ESC HEART FAILURE ORIGINAL RESEARCH ARTICLE ESC Heart Failure 2020; 7: 3320–3328 Published online 13 August 2020 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/ehf2.12944 Effect of insulin on readmission for heart failure following a hospitalization for acute heart failure Vicente Bertomeu‐Gonzalez1,2,3††, Lorenzo Fácila4††, Patricia Palau5,6, Gema Miñana5, Gonzalo Núñez5, Rafael de la Espriella5, Enrique Santas5, Eduardo Núñez5, Vicent Bodí5,6, Francisco Javier Chorro5,6,3, Alberto Cordero1,3, Juan Sanchis5,6,3, Josep Lupón7,8,3, Antoni Bayés‐Genís7,8,3 and Julio Núñez5,6,3* 1 Cardiology Service, Hospital Universitario San Juan de Alicante, Alicante, Spain; 2Clinical Medicine Department, Universidad Miguel Hernández, Alicante, Spain; 3Centro de Investigación Biomédica en Red Enfermedades Cardiovasculares (CIBERCV), Madrid, Spain; 4Cardiology Service, Hospital General Universitario de Valencia, Valencia, Spain; 5 Cardiology Service, Hospital Clínico Universitario, INCLIVA, Valencia, Spain; 6Department of Medicine, Universitat de València, Valencia, Spain; 7Heart Failure Unit, Cardiology Service, Hospital Universitari Germans Trias i Pujol, Barcelona, Spain; 8Department of Medicine, Autonomous University of Barcelona, Barcelona, Spain Abstract Aims Type 2 diabetes mellitus (T2DM) is common in patients with heart failure (HF) and is related with worse outcomes. Insulin treatment is associated with sodium and water retention, weight gain, and hypoglycaemia—all pathophysiological mechanisms related to HF decompensation. This study aimed to evaluate the association between insulin treatment and the risk of 1 year readmission for HF in patients discharged for acute HF. Methods and results We prospectively included 2895 consecutive patients discharged after an episode of acute HF in a single tertiary hospital. Multivariable Cox regression, adapted for competing events, was used to assess the association between insulin treatment and 1 year readmission for HF in patients discharged after acute HF. Participants’ mean age was 73.4 ± 11.2 years, 50.8% were women, 44.7% had T2DM [including 527 (18.2%) on insulin therapy], and 52.7% had preserved ejection fraction. At 1 year follow‐up, 518 (17.9%) patients had died and 693 (23.9%) were readmitted for HF. The crude risk of readmission for HF was higher in patients on insulin, with no differences in 1 year mortality. After multivariable adjustment, patients on insulin were at significantly higher risk of 1 year readmission for HF than patients with diabetes who were not on insulin (hazard ratio 1.28; 95% confidence interval 1.04–1.59, P = 0.022) and patients without diabetes (hazard ratio 1.26; 95% confidence interval 1.02–1.55, P = 0.035). Conclusion Following acute HF, patients with T2DM on insulin therapy are at increased risk of readmission for HF. Further studies unravelling the mechanisms behind this association are warranted. Keywords Insulin therapy; Type 2 diabetes mellitus; Acute heart failure; Hospital readmission Received: 23 April 2020; Revised: 18 June 2020; Accepted: 19 July 2020 *Correspondence to: Julio Núñez, Servicio de Cardiología, Hospital Clínico Universitario, INCLIVA, Valencia, Spain. Tel: +34 652 856 689. Email: yulnunez@gmail.com † These authors contributed equally to this work. Introduction moreover, it is associated with a higher risk of adverse outcomes.5 Recent evidence points to the potential benefits After hospitalization for acute heart failure (AHF), of sodium–glucose cotransporter 2 inhibitors for reducing readmissions are associated with increased mortality and the risk of hospitalizations for HF.6,7 Although strict glycaemic constitute a heavy healthcare burden.1 Current heart failure control is not recommended in patients with HF, insulin (HF) programmes and institutional initiatives thus prioritize remains one of the most common therapies in those who efforts to reduce readmissions rates.2,3 Unfortunately, risk do not achieve optimal metabolic control.8,9 This therapy is factors for identifying patients at higher risk of this outcome still recommended for patients with T2DM and HF,10 and are not well established.4 30–50% of these patients receive it.9 Type 2 diabetes mellitus (T2DM) is a known risk factor for However, insulin is an anabolizing hormone that causes developing HF and is highly prevalent among these patients; sodium and fluid retention, weight gain, and hypoglycaemia, © 2020 The Authors. ESC Heart Failure published by John Wiley & Sons Ltd on behalf of the European Society of Cardiology This is an open access article under the terms of the Creative Commons Attribution‐NonCommercial‐NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non‐commercial and no modifications or adaptations are made. Insulin and heart failure readmissions 3321 with resulting sympathetic nervous system activation.11 Some Endpoints and follow‐up observational studies have found an increased risk of adverse outcomes in patients with HF and on insulin therapy,12,13 The primary endpoint was readmissions for HF at 