Comparison of Three Different Plasma Homocysteine Assays with Gas ...

Clinical Chemistry 45:5 670 – 675 (1999)

Endocrinology and Metabolism

Comparison of Three Different Plasma Homocysteine Assays with Gas Chromatography– Mass Spectrometry Johan B. Ubbink,1* Rhena Delport,1 Reiner Riezler,2 and W.J. Hayward Vermaak1 Increased plasma total homocyst(e)ine3 (tHcy),4 measured either when the subject is fasting or after an oral methionine load test, has been implicated as an independent risk factor for premature vascular disease (1–3 ). A recent metaanalysis of 27 studies estimated that a 5 mmol/L increase in circulating tHcy concentrations increased the risk for coronary heart disease by a factor of 1.7 and the risk for cerebrovascular disease by a factor of 1.5 (4 ). Although a relationship between reductions in plasma tHcy concentrations and a reduced incidence of premature vascular disease has not yet been established, some experts nevertheless recommend the inclusion of plasma tHcy determinations in assessing individual risk profiles (5 ), particularly in patients with existing vascular disease or with a strong family history of premature vascular disorders (6 ). Increasingly, laboratories offer the plasma tHcy assay, and various analytical methods have become available to measure this amino acid. Chromatographic methods have been the methods of choice to determine plasma tHcy concentrations. Most laboratories have used methods based on the derivatization of homocysteine with thiol-specific reagents such as monobromobimane (7 ) or ammonium 7-fluorobenzo-2oxa-1,3-diazole-4-sulfonate (SBDF) (8 ), or by o-phthaldialdehyde derivatization of the primary amine group (9 ). The fluorescent Hcy adduct is then separated from other thiol-containing compounds by HPLC and quantified by fluorescence detection. HPLC methods based on SBDF derivatization seem to be the most popular; data from the European Quality Assessment Scheme for Special Assays in Serum and Urine indicate that 38 of 90 participating

Background: Various methods are available to measure plasma total homocyst(e)ine (tHcy) concentrations, but whether plasma tHcy assays may be used interchangeably is not known. Methods: Results from three different methods [HPLC with fluorescence detection, enzyme immunoassay (EIA), and fluorescence polarization immunoassay (FPIA)] to determine fasting (n 5 163) and post-methionine load (n 5 80) plasma tHcy concentrations were compared with those obtained by gas chromatography– mass spectrometry (GC-MS). Difference plots on nontransformed and log-transformed data were used to assess the agreement between HPLC and GC-MS, EIA and GC-MS, and FPIA and GC-MS. Results: The closest agreement between methods was observed between GC-MS and FPIA for fasting tHcy concentrations, with 95% of the FPIA values between 19% above and 24% below the corresponding GC-MS results. Post-methionine load tHcy concentrations measured by EIA showed the least agreement with GC-MS, with 95% of values measured by EIA ranging between 52% above and 16% below the GC-MS values. With respect to GC-MS, the above-mentioned methods showed a negative bias for fasting tHcy concentrations, but a positive bias for both immunoassays for postmethionine load tHcy concentrations. Conclusions: The agreement among methods is insufficient to allow them to be used interchangeably. The intermethod differences emphasize the need for standardization of plasma tHcy assays. © 1999 American Association for Clinical Chemistry

3 Total homocyst(e)ine refers to the sum of the concentrations of free homocysteine, protein-bound homocysteine, the disulfide homocystine, and the mixed disulfide homocysteine-cysteine.

1 Department of Chemical Pathology, University of Pretoria, P.O. Box 2034, Pretoria 0001, South Africa. 2 Severimed, Wiedaustrasse 202, 48163 Mu¨nster, Germany. *Author for correspondence. Fax 27-12-3283600; e-mail jubbink@ medic.up.ac.za. Received December 14, 1998; accepted March 3, 1999.

