Competing interests: None.
Funding: This study was funded by a grant from the French Ministry of Health.
Statistics from Altmetric.com
Heart failure (HF) is a major public health issue owing to its increasing prevalence and high expenditure related to recurrent hospitalisation. Non-cardiac comorbidities are frequent among patients with HF. Approximately 50% of elderly patients with HF have more than five non-cardiac comorbidities1 and overlapping medical therapies. Diabetes mellitus (DM) is a classic risk factor for obstructive coronary artery disease (CAD) increasing the risk of ischaemic HF.2 However, DM may also alter cardiac structure and function independently of traditional risk factors by promoting hypertrophy and fibrosis.3
The reported prevalence of DM in patients with HF ranges between 16% and 31%,1 4–8 compared to 4–6% in the general population.9 In patients with left ventricular systolic dysfunction, DM is independently associated with mortality.8 10 It has also been reported that DM is a predictor of death in patients with left ventricular dysfunction of ischaemic aetiology, but not in patients with non-ischaemic cardiomyopathy.4 6
Heart failure with preserved ejection fraction (HFPEF), defined by a left ventricular ejection fraction (EF) ⩾50%, is an increasingly prevalent form of HF with poor long-term prognosis as reported by recent studies.11 12 The prognostic implications of DM in patients with HFPEF are not clearly understood, as very few studies have investigated this growing population.13
In the present study, we prospectively followed a large cohort of consecutive patients hospitalised for a first episode of HF in 2000. This study was designed to determine the clinical features of diabetic patients with HFPEF and to assess the prognostic impact of DM on long-term survival in patients with HFPEF.
Population and inclusion criteria
The Somme is a rural and urban department of France with a population of 555 551 inhabitants according to the 1999 census. This department has a total of 11 healthcare establishments managing patients with HF: one university hospital, seven general hospitals, two private clinics and one medium-stay and long-stay unit. General practitioners, cardiologists and internal physicians of these centres have agreed to participate in this study. Consecutive patients over the age of 20 hospitalised for a first episode of HF in all these healthcare establishments during 2000 (from 1 January 2000 through 31 December 2000) were prospectively enrolled. The diagnosis of HF was made by the physician caring for the patient, based on history, symptoms, physical signs and chest x-ray on admission. Two cardiologists specifically recruited for this purpose reviewed, during the index hospitalisation, all medical records in order to validate the diagnosis of HF according to the Framingham criteria amended by the European Society of Cardiology.14 The diagnosis was not validated in 12 of the 811 enrolled patients.
Among the 799 included patients, EF was available in 662 cases (83%). As usually recommended, a cut-off value of 50% was used to distinguish HFPEF from HF with reduced EF.11 15 The 368 patients with EF ⩾50% represented the actual study population.
Clinical data including medical history, traditional cardiovascular risk factors as well as results of complementary investigations were recorded on individual case report forms. Complementary investigations included laboratory tests, ECG, chest x-ray on admission, echocardiography and, in some patients, coronary angiography.
EF was determined during hospitalisation by echocardiography (n = 368) and/or left ventriculography (n = 34). Echocardiograms were recorded according to the guidelines of the American Society of Echocardiography.16 When more than one method was performed, an average EF was calculated.
Criteria used to define DM were use of oral hypoglycaemic drugs or insulin for at least two weeks before the index hospitalisation or documented previous abnormal blood glucose levels (blood glucose ⩾126 mg/dl after an 8-hour fast or blood glucose ⩾200 mg/dl 2 hours after a standard oral glucose tolerance test).17 DM was diagnosed in 96 patients (26%) of the 368 in the study population (EF ⩾50%). For information, among the 294 patients hospitalised with HF and reduced EF during the same period of time there were 75 patients with DM (25.5%). Twenty-five patients (18.2%) out of the 137 without an EF assessment during the index hospitalisation had DM.
Clinical CAD was assumed in patients with a history of ischaemic heart disease, recent documented history of myocardial infarction or angina pectoris, or significant coronary atherosclerosis confirmed by coronary angiography before the index hospitalisation for HF.7 18 A patient was considered to have hypertension in the presence of one of the following criteria: high blood pressure during hospitalisation (>160/95 mm Hg), previous diagnosis of hypertension or normal blood pressure with ongoing antihypertensive therapy.19 Valvular heart diseases were identified on the basis of clinical history, physical examination and echocardiography. Dilated cardiomyopathy, restrictive and hypertrophic cardiomyopathy, constrictive pericarditis and other rare aetiologies of HF were diagnosed in a minority of cases.
