Summary: A recently introduced automated nephelometric immunoassay ..... boratorium e.V. (INSTAND e.V., Düsseldorf, Ger- reference materials (CRM) from ...
Eur J Clin Chem Clin Biochem 1996; 34:847-852 © 1996 by Walter de Gruyter · Berlin · New York
Quantitative Nephelometric Assay for Determining a-Foetoprotein Evaluated Dirk R. Bernard, Joris R. Delanghe, Marc L. De Buyzere and Geert G. Leroux-Roels Laboratorium voor Klinische Biologie, Afdeling Klinische Scheikunde, Universitair Ziekenhuis Gent, Gent, Belgium
Summary: A recently introduced automated nephelometric immunoassay involving shell/core particles for determination of α-foetoprotein in serum and amniotic fluid was evaluated with the Behring Nephelometer analyser II. Method stability was good: reconstituted reagents and calibration curve were stable for at least one week. The intraassay CV varied between 2.3% and 4.0%. The inter-assay CV varied between 3.5% and 4.6%. Samples with afoetoprotein concentrations up to 273 000 μg/l were analysed without high-dose "hook" effect after automatic dilution. No significant interference from haemoglobin, bilirubin, rheumatoid factors, or human anti-mouse antibodies was detected up to concentrations of 0.15 mmol/1 haemoglobin, 268 μιηοΐ/ΐ bilirubin, 470 int. units/1 rheumatoid factor and a titre of 1/1000 human anti-mouse antibodies. Interference due to triacylglycerols depended on the size of triacylglycerol containing particles: for VLDL, interference did not occur up to triacylglycerol levels of 6.0 mmol/1, for chylomicrons interference was already noted at triacylglycerol levels of 1.0 mmol/1. Correlation with a commercial RIA (Kabi Pharmacia) was excellent both for serum (n = 65) and amniotic fluid (n = 100). The effect of the molecular variation of the carbohydrate moiety of α-foetoprotein on the test results was studied using concanavalin A affinity chromatography. The detection of both concanavalin Α-reactive and concanavalin A-nonreactive α-foetoprotein was equivalent by both methods. Multimeric forms of α-foetoprotein were prepared by gel permeation chromatography. The effect of autopolymerization of α-foetoprotein on the nephelometric determination of α-foetoprotein was negligible. We conclude that latex-enhanced immunonephelometry is a rapid, practical, and reliable method for measuring α-foetoprotein in serum and amniotic fluid. Introduction α-Foetoprotein is a glycoprotein with a relative molecular mass of approximately 70000 and a carbohydrate content of about 4%. It is one of the major foetal plasma proteins and is synthesised during embryonic development by the yolk sac, foetal liver and gastrointestinal tract (1). After birth, serum α-foetoprotein concentration decreases and is detectable only in trace amounts after the first year of life (1, 2). Serum α-foetoprotein levels have been used extensively in the screening for, and monitoring of, hepatocellular carcinoma (3, 4) and germ cell tumours containing yolksac or endodermal sinus elements (3, 5). Increased serum α-foetoprotein levels have also been demonstrated in non-malignant liver diseases such as viral hepatitis and liver cirrhosis (3). α-Foetoprotein concentrations in amniotic fluid (6) and maternal serum (1, 6, 7) are widely used for screening for foetal 'malformations. Increased α-foetoprotein concentrations in amniotic fluid and maternal serum are associated with neural tube defects and other foetal disorders such as ventral wall defects, cystic hygroma, teratoma and gastrointestinal malformations (1, 6, 7). Decreased α-foetoprotein concentrations in amniotic fluid and maternal serum are
found in Down 's syndrome, foetal distress and foetal dismise (1,6, 7). Although the carbohydrate content of α-foetoprotein is low, differences in saccharide transferase activity in afoetoprotein-producing cells result in a heterogeneity in the carbohydrate moiety of α-foetoprotein. Differences in the carbohydrate side chains of α-foetoprotein may be determined by the binding of α-foetoprotein to the lectin concanavalin A (8—12). The proportion of α-foetoprotein that does not react to concanavalin A differs in foetal serum and amniotic fluid: 2 to 6% for foetal serum α-foetoprotein and 15 to 45% for amniotic fluid α-foetoprotein (8, 9). The α-foetoprotein produced by hepatocellular carcinoma shows a lower proportion of concanavalin A-non-reactive α-foetoprotein than the afoetoprotein produced by germ cell tumours (10—12). Furthermore, α-foetoprotein is subject to spontaneous and simultaneous polymerization and degradation during storage and dilution in saline solutions (13, 14). This molecular heterogeneity may influence the detection and quantification of α-foetoprotein in immunonephelometric assays.
