Jump to
Abstract
Abstract A preponderance of small, dense LDL has been suggested to be more atherogenic than larger, more buoyant LDL. Although several studies have suggested associations of small, dense LDL with hyperinsulinemia, few data are available on the relation of small, dense LDL to insulin resistance. We examined the association of LDL particle size determined by gradient gel electrophoresis with the rates of whole-body glu-cose uptake (WBGU) as determined by the hyperinsulinemic euglycemic clamp with indirect calorimetry in 87 Finnish normoglycemic men. LDL size was significantly positively correlated with the rates of WBGU (overall, r=.31, P<.01; oxidative, r=.23, P<.05; and nonoxidative, r=.31, P<.01). Rates of WBGU were also significantly lower in subjects with small LDL particles (≤26.0 nm) compared with those in sub-jects with larger LDL particles (>26.0 nm). This relation was not explained by obesity. Serum triglyceride concentrations were found to significantly affect the relationship of LDL particle size to WBGU. Specifically, LDL size was correlated with the rates of WBGU in men with mildly elevated triglyceride levels but not in men with low triglyceride levels. Serum VLDL triglyceride concentration was a substantially stronger determinant of LDL size than were the rates of WBGU. WBGU was not significantly related to LDL size when adjusted for triglycerides. We conclude that a preponderance of small, dense LDL particles is associated with insulin resistance and that serum triglyceride concentration modifies this relationship.
-
Reprint requests to Leena Mykkänen, MD, Department of Medicine, Division of Clinical Epidemiology, University of Texas Health Science Center at San Antonio, 7703 Floyd Curl Dr, San Antonio, TX 78284-7873.
- Received August 29, 1996.
- Accepted November 18, 1996.
An increased concentration of LDL cholesterol is widely recognized as a risk factor for coronary heart disease.1 2 There is considerable heterogeneity in the size and density of LDL particles.3 4 Austin et al5 found that most individuals can be assigned to one of two LDL subclass patterns (A or B). Small, dense LDL particles (pattern B) are thought to be more atherogenic than larger LDL particles.5 6 7
Recently, several clinical and epidemiological studies have shown an association between insulin concentrations and various metabolic and physiological abnormalities, including glucose intolerance, dyslipidemia (specifically, increased serum triglyceride and decreased HDL cholesterol concentrations), and hypertension.8 9 10 11 These disorders are also associated with insulin resistance.12 13 14 15 16 17 18 In two prospective studies, several metabolic disorders related to insulin resistance syndrome have been found to cluster together.19 20
Small, dense LDL has been associated with most individual components of insulin resistance syndrome, including hypertriglyceridemia,4 5 6 7 21 22 23 24 25 low serum HDL cholesterol concentration,21 22 23 24 25 hypertension,22 23 25 diabetes,22 23 24 25 26 and hyperinsulinemia.21 22 23 25 Nevertheless, information about the association between LDL particle size and the degree of insulin resistance is limited. Reaven et al27 showed that subjects with a preponderance of small, dense LDL particles (pattern B) were more insulin resistant and had higher triglyceride concentrations than subjects with larger LDL particles (pattern A). However, they did not examine which of the two, the degree of insulin resistance or triglyceride concentration, was a stronger determinant of LDL particle size. Three studies28 29 30 examined the relation of small, dense LDL particles and insulin resistance in mildly hypertriglyceridemic subjects with NIDDM28 29 or in a combined group of subjects with and without NIDDM30 and gave contradictory results: one study29 did and two studies28 30 did not find a positive correlation between small, dense LDL and insulin resistance.
In the present study, we examined the association between insulin resistance and LDL particle size determined by gradient gel electrophoresis in normoglycemic middle-aged men. We used euglycemic clamp with indirect calorimetry to assess oxidative and nonoxidative glucose disposal. In addition to insulin resistance, we investigated the role of triglyceride and FFA concentrations as determinants of LDL particle size.
Methods
The subjects for this study were randomly selected from a population-based study from eastern Finland aiming to investigate the relationship of insulin resistance with asymptomatic atherosclerosis in middle-aged, healthy men.31 Descriptive data, including measures of insulin resistance and body fat distribution, have been published previously.32 None of the 87 subjects had any chronic disease, any drug treatment that could influence carbohydrate metabolism, any abnormality in an oral glucose tolerance test (impaired glucose tolerance or diabetes according to the criteria of the World Health Organization),33 or hypertension (use of antihypertensive drugs or systolic/diastolic blood pressure >160/95 mm Hg). Furthermore, liver, kidney, and thyroid function tests were normal in every subject. Informed consent was obtained from all subjects. The protocol was approved by the Ethics Committee of the University of Kuopio.
Weight and height were measured in light clothing without shoes. BMI was calculated as weight (kg) divided by height (m) squared and was used as an index of overall adiposity. Waist circumference was measured at the level of the umbilicus with the subject standing and breathing normally. Hip circumference was measured at the level of greatest hip girth. The WHR was used as a measure of body fat distribution.
The subjects selected for the study were admitted to the metabolic ward for 2 days. On day 1, an oral glucose tolerance test (75 g glucose) was performed and samples for blood glucose and plasma insulin were obtained at 0 and 2 hours to exclude those with impaired glucose tolerance or diabetes. On day 2, the euglycemic clamp and indirect calorimetry were performed.
Metabolic Studies
The degree of insulin resistance was evaluated by the euglycemic clamp technique as previously reported.34 After baseline blood collection and measurement of gas exchange (see next paragraph), a priming dose of insulin (Velosulin Human, Novo Nordisk) was administered during the initial 10 minutes in a logarithmically decreasing manner to raise serum insulin immediately to the desired level, where it was maintained by a continuous insulin infusion at a rate of 480 pmol (80 mU)·m2 body surface area−1·min−1. The mean insulin during the last hour of the clamp study was 1.080 pmol/L. Blood glucose was clamped at 5 mmol/L for the next 180 minutes by 20% glucose infused at various rates according to blood glucose measurements performed at 5-minute intervals. The data were calculated for each 20-minute interval during hour 3 of the euglycemic clamp. FFA concentration during the clamp was calculated as the mean of three measurements over 20-minute intervals during hour 3 of the euglycemic clamp.
Indirect calorimetry was performed with a computerized flow-through canopy gas analyzer system (Deltatrac, TM Datex)35 as previously described in detail.36 On the day of the experiment, gas exchange (O2 consumption and CO2 production) was measured in the fasting state and for the last 20 minutes of the euglycemic clamp. Protein, glucose, and lipid oxidation were calculated according to Ferrannini.37 The fraction of carbohydrate nonoxidation during glucose clamp studies was estimated by subtraction of the carbohydrate oxidation rate (determined by indirect calorimetry) from the glucose infusion rate (determined by the euglycemic clamp).
Analytical Methods
Blood glucose in the fasting state and during glucose clamp studies was measured by the glucose oxidase method (Glucose Auto and Stat HGA-1120 analyzer, Daiichi Co). For the determination of plasma insulin, blood was collected in chilled tubes (with EDTA). After centrifugation, the plasma was stored at −70°C until the analysis. Plasma insulin was determined by a radioimmunoassay (Pharmacia Diagnostics AB). Nonprotein urinary nitrogen excretion was measured by an automated Kjeldahl method.38 Serum lipid and lipoprotein levels were determined from fresh serum samples drawn after a 12-hour overnight fast as previously described.31 Serum FFA concentrations in the fasting state and during the clamp were determined by an enzymatic method (Wako Chemicals GmbH).
Aliquots of fasting serum specimens were saved as contingency samples and were frozen at −70°C. The analyses for LDL particle size were done an average of 4 years later. Several studies have suggested that freezing samples does not alter their LDL phenotype.39 40 41 Nondenaturing 3% to 18% polyacrylamide gradient gels (LDL gels) were cast on the basis of a modification42 of published protocols.43 Gels were calibrated by use of (1) thyroglobulin (17.0 nm in diameter, Pharmacia); (2) carboxylated polystyrene microspheres (38.0 nm, Duke Scientific); and (3) two bands of LDL in a lyophilized plasma sample, with diameters of 27.5 and 26.6 nm.42 Samples (4 μL) were subjected to electrophoresis in TBE buffer (90 mmol/L boric acid, 81.5 mmol/L Tris, and 2.5 mmol/L EDTA, pH 8.35) for 3000 V·h. After electrophoresis was completed, LDL cholesterol was stained with Sudan black B39 and the standard proteins with Coomassie brilliant blue R-250 (Sigma).44 After destaining was completed, gels were soaked in the TBE buffer to restore gel size and shape before scanning.
The gels were subjected to densitometric scanning with an LKB-Ultroscan XL laser densitometer with GelScan XL software. Gels were calibrated for size by use of migration distance of each standard relative to thyroglobulin; a quadratic equation in relative migration distance was fitted to the natural logarithms of the diameters of the standards.45 We used a computer program developed in-house that automatically calibrated each gel, subtracted the baseline, and calculated particle diameter for the predominant peak in each sample lane. Samples were assayed at least twice, and mean values were used for analyses; the average coefficient of variation for replicated samples was 0.43% (range, 0% to 1.32%).
For some analyses, we used a cut point of LDL size of 26.0 nm to dichotomize LDL patterns into small, dense and large, buoyant LDLs. This cut point was used rather than 25.5 nm46 because this point best separates the two modes of the LDL size distribution in this group of normoglycemic subjects.47 Small, dense LDLs in this study would therefore include LDL subclass types B and I (intermediate).46 The fact that in this study the LDL size distribution was bimodal with a nadir >25.5 nm suggests that the population sample in the present study was different from earlier studies42 46 ; the previous studies had subjects with a high risk of developing diabetes.
Statistical Analyses
All calculations were performed with the SPSS/PC+ programs (SPSS Inc) and SAS (version 6.09). Data are presented as mean±SEM. The following statistical tests were done: Student’s t test (Table 2⇓), Spearman correlation (nonparametric) (Table 4⇓), and multiple linear regression (Tables 3⇓ and 5⇓). We also used multiple linear regression in which ranks were substituted for actual LDL sizes as an alternative to traditional linear regression (“semiparametric linear regression”).48 Insulin and triglyceride were logarithmically transformed (natural logarithm) for statistical testing to improve their skewness and kurtosis. These variables are presented in their natural units in the tables. Because LDL size in this population was bimodal and could not be successfully transformed to a normal distribution, we emphasized nonparametric methods in this report.
Results
Clinical and metabolic characteristics of the overall population are shown in Table 1⇓. LDL size was significantly inversely correlated with BMI (Spearman correlation coefficient, r=−.24, P<.05), WHR (r=−.30, P<.01), plasma 2-hour insulin concentration (r=−.27, P<.05), and FFA concentration during clamp (r=−.24, P<.05) and positively correlated with the rates of WBGU (overall, r=.31, P<.01; oxidative, r=.23, P<.05; and nonoxidative, r=.31, P<.01). LDL size was also significantly inversely correlated with serum VLDL triglyceride concentration (r=−.60, P<.001) and positively correlated with HDL cholesterol concentration (r=.40, P<.001).
Clinical and Metabolic Characteristics of Subjects (n=87)
LDL cholesterol level was not related to LDL size or WBGU. HDL cholesterol concentration was significantly positively correlated with LDL size and inversely correlated with WHR (r=−.28, P<.01) but was not related to WBGU. VLDL triglyceride concentration was inversely correlated with LDL size and WBGU (overall, r=−.27, P<.05 and nonoxidative, r=−.26, P<.05) and positively correlated with WHR (r=.25, P<.05) and FFA concentration during clamp (r=.50, P<.001). Fasting FFA concentration did not correlate significantly with LDL size or WBGU (overall, r=−.19; oxidative, r=−.14; nonoxidative, r=−.19; P=NS). Lipid oxidation during clamp was not related to LDL size. There was an inverse association between lipid oxidation and oxidative glucose disposal during clamp (r=−.59, P<.001) but no relationship between lipid oxidation during clamp and overall (r=−.10, P=NS) or nonoxidative (r=−.19, P=NS) glucose disposal. FFA concentration during clamp correlated inversely with LDL size and overall (r=−.26, P=.020) and nonoxidative (r=−.30, P=.007) glucose disposal but not with oxidative glucose disposal (r=−.02, P=NS). We have previously reported significant correlations between BMI and WBGU (overall, r=−.30; oxidative, r=−.21; nonoxidative, r=−.25) and between WHR and WBGU (overall, r=−.54; oxidative, r=−.23; nonoxidative, r=−.50).32
In Table 2⇓, we show metabolic characteristics of subjects with different LDL sizes. Subjects with a preponderance of small, dense LDL (predominant LDL peak ≤26.0 nm) had higher VLDL cholesterol and total, LDL, and VLDL triglyceride concentrations and lower HDL cholesterol concentrations than did subjects with larger LDL particles (predominant LDL peak >26.0 nm). In addition, subjects with small LDL particles had lower total, oxidative, and nonoxidative glucose uptake than did subjects with larger LDL particles. However, subjects with small LDL particles did not differ from subjects with larger LDL particles in relation to fasting FFA concentration or lipid oxidation and FFA suppression during clamp. It is noteworthy that subjects with smaller and larger LDL particles did not differ in regard to obesity (BMI) or body fat distribution (WHR).
Clinical and Metabolic Characteristics of the Subjects by LDL Size
We did stepwise multiple linear regression analyses to investigate whether WBGU was an independent determinant of LDL size in the overall study population (Table 3⇓). In model 1, with VLDL triglyceride (natural logarithm) and WBGU as independent variables, the association of WBGU with LDL size was only borderline significant (P=.056) and WBGU independently explained merely 2% of the variance in LDL size. We also did the multiple regression analyses including all the variables associated with LDL size in univariate analyses, ie, VLDL triglyceride level, WBGU, FFA concentration during clamp, WHR, and HDL cholesterol level, as independent variables; then VLDL triglyceride and HDL cholesterol concentrations were independent determinants of LDL size (model 2). In contrast, WBGU was not independently related to LDL size after the effects of VLDL triglyceride and HDL cholesterol levels were accounted for. VLDL triglyceride concentration was overwhelmingly the strongest determinant of LDL particle size, explaining ≈47% of the variance of LDL size in both linear regression models. The results for traditional and semiparametric multiple linear regression analyses were very similar.
Stepwise Multiple Linear Regression Analyses With LDL Size as a Dependent Variable
We found an almost significant interaction (P=.059) between total triglyceride concentration (dichotomized as below or above the median, ie, 1.3 mmol/L) and WBGU when the dependent variable was LDL size in multiple linear regression analyses. Therefore, we also investigated the relationship between LDL size and WBGU, plasma insulin concentration, and serum VLDL triglyceride and HDL cholesterol concentrations by calculating Spearman correlation coefficients separately in men with serum triglyceride concentrations below the median (<1.3 mmol/L; group with low triglyceride levels) and those with serum triglyceride concentrations above the median (≥1.3 mmol/L; group with mildly elevated triglyceride levels) (Table 4⇓). There was no relation between LDL particle size and WBGU and plasma insulin concentration in men with low triglyceride levels. However, VLDL triglyceride concentration was inversely associated with LDL size in men with low triglyceride levels. In contrast, there was a positive correlation between LDL size and overall and nonoxidative glucose disposal (r=.42, P<.01) in men with mildly elevated triglyceride levels. Furthermore, LDL size correlated inversely with VLDL triglyceride concentration and positively with HDL cholesterol concentration in men with mildly elevated triglyceride levels.
Spearman Correlations of LDL Size With Selected Metabolic Variables in Subjects With Low and Mildly Elevated Triglyceride Levels
We also repeated the multiple linear regression analyses separately in subjects with low and those with mildly elevated triglyceride levels (Table 5⇓). Again, VLDL triglyceride level was the strongest determinant of LDL particle size. However, in subjects with triglyceride levels above the median (≥1.3 mmol/L), WBGU was associated with LDL particle size independently of VLDL triglyceride and HDL cholesterol concentrations.
Stepwise Multiple Linear Regression Analyses With LDL Size as a Dependent Variable in Subjects With Low and Mildly Elevated Triglyceride Levels
Discussion
In this report, we have shown an association of insulin resistance with the preponderance of small, dense LDL particles in normoglycemic middle-aged men. Men with small, dense LDL particles had decreased rates of oxidative and nonoxidative glucose disposal compared with men with larger LDL particles. However, this relationship was markedly dependent on serum triglyceride concentration. Insulin resistance was a substantially stronger determinant of LDL size in men with mildly elevated triglyceride levels than in men with low triglyceride levels.
Our results are in accordance with findings of Reaven et al,27 who previously showed that nondiabetic subjects with LDL subclass pattern B are insulin resistant, as measured by the steady-state plasma glucose technique. Previous results on the relation of small, dense LDL particles and insulin sensitivity in mildly hypertriglyceridemic subjects with NIDDM have been contradictory; one study29 showed an association between insulin resistance and small, dense LDL, whereas two studies did not.28 30 This may be due to differences in the methodology to determine LDL subfractions (density gradient ultracentrifugation versus gradient gel electrophoresis) and to the degree of hypertriglyceridemia among the study subjects. Subjects with NIDDM have a preponderance of small, dense LDL particles, particularly if they are hypertriglyceridemic.29 30 42 Furthermore, one of the previous studies30 used 24-hour rather than fasting triglyceride levels.
LDL particle size was inversely related to fasting VLDL triglyceride concentration, which, on the other hand, was associated with insulin resistance. Thus, the relationship between LDL size and insulin resistance could be due to the connection with VLDL triglyceride level. Indeed, VLDL triglyceride concentration was a substantially stronger independent predictor of LDL size than WBGU in these normoglycemic middle-aged men (Table 3⇑). Interestingly, there was a positive correlation between LDL size and WBGU in men with mildly elevated triglyceride concentrations (≥1.3 mmol/L) but not in men with low triglyceride concentrations (<1.3 mmol/L) (Table 4⇑). Previous studies have shown that serum triglyceride concentration and hepatic lipase activity are major determinants of small, dense LDL particles in mildly hypertriglyceridemic diabetic and nondiabetic subjects.29 30 Even mild elevations of triglyceride levels have previously been shown to be associated with insulin resistance.49 In insulin-resistant states, the suppression of hepatic VLDL secretion by insulin is impaired, and this leads to an increase in the number of triglyceride-rich VLDL particles in the circulation.50 51 High levels of VLDL triglyceride may increase neutral lipid exchange between VLDL and LDL, resulting in triglyceride-enriched cholesteryl ester–depleted LDL species.52 53 These LDL species are susceptible to the action of hepatic lipase, which hydrolyzes LDL triglycerides, resulting in the formation of small, dense LDL particles.54
LDL particle size was also inversely related to plasma FFA concentration during clamp, indicating impaired suppression of FFA by insulin and insulin resistance in adipose tissue in subjects with small, dense LDL particles. Elevated plasma FFA levels provide substrate to hepatic VLDL triglyceride synthesis. Furthermore, elevated plasma FFA levels can lead to increased lipid oxidation, which in turn leads to decreased glucose oxidation and insulin resistance via substrate competition.55 56 The association of FFA concentration during clamp and total and nonoxidative glucose disposal rate in the present study is probably due to its association with VLDL triglyceride concentration. In line with this view, the correlation between FFA concentration during clamp and fasting VLDL triglyceride concentration was substantially stronger than correlations between FFA concentration and WBGU. Furthermore, in multiple linear regression analysis in the overall study population, VLDL triglyceride was an independent determinant of LDL particle size, but FFA concentration during clamp and WBGU were not.
LDL particle size was positively correlated with serum HDL cholesterol level. Actually, HDL cholesterol concentration was the next strongest determinant of LDL size in this study. This can be partly explained by the close inverse relationship between HDL cholesterol and VLDL triglyceride concentrations. However, HDL cholesterol concentration was also associated with LDL size independently of VLDL triglyceride levels both in the overall study population and in men with mildly elevated triglyceride levels, which confirms findings in previous studies.24 29 42 This can be related at least in part to hepatic lipase activity. Hepatic lipase modulates both LDL particle composition, resulting in formation of small, dense LDL species,54 and remodeling of HDL particles.57
In conclusion, we have shown that normoglycemic subjects with small, dense LDL particles are insulin resistant. However, the association of LDL size and insulin sensitivity was strongly modified by triglyceride level.
Selected Abbreviations and Acronyms
| BMI | = | body mass index |
| FFA | = | free fatty acid |
| NIDDM | = | non–insulin-dependent diabetes mellitus |
| WBGU | = | whole-body glucose uptake |
| WHR | = | waist-to-hip ratio |
Acknowledgments
This work was supported by grants from the Medical Research Council of the Finnish Academy and grants R01-HL-24799, R37-HL-36820, and P01-HL-45522 from the National Heart, Lung, and Blood Institute. The authors gratefully acknowledge the technical assistance of Mahmood Poushesh, Wendy R. Shelledy, and Susan H. Slifer.
References
This Issue
Jump to
Article Tools
Share this Article
- Relationship of LDL Size to Insulin Sensitivity in Normoglycemic Men
