Long-term LVEF trajectories in patients with type 2 diabetes and heart failure: diabetic cardiomyopathy may underlie functional decline

This study is a T2D-focused post hoc sub analysis of a previously reported cohort [10]. All consecutive ambulatory patients referred to a structured multidisciplinary HF clinic of a university hospital between August 2001 and December 2015, regardless of etiology, were considered for the study. During the 15-year period, clinical pathways and referral geographical area, encompassing ~ 850,000 inhabitants in the northern Barcelona Metro Area, remained stable. The criteria for referral to the HF clinic were HF with at least one hospitalization and/or depressed systolic function [10, 11]. Patients were treated according to a unified protocol. During the baseline visit, patients provided written consent for the use of their clinical data for research purposes.

The main inclusion criteria for this study were having had at least two echocardiography measurements, one at baseline and the other during follow-up. Patients undergoing heart transplantation or cardiac resynchronization therapy after the second echocardiography were censored at the time of the intervention. The study was performed in compliance with the law protecting personal data in accordance with the international guidelines on clinical investigation of the World Medical Association’s Declaration of Helsinki.

The primary outcome of the study was prospectively assess very long-term (up to 15 years) LVEF trajectories in patients with and without T2D and underlying HF. All patients were seen regularly during follow-up visits at the HF clinic according to their clinical needs and treated according to a unified protocol [10, 11]. Follow-up visits included a minimum of 1 visit with a nurse every 3 months and 1 visit with a physician (cardiologist, internist, or family physician) every 6 months, as well as optional visits with specialists in geriatrics, psychiatry, and rehabilitation. During the baseline visit, patients provided written consent for the use of their clinical data for research purposes [10, 11].

LVEF was prospectively evaluated at baseline and at 1, 3, 5, 7, 9, 11, 13, and 15 years of follow-up using two-dimensional echocardiography by image expert cardiologists [10]. LVEF was obtained from apical two- and four-chamber views and was calculated using the Simpson’s method. All echocardiograms were revised for accuracy by expert staff.

A diagnosis of T2D was made when at least one of the following criteria was met: (1) a diagnosis of T2D was previously established and recorded in the patient’s electronic History; (2) fasting plasma glucose ≥ 126 mg/dL or hemoglobin A1C ≥ 6.5% identified by laboratory testing [12]; or (3) had a current prescription for oral hypoglycemic medication or insulin.

Categorical variables were expressed as absolute numbers and percentages. Continuous variables were expressed as the mean (standard deviation) or median (quartiles Q1 to Q3) according to normal or non-normal distributions. Normal distribution was assessed with a normal Q–Q plot. Locally weighted error sum of squares (Loess) curves were plotted for diabetic and non-diabetic patients and for pre-specified study subgroups. Loess regression is a non-parametric approach developed in 1988 [13]. Loess curves are useful for observing a trend or relationship for non-linear data observed over time. Loess moves along the dataset looking at chunks at a time, fitting a set of local regression lines computed on observed data (missing values are omitted) and connecting these lines to make a smooth line. Missing data due to loss to follow-up was assumed to be at random as there was no evidence that not attending the scheduled visit had anything to do with LVEF. Locally weighted regression is based on a weight function, which gives the greatest weight to observations that are closest to the focal observation. Statistical analyses of LVEF change over time were performed using the linear mixed effects (LME) model, which takes into account the group level structure in the data by simultaneous assessing effects within and across groups. LME models incorporate both fixed effects and random effects [14] and describe the relationship between a response and covariates that have been observed along with the response [15]. In this study, LME models were developed to evaluate and compare the effect of time on LVEF change for diabetic and non-diabetic patients and for pre-specified subgroups according to HF etiology, HF duration at baseline and sex. We hypothesized that there are important individual-level effects and that patients have similar rates of change over time. Thus, we fitted the “Random Intercept LME models,” where the measured value of LVEF is assumed to have a set of parameters fixed across individuals, including a specific random effect per individual. Because the form of the Loess curves suggested at least a quadratic in time, all LME models included both the linear term time and the quadratic time ‘timeˆ2’ as fixed effects. By adding the quadratic term ‘timeˆ2’ to the models, we evaluated whether the effect of time significantly changed over time. The Wald Chi-squared test was applied to evaluate the accuracy of the estimates of the effect of time on LVEF values. Comparisons of LVEF between groups were also performed at every timepoint of the study using the paired t-test. Comparisons between included and excluded patients and between alive and dead patients were performed using the Student’s t-test, Mann–Whitney U test, or Chi-squared test as appropriate. Multivariable Cox regression analyses were performed with all-cause death and the composite all-cause death or HF hospitalization as the dependent variables and age, sex, diabetes mellitus, baseline LVEF, ischemic etiology and NYHA functional class as covariates. Recurrent HF-related hospitalizations were analyzed as crude incidence rates (expressed as number of hospitalizations per 100 patient-years) and by multivariable binomial negative regression, expressed as incidence rate ratios (IRRs). Only for the later analyses, out-of-hospital death due to HF was considered as an additional event. In a sensitivity analysis the influence of diabetic control-based on 2662 measurements of HbA1c—on LVEF trajectory and outcomes was also performed within diabetic patients. Optimal glycemic control was considered when median values of HbA1c for each patient were ≤ 7.5%.

