Author + information
- Received July 3, 2017
- Accepted July 31, 2017
- Published online September 25, 2017.
- Laura Ernande, MD, PhDa,b,
- Etienne Audureau, MD, PhDc,
- Christine L. Jellis, MD, PhDd,
- Cyrille Bergerot, MDe,
- Corneliu Henegar, MD, PhDb,
- Daigo Sawaki, MD, PhDb,
- Gabor Czibik, MD, PhDb,
- Chiara Volpi, MDa,
- Florence Canoui-Poitrine, MD, PhDc,
- Hélène Thibault, MD, PhDf,g,
- Julien Ternacle, MDa,b,
- Philippe Moulin, MD, PhDg,h,
- Thomas H. Marwick, MBBS, PhD, MPHi and
- Geneviève Derumeaux, MD, PhDa,b,∗ ()
- aPhysiology Department, DHU Ageing-Thorax-Vessel-Blood, Henri Mondor Hospital, Assistance Publique Hôpitaux de Paris, Créteil, France
- bINSERM U955, Team08, Université Paris-Est Créteil (UPEC), Créteil, France
- cBiostatistics Department, Henri Mondor Hospital, Assistance Publique Hôpitaux de Paris, Créteil, France; CEpiA EA7376, DHU Ageing-Thorax-Vessel-Blood, Université Paris Est (UPEC), Créteil, France
- dCardiology Department, Cleveland Clinic, Cleveland, Ohio
- eCentre d’Investigation Clinique INSERM 1407 Louis Pradel Hospital, Hospices Civils de Lyon, Bron, France
- fService d'Explorations Fonctionnelles Cardiovasculaires, Groupement Hospitalier Est, Hospices Civils de Lyon, Bron, France
- gINSERM UMR-1060, CarMeN Laboratory, Université Claude Bernard Lyon, Lyon, France
- hFédération d’endocrinologie, Hospices Civils de Lyon, Bron, France
- iBaker Heart and Diabetes Institute, Melbourne, Australia
- ↵∗Address for correspondence:
Dr. Geneviève Derumeaux, Institut Mondor de Recherche Biomedicale, Unité Inserm 955-équipe 8, Faculté de médecine de Créteil, 8, rue du Général Sarrail, 94000 Créteil, France.
Background Type 2 diabetes mellitus (T2DM) may alter cardiac structure and function, but obesity, hypertension (HTN), or aging can induce similar abnormalities.
Objectives This study sought to link cardiac phenotypes in T2DM patients with clinical profiles and outcomes using cluster analysis.
Methods Baseline echocardiography and a composite endpoint (cardiovascular mortality and hospitalization) were evaluated in 842 T2DM patients from 2 prospective cohorts. A cluster analysis was performed on echocardiographic variables, and the association between clusters and clinical profiles and outcomes was assessed.
Results Three clusters were identified. Cluster 1 patients had the lowest left ventricular (LV) mass index and ratio between early mitral inflow velocity and mitral annular early diastolic velocity (E/e′) ratio, had the highest left ventricular ejection fraction (LVEF), and were predominantly male with the lowest rate of obesity or HTN. Cluster 2 patients had the highest strain and highest E/e′ ratio, were the oldest, were predominantly female, and had the lowest rate of isolated T2DM (without HTN or obesity). Cluster 3 patients had the highest LV mass index and volumes and the lowest LVEF and strain, were predominantly male, and shared similar age and rate of obesity and HTN as cluster 1 patients. After follow-up of 67 months (interquartile range: 40 to 87), the composite endpoint occurred in 56 of 521 patients (10.8%). Clusters 2 (hazard ratio: 2.37; 95% confidence interval: 1.15 to 4.88) and 3 (hazard ratio: 2.19; 95% confidence interval: 1.00 to 4.82) had a similar outcome, which was worse than cluster 1.
Conclusions Cluster analysis of echocardiographic variables identified 3 different echocardiographic phenotypes of T2DM patients that were associated with distinct clinical profiles and highlighted the prognostic value of LV remodeling and subclinical dysfunction.
- diabetic cardiomyopathy
- diabetic heart disease
- myocardial strain
- subclinical myocardial disease
Insulin resistance and glycemic dysregulation are associated with subtle and progressive modifications in cardiac structure and function. Such a “metabolic exposome” results in left ventricular (LV) remodeling (1–3) and impaired LV systolic and diastolic function (4,5). More specifically, type 2 diabetes mellitus (T2DM), an independent risk factor for heart failure (HF) (6), can induce a diabetic cardiomyopathy (7–9). In T2DM, early cardiac modifications include LV hypertrophy (10), diastolic dysfunction (11), and decreased myocardial strain (12–14), which are precursors for the development of LV remodeling (15) and adverse outcomes, including HF (16) and all-cause mortality and hospitalization (17).
Obesity and hypertension (HTN) are frequent comorbidities of T2DM (18–20), and many of the abnormalities found in diabetic cardiomyopathy are analogous to those described in both obese (21) and hypertensive patients (22,23). Moreover, parameters such as age (1,24) and sex might also influence early cardiac geometry and functional changes. The specific contribution of these causative factors is unclear, as is their synergistic contribution to cardiac dysfunction in T2DM. Whereas classic statistical analyses are built on a priori hypothesis, cluster analysis might improve cardiac phenotyping and provide new insights in heterogeneous patient groups, such as those with T2DM, by providing an innovative exploratory analysis (25). Therefore, we hypothesized that cluster analysis might allow identification of T2DM patient groups (clusters) with common cardiac phenotypes based on ultrasound data and that those cardiac phenotypes are associated with distinct clinical profiles and different prognosis.
This analysis used data from 2 large prospective cohorts (453 from Lyon, France, and 389 from Brisbane, Australia) of asymptomatic T2DM patients free of overt heart disease. Although similar inclusion criteria were shared between the 2 cohorts in terms of age, T2DM, and absence of overt cardiac disease, the different characteristics of the 2 populations, especially the prevalence of obesity and HTN, provided a heterogeneity suitable for cluster analysis (Online Table 1).
Inclusion criteria were T2DM with normal left ventricular ejection fraction (LVEF; >50%) and taking oral hypoglycemic or insulin treatment (12,15,26). Exclusion criteria were type 1 diabetes mellitus; symptoms, signs (clinical or electrocardiographic), or history of heart disease; presence of regional LV wall motion abnormalities; absence of sinus rhythm; history of cardiomyopathy, coronary artery disease, or valvular heart disease; severe renal failure (defined as creatinine clearance <30 ml/min); echocardiographic images unsuitable for quantification; severely uncontrolled diabetes (glycosylated hemoglobin [HbA1c] >12% or glycemia >3 g/l); and uncontrolled resting blood pressure (>180/100 mm Hg). Silent ischemia and/or coronary artery disease was excluded by a negative stress electrocardiogram, echocardiogram, or myocardial perfusion scintigraphy, and/or by normal coronary angiogram.
Ethics committee approval was obtained from both institutions, and all subjects provided informed consent to participate.
Clinical and biological data
All patients underwent physical examination, echocardiography, and biochemical analysis on the same day at inclusion. Clinical data were collected regarding the age, sex, diabetes duration, height, weight, smoking status, and medication use of subjects. A complete physical examination was performed. Hemodynamic parameters, including heart rate and blood pressure, were measured. Obesity was defined as a body mass index (BMI) ≥30 kg/m2. Diagnosis of HTN was based on medical history and the use of antihypertensive medications and/or office blood pressure measurement confirmed by ambulatory blood pressure monitoring (27).
The presence or absence of diabetic retinopathy was assessed by dilated eye examination performed by an ophthalmologist (7). The diagnosis of neuropathy was determined based on interview and physical examination. Blood samples were taken for biochemical analysis of creatinine, cholesterol, triglyceride, HbA1c, and brain natriuretic peptide levels. Estimated glomerular filtration rate was determined using the Modification of Diet in Renal Disease Study equation (28).
Transthoracic echocardiography was performed at rest using a commercially available ultrasound system. Acquisitions from at least 3 consecutive heart beats were stored in raw data format for off-line analysis. Left ventricular end-diastolic volume (LVEDV), left ventricular end-systolic volume (LVESV), and LVEF were calculated using the Simpson biplane method (29). LVEDV and LVESV were indexed to body surface area, and LV mass was determined using the recommended formula (29) and indexed to body surface area.
Using pulsed wave Doppler, mitral inflow velocities and peak early (E) and late (A) diastolic velocities were measured, and E/A ratio was calculated. The mitral annular early diastolic velocity (e′) was assessed at the septal (e′ septal) and lateral (e′ lateral) sites of the mitral annulus using pulsed wave Doppler tissue imaging, and the mean e′ value and E/e′ ratio were calculated (30,31). The left atrial area (LA) was measured in an apical 4-chamber view by planimetry, and the LA volume was measured by the Simpson method and indexed to body surface area (29).
The peak systolic longitudinal strain of the LV basal segments was measured using 2-dimensional speckle tracking imaging on the apical views. The endocardial border was manually traced from an end-systolic frame. The software automatically detected the epicardial border, and the region of interest was manually adjusted to include the entire myocardial wall. The quality of tracking was verified both automatically and visually, and the region of interest was modified and corrected by the observer if judged necessary to obtain optimal tracking. The peak systolic longitudinal strain of the LV basal segments was collected to calculate the mean basal strain value (reported using absolute value).
Major adverse cardiac events were obtained from annual follow-up visits and medical records. The endpoint for the study was a composite of cardiovascular mortality and hospitalization (arrhythmia, HF, coronary artery disease).
The initial analysis sought to assess whether differences in clinical, biological, and echocardiographic features existed between patients with T2DM according to the presence or absence of obesity and/or HTN. Four clinically defined groups were thus constituted: 1) patients with isolated T2DM; 2) T2DM and obesity; 3) T2DM and HTN; and 4) T2DM, obesity, and HTN.
In the second part of the analysis, we sequentially conducted various unsupervised analyses to more deeply investigate the associations between echocardiographic features and the other phenotypic characteristics or major cardiac adverse events. We then sought to identify novel specific patient profiles to improve the characterization of T2DM without overt heart disease. To do so, a principal component analysis was first performed to inform correlations between parameters. A Gabriel biplot was built to project the patients along the principal components axes, based on their own individual echocardiographic characteristics (32). The biplot was used to map cluster solutions by attributing different colors to patient groups according to the results of the cluster analysis conducted afterward.
Agglomerative hierarchical clustering analysis was used to identify echocardiographic patterns that would define subsets of patients. This analysis combines similar subjects based on selected features (here, restricted to echocardiographic parameters, i.e., LV mass indexed to body surface area [LVMi], LVEDV indexed to body surface area, LVESV indexed to body surface area, LVEF, strain, E/A ratio, e′ septal, e′ lateral, and E/e′ mean ratio) to create homogeneous clusters of T2DM patients characterized by an overall similar functional profile. Hierarchical clustering was performed on standardized raw (e′ septal and e′ lateral) or log-transformed (all other variables) values, depending on the normality of distribution, based on Euclidean distance as the similarity metric and using the Ward linkage approach (33). All clustering was performed by 1 investigator (E.A.) blinded to clinical data.
A heat map was generated to illustrate the findings and a dendrogram to demonstrate combinations between patients and distances between groupings, with horizontal branches representing the combination of 2 clusters and vertical branches the degree of dissimilarity between combined clusters. The optimal number of clusters was determined based on statistical indexes, with an a priori minimal cluster size of 100 patients designated to improve stability of the results. Because no unique statistical approach provides definitive guidance on the number of clusters to retain, we used a combination of 30 different tests, as provided by the R function NbClust (34), to determine the optimal number of clusters, finally choosing the most-picked number of clusters across the tests. Radar plots were charted to visualize echocardiographic characteristics across clusters. Finally, clinical and biological characteristics were compared across clusters to characterize their clinical significance.
Cluster analysis was performed on patients with complete information for echocardiographic variables. The proportions of missing values for all variables are shown in Online Figure 1. Although most variables had <5% missing data, 2 echocardiographic variables had higher missing data rates: LA area (37.4%) and LA volume index (57.0%). These 2 variables were deliberately not used for building clusters to preserve an appropriate sample size for clustering analysis. A sensitivity analysis was further conducted to assess the robustness of our findings when also using these variables. To do so, we included all patients with available data for LA area and other variables used for the main analysis (n = 462), while imputing missing LA volumes using k-nearest neighbors methodology.
Descriptive results are presented as percentage for categorical data and mean ± SD or median (interquartile range) for continuous variables, depending on the normality of their distribution as assessed by Shapiro-Wilk test.
Comparisons between groups for continuous data were conducted using 1-way analysis of variance or Kruskal-Wallis tests, followed by post hoc tests for pairwise comparisons in case of overall significance and applying Sidak correction for test multiplicity. Categorical data were compared using the chi-square test or Fisher exact test, as appropriate. Echocardiographic parameters were further compared across groups using linear regression models adjusting for age, sex, and diabetes duration.
Survival analyses based on time-to-event data were performed to assess the prognostic value of the identified clusters or the clinical groups according to presence of obesity and/or HTN. Unadjusted cumulative incidence curves were plotted by the Kaplan-Meier method for the composite study endpoint (cardiovascular hospitalization or death), using log-rank tests to assess significance for group comparison. Cox proportional hazards regression models were also constructed to compute crude hazard ratios (HRs) and 95% confidence intervals (CIs).
A p value < 0.05 was considered significant. Descriptive analyses and comparisons between clinically defined and clusters groups were performed using Stata 14.1 (StataCorp LLC, College Station, Texas). R statistical software (3.1.3) was used for clustering analyses and visualizations (stats, NbClust, mclust, pca3d, d3heatmap, and ggplot2 packages). Radar charts were specifically plotted based on the D3.js library.
The main analyses were performed on 745 of 842 patients (88%) who had complete echocardiographic information. Most variables were similar between excluded and included patients, except for diabetes duration and BMI (Online Table 2).
Table 1 summarizes the characteristics of subgroups of T2DM alone (n = 201 [27.0%]) or associated with both obesity and HTN (n = 213 [28.6%]), obesity (n = 120 [16.1%]), or HTN (n = 211 [28.3%]). The group with T2DM and HTN were older, had longer diabetes duration, and had a higher prevalence of retinopathy or neuropathy compared with the other groups.
Table 2 compares the echocardiographic data among the 4 groups. The presence of obesity and HTN did not significantly alter LV morphology (LVMi and indexed LV volumes) and systolic function (LVEF and longitudinal strain) (Figure 1). In contrast, groups with obesity and HTN had an impact on diastolic function (lower e′ velocities and higher E/e′ ratio, but in normal ranges) compared with the other groups. Despite significant differences between mean values, an important overlap of individual values of diastolic parameters was observed among the 4 groups (Figure 1).
Cluster analysis of echocardiographic phenotypes
Cluster 1 (low comorbidity) had the lowest LVMi and E/e′ ratio values, the highest LVEF, and the second highest strain values. This cluster comprised predominantly male patients, with the lowest rate of obesity or HTN.
Cluster 2 (elderly, diastolic dysfunction) had the highest strain values but the lowest e′ velocities and the highest E/e′ ratio. This cluster gathered the oldest patients and was predominantly female with the lowest rate of isolated T2DM. Blood pressure levels, BMI, and heart rate were the highest compared with the 2 other clusters.
Cluster 3 (hypertrophic systolic dysfunction) had both the highest LVMi and LV volumes and the lowest LVEF and strain. This cluster comprised predominantly males, with similar age and rate of obesity and HTN as cluster 1 (Figure 3).
Additionally, cluster analysis blunted the overlap in patients (Figure 1). Of note, diabetes duration and HbA1c were similar among the 3 clusters.
Follow-up data were available for 521 of the 745 (70%) patients included in the cluster analysis (follow-up duration: 67 months [interquartile range: 40 to 87 months]). The composite endpoint occurred in 56 patients (10.8%), including 3 cardiac deaths (0.6%) and 53 cardiovascular hospitalizations (10.2%). Raw event rates ranged from 4.7% (cluster 1) to 17.9% (cluster 2), and from 5.5% (isolated T2DM) to 17.5% (T2DM + obesity + HTN).
Survival analysis found significant differences across clusters for the composite endpoint (log-rank p = 0.049) (Figure 4A). Using Cox proportional hazards modeling, patients with the best prognosis were from cluster 1, whereas patients from clusters 2 and 3 shared HRs of similar magnitude (cluster 1 HR: 1.00 [reference]; cluster 2 HR: 2.37; 95% CI: 1.15 to 4.88; cluster 3 HR: 2.19; 95% CI: 1.00 to 4.82). Similar results were obtained after adjusting for age, sex, and diabetes duration (data not shown).
The presence of obesity and HTN also had an impact on composite endpoints (log-rank p = 0.02) (Figure 4B). Increasing HRs were observed with accumulating risk factors, with maximal HRs found in T2DM + obesity + HTN patients: isolated T2DM HR: 1.00 (reference); T2DM + obesity HR: 1.56; 95% CI: 0.58 to 4.19; T2DM + HTN HR: 2.25; 95% CI: 1.01 to 5.00; and T2DM + obesity + HTN HR: 3.13; 95% CI: 1.44 to 6.81. Again, similar results were found after adjusting for age, sex, and diabetes duration (data not shown).
Results from the sensitivity analysis conducted using LA area and volume (n = 462) are shown in Online Tables 3 and 4 for cardiac and clinical features and in Online Figures 2 and 3 for principal component analysis and biplot and survival analysis, respectively. Findings were remarkably close to those obtained from the main analysis, identifying 3 clusters with similar key characteristics. Kaplan-Meier survival curves revealed similar trends, although these did not reach statistical significance (log-rank p = 0.14; cluster 1 HR: 1.00 [reference]; cluster 2 HR: 2.25; 95% CI: 0.99 to 5.17; cluster 3 HR: 1.82; 95% CI: 0.72 to 4.56).
In a large prospectively enrolled group of asymptomatic T2DM patients free of overt heart disease, we applied an innovative approach using cluster analysis to identify clusters of patients with homogeneous cardiac phenotypes and sought to determine whether these clusters were associated with distinct clinical profiles and outcomes (Central Illustration). Whereas statistical analysis, based on a priori risk factor groups, resulted in a substantial overlap of the echocardiographic variables, cluster analysis was able to distinguish 3 groups on the basis of echocardiography-related cardiac phenotype. Clinical characteristics varied between these clusters, with phenotypes associated not only with obesity and HTN but also with age, sex, or microvascular complications.
T2DM is an independent risk factor for HF (7,8). In asymptomatic T2DM patients with normal LVEF, subclinical myocardial disease includes increased LV mass (10,35,36) and diastolic dysfunction (37–39), and decreased myocardial strain (12,13). However, T2DM patients are heterogeneous in terms of age, sex, diabetes duration, and associated cardiovascular risk factors, such as obesity or HTN. In this context, interpretation of echocardiographic data in everyday clinical practice in T2DM patients is challenging with regard to the respective impact of these confounders on cardiac abnormalities. Indeed, the coexistence of HTN, obesity, and T2DM has made it difficult to isolate the contribution of the glucometabolic state to myocardial dysfunction, and whether this myocardial dysfunction is triggered by T2DM rather than the synergistic action of these risk factors is debated. This is the first application of cluster analysis to identify distinct cardiac phenotypes in T2DM and to relate these to clinical features and outcomes. Our findings underscored the variety of cardiac profiles in T2DM and the need for improving the detection of subclinical LV dysfunction. Interestingly, using classic statistical analysis, based on a priori hypothesis, we were not able to discriminate patients with “isolated” T2DM (without obesity or HTN) and T2DM patients with obesity and HTN on LV systolic parameters, such as LVEF or longitudinal strain. Furthermore, the distribution of all echocardiographic parameters showed a major overlap between those predefined groups of T2DM patients.
As opposed to classic statistical analysis, cluster analysis is an exploratory technique that provides tools to identify unknown subgroups in order to classify individuals with similar characteristics in the same group (or cluster) and individuals with distinct characteristics into different clusters (25).
In our study, cluster 1 was mostly characterized by the lowest LVMi, the most favorable diastolic parameters with the lowest LA area and E/e′ values, and the highest LVEF. Strain values were the second highest among clusters. This cluster represented the most favorable clinical profile, with the youngest patients and the lowest rate of obesity or HTN. This low comorbidity subset had the best prognosis compared to the 2 other clusters.
Cluster 2 cardiac phenotype was characterized by the highest strain values but the lowest e′ velocities and the highest E/e′ ratio. These patients were the oldest and were predominantly female with the highest prevalence of associated cardiovascular risk factors. This elderly diastolic dysfunction group had a worse prognosis than cluster 1.
Cluster 3 was an interesting group. Patients shared a similar clinical presentation of younger age and prevalent associated cardiovascular risk factors to cluster 1 but had a similar prognosis to cluster 2. Hypertrophy and systolic dysfunction was the unifying phenotype.
Thus, our cluster analysis highlighted the prognostic value of LV remodeling and subclinical systolic dysfunction in T2DM, despite similar clinical profiles of obesity and HTN. This suggested that patients with low strain and/or increased LV mass might be suitable for targeted preventive strategies. These data were in line with previous reports of the association between decreased longitudinal strain and impaired exercise capacity (40), LV remodeling (15), and all-cause mortality (17) in T2DM patients.
Our cluster analysis also supported the finding that diastolic abnormalities commonly observed in T2DM are worsened when HTN and obesity coexist, as in cluster 2. Diastolic function impairment has previously been considered as the first marker of diabetic cardiomyopathy (11,16,37). However, diastolic dysfunction is also related to confounding factors, such as age, obesity, or HTN (38). The clinical profile of cluster 2 reinforced those data, showing the association of the lowest e′ velocities and the highest E/e′ ratio with highest age, blood pressure levels, BMI, and heart rate compared with the 2 other clusters and confirmed the higher risk of HF and death in T2DM patients with diastolic dysfunction (16). It is also remarkable that cluster 2 (mainly women with predominant LV diastolic dysfunction) identified a greater susceptibility of women to cardiovascular impact of T2DM because this cluster presented a prognosis similar to cluster 3 (men with LV hypertrophy and systolic dysfunction) and worse than cluster 1 (men with preserved systolic and diastolic function). This was in line with previous data from the population of the Framingham Heart Study, which highlighted how the relative impact of diabetes on cardiac remodeling (41) and cardiovascular adverse events were greater for women than for men (42). Our data suggested, therefore, that clinicians should pay more attention to the follow-up of diabetic female patients, especially when LV remodeling or diastolic dysfunction is present.
The study population comprised patients recruited into 2 different cohorts. However, those 2 cohorts were well-phenotyped, presented similar inclusion and exclusion criteria, and were prospectively collected in experienced centers. The heterogeneity of the respective study populations reinforced the relevance of our cluster analysis. However, echocardiographic parameters relating to LA size were associated with high missing rates, precluding their direct use for building the clusters. Nonetheless, a sensitivity analysis using those features in a subsample of 462 patients found very similar results for cardiac and clinical phenotypes, reinforcing the robustness of our findings. Finally, a low rate of major cardiac events occurred during follow-up, probably related to the inclusion of a very low-risk population by excluding overt cardiac diseases.
Cluster analysis of echocardiographic variables identified 3 phenotypes among patients with T2DM, underscoring that despite similar clinical profiles, patients with an echocardiographic phenotype characterized by the highest LV mass index and volumes and the lowest LVEF and strain are at higher cardiovascular risk. These findings confirmed the key role of echocardiography to early detect subtle systolic myocardial abnormalities and to better identify echocardiographic phenotypes in T2DM.
COMPETENCY IN MEDICAL KNOWLEDGE: Compared to younger patients without obesity or HTN who have normal LV geometry and function, 2 clusters of patients with T2DM are at higher risk for cardiovascular hospitalization or death: 1) obese, elderly women with hypertension and diastolic dysfunction; and 2) men with LV hypertrophy and systolic dysfunction.
TRANSLATIONAL OUTLOOK: Future studies should evaluate the mechanisms of adverse events in patients with T2DM who have these high-risk characteristics and the impact of specific cardioprotective interventions on survival.
The authors are grateful to the staff of the Centre d’Investigation Clinique 0201, Louis Pradel Hospital, who helped in recruiting the patients and collecting the data.
For supplemental tables and figures as well as a video, please see the online version of this article.
This work was supported by a grant from the Société Francophone du Diabète (formerly the Association of French Language for the Study of Diabetes Mellitus and Metabolic Diseases, grant number D20515) and a grant from the Programme Hospitalier de Recherche Clinique (PHRC 2009-A00089-48). This work was also supported by the French National Agency through the Recherche Hospital-Universitaire-Cardiac & Skeletal Muscle Alteration in Relation to Metabolic Diseases and Ageing: Role of Adipose Tissue (RHU-CARMMA) Grant ANR-15-RHUS-0003. Dr Canoui-Poitrine has epidemiologic expertise and participated on a scientific committee for an observational study in Parkinson's disease, and received honoraria from ABBVIE, France. Dr. Moulin has participated in investigator clinical trials for Sanofi, Pierre Fabre, and Merck Sharp & Dohme; has served on the advisory board of Sanofi; has spoken at symposia for Sanofi, Merck Sharp & Dohme, and Novo Nordisk; and has participated in academic congresses for Janssen, Boehringer, Amgen, and AstraZeneca. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Drs. Ernande and Audureau contributed equally to this work. Drs. Marwick and Derumeaux contributed equally to this work and are joint senior authors. Maurizio Galderisi, MD, served as Guest Editor for this paper.
- Abbreviations and Acronyms
- peak late diastolic velocity
- body mass index
- confidence interval
- peak early diastolic velocity
- mitral annular early diastolic velocity
- e′ lateral
- early diastolic velocity at the lateral site of the mitral annulus
- e′ septal
- early diastolic velocity at the septal site of the mitral annulus
- glycosylated hemoglobin
- heart failure
- hazard ratio
- left atrium
- left ventricle
- left ventricular end-diastolic volume
- left ventricular ejection fraction
- left ventricular end-systolic volume
- left ventricular mass indexed to body surface area
- type 2 diabetes mellitus
- Received July 3, 2017.
- Accepted July 31, 2017.
- 2017 American College of Cardiology Foundation
- Cheng S.,
- Xanthakis V.,
- Sullivan L.M.,
- et al.
- Turkbey E.B.,
- McClelland R.L.,
- Kronmal R.A.,
- et al.
- Gidding S.S.,
- Liu K.,
- Colangelo L.A.,
- et al.
- Heckbert S.R.,
- Post W.,
- Pearson G.D.N.,
- et al.
- Kishi S.,
- Armstrong A.C.,
- Gidding S.S.,
- et al.
- Rydén L.,
- Grant P.J.,
- Anker S.D.,
- et al.
- Yancy C.,
- Jessup M.,
- Bozkurt B.,
- et al.
- Ponikowski P.,
- Voors A.A.,
- Anker S.D.,
- et al.
- Devereux R.B.,
- Roman M.J.,
- Paranicas M.,
- et al.
- Galderisi M.
- Fang Z.Y.,
- Yuda S.,
- Anderson V.,
- Short L.,
- Case C.,
- Marwick T.H.
- From A.M.,
- Scott C.G.,
- Chen H.H.
- Holland D.J.,
- Marwick T.H.,
- Haluska B.A.,
- et al.
- Marwick T.H.,
- Gillebert T.C.,
- Aurigemma G.,
- et al.
- Wong C.Y.,
- O'Moore-Sullivan T.,
- Leano R.,
- Byrne N.,
- Beller E.,
- Marwick T.H.
- Narayanan A.,
- Aurigemma G.P.,
- Chinali M.,
- Hill J.C.,
- Meyer T.E.,
- Tighe D.A.
- Sera F.,
- Jin Z.,
- Russo C.,
- et al.
- Derumeaux G.,
- Ichinose F.,
- Raher M.J.,
- et al.
- Nagueh S.,
- Smiseth O.,
- Appleton A.,
- Byrd B.,
- Dokainish H.,
- Edvardsen T.
- Devereux R.B.,
- Pickering T.G.,
- Harshfield G.A.,
- et al.
- Poirier P.,
- Bogaty P.,
- Garneau C.,
- Marois L.,
- Dumesnil J.G.
- Kosmala W.,
- Jellis C.,
- Marwick T.