JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY NOVEMBER 1994 VOL. 9 1293 Determination of Aluminium in Infusion Solutions by Inductively Coupled Plasma Atomic Emission Spectrometry-a Critical Comparison of Different Emission Lines Sebastian Recknagel Ullrich Rosick and Peter Bratter* Hahn-Meitner lnstitut GmbH Department of Trace Elements in Health and Nutrition Glienicker Strasse 100 D- 74 7 09 Berlin Germany Seven emission lines for the determination of Al in infusion solutions by means of inductively coupled plasma atomic emission spectrometry were compared. For this comparison wavelength scans of different samples were made in order to study the background in the region of each line and to inspect for interference structures arising from molecular bands of water and argon. Background equivalent concentrations and detection limits were calculated for the emission lines at 167 237 394 and 396 nm.The detection limits were 1.2 pg I-' for the 167 nm line 8.8 pg I-' for the 237 nm line 10 pg I-' for the 394 nm line 4.7 pg I-' for the 396 nm line and > 10 pg I -' for the other three lines (226 308 and 309 nm). The A1 content of various infusion solutions was determined on different days using the three emission lines with the lowest detection limits (167 237 and 396 nm). Day-to-day-variation was approximately 5% in all cases. As a final step the Al contamination of 25 infusion solutions for parenteral nutrition was evaluated by using the two most suitable emission lines (1 67 and 396 nm). No statistically significant differences between the results were detected.Advantages and disadvantages of these two emission lines for the determination of Al are discussed. Keywords Inductively coupled plasma atomic emission spectrometry; aluminium determination; infusion solutions It is now well known that for patients with an elevated intravenous supply of A1 and/or a disturbed renal excretion of Al the body burden of the element cannot be disregarded as harmless. There are two main groups of patients in danger from the toxic effects of Al dialysis patients and patients undergoing long-term parenteral nutrition. Therefore it is of vital interest to monitor continuously A1 contamination of the infusion solutions and of the haemodialysis preparations as well as the serum samples of the patients concerned in order to check the actual body burden.The serum A1 level of healthy people today is below 3.5 pg 1-1 (median value 1.5 pg 1-l)l and within the range of the detection limit of inductively coupled plasma atomic emission spectrometry ( ICP-AES).2-7 To deter- mine concentrations of A1 of c 3.5 pg 1-l in serum samples electrothermal atomic absorption spectrometry (ETAAS) is the method of choice. In infusion solutions for total parenteral nutrition A1 concentrations of up to 12000 pg I-' have been found.* Higher concentrations of A1 in infusion solutions can be determined better by ICP-AES rather than by ETAAS because of the greater dynamic range of the former method. Most workers who have described the determination of A1 by ICP-AES in different materials used the 396nm emission line.2,4,9-13 On the other hand the determination of A1 has sometimes been carried out with the 394 line,3.9-11 the 2376*7714 or the 309 nm line.4 Matusiewicz and Barnes successfully used the 308 nm emission line to measure A1 in biological materials after electrothermal vaporization of the ~amp1es.l~ However for the direct determination of A1 in aqueous solutions this line is not appropriate because of the strong OH- inter- ference.16-19 The aim of the present work was to compare the different emission lines that have been proposed to find out which lines are best suited for the determination of A1 in infusion solutions. Experimental Apparatus For ICP measurements the monochromator of a Jobin Yvon 70 Plus spectrometer (Instruments S.A.Longjumeau France) * To whom correspondence should be addressed.with a 40.68 MHz r.f. generator was used. The resolution of the spectrometer was 0.009 nm as given by the manufacturer (grating 2400 lines