1 year. although most of the evidence of this connection comes from Readmission for AHF was defined as an unplanned post hoc analyses in select patients with chronic HF enrolled hospitalization requiring >24 h stay and caused by substan- in randomized clinical trials.13 Little is known about the tive worsening of HF symptoms and/or signs requiring new influence of insulin treatment on the risk of hospitalizations administration of intravenous HF therapies, including for HF, especially in subgroups at higher risk of readmission, inotropes, diuretics, or vasodilators. Electronic clinical such as patients with a recent HF decompensation. We records are shared by all public hospitals in the area and therefore aimed to evaluate the impact of insulin treatment by all general practitioners. Clinical records were reviewed on the risk of readmissions and all‐cause mortality after an to ensure the capture of eventual hospitalizations in other episode of AHF. hospitals. The secondary endpoint was all‐cause mortality at 1 year. The personnel in charge of endpoint adjudications during follow‐up were blinded to T2DM status and treatment. Methods Study design and patients Statistical analysis This prospective observational cohort study included 3054 Continuous variables are presented as mean (± standard consecutive patients admitted with AHF in the cardiology deviation) or median (inter‐quartile range), as appropriate. department of a tertiary referral hospital from 1 January Categorical variables are expressed as absolute and relative 2007 to 1 August 2017. AHF was defined according to clinical frequencies. Baseline continuous variables were compared practice guidelines.14–16 Patients with new‐onset or acutely according to T2DM status (no T2DM, no‐insulin T2DM, and decompensated HF were included in the registry. Patients T2DM on insulin therapy) with the ANOVA or Wilcoxon test, who died during the index hospitalization and those with type as appropriate; discrete variables were compared with the 1 diabetes were excluded from the final analysis (n = 159), χ 2 test. leaving a final sample of 2895 patients. T2DM diagnosis was Cox regression analysis was used for evaluating the risk of defined according to international guidelines: receiving 1 year mortality. An adapted Cox regression model, which medical treatment for T2DM, confirmed fasting glucose levels took into account the effect of all‐cause mortality (method of more than 125 mg/dL, non‐fasting glucose levels of more of Fine and Gray), was used to examine the independent than 140 mg/dL, or glycated haemoglobin of more than association between T2DM status and HF readmissions.18 6.5%.17 Dyslipidaemia was defined as a concentration of total All covariates shown in Table 1 were evaluated in regression cholesterol of 200 mg/dL or more, a concentration of models for prognostic purposes. In an attempt to minimize triglycerides of 150 mg/dL or more, a registered diagnosis the residual confounding and indication bias, the covariates of hypercholesteraemia or hypertriglyceridaemia, or the use included in the multivariable models were selected of lipid‐lowering treatment. based on the biological/clinical plausibility, regardless of During the index hospitalization, data on demographics, the P‐value. The linearity assumption for all continuous medical history, vital signs, 12‐lead electrocardiogram, variables was simultaneously tested and the variable laboratory and echocardiographic parameters, and drug use transformed, if appropriate, with fractional polynomials.19 were routinely recorded using pre‐established registry Several multivariate models were built. The first one questionnaires. Left ventricular ejection fraction (LVEF) included age and sex (Model 1). Model 2 included age, was assessed by two‐dimensional echocardiography and sex, prior admission for AHF, last New York Heart Associa- calculated using the biplane Simpson method (96 ± 24 h after tion (NYHA) class under stable condition, aetiology, admission). HF and anti‐diabetic pharmacological treatments dyslipidaemia, Charlson co‐morbidity index, peripheral were recorded at discharge. oedemas on admission, left bundle branch block, systolic and diastolic blood pressures, atrial fibrillation, heart rate, glomerular filtration rate, haemoglobin, sodium, N‐terminal Ethical considerations pro‐B‐type natriuretic peptide (NT‐proBNP), left ventricular ejection fraction categories (>40%, 40–49%, and ≥50%), The local institutional ethics committee (Hospital Clínico severe tricuspid regurgitation, and left atrial diameter. Universitario de Valencia) approved the study, which Model 3 included the same covariates. of Model 2 and also conformed to the principles outlined in the Declaration of HF and T2DM medications prescribed on discharge Helsinki. All patients gave informed consent. (furosemide equivalent doses, angiotensin‐converting ESC Heart Failure 2020; 7: 3320–3328 DOI: 10.1002/ehf2.12944 3322 V. Bertomeu‐Gonzalez et al. Table 1 Participant characteristics at baseline and outcomes at 12 months Variables No diabetes (N = 1600) No‐insulin T2DM (N = 768) T2DM on insulin (N = 527) P Demographic and medical history Age (years), mean ± SD 73.1 ± 12.5 73.6 ± 9.8 73.7 ± 9.1 0.486 * Men, n (%) 785 (50.9) 339 (44.1) 269 (51.0) 0.005 * ≥2 prior admissions for AHF, n (%) 685 (42.8) 365 (47.5) 298 (56.5) <0.001 * Last stable NYHA class ≥III, n (%) 215 (13.4) 120 (15.6) 118 (22.4) <0.001 * Hypertension, n (%) 1149 (71.8) 649 (84.5) 478 (90.7) <0.001 Dyslipidaemia, n (%) 710 (44.4) 484 (63.0) 357 (67.7) <0.001 Current smoker, n (%) 213 (13.3) 96 (12.5) 50 (9.5) 0.069 * History of CHD, n (%) 437 (27.3) 309 (40.2) 271 (51.4) <0.001 * History of atrial fibrillation, n (%) 774 (48.4) 336 (43.8) 170 (32.3) <0.001 * Charlson index, mean ± SD 1.4 ± 1.4 2.4 ± 1.7 3.3 ± 1.9 <0.001 Physical examination at admission * Heart rate (b.p.m.), mean ± SD 101 ± 30 98 ± 28 93 ± 24 <0.001 Systolic BP (mmHg), mean ± SD 146 ± 33 148 ± 33 149 ± 33 0.215 * Diastolic BP (mmHg), mean ± SD 83 ± 20 81 ± 20 78 ± 18 <0.001 Peripheral oedema, n (%) 908 (56.8) 487 (63.4) 356 (67.6) <0.001 Pleural effusion, n (%) 774 (46.5) 366 (47.7) 255 (48.4) 0.714 Laboratory parameters * Haemoglobin (g/dL), mean ± SD 12.8 ± 1.9 12.4 ± 1.9 11.8 ± 1.9 <0.001 * Creatinine (mg/dL), mean ± SD 1.2 ± 0.6 1.2 ± 0.5 1.4 ± 0.7 <0.001 2 * eGFR (mL/min/1.73 m ), mean ± SD 64.3 ± 29.3 63.7 ± 24.3 54.3 ± 24.4 <0.001 * Serum sodium (mEq/L), mean ± SD 138.8 ± 4.4 138.4 ± 4.5 138.1 ± 4.5 0.005 NT‐proBNP (pg/mL), median (IQR) 4721 (2012–8855) 3992 (1889–7769) 4154 (1937–8012) 0.066 CA 125 (U/mL), median (IQR) 55 (25–126) 54 (25–120) 58 (26–127) 0.900 Echocardiography LVEF categories, n (%) 0.002 ≤40% 512 (32.0) 253 (32.9) 148 (28.1) 41–4% 217 (13.6) 140 (18.2) 100 (19.0) ≥50% 871 (54.4) 375 (48.8) 279 (52.9) LVEF (%), mean ± SD 49.7 ± 15.7 49.0 ± 14.8 50.0 ± 14.7 0.436 LA diameter (mm), mean ± SD 44.2 ± 8.6 43.5 ± 7.3 42.9 ± 7.2 0.002 Tricuspid regurgitation ≥III, mean ± SD 167 (10.4) 64 (8.3) 41 (7.8) 0.097 Treatment at discharge Loop diuretics, n (%) 1584 (99.0) 757 (98.6) 524 (99.4) 0.314 * Furosemide dose (mg/day), mean ± SD 61.9 ± 43.3 66.5 ± 43.7 78.6 ± 42.9 <0.001 Beta‐blockers, n (%) 1123 (70.2) 522 (68.0) 348 (66.2) 0.169 ACEI/ARB, n (%) 1014 (63.41) 533 (69.4) 355 (67.4) 0.010 * MRAs, n (%) 473 (29.6) 271 (35.3) 149 (28.3) 0.007 Statins, n (%) 670 (41.9) 427 (55.6) 303 (57.5) <0.001 Events at 12 month follow‐up, n (%) * Death 279 (17.4) 130 (16.9) 109 (20.7) <0.001 * HF readmissions 332 (20.8) 181 (23.6) 180 (34.2) <0.001 * Death or HF readmission 497 (31.1) 253 (32.9) 243 (46.1) <0.001 ACEI/ARB, angiotensin‐converting enzyme inhibitor/angiotensin receptor blocker; AHF, acute heart failure; BP, blood pressure; CA 125, antigen carbohydrate 125; CHD, coronary heart disease; eGFR, estimated glomerular filtration rate; HF, heart failure; IQR, inter‐quartile range; LA, left atrial; LVEF, left ventricular ejection fraction; MRAs, mineralocorticoid receptor antagonists; NT‐proBNP, N‐terminal pro‐ brain natriuretic peptide; NYHA, New York Heart Association; SD, standard deviation; T2DM, type 2 diabetes mellitus. Comparisons were made among the three categories (no T2DM, no‐insulin T2DM, and T2DM on insulin therapy). Continuous variables 2 were compared with the ANOVA or Kruskal–Wallis test, as appropriate; discrete variables were compared with the χ test. Comparison between continuous variables between no‐insulin T2DM and T2DM on insulin was tested with t‐test or Mann–Whitney–Wilcoxon, as appropriate. Significant differences (P < 0.05) in an additional analysis comparing patients with T2DM with and without insulin therapy. * enzyme inhibitors/angiotensin receptor blockers, mineralo- Release 15 (2017); StataCorp LP, College Station, TX, USA] corticoid receptor antagonists, beta‐blockers, metformin, was used for the analysis. sulfonylureas, and dipeptidyl peptidase‐4 inhibitors). The proportionality assumption for the hazard function over time was tested by means of the Schoenfeld residuals. The discriminatory ability of the multivariable models was Results evaluated with Harrell’s C‐statistic. We set a two‐sided P‐value of <0.05 as the threshold for The mean age of the sample was 73.4 ± 11.2 years, 50.8% statistical significance. Stata 15.1 [Stata Statistical Software, were women, 52.7% showed LVEF of at least 50%, and ESC Heart Failure 2020; 7: 3320–3328 DOI: 10.1002/ehf2.12944 Insulin and heart failure readmissions 3323 44.7% had T2DM. Of the 1295 participants with T2DM, 40.7% The 1 year HF readmission rates were higher in patients (n = 527) were on insulin. with T2DM on insulin compared with patients with T2DM not on insulin and non‐diabetic patients (Table 1). Cumulative incidence curves showed a marked separation from the first month after discharge (Figure 1). After a comprehensive Baseline characteristics across type 2 diabetes and progressive multivariable adjustment, those patients re- mellitus status ceiving insulin were at higher risk for 1 year HF readmissions (Table 2). Compared with non‐T2DM, T2DM on insulin Table 1 shows