Evaluating the Incremental Benefits of Raising High-Density Lipoprotein Cholesterol Levels During Lipid Therapy After Adjustment for the Reductions in Other Blood Lipid Levels FREE

The relationship between the concentration of high-density lipoprotein cholesterol (HDL-C) and the risk of cardiovascular (CV) events has been assessed in several population studies, where it has been found that, for every 1-mg/dL increase (to convert to millimoles per liter, multiply by 0.0259) in HDL-C level, there is a 2% to 3% decrease in the risk of future events.¹ This inverse relationship between HDL-C levels and CV events is independent of low-density lipoprotein cholesterol (LDL-C) levels and remains apparent even when levels of LDL-C have been reduced by aggressive statin treatment to below 70 mg/dL.²

In support of the human epidemiologic observations, numerous studies in animal models have shown that interventions to raise HDL-C levels inhibit the development of atherosclerosis.^3- 6 Furthermore, HDL-C has several known functions with the potential to protect against the development of atherosclerosis and its sequelae.^7- 9 These include promoting the efflux of cholesterol from cells in the artery wall, promotion of endothelial function, and repair and inhibition of thrombosis.^7- 8 High-density lipoprotein cholesterol also stimulates endothelial nitric oxide production and has antioxidant⁸ and anti-inflammatory activities.⁹ Evidence from human intervention studies suggests that increasing the level of HDL-C is associated with a reduction in atheroma progression as measured in coronary^10- 11 and carotid¹² arteries. Thus, there is a compelling case for considering interventions to raise HDL-C levels as a strategy to reduce CV risk.

In contrast to the consistent trial data showing the cardioprotective effects of reducing LDL-C levels with statins, the evidence that raising HDL-C levels translates into a reduction in CV events is at best circumstantial and remains controversial. Indeed, enthusiasm for therapy to raise HDL-C levels has been somewhat dampened by the recent results of treating humans with the cholesteryl ester transfer protein inhibitor torcetrapib. Despite increasing HDL-C levels by more than 60% in these human studies, treatment with torcetrapib had no apparent effect on atherosclerosis and was associated with an increase in major CV events and total mortality.^13- 16 Whether the adverse effects of torcetrapib were related to the inhibition of cholesteryl ester transfer protein, to the associated increase in HDL-C levels, or to an off-target effect of the drug is currently being tested in studies with other cholesteryl ester transfer protein inhibitors that do not share the known off-target effects of torcetrapib.

Despite the adverse results with torcetrapib, numerous epidemiologic and clinical trial studies support the proposition that therapies to raise HDL-C levels have the potential to reduce CV events.^{10- 11,17- 18} On the other hand, in a recently published meta-regression analysis of lipid therapy trials, Briel and coworkers¹⁹ concluded that increasing HDL-C levels did not appear to be associated with a reduction in CV events. To further address this issue, we have analyzed data from individuals treated with lipid-modifying therapy in the Framingham Offspring Study and have tested the hypothesis that a change in the concentration of HDL-C is an inverse and independent predictor of future CV events.

We analyzed public domain data from the Framingham Offspring Study from 1975 through 2003. We focused only on those individuals who started lipid therapy between visits 2 and 6. This strategy provided plasma lipid levels for each individual before therapy was started and at 1 or more follow-up visits. For this report, a CV event included the development of angina pectoris, coronary insufficiency, myocardial infarction, coronary death, transient ischemic attack, and fatal or nonfatal thrombotic stroke as previously defined by the Framingham investigators. Detailed descriptions of the examination procedures and CV events have been reported previously.^20- 21

The change in HDL-C levels after lipid-modifying therapy was calculated for each individual. For individuals with multiple visits while receiving therapy, the average of all available lipid measurements was used to estimate levels during treatment. Similar estimates were also completed for the change in LDL-C and plasma triglyceride levels. After dividing the cohort into quartiles of change in HDL-C levels, we produced Kaplan-Meier curves for each subgroup. Trends across changes in HDL-C level quartiles were assessed using the log-rank χ² test for trend.