mm- '). The monochromator system and the optical interface between the plasma and the entrance to the spectrometer was purged with nitrogen (spectrometer 4 1 min-'; and interface 0.6 1 min-'). The incident power of the r.f. generator was 1.2 kW the argon flow rates were 12 1 rnin-' (outer gas) 0.3 1 rnin-l (intermediate gas) and 0.7 1 min-' (aerosol carrier gas). The intermediate- and aerosol carrier gas flows were controlled by a mass flow controller (MKS Instruments Andover MA USA). For nebulization a concen- tric pneumatic nebulizer (Meinhard type) was used with a sample flow rate of 1 mlmin-1 (Gilson Minipuls 2 peristaltic pump).Instrument control and data aquisition were performed by a Siemens PCD-2M personal computer running with the ISA software J-YESS 4.01. This software package works through peak searching which was used throughout this investigation. In this mode 7-1 1 points around the theoretical line maximum were measured with an integration time of 0.5 s. The actual peak maximum was then calculated using a Gaussian fitting procedure. Reagents and Samples Calibration for the determination was based on a 1 g1-l A1 standard solution [Al(NO3) in 0.5 moll-' HN03 Merck Darmstadt Germany]. The samples to be measured were infusion solutions for parenteral nutrition from different pharmaceutical manufacturers. The determination of A1 is highly susceptible to contami- nation due to the ubiquity of A1 in the laboratory atmos- phere.20,21 Therefore preparation of the samples was carried out under clean-room conditions. The samples were diluted and acidified only there was no digestion of the fluids before measurement.For dilution vessels made of Teflon-tetrafluor- ethylene-perfluorethylene (FEP Nalge Rochester NY USA) were used. These vessels were pre-cleaned with nitric acid (2%) and distilled water. The samples were taken with pre-cleaned pipette tips and diluted with distilled water. Except for the NaHCO solution all samples were acidified (to about pH 1) with HNO (65% Suprapur Merck) before measurement.1294 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY NOVEMBER 1994 VOL. 9 Emission Lines Investigated The six most sensitive emission lines for the determination of Al as given by Boumans,22 are the 396 394 309 308 237 and 226 nm lines.These lines and the emission line 167 nm described by Uehiro et a1.16 were investigated in the present study. As a first step 61 step scans with a step width of 4pm for a wavelength measured in the first optical order (A> 310 nm) and 2pm for wavelengths measured in the second optical order (A < 3 10 nm) were performed for different samples 1 argon; 2 distilled water; 3 500 pg 1-1 of A1 in water; and 4 a 1.25% serum protein solution with an A1 content of 25 p 1-l. A 61-step scan corresponded to a wavelength scan over 0.244nm in the first optical order and over 0.122nm in the second optical order. In a second step slopes of the analytical curves background equivalent concentrations (BEC) and detection limits (DL) were determined for the emission lines 396 394 237 and 167 nm on different days.Background measurement was carried out at the following positions 167 k0.0167 nm (for phosphate solutions; + 0.0557 nm instead of + 0.0167 nm); 237 _+ 0.0250 nm; 394 _+ 0.0670 nm; and 396f0.0535 nm. The third step was the determination of A1 in four infusion solutions calcium gluconate 10% (dilution factor 50) potassium lactate 1 mol 1-1 (20) the artificial colloid solution Gelifundol(10) and NaHCO 8.4% (lo) using the standard additions technique with two additions. The standard additions technique was used in order to avoid sample transport interferences. The samples were measured nine times on 5 d.The determination of A1 in these samples was performed at the emission lines 396 237 and 167 nm. For comparison 25 infusion preparations were analysed at the 396 and the 167nm lines. Comparisons of the determination by ICP-AES with ETAAS have been published else where.