Julián et al. Cardiovasc Diabetol (2020) 19:38 https://doi.org/10.1186/s12933-020-01011-w Cardiovascular Diabetology ORIGINAL INVESTIGATION Open Access Long‑term LVEF trajectories in patients with type 2 diabetes and heart failure: diabetic cardiomyopathy may underlie functional decline María Teresa Julián1,2†, Núria Alonso1,2,3,5*†, Josep Lupón1,4,5,6, Giovana Gavidia‑Bovadilla7,8, Elena Ferrer4, Marta de Antonio1,4, Jorge López‑Ayerbe4, Mar Domingo1, Evelyn Santiago‑Vacas1, Elisabet Zamora1,4,5,6, Pau Codina1,4, Pedro Moliner1,4, Julio Núñez7,8,9, Javier Santesmases1,5, Manel Puig‑Domingo2,5 and Antoni Bayes‑Genis1,4,5,6 Abstract Background: Left ventricular ejection fraction (LVEF) trajectories and functional recovery with current heart failure (HF) management is increasingly recognized. Type 2 diabetes mellitus (T2D) leads to a worse prognosis in HF patients. However, it is unknown whether T2D interferes with LVEF trajectories. The aim of this study was to prospectively assess very long-term (up to 15 years) LVEF trajectories in patients with and without T2D and underlying HF. Methods: Ambulatory patients admitted to a multidisciplinary HF clinic were prospectively evaluated by scheduled two-dimensional echocardiography at baseline, 1 year, and then every 2 years afterwards, up to 15 years. Statistical analyses of LVEF change with time were performed using the linear mixed effects (LME) models, and locally weighted error sum of squares (Loess) curves were plotted. Results: Of the 1921 patients, 461 diabetic and 699 non-diabetic patients with LVEF < 50% were included in the study. The mean number of echocardiography measurements performed in diabetic patients was 3.3 ± 1.6. Early LVEF recovery was similar in diabetic and non-diabetic patients, but Loess curves showed a more pronounced inverted U shape in diabetics with a more pronounced decline after 9 years. LME analysis showed a statistical interaction between T2D and LVEF trajectory over time (p = 0.009), which was statistically significant in patients with ischemic etiologies (p < 0.001). Other variables that showed an interaction between LVEF trajectories and T2D were male sex (p = 0.04) and HF duration (p = 0.008). Conclusions: LVEF trajectories in T2D patients with depressed systolic function showed a pronounced inverted U shape with a marked decline after 9 years. Diabetic cardiomyopathy may underlie the functional decline observed. Keywords: Heart failure, Diabetes mellitus, Ejection fraction, Etiology, Long-term, Follow-up Background Heart failure (HF) is a multidimensional syndrome with *Correspondence: nalonso32416@yahoo.es † a wide variety of symptoms and signs associated with María Teresa Julián and Núria Alonso contributed equally to the present work cardiac functional and structural abnormalities, and car- 1 Heart Failure Unit, Hospital Universitari Germans Trias i Pujol, Badalona, ries a high morbidity and mortality [1, 2]. Patients with Barcelona, Spain type 2 diabetes (T2D) have a two- to fivefold increased Full list of author information is available at the end of the article © The Author(s) 2020. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat​iveco​ mmons​.org/licen​ses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creat​iveco​mmons​.org/publi​cdoma​in/ zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Julián et al. Cardiovasc Diabetol (2020) 19:38 Page 2 of 11 risk of developing HF compared to non-diabetic patients, baseline and the other during follow-up. Patients under- and 15–25% of patients with HF have diabetes. This risk going heart transplantation or cardiac resynchronization of HF remains high despite achieving control of other therapy after the second echocardiography were cen- major cardiovascular risk factors [3–5]. Moreover, the sored at the time of the intervention. The study was per- presence of T2D contributes to worse outcomes in HF. formed in compliance with the law protecting personal Patients with T2D are more likely to be hospitalized for data in accordance with the international guidelines on HF, to be re-admitted for HF, and have a higher risk of clinical investigation of the World Medical Association’s cardiovascular and all-cause mortality than those with- Declaration of Helsinki. out diabetes and HF [6, 7]. Insulin resistance and hyper- glycemia, the key central metabolic disturbances in T2D, Study outcome and follow‑up lead to specific structural and metabolic abnormalities The primary outcome of the study was prospectively that adversely affect myocardial function [8]. assess very long-term (up to 15 years) LVEF trajectories Advances in HF treatment including drug therapy, in patients with and without T2D and underlying HF. devices, coronary revascularization, and valvular repair All patients were seen regularly during follow-up vis- are responsible for systolic function improvement in its at the HF clinic according to their clinical needs and patients with depressed left ventricular ejection frac- treated according to a unified protocol [10, 11]. Follow- tion (LVEF) [3, 9], which might be maintained for a up visits included a minimum of 1 visit with a nurse decade. We recently described the dynamic long-term every 3 months and 1 visit with a physician (cardiologist, trajectory of LVEF in a real-life cohort of HF patients internist, or family physician) every 6 months, as well as with depressed systolic function of diverse etiology. Data optional visits with specialists in geriatrics, psychiatry, from this study revealed that LVEF trajectory is highly and rehabilitation. During the baseline visit, patients pro- variable dependent upon etiology, HF duration, and vided written consent for the use of their clinical data for sex, but in general showed a significant improvement at research purposes [10, 11]. 