- Email: [email protected]
Comparing the ESA HPLC total homocysteine assay with electrochemical detection to the CDC in-house HPLC assay with fluorescence detection
Clinica Chimica Acta 340 (2004) 195 – 200 www.elsevier.com/locate/clinchim
Comparing the ESA HPLC total homocysteine assay with electrochemical detection to the CDC in-house HPLC assay with fluorescence detection Ming Zhang, Christine M. Pfeiffer * Centers for Disease Control and Prevention, 4770 Buford Hwy NE, Atlanta, GA 30341-3724, USA Received 11 September 2003; received in revised form 24 October 2003; accepted 30 October 2003
Abstract Background: Environmental science has developed a simple high performance liquid chromatography (HPLC) assay with electrochemical detection for total homocysteine (tHcy) measurement that does not require derivatization of free thiols. We evaluated this method and compared it with the CDC HPLC assay with fluorescence detection (FD). Methods: tHcy is measured after reduction of disulfides/protein-bound thiols and protein precipitation using four channels of an ESA CoulArray detector. Lhomocystine is used as calibrator, penicillamine as internal standard. Results: Aqueous calibration of the ESA assay resulted in overestimation of tHcy by f 30% compared to the HPLC-FD method. Calibration in plasma alleviated the matrix effect. The within- (n = 3) and between-run (n = 20) imprecision was < 6%, the linearity up to 100 Amol/l was excellent, and the recovery of tHcy added to plasma was nearly complete (98.7% F 2.3%). Good correlation was observed between both methods for 266 plasma samples. The ESA assay showed a minimal negative bias of 0.28 Amol/l (3.3%). Conclusion: The ESA tHcy assay performed well in terms of accuracy and precision, and showed good agreement with the CDC HPLC-FD assay when calibrated in plasma. The major advantage of this assay is that it does not require sample derivatization. Disadvantages include instability of the prepared samples for prolonged storage and matrix effects. Published by Elsevier B.V. Keywords: CoulArray detector; Porous graphite electrode; Method comparison; Matrix effect; Cardiovascular disease
1. Introduction Determination of total homocysteine (tHcy) in plasma has become an important diagnostic procedure in clinical chemistry due to strong evidence supporting the hypotheses that increased homocysteine is an independent determinant of vascular disease and that * Corresponding author. Tel.: +1-770-488-7926; fax: +1-770488-4139. E-mail address: [email protected] (C.M. Pfeiffer). 0009-8981/$ - see front matter. Published by Elsevier B.V. doi:10.1016/j.cccn.2003.10.020
hyperhomocysteinemia often may be amenable to homocysteine-lowering treatment with B vitamins [1]. Recent studies have suggested that increased tHcy may play a role in the development of vascular complications of diabetes [2], and is a strong, independent risk factor for the development of dementia and Alzheimer’s disease [3,4]. Several types of methods are now available to determine tHcy in plasma [5]. Chromatographic methods, especially with fluorometric detection (FD), are still widely used, and immu-
196
M. Zhang, C.M. Pfeiffer / Clinica Chimica Acta 340 (2004) 195–200
noassays have recently become broadly employed because of full automation. HPLC with electrochemical detection (ECD) is also used and has the major advantage of requiring no derivatization of thiols before detection [6,7]. ESA (Chelmsford, MA) has developed a simple method to determine plasma tHcy by HPLC with coulometric ECD on porous graphite electrodes. This type of probe can be used to measure thiols for extended periods before any electrode maintenance is required [8]. The main purpose of this study was to evaluate this new ESA assay. A second purpose was to compare it to a widely used assay that has been previously compared to several other methods [9]. We thus used our in-house HPLC-FD assay as a comparison method [10].
The ESA application note specifies using an aqueous six-point calibration curve (0, 6.25, 12.5, 25, 50, and 100 Amol/l free thiol) with L-homocystine as calibrator. The aqueous calibrators consist of 100 Al K2EDTA, 100 Al water, and 100 Al calibrator. We also evaluated the use of plasma-based calibration to better simulate the composition of the patient samples. The plasma calibrators consist of 100 Al K2EDTA, 100 Al EDTA plasma pool, and 100 Al aqueous calibrator. Peak area ratios are used to calculate tHcy concentrations. In the case of plasma-based calibration, the endogenous amount of tHcy needs to be subtracted, i.e., the peak area ratio tHcy/IS for the plasma calibrator to which 0 Amol/l thiol was added needs to be subtracted from each plasma calibrator to which increasing amounts of thiols were added before plotting the calibration curve.