Bernard et al.: Ncphelomctric α-foetoprotcin determination
848 Assays of α-foctoprotcin in serum and amniotic fluid currently involve radioimmunoassays (RIA) (6) or immunoradiomctric methods (15). More recently, non-radioactive alternative methodologies such as enzyme immunoassays (2, 4), chemiluminescence (16), fluorescence (17), microparticle capture enzyme immunoassays (18, 19) and automated particle-counting immunoassays (20) have been developed. In the present study, we evaluated a recently introduced automated nephelometric method based on shell/core polymer particles coated with anti-ct-foetoprotein antibodies and compared results with a commercially available RIA. Special attention was paid to the comparison between RIA and the nephelometric assay with respect to the molecular heterogeneity of the carbohydrate moiety of α-foetoprotein and the occurrence of a-foetoprotein aggregates, since the latter increase non-specific binding in double antibody RIA assays, resulting in decreased assay sensitivity (21). Materials and Methods Methods Nephelometric method α-Foetoprotein in serum and amniotic fluid was assayed with the Behring N latex AFP test (Behringwerke, Marburg, Germany) based on shell/core particles (mean diameter: 175 nm) coated with murine anti-a-foetoprotein monoclonal antibodies (22). The assay was performed according to the manufacturer's procedure on a Behring Nephelometer Analyser II (Behringwerke, Marburg, Germany). Eighty μΐ of serum was diluted fivefold in N-diluent. After a pre-incubation time of 6 minutes, final readings were taken after 18 minutes at a standard wavelength of 840 nm. Incubation temperature was 37.0 ± 0.1 °C. The pre-incubation period with a volume of 2 μΐ prediluted serum sample was programmed to detect extremely high concentrations of α-foetoprotein (fig. 1). When at the end of the pre-incubation period the light scatter exceeds the prereaction threshold, the reaction is automatically rerun using a higher dilution. For amniotic fluid samples a predilution of 1/2000 in N-diluent was used.
Imprecision Two different serum pools (SI, S2) and two amniotic fluid pools (AI, A2) were prepared. We tested intra-assay reproducibility with a series of ten aliquoted samples from the different pools. Interassay reproducibility was evaluated by assaying serum and amniotic fluid pools for ten consecutive days. • \ Sensitivity and linearity We tested die sensitivity of the nephelometric assay for serum in the standard fivefold dilution as provided by the instrument. To evaluate the linearity of the nephelometric assay, we used serial automated or manual dilutions of amniotic fluid samples, which contained high concentrations of α-foetoprotein (range: 500086 800 μg/l). Amniotic fluid samples in a pre-dilution of 1/5 up to a concentration of 273 000 μg/l were used to test for high dose "hook" effect. Interference studies We studied the effects on α-foetoprotein test results of various potentially interfering substances at different ratios. Effects of haemoglobin, bilirubin, rheumatoid factor, and human anti-mouse antibodies were studied by adding to samples serum that was enriched by these compounds up to final concentrations of 0.15 mmol/1, 268 μτηοΐ/ΐ, 470 int. units/1, and a titre of 1/1000 human anti-mouse antibodies, respectively. The effect of chylomicronaemia on the nephelometric test results was evaluated by adding to normolipaemic serum pools (n = 5), fresh serum, rich in chylomicrons with a final chylomicron level ranging from 0.8 mmol/1 up to 2.75 mmol/1 (expressed as triacylglycerol content). Similarly the effect of verylow density lipoprotein (VLDL) triacylglycerols was tested by adding serum from patients with hyperlipoproteinaemia type IV (final triacylglycerol content up to 6.0 mmol/1) (n = 6). Comparison study For comparison, we assayed α-foetoprotein with an RIA (AFP RIACT, Kabi Pharmacia, Uppsala, Sweden), performed according to the package insert. Radioactivity was measured in an LKB multichannel gamma-counter (1261 Multi Gamma, LKB Wallac, Wallac Oy, Turku, Finland). Affinity chromatography Affinity chromatography of serum and amniotic fluid samples on concanavalin A-Sepharose 4B (Pharmacia Fine Chemicals, Uppsala, Sweden) was performed as described by Buamah et al. (9). The concanavalin Α-reactive and non-reactive fractions were analysed by both RIA and nephelometry.
4000 - End of pre-reaclion σ> W
Stability of calibration and reagents The stability of a calibration curve and reconstituted reagents was evaluated by reanalysing a series of forty serum samples with a single calibration curve and a single pool of reconstituted reagents, stored at 2 to 8 °C after two and seven days.
Gel permeation chromatography
3000 - -
Incubation time [s]
1000
Fig. 1 Typical reaction pattern of Behring N latex AFP determination on a Behring Nephelometer Analyser II. First, a pre-reaction of 360 s is performed with 2 μΐ of prediluted sample (serum or amniotic fluid). After 360 s the additional 80 μΐ of pre-diluted sample is injected to start the main reaction. Final readings (bit signals) occur at 1080s.
To evaluate the effect of spontaneous α-foetoprotein dimer formation in amniotic fluid on test results, 10 arnniotie fluid samples were stored for 72 h at room temperature. Apparent A/r of α-foetoprotein was determined on these 10 stored samples by means of gel permeation chromatography using an isocratic (0.1 mol/l potassium phosphate pH 7.2) HPLC pump system. A Waters 650E Advanced Protein Purification system (Millipore, Milford, MA 01757, USA) was combined with a Protein PAK Glass 300SW column (Waters Nihon Millipore, Tokyo, Japan). Fractions corresponding to a mean apparent Mr of 70 000 and 140 000 were collected in a Retriever II fraction collector (Isco, Inc, Lincoln, NE 68505). In this way, monomeric and dimeric forms of α-foetoprotein were separated and collected. The α-foetoprotein concentration in the obtained fractions was determined using both the nephelometric and RIA assay.