1 year, with an LVEF rise up to decade and a slow decline thereafter [10]. It remains unclear whether the presence Echocardiogram studies of T2D interferes with the improvement and trajectories LVEF was prospectively evaluated at baseline and at 1, 3, of LVEF. 5, 7, 9, 11, 13, and 15 years of follow-up using two-dimen- Accordingly, the aim of the present study was to assess sional echocardiography by image expert cardiologists very long-term (up to 15 years) LVEF trajectories in T2D [10]. LVEF was obtained from apical two- and four- HF patients (from ischemic and non-ischemic etiologies), chamber views and was calculated using the Simpson’s and to compare these trajectories with those observed in method. All echocardiograms were revised for accuracy non-diabetic patients. by expert staff. Methods T2D diagnosis Study design and population A diagnosis of T2D was made when at least one of the This study is a T2D-focused post hoc sub analysis of a following criteria was met: (1) a diagnosis of T2D was previously reported cohort [10]. All consecutive ambu- previously established and recorded in the patient’s elec- latory patients referred to a structured multidisciplinary tronic History; (2) fasting plasma glucose ≥ 126 mg/dL or HF clinic of a university hospital between August 2001 hemoglobin A1C ≥ 6.5% identified by laboratory testing and December 2015, regardless of etiology, were con- [12]; or (3) had a current prescription for oral hypoglyce- sidered for the study. During the 15-year period, clinical mic medication or insulin. pathways and referral geographical area, encompassing ~ 850,000 inhabitants in the northern Barcelona Metro Statistical analysis Area, remained stable. The criteria for referral to the HF Categorical variables were expressed as absolute num- clinic were HF with at least one hospitalization and/or bers and percentages. Continuous variables were depressed systolic function [10, 11]. Patients were treated expressed as the mean (standard deviation) or median according to a unified protocol. During the baseline visit, (quartiles Q1 to Q3) according to normal or non-normal patients provided written consent for the use of their distributions. Normal distribution was assessed with a clinical data for research purposes. normal Q–Q plot. Locally weighted error sum of squares (Loess) curves were plotted for diabetic and non-diabetic Inclusion criteria patients and for pre-specified study subgroups. Loess The main inclusion criteria for this study were having regression is a non-parametric approach developed in had at least two echocardiography measurements, one at 1988 [13]. Loess curves are useful for observing a trend Julián et al. Cardiovasc Diabetol (2020) 19:38 Page 3 of 11 or relationship for non-linear data observed over time. 2662 measurements of HbA1c—on LVEF trajectory and Loess moves along the dataset looking at chunks at a outcomes was also performed within diabetic patients. time, fitting a set of local regression lines computed on Optimal glycemic control was considered when median observed data (missing values are omitted) and con- values of HbA1c for each patient were ≤ 7.5%. necting these lines to make a smooth line. Missing data Statistical analyses were performed using SPSS 24 due to loss to follow-up was assumed to be at random as (SPSS, Inc., Chicago, IL, USA) and R (A Language and there was no evidence that not attending the scheduled Environment for Statistical Computing) by the R Core visit had anything to do with LVEF. Locally weighted Team (2017; R Foundation for Statistical Computing, regression is based on a weight function, which gives the Vienna, Austria). For LME models, we used the nlme R greatest weight to observations that are closest to the package (version 3.1-131.1). Two-sided p < 0.05 was con- focal observation. Statistical analyses of LVEF change sidered statistically significant. The data, analytic meth- over time were performed using the linear mixed effects ods, and study materials will not be made available to (LME) model, which takes into account the group level other researchers for the purposes of reproducing the structure in the data by simultaneous assessing effects results or replicating the procedure. J.L. had full access to within and across groups. LME models incorporate both all data in the study and takes responsibility for its integ- fixed effects and random effects [14] and describe the rity and the data analysis. relationship between a response and covariates that have been observed along with the response [15]. In this study, Results LME models were developed to evaluate and compare A total of 1921 patients were admitted to the HF clinic the effect of time on LVEF change for diabetic and non- from August 2001 to December 2015, and 1656 had an diabetic patients and for pre-specified subgroups accord- LVEF < 50%. After the exclusion criteria were applied, ing to HF etiology, HF duration at baseline and sex. We 1160 patients had a minimum of two LVEF measure- hypothesized that there are important individual-level ments and comprised the study population [10]. Focus- effects and that patients have similar rates of change over ing on T2D patients, a range of 2 to 9 echocardiograms time. Thus, we fitted the “Random Intercept LME mod- were performed per patient, in a total of 1534 patients els,” where the measured value of LVEF is assumed to as shown in Additional file 1: Figure S1. The mean num- have a set of parameters fixed across individuals, includ- ber of echocardiograms performed in T2D patients ing a specific random effect per individual. Because the (3.3 ± 1.6) was less than that performed in non-diabet- form of the Loess curves suggested at least a quadratic in ics (3.8 ± 1.8; p < 0.001) and follow-up was also signifi- time, all LME models included both the linear term time cantly shorter in T2D patients (5.8 ± 3.7 vs. 7 ± 4.1 years; and the quadratic time ‘timeˆ2’ as fixed effects. By add- p < 0.001). ing the quadratic term ‘timeˆ2’ to the models, we evalu- Table 1 shows the clinical, biochemical, and echocardi- ated whether the effect of time significantly changed over ographic characteristics of the studied cohort and treat- time. The Wald Chi-squared test was applied to evalu- ment during follow-up based on the presence or absence ate the accuracy of the estimates of the effect of time on of T2D. The main etiology in diabetic patients was LVEF values. Comparisons of LVEF between groups were ischemic heart disease (64.6%) followed by dilated car- also performed at every timepoint of the study using diomyopathy (12.1%). Medical treatment was optimized the paired t-test. Comparisons between included and according to international guidelines. Although diabetic excluded patients and between alive and dead patients patients received less angiotensin-converting enzyme were performed using the