Admission blood glucose and 10-year mortality among patients with or without pre-existing diabetes mellitus hospitalized with heart failure

Baseline and admission characteristics of patients were extracted from the Heart Failure Survey in Israel (HFSIS) 2003 database as previously described [17]. Briefly, this survey, conducted through March and April 2003, included 4102 patients with a diagnosis of either acute de-novo HF, acute exacerbation of chronic HF, or chronic HF, who were admitted to one of the 25 public hospitals operating in Israel. The criteria used for the diagnosis of HF were symptoms of HF (at rest or during exertion) and objective evidence of cardiac dysfunction at rest [18]. Diagnosis of acute HF or exacerbation of chronic HF was determined by the attending physician based on history, symptoms and physical examination, response to HF therapy, chest radiography, echocardiography, radionuclide studies, cardiac catheterization findings and in-hospital course. Detailed data regarding patient characteristics, in-hospital course and management, pre-hospital and discharge medications and diagnoses were collected and recorded by physicians on pre-specified structured forms. The study was conducted in accordance with the International Conference on Harmonization Guidelines for Good Clinical Practice and the Declaration of Helsinki. The ethics committee at each of the participating hospitals approved the protocol.

We have excluded patients with ABG levels below 70 mg/dL (3.9 mmol/L) or above 600 mg/dL (33 mmol/L) as previously done [7] to ensure a homogenous study population and avoid including patients with possible errors, patients with hypoglycemia or severe hyperosmolar hyperglycemia. Thus, our final analysis included 3993 patients of whom 2182 presented with no pre-existing DM and 1811 presented as patients with DM based on either a history of DM obtained from medical records or the use of anti-diabetic agents on admission. Patients were then sub-divided based on their ABG level similar to previously described thresholds [7, 14, 19]: glucose <110 mg/dL (6.1 mmol/L), 110–140 mg/dL (6.1–7.8 mmol/L), 140–200 md/dL (7.8–11.1 mmol/L) and above 200 mg/dL (11.1 mmol/L).

ABG levels were determined as the first blood glucose test on the day of admission as reported in the medical records. Left ventricular ejection fraction (LVEF) was determined by echocardiography. The LVEF data were collected by medical chart review and recorded only if echocardiography was performed within 1 year prior to or during hospitalization. Out of the 3993 study patients, 2775 had LVEF data. Echocardiography was performed in 1536 patients during index hospitalization. LVEF classes were classified as follows: normal—≥50%; mildly impaired—40 to 49%; moderately impaired—30 to 39%; severely impaired ≤30%. The median (interquartile range [Q1–Q3]) timing of echocardiography was 0 month (0–6 months) prior to index admission.

The end point of this study was all-cause mortality. During the 10 years after index hospitalization, mortality was assessed for all patients by matching their identification numbers with the Israeli National Population Registry.

Continuous variables were expressed as median (interquartile range). Baseline characteristics of the groups (continuous data) were compared by use of the non-parametric Kruskal–Wallis test; categorical variables and frequencies were compared by means of the χ² test. Kaplan–Meier survival curves were produced and compared using the log-rank test. To examine the relationship between ABG level and mortality several models were conducted. First, potential variables (identified in previously published studies [20, 21] as risk factors for mortality) were evaluated by univariate analysis and were selected based on clinical and statistical significance. The values used to dichotomize the variables were taken from previous reports [20, 21]. Second, multi-variable Cox proportional hazards regression analysis was used to assess the independent association between ABG levels and the risk of all-cause mortality. Covariates included in the multivariate models were identified using a best subset procedure, choosing among a wide variety of available baseline measures with the additional stipulation that they needed to be statistically significant with an individual p value <0.05 for inclusion. Thus, all models were adjusted for ABG (analyzed as a categorical variable and as a continuous variable in separate models), age, chronic obstructive pulmonary disease, active malignancy, urea, creatinine, NYHA functional class III and IV versus I and II, acute de novo or acute exacerbation of chronic HF, systolic blood pressure <115 or ≥115 mmHg, [21] hemoglobin, and the following pre-hospital medication use: statins, β-blockers and furosemide. Gender was also forced into the model. In the subgroup of patients with DM we have repeated the multivariable analysis including also oral hypoglycemic drugs and insulin as covariates. Finally, the same multivariable models for 10-year mortality were adjusted for all of the above parameters plus LVEF class for 2775 patients with LVEF data. All statistical tests were two-sided and a p value <0.05 was considered statistically significant. Analyses were carried out with SAS software version 9.4 (SAS Institute, Cary, North Carolina).

We identified 2182 patients with no pre-existing DM and 1811 with DM. Patients were sub-divided based on their ABG levels similar to previously described thresholds (Table 1) [7, 14, 19]. The mean follow-up duration was 4.2 years. Patients with DM were younger and more likely to have a history of hypertension, dyslipidemia, ischemic heart disease, stroke and PVD as compared with patients with no pre-existing DM. Patients with no pre-existing DM were more likely to have a history of atrial fibrillation. As for chronic medication use reported on admission—aspirin, statins, beta-blockers, angiotensin converting enzyme inhibitors (ACEIs) or angiotensin receptor blockers (ARBs), furosemide and thiazides were more frequently used in the DM group. Out of 1811 DM patients, 375 patients were treated with insulin and 1013 were treated with oral hypoglycemic medications. The median ABG level was 114 mg/dL (interquartile range 98–140) and 182 mg/dL (interquartile range 134–255) for the subgroups of patients with no pre-existing DM and those with known DM, respectively (p < 0.001). Hemoglobin and GFR levels were lower in the DM group.

Of note, in the subgroup of patients with no pre-existing DM, elevated ABG levels were associated with higher proportion of females, increased rates of myocardial infarction at presentation, poorer renal function, lower sodium levels and faster heart rates on admission.

The cumulative probabilities of death over 10 years of follow-up in patients with DM versus patients with no pre-existing DM are presented in Fig. 1. By multivariate analysis, patients with DM had a 1.3-fold increased risk of mortality (95% confidence interval 1.22–1.41, p < 0.001).

Among patients with no pre-existing DM the 10-year mortality risk correlated with ABG levels (Fig. 2). Thus, at 10 years of follow-up the cumulative probability of death was 76% for patients with ABG <110 mg/dL, 78% for patients with ABG 110–140 mg/dL, 80% for patients with 140–200 mg/dL and 90% for patients with ABG ≥200 mg/dL (p < 0.001 for the overall comparison).

Consistently, multivariate analysis (Table 2) demonstrated that high ABG levels were associated with an increment risk of death. Patients with ABG levels of 110–140, 140–200 and ≥200 mg/dL had a respective 9% (p = 0.140), 16% (p = 0.031), and 53% (p < 0.001) increased mortality risk compared to patients with ABG <110 mg/dL. Also, the results were consistent when further adjusted for LVEF class (among patients who had LVEF data). In a separate multivariable model (Table 2), each 18-mg/dL (1-mmol/L) increase in glucose level was associated with a 5% increased risk of mortality (p < 0.001); notably, ABG measured as a continuous variable remained an independent predictor of mortality also after excluding patients with ABG levels >200 mg/dL (HR = 1.04 per 18 mg/dL increase in ABG, p = 0.014). In a secondary analysis adjusting also for IHD, MI at presentation, antiplatelet agents and thiazides, results remained the same. In addition, results remained the same when patients with acute MI at presentation were excluded from the analysis. There was no significant interaction between ABG and acute or chronic HF status (ABG-by-acute HF p for interaction = 0.837) and there was no significant interaction between ABG and age (ABG-by-age ≥75 p value for interaction = 0.277), implying that the relationship between high ABG levels and mortality does not change according to acute or chronic HF status or in the different age groups.

Among patients with known DM, Kaplan–Meier survival curves showed that there were no significant differences in mortality risk between the ABG subgroups (p = 0.138, Fig. 3). Likewise, multivariable analysis (Table 3) demonstrated that there were no significant differences in mortality risk when patients with ABG levels 110–140 mg/dL (HR = 0.91, p = 0.35) or ABG 140–200 mg/dL (HR = 1.08, p = 0.38) were compared to patients with ABG <110 mg/dL. An exception was noted in the ABG >200 mg/dL subgroup experiencing a 1.2-fold increased risk of death (95% CI 1.01–1.42, p = 0.032) compared to patients with ABG <110 mg/dL. When adding oral hypoglycemic medications and insulin to the multivariable analyses, results were consistent showing that among patients with DM only those with ABG >200 mg/dL were associated with increased mortality risk (>200 mg/dL versus <110 mg/dL HR = 1.20, p = 0.04). Adjusted analysis of continuous glucose showed that each 18-mg/dL (1-mmol/L) increase in glucose level was associated with a 2% increased risk of mortality (p < 0.001); this result was mainly driven by the effect of ABG >200 mg/dL as this relationship was statistically insignificant when patients with ABG >200 mg/dL were excluded from the analysis (p = 0.283).