849
Bernard et al.: Nephelometric α-foetoprotein determination Desaggregation of the dimeric fraction was performed by treating the samples for 24 h at room temperature with 8 mol/1 urea. Urea was removed by dialysis against saline for 12 h.
Range and linearity
Amniotic fluid was obtained by percutaneous aniniocentesis and centrifuged at 3000 g for 10 min. α-Foetoprotein was determined in 100 amniotic fluid samples (gestational age from 15 to 20 weeks).
For serum samples, the basic measuring range in a standard predilution of 1/5 covers α-foetoprotein concentrations from 5 to 160 μg/l. Serum samples with an afoetoprotein concentration exceeding the initial measuring range are reanalysed by the analyser after a further automatic sample dilution up to 1 : 40000.
In parallel, serum samples from 65 inpatients (36 men, 29 women; age: 44.2 ± 19.8 years) were obtained. According to diagnosis, these patients could be divided into those suffering from hepatocellular carcinoma (n = 10), liver cirrhosis (n = 14), chronic viral hepatitis (n = 5) or other non-malignant liver diseases (n = 19), or teratoma (n = 5) or yolk sac tumours (n = 7), or pregnant women (n = 5). Serum and amniotic fluid samples were stored at —20 °C until analysis.
For amniotic fluid samples, a standard predilution of 1 : 2000 is used, resulting in an assay range of 2000 to 64000 μg/l. Amniotic fluid samples with an α-foetoprotein concentration exceeding the initial measuring range are reanalysed by the analyser after a further automatic sample dilution up to 1 : 40 000.
Procedures followed were in accordance with the Helsinki declaration of 1975, as revised in 1983, and in accordance with the guidelines for research of our institute.
In this way, we found the standard curve of the nephelometric method to be linear from 5 to 86 800 μg/l. Highdose "hook" effects did not occur in the tested concentration range (α-foetoprotein concentrations up to 273 000
Subjects
Statistical e v a l u a t i o n Results are expressed as mean ± SD. Comparison between methods was carried out using Pearson's correlation coefficients after logarithmic transformation of the data when appropriate.
Results Test performance characteristics Stability of calibration and reagents No significant changes in α-foetoprotein concentration were observed when reanalysing forty serum samples with the initial calibration curve and with reconstituted reagents after 2 and 7 days of storage. Imprecision Intra-assay coefficients of variation (CV) for serum and amniotic fluid samples were between 2.3% and 4.0%. The inter-assay CVs for serum and amniotic fluid were between 3.5% and 4.6%. Table 1 summarizes intra- and inter-assay CVs for serum and amniotic fluid pools used. Tab. 1 Reproducibility of nephelometric α-foetoprotein determination.
Sl a
S2
Al
A2
Intra-assay imprecision (n = 10) Mean (μ§/1) 15.7 SD ^g/I) 0.4 CV (%) 2.5
150 6 4.0
13200
20100
Inter-assay imprecision (n = 10) Mean (μ^Ι) 15.9 SD fogfl) 0.7 CV (%) 4.4
143 5 3.5
13600
20000
a
300 2.3
540 4.0
560 2.8
910 4.6
SI, S2: different serum pools; AI, A2: different amniotic fluid pools.
Interferences No interference from haemoglobin (up to a final concentration of 0.15 mmol/1) could be detected for the nephelometric method. Addition of bilirubin (final concentrations up to 268 μπιοΐ/ΐ), rheumatoid factor (final concentrations up to 470 int. units/1) or serum containing human anti-mouse antibodies (final concentrations up to a titre of 1/1000) to serum pools with increased α-foetoprotein did not interfere with the nephelometric assay. Effect of hypertriglyceridaemia on test results depended on the size of the triacylglycerol containing particles. Effects were negligible for VLDL-triacylglycerols, up to triacylglycerol concentrations of 6.0 mmol/1. In contrast, in samples containing chylomicrons, triacylglycerol concentrations of 1 .0 mmol/1 induced a decrease of the α-foetoprotein concentration measured. The presence of interfering chylomicrons can be detected by an increased initial sample turbidity. Correlation studies Sixty-five serum samples were evaluated by both immunonephelometry and RIA. For 29 serum samples yielding a result smaller than the lowest calibration point (5 μg/l) of the nephelometric method, the α-foetoprotein concentration was determined by linear interpolation using raw measurement data (bit value). Both methods correlated well: y (log AFP-Nephelometry, μg/l) = 1.046 χ (log AFP-RIA, μ^) - 0.128 (r = 0.987, n = 65, Syx = 0.212) (fig. 2). With regard to the patients' clinical diagnosis, the correlations were good for patients with liver pathology: y (log AFP-Nephelometry, lig/ΐ) = 1.040 χ (log AFP-RIA, μ§/1) - 0.125 (r = 0.984, n = 48, Syx = 0.212), patients presenting with
Bernard et aL: Nefehelometric α-foetoprotein determination
850 yolk sac tumours: y (log AFP-Nephelometry, μg/l) = 1.055 χ (log AFP-RIA, μ§/1) - 0.133 (r = 0.989, n = 12, Syx = 0.269) and pregnant women: y (log AFPNephelometry, μg/l) = 0.993 χ (log AFP-RIA, μg/l) - 0.024 (r = 0.987, n = 5; Syx = 0.097). A similarly good correlation was obtained when comparing α-foetoprotein concentrations in amniotic fluid samples determined by both methods: y (log AFP-Nephelometry, μg/l) = 0.995 χ (log AFP-RIA, μ^) - 0,011 (r = 0.949, n = 100, Syx = 0.094) (fig. 3). ct-Foetoprotein heterogeneity Concanavalin A separation Four serum samples with pathologic α-foetoprotein concentrations and four amniotic fluid samples were separated by affinity chromatography on concanavalin A bound to Sepharose 4B. The detection of the concanavalin A-non-reactive α-foetoprotein fraction was comparable with both methods: y (concanavalin A-nonreactive fraction, Nephelometry) = 0.90 χ (concanavalin A-non-reactive fraction, RIA) -l· 0.03 (r = 0.88, n = 8, Syx = 0 . 1 5 ) .