4 Nonstandard abbreviations: tHcy, total homocyst(e)ine; SBDF, ammonium 7-fluorobenzo-2-oxa-1,3-diazole-4-sulfonate; GC-MS, gas chromatography–mass spectrometry; EIA, enzyme immunoassay; and FPIA, fluorescence polarization immunoassay.

670

671

Clinical Chemistry 45, No. 5, 1999

laboratories used derivatization with SBDF to measure serum tHcy concentrations (10 ). Gas chromatographic methods to measure tHcy have also been described (11 ), and a few specialist centers measure tHcy by gas chromatography–mass spectrometry (GC-MS) (12 ). The chromatographic methods mentioned above require sophisticated and expensive equipment that generally is not available at routine clinical chemistry laboratories. Furthermore, few chromatographic methods are fully automated, and in all cases, the daily output is relatively low. In contrast, immunoassays usually lend themselves to full automation and also have the potential of a high daily throughput. Therefore, the recently described enzyme-linked immunoassay (EIA) (13 ) and fluorescence polarization immunoassay (FPIA) methods (14 ) for plasma tHcy may become popular with laboratories that offer this assay. Both EIA and FPIA rely on enzymatic conversion of homocysteine to S-adenosylhomocysteine, which is subsequently detected by a competitive immunoassay. Both methods have become commercially available. In this study, we compared a GC-MS method for plasma tHcy with an HPLC method based on thiol derivatization with SBDF, an EIA method, and an FPIA method. This between-method comparison reveals that the above-mentioned methods cannot be used interchangeably and emphasizes the need for standardization of plasma tHcy assays.

and the FPIA were calibrated against calibrators supplied by the respective reagent kit manufacturers. Control specimens were analyzed in each batch, and results were only accepted when the control values were within the range specified by the manufacturer. Controls prepared in house were used for both the GC-MS and the HPLC methods.

statistical methods Difference plots were used to assess the agreement between tHcy results obtained with GC-MS vs HPLC, EIA, and FPIA, respectively (16, 17 ). Fasting and post-methionine load tHcy concentrations were assessed separately. Possible systematic biases between GC-MS vs HPLC, GC-MS vs EIA, and GC-MS vs FPIA, respectively, were assessed by computing the 95% confidence intervals for the mean differences between GC-MS and each of the methods mentioned above. In a subsequent analysis, the data from each assay were log transformed. Mean differences, as well as the limits of agreement (mean difference 6 2 SD), were calculated on the log-transformed data for GC-MS vs HPLC, GC-MS vs EIA, and GC-MS vs FPIA. Antilogs of the mean differences were calculated to assess the mean proportional bias of each method with respect to GC-MS. Antilogs of the limits of agreement were calculated to express these limits as ratios of GC-MS results vs HPLC, EIA and FPIA results, respectively (16, 17 ).

Materials and Methods laboratory analyses In a recent study to determine the effect of vitamin B6 status on homocysteine metabolism (15 ), plasma samples were analyzed for tHcy by the HPLC method of Ubbink et al. (8 ), as well as by the GC-MS method of Allen et al. (12 ). These analyses were performed toward the end of 1995, and aliquots of plasma samples were also stored frozen at 270 °C. Early in 1998, a selection of plasma samples obtained when the participants were fasting (n 5 163) and 6 h after an oral methionine load test (100 mg of lmethionine per kg of body weight; n 5 80) were analyzed in duplicate for tHcy concentrations using the EIA-kit AXH00001 supplied by Axis Biochemicals (Gru¨nerløkka). Using this assay, we performed dilutions manually, but used a Minilab Washer PW40 (Sanofi Diagnostics Pasteur) for the wash steps. The product formed by the enzyme reaction in the assay was measured by a Bio-Rad model 3550 microplate reader. The same samples were also analyzed in duplicate for tHcy by FPIA using reagent kit B3D390E (Abbott Laboratories, Abbott Park, IL) and the Abbott IMx System. For both immunoassays, samples with a tHcy concentration .50 mmol/L were diluted in the appropriate buffer solution and the assay was repeated. The GC-MS and the HPLC assays were calibrated against calibrators prepared independently from crystalline l-homocystine and d,l-homocysteine, respectively. Both were obtained from Sigma Chemicals. Both the EIA

