- Email: [email protected]
Reasons why patients suffering from chronic heart failure at very low risk for mortality die
International Journal of Cardiology 223 (2016) 947–952
Contents lists available at ScienceDirect
International Journal of Cardiology journal homepage: www.elsevier.com/locate/ijcard
Reasons why patients suffering from chronic heart failure at very low risk for mortality die Giulia Russo a,⁎, Giovanni Cioffi b, Giovanni Pulignano c, Giulia Barbati a, Luigi Tarantini d, Donatella Del Sindaco e, Carmine Mazzone a, Antonella Cherubini a, Giorgio Faganello a, Carlo Stefenelli b, Michele Senni f, Andrea Di Lenarda a a
Cardiovascular Center, ASUITS, Italy Cardiology Department, Villa Bianca Hospital Trento, Italy Heart Failure Clinic, Division of Cardiology/C.C.U. San Camillo Hospital, Rome, Italy d Cardiology Department, St. Martino Hospital. Azienda Sanitaria Locale n. 1, Belluno, Italy e Department of Cardiocirculatory Diseases, San Giovanni-Addolorata Hospital, Rome, Italy f Cardiovascular Department, Ospedali Riuniti, Bergamo, Italy b c
a r t i c l e
i n f o
Article history: Received 14 May 2016 Received in revised form 19 August 2016 Accepted 20 August 2016 Available online 23 August 2016 Keywords: Chronic heart failure Mortality Renal dysfunction Prognosis Older age Male gender
a b s t r a c t Background: A proper prognostic stratification is crucial for organizing an effective clinical management and treatment decision-making in patients with chronic heart failure (CHF). In this study, we selected and characterized a sub-group of CHF patients at very low risk for death aiming to assess predictors of death in subjects with an expected probability of 1-year mortality near to 5%. Methods: We used the Cardiac and Comorbid Conditions HF (3C-HF) Score to identify CHF patients with the best mid-term prognosis. We selected patients belonging to the lowest quartile of 3C-HF score (≤9 points). Results: We recruited 1777 consecutive CHF patients at 3 Italian Cardiology Units (age 76 ± 10 years, 43% female, 32% with preserved ejection fraction). Subjects belonging to the lowest quartile of 3C-HF score were 609. During a median follow-up of 21 [12–40] months, 48 of these patients (8%) died, and 561 (92%) survived. The variables that contributed to death prediction by Cox regression multivariate analysis were older age (HR 1.03[CI 1.00– 1.07]; p = 0.04), male gender (HR 2.93[CI 1.50–5.51]; p = 0.002) and a higher degree of renal dysfunction (HR 0.96[CI 0.94–0.98]; p b 0.001). Conclusions: The prognostic stratification of CHF patients by 3C-HF score allows one to select patients at different outcome and to identify the factors associated with death in outliers with a very low mortality risk at mid-term follow-up. The reasons why these patients do not outlive the matching part of subjects who expectedly survive are related to a declined renal function and unmodifiable conditions including older age and male gender. © 2016 Elsevier Ireland Ltd. All rights reserved.
1. Introduction An accurate prognostic stratification is essential for optimizing the clinical management and treatment decision-making of patients with chronic heart failure (HF). In daily clinical practice the assessment of risk for adverse outcomes is often limited and based on individual clinician's capabilities. The combined use of specific and sensible variables tested in prognostic models has appeared as the most appropriate methodology to decipher the natural history of HF syndrome in the single individual [1–7]. By this approach, however, it is usually possible to identify the macro-phenomena leading to the adverse clinical events, while it is “by definition” not possible to characterize the ⁎ Corresponding author at: Cardiovascular Center, ASUITS, Via Slataper 9, 34125 Trieste, Italy. E-mail address: [email protected] (G. Russo).
http://dx.doi.org/10.1016/j.ijcard.2016.08.326 0167-5273/© 2016 Elsevier Ireland Ltd. All rights reserved.
patients defined as “outliers”, who have atypical behaviors and outcomes far from those expected. No data are available on the protecting conditions from adverse clinical events in patients with chronic HF and low risk of death. Accordingly, in this study, among the best available models, we used the Cardiac and Comorbid Conditions HF (3C-HF) Score [8], to predict all-cause mortality in patients with chronic HF. Thus, according to the score, we selected the subgroup of patients with chronic HF and the lowest risk for all-cause mortality with the aim to characterize and assess the predictors of death in those subjects who unexpectedly died during the follow-up. 2. Material and methods Patients for this study were enrolled by 3 Italian HF clinical units into a prospective evaluation and their data collected in the same database. We selected adult subjects in stable clinical conditions [New York Heart Association (NYHA) functional class II or III] who
948
G. Russo et al. / International Journal of Cardiology 223 (2016) 947–952
had previously experienced at least an episode of HF requiring a hospital admission [9]. At enrollment, we considered clinical, laboratory, and echocardiographic data within the last 6 months prior to enrollment. Practical guidelines were used to define severity of valve heart disease [10] and optimization of medical treatment [9]. Left ventricular ejection fraction (LVEF) was measured by echocardiography. Patients with HF symptoms and a LVEF ≥50% (defined as “patients with normal LVEF”) had to show lung congestion (clinical or by chest X-ray), abnormal BNP values or high LV filling pressure at echocardiographic evaluation. History of hypertension was defined a condition preceding the first episode of congestive HF as systolic blood pressure of ≥140 mm Hg and/or a diastolic blood pressure of ≥ 90 mm Hg and/or pharmacologically treated high blood pressure of unknown etiology. Obesity was diagnosed if patients had body mass index ≥30 kg/m2. Cut-off point for defining the condition of “underweight” was 18.5 Kg/m2 [11]. To assess renal function we considered the glomerular filtration rate (GFR) estimated with the CKD-EPI equation. Patients were followed-up at each center for a period lasting 1 year at least. Survival status was ascertained locally by follow-up visits or chart review, telephone interview with the patient, or his/her family, or primary care physician, or by examination of death certificates. Cause of death was physician-reported. Urgent heart transplantation, defined as UNOS status one [12], was counted as a death. Patients who underwent elective heart transplantation were censored seven days after the procedure. Patient follow-up was 100% complete. Patients expressed their general written consent to the anonymous use of data for their care and research purposes. Databases for clinical use were authorized at each center. The study complies with the Declaration of Helsinki; the locally appointed ethics committee approved the cardiovascular registries. For this investigation we initially categorized all the participants according to the 3CHF score classification [8], which considered the following 11 variables: age, NYHA functional class, history of hypertension, atrial fibrillation, diabetes with target organ damage, anemia, chronic kidney dysfunction, LVEF, severe heart valve disease, current treatment with beta-blocker, and current treatment with RAS-inhibitor. Thus, we selected the subgroup of patients belonging to the lowest quartile of the score (≤9 points) having a probability of 1-year mortality which approximated to 5% [8]. Patients with an ascertained diagnosis of hypertrophic or restrictive cardiomyopathy, those with acute myocarditis or with alcohol-induced myocardial injury were excluded from the study. Those subjects who experienced a myocardial infarction and/or underwent a percutaneous transcatheter coronary angioplasty or coronary artery bypass graft or any intervention for valvular disease within 3 months before our first evaluation up were also rejected. Implanted cardioverter defibrillator and/or cardiac resynchronization therapy were not considered contraindications for enrollment.
2.1. Assessment and follow-up A careful history was obtained from all patients: this included an evaluation of NYHA functional class and a complete physical and echocardiographic examination, routine blood tests and standard electrocardiography. During the follow-up, patients received diuretics (flexible regimen) and/or low dose of spironolactone if indicated, and highest tolerated dose of angiotensin-converting enzyme inhibitors (ACE-I), which were replaced on angiotensin type 1 receptor blockers (ARBs) if not tolerated. Carvedilol, metoprolol, bisoprolol or nebivolol were also given at the highest tolerated dose.