Student’s t-test, Mann–Whit- inhibitor/angiotensin II receptor blockers treatment; in ney U test, or Chi-squared test as appropriate. Multi- contrast they received more mineralocorticoid receptor variable Cox regression analyses were performed with antagonist, loop diuretic, digoxin, and ivabradine treat- all-cause death and the composite all-cause death or ments. Comparison between included and not included HF hospitalization as the dependent variables and age, patients due to lack of a second echocardiogram to sex, diabetes mellitus, baseline LVEF, ischemic etiol- address group bias, has been reported elsewhere [10]. No ogy and NYHA functional class as covariates. Recurrent statistical differences were found in the variables mostly HF-related hospitalizations were analyzed as crude inci- associated with LVEF dynamics such as sex, etiology, dence rates (expressed as number of hospitalizations per baseline LVEF, and HF duration. There was also no statis- 100 patient-years) and by multivariable binomial nega- tically significant difference in the prevalence of diabetes: tive regression, expressed as incidence rate ratios (IRRs). included 461 (40%), excluded 118 (43.7%) (p = 0.23). Only for the later analyses, out-of-hospital death due to Baseline characteristics of the study population strati- HF was considered as an additional event. In a sensitiv- fied by HF etiology and diabetic status are presented in ity analysis the influence of diabetic control-based on Additional file 2: Table S1. Co-morbidities such as anemia Julián et al. Cardiovasc Diabetol (2020) 19:38 Page 4 of 11 Table 1 Demographic, clinical, and therapeutic characteristics of patients at baseline and treatments during follow-up Diabetics Non-diabetics p-value N N = 461 N = 699 Age, years 66.4 ± 10.5 64 ± 13.2 0.001 1160 Male 338 (73.3) 549 (78.5) 0.04 1160 White 459 (99.6) 692 (99.0) 0.20 1160 Etiology < 0.001 1160 Ischaemic heart disease 298 (64.6) 361 (51.6) Dilated cardiomyopathy 56 (12.1) 105 (15.0) Hypertensive 33 (7.2) 48 (6.9) Alcohol 16 (3.5) 52 (7.4) Drugs 7 (1.5) 27 (3.9) Valvular 23 (5.0) 49 (7.0) Other 28 (6.1) 57 (8.2) Previous AMI 238 (51.6) 306 (43.8) 0.009 HF duration, months 7 (2–39) 6 (1–41) 0.10 1160 NYHA class 0.001 1160 I 14 (3.0) 46 (6.6) II 318 (60.0) 499 (71.4) III 125 (27.1) 152 (21.7) IV 4 (0.9) 2 (0.3) LVEF, % 30.4 ± 8.4 30.3 ± 8.6 0.96 1160 LVEDD, mm 60.6 ± 7.9 61.8 ± 8.6 0.02 1043 LVESD, mm 48.4 ± 9.4 49.8 ± 9.6 0.02 1027 Hypertension 333 (72.2) 380 (54.4) < 0.001 1160 Anemiaa 232 (52.3) 248 (36.3) < 0.001 1127 Renal ­insufficiencyb 239 (52.2) 249 (36.1) < 0.001 1148 Atrial fibrillation/flutter 80 (17.4) 123 (17.6) 0.92 1160 LBBB 60 (13.0) 95 (13.6) 0.78 1160 Heart rate, bpm 72.2 ± 13.6 69.6 ± 15.1 0.003 1160 Blood pressure, mmHg 126.5 ± 21.9 124.8 ± 21.6 0.19 1160 BMI, kg/m2 27.7 (25–31.2) 26.5 (23.9–29.7) < 0.001 1156 NTproBNP, ng/L 1825 (808–4363) 1530 (611–3344) 0.01 714 HF treatments (F-U), n (%) ACEI or ARB 421 (91.3) 663 (94.8) 0.02 1160 Beta-blocker 436 (94.6) 658 (94.1) 0.75 1160 MRA 333 (72.2) 445 (63.7) 0.002 1160 Loop diuretic 441 (95.7) 621 (88.8) < 0.001 1160 Digoxin 213 (46.2) 264 (37.8) 0.004 1160 Ivabradine 113 (24.5) 122 (17.5) 0.003 1160 Sacubitril/valsartan 14 (3.0) 30 (4.3) 0.27 1160 CRT​ 23 (5.0) 45 (6.4) 0.30 1160 ICD 62 (13.4) 109 (15.6) 0.31 1160 Antidiabetic treatments Oral drugs baseline 248 (53.8) 461 Insulin baseline 175 (38.0) 461 Oral drugs F-U 367 (79.6) 461 Insulin F-U 305 (66.2) 461 Data in mean ± SD, median (IQR) or n (%) ACEI angiotensin converting enzyme inhibitor, ARB angiotensin II receptor blocker, BMI body mass index, CRT​cardiac resynchronization therapy, eGRF estimated glomerular renal filtration (CKD-EPI equation), F-U follow-up, HF heart failure, ICD implantable cardiac defibrillator, LBBB left bundle branch block, LVEF left ventricular ejection fraction, LVEDD left ventricular end-diastolic diameter, LVESD left ventricular end-systolic diameter, MRA mineralocorticoid receptor antagonist, NYHA New York Heart Association, NTproBNP N-terminal pro-brain natriuretic peptide a According to W.H.O. criteria (< 13 g/dL in men and < 12 g/dL in women) b eGFR (CKD-EPI equation) < 60 ml/min/1.73 m2 Julián et al. Cardiovasc Diabetol (2020) 19:38 Page 5 of 11 hypertension, and renal insufficiency were more preva- and non-ischemic etiologies, except regarding the use of lent in diabetic patients, both from ischemic and non- loop diuretics and cardiac devices. Mean LVEF values in ischemic etiologies. Differences between diabetic and diabetic patients were 30.4% ± 8.4 (n = 461), 37.9% ± 11.6 non-diabetic patients were generally similar in ischemic (n = 417), 40.7% ± 12.3 (n = 304), 41.9% ± 12.4 (n = 156), 43% ± 13.2 (n = 92), 42.9% ± 12.4 (n = 53), 42% ± 11.2 (n = 34), 39.2% ± 11 (n = 17), and 40.6% ± 11.7 (n = 12) at baseline and 1, 3, 5, 7, 9, 11, 13, and 15 years, respec- tively. Dynamic trajectories of LVEF based on the presence or absence of T2D are illustrated in Fig. 1. Paired data comparisons showed statistical differences between baseline and 1 year (p < 0.001), and between 1 and 3 years (p < 0.001; Additional file 2: Table S2). As a whole, Loess splines of long-term LVEF trajectories in diabetic patients showed a more pronounced inverted U shape with a marked rise in LVEF during the first year (slightly lower than that of non-diabetics), followed by a more pronounced and earlier LVEF decline (p for trajec- tory < 0.001 for both groups; p for interaction between LVEF trajectory and diabetes = 0.009). Figure 2a, b shows LVEF trajectories relative to the etiology of HF, ischemia versus non-ischemia. Long- term LVEF decline was more apparent in T2D ischemic Fig. 1 Loess spline curves of long-term LVEF trajectories. LVEF patients with HF (p for interaction < 0.001) (Fig. 2a). trajectory comparison between diabetic (orange) vs. non-diabetic An interaction between T2D and LVEF trajectory in (blue) patients. P for trajectory changes on LVEF < 0.001 for both patients with longer HF duration (p = 0.008; Additional groups. P for comparison between groups (interaction between file 3: Figure S2A; Additional file 4: Figure S2B) and in trajectory changes and diabetes) = 0.009. Shaded regions displayed males (p = 0.04; Additional file 5: Figure S3A) was also around curves represent the confidence interval at level = 0.95 observed. In