- Email: [email protected]
Adjustment of direct high-density lipoprotein cholesterol measurements according to intercurrent triglyceride corrects for interference by triglyceride-rich lipoproteins
Journal of Clinical Lipidology (2010) 4, 305–309
New from the Laboratory
Adjustment of direct high-density lipoprotein cholesterol measurements according to intercurrent triglyceride corrects for interference by triglyceride-rich lipoproteins Nimalie J. Perera, MBBS, FRACP, FRCPA, Jennifer C. Burns, BSc, Dip.Ed, Ryle S. Perera, BS, MS, PhD, Barry Lewis, BSc, MB ChB, PhD, MD, FRCP, FRCPath, David R. Sullivan, MBBS, FRACP, FRCPA* Department of Clinical Biochemistry, Royal Prince Alfred Hospital, Missenden Road, Camperdown, NSW 2050, Australia (Drs. N. Perera and Sullivan; and J. C. Burns); Faculty of Medicine, University of Sydney, NSW, Australia (Dr. Lewis); and Faculty of Business and Economics, Macquarie University, NSW, Australia (Dr. R. Perera) KEYWORDS: Cardiovascular disease (CVD); High-density lipoprotein cholesterol (HDL-C); Interference; Prediction formulae; Triglycerides (TG); Triglyceride-rich lipoproteins (TRL)
BACKGROUND: Low plasma levels of high-density-lipoprotein cholesterol (HDL-C) and high triglyceride (TG) are strongly associated with cardiovascular disease (CVD). Clinical recognition of this high-risk population demands accurate measurement of HDL-C, whereas cost and clinical demand dictate that optimal HDL-C measurement requires fully automated methods that avoid manual precipitation. Commercial techniques use specific reagents to selectively expose and ‘‘directly’’ measure cholesterol in HDL. However, these ‘‘direct’’ methods may experience interference from the cholesterol content of triglyceride-rich-lipoproteins (TRL), leading to analytical overestimation of HDL-C, with subsequent underestimation of low-density-lipoprotein cholesterol (LDL-C) and of CVD risk. OBJECTIVE: The aim of this study was to develop a method to overcome this interference. METHODS: Serum/Li1-heparin plasma samples from consecutive patients were analyzed for HDL-C by the comparison of three generations of the Roche Diagnostics, HDL-C assay on a Hitachi-917 or Modular-PPE analyzer. HDL-C measurement was performed before and after removal of TRL by ultracentrifugation (‘‘direct’’ HDL-C and HDL-UC, respectively). We examined the effect of TG on the relationship between HDL-UC and ‘‘direct’’ HDL-C. Analysis of variance multiregression analysis was performed for each generation of the commercial assay. RESULTS: We observed progressive TG interference that increased ‘‘direct’’ HDL-C by 10% to 15% or more in moderately hypertriglyceridemic samples (,600 mg/dL). Predictive equations were derived for each generation of the assay to estimate HDL-C in the absence of TRL. CONCLUSIONS: This study casts doubt on the specificity of ‘‘direct’’ HDL-C assays in the presence of hypertriglyceridemia. The use of assay-specific correction formulae to adjust for interference from TRL reduces the overestimation of HDL-C that influences CVD risk calculation, treatment, and follow-up of patients. Ó 2010 National Lipid Association. All rights reserved.
* Corresponding author. E-mail address: [email protected] Submitted December 30, 2009. Accepted for publication April 13, 2010.
Dyslipidemia involving low high-density lipoprotein cholesterol (HDL-C) and high triglyceride-rich lipoproteins (TRL) is an independent risk factor for atherosclerotic cardiovascular disease (CVD). It is important to recognize
1933-2874/$ - see front matter Ó 2010 National Lipid Association. All rights reserved. doi:10.1016/j.jacl.2010.04.002
306 the presence of this high-risk phenotype to allow detection, treatment, monitoring of progress, and reduction of risk according to recommended lipid cut points and targets.1 This recognition requires accurate laboratory HDL-C measurements, which are used for calculation of low-density lipoprotein cholesterol (LDL-C) and of CVD risk.2 A variety of methods has been used to determine HDL-C, exploiting its physical and chemical characteristics,3 including ultracentrifugation, electrophoresis, chromatography, and selective precipitation with the use of reagents such as heparinMn21, dextran sulfate-Mg21, phosphotungstate-Mg21, or polyethylene glycol (PEG).3-7 These methods are time consuming and are not amenable to full automation. The need for more efficient HDL-C measurements led to the development of several fully automated assays that selectively measure the cholesterol associated with HDL.3 The first-generation HDL-C assay (HDL-C1st, RocheÒ Hitachi 917), which was used in our laboratory until June 2002, was a nonseparation method with a-cyclodextrin, dextran sulfate, magnesium, PEG-modified cholesterol oxidase (CO), and cholesterol esterase (CE) enzymes. The HDL-C second-generation assay (HDL-C2nd, RocheÒ Modular PPE used until March 2007 in our laboratory) and the current third-generation assay (HDL-C3rd, RocheÒ Modular PPE) used no a-cyclodextrin but contains differing concentrations of dextran sulfate, magnesium, and PEG-modified CO and CE enzymes to selectively form water-soluble complexes of non-HDL lipoproteins. These non-HDL complexes, which contain LDL as well as TRL, such as chylomicrons and their remnants, are more resistant to the PEG-modified CO and CE enzymes, resulting in selective catalytic activity in decreasing order of HDL-C . chylomicrons and very low-density lipoprotein . LDL. However, this selectivity is not absolute, leading to concern that ‘‘direct’’ methods may overestimate HDL-C as the result of nonspecific reaction between the HDL-C reagents and the cholesterol content of non-HDL particles, particularly TRLs. The manufacturers state on the package insert that ‘‘to date, there is no model available which can mimic interference by triglycerides’’ and that ‘‘elevated concentrations of free fatty acids and denatured lipoproteins may cause falsely elevated HDL-cholesterol results.’’ It follows from the reciprocal relationship between TG and HDL-C in vivo that TRL levels are increased in precisely those patients in whom HDL-C is likely to be reduced, such as those with insulin resistance or increased risk of CVD.8 Furthermore, overestimation of HDL-C inevitably leads to underestimation of LDL-C when the Friedewald equation is used. This combination of analytical overestimation of HDL-C and underestimation of LDL-C results in underestimation of absolute risk of CVD in patients with this pattern of dyslipidemia. We compared the ‘‘direct’’ HDL-C measurement methods (HDL-C generation first, second, or third) against the same assay performed after the removal of TRL from the sample by ultracentrifugation. We calculated the difference between HDL-C results before and after this procedure and examined the effect of intercurrent TG levels. Analysis of variance
Journal of Clinical Lipidology, Vol 4, No 4, August 2010 nonlinear multiregression analysis was used to generate corrective formulae to reduce the positive bias introduced by TRL. The application of assay-specific corrections could provide more accurate results after the measurement of HDL-C in hypertriglyceridemic patients by the use of homogenous commercial ‘‘direct’’ automated assay systems.
Materials and methods Three sets of consecutive patients (male and female, aged .18 years) who attended an ambulatory CVD riskprevention clinic provided venous samples after a 12-hour fast. HDL-C was measured by use of the generations of HDL-C assay that were then current. Measurements were performed before and after removal of TRL by density ultracentrifugation (d , 1.006) of the samples (HDL-UC). Studies were conducted between January 2000 and June 2009. Additional analysis of the subgroups of patients with fasting ,600 mg/dL was performed. The HDL-C was measured ‘‘directly’’ with the use of a Hitachi-917 or Modular-PPE auto analyzer (RocheÒ Diagnostics, Basel, Switzerland). Reagents included a-cyclodextrin for the HDL-C1st assay and PEG-modified enzymes but not a-cyclodextrin for HDL-C2nd and HDLC3rd assays. Identical direct HDL-C measurements were performed before and after removal TRL and the recovery of HDL as a percentage (HDL-UC/ HDL-C) was compared with that of TG and total cholesterol in the same samples. We calculated the difference between HDL-C and HDLUC results and used Glick interferographs (difference of analytes in mg/dL before and after separation on the y-axis plotted against the likely interferant on the x-axis) to examine the effect of endogenous TG. Nonlinear component regression was performed because TG has a nonlinear effect on the accuracy of direct HDL-C measurements. Equations were generated for each assay of HDL-C to predict the HDL-UC.
Results The HDL-C was measured before and after removal of TRL, and the two sets of results were compared for each generation of the assay. The ‘‘direct’’ HDL-C measurement was consistently high compared with the corresponding HDL-C measurements after ultracentrifugation (HDL-UC). This finding was regarded as a positive bias associated with the direct HDL-C methods because the d , 1.006 fraction contained TRL, but no HDL, on electrophoresis (data not shown). The difference between ‘‘direct’’ HDL-C and HDL-UC measurements increased progressively with increasing TG levels. However, HDL-UC was negatively correlated with TG (HDL-UC 5 20.112 TG 1 1.458 mmol/L, R2 5 0.230, and HDL-UC 5 20.684 TG 1 315 mg/dL, R2 5 0.194). The HDL-C difference (HDL-C 2 HDL-UC) in mg/dL was examined according to increasing TG concentrations by the use of Glick interferographs. Increasing TG
Perera et al
Adjustment of direct HDL-C measurements
concentrations exacerbated interference (Fig. 1, left panels). Even the mild-to-moderate hypertriglyceridemia (,600 mg/dL) that is frequently associated with insulin resistance or the metabolic syndrome caused a substantial level of interference, represented by reductions in HDL-C % recovery, calculated as ([HDL-UC]/[HDL-C] ! 100) of 10% to 40% (Fig. 2). Conversely, recovery of TG or total cholesterol was unimpaired, suggesting that the ultracentrifugation process did not result in loss of recovery of any class of lipoproteins (data not shown). To adjust for the interference caused by TRL cholesterol, analysis of variance multiregression analysis was used to generate prediction formulae for HDL-UC for each
307 assay (Table 1; adjusted R2 0.98–0.99). Contrasting Glick interferographs were plotted to demonstrate the reduced inaccuracy when applying the predictive HDL-C formula for each of the three generations of the HDL-C assay (Fig. 1, right panels). Regression formulae differed when more severely hypertriglyceridemic patients (.600 mg/dL) were excluded (Tables 1 and 2).