2.2. Statistical analysis Categorical variables are presented as percentages, while continuous variables are presented as their means and SD. Categorical variables were compared by the chisquare test and continuous variables by the t-test or the Mann–Whitney U-test. Log cumulative hazard functions were computed by univariable and multivariable Cox proportional hazards analyses (SPSS version 19.0, SPSS Inc. Chicago, Illinois, USA) to identify the predictors of primary end-point (all-cause death). Variables significantly related to the primary end-point in univariate tests (p b 0.05) were included in the multivariable model. The final model was internally validated by means of bootstrap technique (number of samples = 1000, level of confidence interval = 95%). Two sub-analyses considering only patients with reduced LVEF (b50%) and only patients with normal LVEF (≥50%), respectively, were made to assess whether the independent prognostic markers in the low risk population were similar in the two sub-groups with different LVEF phenotype. In order to simplify the clinical usage of the estimated model, Receiver Operating Characteristic (ROC) curve analyses were performed to identify the ‘best’ cut-off points (i.e. values that had the highest sum of sensibility and specificity) of the continuous variables independently associated to the primary end-point which were identified by Cox regression analysis. A score was then defined by summing the significant covariates in the model in their ‘cut-off’ version. Time-dependent ROC curves for censored data (TD-ROC) were estimated, plotted and compared to assess the accuracy of this score and also the accuracy of the single covariates considered separately. TD-ROC curves are calculated taking into account the Inverse Probability of Censoring Weighting (IPCW) estimation of a Cumulative/Dynamic ROC curve: this curve intrinsically depends on the definitions of time-dependent “cases” and “controls”. In our case, we used as time-horizon for the events 24 months of follow up, so defining “case” a subject with an event observed before 24 months and as “control” a subject that survives beyond 24 months. We used for computations the library “timeROC” of the R statistical package, version 3.1.2. A 2-tailed value of p b 0.05 was used to reject the null hypothesis of no difference.
3. Results 3.1. Study population During the enrollment period, 1912 patients with chronic HF were included in the database and 1777 (93%) were eligible for this study; among these 1777 subjects, 609 (34.2%) were selected according to the lowest quartile of 3C-HF score (≤ 9 points) and formed the final study population (Fig. 1). Baseline demographics and clinical characteristics of the 609 patients with the lowest risk of death are reported and compared with those of 1168 patients at higher risk in Table 1. As expected, the former were younger, less affected by diabetes mellitus, atrial fibrillation, anemia or severe chronic renal disease, they were more frequently treated with beta-blockers and ACE-I/ARBs than the latter. Considering the total population of 1777 patients, the probability of all-cause mortality at 1-year follow-up estimated by the 3C-HF logistic score was 15%. During a median follow-up of 21 [12–40] months 398 patients died (22%). All-cause mortality rate was 7.9% in the 609 patients who had the lowest risk of death belonging to the lowest quartile of 3C-HF score, and 30.4% in the 1168 patients at higher risk belonging to the other quartiles (p b 0.0001). The incidence of re-hospitalization (occurred during the same period of observation) was 10.1% in the former and 35% in the latter.
3.2. Picture of the outliers During the follow up, among the 609 patients with the lowest risk of death, 48 (7.9%) died. These patients were older, were more frequently men, had a worse renal function, and a higher prevalence of anemia than the 561 counterparts who survived (92.1%). There was no difference in LVEF and pharmacological therapy between the two groups (Table 2). Mode of death was cardiovascular cause in 63% of cases (progressive pump failure = 27%; sudden death = 53%; acute myocardial infarction 20%), and non-cardiovascular cause in 27% of cases (respiratory disease = 26%; renal failure = 23%; cancer = 33%; other causes = 20%). Cause of death was unknown in the remaining 10% of patients.
3.3. Predictors of death Multivariable Cox regression was performed to characterize the patients who died in spite of the very low risk of death predicted by the prognostic 3C-HF score. Among the covariates age, gender, renal function (measured as GFR) and hemoglobin (b vs N 11 mg/dl), selected by univariate analysis, at multivariate modeling age, gender and GFR were confirmed as independently associated with the study end-point (Table 3). This model was internally validated by a bootstrap technique which confirmed its solidity and constancy: GFR, age and gender appeared as the three strongest prognosticators of mortality (all p = 0.001). The ‘best’ cut-off points of the continuous variables were identified by ROC curve analyses: they were 63 ml/min/1.73 m2 for GFR (AUC 0.60 [CI 0.50–0.72], sensitivity 57% and specificity 63%) and 75 years of age (AUC 0.65 [0.57–0.73], sensitivity 74% and specificity 61%) (Table 4). Fig. 2 shows the results of TD-ROC curve analyses demonstrating a better accuracy of the estimated score (AUC 0.72 [0.62–0.82]) than the 3 prognosticators of death considered separately. Fig. 3 reports the event rates in the 609 study patients stratified by the number of risk factors for death: it ranged from 3% in patients who had none or only 1 risk factor to 20% in those who had the 3 risk factors (male gender, age N 75 years and GFR b 63 ml/min/1.73 m2 coexisted). Based on these results, four sub-groups of patients were identified according to the number of risk factors for death (from zero to 3). Fig. 4 shows Kaplan–Meier all-cause mortality curves of these four sub-groups of patients (difference by Log Rank Test: p b 0.0001).
G. Russo et al. / International Journal of Cardiology 223 (2016) 947–952
949
Fig. 1. Flow chart of patient's selection showing the number of subjects included and excluded according to the criteria for enrolment.
with adverse outcome (HR 4.06 [1.82–9.02], p b 0.001), while no statistical association was found between male gender, older age and death.
3.4. Sub-groups with different LVEF phenotype In these analyses we considered separately patients with reduced LVEF (b 50%) and those with normal LVEF (≥ 50%). During the follow up, 31 of 409 patients (7.6%) with reduced LVEF (mean 37 ± 7%) and 17 of 200 patients (8.5%; p = ns) with normal LVEF (mean 59 ± 7%) died. Considering the three prognosticators of mortality which emerged in the analysis performed in the whole study population, older age (HR 1.15 [1.07–1.23], p b 0.001) and male gender (HR 6.32 [1.91–20.88], p = 0.002) but not GFR predicted death in multivariate Cox analysis performed in patients with normal LVEF. Conversely, in patients with reduced LVEF, lower GFR was strongly and independently associated
4. Discussion It is well known that HF is a serious condition with high mortality despite all treatments. The prognostic stratification of HF patients aims to select subjects at different risk for clinical adverse events and to identify indicators of survival. In addition it allows optimizing patients management including pharmacological treatment, choice of rehabilitation protocols, devices implantation or palliative pathway. In this study, we focused on the sub-group of HF patients with the lowest risk of
Table 1 Baseline clinical characteristics of 609 patients with 3C-HF score ≤9 points (lowest quartile) compared with those of 1168 patients with higher estimated risk for all-cause mortality (3C-HF score N 9 points). Variable
3C-HF score ≤ 9 points n = 609
3C-HF score N 9 points n = 1168
p
Age (years) Age ≥ 70 years (%) Male gender (%) Body mass index (Kg/height2) Obesity (%) History of hypertension (%) Diabetes (%) Diabetes with target organ damage (%) Ischemic etiology of heart failure (%) Prior myocardial infarction (%) Prior coronary revascularization intervention (%) NYHA functional class III/IV (%) Chronic atrial fibrillation (%) Hemoglobin b 11 g/dl (%) Sodium (mEq/l) eGFR (ml/min/1.73m2) CKD (eGFR b60 ml/min/1.73m2) (%) Cardiac resynchronization therapy (%) Implantable cardioverter defibrillator (%) LV ejection fraction (%) Preserved LV ejection fraction (N50%) (%) Severe valve disease (%) Oral therapy (%) ACEi/ARBs Beta-blockers
72 ± 11 63 59 26.6 ± 4.4 21 57 10 9 38 28 24 0 11 7 138 ± 6 139 ± 4 69 ± 20 27 6 6 44 ± 13 27 8 98 92
78 ± 8 84 56 26.3 ± 4.7 19 57 38 35 55 45 42 28 48 29 139 ± 4 55 ± 24 56 7 7 43 ± 16 32 14 77 53
b0.001 b0.001 0.23 0.39 0.45 0.88 b0.001 b0.001 b0.001 b0.001 b0.001 b0.001 b0.001 b0.001 0.18 b0.001 b0.001 0.49 0.55 0.09 0.06 b0.001 b0.001 b 0.001
ACEi = angiotensin-converting enzyme inhibitor; ARB: angiotensin receptor blockers; CKD: chronic kidney disease; eGFR = estimated glomerular filtration rate; LV: left ventricular; NYHA: New York Heart Association.