T2D women, no differences were observed Fig. 2 Loess spline curves of long-term LVEF trajectories based on etiology. a Patients from ischemic aetiology of HF; diabetic (orange) vs. non-diabetic (blue) patients. P value for trajectory changes on LVEF < 0.001 for both groups. P for comparison between groups (interaction between trajectory changes and diabetes) < 0.001. Shaded regions displayed around curves represent the confidence interval at level = 0.95. b Patients from non-ischemic aetiologies. P value for trajectory changes on LVEF < 0.001 for both groups. P for comparison between groups (interaction between trajectory changes and diabetes) = 0.18. Shaded regions displayed around curves represent the confidence interval at level = 0.95 Julián et al. Cardiovasc Diabetol (2020) 19:38 Page 6 of 11 Fig. 3 Loess spline curves of long-term LVEF trajectories in diabetic patients based on clinical modifiers. a Aetiology: ischemic (blue) vs. non-ischemic (yellow) aetiology. P for trajectory changes on LVEF < 0.001 for both groups. P for comparison between groups < 0.001. Shaded regions displayed around curves represent the confidence interval at level = 0.95. b HF Duration: ≤ 12 months (blue) vs. > 12 months (yellow). P value for trajectory changes on LVEF < 0.001 for both groups. P for comparison between groups < 0.001. Shaded regions displayed around curves represent the confidence interval at level = 0.95. c Sex: women (blue) vs. men (yellow). P for trajectory changes on LVEF < 0.001 for both sexes. P for comparison between sexes < 0.001. Shaded regions displayed around curves represent the confidence interval at level = 0.95. d Optimal (blue) vs. non-optimal glycemic (yellow) control. P for trajectory changes on LVEF < 0.001. P for comparison between glycemic control groups 0.35. Shaded regions displayed around curves represent the confidence interval at level = 0.95 in the dynamic trajectories of LVEF (Additional file 6: decrease at the end of follow-up was observed in new- Figure S3B). onset HF patients (≤ 12 months) compared to patients Focusing on diabetic patients (Fig. 3a–c), non-ischemic with longer HF duration at the baseline visit (Fig. 3b; patients showed a significantly higher LVEF upslope p < 0.001); and women showed better LVEF improve- during essentially the entire trajectory (p < 0.001) com- ment than men during nearly all trajectories with a more pared to ischemic patients (Fig. 3a). A more pronounced pronounced inverted U-shape pattern (Fig. 3c; p = 0.03); increase in LVEF during the first year and absence of finally, for those patients with available measurements Julián et al. Cardiovasc Diabetol (2020) 19:38 Page 7 of 11 of HbA1c, no differences in LVEF trajectories were [17, 18]. The transient nature of LVEF recovery has been observed based on the optimal/non-optimal glycemic described previously in patients with depressed systolic control according to the predefined definition (Fig. 3d; function among T2D patients [17–20]. However, none p = 0.32). of these studies compared LVEF trajectories along fol- Causes of death according to the presence of diabetes low-up between HF diabetic and non-diabetic patients. are shown in Additional file 2: Table S3. As expected, Cioffi et al. [19], in a small prospective short-term study mortality was higher in diabetic patients, mainly due to that included patients with LVEF < 40% (22% had T2D), HF progression. Cox regression analyses for all-cause reported LVEF normalization in 22% of non-diabetic HF death and for the composite end-point of all-cause death patients, whereas in diabetic population the LVEF recov- or HF-related hospitalization are shown in Additional ery only occurred in 4% of them. The present is the first file 2: Table S4. Furthermore, a worse prognosis was to report the trajectory of LVEF over very long-term fol- seen in patients with T2D and ischemic heart disease, low-up in patients with T2D and HF compared to non- while non-diabetic patients with non-ischemic etiology diabetic HF patients. showed the best prognosis (Additional file 7: Figure S4A and Additional file 8: Figure S4B). Remarkably, diabetic Differences in long‑term LVEF trajectory between HF patients with non-ischemic etiology had a slightly worse patients according to T2D status prognosis than non-diabetic patients with ischemic etiol- The main finding of the current study was that LVEF ogy. So, the presence of T2D was associated with a higher trajectories in diabetic and non-diabetic patients were morbidity and mortality, regardless of whether HF was significantly different, as diabetic patients had a more of ischemic or non-ischemic etiology. Within diabetic pronounced inverted U shape of LVEF trajectory. This patients, in a sensitivity analysis, those with optimal gly- phenomenon suggests a component of diabetic cardio- cemic control showed better prognosis than those with myopathy. Although the presence of diabetes-related non-optimal control (HR 0.75 [0.55–1.02], p = 0.07 for cardiomyopathy is still controversial, a number of stud- all-cause death and HR 0.75 [0.57–0.97], p = 0.03 for ies in both animals and humans have shown structural the composite end-point respectively, after adjustment and functional changes of the diabetic heart, independ- by age, sex, baseline LVEF, ischemic etiology and NHYA ent of comorbid conditions such as coronary artery dis- functional class. Finally, diabetic patients showed a dou- ease, hypertension, or valvular heart disease [21]. It has ble number of recurrent admissions (23.4 vs. 11.2 per been proposed that the development of diabetic cardio- 100 patient-years, p < 0.001). In multivariable negative myopathy is likely to be multifactorial; several mecha- binomial regression analysis, adjusted IRR was 2.0 (1.15– nisms that include metabolic disturbances, myocardial 3.48), p = 0.02. Within diabetic patients we did not find fibrosis, small vessel disease, autonomic dysfunction, and significant relationship between the degree of glycemic insulin resistance have been implicated [8, 21]. Focus- control and recurrent HF admissions (IRR 0.69 [0.30– ing on structural alterations, the most prominent histo- 1.58], p = 0.37). pathological finding in diabetic patients is fibrosis, which result in anatomic and physiological changes in the Discussion myocardium. Abnormalities in