Statistical analyses were performed using SPSS 24 (SPSS, Inc., Chicago, IL, USA) and R (A Language and Environment for Statistical Computing) by the R Core Team (2017; R Foundation for Statistical Computing, Vienna, Austria). For LME models, we used the nlme R package (version 3.1-131.1). Two-sided p < 0.05 was considered statistically significant. The data, analytic methods, and study materials will not be made available to other researchers for the purposes of reproducing the results or replicating the procedure. J.L. had full access to all data in the study and takes responsibility for its integrity and the data analysis.

The main finding of the current study was that LVEF trajectories in diabetic and non-diabetic patients were significantly different, as diabetic patients had a more pronounced inverted U shape of LVEF trajectory. This phenomenon suggests a component of diabetic cardiomyopathy. Although the presence of diabetes-related cardiomyopathy is still controversial, a number of studies in both animals and humans have shown structural and functional changes of the diabetic heart, independent of comorbid conditions such as coronary artery disease, hypertension, or valvular heart disease [21]. It has been proposed that the development of diabetic cardiomyopathy is likely to be multifactorial; several mechanisms that include metabolic disturbances, myocardial fibrosis, small vessel disease, autonomic dysfunction, and insulin resistance have been implicated [8, 21]. Focusing on structural alterations, the most prominent histopathological finding in diabetic patients is fibrosis, which result in anatomic and physiological changes in the myocardium. Abnormalities in myocardial metabolism, such as impaired myocardial glucose uptake, have been described in ischemic diabetic experimental models [21]. The increased turnover of free fatty acids seen in diabetic patients may further impair myocardial glucose utilization and may have other deleterious consequences such as changes in myocardial gene expression resulting in myocyte hypertrophy with impaired contractile function models [21, 22].

On the other hand, it is acknowledged that altered LV geometry may precede the development of T2D. A recent report found that LV concentric geometry determined by echocardiography and the severity of LV concentricity evaluated as relative wall thickness (RWT) were associated with incident diabetes in the general population [23]. Furthermore, long-term metformin exposure is associated with protective effects in terms of the incidence of new-onset symptomatic HFpEF, LV diastolic dysfunction and hypertrophy in patients with T2DM and hypertension, which might be beneficial for the delay of HFpEF progression [24].

The interaction between LVEF trajectory and diabetes was remarkably centered in HF patients with ischemic etiology. We hypothesized that in ischemic diabetic HF patients, other mechanisms directly or indirectly related to diabetes such as specific diabetic cardiomyopathy may also be implicated. On the other hand, a close link between ischemic heart disease and diabetic myocardiopathy may exist. Coronary artery atherosclerosis, a phenomenon associated with diabetes, is directly related to myocardial ischemia, and consequently, impaired myocardial glucose uptake, increased oxidative stress, and vascular endothelial dysfunction, which may promote the progression of diabetic cardiomyopathy.

LVEF trajectories in diabetic patients are also influenced by HF duration and sex. Regarding gender, better LVEF dynamics were observed in women compared to men, regardless of the presence or absence of diabetes. Men are at higher risk of developing left ventricular systolic dysfunction. The exact underlying mechanism is unclear, but sex-related differences in cardiac remodeling and the protective effects of estrogen against apoptosis may be among the explanations. In addition, we found an interaction between diabetes and LVEF trajectory in males; men with diabetes experienced a more pronounced and earlier LVEF decline than those without diabetes. On the contrary we did not find any relationship between glycemic control and LVEF trajectory. There have been conflicting reports regarding the importance of glycemic control in patients with T2D and HF. Our results strongly suggest that other mechanisms rather than on diabetic status also may explain the differences in long-term LVEF trajectory. We must take into account the fact that comorbidities commonly seen in patients with diabetes, such as hypertension, dyslipidemia, and renal impairment, may accelerate the progression of cardiac dysfunction towards advanced disease. In point of fact, the diabetic HF patients in our study had a higher percentage of hypertension and renal insufficiency than the non-diabetics.

