European Journal of Heart Failure (2009) 11, 170177

doi:10.1093/eurjhf/hfn031

Liver function abnormalities and outcome

in patients with chronic heart failure: data from
the Candesartan in Heart Failure: Assessment
of Reduction in Mortality and Morbidity
(CHARM) program
Larry A. Allen 1, G. Michael Felker 2*, Stuart Pocock 3, John J.V. McMurray 4,
Marc A. Pfeffer 5, Karl Swedberg 6, Duolao Wang 3, Salim Yusuf 7, Eric L. Michelson 8,
and Christopher B. Granger 2, for the CHARM Investigators
1

Division of Cardiology, University of Colorado Denver, Aurora, CO, USA; 2Duke Clinical Research Institute, Duke University Medical Center, Durham, NC, USA; 3London School
of Hygiene and Tropical Medicine, London, UK; 4University of Glasgow, Glasgow, UK; 5Brigham and Womens Hospital, Boston, MA, USA; 6Department of Medicine, Sahlgrenska
stra, Goteburg, Sweden; 7McMaster University, Hamilton, Ontario, Canada; and 8AstraZeneca LP, Wilmington, DE, USA
University Hospital/O
Received 27 March 2008; revised 26 September 2008; accepted 17 November 2008

Aims

The prevalence and importance of liver function test (LFT) abnormalities in a large contemporary cohort of heart
failure patients have not been systematically evaluated.
.....................................................................................................................................................................................
Methods
We characterized the LFTs of 2679 patients with symptomatic chronic heart failure from the Candesartan in Heart
failure: Assessment of Reduction in Mortality and morbidity program (CHARM). We used multivariable modelling to
and results
assess the relationships between baseline LFT values and long-term outcomes. Liver function test abnormalities were
common in patients with chronic heart failure, ranging from alanine aminotransferase elevation in 3.1% of patients to
low albumin in 18.3% of patients; total bilirubin was elevated in 13.0% of patients. In multivariable analysis, elevated
total bilirubin was the strongest LFT predictor of adverse outcome for both the composite outcome of cardiovascular death or heart failure hospitalization (HR 1.21 per 1 SD increase, P , 0.0001) and all-cause mortality
(HR 1.19 per 1 SD increase, P , 0.0001). Even after adjustment for other variables, elevated total bilirubin was
one of the strongest independent predictors of poor prognosis (by global chi-square).
.....................................................................................................................................................................................
Conclusion
Bilirubin is independently associated with morbidity and mortality. Changes in total bilirubin may offer insight into the
underlying pathophysiology of chronic heart failure.

----------------------------------------------------------------------------------------------------------------------------------------------------------Keywords

Heart failure Bilirubin Liver function tests Hepatic congestion Prognosis Death Laboratory tests

Introduction
Chronic heart failure is a systemic clinical syndrome with a variety
of potential effects on other organ systems. Recently, the interactions between heart failure and the kidney and bone marrow
have been the subject of significant interest and investigation,1 3
but the interactions between heart failure and other organ
systems have not been carefully studied. A variety of liver

abnormalities and cardio-hepatic syndromes have been previously

described in patients with heart failure.4,5 Passive hepatic congestion due to increased central venous pressure is believed to
cause elevations of both direct and indirect serum bilirubin
(congestive hepatopathy), with nutmeg liver on pathology.
Impaired perfusion from decreased cardiac output is associated
with acute hepatocellular necrosis (hepatic ischaemia) with
elevations primarily in serum aminotransferases.6 Cardiogenic

* Corresponding author. Tel: 1 919 668 8919, Fax: 1 919 668 7063, Email: michael.felker@duke.edu
Published on behalf of the European Society of Cardiology. All rights reserved. & The Author 2009. For permissions please email: journals.permissions@oxfordjournals.org.

171

Liver function tests in symptomatic heart failure

ischaemic hepatitis (shock liver) may ensue following an episode

of profound hypotension. Bridging fibrosis (cardiac cirrhosis)
can result from prolonged haemodynamic abnormalities, resulting
in an impaired hepatic function with impaired coagulation and
decreased albumin synthesis.4
Although a variety of cardio-hepatic syndromes are recognized
by heart failure clinicians, systematic characterization of liver function abnormalities in a well characterized, contemporary cohort of
patients with heart failure has not been performed. Prior studies
on hepatic abnormalities in heart failure have reported various
patterns in small, selected populations, ranging from predominantly
hepatocellular7 to predominantly cholestatic8,9 to mixed.5,10 The
prognostic importance of observed abnormalities in biochemical
markers of hepatic function [liver function tests (LFTs)] has
varied widely between published studies.
The purpose of this study was to prospectively characterize
LFTs in a large, representative, well-treated cohort of patients
with chronic heart failure (with both preserved and impaired left
ventricular systolic function), as well as explore the prognostic
information such markers provide. Specifically, we evaluated the
association between LFTs and outcome in the Candesartan in
Heart failure: Assessment of Reduction in Mortality and morbidity
(CHARM) program.

