The main findings of this study can be summarized as follows: 1. Diabetic patients with HFpEF show specific characteristics, including higher body mass index, lower prevalence of atrial fibrillation, lower hemoglobin levels and worse renal function. No echocardiographic difference could be detected, but CMR showed a trend towards higher LV mass and more myocardial fibrosis (ECV > 33%). 2. Diabetes is associated with an excess of adverse events (hospitalization for HF and mortality) in HFpEF. 3. Lower levels of HbA1C levels are associated with worse prognosis in diabetic patients with HFpEF.
Characteristics and outcome of diabetic versus nondiabetic HFpEF patients
Regarding clinical characteristics, HFpEF patients with diabetes were younger and more obese than nondiabetic patients. This is consistent with sub studies from large clinical trials [
6‐
8]. A large study examining age-related characteristics in HFpEF also observed that younger patients were more than twice as likely to be obese, and that the prevalence of diabetes ranged from 37% in the younger group versus 18% in the oldest group [
19]. Although the reason for this difference is not completely elucidated, it might reflect that different pathophysiological pathways can lead to the development of HFpEF. The combination of diabetes and obesity, both conditions associated with a release of proinflammatory cytokines and decreased nitric oxide availability, could lead to the development of HFpEF at a younger age through myocardial remodelling and fibrosis [
20]. Supporting this, diabetic patients also exhibited a trend towards higher LV masses and higher levels of myocardial fibrosis than their nondiabetic counterparts, consistently with previous studies [
7,
21,
22]. This can contribute to the worse prognosis conferred by diabetes, as we previously showed that extracellular matrix expansion (higher ECV by CMR) was associated with adverse events in HFpEF [
17].
Atrial fibrillation, on the other hand, was more prevalent in the nondiabetic group. This is consistent with previously published literature [
23‐
25]. AF and HFpEF often coexist and it is still unclear whether one affection leads sequentially to the other. More likely, the two disorders share a common mechanistic substrate, which causes AF and HFpEF [
26,
27] and develop in parallel. A recent meta-analysis underlined that AF was associated with poor prognosis in HFpEF, although it is unclear whether AF is only a marker of more severe heart failure, or a cause of mortality in itself [
28]. Atrial fibrillation is also an age-related marker, hence, it is not surprising that the prevalence of AF is higher in the older nondiabetic group. Studies have also suggested differences in cardiac remodelling, with diabetic patients showing smaller LA volumes, that might contribute to this phenomenon [
23]. However, the presence of AF was retrieved from medical files, patients interrogation, and a standard electrocardiogram at inclusion, but no long term rhythm monitoring was performed. As such, the prevalence of AF and other arrhythmias could have been underestimated in both groups.
The event rate in our study was high compared to clinical trials (16.1 / 100 persons-year overall mortality in the diabetic group versus 6.8–8.8 in pooled data from I-Preserve, Charm-Preserved and TOPCAT [
23]), but similar to a large community based study (15.2/100 persons-year [
29]). Compared to clinical trials, our population is almost 10 years older (76 vs 69 years) had higher NT-proBNP levels (1745 vs 430–581 pg/mL), lower hemoglobin (11 vs 12.9–13.5 g/dL) and worse renal function (50 vs 62.7–71.4 mL/min/1.73 m
2), all parameters associated with adverse events. The association between diabetic status and prognosis (hospitalization for HF and mortality) is consistent with the existing literature [
6,
7,
9,
10,
23]. There are numerous pathophysiologic processes in diabetes that are thought to alter the myocardium resulting in less effective relaxation and contraction, including oxidative stress, inflammation and disorders in calcium transport, as well as alterations in substrate metabolism, and mitochondrial dysfunction [
30,
31]. Furthermore, extra-cardiac effects of diabetes such as decreased arterial compliance, renal angiopathy, and autonomic dysfunction can also accelerate the progression of HFpEF [
31]. In particular, hyperglycemia causes up-regulation of the sodium-glucose cotransporter-2 (SGLT-2) leading to increased proximal renal sodium absorption, volume expansion, and decreased responsiveness to diuretics [
20,
32,
33]. A better understanding of the interplay between diabetes and HF is crucial for the development of new therapies. This has recently been emphasized by the promising results of studies using SGLT-2 inhibitors in diabetic patients with HF [
34,
35]. The results of ongoing randomized controlled trials using SGLT-2i in HFpEF [
4,
20] are eagerly awaited. Nevertheless, a retrospective study showed less impressive effects of SGLT2i on cardiac remodeling in HFpEF compared to HFrEF, tempering enthusiasm for this class of treatment [
35].
Characteristics and outcome of diabetic HFpEF patients according to glycemic control.
While the presence of diabetes conferred a worse prognosis to our HFpEF patients, tight glycemic control did not seem to reverse this association. On the contrary, patients with best controlled diabetes (HbA1C < 7%) were more at risk for adverse events (hospitalization for heart failure and all-cause mortality). Previous studies [
11,
12,
36,
37] showed a U-shaped association between HbA1C and prognosis in heart failure patients, with the lowest risk in the group of patients with HbA1C between 6.5 and 7.5%. However, those studies were either conducted among patients with HFrEF alone, or did not make a distinction between patients according to ejection fraction, while the interplay between diabetes and outcome seems to differ in those populations. In the CHARM trial, the relative risk conferred by diabetes was significantly greater in patients with preserved ejection fraction (EF) than in those with low EF [
8] and a recent study highlighted that the presence of T2D was associated with a reduction of exercise capacity (lower peak VO2) in the LVEF < 40% and LVEF 40–49%, but not in the LVEF > 50% subgroup [
38].
Data about glycemic control and outcome in HFpEF are scarce. A study by Gu et al. [
13] did not find baseline HbA1C to be an independent predictor of outcome, but they analyzed it in the overall population of HF with T2D, and not only in HFpEF. Glycemic variability, however, was associated with outcome in the HFpEF subgroup [
13] and was associated with signs of diastolic dysfunction in patients without HF [
14]. Finally, the GAMIC cohort, a large population-based propensity-matched study of patients with HF [
29] observed an increased mortality and morbidity (hospitalizations and visits) in patients who developed diabetes, particularly in those with a mean HbA1c higher than 7.0%.
How can we explain that, in our population, patients with higher HbA1C levels seem “protected” and suffer from less adverse events, while recent research emphasized the direct role of glucotoxicity on cardiomyocytes in the development of diabetic cardiomyopathy [
31,
39,
40]? Firstly, glucotoxicity plays a part in the pathophysiology of the disease but its role in the evolution of symptoms and outcomes is yet unknown. Heart failure in diabetic patients occurs in a broad context of metabolic disorders including lipotoxicity, glucotoxicity and insulin resistance and resulting in impaired mitochondrial oxidative capacity and increased reactive oxygen species (ROS) production and surely, hyperglycemia is not the only mechanism involved. This is supported by the fact that, before the SGLT-2 inhibitor era, no study could demonstrate a favorable effect of glucose lowering therapies on events related to heart failure [
41]. Conversely, some glucose-lowering therapies, including peroxisome proliferator-activated receptor (PPAR) agonists even increased the risk of heart failure in individuals with type 2 diabetes. Note that those drugs were seldom taken by patients in our cohort (Fig.
3) and cannot solely be responsible for the difference in event-free survival.
For years, it has been assumed that insulin resistance observed in diseases characterized by nutrient excess (ie T2D and obesity), was fundamental to the pathogenesis of these diseases. As stated above, insulin resistance into the heart has been considered to favor myocardial contractile dysfunction and to be involved in the pathophysiology of diabetic cardiomyopathy. However, an alternative view, which recently gained researchers’ interest, is that adaptations occurring in metabolic diseases can be viewed as protective in nature, and that insulin resistance could act as a defense mechanism to prevent or delay pathological intracellular substrate accumulation when substrate uptake exceeds energy demand [
42‐
45]. Fundamental to this hypothesis is that, although these metabolic alterations are deleterious in the long term for complications associated with obesity and diabetes, they provide immediate protection against cell death in response to excess nutrients. Supporting this, it has been shown that cardiac contractile function was preserved, or even improved, in hearts subjected to metabolic and haemodynamic stress when myocardial insulin resistance was induced in response to elevated glucose levels or upon high-fat diet [
46,
47]. Conversely, excessive insulin signaling exacerbates systolic dysfunction when the heart is subjected to pressure overload [
48]. In light of this, the discrepancy between our study and the results of the GAMIC cohort [
29] might be explained by the difference in disease duration. The GAMIC cohort excluded patients with a previous diagnosis of diabetes, while the mean duration of diabetes in our population was 19 ± 8 years. Possibly our results do not apply to new onset diabetes, as the adaptation to excess nutrients have not yet taken place.
In this context of old patients with long standing diabetes, the utility of therapeutically targeting glycemia, particularly through insulin sensitization, is questionable as it may result in exposure of cells and tissues to additional nutrients that will further challenge their survival. This could explain why PPAR agonists, important insulin sensitizers favouring nutrient uptake and storage, have been associated with adverse cardiovascular outcomes in T2D patients. On the other hand, treatment reducing nutrient overload might be beneficial and should be preferred. Metformin, for example, which has shown beneficial effect on mortality in HF patients [
49], although often referred to as insulin sensitizer, has its main glucose-lowering effect via reducing hepatic glucose production. Similarly, SGLT-2 inhibitors lower blood glucose by promoting glycosuria.
This cardioprotective effect of insulin resistance could be involved in the better prognosis observed in heart failure patients with higher BMIs, referred to as the “obesity paradox” [
15,
50]. Although we did not measure insulin resistance per se, we can hypothesize that the group with HbA1C > 7% is more insulin resistant as they are more obese and show higher glycemia levels though intensively treated.
Furthermore, hyperglycemia was shown to be involved in irreversible epigenetic changes, known as “glycemic memory”, and HbA1c at time of the study cannot reflect the whole history of diabetes [
51,
52]
. Similarly, intermittent hyperglycemia, rather than chronic elevation of blood glucose, with a lesser repercussion on HbA1C levels, exacerbates the production of reactive oxygen species, impairs endothelial function and induces cytokines release and contributes to pejorative evolution [
53]. Finally, hypoglycemia could also be involved in the progression of cardiovascular diseases and mortality through sympatho-adrenal response [
54].
In short, together with existing literature, this study underlines that other mechanisms besides glucotoxicity must be involved in the development and worsening of heart failure in diabetic patients, and that the effect of intensive glycemic control on cardiovascular associated morbidity is not fully understood.
Current guidelines recommend that the appropriate target for HbA1C should be individualized based on overall health and life expectancy. As such it is generally accepted that the glycemic goal should be somewhat higher (HbA1C ≤ 8%) in frail older adults with medical and functional comorbidities [
55,
56]. Patients with HFpEF generally match this description (mean age of 78 years and high comorbidity burden in our population). However, these recommendations are based on consensus and there are virtually no trials that have examined glycemic control and complications focusing on the older patients, and even less on older patients with HFpEF. Hence, an important issue that is still unsolved is the optimal target level of HbA1c in that population. Given published data, glycemic variability should be avoided once the optimal target is reached [
13]. Our study is a retrospective analysis of a relatively small population and does not allow answering this question. Furthermore, very few patients in our population had severely uncontrolled diabetes. However, this study generates the hypothesis that low levels of HbA1C are associated with more adverse events, and that physicians should not be too stringent about glycemic control in HFpEF patients with long standing diabetes. In addition, it underlines the need for future studies: fundamental studies to unravel the interaction between diabetes, insulin resistance and heart failure, and clinical studies designed to determine the optimal HbA1C target in HFpEF.