Calculated low-density lipoprotein cholesterol (cLDL-C) may differ from direct measurement (dLDL-C), and this difference may depend on presence of small, dense LDL (sdLDL) particles in addition to variation in triglycerides (TG) and high-density lipoprotein cholesterol (HDL-C) concentrations. The presence of such dependence would offer a simple means to estimate sdLDL. We studied dependence of sdLDL on cLDL-C, dLDL-C, and other variables.
We measured the levels of glucose, creatinine, total cholesterol, TG, HDL-C, and dLDL-C using standardized methods in 297 samples. For sdLDL cholesterol (sdLDL-C), a novel homogeneous assay was used. The cLDL-C was calculated using the Friedewald formula for 220 subjects after excluding for liver or renal disease. Using stepwise regression analysis identified non–HDL-C, cLDL-C, and dLDL-C as significant variables (P < .001; R2 = 0.88). The regression equation was as follows: sdLDL-C (mg/dL) = 0.580 (non–HDL-C) + 0.407 (dLDL-C) – 0.719 (cLDL-C) – 12.05. The sdLDL-C concentration can be estimated from non–HDL-C, dLDL-C, and cLDL-C values. Identification of a simple, inexpensive marker for sdLDL particles provides a cost-effective method for screening cardiovascular disease risk.
Lipid tests, including triglycerides (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C), are well accepted and widely used for the assessment of cardiovascular disease (CVD) and its equivalents, including peripheral vascular disease, abdominal aortic aneurysm, and ischemic cerebral vascular disease.1 Plasma lipoproteins consist of heterogeneous subclasses of particles with varying density, size, electrophoretic mobility, relative lipid-protein proportions, and binding affinity.2 LDL particles are fractionated according to size and density into large, buoyant LDL (lbLDL; diameter ≥25.5 nm) and small, dense LDL (sdLDL; diameter <25.5 nm). The predominance of sdLDL correlates directly with the serum levels of TG and inversely with the serum levels of HDL-C.3 This combined lipid abnormality constitutes the “atherogenic lipoprotein phenotype.” which has an important role in the development of atherosclerosis.
Numerous epidemiologic and pathologic studies have indicated that an increased sdLDL level resulting from changes in abnormality of lipoprotein metabolism closely associates with increased risk of CVD and cerebrovascular disease.3–7 The National Cholesterol Education Program (NCEP)1 and the recently updated guidelines of the National Academy of Clinical Biochemistry expert panel8 accept the predominance of sdLDL as an emerging cardiovascular risk factor. In addition, atherogenic dyslipidemia, as demonstrated by sdLDL, is closely associated with the metabolic syndrome and insulin resistance.9,10 Overproduction of the hepatic TG-enriched large very-low-density lipoprotein (VLDL) causing high generation of sdLDL is an important and early complication of hepatic insulin resistance.11,12 Because insulin resistance has a central role in the pathophysiology of metabolic syndrome, sdLDL measurement may represent a useful clinical tool for characterizing this complex syndrome and its progression, such as type 2 diabetes, fatty liver disease, polycystic ovary syndrome, and some types of cancer.13
Several methods have been developed for the assessment of sdLDL particles, such as density gradient ultracentrifugation,14 gradient gel electrophoresis,15 tube gel electrophoresis,16 and nuclear magnetic resonance.17 The methods can be expensive, time-consuming, and technically demanding, making them too laborious for routine clinical practice or screening a large population. Recently, Denka Seiken developed a simple method to measure sdLDL-cholesterol (sdLDL-C) by a novel homogeneous enzymatic assay (Denka Seiken, Tokyo, Japan).18 Although this method is easier to implement and has the potential for daily clinical use, the reagent cost may be prohibitively expensive for general or screening use.
LDL-C can be determined by calculation using the Friedewald formula (cLDL-C)19 and directly measured (dLDL-C) by using specially designed assays.20,21 With the Friedewald equation, cLDL-C (in mg/dL) = TC – (HDL-C) – (TG/5); a value of TG dividing by 5 represents the VLDL-C concentration. An overproduction of the TG-enriched large VLDL causing high generation of sdLDL might lead to overestimation of the VLDL-C and underestimation of the cLDL-C concentrations. The difference between the cLDL-C and dLDL-C has been ascribed to variation in TG, HDL-C, and, potentially, the presence of sdLDL. We hypothesized that one could use the difference between the calculated and directly measured LDL to estimate the sdLDL-C. If the sdLDL-C concentration can be estimated from the classic lipid measures, it would provide a cost-effective method for screening patients for the risk of CVD.