Methods
Source population
Details of the study design and primary findings of the CHARM program
have been published elsewhere.11 14 Briefly, patients were eligible who
were aged 18 years or older and had symptomatic chronic heart failure
of at least 4 weeks duration. The major exclusion criteria were serum
creatinine 265 mmol/L or more, serum potassium of 5.5 mmol/L or
more, myocardial infarction (MI) or stroke in the prior 4 weeks, or
non-cardiac disease (e.g. cancer) judged to limit 2-year survival; there
were no exclusion criteria specifically related to liver function or
LFTs. The program consisted of three separate trials that shared these
inclusion and exclusion criteria, endpoint definitions, and follow-up
methods. Patients were enrolled (i) with left ventricular ejection fraction
.40% (CHARM-Preserved), (ii) with angiotensin converting enzyme
inhibitor intolerance and left ventricular ejection fraction 40%
(CHARM-Alternative), and (iii) on angiotensin converting enzyme
inhibitor with left ventricular ejection fraction 40% (CHARM-Added).
Patients were enrolled at 618 sites in 26 countries between March 1999
and March 2001, and patients were followed until March 2003 for a
median follow-up of 38 months. The primary combined endpoint of
each constituent trial was time to cardiovascular death or hospitalization
for the management of worsening heart failure (adjudicated by a clinical
events committee) and the primary endpoint of the overall program was
all-cause mortality.
Data from routine laboratory testing at a central core laboratory
were collected on all patients enrolled in the CHARM program in
the USA and Canada (n 2679), who served as the population for
this analysis. This included patients from all three component trials
(n 1085 from CHARM-Preserved, n 931 from CHARM-Added,
n 663 from CHARM-Alternative; median follow-up 34 months).
Baseline laboratory measurements were obtained at the time of randomization. For the purposes of this analysis, we separately analysed
both the primary endpoint of the constituent trials (composite of cardiovascular death or hospitalization for worsening heart failure, 952

events) and the primary endpoint of the overall CHARM program (allcause mortality, 625 events).

Statistical methods
Baseline variables were described using medians and interquartile
ranges for continuous variables and percentages for categorical variables. A Cox proportional hazards model was used to evaluate the
relationship between potential predictor variables of interest (specifically LFTs) and the primary study endpoint (time to cardiovascular
death or heart failure hospitalization). The multivariable modelling
approach used in the CHARM program has been described elsewhere.15 Briefly, a best clinical model (excluding laboratory data)
was created from the overall trial dataset (n 7599) using standard
modelling techniques. Models were built using a forward stepwise variable selection procedure. P , 0.01 was set as level of significance for
including variables in the model due to the large number of candidate
variables being considered. Randomization to candesartan or placebo
was included as a variable in all models. The variables selected for
this model were then used to estimate a final model for the study subcohort for whom laboratory data were available (n 2679). Subsequently, laboratory variables were added to this model one at a
time in a forward stepwise fashion to generate a final model that combined both clinical and laboratory variables. Adjusted hazard ratios
(HRs) for continuous variables were described using standardized
HRs, the HR associated with one standard deviation change in the
variable. The statistical contribution of each variable to the prediction
of outcome was assessed by the chi-square (x2) statistic with one
degree of freedom (the larger the x2, the stronger the statistical association with outcome).
To evaluate the potential effects of non-normal distribution of
laboratory values on our findings, we repeated the modelling with
log transformation on all laboratory variables that had skew
distributions. Additionally, we evaluated the possibility of non-linear
relationships between variables and outcome by repeating the
models with a quadratic term included for each laboratory variable.