2. Materials and methods 2.2. CDC HPLC-FD assay 2.1. ESA HPLC-ECD assay The ESA tHcy assay is not a kit assay but rather an application note published by ESA describing a procedure that can be performed using an ESA eightchannel CoulArray detector (Model 5600) and the CoulArray Windows 32 software. Of the eight channels, only four channels are used for the tHcy assay (at potentials 520, 750, 800, and 820 mV). Any isocratic HPLC system can be used. Our HPLC system comprised two ESA pumps (Model 580) and an ESA autosampler (Model 542). Chromatography was performed on an ESA HR-80 C18 analytical column (4.6 80 mm, 3 Am) in conjunction with a C18 guard column using 0.15 mol/l sodium dihydrogen phosphate, 1.0 mmol/l sodium dodecyl sulfate (SDS), 10% acetonitrile, pH 2.8 as mobile phase at a flow rate of 1 ml/min and a column temperature of 36 jC. Sample preparation is performed by adding 100 Al K 2 EDTA (2 mg/ml), 100 Al water, and 25 Al penicillamine as internal standard (IS) (0.2 mmol/l) to 100 Al EDTA plasma. The mixture of disulfides, mixed disulfides, and protein-bound thiols is reduced by incubating at room temperature for 30 min with tris(2-carboxyethyl)-phosphine hydrochloride (TCEP, 25 Al, 10% w/v). Proteins are precipitated with perchloric acid (HClO4, 500 Al, 0.3 mol/l). After centrifugation (10,000 rpm, 10 min), the supernatant is injected onto the HPLC column for analysis (20 Al, 15-min runtime).
The CDC tHcy HPLC assay is a modification of the method of Vester and Rasmussen [11] and Gilfix et al. [12] and uses TCEP to reduce protein-bound and oxidized thiols, trichloroacetic acid to precipitate proteins, and ammonium 7-fluorobenzo-2-oxa-1,3-diazole-4-sulfonate (SBD-F) to derivatize reduced thiols for fluorescent detection. L-homocystine and L-cysteine are used as calibrators and cystamine dihydrochloride (disulfide) is used as IS. The separation of the obtained thiol-derivatives is performed isocratically within 7 min, with the use of an acetate mobile phase at pH 5.5. Recoveries of thiols added to EDTA plasma at different concentrations are nearly complete (98.7 F 2.5%). Within-run imprecision is between 1.1% and 1.8%; between-run imprecision is between 2.4% and 6.7% [10]. 2.3. Method evaluation experiments Three levels of EDTA plasma QC pools were analyzed by the ESA assay with aqueous and plasmabased calibration over 16 and 20 days, respectively, to assess between-run imprecision and the influence of plasma matrix on the calibration. The same EDTA plasma QC pools were also analyzed in three replicates each for within-run imprecision. These were the same QC pools used for our in-house HPLC-FD method. They were stored at 70 jC for f 2 years. No
M. Zhang, C.M. Pfeiffer / Clinica Chimica Acta 340 (2004) 195–200
deviations from originally determined 2 S.D. limits were observed during this time period. To assess recovery, L-homocystine was added to an EDTA plasma sample with low endogenous tHcy level ( f 6 Amol/l) at seven concentrations in duplicate: 0, 3.125, 6.25, 12.5, 25, 50, and 100 Amol/l free thiol added. Recovery was calculated as the ratio between measured tHcy in each spiked plasma sample divided by expected tHcy (measured tHcy in the unspiked plasma sample + added tHcy) in each spiked plasma sample. To assess dilution linearity, an EDTA plasma sample with relatively high endogenous tHcy concentration ( f 26 Amol/l) was serially diluted with water (1:2, 1:4, 1:8, and 1:16). These tests were performed over 3 days each. To compare the ESA assay to the CDC assay, we used 266 EDTA plasma samples, a convenience subset from the National Health and Nutrition Examination Survey (NHANES) 1999+, which includes an omnibus informed consent and Human Subjects Review protocol. Samples not analyzed simultaneously with both methods were stored at 70 jC for no longer than 1 year between the assays. 2.4. Statistical analysis Where appropriate, data are presented as means, with the variation expressed as standard deviation (S.D.) and coefficient of variance (% CV). Linear regression analysis was used to construct calibration curves and to perform method comparison. Results for
197
the QC pools obtained with the two calibration types (aqueous and plasma-based) were compared by Bonferroni t-test. A p-value of < 0.05 was considered statistically significant. Bland –Altman bias plots were used to assess the agreement between the ESA and CDC assay. Possible systematic biases were assessed by computing the 95% confidence intervals for the mean differences between the two methods (mean difference F 2 S.E.M.). Limits of agreement were assessed by calculating the central 0.95 intervals (mean difference F 2 S.D.). To assess the mean proportional bias between the two methods, we calculated the ratios of the ESA and CDC results expressed as percent.
3. Results and discussion 3.1. Comparison of plasma-based and aqueous calibration Fig. 1 shows calibration curves of the ESA assay performed in water and in plasma. While the plasmabased calibration curve displays a proportional increase in peak area ratio with increasing thiol concentration, the aqueous calibration curve shows lower than expected peak area ratios. This indicates that in aqueous calibration some of the thiol is lost, possibly through unspecific binding to glassware and/or plasticware. The lower slope of the aqueous calibration curve resulted in tHcy concentrations that were f 30% higher than with plasma-based calibration
Fig. 1. ESA assay calibration curve for tHcy added to water or to EDTA plasma. Triangles represent the response in water; solid line represents linear regression for calibration in water. Squares represent the response in plasma after correcting for the response of the endogenous amount of tHcy; broken line represents linear regression for calibration in plasma.
