Coronary artery disease and 10-year outcome after hospital admission for heart failure with preserved and with reduced ejection fraction

Population and study protocol

The ETICS registry has been described previously.18 Briefly, all consecutive patients >20 years old, admitted to hospital during the year 2000 for new-onset HF in the 11 establishments of the Somme department, France, were prospectively enrolled. Patients admitted for worsening chronic HF and patients admitted primarily for acute myocardial infarction were excluded. However, patients in whom the diagnosis of acute coronary syndrome was not suspected at admission but confirmed subsequently based on electrocardiogram (ECG) changes and biomarkers were kept in the analysis. The diagnosis of HF was first established by the attending physician in 811 patients, and then validated by two independent cardiologists in 799 patients using the Framingham criteria amended by the European Society of Cardiology.19 Patients in whom EF was not assessed during the index hospitalization were excluded. Furthermore, patients in whom valvular heart disease was the underlying cause of HF were also excluded. The study population therefore included 591 patients (Figure 1). A cut-off of 50% was used to distinguish HFPEF from HFREF.18, 20 EF was preserved in 320 patients and reduced in 271.

Clinical data, including medical history and cardiovascular risk factors were recorded. Complementary investigations included laboratory tests, ECG, chest X-ray on admission, echocardiography, and in some patients coronary angiography. The performance of coronary angiography during hospitalization was left at the choice of the attending physician. EF was determined during hospitalization by echocardiography and/or left ventriculography.

Coronary artery disease was defined by the presence of documented history of acute coronary syndromes, significant CAD previously confirmed by coronary angiography (reduction of the normal diameter ≥50% for the left main coronary artery and of ≥70% for the right coronary, left anterior descending, and circumflex arteries), or history of coronary revascularization before the index admission for HF.

Prognosis

The vital status was obtained either by a consultation with the general practitioner or the referring cardiologist, or by consulting the civil registry. Ten-year follow-up was 97% complete. Cause of death was ascertained by hospital records, death certificates, and autopsy records or by contacting the patients' physician or the referring cardiologist. Records of death were analysed by two independent physicians in a blinded fashion. If reviewers did not agree, a third physician made the final decision regarding a particular event. The cause of death was classified in one of the following six categories: sudden death, HF-related death, acute myocardial infarction death, other CV death, non-cardiovascular (non-CV) death and unknown cause of death.

Sudden death was defined as an unexpected death in a patient who was previously stable and included observed arrhythmic deaths and sudden deaths not attributable to myocardial infarction or other identifiable causes. Death was considered to be HF-related when preceded by marked aggravation of HF symptoms without response to therapy. Acute myocardial infarction death was defined as death occurring after a definite acute myocardial infarction or death with autopsy evidence of recent myocardial infarction. Other CV deaths were considered deaths by CV causes that did not respond to the previous criteria (cerebrovascular events, renal and abdominal vascular events, aortic dissection and pulmonary embolism).21, 22

Statistical analysis

Continuous variables were expressed as mean ± standard deviation (SD) and were compared between groups by Student t-tests. Categorical variables were summarized by frequency per cents and analysed with chi-square tests.

The risk of all-cause mortality and cause-specific mortality was compared in patients with and without CAD. Analyses of mortality were performed by means of Cox proportional hazards models using a competing risks approach. Each cause of death was separately analysed in a Cox model while patients who died of other causes were censored (as non-events) at the time of death.22, 23 For multivariable analyses, we adopted an epidemiological approach without the use of model-building techniques. To determine the independent association of CAD with overall and cause-specific mortality, we adjusted for covariates selected a priori, considered of potential prognostic impact on the basis of clinical judgment. These covariates were: age, sex, New York Heart Association (NYHA) class, history of hypertension, diabetes mellitus, cancer, stroke, chronic obstructive pulmonary disease, atrial fibrillation at admission, serum sodium, estimated glomerular filtration rate, and admission systolic blood pressure. We tested the proportionality assumption for each variable in single-variable analyses using the test proposed by Grambsch and Therneau,24 with a P-value <0.10 indicating non proportionality. In addition, we performed a competing risk regression based on the proportional subhazards method described by Fine and Gray,25 which is implemented in STATA (version 12; StataCorp LP, College Station, TX, USA) under the stcrreg command. In secondary multivariable analyses we used the 45% cut-off to define HFPEF. All-cause survival curves were plotted using the Kaplan–Meier one-failure function and were compared using log-rank tests. Cumulative incidence curves were used to depict cause-specific mortality and corresponding P values were calculated using the proportional subhazards method by Fine and Gray. For all tests, a two tailed P-value ≤0.05 was considered statistically significant. Statistical analysis was performed with SPSS 13.0 statistical software (SPSS Inc., Chicago, IL, USA) and STATA (version 12, StataCorp LP).