10° 105 1 1 104
ff
103
10°
100
101
102
1Q3
104
a- Foetoprotein in serum (RIA) [pg/l]
105
106
Dimeric a-foetoprotein forms The α-foetoprotein fractions with a mean apparent A/r of 70000 and 140000 were assayed in parallel using both methods. In fresh amniotic fluid samples the number of dimeric α-foetoprotein molecules is low and increases during prolonged storage. Recovery for both the A/r 70000 fraction and Mr 140000 fractions was nearly identical for both nephelometry and RIA: monomeric afoetoprotein: y (α-foetoprotein, Nephelometry, μg/l) = 0.95 χ (α-foetoprotein, RIA, μg/l) + 2.89 (r = 0.98, n = 10, Syx = 3.09); and for the'dimeric form: y (afoetoprotein, Nephelometry, μg/I) = 1.03 χ (a-foetoprotein, RIA, μβ/1) + 1.81 (r = 0.98, n = 10, S^ = 27.3). After desaggregating α-foetoprotein multimers by treating the dimeric fractions with 8 mol/1 urea, no significant increase of the α-foetoprotein concentration was observed.
Discussion The new nephelometric latex α-foetoprotein test allows fast and convenient α-foetoprotein determinations of high analytical quality, with use of small amounts of sample (80 μΐ). Intra- and inter-assay CV's are within acceptable limits and comparable with, or lower than, those observed for RIA and other related immunological methods (2, 6, 15, 16, 19, 20). Samples with a high α-foetoprotein concentration (serum concentration > 160 μg/l and amniotic fluid concentration > 64 000 μg/l) can be rerun automatically by the instrument from a higher sample predilution, which is an advantage with respect to many other techniques. The pre-incubation reaction allows the early detection of extremely high α-foetoprotein concentrations, eliminating the possibility of a high dose "hook" effect.
Fig. 2 Correlation of serum α-foetoprotein concentrations measured by RIA (AFP RIACT) and immunonephelometric assay Interferences resulting from the presence of rheuma(Behring N latex AFP). Data on x-axis and y-axis are given in toid factors, haemoglobin, bilirubin, and human antilogarithmic scales. mouse antibodies are negligible under routine conditions. Although monoclonal antibodies of murine oriS 10, r gin are used, no interference was observed from human anti-mouse antibodies (23). Interference resulting from triacylglycerol-containing particles due to light scattering largely depends on their size (24). As a I? 104 consequence, the presence of post-prandial chyloll ·— ο fl) — •Χ ω microns (particle size: 200-1000 nm) already interferes at a critical triacylglycerol concentration of 1.0 mmol/1. When triacylglycerols are present as VLDL fraction (particle size: 30-200 nm), no inferf103 erence is observed up to triacylglycerol concentrations 103 105 α-Foetoprotein in amniotic fluid (RIA) [pg/l] of 6.0 mmol/1. This particle size related interference has been observed for other latex enhanced immunoFig. 3 Correlation of amniotic fluid α-foetoprotein concentrations chemical techniques as well (25, 26). For serum sammeasured by RJA (AFP RIACT) and immunonephelometric assay (Behring N latex AFP). Data on x-axis and y-axis are given in ples with a turbid appearance, performing the test in logarithmic scales. a 1 :20 sample dilution is recommended in order to
is It
851
Bernard et al.: Nephelometric α-foetoprotein determination
reduce interference. Amniotic fluid samples are not affected by interferences of bilirubin and haemoglobin due to the high sample dilution used (1 :2000). α-Foetoprotein values obtained by immunonephelometry are comparable to those obtained by RIA both for serum and amniotic fluid samples. Although the matrix effects of amniotic fluid and serum samples are clearly different, we could not detect any significant difference in the slope of the regression lines between nephelometry and RIA for either amniotic fluid or serum samples. Although a difference in detection of α-foetoprotein due to the variation of its carbohydrate moiety has been reported (2), no difference in the detection of either the concanavalin A reactive or concanavalin A non-reactive fraction could be demonstrated between RIA and nephelometry. This is in accordance with previous reports that the development of specific antibodies to concanavalin A reactive and concanavalin A non-reactive a-foetoprotein was unsuccessful (27).