198
M. Zhang, C.M. Pfeiffer / Clinica Chimica Acta 340 (2004) 195–200
(Table 1). Since the response of the IS was the same in aqueous and plasma-based calibration, we can conclude that penicillamine cannot correct the observed matrix effect and calibration of the ESA assay has to be performed in plasma. When the ESA assay was calibrated in plasma, it produced virtually the same results as obtained with the CDC assay calibrated in plasma (Table 1 and Fig. 3).
assay earlier: plasma calibration produced more precise results than aqueous calibration [10]. The withinand between-run CVs for the ESA assay were close to reported CVs for HPLC with electrochemical detection, i.e., 3.9% and 5.6%, respectively [13], but were slightly higher than CVs reported for fluorometric detection (mean within-run CV, 1.0%; mean between-run CV, 3.2% Ref. [5]).
3.2. Within- and between-run imprecision
3.3. Linear dynamic range, recovery, and dilution linearity
Plasma QC pools analyzed in three replicates each for within-run imprecision (Table 1) showed comparable results for aqueous and plasma calibration ( < 5%). Analyzed for between-run imprecision over 16 or 20 days, the plasma QC pools showed significantly lower imprecision for plasma calibration ( f 6%) than for aqueous calibration ( f 11%) (Table 1). A similar observation was made with the CDC Table 1 Within- and between-run imprecision of the ESA and CDC tHcy assay and comparison between aqueous and plasma-based calibration QC low
QC medium
QC high
ESA assay, within-run imprecision, aqueous calibration (n = 3) Mean tHcy F S.D. 10.2 F 0.5 14.2 F 0.1 34.3 F 0.6 (Amol/l) CV (%) 4.5 0.8 1.8 ESA assay, within-run imprecision, plasma-based calibration (n = 3) Mean tHcy F S.D. 8.2 F 0.1 11.2 F 0.5 27.3 F 0.6 (Amol/l) CV (%) 1.2 4.4 2.1 ESA assay, between-run imprecision, aqueous calibration (n = 16) Mean tHcy F S.D. 10.5 F 1.1 14.9 F 1.6 35.7 F 4.2 (Amol/l) CV (%) 10.3 10.4 11.8 ESA assay, between-run imprecision, plasma-based calibration (n = 20) Mean tHcy F S.D. 8.1 F 0.5 11.5 F 0.6 27.4 F 1.7 (Amol/l) CV (%) 6.1 5.3 6.1 CDC assay, between-run imprecision, plasma-based calibration (n = 22) Mean tHcy F S.D. 8.6 F 0.5 11.8 F 0.6 27.2 F 1.3 (Amol/l) CV (%) 5.4 5.0 4.7
The linearity of the plasma-based calibration curve with tHcy added up to 100 Amol/l was excellent (mean slope F S.D.: 0.0102 F 0.0006, mean intercept F S.D.: 0.0019 F 0.0055, mean correlation coefficient F S.D.: 0.9997 F 0.0003 over 20 days). The mean recovery ( F S.D.) of L-homocystine added to plasma was nearly complete (98.7 F 2.3% over 3 days). Serial dilution of a plasma sample with increased tHcy concentration resulted in excellent dilution linearity ( y = 0.9958x 0.1343, R2 = 0.9988 over 3 days). The mean ratio ( F S.D.) between the measured and expected tHcy concentration was 0.97 F 0.03. This indicates that while the ESA assay underlies matrix effects, dilution of plasma up to 1:16 with water still leaves sufficient matrix (plasma proteins) in the sample to avoid matrix effects seen with plain water. 3.4. Robustness of the ESA assay The retention times (RT) for tHcy and the IS varied by f 5% over 28 runs performed over 7 months. For tHcy, the mean RT ( F S.D.) was 5.95 F 0.29 min; for the IS, the mean RT ( F S.D.) was 7.81 F 0.30 min. A typically chromatogram for the low QC pool with a tHcy concentration of f 10 Amol/l is presented in Fig. 2. The HPLC guard column needed to be changed every 200 injections to avoid deterioration of chromatographic resolution and peak shapes. The ESA analytical column was used for 400 –600 injections. In comparison, we can use the HPLC column in the CDC assay for f 1500 injections. Because porous graphite electrodes have sufficient surface area, they do not require frequent electrode maintenance such as does noble metal electrodes (gold or platinum). They provide a stable sensor and have to
M. Zhang, C.M. Pfeiffer / Clinica Chimica Acta 340 (2004) 195–200
199
intensity eluting between tHcy and the IS and interfering with resolution and integration. This unknown peak seemed to derive from the IS penicillamine because it appeared in a sample that contained only reagents and the IS, but it did not appear in the reagent blank. 3.5. Comparison of the ESA and CDC assay
Fig. 2. ESA assay representative chromatogram of a plasma sample ( f 10 Amol/l tHcy).
