Author + information
- Received October 19, 2015
- Revision received March 4, 2016
- Accepted March 15, 2016
- Published online May 31, 2016.
- Purav Mody, MDa,
- Parag H. Joshi, MDa,b,
- Amit Khera, MD, MSca,
- Colby R. Ayers, MSa,c and
- Anand Rohatgi, MDa,∗ ()
- aDivision of Cardiology, Department of Internal Medicine, UT Southwestern Medical Center, Dallas, Texas
- bJohns Hopkins Ciccarone Center for the Prevention of Heart Disease, Baltimore, Maryland
- cDepartment of Clinical Sciences, UT Southwestern Medical Center, Dallas, Texas
- ↵∗Reprint requests and correspondence:
Dr. Anand Rohatgi, UT Southwestern Medical Center, 5323 Harry Hines Boulevard, Dallas, Texas 75390-8830.
Background Cholesterol efflux capacity (CEC), which is a key step in the reverse cholesterol transport pathway, is independently associated with atherosclerotic cardiovascular disease (ASCVD). However, whether it predicts ASCVD beyond validated novel risk markers is unknown.
Objectives This study assessed if CEC improved ACSVD risk prediction beyond using coronary artery calcium (CAC), family history (FH), and high-sensitivity C-reactive protein (hs-CRP).
Methods CEC, CAC, self-reported FH, and hs-CRP were assessed among participants without baseline ASCVD who were enrolled in the Dallas Heart Study (DHS). ASCVD was defined as a first nonfatal myocardial infarction (MI) or stroke, coronary revascularization, or cardiovascular death, assessed over a median 9.4 years. Risk prediction was assessed using various modeling techniques and improvements in the c-statistic, the integrated discrimination index (IDI), and the net reclassification index (NRI).
Results The mean age of the population (N = 1,972) was 45 years, 52% had CAC (>0), 31% had FH, and 58% had elevated hs-CRP (≥2 mg/l). CEC greater than the median was associated with a 50% reduced incidence of ASCVD in those with CAC (5.4% vs. 10.5%; p = 0.003), FH (5.8% vs. 10%; p = 0.05), and elevated hs-CRP (3.8% vs. 7.9%; p = 0.004). CEC improved all metrics of discrimination and reclassification when added to CAC (c-statistic, p = 0.004; IDI, p = 0.02; NRI: 0.38; 95% confidence interval [CI]: 0.13 to 0.53), FH (c-statistic, p = 0.006; IDI, p = 0.008; NRI: 0.38; 95% CI: 0.13 to 0.55), or elevated hs-CRP (c-statistic p = 0.008; IDI p = 0.02; NRI: 0.36; 95% CI 0.12 to 0.52).
Conclusions CEC improves ASCVD risk prediction beyond using CAC, FH, and hs-CRP and warrants consideration as a novel ASCVD risk marker.
- cholesterol efflux capacity
- coronary artery calcium
- family history
- high-density lipoprotein
- risk prediction
Low high-density lipoprotein cholesterol (HDL-C) is an important risk marker for atherosclerotic cardiovascular disease (ASCVD). However, recent studies suggest that in the current era of well-treated patients, the association between HDL-C and ASCVD may be attenuated (1–3). In addition, assessment of both low- and high-density lipoprotein particle composition offsets this association completely (4,5), limiting the role of HDL-C as a biomarker of ASCVD risk.
HDL is a complex lipoprotein with heterogeneous composition and functions (6), and static cholesterol concentration of HDL does not capture the diversity inherent in HDL particles. The classic function attributed to HDL is to promote reverse cholesterol transport from the periphery to the liver for elimination from the body. Cholesterol efflux from the macrophage to HDL is the initial key step of reverse cholesterol transport and is associated with atheroprotection in animal studies. We (and others) have shown that macrophage-specific cholesterol efflux capacity (CEC) measured in large human cohorts is inversely associated with both prevalent coronary disease and incident ASCVD events (5,7,8).