*^^^ Results and Discussion 61-Step Scans The results of the 61-step scans for the A1 emission lines investigated are shown in Fig. 1. It can be seen that the emission lines 308 and 309 nm are not sufficently sensitive for the determination of 25 pg 1-1 of A1 in a sample. No difference can be seen between the scans for water and for the 25 pg 1-1 infusion solution. In addition there is strong interference near the peak maxima of the two above mentioned emission lines.In the case of the 308 nm emission line Nygaard et a1.17-19 have described this interference as being due to a molecular band structure of OH-. It was possible to see this interference when measuring argon (without pumping) water and the infusion solution. The interference on the 309 nm emission line is also due to a molecular band structure of OH-. This interference has been mentioned by Winge et a1.24 The 226 nm emission line is also not particularly sensitive and it has a complex background with a weak interference on the peak maximum. This can be seen in Fig. l(b) in the scan of the water sample. Because of these disadvantages of the emission lines discussed above the following investigations were only carried out on the four remaining lines which showed no interference on the peak maxima.The matrix interference of organic sample components on the 167nm line will be dis- cussed later. Sensitivity Background Equivalent Concentrations and Detection Limits Determination of slope of the analytical curve (sensitivity) BEC and DL was achieved by measuring doubly distilled water and three standard solutions (50 200 and 500 pg I-' of Al). Each sample was measured five times apart from water which was evaluated ten times. Measurements were carried out on five successive days with two measurements a day (but on the last day only one measurement). The mean values [+ standard deviation (SD)] calculated from nine independent measurements are summarized in Table 1. As can be seen from the values the emission line at 167 nm had the lowest DL.It was lower than the DL of the 396 nm line by a factor of four. Detection limits of the two other lines were higher than that for 396nm by a factor of two. As expected the SDs of the DLs were approximately 30% for all emission lines. Background equivalent concentrations for the 237 and 396 nm lines were higher than that for 167 nm by a factor of 40 but half as large as that for the 394 nm line. When samples with a high Ca content (> 1000 mg 1-l) were measured BECs of the emission lines at 394.401 and 396.152 nm increased by a factor of five because of the strong Ca emission lines at 393.366 and 396.847 nm which interfere with the two A1 emission lines. The relative standard deviation (RSD) of the slope of the analytical curve is of greater interest than the value of the slope itself because the latter depends on the high voltage of the photomultiplier.The SD is a measure of the run-to-run stability of the measurement on the emission line being exam- ined. It is about 6% for the emission lines at 394 and 396 nm 11% for 237 nm and 26% for 167 nm. The reason for the higher value for the 167nm line is not known. A possible cause could be oxygen in the spectrometer which absorbs light in the far ultraviolet (UV) region (;1<200nm). This was excluded by measuring Ge-containing samples at the 164 nm line [(SD of the slope about 10% (data not published)]. Consequently it was necessary to recalibrate the spectrometer from time to time when measuring at the 167 nm line with the calibration technique. When working with the standard additions technique this problem does not arise as can be seen when the values for the residual error of the method for the above-mentioned emission lines are compared (see Table 1).In this case there were no significant differences between the values for the different emission lines. Within-run and Between-run Variation For four different infusion solutions nine independent determi- nations of A1 were made on five different days on the three different emission lines at 167 237 and 396 nm. The results of these measurements are summarized in Table 2. Statistical analysis with the Bartlett test showed no difference in the variances among the four emission lines (Bartletts x2= 3.3 for two degrees of freedom and therefore p < 0.05). No differences between the relative standard deviations (RSD%) could be Table 1 Comparison of performance of four prominent emission lines for the determination of A1 in pure water samples ~ ~~ Between-run variation* Within-run SD Wavelength/ Detection limit/ BEC/ Sensitivity/ Recovery variation of sensitivity/ of method/ (Yo) arbitrary units per pg 1 - ' I % I-' nm Pg I-' K 3 - l arbitrary units per pg 1-' 167.020 1.2k0.40 9.9 & 1.9 1.4fO.36 99.6 f 4.4 + 0.045 237.312 8.8 k 2.3 424 62 1.0 f 0.12 102.0 _+ 2.9 0.014 394.401 1Of 3.3 864 f 104 0.63 & 0.041 101.2f2.7 7 0.005 396.152 4.7k1.7 446k50 1.15 & 0.0'7 100.7 5 3.2 f 0.010 3.8 2.3 2.6 2.9 * For n = 9 independent measurement cycles.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY NOVEMBER 1994 VOL.9 1295 I / I 60 40 20 0 500 ( C) +- L3V .- 8 600 500 .- +- - 400 300 200 100 960 920 880 840 800 I I 5 15 25 500 ( b ) 1 450 400 350 300 250 600 (dl 100 ’ I I I I 1 660 640 620 35 45 55 Step number Fig. 1 A argon; B water; C sample 1 (1 +4); D 500 pg 1-’ of A1 (for the sake of clarity the individual samples have been shifted) Details of the 61-step scans of four samples at the A1 emission lines (a) 167; (b) 226; (c) 237; ( d ) 308; (e) 309; (f) 394.401; and (g) 396.152 nm.Table 2 Determination of A1 in infusion solutions of the A1 emission lines 167 237 and 396nm; in each case nine independent determi- nations were performed A1 content (+SD)/pg 1-’ Sample Dilution 167.020 nm 237.312 nm 396.152 nm Calcium gluconate 1 + 50 5087 f 249 5008 & 255 5139 f 272 Gelifundol l + 1 0 906f46 865f53 895f27 Potassium lactate 1 + 20 2350 f 101 2370 f 168 2310 f 122 Sodium hydrogen 1 + 10 572 f 22 584 f 28 604 & 32 carbonate shown for the samples.The mean value was 5.2%. Comparison of within-run and between-run variation for the A1 emission lines evaluated did not lead to any conclusion being drawn as to which emission line should be preferred for the determi- nation of A1 in biological samples. Comparison of the 167 and the 396 nm Lines The concentrations of A1 in the samples investigated by means of the two different wavelengths are summarized in Table 3. Agreement was found for most of the preparations. The values measured at the 167 nm line as compared with those measured at the 396 nm line are shown in Fig. 2. The statistical calcu- lation done by means of the reduced major axis method25 shows the equivalence of the results obtained at the two emission lines evaluated.Sample 9 1034 was excluded because the A1 content of this sample was much higher than all of the other values measured. Interferences The interference of Ca on the 396 nm line mentioned above is normally not a problem even when there are high concen- trations of Ca in the solutions being analysed provided that1296 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY NOVEMBER 1994 VOL. 9 Table 3 A1 content (pg 1-l) of various infusion solutions for parenteral nuixition by means of ICP-AES using the emission lines 167 and 396 nm ~ ~~ ~ Sample No. 91009 91020 91021 91023 91027 91028 91029 91034 91038 91039 91041 91042 91043 9 1045 9 1046 91047 91048 91050 91051 91052 91055 91056 91059 91073 91075 Sample (manufacturer) Periplasmal 3.5% (Braun) Sodium hydrogen carbonate 8.4% (Braun) Potassium-L-malate 17.21 % (Braun) Calcium-Braun 10% (Braun) Inzolen PSA (Kohler) Inzolen PSI (Kohler) Inzolen Infantibus (Kohler) Sodium glycerophosphate (Pfrimmer) Gelifundol (Biotest Ph.) Serum protein solution 5% (DRK) Serumar 20 ml (Armour-Pharma) Serumar 250 ml (Amour-Pharma) Vitamin C 500 mg (Pharma Hameln) Calcitrans 10% (Fresenius) Potassium lactate (Fresenius) Calcium-Braun 10% (Braun) Calcium gluconate 10% (Ph.Hameln) Sodium chloride (Pfrimmer) Potassium phosphate (Pfrimmer) Potassium chloride (Pfrimmer) HAES-steril 10% (Fresenius) Lipofundin MCT 20% (Braun) Seltrans (Fresenius) Potassium phosphate (Braun) Sodium glycerophosphate (Pfrimmer) Dilution 1 :2 1:6 1 10 1 10 1:lO 1 10 1 10 1 10 1 4 1:4 1 4 1:3 1 4 1 10 1:lO 1 10 1 10 1:2 1:3 1:2 1:4 1:2 1:3 1:2 1 10 167.020 nm 22 999 1385 4955 2088 886 666 11998 865 61 197 66 986 3934 1940 6372 462 1 12 3758 15 86 39 77 77 1355 396.152 nm 37 92 1 1399 4957 2086 877 590 12143 892 75 235 76 958 3549 2071 6204 4426 12 3853 18 71 32 82 56 141 1 background correction is done on both sides of the A1 peak.However in another investigation problems arose with the determination of A1 in Ca-containing solutions at the 396 nm line. In this determination measurements were made five times at the emission lines 167 and 396 nm and approximately 10% more A1 was found with the 396 nm line than with the 167 nm line in the original calcium gluconate solution (lo% dilution 1 + 49).The reason for this over-estimation has not been clarified yet and needs to be evaluated in a future investigation. Uehiro et al. have described an interference of Fe on the 167 nm emission line.16 The interference factor given by these workers was 0.0018 g of A1 per g of Fe. In an earlier published investigation the present workers found a slightly lower inter- ference factor of 0.0013 g of A1 per g of Fe.23 There is another disturbance near the A1 peak at 167 nm which results from phosphates. This is shown in Fig. 3. The distance between the two peaks is 0.03 nm. Problems can arise when measuring phosphate-containing solutions and the back- ground is extrapolated from the position of the P peak (position A in Fig.3). No problems arose when the correct background measurement position was set to the right of the P peak (position B in Fig. 3). More important than the interference of P are molecular bands that appeared when solutions with a high C content were analysed. The intensity of these molecular bands is dependent not only on the C content of the solution but also on the structure of the organic compound. Alcohols e.g. ethanol and methanol show much higher molecular bands than other organic compounds such as acetic acid or amino acids.23 For example the molecular bands that result from ethanol at a concentration level of 1 moll-' of C in a solution containing 500 pg 1-' of A1 are shown in Fig. 4. Furthermore the intensity of these molecular bands also depends on the aerosol carrier gas flow rate.With a low flow rate (0.5 1-' min) the A1 to C signal ratio is much higher than with a higher flow rate (e.g. 0.8 1 min-'). These molecular bands can cause high values which are misleading when samples are analysed using the standard additions technique. A reduced aerosol carrier gas flow of 0.5 1 min-' was therefore used in the present study for the determination of A1 in infusion solutions with a high C content. In addition to the above-mentioned spectral interferences there are two other possible sources of interference (i) sample transport interferences and (ii) chemical interferences e.g. alkaline interferences because of ionization effects in the plasma. These were avoided by using the standard additions technique throughout this study.Conclusion For the determination of A1 in infusion solutions which are representative of a great variety of biological samples in the 5000 r Al content/pg I-' (12=167 nm) Fig.2 Comparison of methods for determining A1 in infusion solu- tions for parenteral nutrition using the emission lines at 167 and 396 nm 1 13 25 37 49 61 Step number Fig. 3 Interference of P near the A1 emission line at 167 nm solution contained 15.8gl-' of Pod3- and 1.25mg1-' of Al. See text for explanation of positions A and BJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY NOVEMBER 1994 VOL. 9 700 600 r 500 4- .- v) C Q) 4- .E 400 > .