The relative amount of α-foetoprotein dimers in fresh serum and amniotic fluid samples is low (13, 14), so that our observed differences are negligible when interpreting α-foetoprotein test results. Moreover, no differences in test results were obtained in isolated α-foetoprotein aggregates between both methods. Multimeric afoetoprotein is known to be detected less efficiently by double antibody RIA assays by increasing the non-specific binding (21), but did not influence our test results. Desaggregating α-foetoprotein multimers by treatment of the sample with 8 mol/1 urea did not result in significant increases of α-foetoprotein concentrations. In conclusion, irnmunonephelometric determination of α-foetoprotein in both serum and amniotic fluid is a fast, convenient and reliable method, appropriate for quantification of α-foetoprotein in the routine laboratory. Acknowledgements We thank Dr. M. Lammers and Mr. R Loots for their helpful discussions and Behringwerke for providing us with reagent.
References 1. Bock JL. Current issues in maternal serum alpha-fetoprotein screening [review]. Am J Clin Pathol 1992; 97:541-54. 2. Chan DW, Kelsten M, Rock R, Bruzek D. Evaluation of a monoclonal immunoenzymometric assay for alpha fetoprotein. ClinChem 1986; 32:1318-22. 3. Lamerz R. Tumormarker: AFP (Alpha-Fetoprotein). In: Thomas L, editor. Labor und Diagnose. 4th ed. Marburg: Die Medizinische Verlagsgesellschaft, 1992:1153-62. 4. Kelsten ML, Chan DW, Bruzek DJ, Rock RC. Monitoring hepatocellular carcinoma by using a monoclonal immunoenzymometric assay for alpha-fetoprotein. Clin Chem 1988; 34:7681. 5. Barlett NL, Freiha FS, Torti FM. Serum markers in germ cell neoplasms. Hematol Oncol Clin North Am 1991; 115:623-38. 6. Christensen RL, Rea MR, Kessler G, Crane JP, Valdes R Jr. Implementation of a screening program for diagnosing open neural tube defects: selection, evaluation, and utilisation of alpha-fetoprotein. Clin Chem 1986; 32:1812-7. 7. Ashwood ER. Clinical chemistry of pregnancy. In: Burtis CA, Ashwood ER, editors. Tietz Textbook of Clinical Chemistry. 2nd ed. Philadelphia: WB Saunders, 1994: 2107-48. 8. Smith CJ, Kelleher PC, Belanger L, Dallaire L. Reactivity of amniotic fluid alpha-fetoprotein with concanavalin A in diagnosis of neural tube defects. Br Med J 1979; 1:920-1. 9. Buamah P, Taylor P, Ward M. Concanavalin A binding of fetoprotein in amniotic fluids as an aid in the diagnosis of neural tube defects. Clin Chem 1981; 27:1658-60. 10. Buamah PK, Gibb 1, Bates G, Ward AM. Serum alpha fetoprotein heterogeneity as a means of differentiating between primary hepatocellular carcinoma and hepatic secondaries. Clin Chim Acta 1984; 139:313-6. 11. Buamah PK, Cornell C, Skillen AW. Affinity chromatography used in distinguishing alpha-fetoprotein in serum from patients with tumors of hepatic parenchyma and of germ cells. Clin Chem 1984; 30:1257-8. 12. Chang DW, Miao Y-C. Affinity Chromatographie separation of alpha-fetoprotein variants: development of a mini-column procedure, and application to cancer patients. Clin Chem 1986; 32:2143-6. 13. Wu JT, Waterhouse WJ. Identification of alpha-fetoprotein polymers. Artifacts of the isolation procedure. Clin Chim Acta 1982; 125:9-19.
