View Article Online / Journal Homepage / Table of Contents for this issue
1373
ANALYST, AUGUST 1992, VOL. 117
Published on 01 January 1992. Downloaded by University of Prince Edward Island on 25/10/2014 23:56:14.
Determination of Lithium at Ultratrace Levels in Biological Fluids by Flame Atomic Emission Spectrometry. Use of First-derivative Spectrometry Isabel Dol, Moises Knochen and Estela Vieras Universidad de la Republica,Facultad de Quimica, Catedra de Analisis Instrumental, Av. Gral. Flores 2 124, P.O. Box 1757, 7 7800 Montevideo, Uruguay
The use of zero-order and first-derivative flame emission spectrometry has been investigated for the determination of basal concentrations of lithium in serum and urine at the yg dm-3 level. No significant matrix effect was observed; however, it was necessary t o make use of background correction techniques. Two or three wavelength measurements and first-derivative spectrometry were used for this purpose; both methods gave similar results. Detection limits for serum of approximately 0.09 pg dm-3 were found by both zero-order and first-derivative emission measurements, while normal lithium levels were found to be 29.3 and 1.17 yg dm-3 for urine and serum, respectively. Keywords: Lithium; derivative spectrometry; emission spectrometry; background correction
During the last few years, there has been increasing interest in medical and biochemical fields, regarding the role of certain elements at the concentration ranges usually found in the human body (‘normal’ or basal levels).’ This is sometimes related to their action as nutrients, and sometimes to their possible application as indicators for the elucidation of various biological mechanisms, or in the detection of certain health disorders. As basal concentrations are very low, in the parts-perbillion (ppb) to parts-per-million (ppm) range, the determination of these elements is often carried out by atomic spectrometry, generally electrothermal atomic absorption spectrometry (ETAAS). However, despite the widespread use of modern techniques such as this, there are still many laboratories that do not have access to the necessary equipment. Hence there is a trend, especially in developing countries, towards the adoption of simpler techniques. The determination of lithium in biological fluids, in addition to the existing classical interest in psychiatric therapeutics, is important in nephrology, as it allows the study of certain aspects of renal physiology. It has also been stated that serum lithium levels may be increased in certain health disorders, such as chronic renal failure. A current research topic is the elucidation of the mechanisms for sodium and water re-absorption into the various parts of the nephron. Several workers have used lithium clearance as a tool to study the re-absorption of sodium and water into the proximal tubule.2-4 According to the literature, basal lithium levels in serum and urine are at the pg dm-3 (ppb) level. Values reported by different workers cover a wide range, up to 70 ppb. It can be assumed that the highest of the ‘normal’ values reported result from a lack of compensation for interferences, while some variability is also to be expected within individuals belonging to different populations, either by nationality, race or health. Bourret et a1.5 have reported a mean normal value of 8 ppb, with individual values ranging from 2 to 17 ppb, in determinations using ETAAS. The lack of a graphite furnace which is necessary for ETAAS determinations limits or hinders the research of some laboratories. However, despite the high intensity of the lithium emission line at 670.8 nm, little has been reported in the literature concerning its use for the determination of this element at ultratrace levels. Castillo et a1.6 have employed flame atomic emission spectrometry (FAES) for the determination of the normal lithium content of various human tissues and body fluids, finding average normal serum values of 0.8 ppb, one order of magnitude lower than those reported
by Bourret et al.5 The former group, however, did not comment on the possible effects of interferences on their results. It is often stated in the literature that FAES has lower accuracy and is more prone to interference. Taking into account that both aspects could be enhanced by adequate strategies, we have explored the feasibility of using FAES for the determination of lithium in biological fluids (serum and urine) at ultratrace levels. The existence of matrix effects and spectral interferences and the most suitable methods to overcome these problems have been studied. In this context, the use of the first derivative in the wavelength domain has been evaluated for background correction. Derivative spectrometry has been used extensively to overcome additive interferences in molecular absorption spectrophotometry,7-12 but its application in atomic emission spectrometry has received less attention. However, the very small linewidth of atomic emission spectra, compared with the shape of the background spectrum, suggests its use for this application.
Experimental Instrumentation Measurements were made with a Perkin-Elmer (Norwalk, CT, USA) 380 atomic absorption/emission spectrometer in the emission mode, using a three-slot burner, and an air-acetylene flame. The voltage attenuator in the analogue output of the instrument was disabled in order to obtain higher signals, in the 1 V range. An Apple IIe-compatible microcomputer with an Interactive Structures AI-13 12-bit analogue-to-digital ( N D ) interface was used for the acquisition of data from the spectrometer. The interface was controlled by a purpose-modified version of the Vidichart program (Interactive Microware, State College, PA, USA), which was also used for data processing. In the Perkin-Elmer instrument, wavelength scanning is performed by means of a built-in synchronous motor, thus ensuring a constant speed, while the constancy of the data acquisition rate is determined by the crystal-controlled clock of the computer. A Varian (Palo Alto, CA, USA) Model 20 strip-chart recorder was also used.
