Depression increasingly predicts mortality in the course of congestive heart failure

www.elsevier.com/locate/heafai

The European Journal of Heart

Depression increasingly predicts mortality in the course of congestive

heart failure

Jana Jqngera,*, Dieter Schellberga, Thomas Mqller-Tascha, Georg Rauppa, Christian Zugckb,

Armin Haunstetterb, Stephan Zipfela, Wolfgang Herzoga, Markus Haassb,c

aDepartment of General Internal and Psychosomatic Medicine, University of Heidelberg, INF 410, D-69120 Heidelberg, GermanybDepartment of Cardiology, University of Heidelberg, Heidelberg, GermanycDepartment of Cardiology, Theresienkrankenhaus, Mannheim, Germany

Received 8 August 2003; received in revised form 5 May 2004; accepted 24 May 2004

Available online 27 October 2004

at unihd on April 28, 2011

eurjhf.oxfordjournals.orgD

ownloaded from

Abstract

Background: Congestive heart failure (CHF) is frequently associated with depression. However, the impact of depression on prognosis has

not yet been sufficiently established.

Aims: To prospectively investigate the influence of depression on mortality in patients with CHF.

Methods: In 209 CHF patients depression was assessed by the Hospital Anxiety and Depression Scale (HADS-D).

Results: Compared to survivors (n=164), non-survivors (n=45) were characterized by a higher New York Heart Association (NYHA)

functional class (2.8F0.7 vs. 2.5F0.6), and a lower left ventricular ejection fraction (LVEF) (18F8 vs. 23F10%) and peakVO2 (13.1F4.5

vs. 15.4F5.2 ml/kg/min) at baseline. Furthermore, non-survivors had a higher depression score (7.5F4.0 vs. 6.1F4.3) (all Pb0.05). After a

mean follow-up of 24.8 months the depression score was identified as a significant indicator of mortality (Pb0.01). In multivariate analysis

the depression score predicted mortality independent from NYHA functional class, LVEF and peakVO2. Combination of depression score,

LVEF and peakVO2 allowed for a better risk stratification than combination of LVEF and peakVO2 alone. The risk ratio for mortality in

patients with an elevated depression score (i.e. above the median) rose over time to 8.2 after 30 months (CI 2.62–25.84).

Conclusions: The depression score predicts mortality independent of somatic parameters in CHF patients not treated for depression. Its

prognostic power increases over time and should, thus, be accounted for in risk stratification and therapy.

D 2004 European Society of Cardiology. Published by Elsevier B.V. All rights reserved.

Keywords: Congestive heart failure; Depression; Prognosis; Mortality; Risk stratification

1. Introduction

Despite advances in the treatment of congestive heart

failure (CHF), this epidemic continues unabated worldwide

[1,2]. The high morbidity and mortality as well as the

marked decrease in quality of life [3] associated with CHF

1388-9842/$ - see front matter D 2004 European Society of Cardiology. Publishe

doi:10.1016/j.ejheart.2004.05.011

Abbreviations: CHF, Congestive heart failure; LVEF, Left ventricular

ejection fraction; NYHA, New York Heart Association; PeakVO2, peak

oxygen uptake in cardiopulmonary exercise testing; HADS-D, Hospital

Anxiety and Depression Scale in German.

* Corresponding author. Tel.: +49 6221 568657; fax: +49 6221 565749.

E-mail address: [email protected] (J. Jqnger).

points to the need for increased attention towards additional

risk factors for the course of CHF. While depression has

been well-established as an independent risk factor for

development of coronary artery disease [4] and as an

independent predictor for its prognosis [5–7], only recently

have the unfavorable effects that depression exerts on CHF

been reported. Until the studies of Abramson et al. [8] and

Williams et al. [9], it was not well understood that

depression is a risk factor for predisposition to CHF. The

few studies on the prognostic value of depression in CHF

patients have shown inconsistent results. Two studies with

an observation time of 6–12 months found no predictive

value of depression after multivariate controlling for somatic

Failure 7 (2005) 261–267

d by Elsevier B.V. All rights reserved.

