Prognostic Significance of Serum Total Cholesterol and Triglyceride Levels in Patients Hospitalized for Heart Failure With Reduced Ejection Fraction (from the EVEREST Trial)

Prognostic Significance of Serum Total Cholesterol and Triglyceride Levels in Patients Hospitalized for Heart Failure With Reduced Ejection Fraction (from the EVEREST Trial) Stephen J. Greene, MDa, Muthiah Vaduganathan, MD, MPHc, Laura Lupi, MDd, Andrew P. Ambrosy, MDe, Robert J. Mentz, MDf, Marvin A. Konstam, MDg, Savina Nodari, MDd, Haris P. Subacius, MAb, Gregg C. Fonarow, MDh, Robert O. Bonow, MDa, and Mihai Gheorghiade, MDa,*, on behalf of the EVEREST Trial Investigators Lower cholesterol levels are associated with worse outcomes in patients with chronic heart failure (HF) and have been shown to predict in-hospital mortality. The relation between lipid profile and postdischarge outcomes in patients hospitalized for worsening HF is less clear. In this post hoc analysis of the Efficacy of Vasopressin Antagonism in Heart Failure Outcome Study With Tolvaptan (EVEREST), 3,957 patients hospitalized for worsening HF with ejection fractions £40% were examined. Baseline total cholesterol and triglyceride levels were measured <48 hours after admission and evaluated as continuous variables. The primary end points of all-cause mortality and cardiovascular mortality or hospitalization for HF were compared using Cox regression models. Patient characteristics at randomization were also compared among quartiles of total cholesterol. Patients with lower total cholesterol tended to have lower blood pressure, ejection fractions, serum sodium, and albumin, and were more likely to have worse HF functional class, to have higher natriuretic peptide levels, and to have histories of diabetes mellitus, renal insufficiency, and coronary revascularization (all p values <0.001). After adjustment for baseline clinical risk factors, total cholesterol was predictive of all-cause mortality (hazard ratio 0.73, 95% confidence interval 0.63 to 0.85, p <0.001) and cardiovascular mortality or hospitalization for HF (hazard ratio 0.73, 95% confidence interval 0.66 to 0.82, p <0.001) at median follow-up of 9.9 months. Lower baseline triglyceride level was also associated with worse outcomes. In conclusion, lower baseline total cholesterol is correlated with a high-risk patient profile and is a marker of disease severity in patients hospitalized for worsening HF with reduced ejection fraction. Baseline total cholesterol and triglyceride levels are predictive of mortality and HF rehospitalization beyond traditional risk factors. Ó 2013 Elsevier Inc. All rights reserved. (Am J Cardiol 2013;111:574e581) The reciprocal association between traditional cardiovascular risk factors, including hyperlipidemia, and poor clinical a

Center for Cardiovascular Innovation and bDivision of Cardiology, Department of Medicine, Northwestern University Feinberg School of Medicine, Chicago, Illinois; cDepartment of Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts; d Department of Cardiology, University of Brescia, Brescia, Italy; e Department of Medicine, Stanford University School of Medicine, Stanford, California; fDivision of Cardiology, Department of Medicine, Duke University Medical Center, Durham, North Carolina; gDivision of Cardiology, Department of Medicine, Tufts Medical Center, Boston, Massachusetts; and hDivision of Cardiology, Ahmanson-UCLA Cardiomyopathy Center, Los Angeles, California. Manuscript received September 23, 2012; revised manuscript received and accepted October 22, 2012. Financial and material support for the Efficacy of Vasopressin Antagonism in Heart Failure Outcome Study With Tolvaptan (EVEREST) was provided by Otsuka, Inc., Rockville, Maryland. Database management was performed by the sponsor. Mr. Subacius conducted all final analyses for this report with funding from the Center for Cardiovascular Innovation, Northwestern University Feinberg School of Medicine, Chicago, Illinois. See page 580 for disclosure information. *Corresponding author: Tel: þ312-695-0051; fax: þ312-695-1434. E-mail address: [email protected] (M. Gheorghiade). 0002-9149/12/$ - see front matter Ó 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.amjcard.2012.10.042

outcomes in heart failure (HF) has been referred to as “reverse epidemiology.”1 The mechanism underlying this “cholesterol paradox” is presently unclear, but leading hypotheses include relations with persistent malnutrition and inflammation frequently seen in chronic disease states.2,3 The Efficacy of Vasopressin Antagonism in Heart Failure Outcome Study With Tolvaptan (EVEREST) database contains a number of clinically important covariates, including markers of nutrition (i.e., albumin and body mass index), and offers the opportunity to better characterize the prognostic value of baseline serum cholesterol levels on postdischarge outcomes in a large population of patients hospitalized for HF. Specifically, the objectives of the present analysis were (1) to report the distribution of total cholesterol and triglycerides, (2) to describe the baseline clinical characteristics by total cholesterol level, and (3) to assess the association between total cholesterol and triglyceride levels and postdischarge morbidity and mortality in patients hospitalized for HF with reduced ejection fraction (EF). Methods The overall study design4 and primary results5,6 of the EVEREST program have been described in detail previously. www.ajconline.org

Heart Failure/Cholesterol Paradox in Heart Failure

575

Figure 1. Selection of the analytic cohort. LVEF ¼ left ventricular ejection fraction; Q ¼ quartile.