Results The day-to-day variation as observed for the four different plasma tHcy assays is summarized in Table 1, and a comparison of the results obtained by these methods is shown in Table 2. Scatter plots of observed measurement differences (16 ) for fasting tHcy determinations against the mean of GC-MS and the method used in the assay are shown in Fig. 1; Fig. 2 is similar to Fig. 1, except that the scatter plots are derived from plasma tHcy concentrations obtained after methionine loading. The mean (SD) differences between GC-MS and HPLC, GC-MS and EIA, and GC-MS and FPIA are reported in Table 3. Using the standard errors of the mean differences, the 95% confidence intervals were computed; these showed a negative bias for HPLC, EIA, and FPIA with respect to GC-MS when fasting tHcy concentrations were compared. This negative bias was the smallest for FPIA. For post-methionine load tHcy concentrations, HPLC showed a negative bias simiTable 1. Between-day analytical imprecision for plasma tHcy methods. Analytical method

GC-MS HPLC EIA FPIA

n

CV, %

20 11 9 19

8.5 8.4 6.9 4.5

672

Ubbink et al.: Plasma Homocysteine Assays Compared with GC-MS

Table 2. Comparison of results obtained for plasma tHcy. Analytical method

Mean

Fasting tHcy concentrations, mmol/L (n 5 163) GC-MS 10.2 HPLC 8.7 EIA 9.0 FPIA 9.7 Post-methionine load tHcy concentrations, mmol/L (n 5 80) GC-MS 30.8 HPLC 26.0 EIA 35.4 FPIA 32.9

lar to the bias observed for fasting tHcy concentrations; however, both immunoassays now demonstrated a positive bias (Table 3). The central 0.95 interval (mean of the differences 6 2 SD) gives an indication of the agreement between GC-MS and the other methods used to measure plasma tHcy concentrations (16 ). Using this approach, FPIA agreed the best with GC-MS for both fasting and post-methionine load tHcy concentrations (Table 3). Both HPLC and EIA displayed a relatively wider scatter of difference data points. However, this comparison may not be totally appropriate because there appears to be a relationship between the difference and the mean for at least certain scatter plots (HPLC vs GC-MS and EIA vs GC-MS). The data, therefore, were log transformed, and the mean and SD values from the log-transformed data set were used to calculate the limits of agreement (with respect to GC-MS) as described by Bland and Altman (16 ). The limits of agreement were then anti-logged and expressed as intervals (ranges of percentages) by which 95% of the tHcy determinations measured by HPLC, FPIA, and EIA, respectively, were expected to differ from GC-MS (Table 3). When GC-MS vs FPIA was compared with GC-MS vs

SD

Minimum

Maximum

2.8 2.3 2.7 2.8

5.0 5.5 4.2 4.9

21.4 16.8 18.0 19.7

9.2 7.7 12.2 9.9

18.3 13.4 15.2 17.0

55.1 50.4 71.7 69.7

HPLC, it became apparent that the above-mentioned ranges of percentages for the two comparisons were virtually the same size for fasting tHcy concentrations, but that HPLC showed a larger negative mean proportional bias (the antilog of the mean difference calculated from log-transformed data) than FPIA. The agreement between EIA and GC-MS was considerably less than that found for the FPIA and HPLC. For post-methionine load tHcy concentrations, the range of percentages that contains 95% of the data points in the comparison with GC-MS was the narrowest for HPLC, followed by FPIA and then EIA. The mean proportional bias of HPLC vs GC-MS was very similar to that observed for fasting tHcy concentrations. The mean proportional biases for FPIA vs GC-MS and EIA vs GC-MS, respectively, became positive, producing a considerable increase in the upper limits of agreement when the immunological assays were compared with GC-MS (Table 3).

