8.2 + 6.5 +3.8 -1.7 -4.3 +5.2 +1.6 +3.8 -3.2 -0.4 0.0 96 170 +12.4 + 7.5 + 6.8 +3.4 -2.1 -4.6 +3.5 +2.6 +5.6 -1.1 +0.8 +0.8 126 145 +10.5 + 4.7 + 3.8 0.0 -5.5 -8.2 +2.2 +0.9 +4.4 -2.6 -1.4 +0.3 150 130 +21 +15 +10 +6.1 -0.2 -3.3 +6.5 +4.9 +7.9 +0.3 +0.3 +0.8JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 199 1 VOL.6 56 1 0) 800 Is) s - a 600 s ti 0 400 I Drift and Internal Standardization The internal standard used was In. This is added primarily to compensate for changes in the sensitivity of the ICP-MS system. The changes in sensitivity can be time dependent i.e. drift or can be a function of the sample i.e. a matrix effect. The efficacy of the single In internal standard to control drift and matrix effects was tested over a 3 h period by presenting a standard as an unknown after every fifth sample. The results are shown in Table 3. The In- compensated results for the heavier elements Cd and Ba are excellent and the elements from Mn to Mo (mass 55-98) are adequately compensated for with mean drifts between -5.5 and +6%.The signals from the lighter elements A1 and Mg increase during the course of the run to such an extent that the data may be greater than 10% in error and regular recalibration is indicated for these analytes. The In count rate decayed by about 50% during the course of the run. This decay in sensitivity is usually 0 - 0.0 2 0 0 2 02 0 . O H - 0 0 0 0 - 0 . 0 0. 2. - 00 I 1 I attributed to the progressive clogging of the sample orifice and to the defocusing of the ion beam as the run progresses. The decay is doubly exponential and is shown in Fig. 1. This figure also shows that the In internal standard in the tissue digests behaves in the same way as the In in the standard and implies that there is no matrix effect for In in tissue digests.Igarashi et a1.6 have investigated the properties of In and 0 . I 360 480 600 720 840 960 FAAS result/pg g-’ ( b ) 122 0 0. 0 0 0 0 0 0 0.0 0 0 . 0 0 0 0. 0 . 0 0 2 80 I Is) D O a‘ -80 = I -1 14 0 50 100 150 Ti me el a psed/mi n I 1 I 0 I 360 480 600 720 840 960 Mean resultlpg g-’ Fig. 1 Decay of In signal with elapsed time. Closed squares are the top standard repeated at intervals and open squares are samples Fig. 3 (a) Magnesium regression plot (38 samples 17 1; 12 k; 7 m; and 2 h). r=0.91; m=0.92 (CI=O.78-1.05); and (b) Bland and Altman plot for Mg (a) 0 7 120 Is) Is) p 80 e ii 0 ’ 40 0 10.5 0 0 =i 7.0 v) 2 a u v) 4 3.5 0 0. 0 0 . .. -H 1 I I 1 I I 0 25 50 75 100 125 ETAAS resuIt/pg g-’ 202 63 I I 1 0 2.0 4.0 6.0 8.0 10.0 ETAAS result/pg g-’ 0.85 2 2 0 .2 2 2. 0 0 0 2 7 0 m cn 3. -0.3 2 4 2 I cn EI 0.0 a“ 2 2. 0 0 2 . 2 or 2 0 0 0 0 0 0 ,-a- 0. -0.6 1 0 0 0. 0 1 0 I 0 2 4 6 8 10 Mean result/pg g-’ 0.0 0.6 1.2 1.8 2.4 3.0 Mean result/pg g-’ Fig. 2 (a) Aluminium regression plot. The square points represent 20 samples (48 samples 27 1; 12 k; 7 m; and 2 h. See text for identities of 1 k m and h). r=0.998,; rn=1.01 (CI=O.99-1.03); and (b) Bland and Altman plot for A1<3 pg g-I Fig. 4 (a) Manganese regression plot (68 samples 30 1; 20 k; 11 m; and 7 h). r=0.98; rn=0.97 (CI=O.93-1.01); and (6) Bland and Altman plot for Mn562 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 199 1 VOL. 6 -160 other internal standards for the purpose of determining U and Th in lung tissue. These workers found an increase in the In signal over the duration of the first five measure- ments and thereafter a decay to about 50% of the initial value.The In in the standard behaved similarly to that in the digests but the heavy elements U and Th drifted by 10- 1 5% towards higher concentrations over the duration of eight consecutive measurements. The explanation of the initial increase found by Igarashi et a1.6 (but not observed in our own study) is not known but may lie in the fact that they used a concentric type nebulizer with an uptake rate of 1.7 ml min-I compared with our use of a Babington nebulizer