In brief, EVEREST was a prospective, multicenter, multinational, randomized, double-blind, placebo-controlled clinical trial investigating the role of oral tolvaptan, a vasopressin-2 receptor antagonist, on short- and long-term clinical outcomes. Notably, patients were enrolled <48 hours after admission at 359 clinical sites from North America, South America, and Europe from 2003 to 2006. Patients aged >18 years who were hospitalized for worsening HF with EF 40% and New York Heart Association (NYHA) class III or IV functional status presenting with 2 signs or symptoms of fluid overload (i.e., dyspnea, pitting edema, and jugular venous distension) were eligible for enrollment. Exclusion criteria relevant for the present study included acute myocardial infarction at the time of presentation or co-morbid conditions with expected survival <6 months. Patients provided explicit written informed consent for participation that was approved by local institutional review boards or ethics committees. Background HF therapy was left to the discretion of the treating physician, but guidelinebased recommendations for optimal medical management were included in the study protocol. In EVEREST, lipid parameters including total cholesterol and triglycerides were collected at enrollment; a complete lipid panel was not included in the study protocol. These samples were not required to be fasting. Laboratory samples were centrally analyzed at 5 core facilities, and results were cross-validated across centers. Patients with missing baseline lipid parameters (either total cholesterol or triglycerides) were excluded from this analysis (n ¼ 176). The primary predictors of the present study were serum total cholesterol and triglyceride levels, expressed as continuous variables. Because triglyceride data were positively skewed, natural logarithmic transformation was performed for this variable. Additionally, the effect of triglycerides on the composite end point was not linear, and

Figure 2. Distribution and descriptive statistics of total cholesterol (A) and triglyceride (B) levels. Triglyceride level was not normally distributed and thus was log transformed for the purposes of statistical analysis. IQR ¼ interquartile range.

576

The American Journal of Cardiology (www.ajconline.org)

Table 1 Baseline characteristics by baseline total cholesterol quartile Variable

Total cholesterol (mg/dl) Triglycerides (mg/dl) Tolvaptan Age (yrs) Men Ischemic cause of HF Region of origin Eastern Europe North America South America Western Europe Body mass index (kg/m2) Systolic blood pressure (mm Hg) Diastolic blood pressure (mm Hg) Dyspnea Jugular venous distention 10 cm Rales Peripheral edema* Ejection fraction (%) Blood urea nitrogen (mg/dl) Creatinine (mg/dl) Serum sodium (mEq/L) B-type natriuretic peptide (pg/ml)† N-terminal proeB-type natriuretic peptide (pg/ml)z QRS duration on baseline electrocardiogram (ms) Atrial fibrillation on baseline electrocardiogram Albumin (g/dl) New York Heart Association class IV Previous HF hospitalization Previous CAD Previous myocardial infarction Hypertension Hypercholesterolemiax Diabetes mellitus Chronic kidney disease Peripheral vascular disease Previous coronary artery bypass Previous percutaneous coronary intervention Chronic obstructive pulmonary disease Baseline medication use Diuretics Angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers b blockers Mineralocorticoid receptor antagonists Digoxin Intravenous inotropes Statins

Quartile of Total Cholesterol (mg/dl)

p Value

1 (49e128) (n ¼ 992)

2 (129e158) (n ¼ 995)

3 (159e196) (n ¼ 976)

4 (197e437) (n ¼ 994)

108.7  14.7 94.8 (39.7%) 505 (50.9%) 66.5  12.7 809 (81.6%) 655 (66.8%)

143.7  8.7 111.4 (53.0%) 489 (49.1%) 66.0  11.8 765 (76.9%) 638 (65.2%)

176.3  10.8 136.4 (72.4%) 505 (51.7%) 66.7  11.3 700 (71.7%) 626 (65.0%)

235.7  36.0 196.8 (129.2%) 478 (48.1%) 63.9  11.5 672 (67.6%) 650 (66.1%)

201 (20.3%) 494 (49.8%) 160 (16.1%) 137 (13.8%) 28.7  5.8 114.6  18.2 68.7  12.2 875 (90.3%) 345 (35.9%) 786 (80.8%) 830 (85.2%) 25.6  8.1 30 (22e43) 1.4 (1.1e1.8) 139 (136e141) 1,202 (627e216) 6,377 (3,540e12,888)