Discussion This study was unique inasmuch as four different methods for measuring tHcy concentrations were compared

Fig. 1. Scatter plots showing the differences between GC-MS and HPLC, FPIA, or EIA, respectively, for fasting plasma tHcy concentrations (n 5 163). tHcy concentrations are expressed in mmol/L. Dashed lines show the central 0.95 interval of the differences between HPLC, FPIA, or EIA with respect to GC-MS.

673


Fig. 2. Scatter plots showing the differences between GC-MS and HPLC, FPIA, or EIA, respectively, for post-methionine load plasma tHcy concentrations (n 5 80). tHcy concentrations are expressed in mmol/L. Dashed lines show the central 0.95 interval of the differences between HPLC, FPIA, or EIA with respect to GC-MS.

under two commonly encountered circumstances, i.e., fasting and after an oral l-methionine load of 100 mg/kg body weight. This was made possible by a clinical trial in which the effect of vitamin B6 status on homocysteine metabolism was investigated (15 ). During this trial, circulating tHcy concentrations were first measured by HPLC, and later measured again, as part of a comprehensive metabolite assay, by GC-MS according to the method of Allen et al. (12 ). The results obtained by HPLC have been reported previously (15 ). Aliquots of plasma samples were also stored frozen at 270 °C, and a selection of these samples was analyzed by the two immunoassays as these methods became available commercially. Although ;36 months had lapsed between the chromatographic assays and immunoassays, this comparison is valid because of the excellent stability of homocysteine in frozen plasma samples (18 ). The GC-MS method was chosen as “reference method” to which the other assays were compared on the basis of its well-known sensitivity and specificity.

Compared with the other methods, the FPIA assay had the lowest analytical CV, at 4.5%. Considering that the intraindividual biological variation for tHcy has been reported as 9.4% (19 ), the FPIA tHcy assay was the only one that fulfilled the criterion of an analytical CV at least 50% lower than the intraindividual CV (20 ). Universally accepted reference material for homocysteine is not yet available, which implies that none of the four methods were calibrated against appropriately certified reference materials. Both chromatographic methods used calibrators prepared independently: The GC-MS calibrator was prepared in Denver, CO; and the HPLC calibrator was prepared in Pretoria, South Africa. The differences between HPLC and GC-MS presumably are explained by differences in calibrators. Furthermore, Dudman et al. (21 ) recently showed that the SBD-Hcy adduct is light sensitive and that exposure of this compound to fluorescent laboratory light may produce lower estimates of tHcy concentrations. When we performed the HPLC assays in 1995, this was not yet known, and it is

Table 3. Difference plot analyses of results obtained with GC-MS vs HPLC, EIA, or FPIA. Variablea

Mean difference

SD

Central 0.95 interval

Fasting tHcy concentrations (n 5 163) GC-MS vs HPLC 21.42 1.29 24.00 to 1.16 GC-MS vs EIA 21.15 1.44 24.03 to 1.73 GC-MS vs FPIA 20.5 1.1 22.70 to 1.70 Post-methionine load tHcy concentrations (n 5 80) GC-MS vs HPLC 24.78 3.74 212.26 to 2.70 GC-MS vs EIA 4.62 5.42 26.22 to 15.46 GC-MS vs FPIA 2.16 3.51 24.86 to 9.18

SE

95% CIb of the mean

Mean proportional biasc

Upper limit of agreementc

Lower limit of agreementc

0.101 0.113 0.086

21.62 to 21.21 21.38 to 20.93 20.67 to 20.32

26.2% 25.4% 22.3%

10.5% 18.0% 18.6%

232.4% 234.4% 224.2%

0.418 0.606 0.392

25.62 to 23.94 3.40 to 5.83 1.38 to 2.95

27.1% 5.5% 3.0%

6.1% 52.4% 29.3%

232.6% 216.3% 211.5%

a The mean (SD) differences between GC-MS and the other tHcy assays as indicated above were calculated, and the central 0.95 interval was calculated as mean 6 2 SD. b CI, confidence interval. c Calculated from log-transformed data with GC-MS as reference method (see text). The upper and lower limits of agreement define intervals by which 95% of the tHcy concentrations measured by HPLC, EIA, and FPIA, respectively, are expected to differ from tHcy measured by GC-MS.