14. Wu JT, Knight JA. In-vitro stability of human alpha-fetoprotein. Clin Chem 1985; 31:1692-7. 15. Kemp HA, Simpson JSA, Woodhead JS. Automated two-side immunoradiometric assay of human alpha-fetoprotein in maternal serum. Clin Chem 1981; 27:1388-91. 16. Villalta D, Borean M, Santini G. Evaluation of an automated commercial immunoluminometric assay for α-foetoprotein. Eur J Clin Chem Biochem 1991; 29:193-6. 17. Frengen J, Schmid R, Kierulf B, Nustad K, Paus E, Berge A, Lindmo T. Homogeneous immunofluorometric assays of afetoprotein with macroporous, monosized particles and flow cytometry. Clin Chem 1993; 39:2174-81. 18. Gadsen RH, Cate JC. Serum alpha-fetoprotein: I. Evaluation of quantitative assays to automated immunoassay systems. Ann Clin Lab Sc 1991; 21:246-53. 19. Hutton J, Carlson B, Thillen R, Watherholt J. A clinical comparison of the Axsym; AFP assay with the IMx AFP assay [abstract]. Clin Chem 1995; 41:S226. 20. Collet Casart D, Magnusson CGM, Ratcliffe JG, Cambiaso CL, Masson PL. Automated particle-counting immunoassays for alpha-fetoprotein. Clin Chem 1981; 27:64-7. 21. Young J, Reid R, Crawford J. Purification and radioimmunoassay of human alpha-1-fetoprotein: the effect of aggregates on the radioimmunoassay. Clin Chim Acta 1976; 69:11 —20. 22. Kapmeyer W, Pauly H-E, Tuengler P. Automated nephelometric immunoassays with novel shell/core particles. J Clin Lab Anal 1988; 2:76-83. 23. Kinders RJ, Hass GM. interference in immunoassays by human anti-mouse antibodies. Eur J Cancer 1990; 36:647—8. 24. Barrow G. Macromolecules. In: Barrow G, editor. Physical Chemistry. New York: McGraw Hill Book Company, 1961; 641-84. 25. Delanghe JR, Chapelle JP, Vanderschueren SC. Quantitative nephelometric assay for determining myoglobin evaluated. ClinChem 1990; 36:1675-8. 26. Delanghe J, Chapelle JP, El AJlaf M, De Buyzere M. Quantitative turbidimetric assay for determining myoglobin evaluated. Ann Clin Biochem 1991; 28:474-9. 27. Brock DJH, Barron L, van Heyningen V. Approaches to the production of monoclonal antibodies specific for concanavalin A binding and non-binding forms of alpha-fetoprotein. Tumour Biol 1984; 5:171-8.
852 Further readings from this Journal: 28. Costongs GMPJ, Janson PCW. A European quality control programme as a cooperative tool between users and a diagnostic company. Eur J Gin Chem Clin Biochem 1993; 31:851-9. 29. Kruse R, Geilenkeuser W-J, Röhle G. Intel-laboratory surveys of the determination of tumour markers scatter and repeatability of the results. Eur J Clin Chem Clin Biochem 1993; 31:139-46. 30. Bacigalupo MA. lus A, Meroni G, Farina L, lacobello C. Comparison of a time-resolved immunofluorimetric assay with two immunoenzymatic methods for -foetoprotein in human serum. Eur J Clin Chem Clin Biochem 1994: 32:729-31.
Bernard et al.: Nöphelometric -foetoprotein determination
31. van Dalen A, Kessler A-C. A multicentre evaluation of tumour marker determinations using the automatic enzymun-test® systems ES 300 and ES 600/700. Eur J Clin Chem Clin Biochem 1996: 34:377-84. Received April 25/July /, 1996 Corresponding author: Joris R. Delanghe, Laboratorium voor Klinische Biologie, Afdeling Klinische Scheikunde, Universitair Ziekenhuis Gent, De Pintelaan 185, B-9000 Gent, Belgium
Eur J Clin Chem Clin Biochem 1996; 34:853-860 © 1996 by Walter de Gruyter · Berlin · New York
Determination of Reference Method Values by Isotope Dilution-Gas Chromatography/Mass Spectrometry: A Five Years9 Experience of Two European Reference Laboratories1)2) Linda M. Thienpont**2, Benedikt Van Nieuwenhove1, Dietmar Stockt2*3, Hans Reinaner* and Andre P. De Leenheer1 1
2
3
Laboratoria voor Medische Biochemie en voor Klinische Analyse, Faculteit Farmaceutische Wetenschappen, Universiteit Gent, Gent, Belgium Laboratorium voor Analytische Chemie, Faculteit Farmaceutische Wetenschappen, Universiteit Gent, Gent, Belgium3) Institut für Standardisierung und Dokumentation im Medizinischen Laboratorium e. V. (INSTAND e. V.), Johannes-Weyer Straße l, Düsseldorf, Germany
Summary: We report on the cooperation of two European Reference Laboratories for the determination of reference method values in serum based materials intended for use in internal accuracy control and external quality assessment. Reference method values were determined by isotope dilution-gas chromatography/mass spectrometry for aldosterone, cortisol, oestradiol-17ß, progesterone, testosterone, thyroxine, theophylline, cholesterol, creatinine, glucose, total triacylglycerols and uric acid. All determinations were done in parallel in the two laboratories, independently and within certain time constraints. The general measurement design consisted of duplicate measurement of each sample on three different occasions. In each laboratory, rigorous internal quality control was performed according to predefined analytical quality specifications. This was done using certified reference materials. If not available, control materials targeted before by the two laboratories were utilized. Here we present the results of the cooperation during five years. We discuss the precision and accuracy achieved, the between-laboratory agreement and the total analytical error. For the hormones and theophylline, the mean overall coefficient of variation for both laboratories (calculated from the measurements on three days) was always < 2%, for the substrates < 1%. For all substances, the method bias (estimated from several measurement series over the five years) was < 1%, and the average deviation of the results between the two laboratories was < 1.2%. The maximum total analytical error was in all cases < 3%. These data demonstrate that current reference methodology is able to guarantee a stable level of high quality of performance, and that reference laboratories of today are capable of providing, in due time, adequate service in the framework of accuracy-based harmonization of methods in routine laboratory medicine. Introduction laboratories