338 (34.0%) 319 (32.1%) 204 (20.5%) 134 (13.5%) 28.2  5.6 118.5  19.2 71.5  12.8 885 (90.4%) 315 (32.3%) 789 (80.6%) 778 (79.4%) 26.8  8.3 27 (21e38) 1.3 (1.0e1.6) 140 (137e142) 898 (410e1,727) 5,703 (2,761e11,197)

452 (46.3%) 223 (22.9%) 167 (17.1%) 134 (13.7%) 28.4  5.7 123.3  19.9 74.2  12.2 886 (92.1%) 209 (21.9%) 794 (82.5%) 765 (79.5%) 28.5  7.8 24 (19e33) 1.2 (1.0e1.5) 140 (138e143) 592 (261e1,252) 4,839 (2,180e9,570)

582 (58.6%) 150 (15.1%) 145 (14.6%) 117 (11.8%) 29.3  5.3 125.6  19.7 76.3  12.3 900 (91.6%) 172 (17.6%) 802 (81.5%) 745 (75.7%) 29.3  7.6 23 (18e30) 1.2 (1.0e1.4) 141 (138e143) 348 (162e768) 2,148 (949e4,281)

<0.001 <0.001 <0.001 0.424 <0.001 0.686 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001

126 (102e156)

124 (98e153)

121 (96e147)

120 (95e142)

<0.001

291 (29.3%)

311 (31.3%)

299 (30.7%)

233 (23.5%)

<0.001

3.5 (3.2e3.8) 459 (46.4%) 805 (81.4%) 718 (72.5%) 555 (55.9%) 691 (69.7%) 543 (54.8%) 465 (46.9%) 400 (40.4%) 214 (21.6%) 333 (33.6%) 248 (25%) 126 (12.7%)

3.7 (3.3e4.0) 420 (42.3%) 779 (78.5%) 689 (69.2%) 490 (49.2%) 692 (69.5%) 420 (42.5%) 390 (39.2%) 285 (28.6%) 221 (22.3%) 222 (22.3%) 196 (19.7%) 99 (9.9%)

3.8 (3.5e4.2) 374 (38.3%) 761 (78.5%) 691 (70.9%) 476 (48.9%) 700 (71.7%) 411 (42.4%) 343 (35.1%) 211 (21.6%) 200 (20.5%) 150 (15.4%) 138 (14.1%) 86 (8.8%)

4.1 (3.8e4.3) 311 (31.3%) 776 (78.3%) 690 (69.4%) 481 (48.4%) 726 (73.0%) 539 (54.5%) 333 (33.5%) 157 (15.8%) 194 (19.5%) 124 (12.5%) 118 (11.9%) 85 (8.6%)

<0.001 <0.001 0.272 0.343 0.002 0.247 <0.001 <0.001 <0.001 0.455 <0.001 <0.001 0.008

964 (97.3%) 799 (80.6%)

964 (97.2%) 845 (85.2%)

948 (97.5%) 822 (84.6%)

971 (97.7%) 873 (87.9%)

0.887 <0.001

735 511 473 72 479

696 533 526 51 337

666 545 476 27 300

698 562 443 30 254

0.041 0.097 0.002 <0.001 <0.001

(74.2%) (51.6%) (47.7%) (7.3%) (48.3%)

(70.2%) (53.7%) (53.0%) (5.1%) (34.0%)

Data are expressed as mean  SD, as number (percentage), or as median (interquartile range). * Slight, moderate, or marked pedal or sacral edema. † Data available for 681, 707, 709, and 749 patients in quartiles 1 to 4, respectively. z Data available for 387, 362, 334, and 300 patients in quartiles 1 to 4, respectively. x Patient-reported history of hypercholesterolemia.

(68.5%) (56.1%) (49.0%) (2.8%) (30.9%)

(70.2%) (56.5%) (44.6%) (3.0%) (25.6%)

— <0.001 0.354 <0.001 <0.001 0.834 <0.001

Heart Failure/Cholesterol Paradox in Heart Failure

577

Table 2 Causes of death and rehospitalization by baseline total cholesterol quartile Outcome

All-cause mortality Cardiovascular mortality or HF hospitalization Cardiovascular mortality Cardiovascular mortality or cardiovascular hospitalization Worsening HF* HF mortality HF hospitalization Myocardial infarction mortality Myocardial infarction hospitalization Stroke mortality Stroke hospitalization

Quartile of Total Cholesterol (mg/dl)

p Value

1 (49e128) (n ¼ 992)

2 (129e158) (n ¼ 995)

3 (159e196) (n ¼ 976)

4 (197e437) (n ¼ 994)

385 533 298 594 481 187 346 5 9 5 11

290 454 217 506 399 118 307 8 12 6 10

219 372 168 434 317 78 256 6 10 8 12

147 270 119 348 236 49 190 8 11 4 16

(38.8%) (53.7%) (30.0%) (59.9%) (48.5%) (18.9%) (34.9%) (0.5%) (0.9%) (0.5%) (1.1%)

