Impact of diabetes mellitus on mortality in patients with acute heart failure: a prospective cohort study

Background Although more than one-third of the patients with acute heart failure (AHF) have diabetes mellitus (DM), it is unclear if DM has an adverse impact on clinical outcomes. This study compared the outcomes in patients hospitalized for AHF stratified by DM and left ventricular ejection fraction (LVEF). Methods The Korean Acute Heart Failure registry prospectively enrolled and followed 5625 patients from March 2011 to February 2019. The primary endpoints were in-hospital and overall all-cause mortality. We evaluated the impact of DM on these endpoints according to HF subtypes and glycemic control. Results During a median follow-up of 3.5 years, there were 235 (4.4%) in-hospital mortalities and 2500 (46.3%) overall mortalities. DM was significantly associated with increased overall mortality after adjusting for potential confounders (adjusted hazard ratio [HR] 1.11, 95% confidence interval [CI] 1.03–1.22). In the subgroup analysis, DM was associated with higher a risk of overall mortality in heart failure with reduced ejection fraction (HFrEF) only (adjusted HR 1.14, 95% CI 1.02–1.27). Inadequate glycemic control (HbA1c ≥ 7.0% within 1 year after discharge) was significantly associated with a higher risk of overall mortality compared with adequate glycemic control (HbA1c < 7.0%) (44.0% vs. 36.8%, log-rank p = 0.016). Conclusions DM is associated with a higher risk of overall mortality in AHF, especially HFrEF. Well-controlled diabetes (HbA1c < 7.0%) is associated with a lower risk of overall mortality compared to uncontrolled diabetes. Trial registration ClinicalTrial.gov, NCT01389843. Registered July 6, 2011. https://clinicaltrials.gov/ct2/show/NCT01389843

75% patients with HF have ≥ 1 comorbidity, and these comorbidities make overall clinical outcomes worse [6]. In a recent meta-analysis, patients with diabetes mellitus (DM) were suggested to have a two-fold increase in the risk of HF [7]. DM is present in ~ 35% patients hospitalized with acute HF [8]. Multiple factors such as ischemia, hypertension, and extracellular fluid volume expansion are involved in the pathogenesis of HF in DM [9,10]. While DM is associated with an increased cardiovascular morbidity and mortality in patients with chronic HF with reduced left ventricular ejection fraction (HFrEF) [11,12], its independent impact on in-hospital and long-term outcomes after HF hospitalization is unclear. Data from some large registries and clinical trials suggest that DM is associated with worse in-hospital and post-discharge outcomes in patients with acute HF [13][14][15][16][17][18]. Other studies do not suggest a significant association of DM with mortality in patients hospitalized for HF after adjusting for confounding factors [19][20][21][22]. Thus, the independent association of DM with mortality in patients with HF remains unknown. It is also unclear if DM has similar adverse impact across HF subtypes such as HFrEF, HF with preserved ejection fraction (HFpEF), or HF with mid-range ejection fraction (HFmrEF).
We compared acute HF-associated in-hospital and overall all-cause mortality in patients with and without DM using the Korean Acute Heart Failure Registry (KorAHF) [23]. We also compared the outcomes in each HF subtype.

Study population
We evaluated the patients with acute HF enrolled in the KorAHF registry (ClinicalTrial.gov identifier, NCT01389843) [23]. Briefly, the KorAHF registry is a prospective multicenter cohort study of 5625 patients admitted for acute heart failure (AHF) in 10 tertiary university hospitals between March 2011 and February 2014 who have been followed for > 5 years until February 2019. Patients who had signs or symptoms of HF and met ≥ 1 of the following criteria were enrolled in this registry: (1) lung congestion or (2) objective evidence of left ventricular (LV) systolic dysfunction or (3) structural heart disease.
We excluded 210 patients where there was no information on LV ejection fraction (LVEF) and 21 patients who were lost to follow-up. Finally, 5394 patients with AHF and known DM status and LVEF were enrolled for analyses ( Fig. 1).

Data collection and outcome definition
Data were collected at each hospital and entered into a web-based Clinical Research and Trial (iCReaT) system case-report form of the Korea National Institute of Health. Detailed information was collected at the time of admission, and follow-up data were collected from the patients by the attending physician at 30 days and 3, 6, 12, 24, 36, 48, and 60 months after discharge. Data on patient demographics, medical history, physical signs, laboratory test results, electrocardiography, echocardiography, medications, and outcomes were collected. The mortality Fig. 1 Flow chart of the study. KorAHF registry, Korean Acute Heart Failure registry data for patients lost to follow-up was collected from the National Insurance data or National Death Records.