J. Junger et al. / The European Journal of Heart Failure 7 (2005) 261–267262

parameters [10,11], while another showed a predictive value

of depression after 24 months [12]. However, none of these

studies controlled for the established predictors, peakVO2

and left ventricular ejection fraction (LVEF), simultane-

ously. For the latter two a highly predictive value concern-

ing mortality was shown recently in a simple two-variable

model [13].

The first aim of the present study was to test the

hypothesis that depression independently predicts mortality

in patients with CHF, even after adjusting for established

clinical risk variables and concomitant beta-blocking

medication and to show that entering information on

depression score into a model comprising peakVO2 and

LVEF would increase the overall risk stratification. The

second aim was to measure the interaction of the predictive

effect of depression with time.

at unihd on April 28, 2011

eurjhf.oxfordjournals.orgD

ownloaded from

2. Methods

Two hundred and nine patients with stable CHF (New

York Heart Association (NYHA) functional classes I–III)

were prospectively enrolled into the study through the

Department of Cardiology at the University of Heidelberg

between March 1996 and March 1999, after giving written

informed consent. The study complied with the Declaration

of Helsinki and was approved by the Ethics Committee for

human research of the University of Heidelberg.

The cardiac diagnosis was based on left heart catheter-

ization and coronary angiograms prior to enrollment. The

major inclusion criterion was LVEF V45%, as determined

by radionuclide ventriculography [13,14]. Patients in

NYHA functional class IV were excluded, as were those

who had neurological, orthopedic, peripheral vascular or

severe pulmonary diseases, which may have impaired

successful completion of exercise testing (i.e. determination

of peakVO2 by cardiopulmonary exercise testing) [13].

Patients who were not able to speak German fluently were

also excluded.

NYHA functional class was determined by an independ-

ent investigator prior to assessment of somatic variables.

Furthermore, the investigator who applied the standard

Hospital Anxiety and Depression Scale in German (HADS-

D) questionnaires was blinded to the aforementioned data.

All parameters were collected within a period of 48 h.

3. Depression and anxiety

Patients’ self assessment of depression and anxiety were

determined at baseline by the German version of the

Hospital Anxiety and Depression Scale, the HADS-D [15,

16], a validated and reliable 14 item checklist (7 items for

depression and 7 items for anxiety) well-established among

cardiac patients [15]. It has one scale-score for depression

and one for anxiety. Patients with z8 points on the

depression scale are suspected of having depression and

with z10 points of having anxiety [15,16]. As there is no

validated prognostic cutoff point, the median of the

HADS-D depression score was used for survival-analysis.

For analysis of risk stratification, the continuous variable

was used.

4. Predictor and outcome variable

The primary independent, or predictor, variable of

interest was the depression score from the baseline assess-

ment. The primary endpoint was all-cause mortality, defined

as death due to any cause. The dependent variable was time

from baseline to the endpoint (death) or time until trans-

plantation or termination of the study. In this context the

status of censored has to be defined. All patients undergoing

cardiac transplantation were considered as survivors until

the date of the transplantation.

5. Endpoint monitoring

We obtained endpoint information during regular out-

patient visits or by telephone calls to the patients’ home or

their family physician. Information on endpoint status was

available for all patients at the pre-established time point

(i.e. July 2001). Depression was only assessed at baseline.

Mean observation time was 24.8 months (range: minimum

0.6 to maximum 36 months). The predefined endpoint was

all-cause mortality. Death without transplantation was

defined as an outcome event. All patients undergoing

cardiac transplantation were considered as survivors until

the date of their transplantation regardless of the post-

operative outcome.

6. Statistical analysis

Statistical analysis was performed with standard soft-

ware (SAS version 6.09). Spearman rank correlation

coefficient was used as a measure of association between

variables. A nonparametric two-sample Wilcoxon test was

used to test for differences between groups (survivors vs.

non-survivors). Survival curves were calculated using the

Kaplan–Meier method [17]. Multivariate Cox proportional

hazards analysis was used to identify the most important

predictors of mortality [18]. The log of the negative log

survival curves stratified according to the HADS-D

depression median was used to evaluate the basic propor-

tional hazards assumption of the Cox-model for depression.

This assumption was violated, as the curves did not run

parallel (not shown). Standardized Schoenfeld-smoothed

residuals were analysed to demonstrate the functional form

of interaction between depression and its effect on mortality

over time, as proposed by Grambsch and Therneau [19].