myocardial metabolism, Large epidemiological studies have confirmed that T2D is such as impaired myocardial glucose uptake, have been an independent risk factor for the development of HF [3]. described in ischemic diabetic experimental models [21]. It is known that diabetic patients with HF have a higher The increased turnover of free fatty acids seen in diabetic all-cause mortality and cardiovascular mortality than patients may further impair myocardial glucose utiliza- non-diabetic HF patients [5, 6]. Moreover, following the tion and may have other deleterious consequences such first HF hospitalization, the incidence of new-onset dia- as changes in myocardial gene expression resulting in betes was around 2% per year, rising to 3% after 5 years myocyte hypertrophy with impaired contractile function of follow-up, and it was associated with an increased risk models [21, 22]. of death in established HF patients [16]. However, little On the other hand, it is acknowledged that altered LV is known about the comparison of LVEF trajectories over geometry may precede the development of T2D. A recent time between diabetic and non-diabetic patients. Focus- report found that LV concentric geometry determined ing on diabetic patients, the present study expands the by echocardiography and the severity of LV concen- prospective assessment of LVEF trajectories in a consec- tricity evaluated as relative wall thickness (RWT) were utive real-life cohort of HF patients with depressed sys- associated with incident diabetes in the general popula- tolic function and echocardiographic studies performed tion [23]. Furthermore, long-term metformin exposure over 15 years [10]. LVEF is a key predictor of survival and is associated with protective effects in terms of the inci- decreases over time are associated with reduced survival dence of new-onset symptomatic HFpEF, LV diastolic Julián et al. Cardiovasc Diabetol (2020) 19:38 Page 8 of 11 dysfunction and hypertrophy in patients with T2DM and in diabetic patients. This might have important clinical hypertension, which might be beneficial for the delay of implications, and further research is needed for a better HFpEF progression [24]. understanding. Yet, a clear direction is lacking on what to do in patients who experience a decline in LVEF over Factors influencing LVEF trajectory in T2D patients time despite optimal medical treatment. The interaction between LVEF trajectory and diabetes was remarkably centered in HF patients with ischemic Outcomes in patients with and without T2D etiology. We hypothesized that in ischemic diabetic HF Survival and event-free survival curves demonstrated patients, other mechanisms directly or indirectly related that patients with T2D and ischemic heart disease exhib- to diabetes such as specific diabetic cardiomyopathy ited the greatest morbidity and mortality after 15 years of may also be implicated. On the other hand, a close link follow-up. However, the presence of diabetes was asso- between ischemic heart disease and diabetic myocardi- ciated with higher events, regardless of whether HF was opathy may exist. Coronary artery atherosclerosis, a phe- of ischemic or non-ischemic origin. Some studies have nomenon associated with diabetes, is directly related to already described that the prognostic impact of diabetes myocardial ischemia, and consequently, impaired myo- in patients with HF is markedly influenced by the under- cardial glucose uptake, increased oxidative stress, and lying etiology and is particularly deleterious in ischemic vascular endothelial dysfunction, which may promote the heart disease models [4, 25–28]. In a large study pub- progression of diabetic cardiomyopathy. lished by Johansson et al. [28] that included 35.163 HF LVEF trajectories in diabetic patients are also influ- patients (24% had T2D), those with diabetes and ischemic enced by HF duration and sex. Regarding gender, bet- heart disease had the highest mortality. Remarkably, in ter LVEF dynamics were observed in women compared any of these studies, LVEF trajectory has been evaluated. to men, regardless of the presence or absence of diabe- tes. Men are at higher risk of developing left ventricular Future considerations systolic dysfunction. The exact underlying mechanism is Finally, one important question to address in the future unclear, but sex-related differences in cardiac remodeling is how sodium glucose co-transporter 2 (SGLT2) inhibi- and the protective effects of estrogen against apoptosis tors affect LVEF trajectories in T2D with HF. SGLT2 may be among the explanations. In addition, we found inhibitors act in renal glucose metabolism by inhib- an interaction between diabetes and LVEF trajectory iting glucose reabsorption and inducing glycosuria, in males; men with diabetes experienced a more pro- thereby reducing plasma glucose, blood pressure, and nounced and earlier LVEF decline than those without body weight [29]. Several studies in animal models have diabetes. On the contrary we did not find any relationship demonstrated their beneficial effects in left ventricu- between glycemic control and LVEF trajectory. There lar remodeling, cardiac fibrosis, and vascular dysfunc- have been conflicting reports regarding the importance tion [30, 31]. Four large cardiovascular outcomes trials of glycemic control in patients with T2D and HF. Our with SGLT2 inhibitors have shown important benefits results strongly suggest that other mechanisms rather in reducing HF hospitalization by about 30%, even in than on diabetic status also may explain the differences patients without T2D [32–35]. Several possible mecha- in long-term LVEF trajectory. We must take into account nisms have been postulated to explain these benefits of the fact that comorbidities commonly seen in patients SGLT2 inhibitors in HF, but they have not been con- with diabetes, such as hypertension, dyslipidemia, and firmed [36]. We recently used artificial intelligence to renal impairment, may accelerate the progression of car- describe that empagliflozin interacts and blocks the diac dysfunction towards advanced disease. In point of sodium-hydrogen exchanger co-transporter and might fact, the diabetic HF patients in our study had a higher ameliorate cardiomyocyte cell death [37]. In our cohort, percentage of hypertension and renal insufficiency than