Definition of DM and glycemic control
DM was defined as self-reported, history of anti-hyperglycemic agent use, or newly diagnosed during hospitalization [17]. Newly diagnosed DM was defined as a glycated hemoglobin (HbA1c) ≥ 6.5% when measured after a random glucose level ≥ 200 mg/dl at enrollment. We additionally classified DM patients based on HbA1c levels measured at the follow-up visit ≤ 1 year from discharge. We defined well-controlled and uncontrolled DM by an HbA1c < 7.0% and ≥ 7.0% at the follow-up visit, respectively. According to LVEF, we categorized patients with AHF into 3 groups: LVEF < 40% (HFrEF), 40% ≤ LVEF < 50% (HFmrEF), and LVEF ≥ 50% (HFpEF).

Statistical analysis
Baseline characteristics as per DM status were compared using the χ 2 test for categorical variables and the unpaired Student's t-test for continuous variables. Kaplan-Meier survival curves as per DM status were compared using the log-rank test. We used the multivariable Cox proportional hazard regression model to evaluate the association between DM and mortality in patients with AHF. Potential confounders which were different at baseline in patients with and without DM, or were considered clinically significant including age, sex, body mass index (BMI), etiology of HF (ischemic or non-ischemic), prior admission for HF, use of parenteral inotropic agents, serum creatinine concentration (< 2.0 or ≥ 2.0 mg/dL), elevated brain natriuretic peptides (BNP) (≥ 500 pg/ mL) or N-terminal pro-brain natriuretic peptides (NT-proBNP) (≥ 1000 pg/mL), New York Heart Association (NYHA) class (III-IV or I-II) on admission, and smoking status (current or ex-smoker vs. never-smoker) were adjusted for in the multivariable model. An interaction between DM and potential confounders was assessed by adding interaction terms in the Cox proportional hazard regression model. All p-values were two-sided, and p-values < 0.05 were considered statistically significant. Statistical analyses were performed using SAS software version 9.4 (SAS Institute Inc., Cary, NC, USA) and R version 3.6.0 with packages ("survival", and "survminer").

Baseline characteristics
In the study population, 2321 patients with AHF had DM (43.0%) ( Table 1). Patients with DM had a higher prevalence of risk factors like old age, obesity, hypertension, ischemic heart disease, chronic kidney disease, and cerebrovascular disease. Patients with DM had a higher proportion of patients with a BNP ≥ 500 pg/mL or NT-proBNP ≥ 1000 pg/mL, NYHA class III-IV on admission, acute pulmonary edema on chest X-ray, a higher level of systolic blood pressure, C-reactive protein, serum potassium and creatinine concentration, and lower serum sodium concentration and LVEF compared to those without DM. Besides, patients with DM were more likely to be on parenteral diuretics, inotropic agents, and vasodilators. However, aldosterone antagonists were prescribed less frequently in patients with DM.

In-hospital and overall mortality as per DM status
During a median follow-up of 3.5 years, there were 235 (4.4%) deaths during the index hospitalization, and 2500 (46.3%) deaths during the overall follow-up period. Patients with DM had a higher incidence of in-hospital mortality and overall mortality compared to patients without DM (Fig. 2). After adjusting for potential confounders including age, sex, BMI, etiology of heart failure (ischemic vs. non-ischemic), prior admission for HF, parenteral inotropic use, serum creatinine concentration, elevated BNP/NT-proBNP, NYHA class III-IV on admission, and smoking status, DM was still independently associated with overall mortality (adjusted hazard rate [HR] 1.11, 95% confidence interval [CI] 1.03-1.22).

Independent predictors of in-hospital and overall mortality
Results of multivariable Cox proportional hazard regression for in-hospital and overall all-cause mortality are reported in Table 2. DM was not independently associated with an increased in-hospital mortality (HR 0.81, 95% CI 0.61-1.07, p = 0.137). Use of parenteral inotropes, age, ischemic etiology, and a higher serum creatinine concentration also independently predicted in-hospital mortality. DM was an independent predictor for overall mortality (HR 1.11, 95% CI 1.03-1.22, p = 0.013). Other variables, such as old age, male sex, higher BMI, ischemic etiology, acute decompensated HF, use of parenteral inotropes, high concentrations of serum creatinine and BNP/NT-proBNP during index hospitalization, and NYHA class Values are presented as mean ± standard deviation, or n (%) DM diabetes mellitus, SBP systolic blood pressure, DBP diastolic blood pressure, BNP brain natriuretic peptides, NT-proBNP N-terminal pro-brain natriuretic peptides, hsCRP high sensitivity C-reactive protein, CRP C-reactive protein, BUN blood urea nitrogen, LVEDD left ventricular end-diastolic dimension, LVEDV left ventricular enddiastolic volume, LVEF left ventricular ejection fraction, LA left atrium, RVSP right ventricular systolic pressure, ACEIs angiotensin converting enzyme inhibitors, ARBs angiotensin receptor blockers, AAs aldosterone antagonists III-IV on admission also independently predicted higher overall mortality.  Table 3). DM had differential impact on overall mortality as per the HF subtype. In HFrEF, DM was significantly associated with an increased risk of overall mortality after adjusting for potential confounders (adjusted HR 1.14, 95% CI 1.02-1.27). However, DM was not significantly associated with overall mortality in patients with HFm-rEF (adjusted HR 0.99, 95% CI 0.80-1.22) and HFpEF (adjusted HR 1.13, 95% CI 0.96-1.34) ( Table 3). The Kaplan-Meier analysis also revealed significantly worse overall mortality in patients with HFrEF and DM vs. HFrEF and no DM (40.2% vs. 52.7%, log-rank p < 0.001) (Fig. 3).  Table 2 Independent predictors of in-hospital and overall mortality on multivariable Cox proportional hazard regression model a Adjusted for age, sex, body mass index, etiology of heart failure (ischemic vs. non-ischemic), prior admission history due to HF, parenteral inotropics usage, creatinine concentration (< 2.0 vs. ≥ 2.0 mg/dL), elevated BNP (≥ 500) or NT-proBNP (≥ 1000), NYHA class (III-IV or I-II) on admission, and smoking status (current or ex-smoker vs. never-smoker)

Variables
Adjusted HR a P value  Overall mortality as per the prespecified subgroup and glycemic control Figure 4 shows the association between DM and overall mortality in a stratified group as per the potential confounders, including age, sex, ischemic etiology, hypertension, chronic kidney disease, de novo HF, LVEF < 40%, and smoking status. The impact of DM on overall mortality was generally consistent across stratified subgroups (p-interaction ≥ 0.05). However, there was a significant difference in the impact of DM on overall mortality between smoker (current or ex-smoker) and never-smoker (p for interaction = 0.022). Figure 5 shows that patients with uncontrolled DM (HbA1c ≥ 7.0%) had significantly higher overall mortality compared to patients with well-controlled DM (HbA1c < 7.0%) by Kaplan-Meier analysis (44.0% vs. 36.8%, log-rank p = 0.016).

Discussion
The main findings of our study are as follows: (1) patients with AHF and DM have a significantly higher in-hospital and overall mortality vs. patients with AHF and no DM; (2) DM was significantly associated with a higher overall mortality even after adjusting for potential confounding factors including age, sex, BMI, HF etiology, renal function, and HF severity; (3) DM had a significant association with higher overall mortality in HFrEF, but not HFmrEF and HFpEF; (4) patients with poor glycemic control after discharge (HbA1c ≥ 7.0%) had a higher overall mortality vs. patients with adequate glycemic control (HbA1c < 7.0%).
Previous studies in HF have compared the clinical characteristics and outcomes in patients with and without DM. However, there are few reports comparing clinical outcomes stratified by DM in HFpEF [12,24,25]. Moreover, there is no data from a large registry or clinical trials in patients with HFmrEF. The CHARM program demonstrated that DM was significantly associated with a higher mortality and morbidity in HFrEF and HFpEF [11]. Another large-scale study from the I-PRESERVE trial (Irbesartan in Heart Failure with Preserved Ejection Fraction) showed that patients with DM had more significant structural and functional echocardiographic abnormalities and worse clinical outcomes compared to patients without DM in HFpEF [24]. A recent prospective HFpEF study showed a significant association of DM with longterm mortality in women, but not in men [25]. Similar to HFrEF, these studies demonstrate a significant associations of DM with higher mortality in HFpEF. The mechanisms for poor prognosis of HF with DM are unclear.
Some of these mechanisms are: (1) DM causes microangiopathy, myocardial fibrosis, and autonomic neuropathy and these lead to diabetic cardiomyopathy [26]; (2) hyperglycemia leads to lipid accumulation in the heart, and this can cause cellular damage by lipotoxicity [27]; (3) lipid accumulation, collagen deposition and fibrosis, and hyperinsulinemia due to insulin resistance increases risk of hypertrophy of the heart [28,29]; (4) DM may promote extracellular matrix expansion which is associated with a higher mortality in HF [30]; (5) impaired branched-chain amino acids catabolism and insulin signaling are associated with HF [31]; (6) distinct pathways related to inflammation, protein phosphorylation, and neutrophil degranulation are associated with DM in HF [32].
Why DM was not associated with an increased mortality in HFpEF and HFmrEF is unclear. The LVEF cutoff to classify HF in previous studies was different from the current updated guidelines for the diagnosis and treatment of HF that are accepted and used in clinical practice [33]. The CHARM program did not provide detailed echocardiographic data. The I-PRESERVE trial used an LVEF cutoff of 45%, and echocardiographic data were shown for < 20% of the whole study population. Our results require cautious interpretation. In general, patients with DM had a higher overall mortality, but this association was not statistically significant in HFpEF and HFmrEF after adjusting for risk factors such as old age, ischemic etiology, and severity of initial presentation.
Patients with HFmrEF have similar clinical characteristics as patients with HFpEF [34][35][36][37]. Recent studies demonstrate that mortality rates in HFmrEF are similar to those in HFpEF [35][36][37][38]. Although there are no studies on the association of DM with mortality in patients with HFmrEF, our study shows that this association is different Fig. 4 Overall all-cause mortality as per the prespecified subgroup from HFrEF and HFpEF. If HFmrEF is a distinct clinical syndrome or if these patients are in-transition between HFrEF and HFpEF is unknown [39]. Since there were limited patients with HFmrEF in our study, this association needs to be further explored.
Our study has important implications. First, we analyzed one of the largest prospective cohorts comparing the characteristics and clinical outcomes in patients with AHF, with and without DM. Second, our study analyzed baseline echocardiographic findings in all patients, which is unique and challenging to obtain in large HF registries. Third, we evaluated both in-hospital and overall all-cause mortality. This helped estimate both short and long-term impact of DM on mortality in patients with AHF. Fourth, we compared mortality in 3 subtypes of HF based on LVEF. To our best knowledge, this is the first study to evaluate the association of DM with mortality in HFrEF, HFmrEF, and HFpEF. Since the characteristics and prognosis of patients with HFpEF and HFmrEF are unknown, these results may help understand the clinical implications of HFpEF and HFmrEF. Lastly, we also demonstrate that an adequate glycemic control during follow-up was associated with an improved long-term prognosis in patients with AHF and DM.

Limitations
There are several limitations of our study. First, this is an observational study. To evaluate the effect of glycemic control, it has intrinsic limitations of non-randomized comparisons such as the different distribution of other clinical risk factors and the possibility of unmeasured confounding factors. Second, our endpoint was only all-cause mortality. Detailed clinical outcomes such as cardiovascular death and re-hospitalization for HF may help better understand the impact of DM on outcomes in AHF. Third, many recent studies have evaluated the cardiovascular safety of anti-diabetic medications. While dipeptidyl peptidase 4 (DPP-4) inhibitors had a neutral effect, sodium-glucose co-transporter 2 (SGLT2) inhibitors were most favorable among all classes of anti-diabetic medications for reducing the risk of HF [40,41]. However, SGLT2 inhibitors were not prescribed during the enrollment period of our registry. Therefore, our study could not evaluate their effect on HF.

Conclusions
Our study, using large registry data with echocardiographic information from all participants, shows that DM is significantly associated with an increased risk of overall mortality in AHF, especially HFrEF. Overall all-cause mortality as per glycemic control after discharge in patients with DM. *Well-controlled DM as an HbA1c < 7.0% at a follow-up visit within 1 year after discharge; Uncontrolled DM as an HbA1c ≥ 7.0% Well-controlled diabetes (HbA1c < 7.0%) was associated with a lower risk of overall mortality compared with uncontrolled diabetes (HbA1c ≥ 7.0%) in patients with AHF and DM.