Discussion These results suggest that the current automated ‘‘direct’’ HDL-C measurement methods that we studied experience
Figure 1 Plots of pair-wise comparisons of the differences in measured HDL-C before and after ultracentrifugation for generation 1, 2, and 3 HDL-C assays versus the intercurrent TG levels in mg/dL (left) showing positive interference with increasing TG concentrations. The corresponding difference between the calculated HDL-C when the prediction formulae was used and the HDL-C after ultracentrifugation for generation 1, 2, and 3 assays versus the intercurrent TG levels in mmol/L for the same samples (right) showing correction for TG interference. The solid line denotes the forecast trend lines for each data set.
308
Journal of Clinical Lipidology, Vol 4, No 4, August 2010 interference from TRL was described in some, but not all studies.13-16 Because methods involving a-cyclodextrin have been superseded, two more recent studies of ‘‘direct’’ HDL methods have reported positive biases that were attributed to TRL or to the presence of diabetes.17,18 Previous HDL-C methods involved isolation of HDL-C by precipitation of apolipoprotein B-containing TRL particles. In this study we used an identical HDL-C measurement technique (‘‘direct’’ HDL-C assays) before and after the ultracentrifugal removal of the TRL that could potentially interfere with the estimation. Loss of HDL-C was excluded by complete recovery of cholesterol and TG, irrespective of TG level, suggesting that the decrease in observed HDL-C is caused by the removal of d , 1.006 lipoproteins after density separation. Our method would not detect additional interference from TG-containing lipoproteins occupying the 1.006 , d , 1.019 range. Overestimation of HDL-C leads to underdiagnosis of the metabolic syndrome and insulin resistance, as well as underestimation of LDL-C according to the Friedewald equation. These combined effects result in underestimation of absolute risk of CVD, impairing the opportunity to effectively identify and treat patients at risk on the basis of metabolic risk factors. HDL-UC overcomes this problem by selectively removing interfering TRL, but this method is time-consuming and labor-intensive. Manufacturers of ‘‘direct’’ HDL-C assays have acknowledged that TRL cholesterol may interfere with HDL-C analysis and that TG interference studies are based on the use of Intralipid,
Figure 2 HDL-C recovery as a percentage (Recovery %) , calculated as ([HDL-UC]/[HDL-C] ! 100) vs the TG in mg/dL measured in the same samples from subjects with TG , 600 mg/dL. The reduced recovery% is caused by positive interference on the direct HDL-C measurements before removal of TRL particles by ultracentrifugation.
interference from the cholesterol content of TRL, causing a correctable positive bias. This finding may be a ‘‘class-effect’’ that applies to other HDL-C methods, particularly those in which physical separation of HDL does not occur. The problem appeared more marked with the second-generation assay in which a-cyclodextrin had been removed, but this was partially compensated with the third-generation assay, perhaps because of the increase in dextran sulfate concentration. Results from method comparison studies before the year 2000 suggested good agreement between ‘‘separation’’ HDL-C methods and the reference method, even in the presence of Intralipid (Kabivitrum Inc., Alameda, CA) or TRL.9-12 Where positive bias occurred, it was attributed to incomplete precipitation with the comparator method or the presence of apolipoprotein E-containing HDL, but it should be noted that the sources of triglyceride (TG) used in these studies may have had relatively low cholesterol content. Direct methods usually involved the use of a-cyclodextrin at this time, and positive
Table 1 HDL-C prediction formulae for generation 1, 2, and 3 assays using ANOVA multiregression analysis for all subjects (TG range up to 3000 mg/dL) Prediction formula for HDL-C, mmol/L
Number of observations
R2
Predicted HDL-C1st 5 0.95 ! (HDL-C1st) – 1.73 ! (LN TG) mg/dL Predicted HDL-C1st 5 0.92 ! (HDL-C1st) – 0.05 ! (LN TG) mmol/L Predicted HDL-C2nd 5 0.88 ! (HDL-C2nd) – 2.99 ! (LN TG) mg/dL Predicted HDL-C2nd 5 0.87 ! (HDL-C2nd) – 0.04 ! (LN TG) mmol/L Predicted HDL-C3rd 51.02 ! (HDL-C3rd) – 2.65 ! (LN TG) mg/dL Predicted HDL-C3rd 5 0.95 ! (HDL-C3rd) – 0.07 ! (LN TG) mmol/L
201
0.995
101
0.987
105
0.986
ANOVA, analysis of variance; HDL-C, high-density lipoprotein cholesterol; HDL-C1st, generation 1 assay for HDL-C); HDLC2nd, generation 2 assay for HDL-C; HDLC3rd, generation 3 assay for HDL-C; LN, natural log; TG, triglyceride.
Perera et al
Adjustment of direct HDL-C measurements
Table 2 HDL-C prediction formulae for generation 1, 2, and 3 assays using ANOVA multiregression analysis in subjects with TG , 600 mg/dL (6.8 mmol/L). Prediction formula for HDL-C Predicted HDL-C1st 5 0.93 ! (HDL-C1st) – 0.86 ! (LN TG) mg/dL Predicted HDL-C1st 5 0.91 ! (HDL-C1st) – 0.03 ! (LN TG) mmol/L Predicted HDL-C2nd 5 0.89 ! (HDL-C2nd) – 1.92 ! (LN TG) mg/dL Predicted HDL-C2nd 5 0.86 ! (HDL-C2nd) – 0.07 ! (LN TG) mmol/L Predicted HDL-C3rd 5 0.996 ! (HDL-C3rd) – 1.85 ! (LN TG) mg/dL Predicted HDL-C3rd 5 0.943 ! (HDL-C3rd) – 0.04 ! (LN TG) mmol/L
Number of observations
R2
160
0.99
309 1 34.6 mg/dL, R2 5 0.9999). This contrasts the in vivo inverse relationship, which we confirmed in this cohort. We believe that QA materials should be spiked with combinations of lipoproteins in which the reciprocal relationship between HDL and TRL is maintained to ensure that automated HDL-C assays avoid positive interference from TRL cholesterol.
