1. © Copyright 2009 by the American Association for Clinical Chemistry Seven Direct Methods for Measuring HDL and LDL Cholesterol Compared with Ultracentrifugation Reference Measurement Procedures W. Greg Miller, G. L. Myers, I. Sakurabayashi, L. M. Bachmann, S. P. Caudill, A. Dziekonski, S. Edwards, M. M. Kimberly, W. J. Korzun, E. T. Leary, K. Nakajima, M. Nakamura, G. Nilsson, R. D. Shamburek, G. W. Vetrovec,G. R. Warnick, and A. T. Remaley June 2010 http://www.clinchem.org/cgi/reprint/56/6/977 © Copyright 2010 by the American Association for Clinical Chemistry Journal Club

2. © Copyright 2009 by the American Association for Clinical Chemistry Introduction  The objective was to evaluate the performance of direct (homogeneous) methods for measurement of HDL Cholesterol (HDL-C) and LDL Cholesterol (LDL-C) for the following characteristics: Trueness of calibration traceability to the reference measurement procedures Accuracy of individual sample measurements Imprecision of the measurement procedures Specificity for the lipoproteins intended to be measured

3. © Copyright 2009 by the American Association for Clinical Chemistry Introduction (cont)  Challenges for Direct HDL-C and LDL-C Methods Measure the same lipoprotein fractions that are measured by HDL-C and LDL-C reference measurement procedures (RMPs) Not measure anything else Measure accurately for a wide range of clinical conditions with lipoprotein and other metabolic abnormalities

4. © Copyright 2009 by the American Association for Clinical Chemistry Question  What experimental design will ensure the performance of different direct methods will be evaluated correctly?

5. © Copyright 2009 by the American Association for Clinical Chemistry Materials and Methods  Reference Measurement Procedures Performed at the CDC Ultracentrifugation of serum Density >1.006 g/mL “bottom fraction” Cholesterol measured by Abell-Kendall (AK) RMP HDL was isolated by heparin manganese precipitation of apo-B lipoproteins (includes LDL fractions) HDL-C measured in supernatant by AK RMP LDL-C calculated as difference between bottom fraction cholesterol and HDL-C (beta-quantification)

6. © Copyright 2009 by the American Association for Clinical Chemistry Materials and Methods (cont)  Direct measurement procedures Reagent formulations from 7 primary manufacturers that are distributed worldwide under various trade names Performed simultaneously with a Roche Hitachi 917 analyzer using reaction parameters, calibrators, and controls provided by each reagent manufacturer Four frozen serum pools provided by the Centers for Disease Control and Prevention (CDC) were measured at the beginning and end of each run for each method

7. © Copyright 2009 by the American Association for Clinical Chemistry Materials and Methods (cont)  Sample collection and measurements Blood was collected at Virginia Commonwealth University (VCU) and the National Institutes of Health Plain glass tubes, clotted 60-90 minutes at room temperature, centrifuged to recover serum Serum aliquots were stored at 4 ºC until measured Shipped overnight on cold packs to CDC and VCU Direct methods (at VCU) and reference methods (at CDC) were performed on the same day within 24-48 hours of blood collection

8. © Copyright 2009 by the American Association for Clinical Chemistry Question  What kinds of patients need to be included to determine if the direct methods are fit for their intended purpose?

9. © Copyright 2009 by the American Association for Clinical Chemistry Participant Characteristics

10. © Copyright 2009 by the American Association for Clinical Chemistry Figure 1. Box and whisker plot of the differences in percent between the direct and RMP results for HDL-C for each direct method (abbreviations: D is diseased group, N is non-diseased group, De is Denka, Ky is Kyowa, Ro is Roche, Sr is Serotek, Sk is Sekisui, Sy is Sysmex, Um is UMA, Wa is Wako). The median is the center line, the ends of the box represent 25th and 75th percentiles, the end of the lines extend to the 10th and 90th percentiles and individual results are shown beyond the lines. Results

11. © Copyright 2009 by the American Association for Clinical Chemistry Figure 2. Box and whisker plot of the differences in percent between the direct and RMP results for LDL-C for each direct method (same abbreviations as in Figure 1). The median is the center line, the ends of the box represent 25th and 75th percentiles, the end of the lines extend to the 10th and 90th percentiles and individual results are shown beyond the lines. Results

12. © Copyright 2009 by the American Association for Clinical Chemistry Question  How can the major measurement error components that contribute to total error be determined?