However, it remains unknown whether assessment of the reverse cholesterol transport pathway as reflected by CEC could serve as a clinically relevant biomarker in ASCVD risk prediction. Addressing this knowledge gap would support investigation and development of refined bioassays of HDL function for clinical testing. The presence of coronary artery calcium (CAC), a family history (FH) of myocardial infarction (MI), and elevated high-sensitivity C-reactive protein (hs-CRP) reflect mediators of ASCVD risk (subclinical atherosclerosis, inherited risk, and inflammation, respectively) and are validated biomarkers in clinical use that improve ASCVD risk prediction (9–11). We examined the incremental ability of CEC to improve ASCVD risk prediction beyond using CAC, FH, and hs-CRP in a low-risk, population-based cohort.
The Dallas Heart Study (DHS) was a multiethnic, population-based cohort study of Dallas County residents (12). This random probability sample included intentional oversampling of African Americans to make up 50% of the cohort. Participants 30 to 65 years of age underwent body composition assessment by fasting blood and urine collection and dual-energy x-ray absorptiometry. Detailed cardiovascular phenotyping was accomplished by electron-beam computed tomography and magnetic resonance imaging of the heart, and body fat distribution was evaluated by magnetic resonance imaging of the abdomen. Participants with a history of CVD (self-reported history of MI, stroke, arterial revascularization, heart failure, or arrhythmia) or niacin use were excluded, as were those who died within 1 year after enrollment. Participants (N = 2,971) completed risk factor assessment, laboratory testing, and imaging studies between 2000 and 2002, and 2,744 of them completed CAC scans. Of those, 185 were excluded because of a history of CVD, 238 had a lack of adequate follow-up, 263 lacked valid CEC measurements, and 86 had missing covariates, leaving 1,972 participants for the final analysis of cardiovascular outcomes.
Race/ethnicity, medication usage, FH, and smoking status were self-reported. Detailed definitions for the variables of hypertension, metabolic syndrome, and diabetes in the DHS have been previously published (13). FH was defined as any first-degree relative with a history of MI. FH of premature MI was defined as occurring at younger than 50 years in a first-degree male relative or at younger than 55 years in a first-degree female relative (14).
Analytical methods for the biomarkers reported in this study have been previously described, including lipoprotein assessment and hs-CRP (15,16). Electron-beam computed tomographic measurements of CAC were performed in duplicate 1 to 2 minutes apart on an Imatron 150 XP scanner (Imatron Inc., San Bruno, California). CAC scores were determined using the Agatston method and then averaged, with an Agatston score >0 defined as prevalent CAC.
Assessment of lipid variables and efflux capacity
Fasting blood samples were collected by venipuncture into ethylenediaminetetraacetic acid tubes, stored at 4°C for <4 h, and centrifuged. Plasma was removed and stored at −70°C. Plasma lipids, including HDL-C, were measured as described previously (12). CEC was assessed by measuring the efflux of fluorescence-labeled cholesterol from J774 macrophages to apolipoprotein B–depleted plasma in study participants as previously described (5).
The primary endpoint was a composite of a first nonfatal MI, nonfatal stroke, coronary revascularization (percutaneous coronary intervention or coronary artery bypass grafting), or death from cardiovascular causes. Nonfatal endpoints were actively ascertained and adjudicated by 2 cardiologists who were unaware of the measurements of CEC, as previously described (17). The National Death Index was used to determine the vital status for all the participants through December 31, 2010. Death from cardiovascular causes was defined according to the International Classification of Diseases, 10th Revision, codes I00 to I99.