(29.1%) (45.6%) (21.8%) (50.9%) (40.1%) (11.9%) (30.9%) (0.8%) (1.2%) (0.6%) (1.0%)

(22.4%) (38.1%) (17.2%) (44.5%) (32.5%) (8.0%) (26.2%) (0.6%) (1.0%) (0.8%) (1.2%)

(14.8%) (27.2%) (12.0%) (35.0%) (23.7%) (4.9%) (19.1%) (0.8%) (1.1%) (0.4%) (1.6%)

<0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 0.809 0.929 0.655 0.637

* Defined as death from HF, hospitalization for HF, or unscheduled medical office visit for HF.

Table 3 Analysis of primary end points by continuous baseline total cholesterol and triglyceride levels Outcome Total cholesterol All-cause mortality Cardiovascular mortality or HF hospitalization Triglycerides All-cause mortality Cardiovascular mortality or HF hospitalization** Q1 vs Q2 Q2 vs Q3 Q3 vs Q4

Unadjusted HR (95% CI)

p Value

Adjusted HR (95% CI)*

p Value

0.806 (0.778e0.834) 0.836 (0.814e0.859)

<0.001† <0.001x

0.729 (0.629e0.845) 0.733 (0.658e0.817)

<0.001z <0.001k

0.493 (0.431e0.564)

<0.001{

0.634 (0.506e0.793)

<0.001#

1.012 (0.890e1.151) 1.271 (1.112e1.454) 1.389 (1.196e1.613)

0.918 <0.001 <0.001

0.890 (0.780e1.010) 1.180 (1.030e1.360) 0.900 (0.770e1.040)

0.073 0.015 0.159

Q ¼ quartile. HRs and 95% CIs were calculated for every 25 mg/dl increase in total cholesterol or for every unit increase in natural logarithmically transformed triglycerides. HRs were calculated using Cox proportional-hazard models and correspond to the effects of the variable at the beginning of follow-up. Raw coefficients are reported for the time-dependent change in the effect of the primary predictor. * Adjusted for tolvaptan assignment, age (years), gender, geographic region, EF, serum sodium, B-type natriuretic peptide, N-terminal proeB-type natriuretic peptide, blood urea nitrogen, QRS duration on baseline electrocardiogram, NYHA functional class, systolic blood pressure, ischemic cause of HF, serum albumin, body mass index, self-reported history of diabetes, history of hypertension, history of renal insufficiency, history of CAD, atrial fibrillation on baseline electrocardiogram, and use of angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers, b blockers, mineralocorticoid receptor antagonists, digoxin, intravenous inotropic agents, and statins. † Regression coefficient for interaction with ln(time) ¼ 0.003  0.0006 (p <0.001). z Regression coefficient for interaction with ln(time) ¼ 0.002  0.0006 (p <0.001). x Regression coefficient for interaction with ln(time) ¼ 0.003  0.0005 (p <0.001). k Regression coefficient for interaction with ln(time) ¼ 0.002  0.0005 (p <0.001). { Regression coefficient for interaction with time ¼ 0.001  0.0004 (p <0.001). # Regression coefficient for interaction with time ¼ 0.001  0.0004 (p ¼ 0.001). ** Because the effect of triglycerides was not linear, the data were modeled using triglyceride quartiles. Results are reported as comparisons between pairs of adjacent triglyceride quartiles, with risk in the lower quartile compared to risk in the higher quartile.

thus the data were modeled utilizing triglyceride quartiles, with results reported as comparisons between adjacent quartiles, with risk in the lower quartile compared with risk in the higher quartile. For descriptive purposes, baseline characteristics and end point analyses are also presented by quartiles of baseline serum total cholesterol. Total cholesterol was selected because it was considered more reliable than triglyceride level in the setting of nonfasting samples. Demographic characteristics, clinical characteristics, vital signs, other laboratory and diagnostic testing, and enrollment medication use were compared across baseline serum total

cholesterol quartiles. Lipid-lowering agents were divided into statin and nonstatin subgroups for the purposes of this analysis. Fewer than 5% of patients enrolled in EVEREST (n ¼ 194) were taking nonstatin lipid-lowering agents, and their use was not analyzed in this study. Statins used at baseline included atorvastatin, fluvastatin, lovastatin, pravastatin, rosuvastatin, and simvastatin. Figure 1 provides an overview of the patient selection for this post hoc analysis. An independent, blinded clinical events committee adjudicated specific causes of death, hospitalization and unscheduled outpatient visits. All-cause mortality and the

578

The American Journal of Cardiology (www.ajconline.org)

Figure 3. Kaplan-Meier curves for all-cause mortality (A) and cardiovascular mortality or hospitalization for HF (B) by baseline total cholesterol quartile (Q). Times to events were compared using log-rank tests.