- t l - 2 300 200 100 - - - - - - - 0 I I 166.98 167.03 167.08 Wavelengt hlnm 1297 Fig. 4 Variation of aerosol carrier gas flow at the 167 nm emission line sample 500 pg 1-' of A1 in water containing ethanol (1 moll-' of C); A 0.5; B 0.6; C 0.7; and D 0.8 1 min-' gas flow composition of their matrices only two of the seven emission lines evaluated have detection limits below 5 pg 1-I.These two lines 167 and 396 nm are well suited for measuring A1 in biological samples within the existing guidelines for maximum permissible A1 concentrations in infusion solutions (10 pg 1-l). For the determination of A1 in serum of untreated patients ICP-AES is not sensitive enough because the concentration of A1 in serum of healthy people is below 3.5 pg I-' hence ETAAS has to be used. For the analysis of infusion solutions which according to the results presented could contain in some cases widely varying concentrations of Al the use of ICP-AES is more suitable than ETAAS.Compared with the latter ICP-AES has a wider dynamic range and therefore sample preparation is less susceptible to error (e.g. contami- nation) when samples to be analysed have to be highly diluted. Owing to the wide variations in the matrices of the samples analysed throughout this study the standard additions tech- nique was applied. The reason for this was to avoid systematic errors resulting from undetected non-spectral interferences. When series of samples have more uniform matrix composi- tions the less time-consuming calibration technique of using standard solutions can of course be used. Both of the recommended wavelengths have their advantages and disadvantages. The detection limit of the 167nm line is lower than that of the 396 nm line.On the other hand measuring at an emission line in the vacuum UV is more difficult because of possible problems such as air in the system or soiling of optical components of the spectrometer. When samples with a high Fe content or a high content of organic C are to be evaluated it is better to work at the 396 nm line because of the interferences on the 167 nm line. For determining A1 in samples with a high Ca content the 167nm line is more appropriate because of the strong Ca emission line near the A1 line at 396 nm. References 1 Gramm H.-J. Bratter P. Rosick U. Bohge P. and Recknagel S. Infusionsther. Transfusionsmed. in the press. 2 Schramel P. Wolf A. and Klose B. J. J. Clin. Chem. Clin. Biochem. 1980 18 591. 3 Lichte F.E. Hopper S. and Osborn T. W. Anal. Chem. 1980 52 120. 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Sanz-Medel A. Roza R. R. Alonso R. G. Vallina A. N. and Cannata J. J. Anal. At. Spectrom. 1987 2 177. Berner Y. N. Shuler T. R. Nielsen F. H. Flombaum C. Farkouth S. A. and Shike M. Am. J. Clin. Nutr. 1989 50 1079. Coni E. Bellomonte G. and Caroli S. J. Trace Elem. Electrolytes Health Dis. 1993 7 83. Violante N. Petrucci F. Delle Femmine P. and Caroli S. Microchem. J. 1992 46 199. Recknagel S. Bratter P. Chrissafidou A. Gramm H.-J. Kotwas J. and Rosick U. Infusionsther. Transfusionsmed. 1994 21 266. Allain P. and Mauras Y. Anal. Chem. 1979 51 2089. Narayaman P. Csanady G. Wegscheider W. and Knapp G. J. Anal. At. Spectrom. 1989 4 347. Leflon P. Plaquet R. Mornibre A. and Fournier A. Clin. Chim. Acta 1990 191 31. Brenner I. B. ICP Inf. Newsl. 1992 18 289. Arniaud D. ICP In5 Newsl. 1990 16 39. Coni E. Stacchini A. Caroli S. and Falconieri P. J. Anal. At. Spectrom. 1990 5 581. Matusiewicz H. and Barnes R. M. Spectrochim. Acta Part B 1984 39 891. Uehiro T. Morita M. and Fuwa K. Anal. Chem. 1984,56,2020. Nygaard D. D. Chase D. S. Leighty D. A. and Smith S. B. Anal. Chem. 1984 56 424. Nygaard D. D. Sotera J. J. Spectroscopy 1988 314 22. Nygaard D. D. Chase D. S. and Leighty D. A. Appl. Spectrosc. 1983 37 432. Rosick U. Recknagel S. Bratter P. in Mineralstofle und Spurenelemente in der Ernahrung des Menschen ed. Bratter P. and Gramm H.-J. Blackwell Wissenschaft Berlin 1991 p. 104. Ericson S. P. At. Spectrosc. 1992 13 208. Boumans P. W. Line Coincidence Tables for Inductively Coupled Plasma Atomic Emission Spectrometry Pergamon Press Oxford 1980 vol. 1. Recknagel S. Rosick U. Tomiak A. and Bratter P. in 6. Colloquium Atomspek-trometrische Spurenanalytik ed. Welz B. Perkin Elmer GmbH ifberlingen 1991 p. 841. Winge R. K. Peterson J. and Fassel V. A. Appl. Spectrosc. 1979 33 206 Feldmann U. Schneider B. Klinkers H. and Haeckel R. A. J. Clin. Chem. Clin. Biochem. 1981 19 121. Paper 41029731 Received May 18 1994 Accepted July 14 1994