High-density lipoproteins and cardiovascular disease: the plots thicken
- Division of Cardiovascular Medicine, Radcliffe Department of Medicine, University of Oxford, Oxford, UK
- Correspondence to Professor Robin P Choudhury, Division of Cardiovascular Medicine, John Radcliffe Hospital, Oxford OX3 9DU, UK;
Plasma lipoproteins can be separated by ultracentrifugation based on their physical properties, without direct reference to their composition, with ‘high-density’ lipoproteins (HDL) comprising those with density greater than 1.063 g/ml. HDL lipoprotein particles should be explicitly distinguished from HDL-cholesterol, which is but one of the constituents of HDL. Apolipoprotein A-I is the major structural protein of HDL and comprises 70% of total HDL protein. In addition, there is a large number of protein constituents, the functions of which broadly span lipid transport, inflammation, immune function, hormone binding, haemostasis and antioxidant functions.1 Our appreciation of the context-specific roles of these proteins remains rudimentary, and tools to quantify their putative contributions still more so. On the other hand, measurement of HDL-cholesterol is relatively straightforward and has been undertaken widely.
HDL has come to be regarded as atheroprotective, based on three lines of evidence: (1) the inverse relationship between cardiovascular event rates and HDL-cholesterol that is seen in epidemiological studies; (2) animal data showing that the administration of apolipoprotein AI and/or HDL particles reduce atherosclerosis and (favourably remodel atherosclerotic plaques); and (3) a small number of clinical trials suggesting that some drugs that elevate HDL-cholesterol also reduce cardiovascular events or result in plaque regression, assessed by ultrasound or MRI. These observations have raised the possibility that interventions that increase HDL-cholesterol might reduce cardiovascular risk. However, several recent findings have challenged this notion. First, alterations of HDL-cholesterol that are associated with naturally occurring, genetic polymorphisms allow the effects of changes in HDL-cholesterol due to Mendelian randomisation to be mapped to cardiovascular risk. Alterations in HDL-cholesterol by these means do not track with predicted cardiovascular risk. Secondly, niacin (nicotinic acid) is the most efficacious of clinically available drugs to elevate HDL-cholesterol. The AIM–HIGH trial of niacin in patients with low HDL-cholesterol was abandoned due to lack of clinical outcome efficacy. Thirdly, inhibition of cholesteryl ester transfer protein results in marked elevation in HDL-cholesterol. However, two drugs from this class (torcetrapib and dalcetrapib) have also failed to show clinical benefit in phase III trial.
These findings do not easily sit with current thinking, and potentially undermine the utility of HDL-cholesterol as a biomarker and target for intervention.
In a general population, there is a strong inverse relationship between the HDL-cholesterol level and cardiovascular risk.2 This relationship has been shown consistently and is reinforced by a recent evaluation of over 165 000 participants gathered from 37 prospective cohorts, followed for a median of more than 10 years and recording over 10 000 cardiovascular events.3 Significantly, the relationship persists in patients treated with statins to lower low-density lipoprotein (LDL)-cholesterol. Analyses from the Prospective Pravastatin Pooling Project and from the Heart Protection Study both showed a persistent and undiminished relationship between changes in HDL-cholesterol and risk in statin-treated patients, albeit with the absolute risk reduced. A recent meta-analysis of 20 randomised, controlled trials of statins by Jafri et al 4 has also shown a similar relationship. However, a majority of the trials included used low or intermediate intensity statin treatments, so it is important to note that in the TNT trial, HDL-cholesterol levels were associated with major cardiovascular events in patients treated with statins and that this relationship was conserved among patients with LDL-cholesterol levels below 70 mg/dl.5 Of course, these associations do not show that low HDL-cholesterol causes cardiovascular events or that intervening to elevate HDL-cholesterol will translate into clinical benefit and reduced events.
Does low HDL-cholesterol cause cardiovascular events?