674

Ubbink et al.: Plasma Homocysteine Assays Compared with GC-MS

possible that destruction by light could have contributed to the lower tHcy concentrations observed with HPLC. This may be particularly true for those samples analyzed just before recalibration. It should be noted that the analytical imprecision induced by possible light destruction is incorporated in the CV for the quality-control samples (Table 1) because the quality-control samples were always inserted randomly in the daily sequence of samples awaiting HPLC analysis. The analytical CV of the HPLC method compares well with that of GC-MS, indicating that the possible effect of light destruction on the quality of the HPLC results is probably only of minor importance in explaining the negative bias of HPLC vs GC-MS. The fasting plasma tHcy concentrations measured with both immunoassays were lower than those measured with GC-MS. In contrast to HPLC, the results for both immunoassays were higher than the GC-MS results for post-methionine load tHcy concentrations. For GC-MS vs FPIA, a shift of ;11% in the 95% range of agreement was noted. For GC-MS vs EIA, the scatter of data became wider (Table 3). There may be several explanations for the deviations in the limits of agreement between the immunoassays and GC-MS after methionine loading. The independent selection of calibrators for each method, differences in linearity between methods, or the lack of standardization may explain the observed deviations in the limits of agreement. It is also possible that methionine loading increased the concentrations of a cross-reactant in the circulation. Although the mean peak plasma methionine concentrations increased 25-fold, to 0.6 mmol/L, after methionine loading in this study (15 ), the reported interference by methionine was negligible up to 5 mmol/L for both EIA and FPIA, respectively (13, 14 ). It is therefore unlikely that methionine was a cross-reactant in our study, but the appearance of other, as yet unknown, cross-reactants after methionine loading cannot be excluded. In this study, each method used its own calibrator. It may be expected that intermethod agreement will improve when aliquots of the same set of calibrators are used. However, this does not happen in practice, and our study serves as an indication of currently existing interlaboratory and intermethod differences that should be taken into account when results from different centers are interpreted. Although 95% of the fasting tHcy concentrations measured by FPIA fell in the interval of 18.6% above or 24.2% below those measured by GC-MS, this interval may still be too large to allow these methods to be used interchangeably. The interval will probably become smaller if the two methods use the same set of calibrators. Our results indicate that GC-MS and EIA assays for plasma tHcy should not be used interchangeably, and thus support a similar conclusion made by Frantzen et al. (13 ) in their comparison of EIA and HPLC. We also consider the limits of agreement between GC-MS and

HPLC too wide to allow these two methods to be used interchangeably. We conclude that FPIA and HPLC show better agreement with GC-MS than the EIA method. With the possible exception of GC-MS and FPIA for fasting tHcy concentrations, none of the methods should be used interchangeably. Certified reference material is urgently required to improve intermethod and interlaboratory agreement.

We thank L. Goddard and A. Schnell for excellent technical assistance. Axis Biochemicals ASA donated the reagent kits for EIA of tHcy, and Abbott Diagnostics Division donated the reagent kits for FPIA of tHcy. GC-MS analyses were performed by Robert H. Allen and Sally P. Stabler, Division of Hematology, University of Colorado Health Center, Denver, CO.