be typically cleaned by applying a potential of 1200 mV for 5 min after 100 samples have been processed [8]. In the ESA tHcy assay, the CoulArray detector was programmed to apply a voltage of 1000 mV at the end of each injection for 10 s. This allowed continuous usage of the electrode without any maintenance for the duration of the entire study. Because the free thiols are not derivatized in the ESA assay, samples are not stable at room temperature for extended periods of time. Plasma samples need to be measured by HPLC-ECD immediately after preparation or they have to be stored refrigerated overnight or frozen ( 15 jC or colder) for a few days. In samples stored at 15 jC for 14 days, we observed the appearance of an unknown peak with very high
tHcy concentrations determined in 266 apparently healthy subjects were spanning a range of 3– 20 Amol/ l and gave good correlation between the two methods: tHcy (ESA) = 0.9533 tHcy (CDC) + 0.0711 (r2 = 0.9445). To assess the agreement between the two methods, we plotted for each sample the difference of results from the two methods ( y-axis) against the mean of results (x-axis) (Fig. 3). The ESA assay showed a slight negative bias of 0.28 Amol/l (95% confidence interval, 0.375, 0.193), corresponding to a relative bias of 3.3%. The central 0.95 interval (mean difference F 2 S.D.) indicates the agreement between the two methods. Ninety-five percent of tHcy results obtained with the ESA assay were between 1.8 Amol/ l lower and 1.2 Amol/l higher than results obtained with the CDC method. Although some of the plasma samples were stored for up to 1 year between performing the two assays, we do not believe that the storage accounts for the small bias we found. We stratified the entire sample set by length of storage and did not find any differences in bias whether samples were only
Fig. 3. Bland – Altman bias plot showing the difference between the ESA HPLC-ECD and the CDC HPLC-FD assay. Center line represents the mean difference ( 0.28 Amol/l), the bottom and top lines represent the limits of agreement corresponding to the mean difference F2 S.D. ( 1.77, 1.20 Amol/l); n=266 subjects.
200
M. Zhang, C.M. Pfeiffer / Clinica Chimica Acta 340 (2004) 195–200
stored for a few weeks vs. an entire year. While the CDC HPLC-FD assay does not represent a reference method, nor is there any standard reference material with a certified concentration of tHcy available, CDC has participated over the last few years with the HPLCFD assay in several external quality assurance programs and in one international interlaboratory comparison study [9]. As a result, we know that our method is very comparable to the GC/MS assay which can be considered as a reference method.
4. Conclusions The ESA assay showed comparable within- and between-run imprecision (CV < 6%) to other frequently used HPLC assays. The assay displayed a good linear dynamic range, complete recovery of Lhomocystine added to plasma, and satisfactory dilution linearity. As long as the assay is calibrated in plasma, it correlates and agrees well with the CDC assay and produces only a small negative bias of 0.28 Amol/l ( 3.3%). Limitations of the ESA assay are that samples need to be measured by HPLC-ECD shortly after preparation. The lifetime of the guard and analytical column are significantly shorter than with the CDC assay. The major advantage of the ESA assay is that it does not require a derivatization step and thus is cost- and time-saving with regard to the sample preparation. Acknowledgements This study was supported by ESA which provided the mobile phase and columns needed to perform this evaluation. An abstract containing the summary of this evaluation has been presented at Experimental Biology 2003.
References [1] Mangoni AA, Jackson SHD. Homocysteine and cardiovascular disease: current evidence and future prospects. Am J Med 2002;112:556 – 65. [2] Tan KCB, Karmin O, Chow W-S, Ai VHG, Siow YL, Lam KSL. Hyperhomocysteinemia and impaired vasomotor function in type 2 diabetes mellitus. Eur J Clin Invest 2002;32: 328 – 34. [3] Teunissen CE, de Vente J, Steinbusch HWM, De Bruijn C. Biochemical markers related to Alzheimer’s dementia in serum and cerebrospinal fluid. Neurobiol Aging 2002;23: 485 – 508. [4] Seshadri S, Beiser A, Selhub J, Jacques PF, Rosenberg IH, D’Agostino RB, et al. Plasma homocysteine as a risk factor for dementia and Alzheimer’s disease. N Engl J Med 2002; 346:476 – 83. [5] Ducros V, Demuth K, Sauvant MP, Quillard M, Causse´ E, Candito M, et al. Methods for homocysteine analysis and biological relevance of the results. J Chromatogr, B, Biomed Sci Appl 2002;781:207 – 26. [6] Malinow MR, Sexton G, Averbuch M, Grossman M, Wilson D, Upson B. Homocyst(e)inemia in daily practice: levels in coronary artery disease. Coron Artery Dis 1990;1:215 – 20. [7] Solomon BP, Duda CT. Homocysteine determination in plasma. Curr Sep 1998;17:3 – 7. [8] ESA Application Note 70-3994P. [9] Pfeiffer CM, Huff DL, Smith SJ, Miller DT, Gunter EW. Comparison of plasma total homocysteine measurements in 14 laboratories: an international study. Clin Chem 1999;45: 1261 – 8. [10] Pfeiffer CM, Huff DL, Gunter EW. Rapid and accurate HPLC assay for plasma total homocysteine and cysteine in a clinical laboratory setting. Clin Chem 1999;45:290 – 2. [11] Vester B, Rasmussen K. High performance liquid chromatography method for rapid and accurate determination of homocysteine in plasma and serum. Eur J Clin Chem Clin Biochem 1991;29:549 – 54. [12] Gilfix BM, Blank DW, Rosenblatt DS. Novel reductant for determination of total plasma homocysteine. Clin Chem 1997; 43(4):687 – 8. [13] D’Eramo JL, Finkelstein AE, Boccazzi FO, Fridman O. Total homocysteine levels in plasma: high-performance liquid chromatographic determination with electrochemical detection and glassy carbon electrode. J Chromatogr, B, Biomed Sci Appl 1998;720:205 – 10.