Even considering a possible “survival effect,” very long-term survival was accompanied by a progressive LVEF decline, which seemed to be earlier and more pronounced in diabetic patients. We usually continue neurohormonal blockade treatment in all patients irrespective of LVEF increase, as our data support that LVEF improvement represents myocardial remission rather than true myocardial cure, and this issue might be especially important in diabetic patients. This might have important clinical implications, and further research is needed for a better understanding. Yet, a clear direction is lacking on what to do in patients who experience a decline in LVEF over time despite optimal medical treatment.

Survival and event-free survival curves demonstrated that patients with T2D and ischemic heart disease exhibited the greatest morbidity and mortality after 15 years of follow-up. However, the presence of diabetes was associated with higher events, regardless of whether HF was of ischemic or non-ischemic origin. Some studies have already described that the prognostic impact of diabetes in patients with HF is markedly influenced by the underlying etiology and is particularly deleterious in ischemic heart disease models [4, 25‐28]. In a large study published by Johansson et al. [28] that included 35.163 HF patients (24% had T2D), those with diabetes and ischemic heart disease had the highest mortality. Remarkably, in any of these studies, LVEF trajectory has been evaluated.

Finally, one important question to address in the future is how sodium glucose co-transporter 2 (SGLT2) inhibitors affect LVEF trajectories in T2D with HF. SGLT2 inhibitors act in renal glucose metabolism by inhibiting glucose reabsorption and inducing glycosuria, thereby reducing plasma glucose, blood pressure, and body weight [29]. Several studies in animal models have demonstrated their beneficial effects in left ventricular remodeling, cardiac fibrosis, and vascular dysfunction [30, 31]. Four large cardiovascular outcomes trials with SGLT2 inhibitors have shown important benefits in reducing HF hospitalization by about 30%, even in patients without T2D [32‐35]. Several possible mechanisms have been postulated to explain these benefits of SGLT2 inhibitors in HF, but they have not been confirmed [36]. We recently used artificial intelligence to describe that empagliflozin interacts and blocks the sodium-hydrogen exchanger co-transporter and might ameliorate cardiomyocyte cell death [37]. In our cohort, none of the patients were treated with SGLT2 inhibitors, so the results were not influenced by the possible effects of these drugs. On the other hand, in the Prove-HF trial [38], treatment with sacubitril-valsartan was associated with a significant improvement in LVEF during the first year of treatment. In our cohort only 3.8% received such treatment and none of them during the first year of follow-up, so whether this treatment might influence LVEF long-term trajectory is not elucidated yet.

This study had some limitations. Patients were classified into non-diabetic and diabetic subgroups according to their baseline diagnosis and no data on new-onset T2D diagnosis during follow-up were considered. We don’t have complete data of glycemic control of the whole period of the study and the whole cohort. HbA1c began to be routinely assessed around 2006. LVEF was assessed by transthoracic echocardiography in routine clinical care. However, in the current registry, all echocardiograms were scheduled prospectively and at pre-specified intervals and not at the discretion of the patient’s physician, and were not retrospectively analyzed. We acknowledge that intra- and inter-observer variability of echo-derived LVEF is ~ 5%. However, taking into account the large number of studies performed, we may assume that such variability was randomly distributed during follow-up. On the other hand, contrast echocardiography, longitudinal and radial strain analyses, 3-D echocardiography and cardiac magnetic resonance imaging would evaluate left ventricular function and volumes more precisely and may be superior in the evaluation of LV remodeling parameters, but they are not broadly used in clinical practice. As in all reported studies of variations in LVEF during follow-up, our analyses were performed in “completers,” that is, patients with both baseline and at least other echocardiography study available for analysis. We cannot fully exclude some bias in Loess spline curves due to dropout, as we could not statistically distinguish between autonomous time trends and pseudo upward trends because of successive drop out of fatalities with lower initial LVEF values. Missing data because of loss to follow-up was assumed to be random. Furthermore, Loess spline curve estimations at the end of follow-up were less robust due to the limited number of patients. The study cohort was a general HF population treated at a specific multidisciplinary HF clinic in a tertiary care hospital, with most patients referred from the cardiology department; thus, there was a predominance of relatively young men with HF of ischemic etiology, and an almost exclusively white population, so it may not be possible to fully extrapolate the findings to other populations. Of note, a common treatment protocol was applied to all patients, thereby limiting possible bias introduced by different management strategies or treatment protocols.

To provide more insight into understanding of the mechanisms involved in the “U-effect curve” observed in T2D patients, experimental models might be useful. These studies might contribute to assess the effect of hyperglycemia and insulin-resistance in the pathogenesis of diabetic cardiomyopathy.