References 1. Arnesen E, Refsum H, Bønaa KH, Ueland PM, Førde OH, Nordrehaug JE. Serum total homocysteine and coronary heart disease. Int J Epidemiol 1995;24:704 –9. 2. Graham IM, Daly LE, Refsum HM, Robinson K, Brattstro¨m LE, Ueland PM, et al. Plasma homocysteine as a risk factor for vascular disease. The European Concerted Action Project. JAMA 1997;277:1775– 81. 3. Malinow MR, Ducimetiere P, Luc G, Evans AE, Arveiler D, Cambien F, et al. Plasma homocyst(e)ine levels and graded risk for myocardial infarction: findings in two populations at contrasting risk for coronary heart disease. Atherosclerosis 1996;126:27–34. 4. Boushey CJ, Beresford SAA, Omenn GS, Motulsky AG. A quantitative assessment of plasma homocysteine as a risk factor for vascular disease. JAMA 1995;274:1049 –57. 5. Nakayama T, Esaki N, Sugie K, Beresov TT, Tanaka H, Soda K. Purification of bacterial L-methionine-lyase. Anal Biochem 1984; 138:421– 4. 6. Stampfer MJ, Malinow R, Willet WC, Newcomer LM, Upson B, Ullmann D, et al. A prospective study of plasma homocyst(e)ine and risk of myocardial infarction in US physicians. JAMA 1992; 268:877– 81. 7. Jacobsen DW, Gatautis VJ, Green R, Robinson K, Savon SR, Secic M, et al. Rapid HPLC determination of total homocysteine and other thiols in serum and plasma: sex differences and correlation with cobalamin and folate concentrations in healthy subjects. Clin Chem 1994;40:873– 81. 8. Ubbink JB, Vermaak WJH, Bissbort S. Rapid high-performance liquid chromatographic assay for total homocysteine levels in human serum. J Chromatogr 1991;565:441– 6. 9. Hyland K, Bottiglieri T. Measurement of total plasma and cerebrospinal fluid homocysteine by fluorescence following high-performance liquid chromatography and precolumn derivatization with o-phthaldialdehyde. J Chromatogr 1992;579:55– 62. 10. European Quality Assessment Scheme for Special Assays in Serum and Urine. ERNDIM report on samples SA serum 23 and 24. September 1998:5 pp. 11. Kataoka H, Takagi K, Makita M. Determination of total plasma homocysteine and related aminothiols by gas chromatography with flame photometric detection. J Chromatogr 1995;664: 421–5.


12. Allen RH, Stabler SP, Savage DG, Lindenbaum J. Metabolic abnormalities in cobalamin (vitamin B-12) and folate deficiency. FASEB J 1993;7:1344 –53. 13. Frantzen F, Faaren AL, Alfheim I, Nordhei AK. Enzyme conversion immunoassay for determining total homocysteine in plasma or serum. Clin Chem 1998;44:311– 6. 14. Shipchandler MT, Moore EG. Rapid, fully automated measurement of plasma homocysteine with the Abbott IMx Analyzer. Clin Chem 1995;41:991– 4. 15. Ubbink JB, Van der Merwe A, Delport R, Allen RH, Stabler SP, Riezler R, et al. The effect of a subnormal vitamin B-6 status on homocysteine metabolism. J Clin Investig 1996;98:177– 84. 16. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;i:307–10.

675

17. Holis S. Analysis of method comparison studies. Ann Clin Biochem 1996;33:1– 4. 18. Ueland PM, Refsum H, Stabler SP, Malinow MR, Andersson A, Allen RH. Total homocysteine in plasma or serum. Methods and clinical applications. Clin Chem 1993;39:1764 –79. 19. Cobbaert C, Arentsen JC, Mulder P, Hoogerbrugge N, Lindemans J. Significance of various parameters derived from biological variability of lipoprotein(a), homocysteine, cysteine, and total antioxidant status. Clin Chem 1997;43:1958 – 64. 20. Fraser CG, Harris EK. Generation and application of data on biological variation in clinical chemistry. CRC Lab Sci 1989;27: 409 –37. 21. Dudman NPB, Guo XW, Crooks R, Xie L, Silberberg JS. Assay of plasma homocysteine: light sensitivity of the fluorescent 7-benzo2-oxa-1,3-diazole-4-sulfonic acid derivative, and use of appropriate calibrators. Clin Chem 1996;42:2028 –32.