Comparing the ESA HPLC total homocysteine assay with electrochemical detection to the CDC in-house HPLC assay with fluorescence detection Ming Zhang, Christine M. Pfeiffer * Centers for Disease Control and Prevention, 4770 Buford Hwy NE, Atlanta, GA 30341-3724, USA Received 11 September 2003; received in revised form 24 October 2003; accepted 30 October 2003
Abstract Background: Environmental science has developed a simple high performance liquid chromatography (HPLC) assay with electrochemical detection for total homocysteine (tHcy) measurement that does not require derivatization of free thiols. We evaluated this method and compared it with the CDC HPLC assay with fluorescence detection (FD). Methods: tHcy is measured after reduction of disulfides/protein-bound thiols and protein precipitation using four channels of an ESA CoulArray detector. Lhomocystine is used as calibrator, penicillamine as internal standard. Results: Aqueous calibration of the ESA assay resulted in overestimation of tHcy by f 30% compared to the HPLC-FD method. Calibration in plasma alleviated the matrix effect. The within- (n = 3) and between-run (n = 20) imprecision was < 6%, the linearity up to 100 Amol/l was excellent, and the recovery of tHcy added to plasma was nearly complete (98.7% F 2.3%). Good correlation was observed between both methods for 266 plasma samples. The ESA assay showed a minimal negative bias of 0.28 Amol/l (3.3%). Conclusion: The ESA tHcy assay performed well in terms of accuracy and precision, and showed good agreement with the CDC HPLC-FD assay when calibrated in plasma. The major advantage of this assay is that it does not require sample derivatization. Disadvantages include instability of the prepared samples for prolonged storage and matrix effects. Published by Elsevier B.V. Keywords: CoulArray detector; Porous graphite electrode; Method comparison; Matrix effect; Cardiovascular disease
1. Introduction Determination of total homocysteine (tHcy) in plasma has become an important diagnostic procedure in clinical chemistry due to strong evidence supporting the hypotheses that increased homocysteine is an independent determinant of vascular disease and that * Corresponding author. Tel.: +1-770-488-7926; fax: +1-770488-4139. E-mail address: [email protected] (C.M. Pfeiffer). 0009-8981/$ - see front matter. Published by Elsevier B.V. doi:10.1016/j.cccn.2003.10.020
hyperhomocysteinemia often may be amenable to homocysteine-lowering treatment with B vitamins [1]. Recent studies have suggested that increased tHcy may play a role in the development of vascular complications of diabetes [2], and is a strong, independent risk factor for the development of dementia and Alzheimer’s disease [3,4]. Several types of methods are now available to determine tHcy in plasma [5]. Chromatographic methods, especially with fluorometric detection (FD), are still widely used, and immu-
196
M. Zhang, C.M. Pfeiffer / Clinica Chimica Acta 340 (2004) 195–200
noassays have recently become broadly employed because of full automation. HPLC with electrochemical detection (ECD) is also used and has the major advantage of requiring no derivatization of thiols before detection [6,7]. ESA (Chelmsford, MA) has developed a simple method to determine plasma tHcy by HPLC with coulometric ECD on porous graphite electrodes. This type of probe can be used to measure thiols for extended periods before any electrode maintenance is required [8]. The main purpose of this study was to evaluate this new ESA assay. A second purpose was to compare it to a widely used assay that has been previously compared to several other methods [9]. We thus used our in-house HPLC-FD assay as a comparison method [10].
The ESA application note specifies using an aqueous six-point calibration curve (0, 6.25, 12.5, 25, 50, and 100 Amol/l free thiol) with L-homocystine as calibrator. The aqueous calibrators consist of 100 Al K2EDTA, 100 Al water, and 100 Al calibrator. We also evaluated the use of plasma-based calibration to better simulate the composition of the patient samples. The plasma calibrators consist of 100 Al K2EDTA, 100 Al EDTA plasma pool, and 100 Al aqueous calibrator. Peak area ratios are used to calculate tHcy concentrations. In the case of plasma-based calibration, the endogenous amount of tHcy needs to be subtracted, i.e., the peak area ratio tHcy/IS for the plasma calibrator to which 0 Amol/l thiol was added needs to be subtracted from each plasma calibrator to which increasing amounts of thiols were added before plotting the calibration curve.
2. Materials and methods 2.2. CDC HPLC-FD assay 2.1. ESA HPLC-ECD assay The ESA tHcy assay is not a kit assay but rather an application note published by ESA describing a procedure that can be performed using an ESA eightchannel CoulArray detector (Model 5600) and the CoulArray Windows 32 software. Of the eight channels, only four channels are used for the tHcy assay (at potentials 520, 750, 800, and 820 mV). Any isocratic HPLC system can be used. Our HPLC system comprised two ESA pumps (Model 580) and an ESA autosampler (Model 542). Chromatography was performed on an ESA HR-80 C18 analytical column (4.6 80 mm, 3 Am) in conjunction with a C18 guard column using 0.15 mol/l sodium dihydrogen phosphate, 1.0 mmol/l sodium dodecyl sulfate (SDS), 10% acetonitrile, pH 2.8 as mobile phase at a flow rate of 1 ml/min and a column temperature of 36 jC. Sample preparation is performed by adding 100 Al K 2 EDTA (2 mg/ml), 100 Al water, and 25 Al penicillamine as internal standard (IS) (0.2 mmol/l) to 100 Al EDTA plasma. The mixture of disulfides, mixed disulfides, and protein-bound thiols is reduced by incubating at room temperature for 30 min with tris(2-carboxyethyl)-phosphine hydrochloride (TCEP, 25 Al, 10% w/v). Proteins are precipitated with perchloric acid (HClO4, 500 Al, 0.3 mol/l). After centrifugation (10,000 rpm, 10 min), the supernatant is injected onto the HPLC column for analysis (20 Al, 15-min runtime).