none of the patients were treated with SGLT2 inhibitors, the non-diabetics. so the results were not influenced by the possible effects Even considering a possible “survival effect,” very long- of these drugs. On the other hand, in the Prove-HF trial term survival was accompanied by a progressive LVEF [38], treatment with sacubitril-valsartan was associated decline, which seemed to be earlier and more pronounced with a significant improvement in LVEF during the first in diabetic patients. We usually continue neurohormonal year of treatment. In our cohort only 3.8% received such blockade treatment in all patients irrespective of LVEF treatment and none of them during the first year of fol- increase, as our data support that LVEF improvement low-up, so whether this treatment might influence LVEF represents myocardial remission rather than true myo- long-term trajectory is not elucidated yet. cardial cure, and this issue might be especially important Julián et al. Cardiovasc Diabetol (2020) 19:38 Page 9 of 11 Study limitations Conclusions This study had some limitations. Patients were classified LVEF in T2D with depressed systolic function showed into non-diabetic and diabetic subgroups according to an early improvement with a pronounced inverted U their baseline diagnosis and no data on new-onset T2D shape of trajectory and marked LVEF decline in the long diagnosis during follow-up were considered. We don’t term. This interaction was mainly observed in patients have complete data of glycemic control of the whole of ischemic etiology, in whom a component of diabetic period of the study and the whole cohort. HbA1c began cardiomyopathy might be present. In diabetic patients, to be routinely assessed around 2006. LVEF was assessed LVEF trajectories varied upon disease modifiers such as by transthoracic echocardiography in routine clinical HF etiology, HF duration, and sex. care. However, in the current registry, all echocardio- grams were scheduled prospectively and at pre-specified Supplementary information intervals and not at the discretion of the patient’s phy- Supplementary information accompanies this paper at https​://doi. sician, and were not retrospectively analyzed. We org/10.1186/s1293​3-020-01011​-w. acknowledge that intra- and inter-observer variability of echo-derived LVEF is ~ 5%. However, taking into account Additional file 1: Figure S1. Distribution of the number of echocardio‑ grams performed per patient. Number of echocardiograms per patient the large number of studies performed, we may assume ranged from 2 (minimum inclusion criteria) to 9 (patients with all the per that such variability was randomly distributed during protocol pre-specified echocardiograms). follow-up. On the other hand, contrast echocardiogra- Additional file 2: Table S1. Demographic, clinical, and therapeutic char‑ phy, longitudinal and radial strain analyses, 3-D echo- acteristics at baseline and treatments during follow-up according to etiol‑ cardiography and cardiac magnetic resonance imaging ogy of heart failure and presence of diabetes mellitus. Table S2. Paired wise means data analysis in diabetic patients. Table S3. Causes of death of would evaluate left ventricular function and volumes the studied cohort during the 15-year follow-up, according the presence more precisely and may be superior in the evaluation of or absence of diabetes mellitus. Table S4. Multivariable Cox regression LV remodeling parameters, but they are not broadly used analysis for all-cause death and the composite end-point all-cause death or heart failure hospitalization. in clinical practice. As in all reported studies of variations Additional file 3: Figure S2. Loess spline curves of long-term LVEF in LVEF during follow-up, our analyses were performed trajectories based on heart failure duration. Panel A: Patients with HF in “completers,” that is, patients with both baseline and at duration ≤ 12 months; Diabetic (orange) vs. non-diabetic (blue) patients. least other echocardiography study available for analysis. P value for trajectory changes on LVEF < 0.001 for both groups. P for comparison between groups (interaction between trajectory changes and We cannot fully exclude some bias in Loess spline curves diabetes) = 0.22. Shaded regions displayed around curves represent the due to dropout, as we could not statistically distinguish confidence interval at level = 0.95. between autonomous time trends and pseudo upward Additional file 4: Figure S2. Loess spline curves of long-term LVEF trends because of successive drop out of fatalities with trajectories based on heart failure duration. Panel B: Patients with HF duration > 12 months. P value for trajectory changes on LVEF < 0.001 for lower initial LVEF values. Missing data because of loss both groups. P for comparison between groups (interaction between time to follow-up was assumed to be random. Furthermore, and diabetes) = 0.008. Shaded regions displayed around curves represent Loess spline curve estimations at the end of follow-up the confidence interval at level = 0.95. were less robust due to the limited number of patients. Additional file 5: Figure S3. Loess spline curves of long-term LVEF The study cohort was a general HF population treated trajectories based on sex. Panel A: Men. Diabetic (orange) vs. non-diabetic (blue) patients. P value for trajectory changes on LVEF < 0.001 for both at a specific multidisciplinary HF clinic in a tertiary care groups. P for comparison between groups (interaction between trajectory hospital, with most patients referred from the cardiology changes and diabetes) = 0.04. Shaded regions displayed around curves department; thus, there was a predominance of relatively represent the confidence interval at level = 0.95. young men with HF of ischemic etiology, and an almost Additional file 6: Figure S3. Loess spline curves of long-term LVEF trajectories based on sex. Panel B: Women. P value for trajectory changes exclusively white population, so it may not be possible on LVEF <0.001 for both groups. P for comparison between groups (inter‑ to fully extrapolate the findings to other populations. Of action between trajectory changes and diabetes) = 0.14. Shaded regions note, a common treatment protocol was applied to all displayed around curves represent the confidence interval at level = 0.95. patients, thereby