Baseline categorical variables are reported as percentages, and continuous variables are reported as mean ± SD. Kaplan-Meier curves for CEC below the median (vs. above the median) for the overall cohort were compared by the log-rank test and stratified by the presence of prevalent CAC (defined as CAC >0), FH, and elevated hs-CRP (defined as hs-CRP >2 mg/l). Cox proportional-hazards models were used to assess the association between CEC and the time to the first ASCVD event in univariable and multivariable models. The proportional hazards assumption was tested by Schoenfeld residuals. Traditional risk factors included age, sex, race, presence or absence of diabetes, systolic blood pressure, current smoking status, body mass index, total cholesterol level, HDL-C, history of antihypertensive medication use, and statin use. Model overfitting was tested by calculating the shrinkage coefficient of the full model. The shrinkage estimator of van Houwelingen and le Cessie was 0.92, well above the cutoff of 0.85 as outlined by Harrell, and indicated no concern for model overfitting (18). Forward stepwise selection was performed, including all traditional risk factors, CAC, FH, hs-CRP, and CEC, with variables with p < 0.05 retained in the model. Finally, the ability of CEC to improve ASCVD risk prediction beyond using traditional ASCVD risk factors was assessed using indexes of discrimination as measured by the Harrell’s c statistic and integrated discrimination improvement (IDI), and reclassification as measured by the category-less net reclassification index (NRI) (19–21). The Gronnesby and Borgan test was used for assessing model calibration (22). Several pre-specified sensitivity analyses were performed, including removing percutaneous coronary intervention/coronary artery bypass grafting from the primary endpoint, restricting cardiovascular death to fatal MI or fatal stroke, restricting the cohort to age older than 45 years, using risk thresholds for calculation of the categorical NRI, continuous CEC, using categorical definitions of CAC at varying thresholds, and using premature FH instead of FH. Two-sided p values of ≤0.05 were considered to indicate statistical significance. All statistical analyses were performed using SAS software (version 9.3, SAS Institute, Raleigh, North Carolina).
The baseline demographic and clinical characteristics of study participants are listed in Table 1. The mean age of the study population was 44 ± 9.2 years, with 44% men and 47% African Americans. The percentage of participants with prevalent CAC (>0) was 52%. FH and premature FH were reported in 31% and 10% of the participants, respectively. hs-CRP >2 mg/l was noted in 58% of the participants.
Among the 1,972 participants included in analysis, 97 had a first ASCVD event (28 MIs, 32 strokes, 5 coronary artery bypass graft surgeries, 11 percutaneous coronary interventions, and 21 cardiovascular deaths) over a median follow-up of 9.4 years (95% confidence interval [CI]: 9.0 to 9.8). Those with CEC more than the median versus less than the median had a decreased risk of ASCVD (3.1% vs. 6.7%; p = 0.0003) (Figure 1A). Among those with prevalent CAC (n = 1,030), those with CEC more than the median versus less than the median had a decreased risk of ASCVD (5.4% vs. 10.5%; p = 0.003) (Figure 1B). Similar findings were seen among those with FH (n = 621; CEC more than the median vs. less than the median: 5.8% vs. 10%; p = 0.05) (Figure 1C) and elevated hs-CRP (n = 1,148; CEC more than the median vs. less than the median: 3.8% vs. 7.9%; p = 0.004) (Figure 1D).
In a fully adjusted model that included all traditional risk factors, prevalent CAC, FH, and elevated hs-CRP, CEC remained inversely associated with incident ASCVD without attenuation (adjusted hazard ratio [HR]: 0.35; 95% CI: 0.23 to 0.55). Forward stepwise selection retained CEC along with prevalent CAC and FH (Table 2). Subgroup analyses revealed that CEC was inversely associated with incident ASCVD among those with prevalent CAC (adjusted HR: 0.40; 95% CI: 0.25 to 0.64), among those with FH (adjusted HR: 0.31; 95% CI 0.17 to 0.58), and among those with elevated hs-CRP (adjusted HR: 0.37; 95% CI: 0.22 to 0.63) (Figure 2). Statistical interaction tests between CEC and CAC, FH, and hs-CRP were performed and were nonsignificant.
The ability of CEC to improve ASCVD risk prediction beyond using CAC, FH, and hs-CRP was assessed using metrics of calibration, discrimination, and reclassification. All models including CEC were well-calibrated. CEC more than the median versus less than the median improved discrimination indexes as determined by the c-statistic and IDI when added to risk factor–adjusted models, including either prevalent CAC, FH, or elevated hs-CRP (Table 3). With respect to reclassification as determined by the NRI, the addition of CEC led to significant reclassification for all models, including prevalent CAC, FH, and elevated hs-CRP (Table 4). The improvement in reclassification with the addition of CEC was driven by upward reclassification of those with events with minimal change in those without events.
Several sensitivity analyses were conducted with no overall effect on the preceding findings. Excluding coronary revascularization (percutaneous coronary intervention and coronary artery bypass grafting: 16 of 97 events) and restricting death to fatal MI or stroke did not alter the association of CEC with ASCVD when added to prevalent CAC, FH, or elevated hs-CRP. Similarly, findings were unchanged using varying categorical and continuous definitions of CAC, replacing FH with premature FH, or serial adjustment for self-reported exercise activity. Restricting the cohort to age older than 45 years did not alter the association between CEC and ASCVD when adjusted for CAC (fully adjusted HR for cholesterol efflux: 0.38; 95% CI: 0.23 to 0.64). Analysis of CAC and hs-CRP as log-transformed continuous variables did not alter the findings of any of the risk prediction indexes. Using continuous CEC did not alter any findings (Online Figures 1 and 2) and improved both upward and downward reclassification (Online Table 1). The NRI, which was calculated using risk thresholds, did not alter the findings (Online Tables 2 to 4).
In a large, multiethnic population-based cohort, we evaluated the clinical relevance of a measure of reverse cholesterol transport, CEC, on ASCVD in the context of CAC, FH of MI, and hs-CRP. The inverse association of CEC with incident ASCVD was not attenuated when accounting for all 3 risk markers combined. Furthermore, among patients with prevalent CAC, FH, or hs-CRP, CEC was able to meaningfully stratify ASCVD risk.
CAC is a noninvasive measure of calcified coronary atherosclerosis, and an increasing CAC score is directly proportional to coronary plaque burden and short-term ASCVD risk. Multiple prospective studies have demonstrated its ability to improve risk prediction for incident coronary heart disease beyond traditional risk factors, including the DHS (23–26). Based on the consistency of these data, CAC is often used clinically in those in whom statin eligibility is uncertain, a practice given a Class IIb recommendation in the recent 2013 American College of Cardiology/American Heart Association (ACC/AHA) guidelines on risk assessment (9).
Similarly, multiple large population-based studies have demonstrated that FH also improves ASCVD risk prediction after accounting for common cardiovascular risk factors (27,28). FH represents a heritable risk for ASCVD and is easily obtainable by self-report. Along with CAC, FH was given a Class IIb recommendation in those whose statin eligibility remains unclear (9,29). In the DHS, FH was found to be additive to CAC in identifying participants at significantly increased risk, demonstrating that FH and CAC represent distinct pathways leading to ASCVD events (26).
High-sensitivity C-reactive protein is a circulating biomarker that represents inflammatory pathways that lead to and are a part of ASCVD. Multiple studies have demonstrated an association of elevated hs-CRP levels with cardiovascular events and its ability to appropriately reclassify intermediate-risk individuals (30). Because of these data, it has been incorporated into the Reynolds Risk Score in addition to FH (29,31) and the current ACC/AHA guidelines give a Class IIb recommendation for incorporating hs-CRP levels in those with an unclear cardiovascular risk (9).
Reverse cholesterol transport is the key antiatherosclerotic function of HDL, and cholesterol efflux from the macrophage to the circulation is the key first step of this pathway. Unfortunately, circulating HDL-C levels and HDL particle concentration are poor surrogates for CEC (5). Bioassays measuring CEC have been applied to a few large cohorts, which demonstrated inverse associations with prevalent coronary disease and incident cardiovascular death among high-risk individuals (7,32). Within the DHS and EPIC-Norfolk (European Prospective Investigation of Cancer–Norfolk) studies, 2 large population-based cohorts at low baseline risk, CEC was shown to be inversely associated with incident ASCVD, independent of HDL-C and HDL particle composition (5,8).