participants’ characteristics according to T2DM showed and increased risk [hazard ratio (HR) 1.26, 95% confi- categories. People receiving insulin for their diabetes exhib- dence interval (CI) 1.02–1.59, P = 0.035]. Likewise, the excess ited a worse HF risk profile, with a higher proportion of prior of risk was also present when compared with T2DM not on hypertension, dyslipidaemia, prior admission for AHF, ischae- insulin (HR 1.28, 95% CI 1.04–1.59, P = 0.022). The risk of mic heart disease, worse baseline functional NYHA class, HF readmission did not differ significantly between the greater signs of congestion, and more Charlson co‐morbidity. no‐insulin T2DM and no‐T2DM groups (HR 1.02, 95% CI Likewise, haemoglobin, glomerular filtration rate, and serum 0.84–1.25, P = 0.829). Harrell’s C‐statistic of the fully sodium values were lower in this group. At discharge, they adjusted model was 0.773. were more frequently treated with statins and higher loop di- The excess risk attributable to insulin did not differ uretic doses. In contrast, a lower proportion were prescribed significantly by sex or HF treatment nor did it vary across with mineralocorticoid receptor antagonists. No differences key patient subgroups, including those aged 75 years or were found for beta‐blockers. Glycated haemoglobin more, those with LVEF of 50% or more, and those with higher was higher in people with T2DM not receiving insulin than median NT‐proBNP values (Figure 2). When the analysis (7.5 ± 1.4%) compared with those who were (7.1 ± 1.2%; was restricted to patients with T2DM, there was no P < 0.001). The no‐insulin T2DM group was also treated differential effect across the levels of glycated haemoglobin, more frequently with metformin (44.3% vs. 32.4%) and DPP4i regardless of whether the variable was analysed continuously (24.2% vs. 17.0%) than those receiving insulin, with significant (P‐value for interaction = 0.365) or dichotomized at the differences between groups. There were no differences in the median (P‐value for interaction = 0.446). sulfonylurea prescription rates (31.3% vs. 29.9%, P = 0.583). Outcomes at 1 year All‐cause mortality At 1 year follow‐up, 518 (17.9%) of our participants died, Kaplan–Meier curves did not differ significantly among the 693 (23.9%) were readmitted for HF, and 993 (34.3%) T2DM categories (Figure 3). Likewise, multivariable analysis experienced at least one of these two outcomes. did not show differences between these groups (HR 1.03, Figure 1 Cumulative incidence of heart failure (HF) readmissions, stratified by groups according to type 2 diabetes mellitus (T2DM) status and treat- ment (no T2DM, no‐insulin T2DM, and T2DM on insulin therapy). ESC Heart Failure 2020; 7: 3320–3328 DOI: 10.1002/ehf2.12944 3324 V. Bertomeu‐Gonzalez et al. Table 2 Insulin treatment and risk of 1 year HF readmission readmission. These findings point to the potentially harmful HR (95% CI) Omnibus P‐value effect of insulin therapy for T2DM following an episode of Univariate decompensation. No diabetes 1 <0.001 No‐insulin T2DM 1.14 (0.95–1.37) T2DM on insulin 1.72 (1.44–2.06) Multivariate Risk of readmission in patients with established Model 1 a heart failure: what role does type 2 diabetes No diabetes 1 <0.001 mellitus play? No‐insulin T2DM 1.15 (0.95–1.37) T2DM on insulin 1.73 (1.44–2.07) b Model 2 Reducing hospital readmissions is a major objective in No diabetes 1 0.040 managing patients with HF, as rehospitalization is strongly No‐insulin T2DM 0.99 (0.82–1.20) T2DM on insulin 1.27 (1.03–1.56) associated with disease progression and mortality.20 Many c Model 3 strategies have been proposed to reduce the burden of No diabetes 1 0.042 readmissions after an AHF episode,3,21 but predicting which No‐insulin T2DM 0.98 (0.80–1.19) T2DM on insulin 1.26 (1.02–1.55) patients are at higher risk remains challenging, as there are no well‐established predictors for this outcome. CI, confidence interval; HF, heart failure; HR, hazard ratio; T2DM, type 2 diabetes mellitus. Most of the literature suggests that T2DM is an indepen- a Model 1: adjusted for age and sex. dent predictor of mortality in patients with HF, but this b Model 2: adjusted for age, sex, prior admission for acute heart relationship is not necessarily straightforward. Some authors failure, last New York Heart Association class under stable condition, aetiology, dyslipidaemia, Charlson co‐morbidity index, hold that T2DM increases the risk only in patients with peripheral oedemas on admission, left bundle branch block, coronary artery disease,22 whereas others argue that women systolic and diastolic blood pressures, atrial fibrillation, heart bear a disproportionate share of the burden.23 Our group rate, glomerular filtration rate, haemoglobin, sodium, N‐terminal pro‐brain natriuretic peptide, left ventricular ejection fraction, observed a higher risk of death in women with T2DM in a severe tricuspid regurgitation, and left atrial diameter. cohort of patients with HF and preserved ejection fraction.24 c Model 3: adjusted for age, sex, prior admission for acute heart However, emerging evidence suggests that most of T2DM’s failure, last New York Heart Association class under stable condi- tion, aetiology, dyslipidaemia, Charlson co‐morbidity index, prognostic impact is mediated by insulin therapy. To date, peripheral oedemas on admission, left bundle branch block, three post hoc analyses from randomized clinical trials, systolic and diastolic blood pressures, atrial fibrillation, heart rate, plus one observational registry study, have found an indepen- glomerular filtration rate, haemoglobin, sodium, N‐terminal pro‐ brain natriuretic peptide, left ventricular ejection fraction, severe dent association between insulin therapy and adverse tricuspid regurgitation, left atrial diameter, and discharge treat- outcomes in different phenotypes of patients with stable ments (furosemide equivalent doses, angiotensin‐converting HF.13,25–27 Our study corroborates these findings and enzyme inhibitors/angiotensin receptor blockers, mineralocorti- coid receptor antagonists, beta‐blockers, metformin, sulfonyl- extends the implications to the first year following an episode ureas, and dipeptidyl peptidase‐4 inhibitors). Harrell’s C‐statistic of decompensation—a high‐risk period for subsequent HF of Model 3 = 0.773. readmission. Biological plausibility for the association between 95% CI 0.79–1.33, P = 0.840) nor were there variations by insulin and heart failure risk patient subgroup (Figure 4). Congestion explains most of the pathophysiology of AHF syndromes,28 and several pathophysiological mechanisms, probably interrelated, could be behind a causal relationship Discussion between insulin and greater congestion. In this large observational study, there was a strong associa- tion between insulin treatment at discharge and a higher risk Sodium and water retention of 1 year readmission for HF in patients discharged for AHF. This increased risk was consistent across key patient sub- Insulin treatment has well‐known effects on sodium groups. Most published evidence on the association between transport, resulting in sodium and water retention.29 The insulin therapy and adverse events (as a composite outcome) mechanism of insulin’s anti‐natriuretic effects involves the has come from a narrow selection of stable patients enrolled reduction of glycosuria and, in turn, of urinary sodium excre- in clinical trials. In contrast, our sample of patients with AHF tion. In addition, insulin also directly promotes sodium and was much more representative, and we evaluated the water reabsorption in the nephron.29 These anti‐natriuretic specific association between insulin treatment and HF effects, while widely acknowledged by the scientific ESC Heart Failure 2020; 7: 3320–3328 DOI: 10.1002/ehf2.12944 Insulin and heart failure readmissions 3325 Figure 2 The excess risk of heart failure readmission attributable to insulin across patient subgroups [no type 2 diabetes mellitus (T2DM), no‐insulin T2DM, and T2DM on insulin therapy]. ACEI, angiotensin‐converting enzyme inhibitor; ARB, angiotensin receptor blocker; LVEF, left ventricular ejection fraction; MRA, mineralocorticoid receptor antagonist; NT‐proBNP, N‐terminal pro‐B‐type natriuretic peptide. Figure 3 Cumulative mortality incidence, stratified according to type 2 diabetes mellitus (T2DM) status and treatment (no T2DM, no‐insulin T2DM, and T2DM on insulin therapy). community for over 40 years, are probably underestimated. Hypoglycaemia Papers dating back to 1974 warn against insulin’s side effects: ‘In the diabetic with a long‐term uncontrolled disease, it must Insulin is a major determinant of hypoglycaemic episodes. be anticipated that sodium will be avidly retained. Excessive The Action to Control Cardiovascular Risk in Diabetes trial sodium intake might overload the vascular space and was designed to evaluate the effects of intensive glucose precipitate cardiopulmonary complications in the susceptible control for preventing cardiovascular disease. Far from patient’.30 Our results are consistent with these predictions, demonstrating a benefit, the trial found increased mortality demonstrating that patients on insulin therapy showed more in patients with more intensive glycaemic control, which symptoms and signs of congestion. was attributed to higher rates of hypoglycaemic events in ESC Heart Failure 2020; 7: 3320–3328 DOI: 10.1002/ehf2.12944 3326 V. Bertomeu‐Gonzalez et al. Figure 4 The excess mortality risk attributable to insulin across patient subgroups [no type 2 diabetes mellitus (T2DM), no‐insulin T2DM, and T2DM on insulin therapy]. ACEI, angiotensin‐converting enzyme inhibitor; ARB, angiotensin receptor blocker; LVEF, left ventricular ejection fraction; MRA, min- eralocorticoid receptor antagonist; NT‐proBNP, N‐terminal pro‐B‐type natriuretic peptide. that group.31 A recent post hoc analysis of the EMPA‐REG Despite the strong association and biological postulates OUTCOME trial has shown a strong association between that reinforce causality, we cannot rule out the possibility hypoglycaemia and