with an uptake of 0.7 ml min-l. 0 - 0 I I 1 I Comparison of Atomic Absorption and Plasma Mass Spectrometry Aluminium The correlation coefficient r=0.998 equation is Inductively Coupled and the regression Al(ICP-MS)= 1.01( & 0.0 16)Al(ETAAS)+ 0.3( k 0.3) The slope m is not significantly different from unity nor is the intercept significantly different from zero.Fig. 2(a) shows that much of the data are concentrated in a group at a low concentration and this is not ideal for regression analysis which requires an even spread of the data. The regression is heavily influenced by the point at 150 pg g-l. In Fig. 2(b) the distribution of the data has no deleterious effect on the comparison. The mean difference or bias is 0.01 5 pg g-l [confidence interval (CI) from -0.07 to 0.0981 and is not significant. The limit of agreement is from -0.36 to 0.39 pg g-l. Magnesium The correlation coefficient r=0.9 1 and the regression equation is The slope m is not different from unity nor is the intercept Mg(ICP-MS)=0.92( f 0.14)Mg(FAAS)+46( -t 86) significantly different from zero [Fig.3(a)]. Fig. 3(b) shows the mean difference is 4.4 pg g-l (CI from - 15 to 24) i.e. there is no significant bias. The limit of agreement is from - 114 to 122 pg g-l. The poorer level of agreement found for Mg may be attributed at least in part to a combination of drift during the ICP-MS measurement (see Table 3) and the large dilution step ( x 50) necessary for FAAS. Manganese The correlation coefficient r= 0.98 and the regression equation is Mn(1CP-MS)=0.97( f 0.04)Mn(ETAAS)+O. 19( f 0.22) The slope rn is not significantly different from unity and the line passes through the origin [Fig. 4(a)]. Fig. 4(b) shows that the mean difference is - 0.07 pg g-l (CI from - 0.18 to 0.045) i.e.there is no significant bias. The limit of agreement is from -0.98 to 0.85 pg g-l. Iron The correlation coefficient r= 0.99 and the regression equation is Fe(1CP-MS)=0.965( f 0.028)Fe(FAAS)+ 6.9( k 26). The 95% CI for the slope is 0.94-0.99. As the CI does not include 1 m is significantly different from unity (p<0.05) [Fig. 5(a)] and a systematic error in one or other of the methods is indicated. This however is not true. Examina- tion of Fig. 5(b) shows that the differences fall into two populations thus invalidating simple regression analysis. Up to a mean value of about 400 pg g-l [Fig. 5(a)] there is very close agreement between the two methods while above 400 pg g-l the differences suddenly increase [Fig.5(6)]. This is attributed to the fact that dilutions become necessary for FAAS at about this point. Below 400 pg g-l the mean difference is -3.7 pg g-l (CI from - 12 to 4.6) i.e. there is no significant bias. The limit of agreement is from -26 to 19 pg g-*. Above 400 pg g-l there is a significant bias of 33.7 pg g-' (CI 6-61) and a limit of 2400 - w w z 1600 In 2 v) 7 800 a 0 0 (a) 0 0 0 0 0 3 0 0 . 0 . 0 3... 0.34 0.2 2. 2 3 0 07 20 4763 0 500 1000 1500 2000 2500 FAAS resuIt/pg g-' 50 I (c) 1 I 0 19 I 3 2j -50 -1 00 70 140 210 280 350 420 Mean resuIt/pg g-' 160 - 0 0 0 a' 0 0 0 0 0 0. 0. 0 0.0 0 - 0 2 . 0 0 . 0 0 0 0.0. 0 0 - 486250 00 00 0 . 0 0 0 160 c 0) 0 s o 400 800 1200 1600 2000 2400 Mean result/pg g-' Fig. 5 (a) Iron regression plot (75 samples 30 1; 27 k; 1 1 m; and 7 h).r=0.99; m=0.96 (CI=O.94-0.99); (b) Bland and Altman plot for all Fe data; (c) Bland and Altman plot for Fet400 pg g-'; and (d) Bland and Altman plot for Fe>400 p g g-'JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 199 1 VOL. 6 563 agreement of from - 146 to 21 3 pg g-l. This level of bias is unimportant for the work envisaged. Copper The correlation coefficient r= 0.99 and the regression equation is The slope is not significantly different from unity and the line goes through the origin [Fig. 6(a)]. Fig. 6(b) shows that the mean difference is -0.053 pg g-l (CI from -0.34 to 0.22) i.e. there is no significant bias. The limit of agreement is from -2.2 to 2.1 pg g-l. Cu(ICP-MS) = 1 .O 1 ( +- O.O3)Cu( FAAS) - 0.08( k 0.5 6) 0 45 t (a’ 0 3 - 30 3 E d.0 15 0 0. 