composite of cardiovascular mortality and hospitalization for HF, the 2 prespecified EVEREST primary coeend points, were also selected for the present study. End points were analyzed as time to first event. Secondary end points included cardiovascular mortality, worsening HF (defined as death, rehospitalization, or unplanned outpatient visitation), HF rehospitalization, and a composite of cardiovascular mortality and cardiovascular-related hospitalizations. The median follow-up duration was 9.9 months in EVEREST. Categorical variables are reported as number (percentage), while continuous variables are expressed as mean  SD if normally distributed or as median (interquartile range) if not normally distributed. Baseline characteristics across total cholesterol quartiles were compared using 1-way analysis of variance, Kruskal-Wallis tests, and chi-square tests, as appropriate. Univariate and multivariate Cox proportional-hazards models were used to calculate hazard ratios (HRs) and 95% confidence intervals (CIs) for each of the primary predictor variables (as continuous variables). The linearity and proportional-hazards assumptions were tested for the 2 primary end points using Kolmogorov-type supremum tests. Departures from proportional hazards were modeled by introducing time-dependent changes for the effects of the primary predictor in the model. Kaplan-Meier curves were

Figure 4. Kaplan-Meier curves for all-cause mortality (A) and cardiovascular mortality or hospitalization for HF (B) by natural logarithmically transformed baseline triglyceride quartile (Q). Times to events were compared using log-rank tests.

constructed for total cholesterol and triglyceride quartiles, and time to first event was compared using log-rank tests. Twenty-six preselected baseline covariates were included in final multivariate models: tolvaptan assignment, age, gender, region of origin, EF, B-type natriuretic peptide, N-terminal proeB-type natriuretic peptide, serum sodium, blood urea nitrogen, systolic blood pressure, QRS duration on enrollment electrocardiogram, NYHA functional class IV, ischemic cause of HF, serum albumin level, body mass index, baseline co-morbidities (self-reported history of diabetes mellitus, renal insufficiency, hypertension, coronary artery disease [CAD], atrial fibrillation on baseline electrocardiogram), baseline medication use (angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers, b blockers, mineralocorticoid receptor antagonists, digoxin, intravenous inotropes, and statins) and significant interactions. Multicollinearity between primary predictors and control variables did not significantly threaten the results of the multivariate models (tolerance 0.68, variance inflation factor 1.46). Multiple imputation procedure (fully conditional specification methods as implemented in the MI and MIANALYZE procedures in SAS; SAS Institute Inc., Cary, North Carolina) was used for missing covariate data (<5% for all variables).

Heart Failure/Cholesterol Paradox in Heart Failure

Figure 5. Regression plots for albumin by change in baseline total cholesterol (R2 ¼ 0.15) (A) and by change in natural logarithmically transformed baseline triglyceride levels (R2 ¼ 0.10) (B). There was a 0.40  0.015 g/dl (p <0.001) increase in albumin for every 100 mg/dl increase in total cholesterol. There was a 0.34  0.016 g/dl (p <0.001) increase in albumin for every 1 natural logarithmically transformed unit increase in triglycerides.

Prespecified interaction analyses were performed between lipids and (1) tolvaptan assignment, (2) baseline statin therapy, (3) ischemic cause of HF, and (4) history of CAD. All statistical analyses were performed using SAS version 9.3, and p values <0.05 were considered to be statistically significant. Results Patients with missing baseline total cholesterol or triglyceride data (n ¼ 176) were excluded from the primary analysis. In the remaining analytic cohort (n ¼ 3,957), total cholesterol was approximately normally distributed, with a mean of 166  51 mg/dl and a median of 158 mg/dl (interquartile range 128 to 197) (Figure 2). Triglyceride levels were positively skewed, and the variable was natural logarithmically transformed for analytic purposes; the median of the original distribution was 110 mg/dl (interquartile range 83 to 155) (Figure 2). Table 1 lists the baseline characteristics of all patients with available lipid data at the time of enrollment stratified by total cholesterol quartiles. Patients with lower baseline total cholesterol tended to have lower blood pressure, EF, serum sodium, and albumin; higher natriuretic peptide and serum creatinine

579

levels; and wider QRS intervals on electrocardiogram (all p values <0.001). They were more likely to have NYHA class IV HF, to receive angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers and statins, and to have histories of diabetes mellitus, chronic kidney disease, and coronary revascularization (all p values <0.001). Table 2 lists causes of death and rehospitalization by baseline total cholesterol quartile. Outcome analysis is listed in Table 3. Univariate analysis showed baseline total cholesterol and triglycerides to be strongly predictive of the EVEREST primary end points (p <0.001; Figures 3 and 4), with the exception of risk for the composite end point between the 2 lowest triglyceride quartiles. After adjustment for baseline clinical risk factors, total cholesterol and triglyceride levels remained significant independent predictors of outcomes. The reduction in mortality and hospitalization for HF associated with higher baseline lipid levels diminished with time (p 0.001 for all time-dependent components). Interaction analysis revealed 2 significant results: (1) patients receiving background statin therapy experienced a 51% reduction in the hazard of death per 25 mg/dl increase in total cholesterol (HR 0.49, 95% CI 0.40 to 0.60) compared to a 40% reduction in those who were not taking statins (HR 0.60, 95% CI 0.52 to 0.70) (p ¼ 0.003), and (2) the association of total cholesterol with cardiovascular mortality or hospitalization for HF was stronger for patients receiving tolvaptan (HR 0.57, 95% CI 0.51 to 0.65) compared to those who were not (HR 0.64, 95% CI 0.57 to 0.72) (p ¼ 0.018). There were no significant interactions for the remaining combinations of predictors (ischemic cause of HF, history of CAD) and end points (all p values >0.10). The interaction between total cholesterol and tolvaptan was tested and found to be nonsignificant in all multivariate models (p >0.40). Therefore, this parameter was not included in final multivariate models. Discussion In this post hoc analysis, we found that lower serum total cholesterol and triglycerides measured <48 hours after hospital admission for worsening HF were predictive of postdischarge mortality and hospitalization for HF. Although total cholesterol was highly correlated with known risk factors for clinical end points, this lipid parameter retained its prognostic value even after accounting for these traditional covariates. To our knowledge, this is the first comprehensive characterization of the prognostic value of lipids on postdischarge outcomes in a large multinational study of patients hospitalized for HF. In the present analysis, there was a wide range of baseline total cholesterol and triglycerides levels, but it should be noted that most patients were within guideline recommendations.7 However, the mean total cholesterol value and distribution seen here approximate those from a study of ambulatory HF patients, suggesting the existence of similar levels in both the acute and chronic phases.8 Also similar to this chronic HF study, we found lower total cholesterol to be associated with several known predictors of HF and cardiovascular morbidity and mortality, including blood pressure, EF, natriuretic peptide levels, serum creatinine and sodium, QRS duration,

580

The American Journal of Cardiology (www.ajconline.org)

and NYHA class IV symptoms.9 As might be expected, patients with lower total cholesterol were more likely to be receiving background statin therapy. However, patients in the lowest and the highest total cholesterol quartiles were more likely to have self-reported histories of hypercholesterolemia. Together, these findings suggest that increased statin use in this study population was not correlated with a history of hypercholesterolemia per se but rather with other high-risk prognostic characteristics and advanced disease. This study identifies patients hospitalized for HF with reduced EF and with lower baseline total cholesterol and triglycerides as at especially high risk for adverse outcomes. The association of lower baseline total cholesterol and triglycerides with adverse outcomes was strongest in the early postdischarge “vulnerable phase.”10 Interaction analysis demonstrated that although higher total cholesterol was associated with better outcomes regardless of background statin therapy, the relation with all-cause mortality was significantly stronger for patients receiving statins. This may indicate that the ability of hospitalized HF patients to overcome the lipid lowering effects of statins and maintain higher total cholesterol signals improved prognosis. Additional interaction analyses revealed nonsignificant results for HF cause and history of CAD, consistent with previous studies that found lipid levels to be predictive of outcomes in patients with chronic HF of both ischemic and nonischemic origins.11,12 A number of previous studies have revealed an inverse relation between lipid levels and clinical outcomes in patients with stable chronic HF.8,11,12 However, many fewer data are available regarding the proposed relation in patients hospitalized for HF.13 In a retrospective analysis of >17,000 patients admitted for HF included in the Get With the GuidelineseHeart Failure registry, Horwich et al14 found that lower total cholesterol was predictive of in-hospital mortality risk, independent of measured covariates. These data were recently substantiated in a Korean investigation of 2,797 patients hospitalized for HF with a 510-day median follow-up period.15 Although lower total cholesterol was strongly associated with morbidity and mortality in this population, the relation was reduced to a nonsignificant trend after propensity score matching for relevant covariates.15 Our data are also consistent with a number of previous studies suggesting that there may be “reverse epidemiology” in patients with HF in the hospital and outpatient settings, whereby traditional cardiovascular risk factors, including higher cholesterol levels, body mass index, and systolic blood pressure, are paradoxically associated with better outcomes.1 Taken together with our data, these post hoc analyses demonstrate a consistent and strong relation between lower total cholesterol and poor outcomes in patients with HF. The mechanism underlying this association is presently unclear, although a number of plausible hypotheses have been proposed.16e18 Low total cholesterol may reflect a state of malnutrition and cachexia, and lower body mass index and serum albumin are linked with low cholesterol in patients with HF, irrespective of statin therapy.19,20 These same nutritional parameters have also been associated with risk for mortality.21e23 Our study verified a link between lower cholesterol and lower albumin in patients hospitalized for HF (Figure 5). However, in our study, the effect of total