Table 1

Baseline clinical, functional and psychosocial characteristics of the patient

sample

Parameter Value Survivors Non-

survivors

P

Number of

patients (n)

209 164 45

Age (years) 54F10 53F10 55F11 n.s.*

Gender (male %) 86.1 84.8 91.1 n.s.*

Diagnosis (%) y

Coronary artery

disease

26.3 24.4 33.3

Dilated

cardiomyopathy

68.9 70.1 64.4

Other 4.8 5.5 2.3

NYHA functional-

class (%)

z

Class I 11.5 12.8 6.7

Class II 44.5 46.3 37.8

Class III 44.0 40.9 55.5

Peak oxygen uptake

(ml/kg/min)

14.9F5.2 15.4F5.2 13.1F4.5 b0.01

LVEF (%) 22F10 23F10 18F8 b0.01

Medication (%)

ACE-Inhibitor 93 96 93 0.74

Diuretics 84 93 81 0.07

Digitalis 73 82 70 0.13

Beta-blocker 36 27 39 0.16

HADS

Depression total

score

6.4F4.3 6.1F4.3 7.5F4.0 0.02

Depression

scorez8 (%)

30.1 27.4 40.0

Depression score

in NYHA

functional-class

(scorez8 (%))

Class I 2.6F2.5 (4.0)

Class II 5.7F3.7 (23.7)

Class III 8.0F4.4 (43.5)

Anxiety total score 7.0F4.0 6.7F4.1 7.8F3.6 n.s.*

Patients with

anxiety score

z10 (%)

21.5 19.5 28.9

MeansFS.D. or percentage of total patient sample.

* n.s.=not significant.y Distribution between different etiologies not significant.z Distribution between NYHA functional classes not significant.

J. Junger et al. / The European Journal of Heart Failure 7 (2005) 261–267 263

at unihd on April 28, 2011

eurjhf.oxfordjournals.orgD

ownloaded from

Receiver-operating characteristic curves were constructed

by means of plotting true-positive rates (sensitivity) against

false-positive rates (1-specificity) [20,21]. The data are

expressed as meanFstandard deviation (S.D.). A P-value

b0.05 was considered significant.

Table 2

Univariate and multivariate Cox regression analysis of study variables and surviv

Univariate

Chi2 P HR (95% CI)

LVEF 13.70 0.0002 0.93 (0.90–0.9

PeakVO2 9.97 0.0016 0.90 (0.84–0.9

Depression 7.25 0.0071 1.09 (1.02–1.1

7. Results

7.1. Clinical correlation

All results are from Spearman rank correlations. Dem-

ographic data and functional parameters of all CHF patients

at baseline are summarized in Table 1. The majority of the

patients (88.5%) were in NYHA functional-classes II or III.

The advanced stage of disease was also reflected by the fact

that 46% of the patients had a peakVO2 V14 ml/kg/min and

52% had a LVEF of V 20%. Standard medical treatment is

listed in Table 1. No patient received antidepressant

medication.

HADS-D depression score and anxiety score increased,

while LVEF and peakVO2 decreased with NYHA functional

class. The total CHF sample was characterized by signifi-

cantly increased HADS-D depression score and anxiety

scores compared to a healthy reference group [22] (data not

shown).

The HADS-D depression score was just correlated with

LVEF (r=�0.134; P=0.05) and was only weakly correlated

with peakVO2 (r=�0.28; Pb0.01). However, a closer

relationship was observed between depression score and

NYHA functional class (r=0.40; Pb0.01) and anxiety

(r=0.68; Pb0.01).

7.2. Survival analysis

Results are from the nonparametric Wilcoxon two-sample

test. During a mean follow-up of 24.8 months 45 patients

died, all due to cardiac causes. Compared to survivors, non-

survivors were characterized by a higher NYHA functional

class (2.8F0.7 vs. 2.5F0.6) and a lower LVEF (18F8 vs.

23F10%) and peakVO2 (13.1F4.5 vs. 15.4F5.2 ml/kg/min)

(all Pb0.05). Depression score was significantly higher

among non-survivors (7.5F4.0 vs. 6.1F4.3, Pb0.05).