belonging to its organization analytical According to the guidelines of the German Bundesärz- 25 mg for the balance with a readability of 10~5 g, and > 3 mg for the one with a readability of 10~6g), In this way, the inaccuracy of the weighings was < 0.1% (5, 11). In addition, the weighing procedure was kept as short as possible, through not aiming at "round" concentrations (e. g. 1.000 g/1) and restricting the weighings to a series of maximum four at a time. Further dilutions of the stock standard solutions were performed on a gravimetric or a calibrated volumetric basis, with calculation of the concentrations to four significant digits. For each analyte, three working solutions from three different stock solutions were prepared and, before use, cross-checked against each other for accuracy, with acceptance of deviations of maximum 1% on the basis of six measurements. The stock and working solutions were stored under ideal conditions and at regular intervals freshly prepared and checked against the old ones. Rigorous control of all volumetric steps was also performed, i.e. reconstitution of lyophilized serum and sampling of serum and standard solutions (unlabelled and labelled analyte). Laboratory A controlled all sampling steps on a gravimetric basis, respecting the same criteria as described above for gravimetric preparation of the standard solutions. For gravimetric reconstitution, the deviations of the pipetted volume of water from the prescribed volume for reconstitution were taken into account, however, setting the tolerable deviation to a maximum of 0.5%. Laboratory B did sampling of standard solutions on a gravimetric basis, reconstitution and sampling of serum on a calibrated volumetric basis. The accuracy of calibrated volumetric reconstitution was better than 0.3%, therefore, no correction for deviation was necessary. The reference method value for the lyophilized control sera to analyze was always calculated from twelve
857
surement values obtained in the two laboratories for duplicates on three independent occasions. For each batch of lyophilized control material to be analyzed, three different vials were used for the three measurement occasions. This was done to compensate for the vial-to-vial variation in the dry-mass content. The effect of methodological variables during the GC/MS measurements was minimized through injecting according to the bracketing scheme and aiming at ion abundance ratios in calibration and unknown mixtures within a 20% limit. Last but not least, the methods (sample pretreatment, derivatization, m/z values for mass spectrometric monitoring) were optimized to reach good selectivity, sensitivity, accuracy and precision. Further information in this respect has been described for the individual reference methods in I.e. (3 — 11). Internal quality control (accuracy and precision) The analytical quality specifications imposed by INSTAND e. V. urged the reference laboratories to set up rigorous internal quality control rules. The latter concerned limits for the daily deviation from the accepted target (Δ^,^χ) of the materials used for accuracy control (certified CRM/SRM or own controls) and for the overall CV (CVmax) calculated for each set of results obtained on three measurement occasions. For the SRM/ CRM, the certified values with their uncertainties were used, thereby always respecting the stability as guaranteed by the certifying organization (NIST, BCR). For analytes susceptible to degradation with time (i. e. glucose, cholesterol and uric acid) the values from the most recent revision were always considered. The daily accuracy control was organized in such a way that the measurements of the unknowns only were started provided the Alar was < 3% for the hormones/theophylline, and < 2% for the substrates. In addition, fulfillment of the preset limits by INSTAND e. V. (tab. 3) was judged on the basis of the long-term calculated bias (see above). Results for measurements of independently prepared duplicates also had to agree, for the hormones/theophylline analytes within 2.5%, for the substrates within 2%. If not, both aliquots were measured again. In the case that the mean of the two measurements still differed by respectively > 2.5% and > 2%, a fourth set of parallels was prepared on a different day. The same was done when the CV obtained from the six measurements exceeded respectively 3% and 2%. In addition, a fourth measurement day was added when the overall specifications listed in table 3 were not fulfilled. Tables 4, 5 and 6 illustrate typical data for accuracy and precision obtained in the cooperation. The total number of samples measured over five years for each analyte is not always the same, because in certain sera, for some
Thicnpont et al.: Reference method values determined in two European reference laboratories
858
analytes a reference method value was not requested by the manufacturer. Table 4 shows the bias data of laboratory B. Values in the same range applied for laboratory A (not shown). For inteipretation of table 4 the uncer-
Tab. 6 Precision data for the substrates (CV calculated from 6 measurements). Analyte
n1
Coefficient of variation (%) Range
Tab. 4
Reference material or control material
Certified concentration
Cholesterol
SRM 909-al oTMk Λ f\f\f\ SRM 909-a2Ο
4.892 ± 0.061 mmol/1 Λ Λ(Λ -t- (\ (\T% mnif\\/\ 4.40.3 ^ u.u/j mmoi/i
Creatinine
SRJVl 909-al SRM 909-a2
0.084 ± 0.001 mmol/1 0.463 ± 0.006 mmol/1
Glucose
SRM 909-al SRM 909-a2
4.95 ± 0.30 mmol/11 15.41 ± 0.80 mmol/11
own2
1.33 mmol/12
- 0.5
own own
1 7.0 nmol/1 48.5 nmol/1
+ 0. 1 - 0.9
Total triacylglycerols Uric acid Aldosterone Cortisol
Oestradiol-17 Progesterone
Testosterone
Bias (%)
- 0.9 (\ *) u.z + 0.6 + 0.8
Cholesterol Creatinine Glucose Total triacylglycerols Uric acid 1
Range
Mean
Laboratory A
Laboratory B
50 38 38 50
0.4-1.7 0.2-1.9 0.2-1.3 0.2-1.7
0.3-1.5 0.4-2.0 0.4-1.9 0.2-1.8
0.8 0.9 1.0 0.9
48
0.2-1.5
0.3-1.9
0.9
Accuracy data for laboratory B.