The study complied with the Declaration of Helsinki and was approved by local institutional review boards. Informed consent was obtained from all patients. The database was approved by the French Computers and Privacy Commission.

Baseline characteristics

The study population comprised 591 patients (323 males and 268 females). History of CAD was present in 187 patients (32%). Of these, 155 patients (83%) had a previous documented acute coronary syndrome, 30 (16%) had previous coronary bypass grafts and 45 (24%) had previous percutaneous coronary interventions. Patients with CAD (Table 1) were more often males and more had frequently diabetes mellitus, peripheral artery disease, and signs of pulmonary congestion. On admission, EF and estimated glomerular filtration rate were significantly lower in patients with CAD than in patients without CAD (Table 1). Atrial fibrillation was recorded more often in patients without CAD. Among the 591 patients, only eight had an implantable cardioverter–defibrillator. With regard to medical treatment prescribed at discharge in patients who survived the index hospitalization (n = 559), the CAD group had higher prescription rates of angiotensin-converting enzyme inhibitors, beta-blockers, statins, and antiplatelet agents.

More than half of the study population had preserved EF (320 patients had HFPEF and 271 HFREF). Among the 187 patients with CAD, previous documented acute coronary syndromes were more frequently recorded in HFREF than in HFPEF (51% vs. 32%, P = 0.03). In contrast, the proportion of patients with previous myocardial revascularization tended to be higher in HFPEF (coronary bypass grafts: 9.6% vs. 6.4%; P = 0.20 and percutaneous coronary interventions: 14% vs. 10%, P = 0.14). In HFPEF, the precipitating causes of hospital admission for HF were: arrhythmias in 30% of patients, acute coronary syndromes in 14%, worsening renal function in 12%, uncontrolled hypertension in 10%, respiratory infections in 9%, and ‘other’ or ‘unknown’ in the remaining 25% of cases. Among patients with HFREF, the precipitating causes of hospital admission were: acute coronary syndromes (30%), arrhythmias (20%), worsening renal function (13%), respiratory infections (4%), uncontrolled hypertension (3%), and ‘other’ or ‘unknown’ (30%). Compared with HFREF, HFPEF admission to hospital appeared to be triggered more frequently by arrhythmia and less often by acute coronary syndromes (both P < 0.01). The characteristics of the two EF groups according to the presence or absence of CAD followed the characteristics of the overall population and are displayed in Table 2.

Short-term outcome

During the index hospitalization, 32 patients died (23 patients with HFREF and nine patients with HFPEF). The presence of CAD was associated with higher in-hospital mortality rate in HFREF (5.5% vs. 2.7%, P = 0.002). In HFPEF, the presence of CAD did not significantly influence in-hospital mortality (1.9% vs. 0.8%, P = 0.57). While in HFPEF, 1-year all-cause mortality was not influenced by CAD (28% for CAD vs. 21% for non-CAD, P = 0.23), in HFREF, CAD was associated with excess 1-year all-cause mortality (38% for CAD vs. 17% for non-CAD, P < 0.001). In HFPEF, on multivariable analysis, CAD did not increase the risk of 1-year all-cause mortality [adjusted HR 1.19 (0.69–2.07)] and 1-year CV mortality [adjusted HR 1.22 (0.62–2.40)]. In contrast, in HFREF CAD was independently predictive of 1-year all-cause and CV mortality [adjusted HR 2.22 (1.33–3.73) for all-cause death and adjusted HR 2.61 (1.43–4.76) for CV death]. On multivariable analysis, patients with CAD and HFREF were at higher risk of 1-year all-cause death [adjusted HR 1.74 (1.01–3.03)] and 1-year CV death [adjusted HR 1.95 (1.03–3.69)] compared with patients with CAD and HFPEF.

Overall long-term prognostic impact of CAD

During the 10-year follow-up, 459 patients (78%) died. The 10-year all-cause mortality was significantly higher in patients with CAD than in patients without CAD (87% vs. 74%, P < 0.001; Figure 2). The strong association between CAD and all-cause death was still observed on multivariable analysis (Table 3). Causes of death were CV in 269 patients, non-CV in 157 patients and unknown in 33 patients (Figure 3). The cumulative incidence of CV death was greater in patients with CAD than in non-CAD patients (Figure 4). On multivariable analysis, CAD was independently predictive of CV death (Table 3).