cholesterol on clinical end points was found to be independent of serum albumin levels. The post hoc retrospective design of this study makes establishing a causal relation between various lipid indexes and measured outcomes difficult. Although robust multivariate modeling techniques were used to account for known potential confounders in this analysis, there may be other unmeasured parameters that influenced clinical end points. For example, our analysis lacked data regarding markers of inflammation (i.e., C-reactive protein and erythrocyte sedimentation rate). Additionally, the trial database did not include the full traditional lipid panel. This precluded more global measures of lipid status at the time of enrollment, including the cardiac risk ratio. Given that the baseline lipid profiles were obtained in the acute phase, these samples may not have been fasting in many cases. Lipid panel measurements performed at the time of enrollment may also not be entirely reflective of steady-state lipid levels, potentially secondary to fluid shifts and dilutional effects. Last, the goal of this hypothesis-generating work was to investigate lipid parameters as prognostic markers. Further studies are necessary to characterize total cholesterol and triglyceride levels as potential therapeutic targets and to clarify the role of statins and other lipid-lowering therapies in the treatment of patients hospitalized for HF with systolic dysfunction. In patients hospitalized for worsening HF with reduced EF, baseline total cholesterol and triglyceride levels are independently predictive of mortality and hospitalization for HF. Lower baseline total cholesterol correlates with several high-risk prognostic factors and is a marker of disease severity. The therapeutic implications of these findings are presently unclear. Future prospective studies are necessary to better characterize the clinical role of these lipid indices in patients hospitalized for HF. Disclosures Dr. Gheorghiade has been a consultant for Abbott Laboratories, Abbott Park, Illinois; Astellas Pharma Inc., Tokyo, Japan; AstraZeneca, Wilmington, Delaware; Bayer Healthcare, Tarrytown, New York; Corthera, San Carlos, California; Cytokinetics, South San Francisco, California; DebioPharm S.A., Lausanne, Switzerland; Errekappa Terapeutici, Milan, Italy; GlaxoSmithKline, London, United Kingdom; Ikaria, Hampton, New Jersey; Johnson & Johnson, New Brunswick, New Jersey; Medtronic Inc., Minneapolis, Minnesota; Merck & Company, Whitehouse Station, New Jersey; Novartis Pharma AG, Basel, Switzerland; Otsuka Pharmaceuticals, Tokyo, Japan; Palatin Technologies, Cranbury, New Jersey; PeriCor Therapeutics, New York, New York; Protein Design Laboratories, Fremont, California; Sanofi-Aventis, Paris, France; SigmaTau Pharmaceuticals, Gaithersburg, Maryland; Solvay Pharmaceuticals, Brussels, Belgium; Takeda Pharmaceutical, Osaka, Japan; and Trevena Therapeutics, King of Prussia, Pennsylvania. All other authors have no conflicts of interest to disclose. 1. Kalantar-Zadeh K, Block G, Horwich T, Fonarow GC. Reverse epidemiology of conventional cardiovascular risk factors in patients with chronic heart failure. J Am Coll Cardiol 2004;43:1439e1444.

Heart Failure/Cholesterol Paradox in Heart Failure 2. Anker SD, Negassa A, Coats AJ, Afzal R, Poole-Wilson PA, Cohn JN, Yusuf S. Prognostic importance of weight loss in chronic heart failure and the effect of treatment with angiotensin-converting-enzyme inhibitors: an observational study. Lancet 2003;361:1077e1083. 3. Horwich TB, Fonarow GC. Reverse epidemiology beyond dialysis patients: chronic heart failure, geriatrics, rheumatoid arthritis, COPD, and AIDS. Semin Dial 2007;20:549e553. 4. Gheorghiade M, Orlandi C, Burnett JC, Demets D, Grinfeld L, Maggioni A, Swedberg K, Udelson JE, Zannad F, Zimmer C, Konstam MA. Rationale and design of the multicenter, randomized, double-blind, placebo-controlled study to evaluate the Efficacy of Vasopressin antagonism in Heart Failure: Outcome Study With Tolvaptan (EVEREST). J Card Fail 2005;11:260e269. 5. Gheorghiade M, Konstam MA, Burnett JC Jr, Grinfeld L, Maggioni AP, Swedberg K, Udelson JE, Zannad F, Cook T, Ouyang J, Zimmer C, Orlandi C. Short-term clinical effects of tolvaptan, an oral vasopressin antagonist, in patients hospitalized for heart failure: the EVEREST Clinical Status Trials. JAMA 2007;297:1332e1343. 6. Konstam MA, Gheorghiade M, Burnett JC Jr, Grinfeld L, Maggioni AP, Swedberg K, Udelson JE, Zannad F, Cook T, Ouyang J, Zimmer C, Orlandi C. Effects of oral tolvaptan in patients hospitalized for worsening heart failure: the EVEREST Outcome Trial. JAMA 2007;297: 1319e1331. 7. Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults. Executive summary of the third report of the National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III). JAMA 2001;285:2486e2497. 8. Horwich TB, Hamilton MA, Maclellan WR, Fonarow GC. Low serum total cholesterol is associated with marked increase in mortality in advanced heart failure. J Card Fail 2002;8:216e224. 9. O’Connor CM, Abraham WT, Albert NM, Clare R, Gattis Stough W, Gheorghiade M, Greenberg BH, Yancy CW, Young JB, Fonarow GC. Predictors of mortality after discharge in patients hospitalized with heart failure: an analysis from the Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients With Heart Failure (OPTIMIZE-HF). Am Heart J 2008;156:662e673. 10. Gheorghiade M, Bonow RO. Heart failure: early follow-up after hospitalization for heart failure. Nat Rev Cardiol 2010;7:422e424. 11. Afsarmanesh N, Horwich TB, Fonarow GC. Total cholesterol levels and mortality risk in nonischemic systolic heart failure. Am Heart J 2008;152:1077e1083. 12. Rauchhaus M, Clark AL, Doehner W, Davos C, Bolger A, Sharma R, Coats AJ, Anker SD. The relationship between cholesterol and survival in patients with chronic heart failure. J Am Coll Cardiol 2003;42:1933e1940.

581

13. O’Connor CM, Mentz RJ, Cotter G, Metra M, Cleland JG, Davison BA, Givertz MM, Mansoor GA, Ponikowski P, Teerlink JR, Voors AA, Fiuzat M, Wojdyla D, Chiswell K, Massie BM. The PROTECT in-hospital risk model: 7-day outcome in patients hospitalized with acute heart failure and renal dysfunction. Eur J Heart Fail 2012;14:605e612. 14. Horwich TB, Hernandez AF, Dai D, Yancy CW, Fonarow GC. Cholesterol levels and in-hospital mortality in patients with acute decompensated heart failure. Am Heart J 2008;156:1170e1176. 15. Yoon CH, Youn TJ, Ahn S, Choi DJ, Cho GY, Chae IH, Choi J, Cho H, Han S, Cho MC, Jeon ES, Chae SC, Kim JJ, Ryu KH, Oh BH. Low serum total cholesterol level is a surrogate marker, but not a risk factor, for poor outcome in patients hospitalized with acute heart failure: a report from the Korean Heart Failure Registry. J Card Fail 2012;18: 194e201. 16. Rauchhaus M, Koloczek V, Volk H, Kemp M, Niebauer J, Francis DP, Coats AJ, Anker SD. Inflammatory cytokines and the possible immunological role for lipoproteins in chronic heart failure. Int J Cardiol 2000;76:125e133. 17. May HT, Muhlestein JB, Carlquist JF, Horne BD, Bair TL, Campbell BA, Kfoury AG, Lyon JL, Kim H, Renlund DG. Relation of serum total cholesterol, C-reactive protein levels, and statin therapy to survival in heart failure. Am J Cardiol 2006;98:653e658. 18. Ananaba I, Taegtmeyer H. Low serum cholesterol as prognostic indicator in heart failure. J Card Fail 2012;18:596. 19. Horwich TB, Fonarow GC, Hamilton MA, MacLellan WR, Woo MA, Tillisch JH. The relationship between obesity and mortality in patients with heart failure. J Am Coll Cardiol 2001;38:789e795. 20. Araujo JP, Frioes F, Azevedo A, Lourenco P, Rocha-Goncalves F, Ferreira A, Bettencourt P. Cholesterol—a marker of nutritional status in mild to moderate heart failure. Int J Cardiol 2008;129:65e68. 21. Fonarow GC, Srikanthan P, Costanzo MR, Cintron GB, Lopatin M. ADHERE Scientific Advisory Committee and Investigators. An obesity paradox in acute heart failure: analysis of body mass index and inhospital mortality for 108,927 patients in the Acute Decompensated Heart Failure National Registry. Am Heart J 2007;153:74e81. 22. Arques S, Roux E, Sbragia P, Gelisse R, Pieri B, Ambrosi P. Usefulness of serum albumin concentration for in-hospital risk stratification in frail, elderly patients with acute heart failure. Insights from a prospective, monocenter study. Int J Cardiol 2008;125:265e267. 23. Ambrosy AP, Vaduganathan M, Huffman MD, Khan S, Kwasny MJ, Fought AJ, Maggioni AP, Swedberg K, Konstam MA, Zannad F, Gheorghiade M. Clinical course and predictive value of liver function tests in patients hospitalized for worsening heart failure with reduced ejection fraction: an analysis of the EVEREST trial. Eur J Heart Fail 2012;14:302e311.