Anxiety score among these patients was also marginally

higher but the difference was not statistically significant

(7.8F3.6 vs. 6.7F4.1, P=0.103). Depression score and

anxiety score increased concomitantly with rising NYHA

functional class (Table 1).

7.3. Univariate and multivariate analysis

In univariate analysis depression score emerged—next to

NYHA functional class (data not shown), LVEF and

peakVO2—as a significant (Pb0.01) prognostic parameter

al

Multivariate

Chi2 P HR (95% CI)

7) 10.17 0.0014 0.94 (0.90–0.98)

6) 4.66 0.03 0.93 (0.87–0.99)

7) 5.23 0.02 1.08 (1.01–1.15)

J. Junger et al. / The European Journal of Heart Failure 7 (2005) 261–267264

(Table 2). Kaplan–Meier survival curves of depression,

LVEF and peakVO2 divided according to the cutoff-values

(median) are shown in Fig. 1a, b and c. Multivariate Cox

proportional hazard analysis revealed that the depression

score predicted prognosis independent of LVEF and

peakVO2 (Table 2). The combination of depression score,

peakVO2 and LVEF allowed a better risk stratification than

the combination of the established parameters LVEF and

peakVO2 alone (Chi-square 24.93 vs. 20.24, Pb0.001).

Fig. 1. Kaplan–Meier survival curves of the total study population divided

according to cutoff-values (median). (a) HADS-D depression score, cutoff 6

points; Log rank chi-square 11.20, P=0.0008. (b) LVEF, cutoff 20%; Log-

rank chi-square 8.08, P=0.0045. (c) PeakVO2, cutoff 14 ml/min/kg; Log

rank chi-square 6.90, P=0.0086.

Fig. 2. Receiver-operating characteristic curves of prediction of survival at

36 months. The combination of HADS-D depression score, LVEF and

peakVO2 (AUC, 0.831) yields significantly more prognostic information

than combination of LVEF and peakVO2 alone (AUC, 0.774). n=45

patients died within 36 months, n=164 survived; area-test one-tailed

Pb0.0028.

at unihd on April 28, 2011

eurjhf.oxfordjournals.orgD

ownloaded from

Controlling for medication (ACE-inhibitors, diuretics,

digitalis and beta-blockers) and etiolgy (ischaemic vs.

dilated cardiomyopathy) in a multivariable Cox regression

model, the effects of LVEF, peakVO2 and HADS depres-

sion score remained significant and unaltered in size. None

of the former variables controlled for turned out as

significant.

7.4. Receiver-operating characteristic curves for survival at

36 months

After maximum follow-up of 36 months, parameters

identified as independent predictors by multivariate Cox

regression analysis were entered into receiver operating

characteristic analysis: By adding depression score to the

combination of LVEF and peakVO2, risk prediction

improved significantly (Fig. 2).

7.5. Analysis of the time-dependent mortality-risk of

depression

As shown in Table 3, mortality-risk associated with

depression score rises over time. When calculating the

hazard ratios over time, it becomes clear that depression

exerts almost no risk in the first year of follow-up and rises

to an 8-fold level after 30 months (Table 3).

Table 3

Time-dependent risk ratios for mortality for patients with a HADS-D

depression score N6

Survival time in months Chi2 P HR (95% CI)

6 0.01 0.950 1.03 (0.44–2.39)

12 2.71 0.100 1.73 (0.90–3.32)

18 10.18 0.001 2.91 (1.51–5.60)

24 13.34 b0.001 4.90 (2.09–11.45)

24.8 (mean observation time) 13.40 b0.001 5.24 (2.16–12.72)

30 13.01 b0.001 8.22 (2.62–25.84)

J. Junger et al. / The European Journal of Heart Failure 7 (2005) 261–267 265

at unihd on April 28, 2011

eurjhf.oxfordjournals.orgD

ownloaded from

8. Discussion

In this study, depression was an important independent

risk factor for mortality of CHF patients. In multivariate

analysis, depression score predicted prognosis independ-

ent of NYHA functional class, LVEF and peakVO2, as

well as etiology and concomitant medication. The

combination of depression score, LVEF and peakVO2

allowed for a better risk stratification than the combina-

tion of the established parameters LVEF and peakVO2

alone.

In previous studies the investigators focused primarily on

the impact of depression on prognosis without controlling

simultaneously for the established prognostic parameters

LVEF and peakVO2 [10–12]. The studies showed a

predictive value of depression for combined endpoints such

as mortality and re-hospitalization [10] or death and

functional decline [11]. However, after controlling for

multiple variables (such as NYHA functional class, LVEF,

age) and focusing on the single endpoint mortality the

results lost their statistic significance [10,11]. In the more

recent studies of Jiang et al. [10] and Vaccarino et al. [11]

the observation time of CHF patients was restricted to 1

year. Also, Koenig [23] found no prognostic value of

depression after an observation time of 1 year. Faris et al.

[24] showed reduced survival of patients with non-

ischaemic heart failure and depression over a long obser-

vation period of 5 years. However, peakVO2 was available

only in 37% of the study population and the diagnosis of

depression was based on a clinical diagnosis through review

of patients’ medical records, not on an established instru-

ment or questionnaire.

In the present study, patients were followed over a mean

of 24.8 months. The main effect of depression on mortality

began at the end of the first year of observation. This time-

dependent development of the prognostic effect of depres-

sion in patients with CHF was shown here for the first time.

In a previous study we were able to show that patients with

an endstage heart failure and an additional increased

preoperative depression score showed an unfavorable out-

come after heart transplantation, particularly in the long-

term course [25]. Denollet and Brutsaert [26] state an odds

ratio for cardiac death of 7.5 after myocardial infarction

with a reduced LVEF in patients with negative affectivity,

which is part of depressive pathology, after the long

observation period of 6 to 10 years. So, both studies show

depression as a predictor of mortality in the long run.

Abramson et al. [8] found that depression independently

predicts the onset of CHF in patients with isolated systolic

hypertension over an observation period of 4.5 years.

Finally, a study by Fredman et al. [27] showed an

association of depression with mortality in elderly women

that was much stronger after 6 years than after just 2 years.

Thus the effect of depression on mortality seems to develop

over time and might only be recognized in long-term

observation.

There are several mechanisms that could account for the

association between depression and increased heart failure

mortality. In patients with depression, hypercortisolism and

an activation of the sympathetic nervous system can be

observed [28], as well as elevated levels of cytokines [29],

which in turn are predictive factors for CHF mortality

[30,31]. Also, patients with depression show reduced heart

rate variability [32]. Higher platelet activity and consec-

utively an increased risk for cardiovascular events has also

been shown in some studies [4,33]. However, the present

study found no significant difference in the prognostic effect

of depression regarding ischemic or non-ischemic etiology

of heart failure. Likewise, Abramson et al. [8] found that

depression retained its predictive power for the risk of

developing CHF even after controlling for a history of

myocardial infarction.

In addition to these factors, psychological explanations

linking depression and increased mortality in heart failure

are being discussed. Patients with depression are known to

have poor social contacts and support networks [25,34].

Additionally, they show a reduced compliance with treat-

ment because of an appreciable degree of hopelessness,

withdrawal from the social network, with a subsequent loss

of emotional support and possible reductions in the

cognitive functioning essential for following treatment

recommendations [35]. All the above mentioned factors

might increase the risk of non-compliance with medication

regimens and thus increase the risk of cardiac decompensa-

tion [35,36].

8.1. Limitations of the study

This study was performed at a tertiary referral center.

Thus the current CHF sample does not represent the typical

CHF population seen by a general practitioner. The

majority of CHF-patients in a general practitioner’s

practice are characterized by a smaller impairment in

LVEF, a higher prevalence of coronary artery disease, and

an older age [37]. Also, there was a very high proportion

of male patients in this sample. The clinical diagnosis of

depressive disorder cannot be made with the HADS-D.

However, the HADS-D has been shown to be a reliable

and valid screening instrument for depressive symptoms in

general internal patients [15]. Additionally, it seemed

important to select a simple, clinically applicable instru-

ment that could be completed by the patients themselves

and that could easily be used for screening and follow-up

in other centers.

Since the analysis of the prognostic impact of the HADS-

D was restricted to a single timepoint (baseline), it cannot be

ruled out that the depression score further increased

especially in those patients with a depression score above

the median at baseline. Furthermore, as the patients were not

being treated for depression, the present study allows no

conclusion on the prognostic impact of specifically coping

with depression.

J. Junger et al. / The European Journal of Heart Failure 7 (2005) 261–267266

9. Conclusions

The present findings are of major clinical importance.

Screening for depression is a simple, well-established, non-

invasive method that a patient can self-administer. The

HADS-D offers the physician not only information about

the psychological status of a patient with CHF, but also

prognostic information that is independent of somatic

parameters.

While at the physical level excellent pharmacological

treatments are available to improve prognosis, a diagnosis of

depression is also a state of impaired health that needs to be

treated by pharmacological and/or psychotherapeutic

means. As the negative effect of depression on prognosis

seems to evolve slowly, there may be sufficient time to

initiate this treatment. If these findings are confirmed by

future studies, the influence on risk stratification for these

patients should be investigated.

eurjhD

ownloaded from

Acknowledgements

This study was supported by grants from the faculty for

clinical medicine of the University of Heidelberg (projects

32/95 and 158/97).

at unihd on April 28, 2011

f.oxfordjournals.org

References

[1] Ho KK, Pinsky JL, Kannel WB, Levy D. The epidemiology of heart

failure: the Framingham Study. J Am Coll Cardiol 1993;22:6A–13A.

[2] O’Connell JB. The economic burden of heart failure. Clin Cardiol

2000;23(III):6–10.

[3] Juenger J, Schellberg D, Kraemer D, et al. Health related quality of life

in patients with congestive heart failure: comparison with other chronic

diseases and relation to functional variables. Heart 2002;87:235–41.

[4] Rozanski A, Blumenthal JA, Kaplan J. Impact of psychological

factors on the pathogenesis of cardiovascular disease and implications

for therapy. Circulation 1999;99:2192–217.

[5] Ariyo AA, Haan M, Tangen CM, et al. Depressive symptoms and

risks of coronary heart disease and mortality in elderly Americans

Cardiovascular Health Study Collaborative Research Group. Circu-

lation 2000;102:1773–9.

[6] Barefoot JC, Helms MJ, Mark DB, et al. Depression and long-term

mortality risk in patients with coronary artery disease. Am J Cardiol

1996;78:613–7.

[7] Frasure-Smith N, Lesperance F. Coronary artery disease, depression

and social support only the beginning. Eur Heart J 2000;21:1043–5.

[8] Abramson J, Berger AC, Krumholz HM, Vaccarino V. Depression and

risk of heart failure among older persons with isolated systolic

hypertension. Arch Intern Med 2001;161:1725–30.

[9] Williams SA, Kasl SV, Heiat A, Abramson JL, Krumholz HM,

Vaccarino V. Depression and risk of heart failure among the elderly: a

prospective community-based study. Psychosom Med 2002;64:6–12.

[10] Jiang W, Alexander J, Christopher E, et al. Relationship of depression

to increased risk of mortality and rehospitalization in patients with

congestive heart failure. Arch Intern Med 2001;161:1849–56.

[11] Vaccarino V, Kasl SV, Abramson J, Krumholz HM. Depressive

symptoms and risk of functional decline and death in patients with

heart failure. J Am Coll Cardiol 2001;38:199–205.

[12] Murberg TA, Bru E, Svebak S, Tveteras R, Aarsland T. Depressed

mood and subjective health symptoms as predictors of mortality in

patients with congestive heart failure: a two-year follow-up study. Int J

Psychiatry Med 1999;29:311–26.

[13] Zugck C, Krqger C, Kell R, et al. Risk stratification in middle-aged

patients with congestive heart failure: prospective comparison of the

Heart Failure Survival Score (HFSS) and a simplified two-variable

model. Eur J Heart Fail 2001;3:577–85.

[14] Meyer FJ, Borst MM, Zugck C, et al. Respiratory muscle dysfunction

in congestive heart failure: clinical correlation and prognostic

significance. Circulation 2001;103:2153–8.

[15] Herrmann C. International experiences with the Hospital Anxiety and

Depression Scale—a review of validation data and clinical results.

J Psychosom Res 1997;42:17–41.

[16] Bjelland I, Dahl AA, Haug TT, Neckelmann D. The validity of the

Hospital Anxiety and Depression Scale. An updated literature review.

J Psychosom Res 2002;52:69–77.

[17] Kaplan EL, Meier P. Nonparametric estimation from incomplete

observations. J Am Stat Assoc 1958;53:457–81.

[18] Cox DR. Regression Models and Life-Tables. JR Stat Soc 1972;Series

B 34:187–202.

[19] Grambsch P, Therneau T. Proportional hazards tests and diagnostics

based on weighted residuals. Biometrika 1994;81(3):515–26.

[20] Metz CE, Wang P, Kronman H. A new approach for testing the

significance of differences between ROC curves measured from

correlated data. In: Deconinck F, editor. Information processing in

medical imaging. The Hague7 Martinus Nijhoff; 1984. p. 432–45.

[21] Metz CE, Herman BA, Roe CA. Statistical comparison of two ROC-

curve estimates obtained from partially-paired datasets. Med Decis

Making 1998;18:110–21.

[22] Hinz A, Schwarz R. Anxiety and depression in the general population:

normal values in the Hospital Anxiety and Depression Scale.

Psychother Psychosom Med Psychol 2001;51:193–200.

[23] Koenig HG. Depression in hospitalized older patients with congestive

heart failure. Gen Hosp Psychiatry 1998;20:29–43.

[24] Faris R, Purcell H, Henein MY, Coats AJ. Clinical depression is

common and significantly associated with reduced survival in

patients with non-ischaemic heart failure. Eur J Heart Fail 2002;4:

541–51.

[25] Zipfel S, Schneider A, Wild B, et al. Effect of depressive symptoms

on survival after heart transplantation. Psychosom Med 2002;64(5):

740–7.

[26] Denollet J, Brutsaert DL. Personality, disease severity, and the

risk of long-term cardiac events in patients with a decreased

ejection fraction after myocardial infarction. Circulation 1998;97:

167–73.

[27] Fredman L, Magaziner J, Hebel JR, Hawkes W, Zimmerman SI.

Depressive symptoms and 6-year mortality among elderly commun-

ity-dwelling women. Epidemiology 1999;10:54–9.

[28] Musselman DL, Evans DL, Nemeroff CB. The relationship of

depression to cardiovascular disease: epidemiology, biology, and

treatment. Arch Gen Psychiatry 1998;55:580–92.

[29] Gold PW, Chrousos GP. The endocrinology of melancholic and

atypical depression: relation to neurocircuitry and somatic consequen-

ces. Proc Assoc Am Physicians 1999;111:22–34.

[30] Cohn JN, Johnson GR, Shabetai R, et al. Ejection fraction, peak

exercise oxygen consumption, cardiothoracic ratio, ventricular

arrhythmias, and plasma norepinephrine as determinants of prognosis

in heart failure. The V-HeFT VA Cooperative Studies Group.

Circulation 1993;87:VI5–16.

[31] Kell R, Haunstetter A, Dengler TJ, Zugck C, Kqbler W, Haass M. Do

cytokines enable risk stratification to be improved in NYHA functional

class III patients? Comparison with other potential predictors of prog-

nosis. Eur Heart J 2002;23:70–8.

[32] Gorman JM, Sloan RP. Heart rate variability in depressive and anxiety

disorders. Am Heart J 2000;140:S77–83.

J. Junger et al. / The European Journal of Heart Failure 7 (2005) 261–267 267

[33] Musselman DL, Tomer A, Manatunga AK, et al. Exaggerated platelet

reactivity in major depression. Am J Psychiatry 1996;153:1313–7.

[34] Dracup K, Walden JA, Stevenson LW, Brecht ML. Quality of life in

patients with advanced heart failure. J Heart Lung Transplant 1992;

11:273–9.

[35] DiMatteo MR, Lepper HS, Croghan TW. Depression is a risk factor

for noncompliance with medical treatment: meta-analysis of the

effects of anxiety and depression on patient adherence. Arch Intern

Med 2000;160:2101–7.

[36] Michalsen A, Konig G, Thimme W. Preventable causative factors

leading to hospital admission with decompensated heart failure. Heart

1998;80:437–41.

[37] Cowie MR, Mosterd A, Wood DA, et al. The epidemiology of heart

failure. Eur Heart J 1997;18:208–25.

at unihd on April 28, 2011

eurjhf.oxfordjournals.orgD

ownloaded from