CAD and long-term outcome in HFPEF

During the 10-year follow-up, in HFPEF there were more events in the CAD group than in the non-CAD group but the difference did not reach statistical significance (88% vs. 78%; P = 0.19; Figure 2). On multivariable analysis CAD was not independently predictive of all-cause death (Table 3). With regard to cause-specific mortality, there were 146 CV deaths, 94 non-CV patients, and in 18 patients the cause of death remained unknown (Figure 3). The HFPEF patients with CAD had greater cumulative incidence of CV death compared with non-CAD patients (Figure 4). After adjustment for covariates, the prognostic impact of CAD was no longer observed (Table 3). Cumulative incidence curves (Figure 5) demonstrated a greater frequency of sudden death among HFPEF patients with CAD compared with non-CAD patients while the proportion of HF-related death was comparable between CAD and non-CAD patients. On multivariable analysis, CAD remained independently predictive of sudden death [adjusted HR 2.22 (1.05–4.95); Table 3).

CAD and long-term outcome in HFREF

In HFREF, 10-year event-rate was significantly greater in the CAD group than in the non-CAD group (86% vs. 67%, P < 0.001; Figure 2). On multivariable analysis, CAD was independently predictive of all-cause death (Table 3). In the HFREF group, 123 patients died of CV causes, 63 deaths were non-CV, and in 15 cases the causes of death were unknown (Figure 3). Cumulative incidence of CV death was significantly greater in patients with CAD than in patients without CAD (Figure 4). Multivariable analysis confirmed that CAD was independently predictive of CV death in HFREF (Table 3). The HFREF patients with CAD had greater cumulative incidence of HF-related death (Figure 5) and acute myocardial infarction death than patients without CAD. After adjustment for covariates of prognostic impact, CAD was still predictive of HF-related death and acute myocardial infarction death (Table 3).

Long-term prognostic impact of CAD in HFPEF vs. HFREF

Among patients with CAD, the 10-year event-rate was practically identical in the two EF groups (88% in HFPEF and 86% in HFREF; P = 0.72). On multivariable analysis, patients with CAD and HFREF were at higher risk of all-cause death than patients with CAD and HFPEF [adjusted HR 1.48 (1.05–2.07)]. The frequency of 10-year CV death was comparable in CAD patients with HFPEF and with HFREF (56% vs. 60%, P = 0.51). After adjustment for covariates, patients with CAD and HFREF had a greater risk of CV death compared with patients with CAD and HFPEF [adjusted HR 1.68 (1.11–2.54)].

Limitations

Our study did not include all patients with HF from the Somme department during 2000, as less symptomatic patients are not always admitted to hospital. However, restricting the analysis to patients hospitalized for a first episode of HF presumably resulted in a more homogeneous study population. Our patients were enrolled from community hospitals, private clinics, and the university hospital in the Somme department, which should eliminate referral bias of hospital-based studies conducted exclusively in tertiary centres. Despite the fact that physicians who examined the death records had large experience in HF research, and that in cases of disagreement all records were carefully examined by an independent third reviewer who gave the final classification, it is possible that some causes of death may have been misclassified. However, the evenly distributed proportion of unknown causes of death among HFPEF and HFREF patients is reassuring in this regard. The use of a clinical definition for CAD might have led to an underestimation of the true frequency of CAD in the study population. Although current guidelines31 request evidence of diastolic dysfunction to confirm the diagnosis of HFPEF, we did not collect indexes of diastolic function. An overwhelming proportion of patients with HFPEF in the present study had severe symptoms at the time of hospital admission and it was previously shown that such patients have almost always abnormal left ventricular end-diastolic pressures.32 Nevertheless, we cannot rule-out the possibility that a small number of patients with other causes of dyspnoea may have been misclassified as HFPEF. We acknowledge that the time-span of coronary disease and the number of coronary events and revascularizations represent different degrees of risk for HF patients. This information was not collected for our cohort of patients admitted to hospital for a first episode of HF. We could not analyse rehospitalizations for HF during follow-up, because this item was not recorded during follow-up. Finally, we acknowledge that the multiple analyses conducted in the present study might have increased the likelihood of the type I error.