The unexpected magnitude of the association with ASCVD in these studies and the lack of attenuation with traditional risk factors prompted us to further investigate the impact of CAC, FH, and hs-CRP on CEC and ASCVD. When CEC, CAC, FH, and hs-CRP were added to traditional risk factors, efflux remained associated with ASCVD, regardless of how predictors were analyzed. Analysis of the survival curves demonstrated that CEC improved risk stratification among those with either prevalent CAC, FH, or elevated hs-CRP. Those with CEC above the median had a halving or more of their ASCVD risk over almost 10 years. Risk prediction performance measures confirmed that adding CEC to CAC, FH, and hs-CRP improved the ability to predict incident events. Taken together, these findings demonstrate that the pathways that mediate the association between efflux and ASCVD in this study population are not reflected by atherosclerotic pathways reflected by CAC, inherited risk reflected by FH, or inflammatory pathways reflected by CRP (Central Illustration).
The relatively young age of the DHS study population limited generalizability to older age groups. Because of the overall low cardiovascular risk of our study population, there were a small number of ASCVD events. The race/ethnicity distribution of our study sample, with oversampling of African Americans, did not represent the general population. FH in our study population was self-reported, thus making it liable to recall bias and misclassification. However, this is the same information available to practicing clinicians, and the literature has demonstrated that self-reported parental history has a sensitivity of >80% and specificity of >90%, and any misclassification would bias toward the null (33). Ankle–brachial index, the fourth nontraditional risk factor given a Class IIb recommendation for risk factor assessment in addition to CAC, FH, and hs-CRP, was not available in this cohort. Interpretation of the NRI is associated with several documented limitations (34).
Cholesterol efflux represents the first critical step of reverse cholesterol transport, the key antiatherosclerotic action of HDL. Among low-risk individuals, efflux adds to ASCVD risk prediction beyond using CAC, FH, and hs-CRP, which are 3 well-validated and clinically used markers (Central Illustration). Our findings support efforts to standardize CEC methods and develop assays that are amenable for clinical use.
COMPETENCY IN MEDICAL KNOWLEDGE: CEC, a measure of HDL function, improves risk prediction beyond using FH of MI, hs-CRP levels and detection of coronary calcification.
TRANSLATIONAL OUTLOOK: The development of efficient, standardized assays of CEC could provide a practical method for assessment of cardiovascular risk and serve as a surrogate target for the evaluation of novel therapeutic strategies.
The Dallas Heart Study is supported by grants from the Donald W. Reynolds Foundation and the National Center for Advancing Translational Sciences of the National Institutes of Health (NIH) (UL1TR001105). Dr. Rohatgi is supported by the National Heart, Lung, and Blood Institute of the NIH under Award Number K08HL118131 and by the American Heart Association under Award Number 15CVGPSD27030013; has received a research grant from Merck; is a member of the Advisory Board for Cleveland HeartLab; and has been a consultant for Vascular Strategies and CSL Limited. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- American College of Cardiology/American Heart Association
- atherosclerotic cardiovascular disease
- coronary artery calcium
- cholesterol efflux capacity
- confidence interval
- family history of myocardial infarction
- high-density lipoprotein cholesterol
- hazard ratio
- high-sensitivity C-reactive protein
- integrated discrimination index
- myocardial infarction
- net reclassification index
- Received October 19, 2015.
- Revision received March 4, 2016.
- Accepted March 15, 2016.
- American College of Cardiology Foundation
- Guyton J.R.,
- Slee A.E.,
- Anderson T.,
- et al.
- Mackey R.H.,
- Greenland P.,
- Goff D.C. Jr..,
- Lloyd-Jones D.,
- Sibley C.T.,
- Mora S.
- Vallejo-Vaz A.J.,
- Ray K.K.
- Goff D.C. Jr..,
- Lloyd-Jones D.M.,
- Bennett G.,
- et al.
- Perk J.,
- De Backer G.,
- Gohlke H.,
- et al.
- Deo R.,
- Khera A.,
- McGuire D.K.,
- et al.
- Hlatky M.A.,
- Greenland P.,
- Arnett D.K.,
- et al.
- Erbel R.,
- Mohlenkamp S.,
- Moebus S.,
- et al.
- Paixao A.R.,
- Berry J.D.,
- Neeland I.J.,
- et al.
- Nielsen M.,
- Andersson C.,
- Gerds T.A.,
- et al.
- Ridker P.M.,
- Paynter N.P.,
- Rifai N.,
- Gaziano J.M.,
- Cook N.R.