The CDC tHcy HPLC assay is a modification of the method of Vester and Rasmussen [11] and Gilfix et al. [12] and uses TCEP to reduce protein-bound and oxidized thiols, trichloroacetic acid to precipitate proteins, and ammonium 7-fluorobenzo-2-oxa-1,3-diazole-4-sulfonate (SBD-F) to derivatize reduced thiols for fluorescent detection. L-homocystine and L-cysteine are used as calibrators and cystamine dihydrochloride (disulfide) is used as IS. The separation of the obtained thiol-derivatives is performed isocratically within 7 min, with the use of an acetate mobile phase at pH 5.5. Recoveries of thiols added to EDTA plasma at different concentrations are nearly complete (98.7 F 2.5%). Within-run imprecision is between 1.1% and 1.8%; between-run imprecision is between 2.4% and 6.7% [10]. 2.3. Method evaluation experiments Three levels of EDTA plasma QC pools were analyzed by the ESA assay with aqueous and plasmabased calibration over 16 and 20 days, respectively, to assess between-run imprecision and the influence of plasma matrix on the calibration. The same EDTA plasma QC pools were also analyzed in three replicates each for within-run imprecision. These were the same QC pools used for our in-house HPLC-FD method. They were stored at 70 jC for f 2 years. No
M. Zhang, C.M. Pfeiffer / Clinica Chimica Acta 340 (2004) 195–200
deviations from originally determined 2 S.D. limits were observed during this time period. To assess recovery, L-homocystine was added to an EDTA plasma sample with low endogenous tHcy level ( f 6 Amol/l) at seven concentrations in duplicate: 0, 3.125, 6.25, 12.5, 25, 50, and 100 Amol/l free thiol added. Recovery was calculated as the ratio between measured tHcy in each spiked plasma sample divided by expected tHcy (measured tHcy in the unspiked plasma sample + added tHcy) in each spiked plasma sample. To assess dilution linearity, an EDTA plasma sample with relatively high endogenous tHcy concentration ( f 26 Amol/l) was serially diluted with water (1:2, 1:4, 1:8, and 1:16). These tests were performed over 3 days each. To compare the ESA assay to the CDC assay, we used 266 EDTA plasma samples, a convenience subset from the National Health and Nutrition Examination Survey (NHANES) 1999+, which includes an omnibus informed consent and Human Subjects Review protocol. Samples not analyzed simultaneously with both methods were stored at 70 jC for no longer than 1 year between the assays. 2.4. Statistical analysis Where appropriate, data are presented as means, with the variation expressed as standard deviation (S.D.) and coefficient of variance (% CV). Linear regression analysis was used to construct calibration curves and to perform method comparison. Results for
197
the QC pools obtained with the two calibration types (aqueous and plasma-based) were compared by Bonferroni t-test. A p-value of < 0.05 was considered statistically significant. Bland –Altman bias plots were used to assess the agreement between the ESA and CDC assay. Possible systematic biases were assessed by computing the 95% confidence intervals for the mean differences between the two methods (mean difference F 2 S.E.M.). Limits of agreement were assessed by calculating the central 0.95 intervals (mean difference F 2 S.D.). To assess the mean proportional bias between the two methods, we calculated the ratios of the ESA and CDC results expressed as percent.
3. Results and discussion 3.1. Comparison of plasma-based and aqueous calibration Fig. 1 shows calibration curves of the ESA assay performed in water and in plasma. While the plasmabased calibration curve displays a proportional increase in peak area ratio with increasing thiol concentration, the aqueous calibration curve shows lower than expected peak area ratios. This indicates that in aqueous calibration some of the thiol is lost, possibly through unspecific binding to glassware and/or plasticware. The lower slope of the aqueous calibration curve resulted in tHcy concentrations that were f 30% higher than with plasma-based calibration
Fig. 1. ESA assay calibration curve for tHcy added to water or to EDTA plasma. Triangles represent the response in water; solid line represents linear regression for calibration in water. Squares represent the response in plasma after correcting for the response of the endogenous amount of tHcy; broken line represents linear regression for calibration in plasma.
198
M. Zhang, C.M. Pfeiffer / Clinica Chimica Acta 340 (2004) 195–200
(Table 1). Since the response of the IS was the same in aqueous and plasma-based calibration, we can conclude that penicillamine cannot correct the observed matrix effect and calibration of the ESA assay has to be performed in plasma. When the ESA assay was calibrated in plasma, it produced virtually the same results as obtained with the CDC assay calibrated in plasma (Table 1 and Fig. 3).
assay earlier: plasma calibration produced more precise results than aqueous calibration [10]. The withinand between-run CVs for the ESA assay were close to reported CVs for HPLC with electrochemical detection, i.e., 3.9% and 5.6%, respectively [13], but were slightly higher than CVs reported for fluorometric detection (mean within-run CV, 1.0%; mean between-run CV, 3.2% Ref. [5]).
3.2. Within- and between-run imprecision
3.3. Linear dynamic range, recovery, and dilution linearity
Plasma QC pools analyzed in three replicates each for within-run imprecision (Table 1) showed comparable results for aqueous and plasma calibration ( < 5%). Analyzed for between-run imprecision over 16 or 20 days, the plasma QC pools showed significantly lower imprecision for plasma calibration ( f 6%) than for aqueous calibration ( f 11%) (Table 1). A similar observation was made with the CDC Table 1 Within- and between-run imprecision of the ESA and CDC tHcy assay and comparison between aqueous and plasma-based calibration QC low
QC medium
QC high
ESA assay, within-run imprecision, aqueous calibration (n = 3) Mean tHcy F S.D. 10.2 F 0.5 14.2 F 0.1 34.3 F 0.6 (Amol/l) CV (%) 4.5 0.8 1.8 ESA assay, within-run imprecision, plasma-based calibration (n = 3) Mean tHcy F S.D. 8.2 F 0.1 11.2 F 0.5 27.3 F 0.6 (Amol/l) CV (%) 1.2 4.4 2.1 ESA assay, between-run imprecision, aqueous calibration (n = 16) Mean tHcy F S.D. 10.5 F 1.1 14.9 F 1.6 35.7 F 4.2 (Amol/l) CV (%) 10.3 10.4 11.8 ESA assay, between-run imprecision, plasma-based calibration (n = 20) Mean tHcy F S.D. 8.1 F 0.5 11.5 F 0.6 27.4 F 1.7 (Amol/l) CV (%) 6.1 5.3 6.1 CDC assay, between-run imprecision, plasma-based calibration (n = 22) Mean tHcy F S.D. 8.6 F 0.5 11.8 F 0.6 27.2 F 1.3 (Amol/l) CV (%) 5.4 5.0 4.7
The linearity of the plasma-based calibration curve with tHcy added up to 100 Amol/l was excellent (mean slope F S.D.: 0.0102 F 0.0006, mean intercept F S.D.: 0.0019 F 0.0055, mean correlation coefficient F S.D.: 0.9997 F 0.0003 over 20 days). The mean recovery ( F S.D.) of L-homocystine added to plasma was nearly complete (98.7 F 2.3% over 3 days). Serial dilution of a plasma sample with increased tHcy concentration resulted in excellent dilution linearity ( y = 0.9958x 0.1343, R2 = 0.9988 over 3 days). The mean ratio ( F S.D.) between the measured and expected tHcy concentration was 0.97 F 0.03. This indicates that while the ESA assay underlies matrix effects, dilution of plasma up to 1:16 with water still leaves sufficient matrix (plasma proteins) in the sample to avoid matrix effects seen with plain water. 3.4. Robustness of the ESA assay The retention times (RT) for tHcy and the IS varied by f 5% over 28 runs performed over 7 months. For tHcy, the mean RT ( F S.D.) was 5.95 F 0.29 min; for the IS, the mean RT ( F S.D.) was 7.81 F 0.30 min. A typically chromatogram for the low QC pool with a tHcy concentration of f 10 Amol/l is presented in Fig. 2. The HPLC guard column needed to be changed every 200 injections to avoid deterioration of chromatographic resolution and peak shapes. The ESA analytical column was used for 400 –600 injections. In comparison, we can use the HPLC column in the CDC assay for f 1500 injections. Because porous graphite electrodes have sufficient surface area, they do not require frequent electrode maintenance such as does noble metal electrodes (gold or platinum). They provide a stable sensor and have to
M. Zhang, C.M. Pfeiffer / Clinica Chimica Acta 340 (2004) 195–200
199
intensity eluting between tHcy and the IS and interfering with resolution and integration. This unknown peak seemed to derive from the IS penicillamine because it appeared in a sample that contained only reagents and the IS, but it did not appear in the reagent blank. 3.5. Comparison of the ESA and CDC assay
Fig. 2. ESA assay representative chromatogram of a plasma sample ( f 10 Amol/l tHcy).