Mendelian randomisation studies have recently been utilised to investigate the causal relevance of HDL-cholesterol concentration for the risk of cardiovascular events. Single nucleotide polymorphisms (SNP) in specific genes that significantly affect HDL-cholesterol levels (eg, endothelial lipase gene, lecithin-cholesterol acyltransferase) without changing other lipoprotein (LDL-cholesterol or triglyceride) or non-lipoprotein (systolic blood pressure, diabetes mellitus, body mass index) cardiovascular risk factors, might be expected to exert an effect on cardiovascular events. However, the investigation of a number of SNP with apparently isolated HDL-cholesterol-raising properties has revealed no association with cardiovascular events—in other words, in spite of the observation in the study population that HDL-cholesterol inversely correlates with cardiovascular risk, it was not found to be a causative factor.6 ,7 These findings are further supported by rare Mendelian disorders caused by mutations in ABCA1, which result in very low HDL-cholesterol concentrations but which do not result in a clinically significant increased risk of cardiovascular events.8 On the other hand, the polymorphisms that result in the observed changes in HDL-cholesterol are in genes for proteins critical for HDL function. It is not possible currently to show with certainty that the SNP investigated to date affect only measured HDL-cholesterol and have no confounding functional consequences, for example, for cholesterol acceptance/particle clearance. Therefore, further investigation into the mechanisms of HDL-cholesterol function and the specific effects of intervention on the different fractions of HDL are warranted.9
The AIM–HIGH study evaluated the effects of modified release niacin versus a nominal placebo in patients with low HDL-cholesterol and established cardiovascular disease. It was terminated early due to lack of efficacy. In some respects, the study was designed to test the ‘HDL hypothesis’ as patients had low HDL-cholesterol at baseline and the LDL-cholesterol was titrated to equivalence through the differential use of statins and ezetimibe. However, the attained difference in HDL-cholesterol between the treated and placebo groups was a mere 4 mg/dl (42 vs 38 mg/dl). This small difference might reflect regression to the mean in patients selected for low baseline HDL-cholesterol; contributions from other drugs (eg, higher doses of statins in the placebo group) and the unintended effects of low-dose niacin (which was included to induce flushing and maintain blinding) in the nominal placebo group. Furthermore, it is becoming apparent that niacin has non-lipoprotein-mediated effects in inflammatory pathways that are mediated by GPR on leukocytes and that would potentially confound outcomes in a trial of this design.10 ,11 Even with an optimal design, the event rates in AIM–HIGH were too low for differences to have been observed in a trial of that relatively small size.12 We should obtain greater certainty on a potential role for niacin in statin-treated patients when HPS-2 THRIVE reports (projected date 2013).
Cholesteryl ester transfer protein inhibitors
Cholesteryl ester transfer protein (CETP) inhibitors reduce the transfer of neutral lipids (triglycerides and cholesteryl ester) between lipoprotein particles. The net effect is to bias cholesterol into HDL compartments, away from very low-density lipoprotein. Some of the CETP inhibitors can attain a greater than 100% increase in HDL-cholesterol. However, mere accumulation of cholesterol in this compartment does not necessarily imply that there has been any biologically useful intervention (eg, enhanced reverse cholesterol transport) in respect of modifying the biology of atherosclerosis or cardiovascular risk. Imaging studies13 ,14 and phase III clinical trials15 have not shown any benefit from the lipoprotein profile changes achieved with either of the CETP inhibitors reported to date, torcetrapib and dalcetrapib.
Demonstration of a disconnection between measured HDL-cholesterol and risk/outcome clearly has implications for the interpretation of HDL-cholesterol under those specific circumstances, but does not necessarily undermine the broader possibilities for HDL in atheroprotection. Indeed, studies in experimental atherosclerosis continue to show the potential for interventions that affect HDL constituents (and may also result in HDL-cholesterol elevation) to regress atherosclerosis and/or to modify plaque composition favourably. For example, HDL-cholesterol elevation through apolipoprotein A-I administration or synthesis retards atherosclerosis progression in mice and remodels atherosclerotic plaques by increasing smooth muscle cell and collagen content.16 Feig et al 17 have recently shown that apolipoprotein A-I promotes rapid atherosclerosis regression in mice and alters inflammatory properties of plaque monocyte-derived cells through emigration of CD68+ cells, induction of the chemokine receptor CCR7, decreased expression of inflammatory factors and enrichment of markers of the M2 (tissue repair) macrophage state.
Measurements of HDL-cholesterol represent a ‘snap shot’ and do not take into account its composition. This may be important since cholesterol efflux capacity from macrophages, a measure of HDL function, has a strong inverse association with both carotid intima-media thickness and the likelihood of angiographic coronary artery disease, independently of the HDL-cholesterol level.18 Furthermore, additional HDL functions, for example, on inflammation, reverse cholesterol transport, oxidation, endothelial function and emerging possibilities in the regulation of gene transcription,19 etc, are not assessed at all by the measurement of HDL-cholesterol.
Low levels of HDL-cholesterol are associated with increased cardiovascular risk. This relationship remains in patients treated with statins, in whom there remains a significant ‘residual risk’ of cardiovascular disease—recommending HDL-cholesterol elevation as a rational target. However, association does not equate with causation, and recent Mendelian randomisation studies have questioned the nature of the relationship between HDL-cholesterol and cardiovascular risk. Conversely, HDL-raising interventions in animal models of atherosclerosis (especially elevating apolipoprotein A-I) show numerous beneficial effects on atherosclerosis, including lesion regression. If HDL is a therapeutic target, it seems likely to matter which of its many components are targeted. To date, we have measured HDL-cholesterol: (1) because we can and (2) because it has seemed to represent an apparently convenient short-cut to biologically relevant processes. However, this measure is an isolated snap shot that misses functional aspects and ignores numerous other components that may play important roles. Recent data support the notion of HDL complexity and that our model has been overly simplistic. We would not be wise to dismiss the possibilities for this complex particle based merely on an expanding appreciation of our ignorance.
Contributors Design by RPC. Literature review, drafts and revisions by RPC and NR. Final critical review and amendments by RPC.
Funding RPC and NR acknowledge the support of the BHF Centre of Research Excellence, Oxford. RPC is a Wellcome Trust senior research fellow in clinical science. Our laboratory is supported by the Oxford Comprehensive Biomedical Research Centre, NIHR funding scheme.
Competing interests None.
Provenance and peer review Commissioned; internally peer reviewed.