13. © Copyright 2009 by the American Association for Clinical Chemistry Error Component Analysis Possible error components were fit to the following model for ln(concentration): y is the measured value µ is the expected value with the RMP The other factors are error components determined from different aspects of the experimental design and explained briefly in the following tables (see supplemental data for details)

14. © Copyright 2009 by the American Association for Clinical Chemistry Table 1. Error Components for HDL-C Method CV b, %, interassay, frozen pools CV e, %, intraassay, patient samples CV d, %, patient sample-specific effects CV tot, %, combined random effects of CV b, CV e, CV d Mean bias, % % of results outside of upper and lower TE limits HDL-C, nondiseased group Denka1.41.22.32.94.010.4 Kyowa0.9 3.53.72.510.4 Roche1.4 3.84.3−2.4−10.4 Sekisui0.91.13.13.4−1.7−8.2 Serotec1.91.24.24.8−4.8−13.4 Sysmex2.31.01.83.1−5.4−10.9 UMA1.11.25.86.00.713.6 Wako1.71.31.42.64.810.5 HDL-C, diseased group Denka1.28.28.40.418.8 Kyowa1.87.88.12.120.0 Roche2.07.78.1−3.1−17.5 Sekisui1.25.96.1−5.2−16.0 Serotec1.78.69.0−3.0−18.9 Sysmex1.66.16.7−8.6−19.8 UMA1.916.316.4−1.936.3 Wako1.26.06.48.824.0

15. © Copyright 2009 by the American Association for Clinical Chemistry Table 2. Error Components for LDL-C Method CV b, %, interassay, frozen pools CV e, %, intraassay, patient samples CV d, %, patient sample-specific effects CV tot, %, combined random effects of CV b, CV e, CV d Mean bias, % % of results outside of upper and lower TE limits LDL-C, nondiseased group Denka1.32.85.46.20.213.5 Kyowa0.7 3.23.3−1.1−7.5 Roche0.81.63.33.8−6.8−13.3 Sekisui1.21.53.84.2−0.7−8.8 Serotec2.91.40.03.2−6.2−11.9 Sysmex2.30.73.44.2−6.0−13.3 UMA2.20.91.22.6−0.15.3 Wako0.61.82.02.81.16.8 LDL-C, diseased group Denka2.210.510.8−1.522.3 Kyowa1.19.69.7−0.820.4 Roche1.310.010.1−6.3−23.3 Sekisui2.06.06.4−1.7−13.5 Serotec1.39.09.5−11.8−26.6 Sysmex0.910.811.1−7.8−25.9 UMA1.513.814.1−0.431.9 Wako1.86.06.34.118.2

16. © Copyright 2009 by the American Association for Clinical Chemistry Discussion  For Non-Diseased Patients: 6 out of 8 HDLC methods met the National Cholesterol Education Program (NCEP) total error goal of results within 13% of the RMP (see paper for criteria for concentrations <42 mg/dL) 5 out of 8 LDLC methods met the NCEP total error goal of results within 12% of the RMP  For Diseased Patients: None of the HDL-C and LDL-C methods met the NCEP total error goals

17. © Copyright 2009 by the American Association for Clinical Chemistry Discussion (cont)  Examination of Error Components Showed: Measurement inter- and intra-assay imprecision was good (CVs <2.8%) for all methods Sample specific effects dominated the random error components in 6 out of 8 HDL-C and LDL-C methods for the non-diseased group Sample specific effects dominated the random error components in all HDL-C and LDL-C methods for the diseased group Sample specific effects were much larger for the diseased than for the non-diseased groups

18. © Copyright 2009 by the American Association for Clinical Chemistry Question  What causes sample specific effects in direct HDL-C and LDL-C measurements?

19. © Copyright 2009 by the American Association for Clinical Chemistry Sample Specific Effects May Be Caused By:  Any component in the serum that influences the specificity of the reagents for the HDL-C or LDL-C range of lipoprotein particles: Increased, or decreased, amount of a lipoprotein relative to the distribution of lipoproteins present in non- diseased conditions Increased amount of a lipoprotein normally not present, or present in very low concentration, in non-diseased conditions Presence of a protein or other molecule that alters the reagent’s ability to segregate the HDL or LDL lipoproteins for measurement of cholesterol content

20. © Copyright 2009 by the American Association for Clinical Chemistry Discussion (cont)  Examination of Error Components Showed: Mean bias exceeded the NCEP 5% goal for 3 out of 8 HDL-C methods (diseased group) Mean bias exceeded the NCEP 4% goal for 3 out of 8 LDL-C methods (both non-diseased and diseased groups)  Bias was Correlated with Triglycerides (TG) Concentration (see Paper for Details) TG reflects abnormalities in lipoprotein composition that may influence the specificity of direct methods

21. © Copyright 2009 by the American Association for Clinical Chemistry Conclusions  The direct methods performed reasonably well for the non-diseased group  All direct methods had unacceptable total error for the diseased group  Inadequate specificity for the lipoprotein fractions measured by the RMPs was a major limitation  TG concentration may be a surrogate for the presence of abnormal lipoproteins that may influence direct methods

22. © Copyright 2009 by the American Association for Clinical Chemistry Conclusions (cont)  In many cases the differences between the direct method results and RMPs were sufficiently large that they could affect the clinical management of patients  Several patients with a genetic disorder in lipid metabolism had direct method results that differed so markedly from the RMPs that they could have been misdiagnosed  Inquiries regarding discrepancies between direct method results and clinical observations should be followed up with an ultracentrifugation method