Operating conditions In preliminary experiments, the effect of the following variables was studied: type of burner (one- or three-slot); and bandwidth (0.2 or 0.7 nm), optimizing the gas flow rates and
View Article Online
Published on 01 January 1992. Downloaded by University of Prince Edward Island on 25/10/2014 23:56:14.
1374
ANALYST, AUGUST 1992. VOL. 117
burner position for maximum linearity and sensitivity. The three-slot burner was chosen because it provided better stability and precision. The effective bandwidth was chosen to be 0.2 nm in order to reduce background and other spectral interferences. The operating conditions used were as follows: air flow rate, 32 dm3 min-1; acetylene flow rate, 4 dm3 min-1; nebulizer intake rate, 3.8 cm3 min-1 (measured with water at 25 “C); observation height, flame base (burner height in position 7); wavelength, 670.8 nm; spectral bandwidth, 0.2 nm; and scanning speed, 5 nm min-1. Other conditions were as recommended by the manufacturer. The A/D interface setting was adjusted as required according to the manual, considering a 1 V maximum analogue output from the spectrometer. A total of 150 points were read for each spectrum, at a data acquisition rate of 2.88 points s-1. Reagents and Materials A lo00 ppm certified lithium atomic absorption standard solution (Fisher Scientific, Fair Lawn, NJ, USA) was diluted with distilled water as required. Serum and urine samples were obtained from healthy volunteers.
Table 1 Linear regression parameters f. confidence interval (CI) for signal height (background-corrected emission) and first derivative. y = a bx, with b in dm3 pg-l. CI = t s l c w h e r e t is Student’s ‘1’. s is the standard deviation and n the number of measurements
+
Value f CI
Signal heighta b r First derivativea
b r
The analytical signal could be obtained from either the instrument display (emission) or the computer. All of the values reported in this paper were taken from the computer, as it allowed more flexible data handling. The data obtained were signal height (emission) and first derivative for wavelength scans, as shown in Figs. 1 and 2. For data processing, the Vidichart program was used. Raw data were smoothed by means of a 5-point moving average, prior to peak-search and first-derivative calculation. The differentiation algorithm employed by Vidichart performs the difference between successive points. As the experimental points are equally spaced, this is equivalent to an incremental quotient. For background correction, the use of two- and threewavelength measurements was compared with first-derivative spectrometry in the wavelength domain. In the former method, emission readings at 670.8 nm were corrected for linear background at 670.0 and 672.0 nm. For calculation of the first derivative of the emission signal with respect to wavelength, emission was scanned in the range 669-673 nm, and the first derivative was calculated. The peak-to-peak value of the first derivative in the vicinity of the 670.8 nm line was used. Serum and urine samples were diluted with water (2 + 5 and 1 + 1, respectively) prior to measurement. Matrix effects were evaluated by means of the standard additions method (SAM).
Results Linearity
Linearity was studied separately in the range &50 ppb of lithium by means of a calibration graph consisting of seven duplicate points, and in the 0-5 ppb range with six points also in duplicate. Each standard solution was measured four times. Least-squares equations (y = a + bx) are shown in Table 1.
0-5 PPb
1.8 f.7.0 22.31 f.0.35 0.9997
4.0 t 2.5 82.5 f.1.0 0.9998
0.6 f 2.9 9.33 k 0.14 0.9998
2.6 f 1.3 34.1 k 0.6 0.9998
Table 2 Analytical precision for emission measurements performed on urine and in serum using signal height (emission minus background) and first-derivative spectrometry (SD = standard deviation; RSD = relative standard deviation) Urine
SDIpg dm-3 RSD (%)
Methodology
0-50 ppb
Signal height 0.637 1.5
First derivative 0.665 1.5
Serum Signal height 0.046 8.4
First derivative 0.058 8.6
Table 3 Verification of matrix effect in the determination of lithium in urine and serum by background-corrected atomic emission and first-derivative spectrometry. Slopes of the least-squares calibration graphs (Cal) and standard additions method (SAM) graphs are compared by means of a t-test. 0, = a + bx, with b in dm3 pg-1) Urine
b (Cal) b (SAM) t
Signal height 22.31 20.98 0.38
First derivative 9.33 9.30 1.01
Serum Signal height 82.5 83.4 0.396
First derivative 34.1 31.4 2.00
Verification of Matrix Effects
The matrix effects were studied by comparison of the slopes of the regression calibration graphs (described under Linearity) with those of the SAM graphs, under the same operating conditions. Comparison of the slopes of these graphs was performed using a t-test. The results are shown in Table 3. The theoretical value is t (0.025,35) = 2.04. From this result it can be concluded that there is no significant matrix effect. Accuracy
Percentage recoveries were calculated as follows. For serum, three serum aliquots (2 cm3) were diluted with 5 cm3 of distilled water, and with 0.5 and 1 ppb standard solutions, respectively. Recoveries with respect to the dilution with water were 97.2 and 101.4% (signal height), respectively (n = lo), and 106.0 and 101.6% (first derivative, n = 10). Similarly for urine, three aliquots (10 cm3) were diluted 1 + 1 with distilled water, and with 10 and 20 ppb standard solutions, respectively. Recoveries found were 99.1 and 102.7% (signal height, n = lo), respectively, and 102.1 and 100.2% (first derivative). Limits of Detection and Quantification
Precision
Analytical precision was calculated by preparing and measuring ten replicates of serum or urine dilutions, and calculating their respective concentrations by both methods. The results are reported in Table 2.