Results
Baseline clinical characteristics for patients enrolled in the CHARM
program in North America (n 2679), including LFTs, are shown
in Table 1. In general, LFTs (serum values) were within the normal
range for most patients in the study cohort. The most common
LFT abnormalities were low albumin (below lower limit of
normal in 18.3% of patients), elevated alkaline phosphatase
(above upper limit of normal in 14.0% of patients), and elevated
total bilirubin (above upper limit of normal in 13.0% of patients).
Less commonly abnormal were alanine aminotransferase (above
upper limit of normal in 3.1% of patients), aspartate aminotransferase (above upper limit of normal in 4.1% of patients), and direct
bilirubin (above upper limit of normal in 3.7% of patients).
The proportion of patients with impaired left ventricular systolic
function (ejection fraction 0.40, n 1594) who had elevations in
total bilirubin (15.8%) was almost double that of patients with preserved left ventricular systolic function (ejection fraction .0.40,
n 1085; elevated total bilirubin in 9.4%; P , 0.001 for the comparison). Other LFTs were not markedly different in patients with
impaired vs. preserved left ventricular systolic function.
Total bilirubin was modestly higher (mean 11.2 mmol/L) in
patients who had significant evidence of volume overload on

172

L.A. Allen et al.

Table 1 Selected baseline characteristics for CHARM North American Study cohort
Variable

Overall

......................................

CV death or HF hospitalization

..................................................................................

Without event (n 5 1727)

With event (n 5 952)

Median or %

Median or %

......................................

Median or %

IQR

IQR

.....................................
IQR

...............................................................................................................................................................................
Demographic and clinical
Age (years)
Gender (female)

67
33.4%

Race (white)

85.2%

Ejection fraction (left ventricular)

Heart rate (bpm)

0.36
71

Systolic blood pressure

Diastolic blood pressure
NYHA class II

128
72
36.2%

5774

65
34.3%

56 73

84.8%
0.260.5
6480
114 140
6680

0.38
70
130
74
42.7%

69
31.7%

6076

85.8%
0.28 0.52
62 80
118 140
68 80

0.32
72
124
70
24.5%

NYHA class III/IV

63.8%

57.4%

75.6%

Previous hospitalization for HF

Aetiology (ischaemic)

67.8%
58.7%

62.4%
65.0%

77.7%
71.5%

0.220.45
6480
110140
6480

Medical history
Previous myocardial infarction

53.3%

51.1%

57.1%

Hypertension
Diabetes mellitus

66.7%
37.2%

65.8%
30.8%

68.2%
48.8%

Atrial fibrillation

11.4%

25.3%

35.2%

Coronary artery bypass grafting

Implanted cardiac defibrillator

32.3%
4.1%

30.8%
3.5%

34.9%
5.1%

Venous congestion
S3 gallop

40.2%
16.2%

36.1%
12.9%

47.7%
22.1%

Cardiomegaly

12.5%

8.4%

20.1%

12.8%

10.5%

16.8%

Signs

Pulmonary crackles
Laboratory testing (serum)
Bilirubin total (mmol/L)

10

6.8413.68

Bilirubin direct (mmol/L)

ALT (SGPT) (U/L)

2
18

1.714
13 25

AST (SGOT) (U/L)

19

16 24

19

16 24

20

1624

Alkaline phosphatase (U/L)

Albumin (g/L)

81
41

66 100
39 43

79
42

66 100
39 43

83
41

6597
39.544

141

139 142

141

139 142

140

Sodium (mmol/L)
Creatinine (mg/dL)
Urea nitrogen (mmol/L)
Haemoglobin (g/dL)
RDW (%)

1.10
8.27
8.43
14.4

0.91.3
6.511.31
7.89.1
13.515.4

8.55
2
18

1
7.83
8.53
14.1

6.84 13.68

10.26

6.8412

1.71 4
13 25

3.42
17

1.713.42
1325

0.9 1.3
6.5 11.31
7.8 9.1
13.5 15.4

1.20
9.57
8.30
14.8

139142
0.801.30
6.0910.01
7.909.18
6.8412

All variables listed in this table were included in multivariable modelling. Liver function test variables are presented in bold.
CV, cardiovascular; HF, heart failure; IQR, interquartile range; NYHA, New York Heart Association functional class; HF, heart failure; ALT (SGPT), alanine aminotransferase
(serum glutamate-pyruvate transaminase); AST (SGOT), aspartate aminotransferase (serum glutamic-oxaloacetic transaminase); RDW, red cell distribution width.

physical examination (S3 gallop, pulmonary crackles, and/or venous

congestion) compared to those without these three signs (total
bilirubin mean 10.5 mmol/L, P 0.015).