limiting possible bias introduced by dif- Additional file 7: Figure S4. Survival and event-free survival curves related to the presence of diabetes mellitus and to etiology (ischemic vs. ferent management strategies or treatment protocols. non-ischemic). Panel A: All-cause death survival curves. To provide more insight into understanding of the Additional file 8: Figure S4. Survival and event-free survival curves mechanisms involved in the “U-effect curve” observed related to the presence of diabetes mellitus and to etiology (ischemic vs. in T2D patients, experimental models might be useful. non-ischemic). Panel B: Event-free survival curves (composite end-point These studies might contribute to assess the effect of of all-cause death or HF hospitalizations). Diabetic patients from ischemic etiology (dark purple) showed the worse prognosis, while non-diabetic hyperglycemia and insulin-resistance in the pathogenesis from non-ischemic etiology (blue) showed the best. Remarkably diabetic of diabetic cardiomyopathy. patients from non-ischemic etiology (soft orange) showed slightly worse prognosis than non-diabetic patients from ischemic etiology (green). Julián et al. Cardiovasc Diabetol (2020) 19:38 Page 10 of 11 contribution of the Heart Failure Association (HFA) of the ESC. Eur Heart J. Abbreviations 2016;14(37):2129–200. HF: Heart failure; LME: Linear Mixed-Effects; Loess: Locally weighted error sum 3. Kannel WB, McGee DL. Diabetes and cardiovascular disease. The Framing‑ of squares; LVEF: Left ventricular ejection fraction; T2D: Type 2 diabetes. ham study. JAMA. 1979;241:2035–8. 4. De Groote P, Lamblin N, Mouquet F, Plichon D, McFadden E, Van Belle E, Acknowledgements Bauter C. Impact of diabetes mellitus on long-term survival in patients We thank Beatriz González, Roser Cabanes, Margarita Rodríguez, Carmen with congestive heart failure. Eur Heart J. 2004;25:656–62. Rivas, Violeta Díaz, Nuria Benito, and Alba Ros, nurses in the HF unit, for data 5. Rawshani A, Rawshani A, Franzén S, Sattar N, Eliasson B, Svensson AM, collection and their invaluable work in the unit. We also thank physicians Nuria et al. Risk factors, mortality, and cardiovascular outcomes in patients with Vallejo, Francisco Gual, Albert Teis, and Gladys Juncà from the Echocardiogra‑ type 2 diabetes. N Engl J Med. 2018;379:633–44. phy Laboratory of the Cardiology Service for their valuable daily work. We also 6. MacDonald MR, Petrie MC, Varyani F, Ostergren J, Michelson EL, Young JB, wish to acknowledge Redes Temáticas de Investigación Cooperativa en Salud et al. Impact of diabetes on outcomes in patients with low and preserved (RETICS) and Red Cardiovascular (RD12/0042/0047). ejection fraction heart failure: an analysis of the Candesartan in Heart failure: assessment of reduction in mortality and morbidity (CHARM) Authors’ contributions programme. Eur Heart J. 2008;29:1377–85. All authors contributed significantly to the work. MTJ, NA, JL and ABG con‑ 7. Shindler DM, Kostis JB, Yusuf S, Quinones MA, Pitt B, Stewart D, et al. ceived the study and design the manuscript. JL, EF, MA, JLA, MD, ESV, EZ, PC, Diabetes mellitus, a predictor of morbidity and mortality in the Studies PM, JS, ABG contributed to data collection. Statistical analysis was carried out of Left Ventricular Dysfunction (SOLVD) Trials and Registry. Am J Cardiol. by GC and JL. The manuscript was critically revised by MTJ, NA, JL, EF, MA, JLA, 1996;77:1017–20. MD, ESV, EZ, PC, PM, JN, JS, MPD, ABG. All authors read and approved the final 8. Dei Cas A, Khan SS, Butler J, Mentz RJ, Bonow RO, Avogaro A, et al. Impact manuscript. of diabetes on epidemiology, treatment, and outcomes of patients with heart failure. JACC Heart Fail. 2015;3:136–45. Funding 9. Lupón J, Díez-López C, de Antonio M, Domingo M, Zamora E, Moliner Not applicable. P, et al. Recovered heart failure with reduced ejection fraction and out‑ comes: a prospective study. Eur J Heart Fail. 2017;19:1615–23. Availability of data and materials 10. Lupón J, Gavidia-Bovadilla G, Ferrer E, de Antonio M, Perera-Lluna A, The datasets used and/or analysed during the current study are available from López-Ayerbe J, et al. Dynamic trajectories of left ventricular ejection frac‑ the corresponding author on reasonable request. tion in heart failure. J Am Coll Cardiol. 2018;72:591–601. 11. Zamora E, Lupón J, Vila J, et al. Estimated glomerular filtration rate and Ethics approval and consent to participate prognosis in heart failure: value of the modification of diet in renal All participants gave their informed consent, and this study was performed disease study-4, chronic kidney disease epidemiology collaboration, and in compliance with the Helsinki Declaration, and was approved by the local cockroft-gault formulas. J Am Coll Cardiol. 2012;59:1709–15. Ethics Committee. 12. American Diabetes Association. Classification and diagnosis of diabetes. Diabetes Care. 2015;38(Suppl 1):S8–16. Consent for publication 13. Cleveland WS, Devlin SJ. Locally weighted regression: an approach to Not applicable. regression analysis by local fitting. J Am Stat Assoc. 1988;83:596–610. 14. Bates DM. lme4: mixed-effects modeling with R. Springer, 2010. http:// Competing interests lme4.rforg​e.r-proje​ct.org/. Accessed Jan 2016. The authors declare that they have no competing interests. 15. Carey VJ, Wang Y-G. Book review: mixed-effects models in S and S-plus. J Am Stat Assoc. 2001;96:1135. Author details 16. Zareni B, Rørth R, Mogensen UM, Selmer C, Gislason G, Torp-Pedersen 1 Heart Failure Unit, Hospital Universitari Germans Trias i Pujol, Badalona, Barce‑ C. Heart failiure and the prognostic impact and incidence of new-onset lona, Spain. 2 Endocrinology Department, Hospital Universitari Germans Trias i of diabetes mellitus: a nationwide cohort study. Cardiovasc Diabetol. Pujol, Badalona, Barcelona, Spain. 3 Centre of Biomedical Research on Diabetes 2019;18:79. and Associated Metabolic Diseases (CIBERDEM), Instituto de Salud Carlos III, 17. Dunlay SM, Roger VL, Weston SA, Jiang R, Redfield MM. Longitudinal Madrid, Spain. 4 Cardiology Department, Hospital Universitari Germans Trias changes in ejection fraction in heart failure patients with preserved and i Pujol, Badalona, Barcelona, Spain. 5 Department of Medicine, Universitat reduced ejection fraction. Circ Heart Fail. 2012;5:720–6. Autònoma de Barcelona, Barcelona, Spain. 