Exploratory univariate analyses compared changes in other covariates across quartiles of change in HDL-C levels to identify potential confounders. Cox proportional-hazards regressions were used to estimate the association between HDL-C levels and CV events after adjustment for baseline lipid levels and changes following treatment. A second model then adjusted for baseline levels and changes in LDL-C levels. In a third model, additional potential confounders were added to model 2. Potential confounders that might be associated with changes in HDL-C levels and CV events were selected by inspecting the variables in Table 1 (eg, plasma triglyceride levels and use of β-blockers) or on the basis of previously published studies (eg, smoking status, change in blood pressure, age, sex, and prevalent diabetes mellitus or CV disease). Finally, we evaluated possible interactions between levels of HDL-C and other blood lipids.

Patients included in this analysis (n = 454) started lipid-modifying therapy between the second and sixth examination cycles. The type of lipid drug was unknown for 95 patients. Among the remaining 359 patients, 344 (96%) patients were taking 1 drug only, including statins (72%), fibrates (17%), resins (4%), or niacin (7%), and 15 patients (4%) were taking more than 1 drug.

When stratified by the change in HDL-C levels (Table 1), factors associated with larger increases in HDL-C levels included smaller changes in LDL-C levels, larger changes in plasma triglyceride levels, and less frequent use of β-blockers.

During an average follow-up period of 8 (SD, 4) years, 79 individuals experienced a CV event, including 5 coronary deaths, 24 myocardial infarctions, 29 cases of angina or coronary insufficiency, 1 stroke death, 11 thrombotic strokes, and 9 transient ischemic attacks. When we stratified individuals into quartiles of HDL-C change, a significant (log-rank χ² test for trend, P = .006) inverse relationship was observed between changes in HDL-C levels and CV event–free survival (Figure 1). After adjustment for the pretreatment HDL-C levels, age, and sex, the decreased hazard ratio (HR) for cardiovascular events (Table 2) associated with a 5-mg/dL rise in HDL-C level was 0.80 (95% confidence interval [CI], 0.69-0.94) (model 1). Further adjustment for pretreatment LDL-C levels and changes in LDL-C levels (model 2) did not substantially change the HR (0.79; 95% CI, 0.68-0.93). The HR remained stable after the addition of potential confounders such as plasma triglyceride levels, changes in plasma triglyceride levels, the use of β-blockers, prevalent CV disease or diabetes mellitus when lipid therapy was started, and changes in body mass index, systolic blood pressure, or smoking status (0.79; 95% CI, 0.67-0.93) (model 3).

The impact of changes in HDL-C level appeared to be remarkably stable across a wide range of subpopulations, as demonstrated in Figure 2. Overall, a 1% increase in HDL-C level was associated with a 2% drop in CV risk. An important interaction was also observed between the pretreatment levels in LDL-C and changes in HDL-C levels. Stratifying by tertiles of LDL-C levels prior to treatment, the lower the LDL-C, the greater was the risk reduction associated with the increase in HDL-C (Figure 3). This significant interaction (P < .05) was also present after adjustment for all the covariates in model 2 (Table 2), when pretreatment levels of LDL-C and HDL-C and changes in LDL-C levels were included in the model as continuous variables. The resulting HR for a 5-mg/dL increase in HDL-C level can be calculated using the following equation: HR = Exp[5 × βHDL-C + 5 × (βHDL-C × LDL-C) × LDL-C], where Exp indicates exponential, βHDL-C is the estimated β coefficient for changes in HDL-C level, LDL-C is the pretreatment LDL-C level, and βHDL-C × LDL-C is the interaction between changes in HDL-C and pretreatment LDL-C levels. The β coefficient for the interaction was 0.00074 (95% CI, 0.00002-0.0014).

It is well established that HDL-C level is an inverse predictor of CV risk. Whether interventions that increase HDL-C levels reduce the risk of future events is less certain. In this analysis of individuals starting pharmacotherapy for dyslipidemia, there was an inverse relationship between changes in HDL-C levels and CV events. Despite a limited number of CV events to analyze, the greater the increase in HDL-C level, the lower the CV risk—an observation that persisted after adjustment for changes in levels of LDL-C and plasma triglycerides and other potential confounders such as changes in smoking habits or body weight. We found that a 1% increase in HDL-C level was associated with a 2% reduction in CV risk, remarkably consistent with the epidemiologic data from prospective cohort studies of individuals not receiving lipid therapy.¹

As with any therapeutic intervention, analyses of data from a randomized clinical trial based on a clear a priori hypothesis remain the only way to prove causality. In the absence of such data, concerns regarding residual confounding cannot be dismissed. Accordingly, there remains the possibility that changes in factors associated with HDL-C levels and CV events might explain the results observed herein. Although we adjusted for changes in LDL-C and plasma triglyceride levels, as well as a variety of other factors, additional unforeseen confounders remain a concern.

Despite the obvious shortcomings of this analysis, these data provide some of the strongest evidence currently available to support the hypothesis that raising HDL-C levels is associated with a reduction in CV risk. Unlike data from highly selected subjects enrolled in a clinical trial, the data from this study are more representative of individuals treated in a community setting, albeit an extensively studied one. The subjects in the present analyses included those who, at the time they started lipid therapy, had known CV disease and/or diabetes mellitus, or neither condition. Treatment was initiated at a time when therapeutic guidelines in the United States primarily targeted LDL-C levels. Hence, the changes in HDL-C levels were unlikely to be the focus of treatment or to bias the assignment of clinical outcomes.

The conclusion that raising HDL-C levels is associated with lower CV risk appears to be robust across a wide spectrum of patients and does not appear to be uniquely associated with a specific class of drugs, as demonstrated in Figure 2. The broad range of treated patients and drug therapies resulted in a wide range of lipid responses, providing the necessary heterogeneity to demonstrate the association between HDL-C level changes and CV events. Similar levels of heterogeneity are largely lacking in clinical trials testing 1 specific therapy in a relatively homogeneous group of patients. Meta-regression analyses of clinical trials are also often restricted by limited heterogeneity based on the average results reported for each study rather than individual patient data. This may explain the negative results obtained by Briel et al¹⁹ when the results of 108 randomized trials were combined. In the Framingham data set, strong heterogeneity is reflected by the wide variation in observed HDL-C level changes ranging from −37 mg/dL to +35 mg/dL or median values of −6 mg/dL in the lowest quartile to +11 mg/dL in the highest.

The heterogeneity in LDL-C values before treatment was initiated also was useful for identifying an important interaction between baseline LDL-C levels and changes in HDL-C levels. This finding suggests that raising HDL-C levels may be particularly beneficial among individuals with mixed dyslipidemia compared with those with isolated elevations in LDL-C levels. This is consistent with the findings of a secondary analysis of the Scandinavian Simvastatin Study by Ballantyne et al,²² which demonstrated that individuals with the lipid triad (high LDL-C, low HDL-C, and high triglyceride levels) received significantly greater benefits from statin therapy than did individuals with isolated high LDL-C levels. This was despite similar reductions in LDL-C levels. Similarly, a post hoc analysis of the Myocardial Ischaemia Reduction with Aggressive Cholesterol Lowering Study²³ identified that pretreatment HDL-C levels were significantly associated with clinical outcomes but only among those individuals with LDL-C levels below the median for participants in the study. Among those with LDL-C levels above the median, there was no significant relationship between HDL-C levels and ischemic events.

Clinical trial data also support the Framingham Offspring Study results presented herein. In the Arterial Biology for the Investigation of the Treatment Effects of Reducing Cholesterol 2 Study,²⁴ the addition of extended-release niacin to statin therapy slowed the progression of atherosclerosis among individuals with known coronary heart disease. Unfortunately, in that study, there were not enough cardiovascular events observed to confirm a benefit on clinical outcomes. Detailed analyses of clinical outcomes in several lipid therapy trials have also demonstrated that raising HDL-C levels was associated with clinical benefits. These studies include the National Heart, Lung, and Blood Institute Type II Coronary Intervention Study,²⁵ the Lipid Research Clinics Coronary Primary Prevention Trial,²⁶ the Helsinki Heart Study,²⁷ the HDL-Atherosclerosis Treatment Study,²⁸ the Scandinavian Simvastatin Survival Study,²⁹ and the Veterans Affairs High-Density Lipoprotein Intervention Trial.³⁰ Increases in HDL-C levels (5%-26%) were independently associated with a reduction in CV events observed in each trial. Finally, recent post hoc analyses of 1 of the failed torcetrapib trials showed an inverse relationship between changes in HDL-C levels and the percentage of atheroma volume among patients receiving torcetrapib therapy.³¹

In conclusion, although the benefits of raising HDL-C levels remain to be confirmed in randomized clinical trials, it appears that the modest changes in HDL-C levels resulting from treatment with commonly used lipid drugs are associated with a reduction in CV risk independent of the effects on other lipid measures. Among individuals receiving lipid therapy to lower LDL-C levels, raising HDL-C levels should be evaluated further as a secondary goal of treatment.

Correspondence: Steven A. Grover, MD, MPA, McGill University Health Centre, 687 Pine Ave W, V-Bldg, Room V2.07, Montreal, QU H3A 1A1, Canada (steven.grover@mcgill.ca).

Accepted for Publication: July 3, 2009.

Author Contributions: Dr Grover had full access to all the data in the study and takes final responsibility for the decision to submit for publication. Study concept and design: Grover and Joseph. Acquisition of data: Grover and Kaouache. Analysis and interpretation of data: Grover, Kaouache, Joseph, Barter, and Davignon. Drafting of the manuscript: Grover and Davignon. Critical revision of the manuscript for important intellectual content: Grover, Kaouache, Joseph, and Barter. Statistical analysis: Grover, Kaouache, and Joseph. Obtained funding: Grover. Administrative, technical, and material support: Grover and Davignon. Study supervision: Grover.

Financial Disclosure: Dr Grover and Mr Kaouache have received research grants from Merck & Co, Pfizer Inc, Sanofi-Aventis, and AstraZeneca. Dr Grover has received speaker honoraria from Pfizer Inc, Sanofi-Aventis, and Merck & Co and been a consultant to or participated on advisory boards for AstraZeneca, Sanofi-Aventis, Pfizer Inc, and Merck & Co. Dr Barter has received research grants from Merck & Co, Pfizer Inc, and Roche; has received speaker honoraria from Abbott Laboratories, AstraZeneca, Merck & Co, and Pfizer Inc; and has been a consultant to or participated on advisory boards for AstraZeneca, Merck & Co, Pfizer Inc, Roche, and Sanofi-Aventis. Dr Davignon has received research grants from AstraZeneca, Pfizer Inc, and Merck Frosst; has participated in clinical trials for AstraZeneca, Bristol-Myers Squibb, Merck Frosst-Schering, Pharmena, Pfizer Inc, Sanofi-Aventis, and Takeda Pharmaceutical Co Ltd; and has received honoraria as a consultant or advisory board member from Anthera Pharmaceuticals Inc, AstraZeneca, Bristol-Myers Squibb, Life-Cycle Pharma, Merck & Co, Merck Frosst-Schering, Pharmena, Pfizer Inc, Roche, Sanofi-Aventis, Solvay, and Takeda Pharmaceutical Co Ltd.

Funding/Support: This study was supported in part by an unrestricted grant-in-aid from Merck International. The Framingham Offspring Study is conducted and supported by the National Heart, Lung, and Blood Institute (NHLBI) in collaboration with the Framingham Offspring Study Investigators.

Role of the Sponsor: The investigators were responsible for obtaining the data set, data analysis, and preparation of the manuscript independently of the funding source. The sponsor was permitted to review the manuscript, but all final decisions regarding content and the decision to publish remained the responsibility of the investigators. The manuscript was prepared using a limited access data set obtained from the NHLBI and does not necessarily reflect the opinions or views of the Framingham Offspring Study or the NHLBI.

Additional Contributions: Baishali Ambegaonkar, PhD, Vasilisa Sazonov, PhD, and Don Yin, PhD (Merck & Co), assisted in the completion of this study. We thank the Framingham Offspring Study investigators for their important contribution in gathering these data.