Association of liver function tests

with outcomes
Baseline abnormalities in bilirubin, alkaline phosphatase, and
albumin were more common in patients who died or were

hospitalized in follow-up (Figure 1). Because patients are typically

first evaluated in clinical practice based on their clinical characteristics, a best clinical model was initially constructed using clinical
characteristics alone (without laboratory values). In this clinical
model, the most powerful predictors of outcome were diabetes
mellitus, hospitalization for heart failure within the previous 6
months, age, and left ventricular ejection fraction.15 The 36 candidate laboratory variables, including LFTs, were then evaluated to
identify associations between each individual biochemical variable

Liver function tests in symptomatic heart failure

173

Figure 1 Prevalence of baseline liver function test abnormalities stratified by occurrence of cardiovasular death or heart failure hospitalization
during follow-up. Legend: "Total bili, total bilirubin above upper limit of normal; "ALT, alanine transaminase above upper limit of normal; "Alk
Phos, alkaline phosphatase above upper limit of normal; #Albumin, albumin below lower limit of normal; *primary study outcome was CV death
or HF hospitalization at a mean follow up of 3.1 years.

and outcome. In univariate analysis of laboratory predictors

(adjusted for the variables in the final clinical model but not for
other laboratory variables), total bilirubin (HR 1.17 per 1 SD
increase, P , 0.0001), direct bilirubin (HR 1.15 per 1 SD increase,
P , 0.0001), alkaline phosphatase (HR 1.08 per 1 SD increase,
P 0.003), and albumin (HR 0.91 per 1 SD increase, P 0.003)
were significant predictors of outcome (Table 2). In order to
evaluate these laboratory variables in the context of all available
clinical information, a final multivariable model was generated
that included all significant laboratory and clinical predictors
(Table 3). In this final model, total bilirubin (adjusted HR 1.14
per 1 SD increase, P , 0.0001) was among the most highly
significant overall predictors of adverse cardiovascular outcome,
showing greater association with outcome (based on global x2
with one degree of freedom) than many traditional measures of
risk, including New York Heart Association functional class,
left ventricular ejection fraction, serum creatinine, and diabetes
mellitus. Of clinical variables tested, only age, cardiomegaly,
hospitalization for heart failure in the previous 6 months, and
red cell distribution width showed a greater independent
association with outcome than total bilirubin. Adjusted HRs by
quintiles of total bilirubin are shown in Figure 2A.
To further evaluate the association between bilirubin and clinical
outcomes, we repeated this modelling procedure using the endpoint of all-cause mortality (625 events) rather than the composite
of cardiovascular death or heart failure hospitalization. In this
mortality analysis, total bilirubin remained a highly significant

independent predictor of outcome (adjusted HR 1.19 per 1 SD

increase, 95% confidence interval 1.11 1.27, P , 0.0001) after
adjustment for other clinical and laboratory measures (Figure 2B).
The global x2 for total bilirubin was the second highest among
laboratory predictors and the fourth highest among all predictors
of all-cause mortality, behind age, left ventricular ejection fraction,
and serum chloride.
None of the LFT values were normally distributed in the study
population, so we repeated the modelling procedure using log
transformation of individual LFT variables prior to insertion in
the model. This did not result in any substantive change in our findings, with total bilirubin remaining a powerful predictor of adverse
outcome (HR 1.44 per 1 SD increase, P , 0.0001) using the model
with log-transformed variables. Inclusion of a quadratic term in the
multivariable model (total bilirubin squared) did not change the
statistical significance of the relationship, suggesting that there
was no statistical evidence of a non-linear or j-shaped relationship
between bilirubin and outcomes.
Although HRs for continuous predictors expressed as per SD
increase provide standardization and maximize predictive power,
such information may be difficult to use in routine clinical practice.
To simplify interpretation, total bilirubin was dichotomized
into normal (17 mmol/L) and abnormal (.17 mmol/L). After
adjustment, the HR for an abnormal total bilirubin was 1.55 (CI
1.30 1.84, P , 0.001) for cardiovascular death or heart failure
hospitalization, and 1.60 (CI 1.301.97, P , 0.001) for all cause
mortality.

174

L.A. Allen et al.

Table 2 Univariate hazard ratios for cardiovascular death or heart failure hospitalization for laboratory variables
(adjusted for clinical variables)
Variablea

Mean

SD

Hazard ratiob

Chi-square

P-value

RDW (%)
Uric acid (mmol/L)

14.7
417

1.8
127

1.23
1.24

48.79
44.49

,0.0001
,0.0001

...............................................................................................................................................................................

Urea nitrogen (mmol/L)

9.68

5.11

1.19

40.62

,0.0001

Creatinine (mg/dL)
Bilirubin total (mmol/L)

1.17
11.1

0.63
6.5

1.09
1.17

33.24
31.84

,0.0001
<0.0001

3.0

2.6

1.15

28.65

<0.0001

25.5
1.11

8.6
0.19

0.83
1.15

24.91
16.41

,0.0001
,0.0001

Bilirubin direct (mmol/L)

Lymphocytes (%)
Phosphorus inorganic (mmol/L)
Neutrophils segmented (%)
MCHC (mmol/L)
Chloride (mmol/L)
Haemoglobin (mmol/L)

64.5

9.4

1.15

15.18

,0.0001

20.7
102.5

0.7
3.8

0.88
0.88

13.99
13.99

0.0002
0.0002

0.88

12.76

0.0004

Alkaline phosphatase (U/L)

Albumin (g/L)

87.3
41.2

39.0
3.2

1.08
0.91

9.11
8.63

0.003
0.003

MCH (pg)

30.6

2.3

0.91

8.24

0.004

Globulin total (g/L)

White cell count (109/L)

31.1
7.4

5.0
2.2

1.09
1.08

7.43
6.92

0.006
0.009

Haematocrit (%)
Creatine kinase (U/L)
Monocytes (%)
MCV (fL)
Glycohaemoglobin A1C (%)

8.43

1.00

40.8

4.7

0.91

6.62

0.01

110.9
6.4

109.0
2.6

0.92
1.07

4.23
3.66

0.04
0.06

91.9

5.9

0.94

3.50

0.06

1.52

1.07

3.18

0.07

7.01

The univariate hazard ratio for sodium was 0.97. Liver function test variables are presented in bold. Abbreviations same as Table 1.
a
All variables with P , 0.1 are shown.
b
Hazard ratios shown are standardized hazard ratios (the HR per 1 standard deviation).

Discussion
The principal findings of our analysis are that mild abnormalities of
LFTs are relatively common in patients with chronic heart failure,
with a predominately cholestatic pattern of greater elevation
in bilirubin than transaminases. Additionally, total bilirubin is a
strong, independent predictor of adverse prognosis, even after
adjustment for a wide variety of demographic, clinical, and
biochemical variables. These data provide a comprehensive
evaluation of the prevalence and prognostic importance of LFT
abnormalities in chronic heart failure with both preserved and
impaired left ventricular systolic function.
Prior studies investigating biochemical abnormalities of liver
function have been relatively small and evaluated highly selected
populations. Kubo et al.8 analysed biochemical profiles and intracardiac haemodynamics in 133 patients with stable chronic congestive heart failure due to systolic dysfunction, and reported that
changes in LFTs were common but typically of small magnitude,
particularly in those patients with a cardiac index of .1.5 L/min/m2.
Mild elevations in alkaline phosphatase and total bilirubin, as well
as mild decreases in albumin, were frequent in this population
regardless of haemodynamics. With more significant decreases in
cardiac output and associated increases in cardiac filling pressures,
elevations in transaminases, lactate dehydrogenase, and total
bilirubin were seen, but did not correlate with clinically apparent

hepatic disease. A retrospective study of 110 patients hospitalized

with heart failure with both impaired and preserved left ventricular
function confirmed the predominantly cholestatic pattern and
showed that increases in LFTs were most strongly correlated
with severity of tricuspid regurgitation.9 A longitudinal prospective
study of 552 chronic heart failure patients found that all abnormal
LFTs were significantly related to mortality, but that aspartate
transaminase (x2 17.36, P , 0.001) accounted for the greatest
variance followed by total bilirubin (x2 14.35, P , 0.001).10
Variability between published studies may relate to differences in
the population studied (acute vs. chronic heart failure) or to
greater use of contemporary medical therapy (such as b-blockers)
in the CHARM program. CHARM excluded patients with acute
decompensated heart failure, symptomatic hypotension, recent
MI, or serum creatinine .3.0 mg/dL at the time of enrolment,
and only 2.6% of subjects had New York Heart Association
functional class IV heart failure.11 Thus, as was confirmed in our
results, hypotension-related elevations in transaminases would
have been unexpected at the time of enrolment in an ambulatory
population of heart failure patients.
Clinically, the validation of elevated bilirubin as a strong
independent predictor of outcome has potentially important
value for risk stratification of chronic heart failure patients.
Unlike the only other published study of LFT association with
outcome,10 our study had sufficient numbers of events to

175

Liver function tests in symptomatic heart failure

Table 3 Final multivariable model for cardiovascular death or heart failure hospitalization (including laboratory
variables and clinical variables)
Variables

Hazard ratioa

95%CI

Chi-square

P-value

Age (per 10 years over age 60)

Cardiomegaly

1.32
1.56

1.211.44
1.331.84

41.32
28.15

,0.0001
,0.0001

...............................................................................................................................................................................

RDW

1.17

1.101.25

24.78

,0.0001

Bilirubin total
Prior HF Hosp within 6 months

1.14
1.53

1.081.21
1.281.84

20.72
21.11

<0.0001
,0.0001

NYHA class IV

2.09

1.522.89

20.43

,0.0001

NYHA class III

Ejection fraction (per 5% decrease below 45)

1.42
1.59

1.211.66
1.291.97

18.88
18.24

,0.0001
,0.0001

Diabetes: other

1.07

1.041.11

17.62

,0.0001

Diagnosis of HF over 2 years ago

Lymphocytes

1.36
0.86

1.171.57
0.800.93

16.62
15.09

,0.0001
,0.0001

Randomization to candesartan

0.79

0.690.90

12.59

0.0004

Uric acid
Diabetes: insulin treated

1.13
1.31

1.061.21
1.101.56

12.20
8.84

0.0005
0.003

Glycohaemoglobin A1c

1.12

1.041.22

8.46

0.004

Bundle branch block abnormal

Haemoglobin

1.24
0.90

1.071.44
0.840.97

8.34
8.15

0.004
0.004

Creatinine

1.13

1.031.23

6.79

0.009

Phosphorus inorganic

1.08

1.011.16

5.45

0.02

SDs are given in Table 2. Liver function test variables are presented in bold. Abbreviations same as Table 1.
a
Hazard ratios for laboratory parameters shown as standardized hazard rations (HR per 1 SD).

investigate a wide range of prognostically important clinical and

laboratory variables. In final analysis, total bilirubin was among
the most powerful contributors to the prognostic model for
both the composite of cardiovascular death or worsening heart
failure (fifth most powerful overall predictor) and for all-cause
mortality (fourth most powerful overall predictor). Total bilirubin
is a widely available, low-cost, non-invasive test, which is often
measured in the course of routine inpatient care. Our results
suggest that bilirubin in combination with automated blood
count, basic metabolic panel, and some straightforward clinical
characteristics (age, recent hospitalization, New York Heart
Association functional class) may provide a powerful estimation
of overall risk of morbidity and mortality in patients with
symptomatic chronic heart failure.

Mechanisms
Our results are consistent with existing knowledge of hepatic
pathophysiology. Compared to other organs such as the kidney,
the livers complex dual blood supply makes it relatively resistant
to hepatocyte necrosis from haemodynamic perturbations. Thus,
significant transaminase elevations are expected only in cases of
marked hypotension or hypoperfusion. Previous haemodynamic
data suggest that the elevated right atrial pressure may contribute
to cholestatic abnormalities of liver function in patients with heart
failure.8 Elevated central venous pressure is transmitted directly to
the hepatic veins, leading to impairment of hepatocyte function.4
Additionally, the extent of biochemical liver abnormalities may
be related to the severity of tricuspid regurgitation.9 Right

ventricular dysfunction, typically the cause of elevated venous

pressure and tricuspid regurgitation, has been shown to be a significant marker for adverse outcomes in heart failure patients.16
In our study, total bilirubin was significantly higher in patients
who had evidence of volume overload on physical examination
compared to patients who did not. It seems likely that the prognostic importance of total bilirubin in our study at least in part
reflects greater elevations in central venous pressure (whether
due to right ventricular dysfunction or tricuspid regurgitation)
and therefore more severe haemodynamic perturbation. Other
signs of elevated central venous pressure, such as elevated
jugular venous pressure, have also been shown to be associated
with adverse prognosis in chronic heart failure.17 Even though
signs of congestion were similar for patients with preserved and
impaired left ventricular systolic function, those with impaired
left ventricular systolic function had significantly higher total
bilirubin, arguing that the low output may also contribute to the
cholestatic abnormalities seen in these heart failure patients.
Alternatively, this lack of difference in signs of volume overload
between patients with preserved and impaired systolic function
may reflect the relatively poor reproducibility of the physical
examination.
Abnormalities of the haematologic system in advanced heart
failure provide a potential, but less likely, mechanism for elevations
of bilirubin. Anaemia is a well-known predictor of adverse outcome
in heart failure patients,1 and haemolysis increases indirect bilirubin.
In our study, total and indirect bilirubin were better predictors
of outcome than direct bilirubin. However, total bilirubin was

176

L.A. Allen et al.

Figure 2 Adjusted hazard ratios by quintile of total bilirubin for (A) cardiovascular death or heart failure hospitalization and (B) all-cause
mortality in CHARM.

not significantly correlated with haemoglobin or red cell distribution width (Pearson correlation coefficients 0.18 and 0.20,
respectively).
Of note, bilirubin has recently been shown to be inversely
associated with the risk of MI in a cohort from the Framingham
Offspring study.18 Bilirubin may act as an anti-oxidant and free
radical scavenger, and has been hypothesized to decrease rates
of incidence MI through anti-atherosclerotic properties. In contrast

to patients enrolled in the Framingham Offspring study, the

patients enrolled in CHARM already had significant heart failure
at the time of study entry, with high rates of concomitant preexisting coronary disease (67% with ischaemic cardiomyopathy).
Thus, it is likely that any potential benefits related to the antioxidant effects of bilirubin are outweighed by elevated bilirubins
reflection of elevated right-sided filling pressures in patients with
greater degrees of heart failure.

177

Liver function tests in symptomatic heart failure

Limitations and generalizability

References

Other than the baseline LFTs, we do not have additional information on liver pathology, such as serial LFTs, hepatitis serologies,
liver imaging, haemodynamic measures, or histologic tissue. Eligibility criteria excluding patients with non-cardiac disease judged
to limit 2 year survival should have limited the potential impact
cancer and cirrhosis might have had on our findings; in follow-up,
there were only 45 deaths due to cancer. Patients enrolled in the
CHARM program were symptomatic but stable outpatients at the
time of enrolment. Consequently, our findings do not reflect either
the clinical characteristics or the LFT profiles of patients hospitalized with acute decompensated heart failure. As in other randomized clinical trials, the very elderly and patients with severe
comorbid conditions were less likely to be enrolled in the
CHARM program. Because the hyperbilirubinaemia of Gilberts
does not carry with it the same negative predictive value as the
hyperbilirubinaemia of advancing heart failure, the potential for
spurious prognostication exists in the application of these findings
to heart failure patients who also have Gilberts. Given the independent predictive value of hyponatraemia seen in other survival
models of chronic heart failure,19,20 it was surprising that the
sodium was not retained in the final CHARM risk model. We
speculate that colinearity between the sodium and other predictor
variables may in part explain its absence from the final multivariable model. This study is a retrospective review of a clinical trial
dataset, and as such is subject to the limitations of this type of
analysis. Importantly, however, the CHARM cohort is generally
representative of the epidemiology and treatment milieu of contemporary heart failure, with a high proportion of patients with
preserved systolic function and widespread use of contemporary
medical therapy. Additionally, the large number of events in our
cohort (952 for the composite endpoint and 625 events for
all-cause mortality) greatly strengthened our ability to make conclusions about the relative prognostic information contained in a
variety of clinical and laboratory variables.

1. Felker GM, Allen LA, Pocock SJ, Shaw LK, McMurray JJ, Pfeffer MA, Swedberg K,
Wang D, Yusuf S, Michelson EL, Granger CB. Red cell distribution width as a
novel prognostic marker in heart failure: data from the CHARM Program and
the Duke Databank. J Am Coll Cardiol 2007;50:40 47.
2. de Silva R, Nitikin NP, Witte KK, Rigby AS, Goode K, Bhandari S, Clark AL,
Cleland JG. Incidence of renal dysfunction over 6 months in patients with
chronic heart failure due to left ventricular systolic dysfunction: contributing
factors and relationship to prognosis. Eur Heart J 2006;27:569581.
3. Damman K, Navis G, Voors AA, Asselbergs FW, Smilde TD, Cleland JG, van
Veldhuisen DJ, Hillege HL. Worsening renal function and prognosis in heart
failure: systematic review and meta-analysis. J Card Fail 2007;13:698 700.
4. Giallourakis CC, Rosenberg PM, Friedman LS. The liver in heart failure. Clin Liver
Dis 2002;6:947 967. viii ix.
5. Naschitz JE, Slobodin G, Lewis RJ, Zuckerman E, Yeshurun D. Heart diseases
affecting the liver and liver diseases affecting the heart. Am Heart J 2000;140:
111 120.
6. Seeto RK, Fenn B, Rockey DC. Ischemic hepatitis: clinical presentation and
pathogenesis. Am J Med 2000;109:109 113.
7. Killip T, Payne MA. High serum transaminase activity in heart disease: circulatory
failure and hepatic necrosis. Circulation 1960;21:273 293.
8. Kubo SH, Walter BA, John DH, Clark M, Cody RJ. Liver function abnormalities in
chronic heart failure. Influence of systemic hemodynamics. Arch Intern Med 1987;
147:1227 1230.
9. Lau GT, Tan HC, Kritharides L. Type of liver dysfunction in heart failure and
its relation to the severity of tricuspid regurgitation. Am J Cardiol 2002;90:
1405 1409.
10. Batin P, Wickens M, McEntegart D, Fullwood L, Cowley AJ. The importance of
abnormalities of liver function tests in predicting mortality in chronic heart
failure. Eur Heart J 1995;16:1613 1618.
11. Pfeffer MA, Swedberg K, Granger CB, Held P, McMurray JJ, Michelson EL,
Olofsson B, Ostergren J, Yusuf S, Pocock S. Effects of candesartan on mortality
and morbidity in patients with chronic heart failure: the CHARM-Overall
programme. Lancet 2003;362:759 766.
12. Granger CB, McMurray JJ, Yusuf S, Held P, Michelson EL, Olofsson B, Ostergren J,
Pfeffer MA, Swedberg K. Effects of candesartan in patients with chronic heart
failure and reduced left-ventricular systolic function intolerant to
angiotensin-converting-enzyme inhibitors: the CHARM-Alternative trial. Lancet
2003;362:772 776.
13. McMurray JJ, Ostergren J, Swedberg K, Granger CB, Held P, Michelson EL,
Olofsson B, Yusuf S, Pfeffer MA. Effects of candesartan in patients with chronic
heart failure and reduced left-ventricular systolic function taking angiotensinconverting-enzyme inhibitors: the CHARM-Added trial. Lancet 2003;362:
767 771.
14. Yusuf S, Pfeffer MA, Swedberg K, Granger CB, Held P, McMurray JJ, Michelson EL,
Olofsson B, Ostergren J. Effects of candesartan in patients with chronic heart
failure and preserved left-ventricular ejection fraction: the CHARM-Preserved
Trial. Lancet 2003;362:777 781.
15. Pocock SJ, Wang D, Pfeffer MA, Yusuf S, McMurray JJ, Swedberg KB, Ostergren J,
Michelson EL, Pieper KS, Granger CB. Predictors of mortality and morbidity in
patients with chronic heart failure. Eur Heart J 2006;27:65 75.
16. Sun JP, James KB, Yang XS, Solankhi N, Shah MS, Arheart KL, Thomas JD,
Stewart WJ. Comparison of mortality rates and progression of left ventricular
dysfunction in patients with idiopathic dilated cardiomyopathy and dilated
versus nondilated right ventricular cavities. Am J Cardiol 1997;80:1583 1587.
17. Drazner MH, Rame JE, Stevenson LW, Dries DL. Prognostic importance of
elevated jugular venous pressure and a third heart sound in patients with heart
failure. N Engl J Med 2001;345:574 581.
18. Lin JP, ODonnell CJ, Schwaiger JP, Cupples LA, Lingenhel A, Hunt SC, Yang S,
Kronenberg F. Association between the UGT1A1*28 allele, bilirubin levels, and
coronary heart disease in the Framingham Heart Study. Circulation 2006;114:
1476 1481.
19. Levy WC, Mozaffarian D, Linker DT, Sutradhar SC, Anker SD, Cropp AB,
Anand I, Maggioni A, Burton P, Sullivan MD et al The Seattle Heart Failure
Model: prediction of survival in heart failure. Circulation 2006;113:1424 1433.
20. Aaronson KD, Schwartz JS, Chen TM, Wong KL, Goin JE, Mancini DM. Development and prospective validation of a clinical index to predict survival in
ambulatory patients referred for cardiac transplant evaluation. Circulation 1997;
95:2660 2667.

Conclusions
Bilirubin is frequently elevated in patients with symptomatic
chronic heart failure and is strongly associated with morbidity
and mortality after adjusting for a wide array of other clinical and
laboratory predictors. Further research into the complex relationship between cardiac and hepatic function in heart failure may
improve both understanding of pathophysiology and the clinical
care of heart failure patients.
Conflict of interest: none declared.

Funding
The CHARM program was funded by AstraZeneca. Drs M.A.P., K.S.,
J.J.V.M., S.Y., C.B.G., and S.P. have served as consultants to or received
research grants and honoraria from AstraZeneca. Dr E.L.M. is an
employee of AstraZeneca.