be typically cleaned by applying a potential of 1200 mV for 5 min after 100 samples have been processed [8]. In the ESA tHcy assay, the CoulArray detector was programmed to apply a voltage of 1000 mV at the end of each injection for 10 s. This allowed continuous usage of the electrode without any maintenance for the duration of the entire study. Because the free thiols are not derivatized in the ESA assay, samples are not stable at room temperature for extended periods of time. Plasma samples need to be measured by HPLC-ECD immediately after preparation or they have to be stored refrigerated overnight or frozen ( 15 jC or colder) for a few days. In samples stored at 15 jC for 14 days, we observed the appearance of an unknown peak with very high
tHcy concentrations determined in 266 apparently healthy subjects were spanning a range of 3– 20 Amol/ l and gave good correlation between the two methods: tHcy (ESA) = 0.9533 tHcy (CDC) + 0.0711 (r2 = 0.9445). To assess the agreement between the two methods, we plotted for each sample the difference of results from the two methods ( y-axis) against the mean of results (x-axis) (Fig. 3). The ESA assay showed a slight negative bias of 0.28 Amol/l (95% confidence interval, 0.375, 0.193), corresponding to a relative bias of 3.3%. The central 0.95 interval (mean difference F 2 S.D.) indicates the agreement between the two methods. Ninety-five percent of tHcy results obtained with the ESA assay were between 1.8 Amol/ l lower and 1.2 Amol/l higher than results obtained with the CDC method. Although some of the plasma samples were stored for up to 1 year between performing the two assays, we do not believe that the storage accounts for the small bias we found. We stratified the entire sample set by length of storage and did not find any differences in bias whether samples were only
Fig. 3. Bland – Altman bias plot showing the difference between the ESA HPLC-ECD and the CDC HPLC-FD assay. Center line represents the mean difference ( 0.28 Amol/l), the bottom and top lines represent the limits of agreement corresponding to the mean difference F2 S.D. ( 1.77, 1.20 Amol/l); n=266 subjects.
200
M. Zhang, C.M. Pfeiffer / Clinica Chimica Acta 340 (2004) 195–200
stored for a few weeks vs. an entire year. While the CDC HPLC-FD assay does not represent a reference method, nor is there any standard reference material with a certified concentration of tHcy available, CDC has participated over the last few years with the HPLCFD assay in several external quality assurance programs and in one international interlaboratory comparison study [9]. As a result, we know that our method is very comparable to the GC/MS assay which can be considered as a reference method.
4. Conclusions The ESA assay showed comparable within- and between-run imprecision (CV < 6%) to other frequently used HPLC assays. The assay displayed a good linear dynamic range, complete recovery of Lhomocystine added to plasma, and satisfactory dilution linearity. As long as the assay is calibrated in plasma, it correlates and agrees well with the CDC assay and produces only a small negative bias of 0.28 Amol/l ( 3.3%). Limitations of the ESA assay are that samples need to be measured by HPLC-ECD shortly after preparation. The lifetime of the guard and analytical column are significantly shorter than with the CDC assay. The major advantage of the ESA assay is that it does not require a derivatization step and thus is cost- and time-saving with regard to the sample preparation. Acknowledgements This study was supported by ESA which provided the mobile phase and columns needed to perform this evaluation. An abstract containing the summary of this evaluation has been presented at Experimental Biology 2003.
References [1] Mangoni AA, Jackson SHD. Homocysteine and cardiovascular disease: current evidence and future prospects. Am J Med 2002;112:556 – 65. [2] Tan KCB, Karmin O, Chow W-S, Ai VHG, Siow YL, Lam KSL. Hyperhomocysteinemia and impaired vasomotor function in type 2 diabetes mellitus. Eur J Clin Invest 2002;32: 328 – 34. [3] Teunissen CE, de Vente J, Steinbusch HWM, De Bruijn C. Biochemical markers related to Alzheimer’s dementia in serum and cerebrospinal fluid. Neurobiol Aging 2002;23: 485 – 508. [4] Seshadri S, Beiser A, Selhub J, Jacques PF, Rosenberg IH, D’Agostino RB, et al. Plasma homocysteine as a risk factor for dementia and Alzheimer’s disease. N Engl J Med 2002; 346:476 – 83. [5] Ducros V, Demuth K, Sauvant MP, Quillard M, Causse´ E, Candito M, et al. Methods for homocysteine analysis and biological relevance of the results. J Chromatogr, B, Biomed Sci Appl 2002;781:207 – 26. [6] Malinow MR, Sexton G, Averbuch M, Grossman M, Wilson D, Upson B. Homocyst(e)inemia in daily practice: levels in coronary artery disease. Coron Artery Dis 1990;1:215 – 20. [7] Solomon BP, Duda CT. Homocysteine determination in plasma. Curr Sep 1998;17:3 – 7. [8] ESA Application Note 70-3994P. [9] Pfeiffer CM, Huff DL, Smith SJ, Miller DT, Gunter EW. Comparison of plasma total homocysteine measurements in 14 laboratories: an international study. Clin Chem 1999;45: 1261 – 8. [10] Pfeiffer CM, Huff DL, Gunter EW. Rapid and accurate HPLC assay for plasma total homocysteine and cysteine in a clinical laboratory setting. Clin Chem 1999;45:290 – 2. [11] Vester B, Rasmussen K. High performance liquid chromatography method for rapid and accurate determination of homocysteine in plasma and serum. Eur J Clin Chem Clin Biochem 1991;29:549 – 54. [12] Gilfix BM, Blank DW, Rosenblatt DS. Novel reductant for determination of total plasma homocysteine. Clin Chem 1997; 43(4):687 – 8. [13] D’Eramo JL, Finkelstein AE, Boccazzi FO, Fridman O. Total homocysteine levels in plasma: high-performance liquid chromatographic determination with electrochemical detection and glassy carbon electrode. J Chromatogr, B, Biomed Sci Appl 1998;720:205 – 10.