In this study, we developed an equation for the estimation of sdLDL-C by using lipids usually measured in routine clinical laboratories.
Materials and Methods
Biochemical Characteristics of Samples
The study was approved by the institutional review board committee of the Faculty of Medicine, Ramathibodi Hospital, Mahidol University, Bangkok, Thailand. A total of 336 patients (140 men and 196 women) attending the outpatient clinics of Ramathibodi Hospital were enrolled in the study. A blood sample was obtained from each subject after 10 to 12 hours of overnight fasting. By using lipoprotein electrophoresis (product No. 4124, Sebia, Norcross, GA), we excluded the serum samples from 39 patients owing to the presence of chylomicrons or a TG level of 400 mg/dL (4.52 mmol/L) or more.
Classification of Chronic Kidney Disease
The renal function status of individual patients was classified by using the estimated glomerular filtration rate (eGFR) into 5 stages of renal dysfunction ranging from 1 to 5: normal, minimally impaired, moderately impaired, severely impaired, and failure. The eGFR was calculated from the serum creatinine concentration (traceable to the isotope dilution mass spectrometry reference method), age, and sex according to the new Chronic Kidney Disease Epidemiology Collaboration equation.22
Classification of Glucose Metabolism
The status of glucose metabolism of individual patients was classified by using the plasma glucose level. Fasting glucose categories were defined according to the American Diabetes Association criteria for the diagnosis of diabetes: normal glucose regulation (<100 mg/dL [5.55 mmol/L]), impaired fasting glucose level (100 to <126 mg/dL [5.55–7.00 mmol/L]), and diabetes mellitus (≥126 mg/dL [7.00 mmol/L]).23
All serum samples were analyzed for creatinine, TC, TG, HDL-C, dLDL-C, and sdLDL-C within 3 hours after sampling and for lipoprotein electrophoresis within 1 day, with the samples refrigerated between 2°C and 8°C. Sodium fluoride plasma samples were used for glucose analysis.
The levels of glucose, creatinine, TC, TG, HDL-C, and dLDL-C were measured on the Siemens Dimension RxL Max by using the Siemens enzymatic methods (Siemens Medical Solution Diagnostics, Tarrytown, NY). Our laboratory is standardized for the determination of TC, TG, and HDL-C by the Centers for Disease Control and Prevention-National Heart, Lung, and Blood Institute Lipid Standardization Program. The accuracy and precision of the measurements during the study were within the acceptable criteria of the National Cholesterol Education Program (NCEP).24
For the dLDL-C assay (Siemens Medical Solution Diagnostics), the method uses a reagent 1 containing a detergent that solubilizes only non-LDL particles. The cholesterol released is consumed by cholesterol esterase and cholesterol oxidase in a non–color-forming reaction. The second detergent contained in reagent 2 solubilizes the remaining LDL particles. The soluble LDL-C is then oxidized by the action of cholesterol esterase and cholesterol oxidase forming cholestenone and hydrogen peroxide. The enzymatic action of peroxidase on hydrogen peroxide in the presence of N, N-bis (4-sulfobutyl)-m-toluidine, disodium salt, and 4-aminoantipyrine generates a colored product.
For sdLDL-C, a novel homogeneous enzymatic assay (Randox Laboratories, Antrim, Northern Ireland) was used. This assay uses 2 liquid, ready-to-use reagents containing a polyoxyethylene benzylphenyl ether selectively decomposing chylomicrons, VLDL, and HDL; a polyoxyethylene distyrenelphenyl ether selectively binding to sdLDL to protect it from the action of enzymes; and sphingomyelinase possessing higher affinity to LDL larger than sdLDL. These ingredients eliminate the lipoproteins other than sdLDL. At the first step, cholesterol from such non-sdLDL lipoproteins is degraded to water and oxygen by enzymatic reactions. At the second step, cholesterol is released from sdLDL and leads to color development. We developed the user-defined sdLDL-C method on the Dimension RxL Max, performed according to the manufacturer’s specifications.
We calculated the cLDL-C (in mg/dL) by using the Friedewald formula: cLDL-C = TC – HDL-C – (TG/5). We estimated the lbLDL-C by subtracting the sdLDL-C concentration from the dLDL-C concentration.
Statistical analysis was carried out using SPSS software, version 12.0 (SPSS, Chicago, IL). All data for individual patients by sex are given as number (percentage) of men, whereas those for biochemical tests are given as the mean and SD. To study the effects of sex, age, renal function, and glucose metabolism on the sdLDL-C concentrations, we used univariable analysis for the assessment of the correlations between the variables. The stepwise multivariate regression analysis was performed to elucidate factors related to sdLDL-C concentrations in all subjects. We built 2 models of multivariate regression analysis: Model 1 included age, glucose, creatinine, TC, TG, HDL-C, cLDL-C, and dLDL-C. Model 2 was identical to model 1 except that we substituted non–HDL-C (TC minus HDL-C) for TC and HDL-C. We examined the influence of factors such as sex, age, renal function, and glucose metabolism on the relationship between measured and calculated sdLDL-C by comparing the slopes and intercepts of the regression equations for the individual subgroups. Outcomes were considered statistically significant when P values were less than .05.
Characteristics of the Study Samples
A total of 297 samples (from 115 men and 182 women) were included in the study. Biochemical characteristics of individual samples are summarized in Table 1. The mean age of the patients was 59.7 years. The TG, TC, HDL-C, dLDL-C, and sdLDL-C concentrations among the patients in the present study were 37 to 397 (0.42–4.49 mmol/L), 76 to 575 (1.97–14.90 mmol/L), 18 to 112 (0.47–2.91 mmol/L), 36 to 522 (0.92–13.52 mmol/L), and 7–187 mg/dL (0.18 to 4.84 mmol/L), respectively.
Table 2 shows univariable analyses of sdLDL-C. The mean values for sdLDL-C were unaffected by sex (P = .091) but were affected by age group (P = .015), stratified by decade. The presence or absence of renal dysfunction and the impaired fasting plasma glucose level did not statistically affect the sdLDL-C concentration (P > .08). However, the estimated marginal means of the sdLDL-C level were likely elevated with increasing degrees of impaired glucose metabolism.
Multiple Regression Analyses for sdLDL-C Concentrations
The presence of liver and kidney diseases may interfere with dLDL-C and HDL-C assays. To reduce the effect of both diseases, we excluded the serum samples with alanine aminotransferase concentrations more than twice the upper limit of normal and an eGFR less than 60 mL/min/1.73 m2. The total number of patient samples was 220 (from 78 men and 142 women).
Multiple regression analysis with sdLDL-C concentration as the dependent variable and with age, glucose, creatinine, TC, TG, HDL-C, cLDL-C, and dLDL-C as independent variables was performed. The stepwise regression analysis identified TC, HDL-C, cLDL-C, and dLDL-C (model 1) as significant variables (P < .001; R2 = 0.879) and the standard error of the estimate as 9.18 mg/dL (0.238 mmol/L), as shown in Table 3. The best fit of the linear regression equation was as follows:
The coefficients for TC and HDL-C were near the same magnitude but with the opposite signs. Therefore, we substituted non–HDL-C for the independent variables TC and HDL-C. By using non–HDL-C (model 2), the stepwise regression analysis identified non–HDL-C, cLDL-C, and dLDL-C as significant variables (P < .001; R2 = 0.878 and the standard error of the estimate as 9.206 mg/dL (0.238 mmol/L; Table 3). The linear regression equation was as follows:
In addition, non–HDL-C provided the strongest relationship (correlation coefficient, r = 0.866) with sdLDL-C, followed by dLDL-C (r = 0.823), TC (r = 0.807), cLDL-C (r = 0.758), and HDL-C (r = −0.230). Therefore, the linear equation using the non–HDL-C was selected for calculating the sdLDL-C concentration.
Relationship Between the Calculation and the Homogeneous Assays for sdLDL-C
In Figure 1, we show the association between the measured and the calculated sdLDL-C concentrations obtained from overall study samples. For the scatter plot (Figure 1A), the least-squares regression statistics obtained between the calculated (y) and the measured (x) sdLDL-C values were ymg/dL= 0.866x + 7.59 or ymmol/L= 0.866x + 0.197 (95% confidence interval [CI], 0.829–0.904 for slope and 5.47–9.71 mg/dL [0.14–0.251 mmol/L] for the y-intercept) with a correlation coefficient of 0.935. In addition, the paired t test showed no significant mean difference (P = .111). Besides the use of simple linear regression to evaluate correlation, we also used the other regression statistics, Deming and Passing-Bablok. Deming regression statistics were ymg/dL= 0.92x + 4.89 or ymmol/L= 0.92x + 0.127 (95% CI, 0.87–0.97 for slope and 1.54–6.75 mg/dL [0.06–0.19 mmol/L] for the y-intercept. Passing-Bablok regression statistics were ymg/dL= 0.92x + 4.67 or ymmol/L= 0.92x + 0.12 (95% CI, 0.88–0.96 for slope and 2.82–6.50 mg/dL (0.07–0.17 mmol/L) for the y-intercept. The Bland-Altman plot for the comparison between the calculated and measured sdLDL-C is shown in Figure 1B. The average difference reported from the calculated sdLDL-C minus the measured sdLDL-C and the SD of the difference were 0.85 mg/dL (0.02 mmol/L) and 9.2 mg/dL (0.238 mmol/L), respectively.
Effect of TG, TC, HDL-C, and LDL-C Concentrations on the Estimated sdLDL-C
Figure 2 illustrates the difference between the calculated sdLDL-C and the measured sdLDL-C against the independent variables of lipid concentrations (TG, Figure 2A; TC, Figure 2B; HDL-C, Figure 2C; and dLDL-C, Figure 2D). The effect the major lipids, TG ranging from 37 to 397 mg/dL (0.42–4.49 mmol/L), TC ranging from 76 to 575 mg/dL (1.97–14.90 mmol/L), HDL-C ranging from 18 to 112 mg/dL (0.47–2.91 mmol/L), and dLDL-C ranging from 36 to 522 mg/dL (0.92–13.52 mmol/L), did not significantly affect the bias error of the calculated sdLDL-C (R2 = 0.002 and P = .451; R2 = 0.0045 and P = .443; R2 = 0.0129 and P = .063; and R2 = 0.0032 and P = .520, respectively).
We also performed a correlation between measured sdLDL-C and TG concentrations as shown in Figure 3. As it was apparent that sdLDL-C levels correlated well (r = 0.489) with TG concentrations less than 200 mg/dL (2.26 mmol/L), as shown in Figure 3A. With increasing TG levels ranging from 200 to 400 mg/dL (2.26–4.52 mmol/L; Figure 3B) and more than 400 mg/dL (4.52 mmol/L; Figure 3C), the correlation values were decreased to r = 0.216 and r = 0.196, respectively.
Regression Analyses for Subgroups
Table 4 shows the regression analyses between the measured and the calculated sdLDL-C values obtained from the non–HDL-C equation for subgroups. The relationship by least-squares regression analysis was tight and consistent across subgroups of sex, age group, chronic kidney disease stages, and fasting plasma glucose categories. The slopes and y-intercepts showed consistent direction, ranging from 0.790 to 0.966 and 3.26 to 13.37 mg/dL (0.084–0.346 mmol/L), respectively. There were no significant differences in the slope or intercept for the regression equations for any of the subgroup comparisons (P > .37). All correlation coefficient values were greater than 0.90. The mean bias between measured and calculated sdLDL-C values was small, ranging from −0.63 to 5.50 mg/dL (–0.02–0.14 mmol/L).
Several epidemiologic studies have demonstrated that many patients with CVD had LDL-C levels in the same range compared with healthy subjects, whereas the distribution of LDL particle size shifted toward smaller.25–28 Thus, sdLDL particles are believed to be more atherogenic compared with lbLDL particles. The sdLDL particles more readily penetrate the arterial wall and show a higher affinity for the intimal proteoglycans, a more prolonged plasma half-life, a lower binding affinity for the LDL receptor, and a lower resistance to oxidative stress than lbLDL. Koba et al29 found that the cholesterol concentration carried on sdLDL was significantly higher in severe coronary heart disease than in mild disease and that its concentrations were associated with the severity of coronary atherosclerosis independent of the levels of LDL-C, HDL-C, apolipoprotein B, and non–HDL-C in the overall group of patients with CVD. Furthermore, sdLDL-C may be a good biomarker to assess response to therapeutic interventions in patients with type 2 diabetes with dyslipidemia.30
Many available measurements of sdLDL require special equipment and a lengthy analytic time and are, therefore, unsuitable for general clinical use and screening. Moreover, currently available methods vary considerably, limiting their clinical usefulness.31 We determined that one can easily estimate sdLDL-C from measurements of classic plasma lipids. Our equation (in mg/dL), sdLDL-C = 0.580 (non–HDL-C) + 0.407 (dLDL-C) – 0.719 (cLDL-C) – 12.05, was developed on the hypothesis that the inaccuracy of calculated LDL-C by using the Friedewald formula is linearly related to TG and HDL-C levels, which frequently associate with predominance of sdLDL particles.
The estimated sdLDL-C according to our equation seemed reliable across a wide spectrum of sdLDL-C from 7 to as high as 187 mg/dL (0.18–4.84 mmol/L). The results clearly support a strong linear relationship between measured and calculated sdLDL-C values, with an R2 of 0.88. The differences between measured and calculated sdLDL-C values were independent of the TG, TC, HDL-C, and LDL-C concentrations usually found in clinical practice (Figure 2). Moreover, the differences between various subgroups, such as sex, age, chronic kidney disease stages, and fasting plasma glucose categories, were not statistically significant (Table 4), which demonstrates that this equation may be reliable for the general population.
Our results showed that non–HDL-C levels had the strongest relationship with sdLDL-C levels (r = 0.866) compared with other lipid measures. Recent observational and interventional studies suggest that the predictive value of non–HDL-C for cardiovascular risk and mortality is better than that of LDL-C.32–34 In addition, the close association between non–HDL-C and sdLDL-C adds additional support for using the non–HDL-C level as a predictor of CVD mortality.
In calculating sdLDL-C, 3 of the parameters (TC, TG, and HDL-C) are currently standardized, while that of LDL-C is not. Although most direct measurement methods for LDL-C have received Centers for Disease Control and Prevention certification from the Cholesterol Reference Method Laboratory Network,35 the accuracy may vary. Miller et al36 found that 7 direct methods for measuring LDL-C failed to meet the NCEP total error goals, especially for samples from patients with CVD and/or dyslipidemia. Variation in the results for LDL-C may have contributed to the variation found in our study for the calculated sdLDL. The coefficients from this study are applicable to the dLDL-C based on the particular liquid selective detergent method used in the study. Other methods, such as elimination, selective solubilization, and enzyme selective protecting methods, for dLDL-C may show different results. Extending these results to other methods for dLDL-C would require performance with those reagents.
Although the importance of measurement of sdLDL has been well recognized, there has been no standard assay procedure in general clinical use. We developed a convenient equation for calculating sdLDL-C in serum samples from commonly available measurements of non–HDL-C and direct and calculated LDL-C values by using the Friedewald formula. In addition, a simple method for the calculation of sdLDL-C in serum samples without requiring specialized laboratory measurement will provide a cost-effective method for screening patients for the risk of CVD. Furthermore, it may be of particular importance in clinical practice and public health for screening for abnormalities in the metabolism of lipoproteins. The identification of a simple, inexpensive marker for sdLDL particles may preselect patients who would most benefit from a more definitive subfraction workup. The findings of this study can have an important role in helping clinicians to embrace the calculated sdLDL-C as a new addition approach to CVD risk assessment.
The study did not evaluate the performance of our proposed equation in specific patient groups with abnormal lipoprotein metabolism such as CVD, kidney disease, diabetes mellitus, metabolic syndrome, and type III hyperlipoproteinemia. Future studies are needed to add confirmation that the equation can provide the potential application in all populations.
sdLDL-C seems to be an independent risk factor for CVD, independent of LDL-C, because patients with CVD may have LDL-C within the reference interval but increased levels of sdLDL.25–28 Non-HDL and apolipoprotein B may represent improved measurement over LDL-C, but they do not necessarily capture the increased risk associated with increased sdLDL levels.29
The sdLDL-C concentration can be estimated from the classic lipid measures of non–HDL-C and calculated and direct LDL-C levels. Our equation (in mg/dL) was sdLDL-C = 0.580 (non–HDL-C) + 0.407 (dLDL-C) – 0.719 (cLDL-C) – 12.05. This linear regression equation can be used to calculate the sdLDL-C concentration, which did not differ significantly across subgroups based on sex, age group, chronic kidney disease stages, and fasting plasma glucose categories. There was a suggestion that using the calculated sdLDL-C in serum samples as a means of assessing CVD risk should be considered worldwide in clinical practice.
Upon completion of this activity you will be able to:
describe the important role of small, dense low-density lipoprotein (sdLDL) in the development of atherosclerosis.
predict a simple inexpensive marker for sdLDL particles for screening of abnormalities in the metabolism of lipoproteins.
discuss the effect of lipoprotein particle composition contributed to the variation for calculating sdLDL-cholesterol.
The ASCP is accredited by the Accreditation Council for Continuing Medical Education to provide continuing medical education for physicians. The ASCP designates this journal-based CME activity for a maximum of 1 AMA PRA Category 1 Credit ™ per article. Physicians should claim only the credit commensurate with the extent of their participation in the activity. This activity qualifies as an American Board of Pathology Maintenance of Certification Part II Self-Assessment Module.
The authors of this article and the planning committee members and staff have no relevant financial relationships with commercial interests to disclose.
Questions appear on p 152. Exam is located at www.ascp.org/ajcpcme.