References
86
0.98
101
0.99
ANOVA, analysis of variance; HDL-C, high-density lipoprotein cholesterol; HDL-C1st, generation 1 assay for HDL-C); HDLC2nd, generation 2 assay for HDL-C; HDLC3rd, generation 3 assay for HDL-C; LN, natural log; TG, triglyceride.
which does not contain cholesterol. We have demonstrated that the progressive modifications of ‘‘direct’’ HDL-C assays have not completely overcome this problem. This is because the interference depends on the unpredictable variation in the TG levels in individual patient specimens. It is necessary for current HDL-C assays to be made more selective for HDL particles, free from interference by TRL cholesterol, while maintaining the efficiency and precision of automated methods. Until this is achieved, our study provides a means of correcting for this bias by the use of assay-specific adjustment formulae. The betweengeneration differences and those after exclusion of severe hypertriglyceridemia show that the interference is assay and range specific. Therefore, we encourage manufacturers of individual ‘‘direct’’ HDL-C assays to generate formulae that are appropriate for their assay, particularly in the TG range in which results are required for estimation of LDL-C. Current quality assurance (QA) materials are not representative of the typical metabolic syndrome phenotype. Materials for the RCPA General Serum Chemistry and Special Lipids external QA Programmes are spiked with increasing concentrations of both TRL and HDL, leading to a positive association between TG and HDL-C levels (TG 5 0.54 ! HDL-C 1 0.35 mmol/L and TG 5 0.60 ! HDL-C
1. Barter PJ. HDL cholesterol testing: implications for clinical management. Aust Presc. 1994;17:99–102. 2. Marques-Vidal P, Ferrario M, Kuulasmaa K, et al. Quality Assessment of Data on HDL cholesterol in the WHO Monica Project. Lyon, France: World Health Organization Publications; 1994. 3. Warnick GR, Nauck M, Rifai N. Evolution of methods for measurement of HDL-cholesterol: from ultracentrifugation to homogenous assays. Clin Chem. 2001;47:1579–1596. 4. Manual of Laboratory Operations. Lipid Research Clinics Program Lipid and lipoprotein analysis DHEW Publication Number (NIH) 75:628:59. Bethesda, MD: NIH; 1974. 5. Mahley RW, Weisgraber KH. An electrophoretic method for the quantitative isolation of human and swine plasma lipoproteins. Biochemistry. 1974;13:1964–1969. 6. Warnick GR, Benderson J, Albers JJ. Dextran sulfate-Mg21 precipitation procedure for quantitation of high-density-lipoprotein cholesterol. Clin Chem. 1982;28:1379–1388. 7. Burstein M, Scholnick HR. Lipoprotein-polyanion-metal interactions. Adv Lipid Res. 1973;11:67–108. 8. Phillips NR, Havel RJ, Kane JP. Serum apolipoprotein A-I levels: relationship to lipoprotein lipid levels and selected demographic variables. Am J Epidemiol. 1982;116:302–313. 9. Rifai N, Cole TG, Iannotti E, et al. Assessment of interlaboratory performance in external proficiency testing programs with a direct HDL-cholesterol assay. Clin Chem. 1998;44:1452–1458. 10. Nauk M, Marz W, Haas B, et al. Homogeneous assay for direct determination of high-density lipoprotein cholesterol evaluated. Clin Chem. 1996;42:424–429. 11. Harris N, Galpchian V, Rifai N. Three routine methods for measuring high-density lipoprotein cholesterol compared with the Reference Method. Clin Chem. 1996;42:738–743. 12. de Keijer MH, Elbers D, Baadenhuijsen H, et al. Evaluation of five different high-density lipoprotein cholesterol assays; the most precise are not the most accurate. Ann Clin Biochem. 1999;36:168–175. 13. Hubbard RS, Hirany SV, Devaraj S, et al. Evaluation of a rapid homogeneous method for direct measurement of high-density lipoprotein cholesterol. Am J Clin Pathol. 1998;110:495–502. 14. Okazaki M, Sasamoto K, Muramatsu T, et al. Evaluation of precipitation and direct methods for HDL-cholesterol assay by HPLC. Clin Chem. 1997;43:1885–1890. 15. Okada M, Matsui H, Fujiwara A. Direct measurement of HDL cholesterol: method eliminating apolipoprotein E-rich particles. J Clin Lab Anal. 2001;15:223–229. 16. Eglof M, Eglise D, Duvillard L, et al. Multicentre evaluation on different analysers of three methods for direct HDL-cholesterol assay. Ann Biol Clin. 1999;57:561–572. 17. Zhao W, Chaffin C, Desmond RA, et al. Overestimation of HDLcholesterol using a homogeneous ‘‘direct’’ assay. J Clin Lab Anal. 2004;18:42–44. 18. Saeed BO, Smart P, Keeka G, et al. Comparison of two direct methods for HDL cholesterol measurement with an indirect precipitation method in diabetic patients. Diab Nutr Metab. 2002;15:169–172.
New from the Laboratory
Adjustment of direct high-density lipoprotein cholesterol measurements according to intercurrent triglyceride corrects for interference by triglyceride-rich lipoproteins Nimalie J. Perera, MBBS, FRACP, FRCPA, Jennifer C. Burns, BSc, Dip.Ed, Ryle S. Perera, BS, MS, PhD, Barry Lewis, BSc, MB ChB, PhD, MD, FRCP, FRCPath, David R. Sullivan, MBBS, FRACP, FRCPA* Department of Clinical Biochemistry, Royal Prince Alfred Hospital, Missenden Road, Camperdown, NSW 2050, Australia (Drs. N. Perera and Sullivan; and J. C. Burns); Faculty of Medicine, University of Sydney, NSW, Australia (Dr. Lewis); and Faculty of Business and Economics, Macquarie University, NSW, Australia (Dr. R. Perera) KEYWORDS: Cardiovascular disease (CVD); High-density lipoprotein cholesterol (HDL-C); Interference; Prediction formulae; Triglycerides (TG); Triglyceride-rich lipoproteins (TRL)
BACKGROUND: Low plasma levels of high-density-lipoprotein cholesterol (HDL-C) and high triglyceride (TG) are strongly associated with cardiovascular disease (CVD). Clinical recognition of this high-risk population demands accurate measurement of HDL-C, whereas cost and clinical demand dictate that optimal HDL-C measurement requires fully automated methods that avoid manual precipitation. Commercial techniques use specific reagents to selectively expose and ‘‘directly’’ measure cholesterol in HDL. However, these ‘‘direct’’ methods may experience interference from the cholesterol content of triglyceride-rich-lipoproteins (TRL), leading to analytical overestimation of HDL-C, with subsequent underestimation of low-density-lipoprotein cholesterol (LDL-C) and of CVD risk. OBJECTIVE: The aim of this study was to develop a method to overcome this interference. METHODS: Serum/Li1-heparin plasma samples from consecutive patients were analyzed for HDL-C by the comparison of three generations of the Roche Diagnostics, HDL-C assay on a Hitachi-917 or Modular-PPE analyzer. HDL-C measurement was performed before and after removal of TRL by ultracentrifugation (‘‘direct’’ HDL-C and HDL-UC, respectively). We examined the effect of TG on the relationship between HDL-UC and ‘‘direct’’ HDL-C. Analysis of variance multiregression analysis was performed for each generation of the commercial assay. RESULTS: We observed progressive TG interference that increased ‘‘direct’’ HDL-C by 10% to 15% or more in moderately hypertriglyceridemic samples (,600 mg/dL). Predictive equations were derived for each generation of the assay to estimate HDL-C in the absence of TRL. CONCLUSIONS: This study casts doubt on the specificity of ‘‘direct’’ HDL-C assays in the presence of hypertriglyceridemia. The use of assay-specific correction formulae to adjust for interference from TRL reduces the overestimation of HDL-C that influences CVD risk calculation, treatment, and follow-up of patients. Ó 2010 National Lipid Association. All rights reserved.
* Corresponding author. E-mail address: [email protected] Submitted December 30, 2009. Accepted for publication April 13, 2010.
Dyslipidemia involving low high-density lipoprotein cholesterol (HDL-C) and high triglyceride-rich lipoproteins (TRL) is an independent risk factor for atherosclerotic cardiovascular disease (CVD). It is important to recognize
1933-2874/$ - see front matter Ó 2010 National Lipid Association. All rights reserved. doi:10.1016/j.jacl.2010.04.002
306 the presence of this high-risk phenotype to allow detection, treatment, monitoring of progress, and reduction of risk according to recommended lipid cut points and targets.1 This recognition requires accurate laboratory HDL-C measurements, which are used for calculation of low-density lipoprotein cholesterol (LDL-C) and of CVD risk.2 A variety of methods has been used to determine HDL-C, exploiting its physical and chemical characteristics,3 including ultracentrifugation, electrophoresis, chromatography, and selective precipitation with the use of reagents such as heparinMn21, dextran sulfate-Mg21, phosphotungstate-Mg21, or polyethylene glycol (PEG).3-7 These methods are time consuming and are not amenable to full automation. The need for more efficient HDL-C measurements led to the development of several fully automated assays that selectively measure the cholesterol associated with HDL.3 The first-generation HDL-C assay (HDL-C1st, RocheÒ Hitachi 917), which was used in our laboratory until June 2002, was a nonseparation method with a-cyclodextrin, dextran sulfate, magnesium, PEG-modified cholesterol oxidase (CO), and cholesterol esterase (CE) enzymes. The HDL-C second-generation assay (HDL-C2nd, RocheÒ Modular PPE used until March 2007 in our laboratory) and the current third-generation assay (HDL-C3rd, RocheÒ Modular PPE) used no a-cyclodextrin but contains differing concentrations of dextran sulfate, magnesium, and PEG-modified CO and CE enzymes to selectively form water-soluble complexes of non-HDL lipoproteins. These non-HDL complexes, which contain LDL as well as TRL, such as chylomicrons and their remnants, are more resistant to the PEG-modified CO and CE enzymes, resulting in selective catalytic activity in decreasing order of HDL-C . chylomicrons and very low-density lipoprotein . LDL. However, this selectivity is not absolute, leading to concern that ‘‘direct’’ methods may overestimate HDL-C as the result of nonspecific reaction between the HDL-C reagents and the cholesterol content of non-HDL particles, particularly TRLs. The manufacturers state on the package insert that ‘‘to date, there is no model available which can mimic interference by triglycerides’’ and that ‘‘elevated concentrations of free fatty acids and denatured lipoproteins may cause falsely elevated HDL-cholesterol results.’’ It follows from the reciprocal relationship between TG and HDL-C in vivo that TRL levels are increased in precisely those patients in whom HDL-C is likely to be reduced, such as those with insulin resistance or increased risk of CVD.8 Furthermore, overestimation of HDL-C inevitably leads to underestimation of LDL-C when the Friedewald equation is used. This combination of analytical overestimation of HDL-C and underestimation of LDL-C results in underestimation of absolute risk of CVD in patients with this pattern of dyslipidemia. We compared the ‘‘direct’’ HDL-C measurement methods (HDL-C generation first, second, or third) against the same assay performed after the removal of TRL from the sample by ultracentrifugation. We calculated the difference between HDL-C results before and after this procedure and examined the effect of intercurrent TG levels. Analysis of variance
Journal of Clinical Lipidology, Vol 4, No 4, August 2010 nonlinear multiregression analysis was used to generate corrective formulae to reduce the positive bias introduced by TRL. The application of assay-specific corrections could provide more accurate results after the measurement of HDL-C in hypertriglyceridemic patients by the use of homogenous commercial ‘‘direct’’ automated assay systems.
Materials and methods Three sets of consecutive patients (male and female, aged .18 years) who attended an ambulatory CVD riskprevention clinic provided venous samples after a 12-hour fast. HDL-C was measured by use of the generations of HDL-C assay that were then current. Measurements were performed before and after removal of TRL by density ultracentrifugation (d , 1.006) of the samples (HDL-UC). Studies were conducted between January 2000 and June 2009. Additional analysis of the subgroups of patients with fasting ,600 mg/dL was performed. The HDL-C was measured ‘‘directly’’ with the use of a Hitachi-917 or Modular-PPE auto analyzer (RocheÒ Diagnostics, Basel, Switzerland). Reagents included a-cyclodextrin for the HDL-C1st assay and PEG-modified enzymes but not a-cyclodextrin for HDL-C2nd and HDLC3rd assays. Identical direct HDL-C measurements were performed before and after removal TRL and the recovery of HDL as a percentage (HDL-UC/ HDL-C) was compared with that of TG and total cholesterol in the same samples. We calculated the difference between HDL-C and HDLUC results and used Glick interferographs (difference of analytes in mg/dL before and after separation on the y-axis plotted against the likely interferant on the x-axis) to examine the effect of endogenous TG. Nonlinear component regression was performed because TG has a nonlinear effect on the accuracy of direct HDL-C measurements. Equations were generated for each assay of HDL-C to predict the HDL-UC.
Results The HDL-C was measured before and after removal of TRL, and the two sets of results were compared for each generation of the assay. The ‘‘direct’’ HDL-C measurement was consistently high compared with the corresponding HDL-C measurements after ultracentrifugation (HDL-UC). This finding was regarded as a positive bias associated with the direct HDL-C methods because the d , 1.006 fraction contained TRL, but no HDL, on electrophoresis (data not shown). The difference between ‘‘direct’’ HDL-C and HDL-UC measurements increased progressively with increasing TG levels. However, HDL-UC was negatively correlated with TG (HDL-UC 5 20.112 TG 1 1.458 mmol/L, R2 5 0.230, and HDL-UC 5 20.684 TG 1 315 mg/dL, R2 5 0.194). The HDL-C difference (HDL-C 2 HDL-UC) in mg/dL was examined according to increasing TG concentrations by the use of Glick interferographs. Increasing TG
Perera et al
Adjustment of direct HDL-C measurements
concentrations exacerbated interference (Fig. 1, left panels). Even the mild-to-moderate hypertriglyceridemia (,600 mg/dL) that is frequently associated with insulin resistance or the metabolic syndrome caused a substantial level of interference, represented by reductions in HDL-C % recovery, calculated as ([HDL-UC]/[HDL-C] ! 100) of 10% to 40% (Fig. 2). Conversely, recovery of TG or total cholesterol was unimpaired, suggesting that the ultracentrifugation process did not result in loss of recovery of any class of lipoproteins (data not shown). To adjust for the interference caused by TRL cholesterol, analysis of variance multiregression analysis was used to generate prediction formulae for HDL-UC for each
307 assay (Table 1; adjusted R2 0.98–0.99). Contrasting Glick interferographs were plotted to demonstrate the reduced inaccuracy when applying the predictive HDL-C formula for each of the three generations of the HDL-C assay (Fig. 1, right panels). Regression formulae differed when more severely hypertriglyceridemic patients (.600 mg/dL) were excluded (Tables 1 and 2).
Discussion These results suggest that the current automated ‘‘direct’’ HDL-C measurement methods that we studied experience
Figure 1 Plots of pair-wise comparisons of the differences in measured HDL-C before and after ultracentrifugation for generation 1, 2, and 3 HDL-C assays versus the intercurrent TG levels in mg/dL (left) showing positive interference with increasing TG concentrations. The corresponding difference between the calculated HDL-C when the prediction formulae was used and the HDL-C after ultracentrifugation for generation 1, 2, and 3 assays versus the intercurrent TG levels in mmol/L for the same samples (right) showing correction for TG interference. The solid line denotes the forecast trend lines for each data set.
308
Journal of Clinical Lipidology, Vol 4, No 4, August 2010 interference from TRL was described in some, but not all studies.13-16 Because methods involving a-cyclodextrin have been superseded, two more recent studies of ‘‘direct’’ HDL methods have reported positive biases that were attributed to TRL or to the presence of diabetes.17,18 Previous HDL-C methods involved isolation of HDL-C by precipitation of apolipoprotein B-containing TRL particles. In this study we used an identical HDL-C measurement technique (‘‘direct’’ HDL-C assays) before and after the ultracentrifugal removal of the TRL that could potentially interfere with the estimation. Loss of HDL-C was excluded by complete recovery of cholesterol and TG, irrespective of TG level, suggesting that the decrease in observed HDL-C is caused by the removal of d , 1.006 lipoproteins after density separation. Our method would not detect additional interference from TG-containing lipoproteins occupying the 1.006 , d , 1.019 range. Overestimation of HDL-C leads to underdiagnosis of the metabolic syndrome and insulin resistance, as well as underestimation of LDL-C according to the Friedewald equation. These combined effects result in underestimation of absolute risk of CVD, impairing the opportunity to effectively identify and treat patients at risk on the basis of metabolic risk factors. HDL-UC overcomes this problem by selectively removing interfering TRL, but this method is time-consuming and labor-intensive. Manufacturers of ‘‘direct’’ HDL-C assays have acknowledged that TRL cholesterol may interfere with HDL-C analysis and that TG interference studies are based on the use of Intralipid,
Figure 2 HDL-C recovery as a percentage (Recovery %) , calculated as ([HDL-UC]/[HDL-C] ! 100) vs the TG in mg/dL measured in the same samples from subjects with TG , 600 mg/dL. The reduced recovery% is caused by positive interference on the direct HDL-C measurements before removal of TRL particles by ultracentrifugation.
interference from the cholesterol content of TRL, causing a correctable positive bias. This finding may be a ‘‘class-effect’’ that applies to other HDL-C methods, particularly those in which physical separation of HDL does not occur. The problem appeared more marked with the second-generation assay in which a-cyclodextrin had been removed, but this was partially compensated with the third-generation assay, perhaps because of the increase in dextran sulfate concentration. Results from method comparison studies before the year 2000 suggested good agreement between ‘‘separation’’ HDL-C methods and the reference method, even in the presence of Intralipid (Kabivitrum Inc., Alameda, CA) or TRL.9-12 Where positive bias occurred, it was attributed to incomplete precipitation with the comparator method or the presence of apolipoprotein E-containing HDL, but it should be noted that the sources of triglyceride (TG) used in these studies may have had relatively low cholesterol content. Direct methods usually involved the use of a-cyclodextrin at this time, and positive
Table 1 HDL-C prediction formulae for generation 1, 2, and 3 assays using ANOVA multiregression analysis for all subjects (TG range up to 3000 mg/dL) Prediction formula for HDL-C, mmol/L
Number of observations
R2
Predicted HDL-C1st 5 0.95 ! (HDL-C1st) – 1.73 ! (LN TG) mg/dL Predicted HDL-C1st 5 0.92 ! (HDL-C1st) – 0.05 ! (LN TG) mmol/L Predicted HDL-C2nd 5 0.88 ! (HDL-C2nd) – 2.99 ! (LN TG) mg/dL Predicted HDL-C2nd 5 0.87 ! (HDL-C2nd) – 0.04 ! (LN TG) mmol/L Predicted HDL-C3rd 51.02 ! (HDL-C3rd) – 2.65 ! (LN TG) mg/dL Predicted HDL-C3rd 5 0.95 ! (HDL-C3rd) – 0.07 ! (LN TG) mmol/L
201
0.995
101
0.987
105
0.986
ANOVA, analysis of variance; HDL-C, high-density lipoprotein cholesterol; HDL-C1st, generation 1 assay for HDL-C); HDLC2nd, generation 2 assay for HDL-C; HDLC3rd, generation 3 assay for HDL-C; LN, natural log; TG, triglyceride.
Perera et al
Adjustment of direct HDL-C measurements
Table 2 HDL-C prediction formulae for generation 1, 2, and 3 assays using ANOVA multiregression analysis in subjects with TG , 600 mg/dL (6.8 mmol/L). Prediction formula for HDL-C Predicted HDL-C1st 5 0.93 ! (HDL-C1st) – 0.86 ! (LN TG) mg/dL Predicted HDL-C1st 5 0.91 ! (HDL-C1st) – 0.03 ! (LN TG) mmol/L Predicted HDL-C2nd 5 0.89 ! (HDL-C2nd) – 1.92 ! (LN TG) mg/dL Predicted HDL-C2nd 5 0.86 ! (HDL-C2nd) – 0.07 ! (LN TG) mmol/L Predicted HDL-C3rd 5 0.996 ! (HDL-C3rd) – 1.85 ! (LN TG) mg/dL Predicted HDL-C3rd 5 0.943 ! (HDL-C3rd) – 0.04 ! (LN TG) mmol/L
Number of observations
R2
160
0.99
309 1 34.6 mg/dL, R2 5 0.9999). This contrasts the in vivo inverse relationship, which we confirmed in this cohort. We believe that QA materials should be spiked with combinations of lipoproteins in which the reciprocal relationship between HDL and TRL is maintained to ensure that automated HDL-C assays avoid positive interference from TRL cholesterol.
