LDL Cholesterol: What Is the Best Way to Measure It?

In this issue of the journal, H. Ikezaki and colleagues, a group led by Dr. Ernst Schaefer, describe an interesting study in which they compare 2 direct LDL cholesterol (dLDL-C)2 tests and 2 equations for calculating LDL-C (Friedewald and Martin) for cardiovascular disease (CVD) risk prediction in the Framingham Offspring Study (1). Besides the important observation that 1 dLDL-C assay (Denka-Seiken) showed a much stronger association with CVD than the other dLDL-C test (Kyowa-Medex) or calculated LDL-C, the study raises several important issues that if resolved could improve our ability to predict CVD in the future. Although there is much controversy about the clinical utility of various CVD risk markers, there is almost universal agreement that LDL plays a central role in the pathogenesis of atherosclerosis. Despite this agreement, the pooled-cohort risk equations for calculating CVD risk in the most recent 2018 ACC/AHA guidelines (2) use total cholesterol and HDL cholesterol (HDL-C) as the only 2 lipid measures for CVD risk. Age, sex, gender, race, blood pressure, hypertension treatment, diabetes, and smoking status are the other inputs into the equation, but age is the predominant factor. LDL-C over 190 or <70 mg/dL are also used to identify at the outset very high-risk and low-risk individuals, who do not need to have their CVD risk calculated. If LDL is so central to the development of CVD, why then is LDL-C not used more directly to calculate CVD risk? As will be discussed below and as described in the report by H. Ikezaki et al. (1), analytical issues related to the accurate measurement of LDL-C constitute one likely explanation but probably not the only one. In this study of 3147 participants from Framingham Offspring Study with a median follow-up of 16 years, there were 677 CVD events. By univariate analysis, all the traditional CVD risk markers were associated with CVD, but after multivariate adjustment, with standard risk factors, nonHDL-C, and triglycerides, only the Denka-Seiken dLDL-C test was strongly associated with CVD (hazard ratio, 1.33–1.53, depending on model and end point). This finding was true whether CVD was defined more broadly as inclusive CVD or just as hard CVD (myocardial infarction plus stroke) or as hard CVD plus coronary revascularization. Based on net reclassification, the Denka-Seiken assay correctly reclassified approximately 6% of patients, suggesting that it would be a valuable addition to CVD risk assessment. High-sensitivity C-reactive protein, as has been previously described, also improved net reclassification to a similar degree. In contrast, the Kyowa-Medex dLDL-C test and the 2 calculated LDL-C tests, which all showed a weaker association with CVD by univariate analysis, were not associated at all with CVD after multivariate adjustment. What could be the explanation? As described in this report, 1 possibility for the calculated LDL-C tests is their poorer accuracy. It has been known since its inception that the weak link in the Friedewald equation (3) is the estimate of very low-density lipoprotein cholesterol (VLDL-C), which is calculated by dividing triglycerides by 5. For this reason, this calculation is not recommended for triglycerides more than 400 mg/dL, but even lower concentrations of triglycerides can cause significant errors. Recently, Martin et al. (4) have described a new equation that uses an optimized triglyceride “factor” or denominator that differs depending on triglyceride and nonHDL-C concentrations. When compared to β quantification, an ultracentrifugation–precipitation reference method, it is more accurate than the Friedewald equation, but it is still only an estimate of LDL-C. Like the Friedewald equation, it also depends on the measurement of total cholesterol, triglycerides, and HDL-C, which all contribute to its inaccuracy. Thus, the inaccuracy of estimating LDL-C by various equations could account for its poorer association with CVD and the reason that LDL-C is not used more directly in CVD risk estimation, even though calculation of LDL-C is still in widespread use. Direct lipoprotein assays, which are homogeneous assays that can work directly on serum or plasma, were developed over 10 years ago. Direct HDL-C assays are universally performed by routine clinical laboratories, because the test is fully automated and thus avoids the labor-intensive manual precipitation step that was previously used to remove LDL before measuring HDL-C. dLDL-C tests have also been developed, but they are not as widely used, because of the extra cost to do the assay compared to the free LDL-C equation and because of the concerns about their analytical performance. Several studies have shown that although dLDL-C tests work well on normolipidemic individuals, they often do not match well with LDL-C as measured by β quantification on dyslipidemic