Analytical detection (30) and quantification (100) limits were calculated by repeated measurements (n = 10) of the respective analytical signals (background-corrected signal height and first derivative) obtained for serum samples with very low lithium concentrations.
View Article Online
1375
ANALYST, AUGUST 1992, VOL. 117
Results (detection and quantification levels, respectively) were: signal height, 0.089 and 0.41 pg dm-3; and first derivative, 0.097 and 0.50 pg dm-3.
Published on 01 January 1992. Downloaded by University of Prince Edward Island on 25/10/2014 23:56:14.
Normal Values For the population studied, normal values found for urine and serum were: urine, 29.3 pg dm-3 [n = 10, relative standard deviation (RSD) = 40%]; and serum, 1.17 pg dm-3 ( n = 10, RSD = 40%).
Discussion The determination of lithium at ultratrace levels by atomic emission spectrometry (AES) involves the risk of several types of interference, namely spectral, ionization and chemical interferences, and matrix effects. Spectral interferences are additive, while the remainder are mu1tiplicative, thus allowing their detection and correction by means of the SAM or by using spectroscopic buffers for ionization interferences. However, spectral interferences require more complex procedures for their detection and are not always easy to correct. Such interferences may arise from emission lines from concomitants, bands originating in metal oxides or other transient molecular species in the flame, or continua from various origins. Monochromators d o not behave as ideal devices, and stray-light levels found in commercial instruments, negligible for most routine applications, may turn out to be intolerable in the attempted detection of ultratrace levels. The first-derivative method proposed in this work seems to be adequate for background correction, even though it may be better to resort to the second derivative in order to cope with the sloping background. However, with the use of peak-to-peak first-derivative measurements good background suppression can be obtained together with an acceptable signal-to-noise ratio. The lack of a significant matrix effect is noteworthy, although it could be supposed that the ionization effect at least could be due to the presence of the saline matrix. However, lithium exhibits the highest ionization potential of the alkali metals (5.39 eV); hence, the number of atoms in the ionized state at the usual temperatures of the air-acetylene flame is less than 5%.13 Chemical interferences are more complex in nature. It is known that the lithium in the flame is in equilibrium with its hydroxide: Li
Evaluation and Control of Spectral Interference Owing to the high concentrations of organic matter and the various cations present in biological samples, a significant background emission could be expected. In order to assess the amount and influence of these emissions, emission spectra were obtained by scanning several nanometres t o each side of the nominal wavelength. The microcomputer and AID interface were employed for this purpose. In this way the form and magnitude of the spectral background were evaluated. Some of these spectra can be seen in Figs. l(a) and 2(a). The existence of a significant spectral background was verified for serum and, to a lesser extent, for urine analyses. In order to obtain the most appropriate correction method for the background, emission spectra were recorded within an interval 2 nm to each side of the nominal wavelength. It was found that, as expected, the background intensity is lower for a 0.2 nm bandwidth than for a 0.7 nm bandwidth, and that it is not constant with wavelength. The background for urine samples was comparatively less important than that for serum samples, and it was also fairly flat surrounding the Li line. Hence, two-wavelength correction by measuring the background at 672 nm was effective. However, the background was found to be very intense and sloping for serum samples. If two-wavelength background correction is used, biased results will be obtained; however, this method may still be effective for some purposes.
600
G .-I-’ 2
L. 2
+ 400 I
c
.-0 v)
.E 300
E a
.-
I-’
a
+ H 2 0 e LiOH + H
According to the literature the [LiOH] : [Li] ratio could reach a value of 10.14 From the above equation it could be inferred that this ratio depends proportionally on the [H20]: [HI ratio, so that the proportion of atomic lithium could be enhanced by raising the concentration of H , a situation observed at the base of the flame. This was verified experimentally. Determination of lithium by AES is most often carried out by using a dinitrogen oxide-acetylene flame, where lower interferences have been observed.
Evaluation of the Influence of Sodium and Potassium In previous work dealing with the therapeutic levels of lithium (approximately 5 ppm) in both serum and urine15 it was found that emission values decreased as the potassium concentration increased, contrary to what is expected on the basis of ionization. This phenomenon was verified by Thompson and Cummings,l6 who assigned it purely to chemical effects, excluding ionization. A similar study for ultratrace levels would be desirable; however, lithium contamination introduced by even ultrapure-grade potassium and sodium salts at this concentration level makes such a determination difficult to perform.
500
C
200
I
100
I
80
40 x $
0
0
- 40
-80
-120
,
669.0
669.8
670.6
671.4
672.2
Wavelengthhm
Fig. 1 (a) Smoothed emission spectrum of a urine sample diluted with water (1 1). (b) First-derivative emission spectrum of a urine sample diluted with water (1 + 1). For details, see text
+
View Article Online
1376
ANALYST, AUGUST 1992, VOL. 117 550
standards calibration graph, as no matrix effect could be detected. However, there is significant background emission caused by organic matter and the high concentration of other elements. This could be minimized by working with small spectral bandwidths (0.2 nm) and by the use of two- or three-wavelength correction or derivative spectrometry. The use of computer-assisted derivative spectrometry is promising in the sense that it is more independent of the operator’s criterion for selecting the wavelengths used for background estimation. It presents the additional advantage of being almost independent of the wavelength calibration of the monochromator. The scan does not need an exact pre-set wavelength, as is necessary for zero-order emission, where the monochromator should be tuned to the line peak. The results could be enhanced by adequate selection of the algorithms and of the number of points employed for the calculation of the derivative. Derivative spectrometry affords less significant figures than emission spectrometry; it is therefore necessary that data acquisition be made with the highest possible signal-to-noise ratio. The concentration values found in the serum samples are lower than those reported by Bourret et a1.5 and by Matusiewicz17 and approximately the same order of magnitude as the lithium content found in local tap water.15 However, the levels found in urine from healthy individuals were higher than those reported by Matusiewicz.17
(4
Published on 01 January 1992. Downloaded by University of Prince Edward Island on 25/10/2014 23:56:14.
A
450
!
References
1
-20 669.6
I
I
I
I
670.4
671.2
672.0
672.8
Wavelengthhm Fig. 2 ( a ) Smoothed emission spectrum of a serum sample diluted with water (2 + 5). A, Lithium peak. (b) First-derivative emission spectrum of a serum sample diluted with water (2 + 5). For details, see text. A and B are the extremes used for the measurements
Spectral interferences may also arise due to the series of bands surrounding the analytical line, including one band that overlaps it. This band interference is particularly troublesome when working at high sensitivity levels, such as in determinations performed in sera, and was found to originate in the flame itself, independently of the aspiration of water. The interference was tentatively assigned to molecular bands originating in species such as C2. This interference, additive in nature, cannot be compensated for by using the SAM, nor can it be corrected for by first-derivative spectrometry because of its structure which presents small bandwidths. However, its influence could be estimated from blank measurements and this correction could be used to compensate for it.
Conclusions It has been shown that it is possible to determine lithium in urine or serum directly at the ppb level versus an aqueous
1 Dawson, J. B., Fresenius’ 2. Anal. Chem., 1986, 324, 463. 2 Thomsen, K., Nephron, 1984, 37, 217. 3 Koomans, H . A . , Boer, W. H., and Dorhout Mees, E. J., Kidney Int., 1989, 36,2. 4 Durr, J. A . , Miller, N. L., and Alfrey, A . C., Kidney Znt. Suppl., 1990, 37, 58. 5 Bourret, E., Moynier, I., Bardet, L., and Fussellier, M., Anal. Chim. Acta, 1985, 172, 157. 6 Castillo, J . R., Fernandez, A . , and Bona, M. A . , At. Spectrosc., 1987, 8, 109. 7 Morrey, J. R., Anal. Chem., 1968,40, 905. 8 O’Haver, T. C., Anal. Chem., 1979,51, 91A. 9 Talsky, G., Mayring, L., and Kreuzer, H., Angew. Chem., Znt. Ed. Engl., 1978, 17, 785. 10 Cahill, J . E., Am. Lab., 1979, 11, 79. 11 Traveset, R . , Such, V., Gonzalo, R., and Gelpi, E., J. Pharm. Sci., 1980, 69, 629. 12 Levillain, P . , and Fompeydie, D., Analusis, 1986, 14, 1. 13 Flame Emission and Atomic Absorption Spectrometry, eds., Dean, J. A . , and Rains, T. C., Marcel Dekker, New York and London, 1975, vol. 3, p. 8. 14 Flame Emission and Atomic Absorption Spectrometry, eds. Dean, J. A., and Rains, T. C., Marcel Dekker, New York and London, 1969, vol. 1, pp. 116-121. 15 Dol, I., unpublished work. 16 Thompson, K. C., and Cummings, P. M., Analyst, 1984, 109, 511. 17 Matusiewicz, H., Anal. Chim. Acta, 1982, 136,215.
Paper 1I05194F Received October 14, 1991 Accepted January 13, 1992
1373
ANALYST, AUGUST 1992, VOL. 117
Published on 01 January 1992. Downloaded by University of Prince Edward Island on 25/10/2014 23:56:14.
Determination of Lithium at Ultratrace Levels in Biological Fluids by Flame Atomic Emission Spectrometry. Use of First-derivative Spectrometry Isabel Dol, Moises Knochen and Estela Vieras Universidad de la Republica,Facultad de Quimica, Catedra de Analisis Instrumental, Av. Gral. Flores 2 124, P.O. Box 1757, 7 7800 Montevideo, Uruguay
The use of zero-order and first-derivative flame emission spectrometry has been investigated for the determination of basal concentrations of lithium in serum and urine at the yg dm-3 level. No significant matrix effect was observed; however, it was necessary t o make use of background correction techniques. Two or three wavelength measurements and first-derivative spectrometry were used for this purpose; both methods gave similar results. Detection limits for serum of approximately 0.09 pg dm-3 were found by both zero-order and first-derivative emission measurements, while normal lithium levels were found to be 29.3 and 1.17 yg dm-3 for urine and serum, respectively. Keywords: Lithium; derivative spectrometry; emission spectrometry; background correction
During the last few years, there has been increasing interest in medical and biochemical fields, regarding the role of certain elements at the concentration ranges usually found in the human body (‘normal’ or basal levels).’ This is sometimes related to their action as nutrients, and sometimes to their possible application as indicators for the elucidation of various biological mechanisms, or in the detection of certain health disorders. As basal concentrations are very low, in the parts-perbillion (ppb) to parts-per-million (ppm) range, the determination of these elements is often carried out by atomic spectrometry, generally electrothermal atomic absorption spectrometry (ETAAS). However, despite the widespread use of modern techniques such as this, there are still many laboratories that do not have access to the necessary equipment. Hence there is a trend, especially in developing countries, towards the adoption of simpler techniques. The determination of lithium in biological fluids, in addition to the existing classical interest in psychiatric therapeutics, is important in nephrology, as it allows the study of certain aspects of renal physiology. It has also been stated that serum lithium levels may be increased in certain health disorders, such as chronic renal failure. A current research topic is the elucidation of the mechanisms for sodium and water re-absorption into the various parts of the nephron. Several workers have used lithium clearance as a tool to study the re-absorption of sodium and water into the proximal tubule.2-4 According to the literature, basal lithium levels in serum and urine are at the pg dm-3 (ppb) level. Values reported by different workers cover a wide range, up to 70 ppb. It can be assumed that the highest of the ‘normal’ values reported result from a lack of compensation for interferences, while some variability is also to be expected within individuals belonging to different populations, either by nationality, race or health. Bourret et a1.5 have reported a mean normal value of 8 ppb, with individual values ranging from 2 to 17 ppb, in determinations using ETAAS. The lack of a graphite furnace which is necessary for ETAAS determinations limits or hinders the research of some laboratories. However, despite the high intensity of the lithium emission line at 670.8 nm, little has been reported in the literature concerning its use for the determination of this element at ultratrace levels. Castillo et a1.6 have employed flame atomic emission spectrometry (FAES) for the determination of the normal lithium content of various human tissues and body fluids, finding average normal serum values of 0.8 ppb, one order of magnitude lower than those reported
by Bourret et al.5 The former group, however, did not comment on the possible effects of interferences on their results. It is often stated in the literature that FAES has lower accuracy and is more prone to interference. Taking into account that both aspects could be enhanced by adequate strategies, we have explored the feasibility of using FAES for the determination of lithium in biological fluids (serum and urine) at ultratrace levels. The existence of matrix effects and spectral interferences and the most suitable methods to overcome these problems have been studied. In this context, the use of the first derivative in the wavelength domain has been evaluated for background correction. Derivative spectrometry has been used extensively to overcome additive interferences in molecular absorption spectrophotometry,7-12 but its application in atomic emission spectrometry has received less attention. However, the very small linewidth of atomic emission spectra, compared with the shape of the background spectrum, suggests its use for this application.
Experimental Instrumentation Measurements were made with a Perkin-Elmer (Norwalk, CT, USA) 380 atomic absorption/emission spectrometer in the emission mode, using a three-slot burner, and an air-acetylene flame. The voltage attenuator in the analogue output of the instrument was disabled in order to obtain higher signals, in the 1 V range. An Apple IIe-compatible microcomputer with an Interactive Structures AI-13 12-bit analogue-to-digital ( N D ) interface was used for the acquisition of data from the spectrometer. The interface was controlled by a purpose-modified version of the Vidichart program (Interactive Microware, State College, PA, USA), which was also used for data processing. In the Perkin-Elmer instrument, wavelength scanning is performed by means of a built-in synchronous motor, thus ensuring a constant speed, while the constancy of the data acquisition rate is determined by the crystal-controlled clock of the computer. A Varian (Palo Alto, CA, USA) Model 20 strip-chart recorder was also used.
Operating conditions In preliminary experiments, the effect of the following variables was studied: type of burner (one- or three-slot); and bandwidth (0.2 or 0.7 nm), optimizing the gas flow rates and
View Article Online
Published on 01 January 1992. Downloaded by University of Prince Edward Island on 25/10/2014 23:56:14.
1374
ANALYST, AUGUST 1992. VOL. 117
burner position for maximum linearity and sensitivity. The three-slot burner was chosen because it provided better stability and precision. The effective bandwidth was chosen to be 0.2 nm in order to reduce background and other spectral interferences. The operating conditions used were as follows: air flow rate, 32 dm3 min-1; acetylene flow rate, 4 dm3 min-1; nebulizer intake rate, 3.8 cm3 min-1 (measured with water at 25 “C); observation height, flame base (burner height in position 7); wavelength, 670.8 nm; spectral bandwidth, 0.2 nm; and scanning speed, 5 nm min-1. Other conditions were as recommended by the manufacturer. The A/D interface setting was adjusted as required according to the manual, considering a 1 V maximum analogue output from the spectrometer. A total of 150 points were read for each spectrum, at a data acquisition rate of 2.88 points s-1. Reagents and Materials A lo00 ppm certified lithium atomic absorption standard solution (Fisher Scientific, Fair Lawn, NJ, USA) was diluted with distilled water as required. Serum and urine samples were obtained from healthy volunteers.
Table 1 Linear regression parameters f. confidence interval (CI) for signal height (background-corrected emission) and first derivative. y = a bx, with b in dm3 pg-l. CI = t s l c w h e r e t is Student’s ‘1’. s is the standard deviation and n the number of measurements
+
Value f CI
Signal heighta b r First derivativea
b r
The analytical signal could be obtained from either the instrument display (emission) or the computer. All of the values reported in this paper were taken from the computer, as it allowed more flexible data handling. The data obtained were signal height (emission) and first derivative for wavelength scans, as shown in Figs. 1 and 2. For data processing, the Vidichart program was used. Raw data were smoothed by means of a 5-point moving average, prior to peak-search and first-derivative calculation. The differentiation algorithm employed by Vidichart performs the difference between successive points. As the experimental points are equally spaced, this is equivalent to an incremental quotient. For background correction, the use of two- and threewavelength measurements was compared with first-derivative spectrometry in the wavelength domain. In the former method, emission readings at 670.8 nm were corrected for linear background at 670.0 and 672.0 nm. For calculation of the first derivative of the emission signal with respect to wavelength, emission was scanned in the range 669-673 nm, and the first derivative was calculated. The peak-to-peak value of the first derivative in the vicinity of the 670.8 nm line was used. Serum and urine samples were diluted with water (2 + 5 and 1 + 1, respectively) prior to measurement. Matrix effects were evaluated by means of the standard additions method (SAM).
Results Linearity
Linearity was studied separately in the range &50 ppb of lithium by means of a calibration graph consisting of seven duplicate points, and in the 0-5 ppb range with six points also in duplicate. Each standard solution was measured four times. Least-squares equations (y = a + bx) are shown in Table 1.
0-5 PPb
1.8 f.7.0 22.31 f.0.35 0.9997
4.0 t 2.5 82.5 f.1.0 0.9998
0.6 f 2.9 9.33 k 0.14 0.9998
2.6 f 1.3 34.1 k 0.6 0.9998
Table 2 Analytical precision for emission measurements performed on urine and in serum using signal height (emission minus background) and first-derivative spectrometry (SD = standard deviation; RSD = relative standard deviation) Urine
SDIpg dm-3 RSD (%)
Methodology
0-50 ppb
Signal height 0.637 1.5
First derivative 0.665 1.5
Serum Signal height 0.046 8.4
First derivative 0.058 8.6
Table 3 Verification of matrix effect in the determination of lithium in urine and serum by background-corrected atomic emission and first-derivative spectrometry. Slopes of the least-squares calibration graphs (Cal) and standard additions method (SAM) graphs are compared by means of a t-test. 0, = a + bx, with b in dm3 pg-1) Urine
b (Cal) b (SAM) t
Signal height 22.31 20.98 0.38
First derivative 9.33 9.30 1.01
Serum Signal height 82.5 83.4 0.396
First derivative 34.1 31.4 2.00
Verification of Matrix Effects
The matrix effects were studied by comparison of the slopes of the regression calibration graphs (described under Linearity) with those of the SAM graphs, under the same operating conditions. Comparison of the slopes of these graphs was performed using a t-test. The results are shown in Table 3. The theoretical value is t (0.025,35) = 2.04. From this result it can be concluded that there is no significant matrix effect. Accuracy
Percentage recoveries were calculated as follows. For serum, three serum aliquots (2 cm3) were diluted with 5 cm3 of distilled water, and with 0.5 and 1 ppb standard solutions, respectively. Recoveries with respect to the dilution with water were 97.2 and 101.4% (signal height), respectively (n = lo), and 106.0 and 101.6% (first derivative, n = 10). Similarly for urine, three aliquots (10 cm3) were diluted 1 + 1 with distilled water, and with 10 and 20 ppb standard solutions, respectively. Recoveries found were 99.1 and 102.7% (signal height, n = lo), respectively, and 102.1 and 100.2% (first derivative). Limits of Detection and Quantification
Precision
Analytical precision was calculated by preparing and measuring ten replicates of serum or urine dilutions, and calculating their respective concentrations by both methods. The results are reported in Table 2.
Analytical detection (30) and quantification (100) limits were calculated by repeated measurements (n = 10) of the respective analytical signals (background-corrected signal height and first derivative) obtained for serum samples with very low lithium concentrations.
View Article Online
1375
ANALYST, AUGUST 1992, VOL. 117
Results (detection and quantification levels, respectively) were: signal height, 0.089 and 0.41 pg dm-3; and first derivative, 0.097 and 0.50 pg dm-3.
Published on 01 January 1992. Downloaded by University of Prince Edward Island on 25/10/2014 23:56:14.
Normal Values For the population studied, normal values found for urine and serum were: urine, 29.3 pg dm-3 [n = 10, relative standard deviation (RSD) = 40%]; and serum, 1.17 pg dm-3 ( n = 10, RSD = 40%).
Discussion The determination of lithium at ultratrace levels by atomic emission spectrometry (AES) involves the risk of several types of interference, namely spectral, ionization and chemical interferences, and matrix effects. Spectral interferences are additive, while the remainder are mu1tiplicative, thus allowing their detection and correction by means of the SAM or by using spectroscopic buffers for ionization interferences. However, spectral interferences require more complex procedures for their detection and are not always easy to correct. Such interferences may arise from emission lines from concomitants, bands originating in metal oxides or other transient molecular species in the flame, or continua from various origins. Monochromators d o not behave as ideal devices, and stray-light levels found in commercial instruments, negligible for most routine applications, may turn out to be intolerable in the attempted detection of ultratrace levels. The first-derivative method proposed in this work seems to be adequate for background correction, even though it may be better to resort to the second derivative in order to cope with the sloping background. However, with the use of peak-to-peak first-derivative measurements good background suppression can be obtained together with an acceptable signal-to-noise ratio. The lack of a significant matrix effect is noteworthy, although it could be supposed that the ionization effect at least could be due to the presence of the saline matrix. However, lithium exhibits the highest ionization potential of the alkali metals (5.39 eV); hence, the number of atoms in the ionized state at the usual temperatures of the air-acetylene flame is less than 5%.13 Chemical interferences are more complex in nature. It is known that the lithium in the flame is in equilibrium with its hydroxide: Li
Evaluation and Control of Spectral Interference Owing to the high concentrations of organic matter and the various cations present in biological samples, a significant background emission could be expected. In order to assess the amount and influence of these emissions, emission spectra were obtained by scanning several nanometres t o each side of the nominal wavelength. The microcomputer and AID interface were employed for this purpose. In this way the form and magnitude of the spectral background were evaluated. Some of these spectra can be seen in Figs. l(a) and 2(a). The existence of a significant spectral background was verified for serum and, to a lesser extent, for urine analyses. In order to obtain the most appropriate correction method for the background, emission spectra were recorded within an interval 2 nm to each side of the nominal wavelength. It was found that, as expected, the background intensity is lower for a 0.2 nm bandwidth than for a 0.7 nm bandwidth, and that it is not constant with wavelength. The background for urine samples was comparatively less important than that for serum samples, and it was also fairly flat surrounding the Li line. Hence, two-wavelength correction by measuring the background at 672 nm was effective. However, the background was found to be very intense and sloping for serum samples. If two-wavelength background correction is used, biased results will be obtained; however, this method may still be effective for some purposes.
600
G .-I-’ 2
L. 2
+ 400 I
c
.-0 v)
.E 300
E a
.-
I-’
a
+ H 2 0 e LiOH + H
According to the literature the [LiOH] : [Li] ratio could reach a value of 10.14 From the above equation it could be inferred that this ratio depends proportionally on the [H20]: [HI ratio, so that the proportion of atomic lithium could be enhanced by raising the concentration of H , a situation observed at the base of the flame. This was verified experimentally. Determination of lithium by AES is most often carried out by using a dinitrogen oxide-acetylene flame, where lower interferences have been observed.
Evaluation of the Influence of Sodium and Potassium In previous work dealing with the therapeutic levels of lithium (approximately 5 ppm) in both serum and urine15 it was found that emission values decreased as the potassium concentration increased, contrary to what is expected on the basis of ionization. This phenomenon was verified by Thompson and Cummings,l6 who assigned it purely to chemical effects, excluding ionization. A similar study for ultratrace levels would be desirable; however, lithium contamination introduced by even ultrapure-grade potassium and sodium salts at this concentration level makes such a determination difficult to perform.
500
C
200
I
100
I
80
40 x $
0
0
- 40
-80
-120
,
669.0
669.8
670.6
671.4
672.2
Wavelengthhm
Fig. 1 (a) Smoothed emission spectrum of a urine sample diluted with water (1 1). (b) First-derivative emission spectrum of a urine sample diluted with water (1 + 1). For details, see text
+
View Article Online
1376
ANALYST, AUGUST 1992, VOL. 117 550
standards calibration graph, as no matrix effect could be detected. However, there is significant background emission caused by organic matter and the high concentration of other elements. This could be minimized by working with small spectral bandwidths (0.2 nm) and by the use of two- or three-wavelength correction or derivative spectrometry. The use of computer-assisted derivative spectrometry is promising in the sense that it is more independent of the operator’s criterion for selecting the wavelengths used for background estimation. It presents the additional advantage of being almost independent of the wavelength calibration of the monochromator. The scan does not need an exact pre-set wavelength, as is necessary for zero-order emission, where the monochromator should be tuned to the line peak. The results could be enhanced by adequate selection of the algorithms and of the number of points employed for the calculation of the derivative. Derivative spectrometry affords less significant figures than emission spectrometry; it is therefore necessary that data acquisition be made with the highest possible signal-to-noise ratio. The concentration values found in the serum samples are lower than those reported by Bourret et a1.5 and by Matusiewicz17 and approximately the same order of magnitude as the lithium content found in local tap water.15 However, the levels found in urine from healthy individuals were higher than those reported by Matusiewicz.17
(4
Published on 01 January 1992. Downloaded by University of Prince Edward Island on 25/10/2014 23:56:14.
A
450
!
References
1
-20 669.6
I
I
I
I
670.4
671.2
672.0
672.8
Wavelengthhm Fig. 2 ( a ) Smoothed emission spectrum of a serum sample diluted with water (2 + 5). A, Lithium peak. (b) First-derivative emission spectrum of a serum sample diluted with water (2 + 5). For details, see text. A and B are the extremes used for the measurements
Spectral interferences may also arise due to the series of bands surrounding the analytical line, including one band that overlaps it. This band interference is particularly troublesome when working at high sensitivity levels, such as in determinations performed in sera, and was found to originate in the flame itself, independently of the aspiration of water. The interference was tentatively assigned to molecular bands originating in species such as C2. This interference, additive in nature, cannot be compensated for by using the SAM, nor can it be corrected for by first-derivative spectrometry because of its structure which presents small bandwidths. However, its influence could be estimated from blank measurements and this correction could be used to compensate for it.
Conclusions It has been shown that it is possible to determine lithium in urine or serum directly at the ppb level versus an aqueous
1 Dawson, J. B., Fresenius’ 2. Anal. Chem., 1986, 324, 463. 2 Thomsen, K., Nephron, 1984, 37, 217. 3 Koomans, H . A . , Boer, W. H., and Dorhout Mees, E. J., Kidney Int., 1989, 36,2. 4 Durr, J. A . , Miller, N. L., and Alfrey, A . C., Kidney Znt. Suppl., 1990, 37, 58. 5 Bourret, E., Moynier, I., Bardet, L., and Fussellier, M., Anal. Chim. Acta, 1985, 172, 157. 6 Castillo, J . R., Fernandez, A . , and Bona, M. A . , At. Spectrosc., 1987, 8, 109. 7 Morrey, J. R., Anal. Chem., 1968,40, 905. 8 O’Haver, T. C., Anal. Chem., 1979,51, 91A. 9 Talsky, G., Mayring, L., and Kreuzer, H., Angew. Chem., Znt. Ed. Engl., 1978, 17, 785. 10 Cahill, J . E., Am. Lab., 1979, 11, 79. 11 Traveset, R . , Such, V., Gonzalo, R., and Gelpi, E., J. Pharm. Sci., 1980, 69, 629. 12 Levillain, P . , and Fompeydie, D., Analusis, 1986, 14, 1. 13 Flame Emission and Atomic Absorption Spectrometry, eds., Dean, J. A . , and Rains, T. C., Marcel Dekker, New York and London, 1975, vol. 3, p. 8. 14 Flame Emission and Atomic Absorption Spectrometry, eds. Dean, J. A., and Rains, T. C., Marcel Dekker, New York and London, 1969, vol. 1, pp. 116-121. 15 Dol, I., unpublished work. 16 Thompson, K. C., and Cummings, P. M., Analyst, 1984, 109, 511. 17 Matusiewicz, H., Anal. Chim. Acta, 1982, 136,215.
Paper 1I05194F Received October 14, 1991 Accepted January 13, 1992
