Method of LDL cholesterol measurement influences classification of LDL cholesterol treatment goals: clinical research study

Abstract

Background: Low-density lipoprotein cholesterol (LDL-C) has been clearly associated with the risk of developing coronary heart disease. The best and most convenient method for determining LDL-C has come under increased scrutiny in recent years. We present comparisons of the Friedewald calculated LDL-C (C-LDL-C) and direct LDL-C (D-LDL-C) using 3 different homogenous assays. This highlights differences between the 2 methods of LDL-C measurement and how this affects the classification of samples into different LDL-C treatment goals as determined by the National Cholesterol Education Program Adult Treatment Panel III guidelines thus potentially affecting treatment strategies.

Methods: Lipid profiles of a total of 2208 clinic patients were retrieved from the Central Arkansas VA Healthcare System clinical laboratory database. Samples studied were of 1-week period during the 3 periods studied: 2000 (period 1), 2002 (period 2), and 2005 (period 3). Different homogenous assays for D-LDL-C measurement were used for each of the 3 periods.

Results: There is a fundamental disagreement between D-LDL-C and C-LDL-C, although Pearson correlation coefficients are 0.93, 0.97, and 0.98 for periods 1, 2, and 3, respectively. Using the model for period 1, when C-LDL-C is 70 mg/dL, the predicted D-LDL-C is 95 mg/dL (36% higher). The differences between C-LDL-C and predicted D-LDL-C progressively decrease at higher LDL-C cut points. In the assay used in period 3, there are 290 samples with D-LDL-C values between 100 and 130 mg/dL. Of these, only 182 samples show agreement with C-LDL-C values, whereas 90 samples with a D-LDL-C in the 100- to 130-mg/dL range are in the 70- to 100-mg/dL range using the C-LDL-C assay. Although the κ statistics suggests the LDL-C measures have relatively high levels of agreement, the significant generalized McNemar tests (P < 0.01) provide additional evidence of disagreement between C-LDL-C and D-LDL-C during all the 3 periods.

Conclusions: Our results highlight D-LDL-C measurements using 3 different assays during 3 different periods. In all assays, there is a substantial lack of agreement between D-LDL-C and C-LDL-C, which, in most cases, resulted in higher D-LDL-C values than C-LDL-C. This leads to clinically significant misclassification of patient's LDL-C to a different LDL-C treatment goal, which would potentially result in more drug usage, thus exposing patients to more potential adverse effects and at a much greater cost with little evidence of benefit.

Figure 1

Intercept and slope parameters estimated from linear regression models are shown for each time period (solid round symbols) with corresponding 99.99% confidence regions (indicated by the surrounding ellipses) A point representing the hypothetical regression parameters (β₀=0 and β₁=1) that one would expect if the two measures agreed is presented for comparison (solid square symbol). None of the 99.99% confidence regions contain this hypothetical point, providing evidence that D-LDL-C and C-LDL-C measures do not agree and that the extent of lack of agreement differs by time periods.

Figure 2

Graph of D-LDL-C and C-LDL-C difference with corresponding triglyceride values for the periods 2000 (solid black line), 2002 (dashed gray line) and 2005 (dashed black line) using 3 different assays. In the parenthesis are shown the mean and the standard deviation of D-LDL-C and C-LDL-C difference for the 3 corresponding assay results. This Figure illustrates that differences between D-LDL-C and C-LDL-C increase as triglyceride levels increase, most striking in the 2000 time period.