individuals (5, 6). The 2 dLDL-C assays used in this study have been directly compared, and the Denka-Seiken assay appears to be more accurate on dyslipidemic samples (6), which could account for the findings in this study. The other possibility is that the accuracy of the 2 direct assays could have been differentially affected by the fact that samples tested in this study were frozen. All dLDL-C assays in use today have been shown to have poor commutability on frozen specimens (7), which has hampered standardization efforts for LDL-C. So another possibility is that the accuracy of the Kyowa-Medex dLDL-C test was more adversely affected from freezing than the Denka-Seiken assay. It is also important to point out that LDL-C is not synonymous with LDL and there are other measures of LDL, such as apoB and LDL-particle number, which when discordant with LDL-C are more predictive of CVD (8, 9). Even nonHDL-C, which includes cholesterol not only on LDL but also on proatherogenic triglyceride-rich lipoproteins, is also usually more predictive of CVD than LDL-C (10). Strong arguments can be made for shifting our focus away from LDL-C to other measures of LDL, but so far this has not happened to a great degree, although the 2018 American College of Cardiology/American Heart Association guidelines now endorse the use of apoB as a risk enhancer for intermediate risk patients (1). It is often not recognized that the measurement of cholesterol on LDL creates an inherent bias that favors larger LDL particles because of their greater cholesterol-carrying capacity. This knowledge is important, because it has been known for many years that smaller LDL particles are more closely associated with CVD, possibly because they can more readily infiltrate into the vessel wall and have an increased propensity for oxidation (11). Small dense LDL particles are commonly found in the so-called “atherogenic phenotype,” which typically occurs in patients with metabolic syndrome and hypertriglyceridemia, which drives down LDL-C but increases other proatherogenic particles like small dense LDL. It has been difficult to translate these findings into a diagnostic test, because traditionally small dense LDL were separated by ultracentrifugation, but recently a direct small dense LDL-C test has been developed by Denka-Seiken. It has been shown to be more strongly associated with CVD than LDL-C in at least 3 well-known clinical studies, namely the Framingham Offspring Study (12), Multi-Ethnic Study of Atherosclerosis (13), and Atherosclerosis Risk in Communities Study (14), and was recently approved by the Food and Drug Administration. Interestingly, the total dLDL-C by Denka-Seiken has a known bias and underestimates cholesterol on large LDL size subfractions by almost 30% but fully measures cholesterol on midsize and small LDL particles (15). Hence, this assay is possibly superior to other total measures of LDL-C as a CVD risk marker because of its increased selectivity toward smaller LDL size subfractions, but this superiority remains to be proven. So how do we move forward? As the authors point out, this relatively small study of mostly whites in the US should be replicated in larger prospective studies containing different populations. Ideally, future studies should be done on fresh specimens to rule out the possibility of artifacts from freezing. A particular emphasis should be placed on discordant results between the different measures of LDL-C to get possible clues on why 1 LDL-C assay is performing better than the other for CVD risk prediction. It will also be important to continue to compare the Denka-Seiken dLDL-C assay to other measures of LDL, including small dense LDL, to see if it outperforms these assays. Because it will likely take a lot effort and several years before there would be any change to our current pooled-cohort risk equations, it would be useful in the meantime to determine if the Denka-Seiken assay or other dLDL-C tests decrease the misclassifications of patients, using the 3 current LDL cutpoints (70, 100, and 190 mg/dL) recommended by the American Heart Association and American College of Cardiology for managing patients (1). To do this, the dLDL-C assays should be compared to β quantification. Differences from β quantification should be carefully investigated because although it is considered a “gold standard” reference method, it can include cholesterol from remnant lipoproteins and other lipoprotein particles and thus has its own limitations. One of the early pioneers in lipoprotein research, Dr. Fredrickson, once said, “In all fields of science, technological changes have dictated the forward movement in research” (16). Hopefully, the recent technological advances related to the direct measurement of LDL-C and its subfractions will in the future provide us another test for the prediction of CVD. direct LDL cholesterol cardiovascular disease HDL cholesterol.