0 0 0 - 0 3 0 00 0 02.2.. 2 232 0 240 2 0 4 70 1 7 0.0 0 0 0 0. o r n o 0 0 . 0 2 0 0 . 0 - 032 20 0 0 2 0 0 2 ....2.... 0 0 O H 0 0 0 0 0 Fig. 6 (a) Copper regression plot (62 samples 17 1; 28 k; 12 m; and 5 h). r=0.99; m=l.Ol (CI=O.98-1.04); and (b) Bland and Altman plot for Cu (a) 0 0 0 0 0 0 0 2 0 . 0. 002 aa 0 . 2 2. 0 . 100 200 300 400 500 FAAS result/pg g-‘ 35 I 1 -35 t 0 0 1 1 I I I 100 150 200 250 300 Mean result/pg g-’ Fig. 7 (a) Zinc regression plot (3 1 samples 13 1; 12 k; 4 m; and 2 h). r=0.98; m= 1.01 (CI=0.95-1.07); and (b) Bland and Altman plot for Zn 240 1 (a) 0 0 0 0 0 0 0. 0.0 0 0 0. a.3. 32. 0 1 1 1 0 50 100 150 200 250 FAAS result/pg g-’ l2 t (b’ 1 1 1 1 1 1 200 250 0 50 100 150 Mean result/pg g-’ Fig.8 (a) Cadmium regression plot (29 kidney samples). r= 1 .OO; m= 1.00 (CI=O.98-1.01); and (b) Bland and Altman plot for Cd Zinc The correlation coefficient r= 0.98 and the regression equation is Zn(1CP-MS)= 1.01( *0.06)Zn(FAAS)+7.5( f 12) showing that m is not significantly different from unity and that the line passes through the origin [Fig. 7(a)]. Fig. 7(b) shows the mean difference to be -7.9 pg g-l (CI from - 13 to -2.7) i.e. there is a positive bias towards ICP-MS. An explanation of the positive bias towards ICP-MS could be the presence of the polyatomic species 34S02+ arising from the 1% sulphur content of dry tissue. This explanation was tested by analysing a solution that matched the tissue digests in sulphur content (8 mmol dm-3 sulphuric acid).The 34SOtf signal at m/z 66 was negligible and cannot explain the bias. The limit of agreement is from -20 to 10 Pug g-l. Cadmium The correlation coefficient r= 1 .OO and the regression equation is indicating that m is not different from unity and that the intercept is zero [Fig. 8(a)]. Fig. 8(b) shows that the mean difference is 1.1 pg g-’ (CI from -0.06 to 2.1) and the limit of agreement from - 1.7 to 3.9. Cd(1CP-MS)= 1.00( f 0.016)Cd(FAAS)-0.80( * 1.8) Conclusion The statistical test of Bland and Altman5 is superior to regression analysis and revealed small biases for Fe (at high concentration) and Zn. These constant differences between the two methods are small but nevertheless irritating. The ICP-MS technique has been shown to give results for Al Mg Mn Fe Cu Zn and Cd that are generally in good agreement with reference materials and with standard methods of AAS.Analysis of reference materials IAEA H 4 IAEA H 8 and NIST SRM 1577a indicate that the ICP-MS564 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 199 1 VOL. 6 technique described here is also applicable to Co Mo Rb and Sr. The ICP-MS method is much faster and is comparable in accuracy to AAS. The increased dynamic range of the ICP-MS method is also advantageous as dilution errors are eliminated. The main problem that has to be guarded against is drift. In this work drift was controlled by a single internal standard and regular re- calibration. However other methods of control perhaps using multiple internal standards or isotope dilution should be considered. The authors thank Dr. Gordon D. Murray Medical Statistics Unit Department of Surgery Western Infirmary Glasgow for his assistance with the statistical aspects of this paper. References 1 2 3 4 5 6 Lyon T. D. B. Fell G. S. Hutton R. C. and Eaton A. N. J. Anal. At. Spectrom. 1988 3 265. Beauchemin D. McLaren J. W. and Berman S. S. J. Anal. At. Spectrom. 1988 3 775. Ridout P. S. Jones H. R. and Williams J. G. Analyst 1988 113 1383. Friel J. K. Skinner C. S. Jackson S. E. and Longench H. P. Analyst 1990 115 269. Bland J. M. and Altman D. G. Lancet 1986 i 307. Igarashi Y. Kawamura H. Shiraishi K. and Takaku Y. J. Anal. At. Spectrom. 1989 4 571. Paper 0/05 5 3 3 F Received December 1 Oth 1990 Accepted June 7th 1991