an increased risk of HF readmissions and that this relationship is confounded by other factors. Insulin myocardial infarction.32 Although the statistical association is used to treat cases of T2DM that are more severe and have between hypoglycaemia and cardiovascular mortality is a longer duration. In HF clinical trials, patients receiving strong, the mechanisms underpinning it are not well insulin for T2DM were at a more advanced stage in terms understood. The counterregulatory response stimulates the of NYHA functional class, NT‐proBNP values, and congestion sympathetic nervous system, which can lead to increased signs compared with those without insulin treatment.13 preload, myocardial ischaemia, cardiac arrhythmias, Despite multivariable adjustment, we cannot rule out the and prothrombotic state—all factors related to HF presence of some residual confounding. A randomized clinical decompensation.33,34 In addition, some evidence suggests trial designed specifically to assess the effects of that in patients with chronic insulin resistance, insulin may de‐intensifying insulin treatment in HF patients with T2DM be associated with the activation of the sympathetic nervous would provide valuable information for better managing system, even without the mediation of hypoglycaemia.29 these patients. With the existing evidence, we propose to counteract the excess of risk associated to insulin therapy with a careful adjustment of diuretics, the use of other anti‐diabetics with natriuretic effect such as sodium–glucose Cardiac hypocontractility cotransporter 2 inhibitors, and a close monitoring of body weight and other decompensation parameters of HF. Hyperinsulinaemia has been associated with decreased myo- cardial contractility, despite elevated catecholamine levels. The mechanisms have yet to be fully elucidated, but some Limitations of the implicated pathophysiological pathways include dysfunctional calcium handling within cardiomyocytes, Our study has several limitations. First, it is a single‐centre catecholamine‐induced β‐adrenergic receptor desensitiza- observational study with several potential confounders. tion, and direct inhibition of cardiac adrenergic signalling. Re- Although the analysis used rigorous multivariable adjustment, inforcing these findings, blocking the insulin receptor‐induced the effects of unmeasured confounders cannot be fully adrenergic signalling has been shown to ameliorate cardiac excluded. Some important data were not registered, dysfunction.35 including the duration of T2DM and HF, the doses and type ESC Heart Failure 2020; 7: 3320–3328 DOI: 10.1002/ehf2.12944 Insulin and heart failure readmissions 3327 of insulin regimen, the number of hypoglycaemic episodes, findings warrant further investigation into the relationship and changes over time in treatment (HF and anti‐diabetes) between insulin treatment and prognosis following an epi- and glycaemic control. Observational studies are likewise sode of AHF. unable to establish causality, although they do provide real‐world data, which are particularly necessary when studying patients who are under‐represented in clinical trials. Another limitation was that no uniform screening for T2DM Conflict of interest was used, raising the possibility that some borderline cases may have been misdiagnosed. Finally, given the underuse of None declared. sodium–glucose cotransporter 2 inhibitors and glucagon‐like peptide‐1 receptor agonists at the time of this study, we could not evaluate the influence of these pharmacological Funding groups on the present findings. This work was supported in part by grants from Centro de Investigación Biomédica en Red Enfermedades Conclusions Cardiovasculares (CIBERCV) (16/11/00420 and 16/11/ 00403), Federación Española de Enfermedades Raras Insulin therapy is associated with an increased risk of 1 year (FEDER), and European Observation Network for Territorial HF readmission in patients with AHF and T2DM. The present Development and Cohesion (PIE15/00013). References 1. Santas E, Valero E, Mollar A, García‐Blas FA, Ponikowski P, Sabatine MS, Anand 10. American Diabetes Association. 9. Phar- S, Palau P, Miñana G, Núñez E, Sanchis J, IS, Bělohlávek J, Böhm M, Chiang C‐E, macologic approaches to glycemic treat- Chorro FJ, Núñez J. Burden of recurrent Chopra VK, de Boer RA, Desai AS, Diez ment: Standards of Medical Care in hospitalizations following an admission M, Drozdz J, Dukát A, Ge J, Howlett JG, Diabetes—2019. Diabetes Care 2019; for acute heart failure: preserved versus Katova T, Kitakaze M, Ljungman CEA, 42: S90–S102. reduced ejection fraction. Rev Esp Cardiol Merkely B, Nicolau JC, O’Meara E, Petrie 11. Rowe JW, Young JB, Minaker KL, Ste- (Engl Ed) 2017; 70: 239–246. MC, Vinh PN, Schou M, Tereshchenko S, vens AL, Pallotta J, Landsberg L. Effect 2. Horne BD, Roberts CA, Rasmusson KD, Verma S, Held C, DeMets D, Docherty of insulin and glucose infusions on sym- Buckway J, Alharethi R, Cruz J, Evans KF, Jhund PS, Bengtsson O, Sjöstrand pathetic nervous system activity in nor- RS, Lloyd JF, Bair TL, Kfoury AG, Lappé M, Langkilde AM, DAPA‐HF Trial Com- mal man. Diabetes 1981; 30: 219–225. DL. Risk score‐guided multidisciplinary mittees and Investigators. Dapagliflozin 12. Pocock SJ, Wang D, Pfeffer MA, Yusuf S, team‐based Care for Heart Failure Inpa- in patients with heart failure and re- McMurray JJV, Swedberg KB, Östergren tients is associated with lower 30‐day re- duced ejection fraction. N Engl J Med J, Michelson EL, Pieper KS, Granger CB. admission and lower 30‐day mortality. 2019; 381: 1995–2008. Predictors of mortality and morbidity in Am Heart J 2020; 219: 78–88. 7. Zinman B, Wanner C, Lachin JM, patients with chronic heart failure. Eur 3. Ryan J, Kang S, Dolacky S, Ingrassia J, Fitchett D, Bluhmki E, Hantel S, Heart J 2006; 27: 65–75. Ganeshan R. Change in readmissions Mattheus M, Devins T, Johansen OE, 13. Cosmi F, Shen L, Magnoli M, Abraham and follow‐up visits as part of a heart Woerle HJ, Broedl UC, Inzucchi SE. WT, Anand IS, Cleland JG, Cohn JN, failure readmission quality improvement Empagliflozin, cardiovascular outcomes, Cosmi D, De Berardis G, Dickstein K, initiative. Am J Med 2013; 126: 989, and mortality in type 2 diabetes. N Engl Franzosi MG, Gullestad L, Jhund PS, e1–994. J Med Massachusetts Medical Society Kjekshus J, Køber L, Lepore V, Lucisano 4. Ross JS, Mulvey GK, Stauffer B, Patlolla 2015; 373: 2117–2128. G, Maggioni AP, Masson S, McMurray V, Bernheim SM, Keenan PS, Krumholz 8. Zareini B, Rørth R, Holt A, Mogensen JJV, Nicolucci A, Petrarolo V, Robusto HM. Statistical models and patient UM, Selmer C, Gislason G, Schou M, F, Staszewsky L, Tavazzi L, Teli R, predictors of readmission for heart Køber L, Torp‐Pedersen C, Lamberts M, Tognoni G, Wikstrand J, Latini R. Treat- failure. a systematic review. Arch Intern Kristensen SL. Heart failure and the ment with insulin is associated with Med 2008; 168: 1371–1386. prognostic impact and incidence of worse outcome in patients with chronic 5. Al‐Jarallah M, Rajan R, Al‐Zakwani I, new‐onset of diabetes mellitus: a nation- heart failure and diabetes. Eur J Heart Dashti R, Bulbanat B, Ridha M, wide cohort study. Cardiovasc Diabetol Fail 2018; 20: 888–895. Sulaiman K, Alsheikh‐Ali AA, 2019; 18: 79. 14. , Dickstein K, Cohen‐Solal A, Filippatos Panduranga P, AlHabib KF, AlSuwaidi 9. Harjola V‐P, Follath F, Nieminen MS, G, McMurray JJV, Ponikowski P, Poole‐ J, Al‐Mahmeed W, AlFaleh H, Elasfar A, Brutsaert D, Dickstein K, Drexler H, Wilson PA, Stromberg A, van Veldhuisen Al‐Motarreb A, Bazargani N, Asaad N, Hochadel M, Komajda M, Lopez‐Sendon DJ, Atar D, Hoes AW, Keren A, Mebazaa Amin H. Impact of diabetes on mortality JL, Ponikowski P, Tavazzi L. Characteris- A, Nieminen M, Priori SG, Swedberg K, and rehospitalization in acute heart fail- tics, outcomes, and predictors of mortal- Vahanian A, Camm J, De Caterina R, ure patients stratified by ejection frac- ity at 3 months and 1 year in patients Dean V, Dickstein K, Filippatos G, tion. ESC Hear Fail 2019: ehf2.12538. hospitalized for acute heart failure. Eur Funck‐Brentano C, Hellemans I, 6. McMurray JJV, Solomon SD, Inzucchi J Heart Fail John Wiley & Sons, Ltd Kristensen SD, McGregor K, Sechtem U, SE, Køber L, Kosiborod MN, Martinez 2010; 12: 239–248. Silber S, Tendera M, Widimsky P, ESC Heart Failure 2020; 7: 3320–3328 DOI: 10.1002/ehf2.12944 3328 V. Bertomeu‐Gonzalez et al. Zamorano JL, Tendera M, Auricchio A, 17. American Diabetes Association. 2. Classi- Heart Fail John Wiley and Sons Ltd Bax J, Bohm M, Corra U, della Bella P, fication and diagnosis of diabetes: Stan- 2019; 21: 974–984. Elliott PM, Follath F, Gheorghiade M, dards of Medical Care in Diabetes— 27. Huynh T, Harty BJ, Claggett B, Fleg JL, Hasin Y, Hernborg A, Jaarsma T, 2019. Diabetes Care American Diabetes McKinlay SM, Anand IS, Lewis EF, Jo- Komajda M, Kornowski R, Piepoli M, Association Inc. 2019; 42: S13–S28. seph J, Desai AS, Sweitzer NK, Prendergast B, Tavazzi L, Vachiery J‐L, 18. Fine JP, Gray RJ. A proportional hazards EileenO’Meara PB, Pfeffer MA, Rouleau Verheugt FWA, Zannad F. ESC Guide- model for the subdistribution of a com- J‐L. Comparison of outcomes in patients lines for the diagnosis and treatment of peting risk. J Am Stat Assoc Taylor & with diabetes mellitus treated with ver- acute and chronic heart failure 2008: Francis, Ltd. American Statistical Associ- sus without insulin + heart failure with The Task Force for the Diagnosis and ation 1999; 94: 496–509. preserved left ventricular ejection frac- Treatment of Acute and Chronic Heart 19. Harrell FE. Regression Modeling Strate- tion (from the TOPCAT study). Am J Failure 2008 of the European Society of gies: With Applications to Linear Models, Cardiol 2019; 123: 611–617. Cardiology. Developed in collaboration Logistic Regression, and Survival Analy- 28. Costanzo MR. Verdict in: congestion with the Heart Failure Association of sis. New York: Springer; 2001. guilty! JACC Hear Fail 2015; 3: 762–764. the ESC (HFA) and endorsed by the Eu- 20. Gheorghiade M, Vaduganathan M, 29. Brands MW. Role of insulin‐mediated ropean Society of Intensive Care Medi- Fonarow GC, Bonow RO. Rehospitaliza- antinatriuresis in sodium homeostasis cine (ESICM). Eur Heart J 2008; 29: tion for heart failure: problems and per- and hypertension. Hypertension 2018; 2388–2442. https://doi.org/10.1093/ spectives. J Am Coll Cardiol 2013; 61: 72: 1255–1262. eurheartj/ehn309 391–403. 30. Saudek CD, Boulter PR, Knopp RH, Arky 15. McMurray JJV, Adamopoulos S, Anker 21. Psotka MA, Teerlink JR. Strategies to RA. Sodium retention accompanying in- SD, Auricchio A, Bohm M, Dickstein K, prevent postdischarge adverse events sulin treatment of diabetes mellitus. Di- Falk V, Filippatos G, Fonseca C, Gomez‐ among hospitalized patients with heart abetes 1974; 23: 240–246. Sanchez MA, Jaarsma T, Kober L, Lip failure. Heart Fail Clin 2013; 9: 31. Action to Control Cardiovascular Risk in GYH, Maggioni AP, Parkhomenko A, 303–320. Diabetes Study Group, Gerstein HC, Pieske BM, Popescu BA, Ronnevik PK, 22. Dries DL, Sweitzer NK, Drazner MH, Miller ME, Byington RP, Goff DC, Bigger Rutten FH, Schwitter J, Seferovic P, Stevenson LW, Gersh BJ. Prognostic im- JT, Buse JB, Cushman WC, Genuth S, Is- Stepinska J, Trindade PT, Voors AA, pact of diabetes mellitus in patients with mail‐Beigi F, Grimm RH, Probstfield JL, Zannad F, Zeiher A, Bax JJ, Baumgartner heart failure according to the etiology of Simons‐Morton DG, Friedewald WT. Ef- H, Ceconi C, Dean V, Deaton C, Fagard R, left ventricular systolic dysfunction. J fects of intensive glucose lowering in Funck‐Brentano C, Hasdai D, Hoes A, Am Coll Cardiol 2001; 38: 421–428. type 2 diabetes. N Engl J Med 2008; Kirchhof P, Knuuti J, Kolh P, McDonagh 23. Gustafsson I, Brendorp B, Seibaek M, 358: 2545–2559. T, Moulin C, Popescu BA, Reiner Z, Burchardt H, Hildebrandt P, Køber L, 32. Fitchett D, Inzucchi SE, Wanner C, Sechtem U, Sirnes PA, Tendera M, Torp‐Pedersen C, Danish Investigations Mattheus M, George JT, Vedin O, Torbicki A, Vahanian A, Windecker S, of Arrhythmia and Mortality on Zinman B, Johansen OE. Relationship McDonagh T, Sechtem U, Bonet LA, Dofetilide Study Group. Influence of dia- between hypoglycaemia, cardiovascular Avraamides P, Ben Lamin HA, Brignole betes and diabetes–gender interaction outcomes, and empagliflozin treatment ® M, Coca A, Cowburn P, Dargie H, Elliott on the risk of death in patients hospital- in the EMPA‐REG OUTCOME trial. P, Flachskampf FA, Guida GF, Hardman ized with congestive heart failure. J Am Eur Heart J Oxford University Press S, Iung B, Merkely B, Mueller C, Nanas Coll Cardiol 2004; 43: 771–777. (OUP) 2020; 41: 209–217. JN, Nielsen OW, Orn S, Ponikowski P. 24. Palau P, Bertomeu‐González V, Sanchis 33. Stahn A, Pistrosch F, Ganz X, Teige M, ESC Guidelines for the diagnosis and J, Soler M, de la Espriella R, Domínguez Koehler C, Bornstein S, Hanefeld M. Re- treatment of acute and chronic heart E, Santas E, Núñez E, Chorro FJ, Miñana lationship between hypoglycemic epi- failure 2012: The Task Force for the Di- G, Bayés‐Genís A, Núñez J. Differential sodes and ventricular arrhythmias in agnosis and Treatment of Acute and prognostic impact of type 2 diabetes patients with type 2 diabetes and cardio- Chronic Heart Failure 2012 of the Euro- mellitus in women and men with heart vascular diseases: silent hypoglycemias pean Society of Cardiology. Developed failure with preserved ejection fraction. and silent arrhythmias. Diabetes Care in collaboration with the Heart Failure Rev Española Cardiol (English Ed 2014; 37: 516–520. Association (HFA) of the ESC. Eur Heart Elsevier BV) 2020; 73: 463–470. 34. Dandona P, Chaudhuri A, Dhindsa S. J 2012; 33: 1787–1847. https://doi.org/ 25. Smooke S, Horwich TB, Fonarow GC. Proinflammatory and prothrombotic ef- 10.1093/eurheartj/ehs104 Insulin‐treated diabetes is associated fects of hypoglycemia. Diabetes Care 16. Ponikowski P, Voors AA, Anker SD, with a marked increase in mortality in American Diabetes Association 2010; 33: Bueno H, Cleland JGF, Coats AJS, Falk patients with advanced heart failure. 1686–1687. V, González‐Juanatey JR, Harjola V‐P, Am Heart J 2005; 149: 168–174. 35. Wang Q, Liu Y, Fu Q, Xu B, Zhang Y, Kim Jankowska EA, Jessup M, Linde C, 26. Shen L, Rørth R, Cosmi D, Kristensen S, Tan R, Barbagallo F, West T, Anderson Nihoyannopoulos P, Parissis JT, Pieske SL, Petrie MC, Cosmi F, Latini R, Køber E, Wei W, Abel ED, Xiang YK. Inhibiting B, Riley JP, Rosano GMC, Ruilope LM, L, Anand IS, Carson PE, Granger CB, insulin‐mediated β2‐adrenergic receptor Ruschitzka F, Rutten FH, van der Meer Komajda M, McKelvie RS, Solomon SD, activation prevents diabetes‐associated Peter. 2016 ESC Guidelines for the diag- Staszewsky L, Swedberg K, Huynh T, cardiac dysfunction. Circulation 2017; nosis and treatment of acute and chronic Zile MR, Jhund PS, McMurray JJV. Insu- 135: 73–88. heart failure. Eur Heart J 2016; 37: lin treatment and clinical outcomes in 2129–2200. https://doi.org/10.1093/ patients with diabetes and heart failure eurheartj/ehw128 with preserved ejection fraction. Eur J ESC Heart Failure 2020; 7: 3320–3328 DOI: 10.1002/ehf2.12944