6 CIBER Cardiovascular, Instituto de 18. de Groote P, Fertin M, Duva Pentiah A, Goéminne C, Lamblin N, Bauters C. Salud Carlos III, Madrid, Spain. 7 Department of e‑Health, Eurecat, Technologi‑ Long-term functional and clinical follow-up of patients with heart failure cal Center of Catalonia, Barcelona, Spain. 8 Cardiology Department, Hospital with recovered left ventricular ejection fraction after β-blocker therapy. Clínico Universitario, INCLIVA, Valencia, Spain. 9 Department of Medicine, Circ Heart Fail. 2014;7:434–9. Universitat de València, Valencia, Spain. 19. Cioffi G, Stefenelli C, Tarantini L, Opasich C. Chronic left ventricular failure in the community: prevalence, prognosis, and predictors of the complete Received: 11 January 2020 Accepted: 7 March 2020 clinical recovery with return of cardiac size and function to normal in patients undergoing optimal therapy. J Card Fail. 2004;10:250–7. 20. Merlo M, Stolfo D, Anzini M, et al. Persistent recovery of normal left ventricular function and dimension in idiopathic dilated cardiomyopathy during long-term follow-up: does real healing exist? J Am Heart Assoc. References 2015;4:e001504. 1. Yancy CW, Jessup M, Bozkurt B, Butler J, Casey DE Jr, Drazner MH, Ameri‑ 21. Fang ZY, Prins JB, Marwick TH. Diabetic cardiomyopathy: evidence, can College of Cardiology Foundation, American Heart Association Task mechanisms, and therapeutic implications. Endocr Rev. 2004;25:543–67. Force on Practice Guidelines, et al. 2013 ACCF/AHA guideline for the 22. Young ME, McNulty P, Taegtmeyer H. Adaptation and maladaption management of heart failure: a report of the American College of Cardiol‑ of the heart in diabetes: part II: potential mechanism. Circulation. ogy Foundation/American Heart Association Task Force on Practice 2002;105:1861–70. Guidelines. J Am Coll Cardiol. 2013;2013(62):e147–239. 23. Modin D, Møgelvang R, Jørgensen PG, Jensen MT, Seferovic JP, Biering- 2. Ponikowski P, Voors AA, Anker SD, Bueno H, Cleland JGF, Falk V, Authors/ Sørensen T. Left ventricular concentric geometry predicts incident Task Force Members, Document Reviewers, et al. 2016 ESC guidelines for diabetes mellitus independent of established risk factors in the general the diagnosis and treatment of acute and chronic heart failure: The Task population: the Copenhagen City Heart Study. Cardiovasc Diabetol. Force for the diagnosis and treatment of acute and chronic heart failure 2019;18:37. of the European Society of Cardiology (ESC). Developed with the special Julián et al. Cardiovasc Diabetol (2020) 19:38 Page 11 of 11 24. Gu J, Yin ZF, Zhang JF, Wang CQ. Association between long-term prescrip‑ 33. Neal B, Perkovic V, Mahaffey KW, de Zeeuw D, Fulcher G, Erondu N, Shaw tion of metformin and the progression of heart failure with preserved W, Law G, Desai M, Matthews DR, CANVAS Program Collaborative Group. ejection fraction in patients with type 2 diabetes mellitus and hyperten‑ Canagliflozin and cardiovascular and renal events in type 2 diabetes. N sion. Int J Cardiol. 2019. https​://doi.org/10.1016/j.ijcar​d.2019.11.087. Engl J Med. 2017;377:644–57. 25. Lopaschuk GD, Ussher JR, Folmes CD, Jaswal JS, Stanley WC. Myocardial 34. Wiviott SD, Raz I, Bonaca MP, Mosenzon O, Kato ET, Cahn A, et al. Dapa‑ fatty acid metablism in health and disease. Physiol Rev. 2010;90:207–58. gliflozin and cardiovascular outcomes in type 2 diabetes. N Engl J Med. 26. Dries DL, Sweitzer NK, Drazner MH, Stevenson LW, Gersh BJ. Prognostic 2019;380:347–57. impact of diabetes mellitus in patients with heart failure according to 35. McMurray JJV, Solomon SD, Inzucchi SE, Køber L, Kosiborod MN, Martinez the etiology of left ventricular systolic dysfunction. J Am Coll Cardiol. FA, et al. Dapagliflozin in patients with heart failure and reduced ejection 2001;38:421–8. fraction. N Engl J Med. 2019;38:1995–2008. 27. Domanski M, Krause-Steinrauf H, Deedwania P, Follmann D, Ghali JK, 36. Verma S, McMurray JJV. SGLT2 inhibitors and mechanisms of cardiovascu‑ Gilbert E, et al. The effect of diabetes on outcomes of patients with lar benefit: a state-of-the-art review. Diabetologia. 2018;61:2108–17. advanced heart failure in the BEST trial. J Am Coll Cardiol. 2003;42:914–22. 37. Oriol I, Santiago-Vacas E, Yurista S, Lupón P, Packer M, Heymans S, et al. 28. Johansson I, Dahlström U, Edner M, Näsman P, Rydén L, Norhammar Unraveling the molecular mechanism of action of empagliflozin in heart A. Prognostic implications of type 2 diabetes mellitus in ischemic and failure with reduced ejection fraction with or without diabetes. JACC nonischemic heart failure. J Am Coll Cardiol. 2016;68:1404–16. Basic Transl Sci. 2019. https​://doi.org/10.1016/j.jacbt​s.2019.07.010. 29. Hasan FM, Alsahli M, Gerich JE. SGLT2 inhibitors in the treatment of type 2 38. Januzzi JL Jr, Prescott MF, Butler J, Felker GM, Maisel AS, McCague K, et al. diabetes. Diabetes Res Clin Pract. 2014;104:297–322. Association of change in N-Terminal Pro-B-Type natriuretic peptide fol‑ 30. Lin B, Koibuchi N, Hasegawa Y, Sueta D, Toyama K, Uekawa K, et al. lowing initiation of sacubitril–valsartan treatment with cardiac structure Glycemic control with empagliflozin, a novel selective SGLT2 inhibitor, and function in patients with heart failure with reduced ejection fraction. ameliorates cardiovascular injury and cognitive dysfunction in obese and JAMA. 2019;2:1–11. type 2 diabetic mice. Cardiovasc Diabetol. 2014;13:148. 31. Zhang N, Feng B, Ma X, Sun K, Xu G, Zhou Y. Dapagliflozin improves left ventricular remodeling and aorta sympathetic tone in a pig model Publisher’s Note of heart failure with preserved ejection fraction. Cardiovasc Diabetol. Springer Nature remains neutral with regard to jurisdictional claims in pub‑ 2019;18:107. lished maps and institutional affiliations. 32. Zinman B, Wanner C, Lachin JM, Fitchett D, Bluhmki E, Hantel S, et al. Empagliflozin, cardiovascular outcomes, and mortality in type 2 diabetes. N Engl J Med. 2015;373:2117–28. Ready to submit your research ? Choose BMC and benefit from: • fast, convenient online submission • thorough peer review by experienced researchers in your field • rapid publication on acceptance • support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations • maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions