Background
Aortic stenosis (AS) is initially a disease of the heart valve but its prognosis depends greatly on the health of the myocardium. Sustained pressure overload by AS induces ventricular hypertrophy and myocardial fibrosis that lead to ventricular decompensation [
1,
2]. Myocardial fibrosis is commonly observed in two forms, diffuse interstitial and focal replacement fibrosis. Both forms of fibrosis can be imaged noninvasively with cardiac magnetic resonance (CMR) with gadolinium-based contrast agents: diffuse interstitial fibrosis is quantified by extracellular volume fraction (ECV) on T1 mapping and replacement fibrosis by late gadolinium enhancement (LGE). The former is partially reversible while the latter remains even after relief of pressure overload by aortic valve replacement (AVR) in AS [
3]. Both increased ECV and LGE are associated with worse prognosis in patients with AS [
4‐
6].
‘Diabetic cardiomyopathy’ was first described from autopsies of diabetic patients who manifested with heart failure but had no evidence of coronary problems, valvular disease or hypertension [
7]. Subsequent investigations demonstrated that diabetic patients have increased myocardial fibrosis explained by multiple biological and molecular mechanisms [
8]. Most studies on the interaction of diabetes with AS have focused on the progression of valvular stenosis [
9‐
11]; however, there have been only few studies that addresses the impact of diabetes on myocardial health in AS patients [
12‐
14].
Considering that diabetes is associated with worse prognosis in AS patients [
15,
16], we hypothesized that diabetes would aggravate the degree of myocardial fibrosis in AS patients. The objective of this study was two-fold; first, to elucidate the prognostic impact of diabetes in AS patients and second, to dissect its underlying mechanisms using comprehensive noninvasive imaging and plasma proteomics.
Methods
Study population
This study utilized two prospective cohorts of AS patients: the imaging cohort for the assessment of the myocardial health using CMR and echocardiography with long term follow-up for clinical events, and a biomarker cohort for the assessment of enriched circulating proteins using multiplex proximity extension assay.
The imaging cohort consisted of 253 patients with moderate or severe AS prospectively enrolled from 2011 to 2015 at three tertiary medical centers in Korea. All participants in this cohort underwent comprehensive echocardiography and CMR. The biomarker cohort consisted of 100 patients with severe AS undergoing surgical AVR enrolled prospectively from 2018 to 2021 at Seoul National University Hospital (Additional file
1: Fig. S1). Detailed inclusion and exclusion criteria are in Additional file
1: Additional Methods. Two separate cohorts were used in the study because patients in the imaging cohort did not undergo blood sample collection and most of the patients in the biomarker cohort did not undergo CMR and were followed for less than one year (median follow-up 6.6 [IQR 3.9–12.2] months). The study complied with the
Declaration of Helsinki, and the cohorts were approved by the institutional review boards of each institution, with all study subjects providing written informed consent before enrollment.
Diabetic status was defined as either: (i) treatment for diabetes at enrollment with lifestyle modification, oral hypoglycemic agents, or insulin, or (ii) HbA1c ≥ 6.5% in those not previously diagnosed with diabetes and with HbA1c levels available at the period of enrollment. Medication status at enrollment was assessed for the diabetes group. Ischemic heart disease (IHD) was defined as either a history of myocardial infarction or coronary intervention, or significant coronary artery stenosis (> 50%) on angiography prior to AVR, or coronary intervention or concomitant coronary artery bypass grafting performed together with AVR.
Echocardiographic evaluation
Echocardiography was performed using commercially available machines and the severity of AS was determined according to the contemporary guidelines [
17]. The left ventricular (LV) chamber size, and systolic and diastolic function were also evaluated and categorized according to the latest guidelines [
18,
19]. AS severity was defined based on aortic valve area (severe AS, aortic valve area ≤ 1 cm
2; moderate AS, aortic valve area > 1 cm
2 and ≤ 1.5 cm
2) [
20].
CMR analysis
In the imaging cohort, CMR was performed at a median interval of 11 days (interquartile range 2–29 days) from the date of echocardiography. The details of the scanners, field strengths, T1 mapping sequences, contrast agents, and summary of imaging analyses for each study center are presented in Additional file
1: Table S1 [
6]. Briefly, CMR scans consisted of balanced steady-state free precession cine images, pre- and post-gadolinium T1 mapping, and LGE images. The chamber sizes and myocardial mass were quantified according to a standardized protocol [
5]. The degree of diffuse interstitial fibrosis was assessed by calculating ECV from the pre- and post-gadolinium T1 values measured at the short-axis mid-ventricular septum and blood pool, and hematocrit from blood samples at the time of CMR [
2,
21]. Infarct-related LGE was excluded while non-infarct LGE was included in the ECV assessment [
5]. The presence and extent of LGE was assessed on short-axis images acquired by phase sensitive inversion recovery sequence; the regions of LGE were drawn semi-automatically as pixels of the myocardium with a signal intensity > 5 standard deviations of the normal remote myocardium and the LGE% was calculated by dividing the LGE area by the total LV myocardial area.
Plasma proteomics assay
Blood samples from the biomarker cohort were collected preoperatively in EDTA bottles, divided into plasma and buffy coat layers with centrifugation, and then stored at – 80 °C. For plasma proteomics analysis, deep frozen plasma samples were shipped to Olink Proteomics (Uppsala, Sweden) and the plasma levels of 92 protein biomarkers were measured using commercially available multiplex proximity extension assay kits (Cardiovascular III Panel, Additional file
1: Table S2). This high-throughput technique utilizes immunoassay with oligonucleotide-labeled antibodies followed by real-time polymerase chain reaction for simultaneous quantification of target proteins with high specificity and scalability [
22]. After normalization, plasma levels were expressed for each protein in relative quantification units called normalized protein expression (NPX) using the Log
2 scale (1 NPX difference equaling twofold change in protein concentration).
Clinical outcomes
The clinical outcomes of interest in this study were unexpected hospitalization for heart failure that necessitated intravenous diuretics and all-cause mortality. These outcomes were assessed in the imaging cohort by review of medical records, reports from family members, and official mortality data from Statistics Korea. Patients were followed from the date of CMR to the last clinical follow-up or death.
Statistical analysis
Continuous data are presented as mean ± standard deviation or median (interquartile range) depending on the normality of distribution, and categorical data as number (%). Characteristics were compared between the groups using the t-test (or Mann–Whitney test for non-normally distributed continuous variables) or the chi-square test. Comparisons of groups according to diabetes medication was conducted using the Kruskal–Wallis test. Variables associated with increased diffuse interstitial or replacement fibrosis were analyzed using logistic regression and the degree of association expressed in odds ratio (OR) with 95% confidence interval (CI). Multivariable models were constructed with the stepwise backward selection method using the Akaike information criterion or inclusion of clinically important variables such as age, sex, diabetes, hypertension, atrial fibrillation, IHD, and peak aortic velocity.
Comparison of plasma biomarker levels according to the diabetic status was performed using the Welch’s two-sample t-test, adjusting for multiple testing with the Benjamini–Hochberg method. The adjusted p-values represent the false discovery rate and p-values < 0.05 were considered significant. Logistic regression was used to assess the association of plasma biomarkers with diabetic status, adjusting for age, sex, hypertension, atrial fibrillation, IHD, and peak aortic velocity. Functional enrichment analyses were performed using g:Profiler with Gene Ontology terms.
Kaplan–Meier survival curves with log-rank tests were used to compare event-free survival according to the presence of diabetes. Cox proportional-hazards regression analyses were used to assess predictors of the endpoints and the effect size expressed as hazard ratio (HR) with 95% CI. The final multivariable model was constructed with stepwise backward selection from clinically important variables such as age, sex, diabetes, hypertension, atrial fibrillation, stroke, IHD, peak aortic velocity, LV ejection fraction by echocardiography, and AVR. Two-sided p-values < 0.05 were considered statistically significant. Analyses were conducted using R version 4.0 (Vienna, Austria) or SPSS version 25 (Chicago, USA).
Discussion
Herein, we demonstrated that AS patients with diabetes compared to non-diabetic patients had increased diffuse interstitial and replacement fibrosis by CMR analysis of the myocardium. An in-depth investigation of plasma proteomics demonstrated that factors related to proinflammatory response and extracellular matrix components were enriched in diabetic AS patients. These diabetic AS patients had a significantly higher incidence of heart failure and death than the non-diabetic patients. These results suggest that diabetes is associated with effects that potentiate systemic proinflammatory and profibrotic milieu in the pressure-overloaded myocardium, which translates into worse clinical outcomes even after AVR.
Diabetes mellitus is a systemic disease that affects the myocardium directly [
8]. Patients with otherwise uncomplicated diabetes show reduced cardiopulmonary performance [
24,
25], which is associated with increased myocardial fibrosis [
26]. Microvascular function is also impaired in diabetes [
27], and therefore, the myocardium may be more vulnerable to ischemic insult [
28]. Studies utilizing CMR in the general population have suggested that diffuse interstitial and replacement fibrosis are increased in diabetic compared to non-diabetic subjects, and both forms of fibrosis are associated with heart failure and mortality events [
29‐
31].
There have been few studies on how diabetes impacts myocardial remodeling in AS. We found that the degree of myocardial fibrosis and diastolic dysfunction is significantly more advanced in diabetic AS patients. An invasive histological study using myocardial biopsies in 60 AS patients undergoing AVR suggested that patients with concomitant AS and diabetes had more myocardial fibrosis and higher cardiomyocyte stiffness [
13]. Using noninvasive imaging to examine the entire myocardium, we show that the histologic evidence from the previous study [
13] holds true in a larger AS population.
In AS patients, diabetes is independently associated with an increase in both mid-term and long-term mortality in those undergoing AVR, as well as in those with asymptomatic AS patients managed conservatively [
15,
16,
32]. Herein, we suggest the missing link between diabetes and outcome in AS by demonstrating the association between diabetes and myocardial fibrosis. Moreover, the need for more intensive diabetes treatment as a marker of diabetes severity was associated with a greater degree of myocardial fibrosis and LV diastolic dysfunction, especially in insulin-treated diabetic patients. Previous studies have also shown that the prognosis of diabetic AS patients even after AVR are significantly worse especially in those treated with insulin [
16,
32]. This suggests that factors other than the stenotic valve in AS is responsible for the worse prognosis in diabetic AS patients and that relief of the stenotic valve may not be enough for optimal outcome.
In the pressure-overloaded heart, myocardial fibrosis is an important driver of the progression from compensated hypertrophy to heart failure with diastolic and systolic dysfunction, findings demonstrated in both histological and imaging studies [
1,
33,
34]. In the current study, AS patients with diabetes had greater replacement and diffuse interstitial fibrosis compared to non-diabetic counterparts. Although the difference in both measures of fibrosis between the diabetics versus non-diabetics may seem small, this difference is clinically significant as shown by previous investigations demonstrating 11% and 10% higher mortality with each 1% increase of %LGE and %ECV, respectively [
4,
5]. Furthermore, there is a characteristic non-linear and threshold effect of %LGE and %ECV on outcome in AS [
6], suggesting that even a small increase of fibrosis can lead to significantly worse outcomes. Diabetes is also associated with poor LV mass regression after AVR [
14], suggesting that it continues to affect myocardial remodeling after relief of pressure overload.
This leads us to question whether and how diabetes changes the systemic milieu and ultimately, the myocardium. Circulating protein biomarkers provide important clues to pathophysiological mechanisms of diseases, and high-throughput proteomic methods can measure a multitude of proteins simultaneously. In previous plasma proteome studies of AS patients, higher GDF-15 was associated with poor LV reverse-remodeling and increased mortality after AVR [
35,
36]. In plasma proteome analysis of patients with heart failure, diabetic patients had higher circulating GDF-15 and galectin-4 levels [
23], and over-representation of pathways related to inflammation, cardiac remodeling, and fibrosis [
23,
37]. We found that E-selectin, interleukin-1 receptor type 1, interleukin-1 receptor type 2, galectin-3, galectin-4, intercellular adhesion molecule 2, integrin beta-2, GDF-15, and cathepsin D levels were significantly upregulated in diabetic AS patients, proteins which have been implicated in inflammation, cardiac fibrosis and remodeling, atherosclerosis, and heart failure [
38‐
44]. Furthermore, over-representation analyses of the plasma proteome demonstrated that pathways related to neutrophil activation, interleukin-1 and amplification of inflammation, leukocyte migration, and extracellular matrix were enriched in diabetic AS patients. Our study supports that upregulation of signals related to inflammation and extracellular matrix expansion are important pathophysiological processes systemically mediated by diabetes in AS patients, which in turn, may aggravate the degree of myocardial fibrosis and lead to worse outcomes.
Our study suggests that clinicians should be aware of the significantly higher clinical events in AS patients with diabetes. According to our analysis, diabetes not only damages the stenotic valve [
9‐
11], but also, the health of the myocardium, with its effect pervading even after AVR. In particular, diabetes patients treated with insulin may be at the highest risk. Our findings suggest that the changes in the myocardium by AS are already more advanced in diabetes patients and it can be assumed that the systemic proinflammatory−profibrotic milieu will continue with diabetes even after AVR. This also suggests that AVR may not be the final treatment for AS when the patient has diabetes. Considering the systemic proinflammatory−profibrotic environment by diabetes, our findings call for further studies on whether anti-inflammatory approaches may be beneficial in AS patients with concomitant diabetes, to prevent adverse myocardial remodeling and to improve clinical outcomes. Exciting options testing this idea have been developed with promising outcomes, such as the monoclonal antibody targeting interleukin-1β [
45]. The anti-diabetic medications alleviate inflammation by differing degrees [
46], which may also affect the diabetes-related myocardial remodeling and fibrosis, especially in the pressure-overloaded myocardium of AS patients. Additionally, whether individualizing the treatment strategy would be more beneficial in severe AS patients with concomitant diabetes should be evaluated in future trials.
This study has some limitations that needs to be considered. First, because of the shortage of resources, two separate imaging and biomarker cohorts were used for analysis, and analysis of the direct association of plasma proteins with noninvasive measures of fibrosis on CMR could not be performed. Second, information on the diabetes control status assessed by HbA1c or FBS values was not available for all patients and information on the duration of diabetes was unavailable. Thus, we used the diabetes medication status as a surrogate marker of the severity of diabetes, but its limitations must be acknowledged. Also, it should be considered that the number of patients on insulin was small to make any definite statements regarding the relationship between myocardial fibrosis and the type of medications for diabetes. Lastly, we used a select biomarker panel of 92 proteins, and future studies utilizing a more comprehensive set of proteins may provide more pathophysiological information as well as therapeutic targets.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.