Medical treatment records were completed at discharge. Prescription of the main therapeutic classes in HF was recorded.
One-year, three-year and five-year overall and cardiovascular mortality were determined. Sudden death was classified as cardiovascular death. The vital status was obtained either by a consultation with the general practitioner or the referring cardiologist or by consulting the civil registry. Cause of death was ascertained by hospital records, death certificates and autopsy records or by contacting the patients’ physicians. No patients were lost to follow-up at three years. Three patients (0.8%) were lost to follow-up at five years.
Continuous variables were expressed as mean (SD) and compared between groups by a Student t test. Categorical variables were summarised by frequency percentages and analysed by a χ2 test. Unadjusted survival curves were generated using Kaplan-Meier survival estimates. Differences in time to death between groups were analysed using a two-sided log-rank test. Univariate and multivariable analyses of mortality were performed using Cox proportional hazards models. Covariates considered of potential prognostic impact (age, gender, ischaemic aetiology, history of hypertension, stroke, cancer, creatinine on admission) and factors associated with five-year overall mortality on univariate analysis (p value ⩽0.10) were entered in the multivariable models. Age and creatinine were analysed as continuous variables, while gender, DM, ischaemic aetiology, history of hypertension, atrial fibrillation, stroke, chronic obstructive pulmonary disease (COPD) and cancer were analysed as dichotomous variables. A separate Cox model was constructed in patients in whom body mass index (BMI) was available (76%, n = 293). BMI was entered in the model as a continuous covariate. For all tests, a p value ⩽0.05 was considered statistically significant. Adjusted survival curves were constructed after adjustment for all covariates in the multivariable model.
Data on survival of the two groups were compared with the expected survival of the age-matched and gender-matched population of the Somme department, respectively. Control data were obtained from French life tables of the Somme department for 1999 provided by the French Institute of Statistics (INSEE) and represent the survival of the Somme general population. The relative survival was computed as the ratio of the observed to expected survival (observed number of deaths/expected number of deaths in the general population). Data were entered into a SPSS 13.0 statistical software file on a PC (SPSS Inc, Chicago, IL, USA).
The study conforms to the principles outlined in the Declaration of Helsinki and was approved by local institutional review boards. Informed consent was obtained from the patients before entering data into the electronic database. The database was approved by the CNIL (French computers and privacy commission).
Baseline characteristics and treatment
Three hundred sixty eight patients were included (mean age 76 (10) years). Fifty-three per cent of patients were women (female/male sex ratio = 1.13). DM was diagnosed in 96 patients (26%). In the diabetic group, 74 patients (77%) were treated by antidiabetic medication and 22 patients (23%) were treated by diet alone. Insulin was prescribed in 32 patients and oral hypoglycaemic agents in 42 patients (sulfonylurea derivates in 35 patients, metformin in 23 patients and other hypoglycaemic drugs in four patients). In patients with clinical CAD (n = 103, 28%), the prevalence of DM was 36% compared to 22% in patients without the diagnosis of clinical CAD.
Table 1 presents the baseline features of the study population according to diabetic status. Patients with DM were significantly younger, with a higher BMI, more cardiovascular risk factors and a higher rate of clinical CAD compared to patients without DM. Stroke tended to be more prevalent among patients with DM. The mean EF in the diabetic group was lower than in patients without DM.
Significant differences in the prescription of HF therapeutic classes were observed between the two groups (table 1). ACE-inhibitors, calcium channel blockers, statins and nitrates were prescribed at discharge significantly more frequently in patients with DM.
Prognostic impact of DM in patients with HFPEF
During the five-year follow-up, 208 patients died (43.5%). One-year, three-year and five-year survival rates in patients with DM were 77%, 50% and 32%, respectively, compared to 78%, 61% and 47% in non-diabetic patients, respectively. Compared to the expected survival of the age-matched and gender-matched general population (fig 1A), the one-year, three-year and five-year survival rates in HFPEF patients with DM were dramatically lower (77% vs 96%, 50% vs 88%, and 32% vs 79%, respectively). One-year, three-year and five-year relative survival rates (observed/expected survival) of patients with DM were also dramatically lower than those of the non-diabetic group (80% vs 84%; 57% vs 75% and 41% vs 68%, respectively; fig 1B).
DM was associated with a poorer prognosis on Cox univariate analysis (hazard ratio (HR) 1.43, 95% confidence interval (CI) 1.07 to 1.92, p = 0.016). The association was stronger after adjustment for age and gender (HR 1.92, 95% CI 1.41 to 2.61, p<0.001) and after adjustment for age, gender, ischaemic aetiology, history of hypertension, atrial fibrillation, stroke, COPD, cancer and creatinine on admission (table 2 and fig 2A). Older age, stroke, COPD, cancer and higher blood creatinine on admission were also independent predictors of five-year overall mortality in patients with HFPEF (table 2). Lower BMI was not associated with increased mortality on multivariable analysis (table 2).
We further investigated the prognostic impact of DM in patients with and without clinical CAD. On multivariable analysis, after adjustment for covariates of potential prognostic impact, DM was associated with an increased risk of death in patients without clinical CAD (HR 1.85, 95% CI 1.22 to 2.82, p = 0.004; fig 2B) as well as in patients with clinical CAD (HR 1.82, 95% CI 1.02 to 3.25, p = 0.04; fig 2C).
In patients with DM, cardiovascular causes were responsible for more than 60% of deaths. Cardiovascular mortality rates in patients with DM were significantly higher than in non-diabetic patients (17% vs 13% at one year, 29% vs 22% at three years and 44% vs 29% at five years; p = 0.014).
The present study shows that DM is a frequent comorbidity in patients with HFPEF with a powerful prognostic impact. DM was associated with an increased risk of death in our study population with a low five-year survival of 32% and a 60% excess mortality compared to the expected survival. In diabetic patients the cause of death was mostly cardiovascular. The adverse prognostic impact of DM was observed in both patients with and without clinical CAD. ACE inhibitors, calcium channel blockers, statins and nitrates were prescribed more frequently in patients with DM.
Diabetes and HF
DM is recognised as an independent risk factor for the development of HF.2 20 It accelerates coronary atherosclerosis and increases the risk of myocardial infarction and ischaemic cardiomyopathy. Independently of obstructive CAD, DM may induce a specific cardiomyopathy characterised by myocardial hypertrophy, fibrosis and microangiopathy.3 In a recent report, HF was associated with the development of new-onset DM.21
The reported prevalence of DM in patients with a diagnosis of HF ranges between 16% and 31%.1 4–8 This information is mainly derived from selected HF populations4–6 8 and a few population-based studies.1 7 HFPEF is an increasingly prevalent form of HF, representing almost 50% of all cases.11 Until recently, it was accepted that HFPEF had a much better prognosis than systolic HF. Two large observational studies have discredited this notion, as they independently reported similar mortality rates in patients with preserved and reduced systolic function.11 12 Moreover, it was observed that the prevalence of DM has significantly increased over time among patients with HF.7 11 A better understanding of comorbid conditions associated with HFPEF appears to be of particular importance.
To our knowledge, the prevalence of DM in patients with HFPEF has not been previously reported. The prevalence of DM in our study was 26%. As expected, the prevalence of DM was higher in our patients with clinical CAD (36% vs 22%).
In our study, patients with DM were younger but presented more risk factors and a higher prevalence of clinical CAD compared with patients without DM. The prevalence of CAD in diabetic patients in our study is lower compared with other reports.7 However this issue was poorly studied in patients with HFPEF. In a study including selected patients with systolic HF, 42% patients with DM had no angiographic evidence of CAD.4 Data on the prevalence of HF with normal coronary angiogram in patients with HFPEF are lacking to our knowledge.
ACE inhibitors, calcium channel blockers and statins were more frequently prescribed in patients with DM. ACE inhibitors remain underused in this group of patients at high cardiovascular risk. Despite evidence of the beneficial effect of β-blockers in diabetic patients with HF,22 the prescription rate of 32% in our study is disappointingly low.
Prognostic impact of DM
The impact of DM on long-term mortality has been analysed in several clinical trials conducted in selected populations with systolic HF and in population-based studies with discordant results. DM was a predictor of increased mortality in the retrospective analyses of the Studies of Left Ventricular Dysfunction,10 β-Blocker Evaluation of Survival trial8 and Danish Investigations of Arrhythmia and Mortality of Dofetilide in Congestive Heart Failure5 databases. This finding was not confirmed by other clinical trial reports.23 24 Population-based studies have also reported contradictory findings.4 7 25 The explanation, although still unclear, may be related to differences in study populations, inclusion criteria and variable prevalence and diagnostic criteria for CAD. Recently, Berry et al reported that in patients with HF, DM and abnormal glucose tolerance were both associated with increased in-hospital and long-term mortality and had comparable predictive power for these outcomes.26
Very few population-based studies have investigated the prognostic impact of DM in “real-life” patients with HFPEF.13 In our study, after adjustment for covariates of prognostic importance, DM was responsible for an increased mortality risk over a five-year follow-up. An impressive excess mortality was observed in patients with DM compared to the expected survival of the general population.
It has been previously suggested that the impact of DM on the prognosis of patients with systolic HF is modulated by the presence or absence of CAD. In patients with left ventricular systolic dysfunction of ischaemic aetiology, DM was found to be a powerful predictor of death, but no association was observed between DM and mortality in patients with non-ischaemic cardiomyopathy.4 6 8 In a recent study including patients with preserved and reduced EF, DM was associated with mortality in both groups.13 In this study, DM was not associated with an increased risk of death in patients with CAD. In a retrospective community study including patients with new-onset HF, the risk of death related to DM was highest in patients with no evidence of clinical CAD.7
In our study, DM had a comparable pejorative prognostic impact in patients with clinical CAD and patients without clinical CAD. Among diabetic patients without a clinical diagnosis of CAD some might have asymptomatic CAD or silent myocardial ischaemia. It has been reported that the outcome of diabetic individuals without any cardiovascular disease is comparable with that of non-diabetic patients with confirmed cardiovascular disease,27 and severe coronary atherosclerosis is frequent in diabetic individuals without a clinical diagnosis of CAD.28 Autopsy data from a cohort of elderly residents of Rochester, Minnesota (mean age 75 (11) years) show a high prevalence (⩾40%) of high-grade coronary atherosclerosis among elderly non-diabetic individuals without an antemortem clinical diagnosis of CAD.28 Our data support the need for aggressive management of diabetic patients even in the absence of diagnosed CAD. Efforts must be pursued in this direction, as ACE inhibitors, angiotensin-receptor blockers, β-blockers and statins are still largely underused in diabetic patients with HF.
This study was exclusively hospital-based, with the advantage of allowing prospective recruitment of all individuals hospitalised for a first episode of HF. Patients with HF treated exclusively on an outpatient basis or in a long-term care facility were therefore not included. Our patients were enrolled from all establishments hospitalising patients with HF—community hospitals, private clinics and one teaching hospital. This recruitment should reduce referral bias of hospital-based studies performed exclusively in tertiary centres. Echocardiography was not systematically performed during the first hours after the patients’ arrival, as in all epidemiological studies of this type. However, EF was estimated in 83% of patients, an acceptable figure compared to most studies.12 18 As coronary angiography was not systematically performed in this elderly population with incident HF, the diagnosis of CAD was based on clinical data, as in most epidemiological studies.18 Non-invasive evaluation of reversible myocardial ischaemia was not performed and troponin or creatine kinase isoenzyme levels on admission were not available in our study population. The proportion of 40.9% of diabetic patients receiving aspirin in our cohort is low. The low prescription rate of aspirin may in part be explained by the use of anticoagulants for atrial fibrillation in 30% of diabetic patients because in clinical practice, physicians are often reluctant to associate oral anticoagulants and aspirin in elderly diabetic patients with atrial fibrillation.
Despite the high prevalence and burden of HFPEF, current treatment guidelines are still empirical. Ongoing clinical trials will probably define a better therapeutic approach. In patients with HFPEF, DM is a frequent comorbidity with adverse prognostic impact. DM is associated with a poor prognosis in HFPEF patients with a clinical diagnosis of CAD as well as in patients without clinical CAD. Further studies are needed to establish the role of lifesaving therapies in diabetic patients with HFPEF.
Competing interests: None.
Funding: This study was funded by a grant from the French Ministry of Health.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.