Clinical Feature: HDL and Cardiovascular Disease Risk—Risk Marker or Risk Factor?

Low HDL cholesterol (HDL-C) is an established biomarker for the future development of atherosclerosis and atherosclerotic cardiovascular disease (CVD) events in population-based observational studies and clinical trials of cholesterol-lowering therapies and this includes coronary heart disease (CHD) patients with low levels of LDL cholesterol (LDL-C) on statin therapy.^1-3 Because multiple clinical trials designed to increase HDL-C and reduce CVD risk have not demonstrated efficacy and instead have shown potential harm,4-6 there has been widespread confusion regarding the importance of HDL as a biomarker of risk and as a potential target for therapeutic intervention.⁷ Confusion around the importance of HDL in atheroprotection has been magnified by a Mendelian randomization study that utilized HDL-C as an intermediary biomarker of CVD risk.⁸

This perspective discusses the limitations of HDL-C as a biomarker of risk, and misguided attempts to reduce atherosclerosis through effectuating changes in the cholesterol content of HDL particles.

HDL-C: A Biomarker of Atherogenic Lipoproteins
Low HDL-C levels are commonly associated with cholesterol-depleted HDL particles and cholesterol-depleted LDL particles.⁹ Due to conformational changes in apolipoprotein (apoB) on the surface of small LDL particles, these cholesteroldepleted particles have reduced interactions with LDL receptors, which results in delayed hepatic clearance of these atherogenic lipoproteins from the bloodstream. In a primary prevention trial of individuals with reduced HDL-C levels, baseline and on-trial CHD events were linearly related to apoB levels.¹⁰ It remains uncertain from current observational and clinical trial data whether the increased CVD risk associated with low levels of HDL-C is anything more than an excess concentration of atherogenic apoBcontaining particles.

HDL Particles: A Biomarker of CVD Risk
For 50 years, "reverse cholesterol transport" has been considered the major anti-atherogenic function of HDL.^11,12 Only 3-5% of the mass of HDL particles, however, is derived from macrophage cholesterol efflux.¹³ This reliance on HDL-C has resulted in an under appreciation of the contribution of protein-enriched, cholesterol-depleted HDL subclasses in modulating critical atheroprotective functions.¹⁴ Specifically, certain populations of small HDL particles have stronger anti-oxidant, antiinflammatory and anti-infective properties than other constituents of HDL.¹⁵

Several prospective population studies^16-18 and clinical trials of lipid modifyingtherapies^19,20 have reported that the concentrations of small HDL particles and/ or total HDL particles are more robust predictors of CVD risk than total HDL-C. These studies suggest that the protein composition or proteome of certain subpopulations of HDL particles may have more important contributions to HDLassociated CVD risk than can be deduced from the cholesterol carrying capacity of circulating HDL particles alone.

Mendelian randomization studies have been considered to provide an "unbiased" evaluation of biomarkers and their association with the development of disease. Two recent GWAS studies provide an example of conflicting conclusions related to HDL associated CVD risk however.^8,21 In a combination of five cardiovascular disease case-control studies (4,658 cases and 11,459 controls), two single nucleotide polymorphisms (SNP) in the phospholipid transfer protein (PLTP) gene associated with reduced PLTP activity were associated with reduced CVD risk despite the lack of influence of this enzyme on the cholesterol content of HDL.²¹ In contrast, the PLTP polymorphism was associated with higher concentrations of small HDL particles and total numbers of HDL particles. In the one Mendelian randomization study that had previously used a genome-wide association approach to identify SNPs that affect blood lipid concentrations,^22,23 a polymorphism in the endothelial lipase gene (LIPG Asn396Ser) found in 20 studies (20,913 myocardial infarction cases, 95,407 controls) as well as 14 other pathways (12,482 cases of myocardial infarction and 41,331 controls exclusively associated with HDL-C), were investigated for their association with myocardial infarction.⁸ Despite having high HDL-C levels, there was no association with CVD risk. In this study, the authors assumed that "plasma HDL-C" was the biomarker “directly involved” in the pathogenesis of CHD, and concluded that inherited variation in plasma HDL-C was a surrogate measure of disease.

The Multi-Ethnic Subclinical Atherosclerosis (MESA) study¹⁸ investigated the relative importance of several HDL measures (HDL-C, HDL size and HDL particle concentration) as to risk of subclinical atherosclerosis and incident CVD events. In multivariate models that included major risk factors, lipids and lipoprotein subclasses, HDL particle concentration but not HDL-C was inversely associated with carotid intima medial thickness and incident CHD events. Two other prospective population studies have reported that HDL-P is a more accurate predictor of CVD events in a subset of metabolic syndrome patients (Multiple Risk Factor Collaboration Trial,¹⁶ and the EPIC Norfolk Study).¹⁷

Several pharmacological agents are known to increase the cholesterol content of HDL (CETP inhibitors, niacin)^24,25 and other agents increase the number of HDL particles, but are less effective in loading the particles with cholesterol (PPAR-a agonists).¹⁹ In the Veterans Administration HDL Intervention Trial (VAHIT), gemfibrozil therapy reduced LDL-P by percent, increased total HDL particles by 10% and small HDL particle subclass by 21%. In the initial report of this trial, on-trial changes in HDL-C explained only 17% of the CVD risk.²⁶ A nested-case control study from VA-HIT investigated the CVD risk associated with baseline and on-trial changes in lipids and lipoprotein subclasses.²⁰ In multivariate models that included treatment group, conventional risk factors and lipid and lipoprotein concentrations, every one standard deviation change in HDL-C was unrelated to CHD events (p=0.14); whereas HDL-P was inversely related with CHD events at baseline (odds ratio: 0.78 [0.69-0.90] ) and on-trial (odds ratio: 0.71 [0.61-0.81.]¹⁹ In an analysis of HDL subclasses, baseline and on-trial risk were associated with small HDL subclasses (0.71[0.60-0.84 and 0.67[0.57-079] respectively) and medium subclasses (0.82[0.70-0.96] and 0.82[0.69-0.97]) but not by large HDL subclasses (0.95 [082-1.11] ) and 0.92 [0.79-1.07] respectively).

From the available data, HDL-C remains a robust marker of elevated apoB-containing lipoproteins and CVD risk; however, HDL-C is not a useful biomarker of HDL functionality. Multiple HDL subclasses are involved in anti-atherogenic mechanisms. HDL-P assessment provides one clinically available measure that appears to more precisely predict CVD risk. From my perspective, future trials that investigate HDL-modifying therapies should assess HDL-P and, in particular, certain functional HDL subclasses.

Disclosure statement: Dr. Rosenson has received grants from Amgen, Genentech and Hoffman- LaRoche. Dr. Rosenson has been an adviser to Abbott Laboratories, Amarin Corporation, Amgen, Genentech, Hoffman-LaRoche, Kowa Pharmaceuticals America, LipoScience Inc. and Sanofi-Aventis. Dr. Rosenson has stock options in LipoScience Inc.