950
G. Russo et al. / International Journal of Cardiology 223 (2016) 947–952
Table 2 Clinical characteristics at entry of 609 patients with 3C-HF score ≤9 points (lowest quartile) divided according to the outcome (all-cause mortality). During the follow up 48 patients died and 561 survived. Variable
Deceased n = 48
Survivors n = 561
p
Age (years) Age ≥ 70 years (%) Male gender (%) Body mass index (Kg/height2) Obesity (%) Underweight (%) History of hypertension (%) Diabetes (%) Diabetes with target organ damage (%) Ischemic etiology of heart failure (%) Prior myocardial infarction (%) Prior coronary revascularization intervention (%) Chronic atrial fibrillation (%) NYHA functional class III/IV at baseline (%) Hemoglobin b11 g/dl (%) Sodium (mEq/l) eGFR (ml/min/1.73m2) CKD (eGFR b60 ml/min/1.73m2) Prognostic 3C-HF score Implantable cardioverter defibrillator (%) Cardiac resynchronization therapy (%) LV ejection fraction (%) Preserved LV ejection fraction (N50%) (%) Severe valve disease (%) Oral therapy (%) ACEi/ARBs Beta-blockers
77 ± 8 81 75 25.7 ± 4.4 20 0 71 8 7 33 25 20 5 0 17 139 ± 6 139 ± 4 57 ± 23 52 5.4 ± 2.7 8 11 44 ± 13 31 10
71 ± 11 62 58 26.7 ± 4.1 21 2 56 10 9 38 28 24 19 0 7 139 ± 5 70 ± 19 25 4.2 ± 3.0 6 8 44 ± 13 28 8
b0.001 b0.001 0.02 0.18 0.86 0.45 0.06 0.66 0.77 0.53 0.68 0.63 0.07 0.91 0.01 0.87 b0.001 b0.001 0.006 0.53 0.55 0.99 0.68 0.62
88 75
92 83
0.23 0.19
CKD: chronic kidney disease; eGFR = estimated glomerular filtration rate; LV: left ventricular.
death (selected among a very large cohort of HF subjects using a simple and accurate prognostic clinical score) identifying outliers who unexpectedly died and recognizing the clinical variables which contribute to the mortality prediction. Our study preliminary showed that all-cause mortality rate was 4-fold and re-hospitalization rate was 3.5-fold lower in the 609 study patients at low risk than in the 1168 patients at higher risk. Cox regression analyses showed that the reasons why these HF patients at low risk for all-cause mortality unexpectedly died were worsen renal function, older age and male sex. Renal function is known to be a strong prognostic factor for short and long-term mortality in HF patients [13] and it has a positive correlation with decreasing value of GFR over time. Moreover, chronic HF syndrome is commonly associated with renal dysfunction. Although these two pathological conditions often co-exist in HF patients, the degree of renal dysfunction still remains a strong independent predictor of mortality [14–18]. This finding has been demonstrated both in unselected HF community populations [14–18] and in high risk populations selected for having severe impairment of renal function at baseline evaluation [19]. In our low risk population, the mean values of GFR (69 ± 20 ml/min/1.73 m2) were close to those found in the general population matched for age and sex, and CKD could be diagnosed in only one fourth of our patients. Despite this circumstance, in our study the mortality was strongly associated with the degree of renal function, although it was compromised in a mild way. In fact the cut-off value of GFR by which patients who experienced adverse clinical events could be
Table 3 Variables included in the multivariate Cox proportional hazard analysis that contributed to death prediction. Hazard ratio (HR) represents the increase “per unit” for each variable. Multivariate analysis
HR
95% C.I.
p
Glomerular filtration rate (ml/min/1.73 m2) Male Gender (%) Age (years) Hemoglobin b11 g/dl (%)
0.96 2.93 1.03 1.59
(0.94–0.98) (1.50–5.51) (1.00–1.07) (0.73–3.45)
b0.001 0.002 0.04 0.24
identified was 63 ml/min/1.73 m2. This agreed with a meta-analysis of Hallan et coll. [20]. The authors documented that the mortality risk is increased when GFR decrease below 60 ml/min/1.73 m2, that is the threshold we found in our work to characterized our patients at higher risk to die. Moreover, in their oldest patients, the absolute mortality was higher than in the youngest ones, probably due to the presence of major comorbidities. We found the same results with age: in a context of low risk, the older is our HF patient, the higher is his overall mortality. In general population and high-risk cohorts, men had higher allcause mortality and cardiovascular mortality at all levels of estimated GFR. This finding could be explained by the duration and severity of few risk factors such as diabetes, hypertension and obesity [20]. We can confirm this by our results: men with chronic HF, even if the risk of death is low, demonstrate a lower survival than women. Kurth and coll. [21] in a large prospective white women study did not observe any increased risk of adverse cardiovascular event including death for women with GFR between 60 and 89 ml/min/1.73 m2, compared to those who had GFR b 60/m/min/1.73 m2. A further explanation for the higher mortality in men and old patients could be given by the fact that these data come from three different registries and they are collected from clinical routine. In fact, real life HF patients are older than those described in clinical studies.
Table 4 Accuracy for the prediction of death when the 3 predicting variables were considered alone or as elements of the predictive score. Variables
Area under the TD-ROC curve
95% confidence intervals
Age (years) (≤75 vs N75) Male gender Glomerular filtration rate (b63 vs ≥63 ml/min/1.73 m2) Predictive score for all-cause death⁎
0.65 0.58 0.60
0.57–0.73 0.48–0.69 0.50–0.72
0.72
0.62–0.82
*p = 0.05 vs age (alone). *p = 0.01 vs male gender (alone). *p = 0.02 vs glomerular filtration rate (alone).
G. Russo et al. / International Journal of Cardiology 223 (2016) 947–952
951
Fig. 2. Time-dependent Receiving operating characteristic (TD-ROC) curve analysis comparing the accuracy of the multivariate model generated by the regression equation of the multivariate Cox analysis and the 3 prognosticators of death considered separately: male gender, older age and estimated glomerular filtration rate (eGFR).
In a recent paper by our group on high risk HF patients [22] one of the most important condition independently associated with an adverse clinical outcome was the higher degree of renal dysfunction. Finally, we affirm that even in a low-risk group of HF patients, kidney function is strongly associated with mid-term mortality. This association remains strong in the presence of older age, male gender, cardiovascular phenotypes, the presence of co-morbidities and previous hospitalizations. It is important to close follow-up and to preserve renal function even in patients with HF scored at low-risk of dying. A final consideration regards the stimulating results derived from the analyses performed in the sub-groups of patients with different LVEF phenotype, showing that the prognostic value of the low GFR is particularly high (if not exclusive) in patients with reduced LVEF while the clinical relevance of male gender and older age (which is very high in the sub-group of patients with normal LVEF) is lost in these subjects. These findings have not to be considered as inconsistencies but expected results mainly deriving from the circumstance that the condition defined as “HF with normal LVEF” is a spectrum of overlapping syndromes, each one having different aetiologies and pathophysiologic pathways. These could be the reasons why our two study
sub-groups (normal and reduced LVEF) have two different prognosis and respond unequally at the same medical treatments [23]. In accordance to this understanding, Sanders-van Wijk et al. [24] recently demonstrated that different biomarkers are activated in patients with normal and in those with reduced LVEF, meaning that different pathophysiologic processes are involved in the different LVEF phenotypes of HF. Furthermore, among lots of analyzed biomarkers, Cyst-C (inflammation marker of renal function) had less prognostic impact on patients with normal than those with reduced LVEF, independent of the traditional clinical risk factors. This result is in line with our results, highlighting the prognostic importance of renal dysfunction in patients with reduced more than in those with normal LVEF.
