to be small, and these methods probably have a higher degree of accuracy than ... before analysis by gas chromatography-mass spectrometry. (GC-MS).
dIN. Cl-EM. 27/5, 733-735 (1981)
Accuracy of Some Routine Methods Used in Clinical Chemistry as Judged by Isotope Dilution-Mass Spectrometry I. BjOrkhem, A. Bergman, 0. Falk, A. Kallner,1 0. Lantto, L. Svensson, E. Akerlof, and R. Blomstrand Serum from patients was pooled, filtered, dispensed, and frozen. This pooled specimen was used for accuracy control in 64 participating laboratories in Sweden. Mean values (“state-of-the-art” values) were obtained for creatinine, cholesterol, glucose, urea, uric acid, and cortisol. These values were compared with values obtained with highly accurate reference methods based on isotope dilution-mass spectrometry. Differences were marked in the case of determination of creatinine and cortisol. Concerning the other components, the differences between the state-of-the-art value and the values obtained with the reference methods were negligible. Moreover, the glucose oxidase and the oxime methods for determination of glucose and urea were found to give significantly lower values than the hexokinase and urease methods, respectively. We conclude that methods with a higher degree of accuracy are required for routine determination of creatinine and cortisol. Addftlonal Keyphrases: reference methods analytical error quality control . accuracy of commonly used methods
proficiency testing Isotope dilution-mass spectrometry is an analytical technique with unusually high accuracy. It has been put forward that this technique should be ideal for development of “definitive” methods, i.e., methods with no known source of inaccuracy (1). We have developed several reference methods based on this principle (2-9). All possible sources of errors, such as errors due to pipettings, have not yet been eliminated and thus these methods cannot be regarded as definitive at the present state. The remaining errors are, however, judged to be small, and these methods probably have a higher degree of accuracy than methods based on most other principles. For example, the methods for determination of cholesterol and cortisol have been compared with similar isotope-dilution methods developed at the U.S. National Bureau of Standards (10) and at the University of Bonn (11), and the results were always within 1-2% (unpublished work). A further development of the reference methods into definitive methods would seriously affect their practicability. Definitive methods (10), in which by definition all possible sources of inaccuracy have been eliminated, may impose practical difficulties that preclude their routine use. Except for a small pilot study (12), our reference methods have so far only been used to assure the accuracy of different routine methods within our own laboratory. The goal of the present work was to evaluate the accuracy of laboratory performance on a national scale. A frozen serum pool from patients was prepared, and creatinine, cholesterol, glucose, urea, uric acid, and cortisol were determined with methods based on isotope dilution-mass spectrometry. This serum pool was
Department of Clinical Chemistry, Huddinge Hospital, Karolinska Institutet, S-141 86 Huddinge, Sweden. 1 Present address: Dept. of Clin. Chem., Karolinska Hospital, S-104 01 Stockholm, Sweden. Received June 17, 1980; accepted Feb. 9, 1981.
then distributed to most clinical laboratories in Sweden and assayed with their different routine methods. The “stateof-the-art” values obtained were then compared with the corresponding spectrometry.
values
obtained
by isotope
dilution-mass
Materials and Methods Isotope Dilution-Mass Fragmentography)
SpeCtrometry
Mass-fragmentographic
determination
concentration
(Mass of the cholesterol
in the serum pool was performed, with 2H7-
labeled cholesterol (obtained from Applied Science Laboratories, State College, PA 16801) as internal standard (cf. ref. 3). This compound had a considerably higher isotopic purity than the 2H4-labeled material used previously (3). The cholesterol in the samples was converted into the trimethylsilyl ether derivative before analysis by gas chromatography-mass spectrometry (GC-MS) in the present study. This derivatization improved the chromatography and facilitated the quantitation. The molecular ionsat mle 458 and mle 465 were used in the quantitation with an LKB 2091 instrument, under the same chromatographic conditions as previously (3). With this instrument a higher precision was obtained (cf. ref. 3 and Table 1), with no change in accuracy. Mass fragmentography of glucose with 2H7-glucose as internal standard was performed as described previously (4). Urea was determined as previously described by coupling with 5,5-diallylmalonic acid to yield a barbituric acid, with [15NaJurea as internal standard (5). The methyl ester of the barbiturate was analyzed by GC-MS. Mass fragmentography of urate was performed, with [‘5N2juric acid as the internal standard (9). The compound was isolated by means of ionexchange chromatography and converted into the tetra-trimethylsilyl derivative before GC-MS. Mass fragmentography of creatinine was performed, with (‘5N2lcreatinine as the internal standard (6). The compound was isolated from serum by “high-performance” liquid chromatography and converted into the di-trifluoro acetate of the (2-hydroxypropyl)ethyl derivative before GC-MS. Mass fragmentography of cortisol was performed with use of [2H4]cortisol as internal standard (KOR-Isotopes, Cambridge, MA 02142) essentially according to the previously prescribed procedure (8). Use of [2H4lcortisol instead of the (4-’4Cjcortisol used in the previous work increased the linear range. The ions at m/e 605 and m/e 609 were used in the mass-fragmentographic analysis of the methoxime-trimethylsilyl ether derivative (cf. ref. 8). In general, 20 independent replicates were analyzed, 10 on each day.
Accuracy Control of Different Routine Methods The reference serum, frozen at -70 #{176}C, was distributed to the 64 participating laboratories in 10-mL samples packed in solid carbon dioxide. The samples arrived frozen at each laboratory and were stored at -20 #{176}C until the day of analysis, then thawed at room temperature. In most cases, the analyses were performed immediately thereafter, but in a few cases the serum was stored for one or two days at 4 #{176}C before analysis. CLINICAL CHEMISTRY,
Vol. 27, No. 5, 1981
733
Table 1. Results of Determination of Components in the Reference MaterIal by Isotope DIlution-Mass Spectrometry and by VarIous RoutIne Methods Det.rmlnation
No. Component
Units
Cholesterol Glucose Urea Urate Creatinine
mmol/L mmol/L mmol/L mol/L .zmol/L
Cortisol
nmol/L
a
Standard error ofthe mean.
Each component routine method
was of
b
detns
by Isotope
Total
mean
0.02
5.79
0.03
17 10
6.39 284
0.05
43 64 41
4
43
20 9
78.7 436
0.05 4
64 18
high, the of 16 rep-
respectively.
Results Table 1 summarizes the results of the comparison between the mass-fragmentographic method and the routine methods for each compound studied. The concentration of cholesterol obtained with the massfragmentographic method (4.47 ± 0.02 mmol/L, mean ± SEM) was almost identical with the state-of-the-art value (4.52± 0.04 mmol/L, mean ± SEM). The interlaboratory CV was 5.0%. No significant2 differences were observed between the means of different methods (enzymic methods, n = 37, and methods based on the Liebermann-Burchard reaction, n = 6).
The concentration of glucose obtained with the massfragmentographic method (5.79 ± 0.03 mmol/L) was not significantly different from the state-of-the-art value (5.72 ± 0.05 mmol/L). The interlaboratory CV was 6.9%. The mean value for laboratories using different modifications of the glucose oxidase method (n = 40) was 5.60 ± 0.06 mmol/L, a value significantly different from the reference value. The mean value for laboratories using different modifications of the hexokinase method (n = 17) was 5.97 ± 0.09 mmol/L. This value was not significantly different from the reference value but was significantly (p
