The metabolic heterogeneity of HDL particles may underlie the inconsistency between epidemiological studies, consistently showing independent risk prediction, and experimental approaches in clinical trials of lipid treatments. HDL comprises a diverse group of lipoproteins with substantial differences in size and density, and composition of lipids and proteins that influence the functional properties and metabolism of the particles. Thus, it is likely that subpopulations of HDL exist with more or less anti-atherogenic potential.^12–15 Several large-scale epidemiologic studies have investigated the risk of CHD when HDL was separated by size. In some studies the concentration of small size HDL, or increase in small size HDL caused by gemfibrozil, is associated with lower incidence of CHD;^16–18 in other studies, large size HDL had the protective associations;^19,20 and in another cohort study, very large HDL particle size was directly associated with incidence of CAD.²¹ Thus, it remains inconclusive whether any of these techniques lead to any gain in information in terms of the identification of HDL subclasses with variable anti-atherogenic potential. Efforts to identify characteristics that may modulate the functional properties and metabolism of the HDL particle are important to improve the understanding of the atherosclerotic process and to prevent and treat cardiovascular diseases.

We analyzed whether lifestyle factors and biomarkers known to be associated with cardiovascular risk were differentially associated with levels of HDL-C with and without apoC-III. Compared with normal weight, overweight and obesity was associated with 7% (P=0.0002) and 12% (P=0.01) lower levels of HDL-C without apoC-III, whereas elevated body mass index was statistically insignificantly associated with higher levels of HDL-C with apoC-III (Figure 2). Alcohol was associated with approximately 3% (P=0.04 and 0.01, respectively) higher levels of both HDL-C types per each 15 g/day. Smokers had 1% (P=0.04) higher levels of HDL-C with apoC-III compared with nonsmokers. Levels of HDL-C without apoC-III were 9% (P=0.02) higher in premenopausal women and postmenopausal women who received hormone replacement than postmenopausal women who did not take hormones. We evaluated cardiovascular biomarkers in separate models and found that higher triglyceride levels were associated wit h higher levels of HDL-C with apoC-III, and lower levels of HDL-C without apoC-III (both P = 0.001).

Investigations of the metabolic heterogeneity of lipid particles are potentially valuable to improve understanding of the atheroprotective or nonprotective effects of HDL. Although previous large-scale studies have evaluated HDL subpopulations separated by size,²¹ whether such measurements improve cardiovascular risk prediction remains uncertain as the findings have been inconsistent.^15,35 Other more experimental subclassifications include the effect of HDL on cholesterol efflux or anti-inflammatory activities of HDL.³⁶ These new experimental assays are of scientific interest and suggest that the measure of total HDL-C may be diluted due to a mixing of cholesterol distributed in both anti- and proatherogenic HDL particles. However, so far, the concept and understanding of what makes a dysfunctional or even proinflammatory HDL subtype remains elusive.^14,37 Our data suggest that apoC-III may confer atherogenic properties to HDL that potentially could overcome other beneficial components. We observed that the apoC-III concentration in HDL was not significantly associated with CHD, indicating that the concentration of HDL-C with any apoC-III may be more relevant to the risk of CHD than how much apoC-IIII is in the HDL. Previous studies have only addressed the latter question. In the CLAS trial, the concentration of apoC-III in HDL was inversely associated with the progression of CAD in the drug-treated group of CAD patients only,³⁸ whereas a direct, but not statistically significant, association with CHD and re-current events was reported in two larger studies.^26,39 These studies were either cross sectional³⁹ or conducted in a patient population with existing CVD.²⁶ Because the concentration of apoC-III may be affected by disease status, it may be particularly important to study this in a prospective setting in populations that did not have clinical CVD at baseline. For the metabolism of entire lipoprotein particles, it is likely that the presence (if any) versus absence of apoC-III may determine the downstream interactions with receptors and enzymes.^22,40. Kawakami et al²⁵ reported that HDL without apoC-III, but not HDL with apoC-III, limits the proinflammatory adhesion of human monocytes to endothelial cells. ApoC-III also plays an important role in the catabolism of triglyceride-rich lipoproteins through the inhibition of clearance of plasma VLDL and LDL by the liver.^41–43 It is possible that apoC-III functions similarly in HDL circulating in blood, impairing delivery of HDL-C to the liver. However, it remains a possibility that apoC-III is a marker for other attributes of HDL that are related to atherosclerosis.

Lifestyle factors may modulate the distribution of cholesterol within the two HDL fractions. We found that alcohol intake was similarly associated with both HDL-C subfractions, whereas body weight and estrogens were only associated with HDL-C without apoC-III. Other unmeasured confounders cannot be excluded.

We used the concentration of cholesterol in HDL as our measure of HDL. It is possible that our results could be refined by assessing alternative measures, such as the concentration of apoA-I, the major apolipoprotein component of HDL, instead. It is among our study limitations that we only had one assessment of the lipid subfractions. Thus, our findings cannot determine whether changes in the proportions of cholesterol transported in HDL particles with and without apoC-III are causally related to risk of CHD. Furthermore, the immunoaffinity chromatography approach is a lengthy procedure and nondifferential measurement error might have diluted our relative risk estimates. Earlier studies of apoC-III in lipoprotein fractions have reported concern about redistribution of apoC-III from apoB-containing lipoproteins to HDL during storage.⁴⁴ We did not detect instability in our measures of HDL-C with (any) apoC-III and HDL-C without apoC-III in our own assessment of samples that were analyzed both fresh and after freezer storage. Use of different methodologies for the separation of the apoC-III—containing lipoprotein fractions may be one explanation.

Although our nested case-control study was prospective in nature, undiagnosed illness at baseline might create a spurious association. We compared the associations from analyses in strata of 5 years of follow-up. The results were similar to those presented here both when cases that occurred during the first five and subsequent years of follow-up were considered (data not shown).