Fig. 3. Rates of all-cause mortality in the 609 patients with chronic heart failure at very low risk for death belonging to the lowest quartile of 3C-HF score (≤9 points) stratified by the number of risk factors for death: male gender, glomerular filtration rate b 63 ml/min/ 1.73 m2 and age N 75 years.
Fig. 4. All-cause mortality curves (Kaplan–Meier plots) of 609 study patients at very low risk for 1-year mortality (selected for having ≤9 points of 3C-HF score), divided in 4 subgroups according to the number of risk factors for death including male gender, glomerular filtration rate b 63 ml/min/1.73 m2 and age N 75 years.
4.1. Study limitations and strengths The main strengths of our work consist of the very large number of participants of both genders, the complete nature of the dataset and the ability to adjust for several CV risk factors and potential
952
G. Russo et al. / International Journal of Cardiology 223 (2016) 947–952
confounders. Furthermore, we used the 3C-HF score [8] in selecting our patients. This score is based on easy-to-obtain cardiac and comorbid variables, and has an excellent accuracy in prognosticating all-cause mortality of HF patients in daily practice. The use of 3C-HF score appears correct because it derives by a prospective multicenter study, which included unselected patients of the real word and considered both patients with depressed and preserved LVEF. The main limitation of the study derives from the observational nature and the design of this study, which did not allow us to assess the role on mortality of some variables such as the patient's frailty and the degree of hemodynamic impairment. Also, in our study we did not present data on biomarkers (such as NT-proBNP or Troponin) because they were not measured routinely at baseline and during follow-up period and were available only in a minority (about one third) of study patients. 5. Conclusions The prognostic stratification of chronic HF patients allows in daily practice to select patients at different risk for death and identify prognosticators of mortality in outliers at very low risk of death. Patients with chronic HF at very low risk for all-cause mortality who die are older, more frequently male and have slightly lower GFR values than counterparts who survive. These results imply that checking renal function, discontinuing the progression of renal dysfunction and/or reversing renal failure have to be the primary goal for the clinical management of these patients. Conflicts of interest The authors report no relationships that could be construed as a conflict of interest. References [1] R. Vazquez, A. Bayes-Genis, I. Cygankiewicz, et al., The MUSIC risk score: a simple method for predicting mortality in ambulatory patients with chronic heart failure, Eur. Heart J. 30 (2009) 1088–1096. [2] F. Alla, S. Briançon, Y. Juillière, P.M. Mertes, J.P. Villemot, F. Zannad, Differential clinical prognostic classifications in dilated and ischemic advanced heart failure: the EPICAL study, Am. Heart J. 139 (2000) 895–904. [3] M.T. Kearney, J. Nolan, A.J. Lee, et al., A prognostic index to predict long-term mortality in patients with mild to moderate chronic heart failure stabilised on angiotensin converting enzyme inhibitors, Eur. J. Heart Fail. 5 (2003) 489–497. [4] W.C. Levy, D. Mozaffarian, D.T. Linker, et al., The Seattle heart failure model: prediction of survival in heart failure, Circulation 113 (2006) 1424–1433.
[5] K.D. Aaronson, J.S. Schwatz, T.M. Chen, K.L. Wong, J.E. Goin, D.M. Mancini, Development and prospective validation of a clinical index to predict survival in ambulatory patients referred to cardiac transplant evaluation, Circulation 95 (1997) 2660–2667. [6] C. Campana, A. Gavazzi, C. Berzuini, et al., Predictors of prognosis in patients awaiting heart transplantation, J. Heart Lung Transplant. 12 (1993) 756–765. [7] L. Manzano, D. Babalis, M. Roughton, et al., Predictors of clinical outcomes in elderly patients with heart failure, Eur. J. Heart Fail. 13 (2011) 528–536. [8] M. Senni, P. Parrella, R. De Maria, C. Cottini, M. Bohm, P. Polikowski, et al., Predicting heart failure outcome from cardiac and comorbid conditions: the 3C-HF score, Int. J. Cardiol. 163 (2013) 206–211. [9] W.J. Remme, K. Swedberg, Task force for the diagnosis and treatment of chronic heart failure, European society of cardiology. Guidelines for the diagnosis and treatment of chronic heart failure, Eur. Heart J. 22 (2001) 1527–1560. [10] ACC/AHA practice guidelines. Guidelines for the management of patients with valvular heart disease, Circulation 98 (1998) 1949–1984. [11] WHO Expert Consultation, Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies, Lancet 363 (2004) 157–163. [12] http://www.unos.org/PoliciesandBylaws2/policies/pdfs/policy_9.pdf at http:// www.unos.org/policiesandbylaws/policies.asp?resource (18 September 2009). [13] S.K. LöfmanI, I. Hagerman, et al., Prevalence and prognostic impact of kidney disease on heart failure patients, BMJ 3 (2016), e000324. [14] D.L. Dries, D.V. Exner, M.J. Domanski, B. Greenberg, L.W. Stevenson, The prognostic implications of renal insufficiency in asymptomatic and symptomatic patients with left ventricular systolic dysfunction, J. Am. Coll. Cardiol. 35 (2000) 681–689. [15] H.L. Hillege, A.R. Girbes, P.J. de Kam, et al., Renal function, neurohormonal activation, and survival in patients with chronic heart failure, Circulation 102 (2000) 203–210. [16] A. Al-Ahmad, W.M. Rand, G. Manjunath, et al., Reduced kidney function and anemia as risk factors for mortality in patients with left ventricular dysfunction, J. Am. Coll. Cardiol. 38 (2001) 955–962. [17] H.L. Hillege, D. Nitsch, M.A. Pfeffer, et al., Renal function as a predictor of outcome in a broad spectrum of patients with heart failure, Circulation 113 (2006) 671–678. [18] G.L. Smith, J.H. Lichtman, M.B. Bracken, et al., Renal impairment and outcomes in heart failure: systematic review and meta-analysis, J. Am. Coll. Cardiol. 47 (2006) 1987–1996. [19] G. Cioffi, A. Mortara, A. Di Lenarda, et al., Italian acute heart failure survey. Clinical features, and in-hospital and 1-year mortalities of patients with acute heart failure and severe renal dysfunction. Data from the Italian registry IN-HF outcome, Int. J. Cardiol. 168 (2013) 3691–3697. [20] S.I. Hallan, K. Matsushita, Y. Sang, et al., Age and association of kidney measures with mortality and end-stage renal disease, JAMA 308 (2012) 2349–2360. [21] T. Kurth, P.E. De Jong, N.R. Cook, et al., Kidney function and risk of cardiovascular disease and mortality in women: a prospective cohort study, BMJ 338 (2009) b2392. [22] G. Cioffi, G. Pulignano, G. Barbati, et al., Reasons why patients suffering from chronic heart failure at very high risk for death survive, Int. J. Cardiol. 177 (1) (2014) 213–218. [23] D.P. Kao, J.D. Lewsey, I.S. Anand, et al., Characterization of subgroups of heart failure patients with preserved ejection fraction with possible implications for prognosis and treatment response, Eur. J. Heart Fail. 17 (2015) 925–935. [24] S. Sanders-van Wijk, V. van Empel, N. Davarzani, M.T. Maeder, et al., TIME-CHF investigators. Circulating biomarkers of distinct pathophysiological pathways in heart failure with preserved vs. reduced left ventricular ejection fraction, Eur. J. Heart Fail. 17 (2015) 1006–1014.
Contents lists available at ScienceDirect
International Journal of Cardiology journal homepage: www.elsevier.com/locate/ijcard
Reasons why patients suffering from chronic heart failure at very low risk for mortality die Giulia Russo a,⁎, Giovanni Cioffi b, Giovanni Pulignano c, Giulia Barbati a, Luigi Tarantini d, Donatella Del Sindaco e, Carmine Mazzone a, Antonella Cherubini a, Giorgio Faganello a, Carlo Stefenelli b, Michele Senni f, Andrea Di Lenarda a a
Cardiovascular Center, ASUITS, Italy Cardiology Department, Villa Bianca Hospital Trento, Italy Heart Failure Clinic, Division of Cardiology/C.C.U. San Camillo Hospital, Rome, Italy d Cardiology Department, St. Martino Hospital. Azienda Sanitaria Locale n. 1, Belluno, Italy e Department of Cardiocirculatory Diseases, San Giovanni-Addolorata Hospital, Rome, Italy f Cardiovascular Department, Ospedali Riuniti, Bergamo, Italy b c
a r t i c l e
i n f o
Article history: Received 14 May 2016 Received in revised form 19 August 2016 Accepted 20 August 2016 Available online 23 August 2016 Keywords: Chronic heart failure Mortality Renal dysfunction Prognosis Older age Male gender
a b s t r a c t Background: A proper prognostic stratification is crucial for organizing an effective clinical management and treatment decision-making in patients with chronic heart failure (CHF). In this study, we selected and characterized a sub-group of CHF patients at very low risk for death aiming to assess predictors of death in subjects with an expected probability of 1-year mortality near to 5%. Methods: We used the Cardiac and Comorbid Conditions HF (3C-HF) Score to identify CHF patients with the best mid-term prognosis. We selected patients belonging to the lowest quartile of 3C-HF score (≤9 points). Results: We recruited 1777 consecutive CHF patients at 3 Italian Cardiology Units (age 76 ± 10 years, 43% female, 32% with preserved ejection fraction). Subjects belonging to the lowest quartile of 3C-HF score were 609. During a median follow-up of 21 [12–40] months, 48 of these patients (8%) died, and 561 (92%) survived. The variables that contributed to death prediction by Cox regression multivariate analysis were older age (HR 1.03[CI 1.00– 1.07]; p = 0.04), male gender (HR 2.93[CI 1.50–5.51]; p = 0.002) and a higher degree of renal dysfunction (HR 0.96[CI 0.94–0.98]; p b 0.001). Conclusions: The prognostic stratification of CHF patients by 3C-HF score allows one to select patients at different outcome and to identify the factors associated with death in outliers with a very low mortality risk at mid-term follow-up. The reasons why these patients do not outlive the matching part of subjects who expectedly survive are related to a declined renal function and unmodifiable conditions including older age and male gender. © 2016 Elsevier Ireland Ltd. All rights reserved.
1. Introduction An accurate prognostic stratification is essential for optimizing the clinical management and treatment decision-making of patients with chronic heart failure (HF). In daily clinical practice the assessment of risk for adverse outcomes is often limited and based on individual clinician's capabilities. The combined use of specific and sensible variables tested in prognostic models has appeared as the most appropriate methodology to decipher the natural history of HF syndrome in the single individual [1–7]. By this approach, however, it is usually possible to identify the macro-phenomena leading to the adverse clinical events, while it is “by definition” not possible to characterize the ⁎ Corresponding author at: Cardiovascular Center, ASUITS, Via Slataper 9, 34125 Trieste, Italy. E-mail address: [email protected] (G. Russo).
http://dx.doi.org/10.1016/j.ijcard.2016.08.326 0167-5273/© 2016 Elsevier Ireland Ltd. All rights reserved.
patients defined as “outliers”, who have atypical behaviors and outcomes far from those expected. No data are available on the protecting conditions from adverse clinical events in patients with chronic HF and low risk of death. Accordingly, in this study, among the best available models, we used the Cardiac and Comorbid Conditions HF (3C-HF) Score [8], to predict all-cause mortality in patients with chronic HF. Thus, according to the score, we selected the subgroup of patients with chronic HF and the lowest risk for all-cause mortality with the aim to characterize and assess the predictors of death in those subjects who unexpectedly died during the follow-up. 2. Material and methods Patients for this study were enrolled by 3 Italian HF clinical units into a prospective evaluation and their data collected in the same database. We selected adult subjects in stable clinical conditions [New York Heart Association (NYHA) functional class II or III] who
948
G. Russo et al. / International Journal of Cardiology 223 (2016) 947–952
had previously experienced at least an episode of HF requiring a hospital admission [9]. At enrollment, we considered clinical, laboratory, and echocardiographic data within the last 6 months prior to enrollment. Practical guidelines were used to define severity of valve heart disease [10] and optimization of medical treatment [9]. Left ventricular ejection fraction (LVEF) was measured by echocardiography. Patients with HF symptoms and a LVEF ≥50% (defined as “patients with normal LVEF”) had to show lung congestion (clinical or by chest X-ray), abnormal BNP values or high LV filling pressure at echocardiographic evaluation. History of hypertension was defined a condition preceding the first episode of congestive HF as systolic blood pressure of ≥140 mm Hg and/or a diastolic blood pressure of ≥ 90 mm Hg and/or pharmacologically treated high blood pressure of unknown etiology. Obesity was diagnosed if patients had body mass index ≥30 kg/m2. Cut-off point for defining the condition of “underweight” was 18.5 Kg/m2 [11]. To assess renal function we considered the glomerular filtration rate (GFR) estimated with the CKD-EPI equation. Patients were followed-up at each center for a period lasting 1 year at least. Survival status was ascertained locally by follow-up visits or chart review, telephone interview with the patient, or his/her family, or primary care physician, or by examination of death certificates. Cause of death was physician-reported. Urgent heart transplantation, defined as UNOS status one [12], was counted as a death. Patients who underwent elective heart transplantation were censored seven days after the procedure. Patient follow-up was 100% complete. Patients expressed their general written consent to the anonymous use of data for their care and research purposes. Databases for clinical use were authorized at each center. The study complies with the Declaration of Helsinki; the locally appointed ethics committee approved the cardiovascular registries. For this investigation we initially categorized all the participants according to the 3CHF score classification [8], which considered the following 11 variables: age, NYHA functional class, history of hypertension, atrial fibrillation, diabetes with target organ damage, anemia, chronic kidney dysfunction, LVEF, severe heart valve disease, current treatment with beta-blocker, and current treatment with RAS-inhibitor. Thus, we selected the subgroup of patients belonging to the lowest quartile of the score (≤9 points) having a probability of 1-year mortality which approximated to 5% [8]. Patients with an ascertained diagnosis of hypertrophic or restrictive cardiomyopathy, those with acute myocarditis or with alcohol-induced myocardial injury were excluded from the study. Those subjects who experienced a myocardial infarction and/or underwent a percutaneous transcatheter coronary angioplasty or coronary artery bypass graft or any intervention for valvular disease within 3 months before our first evaluation up were also rejected. Implanted cardioverter defibrillator and/or cardiac resynchronization therapy were not considered contraindications for enrollment.
2.1. Assessment and follow-up A careful history was obtained from all patients: this included an evaluation of NYHA functional class and a complete physical and echocardiographic examination, routine blood tests and standard electrocardiography. During the follow-up, patients received diuretics (flexible regimen) and/or low dose of spironolactone if indicated, and highest tolerated dose of angiotensin-converting enzyme inhibitors (ACE-I), which were replaced on angiotensin type 1 receptor blockers (ARBs) if not tolerated. Carvedilol, metoprolol, bisoprolol or nebivolol were also given at the highest tolerated dose.
2.2. Statistical analysis Categorical variables are presented as percentages, while continuous variables are presented as their means and SD. Categorical variables were compared by the chisquare test and continuous variables by the t-test or the Mann–Whitney U-test. Log cumulative hazard functions were computed by univariable and multivariable Cox proportional hazards analyses (SPSS version 19.0, SPSS Inc. Chicago, Illinois, USA) to identify the predictors of primary end-point (all-cause death). Variables significantly related to the primary end-point in univariate tests (p b 0.05) were included in the multivariable model. The final model was internally validated by means of bootstrap technique (number of samples = 1000, level of confidence interval = 95%). Two sub-analyses considering only patients with reduced LVEF (b50%) and only patients with normal LVEF (≥50%), respectively, were made to assess whether the independent prognostic markers in the low risk population were similar in the two sub-groups with different LVEF phenotype. In order to simplify the clinical usage of the estimated model, Receiver Operating Characteristic (ROC) curve analyses were performed to identify the ‘best’ cut-off points (i.e. values that had the highest sum of sensibility and specificity) of the continuous variables independently associated to the primary end-point which were identified by Cox regression analysis. A score was then defined by summing the significant covariates in the model in their ‘cut-off’ version. Time-dependent ROC curves for censored data (TD-ROC) were estimated, plotted and compared to assess the accuracy of this score and also the accuracy of the single covariates considered separately. TD-ROC curves are calculated taking into account the Inverse Probability of Censoring Weighting (IPCW) estimation of a Cumulative/Dynamic ROC curve: this curve intrinsically depends on the definitions of time-dependent “cases” and “controls”. In our case, we used as time-horizon for the events 24 months of follow up, so defining “case” a subject with an event observed before 24 months and as “control” a subject that survives beyond 24 months. We used for computations the library “timeROC” of the R statistical package, version 3.1.2. A 2-tailed value of p b 0.05 was used to reject the null hypothesis of no difference.
3. Results 3.1. Study population During the enrollment period, 1912 patients with chronic HF were included in the database and 1777 (93%) were eligible for this study; among these 1777 subjects, 609 (34.2%) were selected according to the lowest quartile of 3C-HF score (≤ 9 points) and formed the final study population (Fig. 1). Baseline demographics and clinical characteristics of the 609 patients with the lowest risk of death are reported and compared with those of 1168 patients at higher risk in Table 1. As expected, the former were younger, less affected by diabetes mellitus, atrial fibrillation, anemia or severe chronic renal disease, they were more frequently treated with beta-blockers and ACE-I/ARBs than the latter. Considering the total population of 1777 patients, the probability of all-cause mortality at 1-year follow-up estimated by the 3C-HF logistic score was 15%. During a median follow-up of 21 [12–40] months 398 patients died (22%). All-cause mortality rate was 7.9% in the 609 patients who had the lowest risk of death belonging to the lowest quartile of 3C-HF score, and 30.4% in the 1168 patients at higher risk belonging to the other quartiles (p b 0.0001). The incidence of re-hospitalization (occurred during the same period of observation) was 10.1% in the former and 35% in the latter.
3.2. Picture of the outliers During the follow up, among the 609 patients with the lowest risk of death, 48 (7.9%) died. These patients were older, were more frequently men, had a worse renal function, and a higher prevalence of anemia than the 561 counterparts who survived (92.1%). There was no difference in LVEF and pharmacological therapy between the two groups (Table 2). Mode of death was cardiovascular cause in 63% of cases (progressive pump failure = 27%; sudden death = 53%; acute myocardial infarction 20%), and non-cardiovascular cause in 27% of cases (respiratory disease = 26%; renal failure = 23%; cancer = 33%; other causes = 20%). Cause of death was unknown in the remaining 10% of patients.
3.3. Predictors of death Multivariable Cox regression was performed to characterize the patients who died in spite of the very low risk of death predicted by the prognostic 3C-HF score. Among the covariates age, gender, renal function (measured as GFR) and hemoglobin (b vs N 11 mg/dl), selected by univariate analysis, at multivariate modeling age, gender and GFR were confirmed as independently associated with the study end-point (Table 3). This model was internally validated by a bootstrap technique which confirmed its solidity and constancy: GFR, age and gender appeared as the three strongest prognosticators of mortality (all p = 0.001). The ‘best’ cut-off points of the continuous variables were identified by ROC curve analyses: they were 63 ml/min/1.73 m2 for GFR (AUC 0.60 [CI 0.50–0.72], sensitivity 57% and specificity 63%) and 75 years of age (AUC 0.65 [0.57–0.73], sensitivity 74% and specificity 61%) (Table 4). Fig. 2 shows the results of TD-ROC curve analyses demonstrating a better accuracy of the estimated score (AUC 0.72 [0.62–0.82]) than the 3 prognosticators of death considered separately. Fig. 3 reports the event rates in the 609 study patients stratified by the number of risk factors for death: it ranged from 3% in patients who had none or only 1 risk factor to 20% in those who had the 3 risk factors (male gender, age N 75 years and GFR b 63 ml/min/1.73 m2 coexisted). Based on these results, four sub-groups of patients were identified according to the number of risk factors for death (from zero to 3). Fig. 4 shows Kaplan–Meier all-cause mortality curves of these four sub-groups of patients (difference by Log Rank Test: p b 0.0001).
G. Russo et al. / International Journal of Cardiology 223 (2016) 947–952
949
Fig. 1. Flow chart of patient's selection showing the number of subjects included and excluded according to the criteria for enrolment.
with adverse outcome (HR 4.06 [1.82–9.02], p b 0.001), while no statistical association was found between male gender, older age and death.
3.4. Sub-groups with different LVEF phenotype In these analyses we considered separately patients with reduced LVEF (b 50%) and those with normal LVEF (≥ 50%). During the follow up, 31 of 409 patients (7.6%) with reduced LVEF (mean 37 ± 7%) and 17 of 200 patients (8.5%; p = ns) with normal LVEF (mean 59 ± 7%) died. Considering the three prognosticators of mortality which emerged in the analysis performed in the whole study population, older age (HR 1.15 [1.07–1.23], p b 0.001) and male gender (HR 6.32 [1.91–20.88], p = 0.002) but not GFR predicted death in multivariate Cox analysis performed in patients with normal LVEF. Conversely, in patients with reduced LVEF, lower GFR was strongly and independently associated
4. Discussion It is well known that HF is a serious condition with high mortality despite all treatments. The prognostic stratification of HF patients aims to select subjects at different risk for clinical adverse events and to identify indicators of survival. In addition it allows optimizing patients management including pharmacological treatment, choice of rehabilitation protocols, devices implantation or palliative pathway. In this study, we focused on the sub-group of HF patients with the lowest risk of
Table 1 Baseline clinical characteristics of 609 patients with 3C-HF score ≤9 points (lowest quartile) compared with those of 1168 patients with higher estimated risk for all-cause mortality (3C-HF score N 9 points). Variable
3C-HF score ≤ 9 points n = 609
3C-HF score N 9 points n = 1168
p
Age (years) Age ≥ 70 years (%) Male gender (%) Body mass index (Kg/height2) Obesity (%) History of hypertension (%) Diabetes (%) Diabetes with target organ damage (%) Ischemic etiology of heart failure (%) Prior myocardial infarction (%) Prior coronary revascularization intervention (%) NYHA functional class III/IV (%) Chronic atrial fibrillation (%) Hemoglobin b 11 g/dl (%) Sodium (mEq/l) eGFR (ml/min/1.73m2) CKD (eGFR b60 ml/min/1.73m2) (%) Cardiac resynchronization therapy (%) Implantable cardioverter defibrillator (%) LV ejection fraction (%) Preserved LV ejection fraction (N50%) (%) Severe valve disease (%) Oral therapy (%) ACEi/ARBs Beta-blockers
72 ± 11 63 59 26.6 ± 4.4 21 57 10 9 38 28 24 0 11 7 138 ± 6 139 ± 4 69 ± 20 27 6 6 44 ± 13 27 8 98 92
78 ± 8 84 56 26.3 ± 4.7 19 57 38 35 55 45 42 28 48 29 139 ± 4 55 ± 24 56 7 7 43 ± 16 32 14 77 53
b0.001 b0.001 0.23 0.39 0.45 0.88 b0.001 b0.001 b0.001 b0.001 b0.001 b0.001 b0.001 b0.001 0.18 b0.001 b0.001 0.49 0.55 0.09 0.06 b0.001 b0.001 b 0.001
ACEi = angiotensin-converting enzyme inhibitor; ARB: angiotensin receptor blockers; CKD: chronic kidney disease; eGFR = estimated glomerular filtration rate; LV: left ventricular; NYHA: New York Heart Association.
950
G. Russo et al. / International Journal of Cardiology 223 (2016) 947–952
Table 2 Clinical characteristics at entry of 609 patients with 3C-HF score ≤9 points (lowest quartile) divided according to the outcome (all-cause mortality). During the follow up 48 patients died and 561 survived. Variable
Deceased n = 48
Survivors n = 561
p
Age (years) Age ≥ 70 years (%) Male gender (%) Body mass index (Kg/height2) Obesity (%) Underweight (%) History of hypertension (%) Diabetes (%) Diabetes with target organ damage (%) Ischemic etiology of heart failure (%) Prior myocardial infarction (%) Prior coronary revascularization intervention (%) Chronic atrial fibrillation (%) NYHA functional class III/IV at baseline (%) Hemoglobin b11 g/dl (%) Sodium (mEq/l) eGFR (ml/min/1.73m2) CKD (eGFR b60 ml/min/1.73m2) Prognostic 3C-HF score Implantable cardioverter defibrillator (%) Cardiac resynchronization therapy (%) LV ejection fraction (%) Preserved LV ejection fraction (N50%) (%) Severe valve disease (%) Oral therapy (%) ACEi/ARBs Beta-blockers
77 ± 8 81 75 25.7 ± 4.4 20 0 71 8 7 33 25 20 5 0 17 139 ± 6 139 ± 4 57 ± 23 52 5.4 ± 2.7 8 11 44 ± 13 31 10
71 ± 11 62 58 26.7 ± 4.1 21 2 56 10 9 38 28 24 19 0 7 139 ± 5 70 ± 19 25 4.2 ± 3.0 6 8 44 ± 13 28 8
b0.001 b0.001 0.02 0.18 0.86 0.45 0.06 0.66 0.77 0.53 0.68 0.63 0.07 0.91 0.01 0.87 b0.001 b0.001 0.006 0.53 0.55 0.99 0.68 0.62
88 75
92 83
0.23 0.19
CKD: chronic kidney disease; eGFR = estimated glomerular filtration rate; LV: left ventricular.
death (selected among a very large cohort of HF subjects using a simple and accurate prognostic clinical score) identifying outliers who unexpectedly died and recognizing the clinical variables which contribute to the mortality prediction. Our study preliminary showed that all-cause mortality rate was 4-fold and re-hospitalization rate was 3.5-fold lower in the 609 study patients at low risk than in the 1168 patients at higher risk. Cox regression analyses showed that the reasons why these HF patients at low risk for all-cause mortality unexpectedly died were worsen renal function, older age and male sex. Renal function is known to be a strong prognostic factor for short and long-term mortality in HF patients [13] and it has a positive correlation with decreasing value of GFR over time. Moreover, chronic HF syndrome is commonly associated with renal dysfunction. Although these two pathological conditions often co-exist in HF patients, the degree of renal dysfunction still remains a strong independent predictor of mortality [14–18]. This finding has been demonstrated both in unselected HF community populations [14–18] and in high risk populations selected for having severe impairment of renal function at baseline evaluation [19]. In our low risk population, the mean values of GFR (69 ± 20 ml/min/1.73 m2) were close to those found in the general population matched for age and sex, and CKD could be diagnosed in only one fourth of our patients. Despite this circumstance, in our study the mortality was strongly associated with the degree of renal function, although it was compromised in a mild way. In fact the cut-off value of GFR by which patients who experienced adverse clinical events could be
Table 3 Variables included in the multivariate Cox proportional hazard analysis that contributed to death prediction. Hazard ratio (HR) represents the increase “per unit” for each variable. Multivariate analysis
HR
95% C.I.
p
Glomerular filtration rate (ml/min/1.73 m2) Male Gender (%) Age (years) Hemoglobin b11 g/dl (%)
0.96 2.93 1.03 1.59
(0.94–0.98) (1.50–5.51) (1.00–1.07) (0.73–3.45)
b0.001 0.002 0.04 0.24
identified was 63 ml/min/1.73 m2. This agreed with a meta-analysis of Hallan et coll. [20]. The authors documented that the mortality risk is increased when GFR decrease below 60 ml/min/1.73 m2, that is the threshold we found in our work to characterized our patients at higher risk to die. Moreover, in their oldest patients, the absolute mortality was higher than in the youngest ones, probably due to the presence of major comorbidities. We found the same results with age: in a context of low risk, the older is our HF patient, the higher is his overall mortality. In general population and high-risk cohorts, men had higher allcause mortality and cardiovascular mortality at all levels of estimated GFR. This finding could be explained by the duration and severity of few risk factors such as diabetes, hypertension and obesity [20]. We can confirm this by our results: men with chronic HF, even if the risk of death is low, demonstrate a lower survival than women. Kurth and coll. [21] in a large prospective white women study did not observe any increased risk of adverse cardiovascular event including death for women with GFR between 60 and 89 ml/min/1.73 m2, compared to those who had GFR b 60/m/min/1.73 m2. A further explanation for the higher mortality in men and old patients could be given by the fact that these data come from three different registries and they are collected from clinical routine. In fact, real life HF patients are older than those described in clinical studies.
Table 4 Accuracy for the prediction of death when the 3 predicting variables were considered alone or as elements of the predictive score. Variables
Area under the TD-ROC curve
95% confidence intervals
Age (years) (≤75 vs N75) Male gender Glomerular filtration rate (b63 vs ≥63 ml/min/1.73 m2) Predictive score for all-cause death⁎
0.65 0.58 0.60
0.57–0.73 0.48–0.69 0.50–0.72
0.72
0.62–0.82
*p = 0.05 vs age (alone). *p = 0.01 vs male gender (alone). *p = 0.02 vs glomerular filtration rate (alone).
G. Russo et al. / International Journal of Cardiology 223 (2016) 947–952
951
Fig. 2. Time-dependent Receiving operating characteristic (TD-ROC) curve analysis comparing the accuracy of the multivariate model generated by the regression equation of the multivariate Cox analysis and the 3 prognosticators of death considered separately: male gender, older age and estimated glomerular filtration rate (eGFR).
In a recent paper by our group on high risk HF patients [22] one of the most important condition independently associated with an adverse clinical outcome was the higher degree of renal dysfunction. Finally, we affirm that even in a low-risk group of HF patients, kidney function is strongly associated with mid-term mortality. This association remains strong in the presence of older age, male gender, cardiovascular phenotypes, the presence of co-morbidities and previous hospitalizations. It is important to close follow-up and to preserve renal function even in patients with HF scored at low-risk of dying. A final consideration regards the stimulating results derived from the analyses performed in the sub-groups of patients with different LVEF phenotype, showing that the prognostic value of the low GFR is particularly high (if not exclusive) in patients with reduced LVEF while the clinical relevance of male gender and older age (which is very high in the sub-group of patients with normal LVEF) is lost in these subjects. These findings have not to be considered as inconsistencies but expected results mainly deriving from the circumstance that the condition defined as “HF with normal LVEF” is a spectrum of overlapping syndromes, each one having different aetiologies and pathophysiologic pathways. These could be the reasons why our two study
sub-groups (normal and reduced LVEF) have two different prognosis and respond unequally at the same medical treatments [23]. In accordance to this understanding, Sanders-van Wijk et al. [24] recently demonstrated that different biomarkers are activated in patients with normal and in those with reduced LVEF, meaning that different pathophysiologic processes are involved in the different LVEF phenotypes of HF. Furthermore, among lots of analyzed biomarkers, Cyst-C (inflammation marker of renal function) had less prognostic impact on patients with normal than those with reduced LVEF, independent of the traditional clinical risk factors. This result is in line with our results, highlighting the prognostic importance of renal dysfunction in patients with reduced more than in those with normal LVEF.
Fig. 3. Rates of all-cause mortality in the 609 patients with chronic heart failure at very low risk for death belonging to the lowest quartile of 3C-HF score (≤9 points) stratified by the number of risk factors for death: male gender, glomerular filtration rate b 63 ml/min/ 1.73 m2 and age N 75 years.
Fig. 4. All-cause mortality curves (Kaplan–Meier plots) of 609 study patients at very low risk for 1-year mortality (selected for having ≤9 points of 3C-HF score), divided in 4 subgroups according to the number of risk factors for death including male gender, glomerular filtration rate b 63 ml/min/1.73 m2 and age N 75 years.
4.1. Study limitations and strengths The main strengths of our work consist of the very large number of participants of both genders, the complete nature of the dataset and the ability to adjust for several CV risk factors and potential
952
G. Russo et al. / International Journal of Cardiology 223 (2016) 947–952
confounders. Furthermore, we used the 3C-HF score [8] in selecting our patients. This score is based on easy-to-obtain cardiac and comorbid variables, and has an excellent accuracy in prognosticating all-cause mortality of HF patients in daily practice. The use of 3C-HF score appears correct because it derives by a prospective multicenter study, which included unselected patients of the real word and considered both patients with depressed and preserved LVEF. The main limitation of the study derives from the observational nature and the design of this study, which did not allow us to assess the role on mortality of some variables such as the patient's frailty and the degree of hemodynamic impairment. Also, in our study we did not present data on biomarkers (such as NT-proBNP or Troponin) because they were not measured routinely at baseline and during follow-up period and were available only in a minority (about one third) of study patients. 5. Conclusions The prognostic stratification of chronic HF patients allows in daily practice to select patients at different risk for death and identify prognosticators of mortality in outliers at very low risk of death. Patients with chronic HF at very low risk for all-cause mortality who die are older, more frequently male and have slightly lower GFR values than counterparts who survive. These results imply that checking renal function, discontinuing the progression of renal dysfunction and/or reversing renal failure have to be the primary goal for the clinical management of these patients. Conflicts of interest The authors report no relationships that could be construed as a conflict of interest. References [1] R. Vazquez, A. Bayes-Genis, I. Cygankiewicz, et al., The MUSIC risk score: a simple method for predicting mortality in ambulatory patients with chronic heart failure, Eur. Heart J. 30 (2009) 1088–1096. [2] F. Alla, S. Briançon, Y. Juillière, P.M. Mertes, J.P. Villemot, F. Zannad, Differential clinical prognostic classifications in dilated and ischemic advanced heart failure: the EPICAL study, Am. Heart J. 139 (2000) 895–904. [3] M.T. Kearney, J. Nolan, A.J. Lee, et al., A prognostic index to predict long-term mortality in patients with mild to moderate chronic heart failure stabilised on angiotensin converting enzyme inhibitors, Eur. J. Heart Fail. 5 (2003) 489–497. [4] W.C. Levy, D. Mozaffarian, D.T. Linker, et al., The Seattle heart failure model: prediction of survival in heart failure, Circulation 113 (2006) 1424–1433.
[5] K.D. Aaronson, J.S. Schwatz, T.M. Chen, K.L. Wong, J.E. Goin, D.M. Mancini, Development and prospective validation of a clinical index to predict survival in ambulatory patients referred to cardiac transplant evaluation, Circulation 95 (1997) 2660–2667. [6] C. Campana, A. Gavazzi, C. Berzuini, et al., Predictors of prognosis in patients awaiting heart transplantation, J. Heart Lung Transplant. 12 (1993) 756–765. [7] L. Manzano, D. Babalis, M. Roughton, et al., Predictors of clinical outcomes in elderly patients with heart failure, Eur. J. Heart Fail. 13 (2011) 528–536. [8] M. Senni, P. Parrella, R. De Maria, C. Cottini, M. Bohm, P. Polikowski, et al., Predicting heart failure outcome from cardiac and comorbid conditions: the 3C-HF score, Int. J. Cardiol. 163 (2013) 206–211. [9] W.J. Remme, K. Swedberg, Task force for the diagnosis and treatment of chronic heart failure, European society of cardiology. Guidelines for the diagnosis and treatment of chronic heart failure, Eur. Heart J. 22 (2001) 1527–1560. [10] ACC/AHA practice guidelines. Guidelines for the management of patients with valvular heart disease, Circulation 98 (1998) 1949–1984. [11] WHO Expert Consultation, Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies, Lancet 363 (2004) 157–163. [12] http://www.unos.org/PoliciesandBylaws2/policies/pdfs/policy_9.pdf at http:// www.unos.org/policiesandbylaws/policies.asp?resource (18 September 2009). [13] S.K. LöfmanI, I. Hagerman, et al., Prevalence and prognostic impact of kidney disease on heart failure patients, BMJ 3 (2016), e000324. [14] D.L. Dries, D.V. Exner, M.J. Domanski, B. Greenberg, L.W. Stevenson, The prognostic implications of renal insufficiency in asymptomatic and symptomatic patients with left ventricular systolic dysfunction, J. Am. Coll. Cardiol. 35 (2000) 681–689. [15] H.L. Hillege, A.R. Girbes, P.J. de Kam, et al., Renal function, neurohormonal activation, and survival in patients with chronic heart failure, Circulation 102 (2000) 203–210. [16] A. Al-Ahmad, W.M. Rand, G. Manjunath, et al., Reduced kidney function and anemia as risk factors for mortality in patients with left ventricular dysfunction, J. Am. Coll. Cardiol. 38 (2001) 955–962. [17] H.L. Hillege, D. Nitsch, M.A. Pfeffer, et al., Renal function as a predictor of outcome in a broad spectrum of patients with heart failure, Circulation 113 (2006) 671–678. [18] G.L. Smith, J.H. Lichtman, M.B. Bracken, et al., Renal impairment and outcomes in heart failure: systematic review and meta-analysis, J. Am. Coll. Cardiol. 47 (2006) 1987–1996. [19] G. Cioffi, A. Mortara, A. Di Lenarda, et al., Italian acute heart failure survey. Clinical features, and in-hospital and 1-year mortalities of patients with acute heart failure and severe renal dysfunction. Data from the Italian registry IN-HF outcome, Int. J. Cardiol. 168 (2013) 3691–3697. [20] S.I. Hallan, K. Matsushita, Y. Sang, et al., Age and association of kidney measures with mortality and end-stage renal disease, JAMA 308 (2012) 2349–2360. [21] T. Kurth, P.E. De Jong, N.R. Cook, et al., Kidney function and risk of cardiovascular disease and mortality in women: a prospective cohort study, BMJ 338 (2009) b2392. [22] G. Cioffi, G. Pulignano, G. Barbati, et al., Reasons why patients suffering from chronic heart failure at very high risk for death survive, Int. J. Cardiol. 177 (1) (2014) 213–218. [23] D.P. Kao, J.D. Lewsey, I.S. Anand, et al., Characterization of subgroups of heart failure patients with preserved ejection fraction with possible implications for prognosis and treatment response, Eur. J. Heart Fail. 17 (2015) 925–935. [24] S. Sanders-van Wijk, V. van Empel, N. Davarzani, M.T. Maeder, et al., TIME-CHF investigators. Circulating biomarkers of distinct pathophysiological pathways in heart failure with preserved vs. reduced left ventricular ejection fraction, Eur. J. Heart Fail. 17 (2015) 1006–1014.