Analyte
Mean
Coefficient of variation (%)
0.8 0.9 0.7 0.8 ' 0.8
Number of samples measured
- 0.6 tainties around the certified values of the SRM/CRM are - 0.4
also mentioned. From these figures, it can be derived that the bias never exceeded the uncertainty limits of the matrix reference materials, indicating good accuracy of SRM 909-al 0.234 ± 0.003 mmol/1 - 0.4 the reference method measurements. At the same time, SRM 909-a2 0.525 ± 0.009 mmol/1 ± 0.0 they demonstrate that the 0.86%/0.90% limit imposed o\vn 0.777 nmol/1 -0.1 by INSTAND e. V. was respected. Only for glucose the CRM 192 273 ± 6 nmol/1 +0.3 SRM 909 al/a2 do not allow adequate accuracy assessCRM 193 763 ± 14 nmol/1 + 0.5 ment, since the uncertainties account for the degradation own 469 nmol/1 + 0.6 own 770 nmol/1 + 0.6 of glucose with time and therefore are relatively high. The same applies for the use of "own control materials" own 0.955 nmol/1 -0.4 own 1.84 nmol/1 -0.2 for accuracy estimation, since their target values obCRM 347 1 0. 1 3 ± 0.2 1 nmol/1 + 0.5 tained by only two laboratories might be less reliable. CRM 348 40.3 ± 1.0 nmol/1 + 0.8 However, it is our opinion that in the absence of officown 3 1.8 nmol/1 +0.5 ially certified materials, their use is justified. own 54.1 nmol/1 — 0.4
The precision data in table 5 (for the hormones/theophylline) and in table 6 (for the substrates) represent the Thyroxine own 90.1 nmol/1 +0.1 overall CV achieved in each laboratory (mean and own 193 nmol/1 - 0.6 range). This CV consists of the combination of the beTheophylline own 82.3 μιηοΐ/ΐ + 0.3 tween day measurement CV and the dry-mass variability 1 Values from the most recent recertification (February 24, 1993). of the lyophilized samples. The mean CV never exHowever, for calculation of the deviation of each measurement ceeded 2% for the hormones/theophylline, 1% for the campaign, the certified values valid at that time were used. substrates. In addition, it could be observed that the CV 2 Own control material. The certified concentration corresponds to the mean of the reference method values determined by the two was independent from the concentrations measured. laboratories (n = 12). Only for the lowest concentrated samples was a slightly higher CV obtained. From the above data on bias and Tab. 5 Precision data for the hormones and theophylline (CV cal- overall CV, it can be concluded that the total analytical culated from 6 measurements). error preset by INSTAND e. V. was respected. It should be further noted that only in < 2% (for the hormones/ Analyte n1 Coefficient of Coefficient of theophylline), < 5% (for the substrates), did the overall variation (%) variation (%) CV exceed the predefined values of respectively 3% and Range Mean Range Mean 2%, making a fourth measurement occasion necessary. Laboratory A
Aldosterone Cortisol Oestradiol-17 Progesterone Testosterone Thyroxine Theophylline 1
16 31 23 26 33 28 21
0.8-2.3 0.4-1.8 0.5-2.4 0.4-2.4 0.3-1.2 0.7-2.0 0.2-2.9
Number of samples measured
.9 .0 .0 .4 ().7 .2 .4
Laboratory B 0.7-2.4 0.5-1.7 0.5-1.8 0.4-2.3 0.3-1.7 0.5-2.4 0.4-1.5
.6 .2 .2 .3 .2 .6 (18
Between-laboratory agreement The between-laboratory agreement was derived from the difference plots shown in figures 1 and 2. The deviations were calculated from the difference of the results of laboratory B from those of laboratory A, expressed in % of the mean result. A visual inspection of the figures reveals that for ail analytes, the deviation of the results is
859
Thienpont et al.: Reference method values determined in two European reference laboratories
(confidence interval 99.7%, α = 0.05) would have resulted.
4-
Conclusion The analytical performance described here, achieved by two European reference laboratories over a period of five years, proves that current reference methodology is able to guarantee a stable level of high quality. This is primarily due to the use of officially accepted analytical measurement principles as the basis of the reference methods. Other prerequisites are that traceability of the measurements to the true value is realized by the use of
Ικ
0 0 °
0 - αCM°°oa α α α ο σ ^ .£ -4 α -R 1.0
2.0
0.0
Aldosterone [nmol/l]
0.4
0.8
OQ
0 - SrfD 0°
D
o
^3"* Q
>ο
2
0.0
4 -? |1
c
-4 -
2
R .
1.2
Q
41 0 -
i:·..·.·
c -4 -
'm