References
86
0.98
101
0.99
ANOVA, analysis of variance; HDL-C, high-density lipoprotein cholesterol; HDL-C1st, generation 1 assay for HDL-C); HDLC2nd, generation 2 assay for HDL-C; HDLC3rd, generation 3 assay for HDL-C; LN, natural log; TG, triglyceride.
which does not contain cholesterol. We have demonstrated that the progressive modifications of ‘‘direct’’ HDL-C assays have not completely overcome this problem. This is because the interference depends on the unpredictable variation in the TG levels in individual patient specimens. It is necessary for current HDL-C assays to be made more selective for HDL particles, free from interference by TRL cholesterol, while maintaining the efficiency and precision of automated methods. Until this is achieved, our study provides a means of correcting for this bias by the use of assay-specific adjustment formulae. The betweengeneration differences and those after exclusion of severe hypertriglyceridemia show that the interference is assay and range specific. Therefore, we encourage manufacturers of individual ‘‘direct’’ HDL-C assays to generate formulae that are appropriate for their assay, particularly in the TG range in which results are required for estimation of LDL-C. Current quality assurance (QA) materials are not representative of the typical metabolic syndrome phenotype. Materials for the RCPA General Serum Chemistry and Special Lipids external QA Programmes are spiked with increasing concentrations of both TRL and HDL, leading to a positive association between TG and HDL-C levels (TG 5 0.54 ! HDL-C 1 0.35 mmol/L and TG 5 0.60 ! HDL-C
1. Barter PJ. HDL cholesterol testing: implications for clinical management. Aust Presc. 1994;17:99–102. 2. Marques-Vidal P, Ferrario M, Kuulasmaa K, et al. Quality Assessment of Data on HDL cholesterol in the WHO Monica Project. Lyon, France: World Health Organization Publications; 1994. 3. Warnick GR, Nauck M, Rifai N. Evolution of methods for measurement of HDL-cholesterol: from ultracentrifugation to homogenous assays. Clin Chem. 2001;47:1579–1596. 4. Manual of Laboratory Operations. Lipid Research Clinics Program Lipid and lipoprotein analysis DHEW Publication Number (NIH) 75:628:59. Bethesda, MD: NIH; 1974. 5. Mahley RW, Weisgraber KH. An electrophoretic method for the quantitative isolation of human and swine plasma lipoproteins. Biochemistry. 1974;13:1964–1969. 6. Warnick GR, Benderson J, Albers JJ. Dextran sulfate-Mg21 precipitation procedure for quantitation of high-density-lipoprotein cholesterol. Clin Chem. 1982;28:1379–1388. 7. Burstein M, Scholnick HR. Lipoprotein-polyanion-metal interactions. Adv Lipid Res. 1973;11:67–108. 8. Phillips NR, Havel RJ, Kane JP. Serum apolipoprotein A-I levels: relationship to lipoprotein lipid levels and selected demographic variables. Am J Epidemiol. 1982;116:302–313. 9. Rifai N, Cole TG, Iannotti E, et al. Assessment of interlaboratory performance in external proficiency testing programs with a direct HDL-cholesterol assay. Clin Chem. 1998;44:1452–1458. 10. Nauk M, Marz W, Haas B, et al. Homogeneous assay for direct determination of high-density lipoprotein cholesterol evaluated. Clin Chem. 1996;42:424–429. 11. Harris N, Galpchian V, Rifai N. Three routine methods for measuring high-density lipoprotein cholesterol compared with the Reference Method. Clin Chem. 1996;42:738–743. 12. de Keijer MH, Elbers D, Baadenhuijsen H, et al. Evaluation of five different high-density lipoprotein cholesterol assays; the most precise are not the most accurate. Ann Clin Biochem. 1999;36:168–175. 13. Hubbard RS, Hirany SV, Devaraj S, et al. Evaluation of a rapid homogeneous method for direct measurement of high-density lipoprotein cholesterol. Am J Clin Pathol. 1998;110:495–502. 14. Okazaki M, Sasamoto K, Muramatsu T, et al. Evaluation of precipitation and direct methods for HDL-cholesterol assay by HPLC. Clin Chem. 1997;43:1885–1890. 15. Okada M, Matsui H, Fujiwara A. Direct measurement of HDL cholesterol: method eliminating apolipoprotein E-rich particles. J Clin Lab Anal. 2001;15:223–229. 16. Eglof M, Eglise D, Duvillard L, et al. Multicentre evaluation on different analysers of three methods for direct HDL-cholesterol assay. Ann Biol Clin. 1999;57:561–572. 17. Zhao W, Chaffin C, Desmond RA, et al. Overestimation of HDLcholesterol using a homogeneous ‘‘direct’’ assay. J Clin Lab Anal. 2004;18:42–44. 18. Saeed BO, Smart P, Keeka G, et al. Comparison of two direct methods for HDL cholesterol measurement with an indirect precipitation method in diabetic patients. Diab Nutr Metab. 2002;15:169–172.