Impaired myocardial deformation and ventricular vascular coupling in obese adolescents with dysglycemia

A total of 39 obese adolescents with BMI ≥ 95th percentile were recruited for this observational cross-sectional study, as previously reported [16]. All subjects were adolescents recruited from Endocrine and Primary care clinics affiliated within the New York University Langone Medical Center and Health and Hospital Corporation Hospitals (Bellevue and Woodhull) seen for obesity related concerns, namely prediabetes and/or type 2 diabetes, dyslipidemia or hypertension. Written informed consent and age-appropriate assent were obtained from all the subjects before participation and the study was approved by the New York University Institutional Review Board. Recruitment began in October 2011 and ended in February 2014. All study measurements were conducted at the Clinical and Translation Research Center (CTSI) of New York University Langone Medical Center. Participants were instructed to come after a 12 h overnight fast to the Clinical Research Center at the NYU Clinical and Translational Science Institute. A total of n = 6 participants was previously diagnosed with T2DM. All remaining participants not diagnosed with T2DM (n = 33) underwent an oral glucose tolerance test (OGTT) using 75 g of glucose. Samples for glucose, insulin, and cytokines were stored at − 80 °C at the Research Center of New York University.

Key inclusion criteria were: (1) age 12–18 years and (2) Tanner stage II or above. Key exclusion criteria were: (1) Use of medications known to affect insulin sensitivity such as high-dose inhaled glucocorticoids, thiazolidinediones and atypical antipsychotics, (2) medications for treatment of dyslipidemia such as statin therapy, bile acids and fibrates, (3) pregnancy, (4) significant psychiatric illness, (5) blood pressure over 90% for height (6) renal insufficiency based on Schwartz equation as > 0.55 for females and males > 0.7 ages 13–18 years and (7) nonalcoholic fatty liver (NAFLD: AST and ALT ≥ threefold higher value above normal at baseline.

Adolescents were grouped by their OGTT results as obese normoglycemia (ONG; n = 25) and obese dysglycemia (ODG; n = 14). Abnormal glucose tolerance was defined as either (1) impaired fasting glucose (fasting glucose level ≥ 100 mg/dl) or (2) impaired glucose tolerance (2-h postprandial glucose ≥ 140 mg/dl after glucose challenge). None of the adolescents who underwent the OGTT were diagnosed with T2DM after the OGTT.

Adolescents diagnosed with T2DM did not undergo an OGTT and were categorized as ODG. The adolescents with T2DM were treated with metformin and/or insulin. Those on metformin were asked to stop metformin 72 h prior to their study day as metformin can affect vascular studies (flow mediated dilatation), which were performed during the study visit, previously reported [16].

Healthy age and sex-matched controls were derived from the Pediatric Cardiology program at the collaborating center, as part of a registry collecting normative data of strain parameters. The control group consisted of children and adolescents who responded to an institutional IRB approved advertisement for participation. In the control group, 2D-STE was performed for research purposes. The control inclusion criteria included: (1) age 1 to 18 years; (2) no history of heart disease, hypertension, or any other systemic disease. Age appropriate assents were signed by the adolescents with parental consent. Data collected at the time of 2D-STE included gender, date of birth, height, weight, heart rate, and systemic blood pressure. BSA was calculated using the Haycock formula. All participants in the control group had no history of congenital heart disease and were lean BMI < 85% with no signs of glucose intolerance. Males were observed to be heavier in the control cohort (M: F; mean ± SD − 60.5 ± 2.1 and 59.5 ± 8.6 kg respectively). Glycemic status was not established in the control using a glucose tolerance test though there was no history of polyuria or polydipsia elicited by the adolescents.

Transthoracic echocardiography images were obtained by an experienced sonographer on a Philips iE33 ultrasound imaging system with a 5 MHz probe (Philips Healthcare, Andover, Massachusetts). Standard echocardiography acquisition included two-dimensional, color flow Doppler, pulsed Doppler, and continuous wave Doppler. 3–5 cardiac cycles were acquired for each measurement with the patient in supine and left lateral decubitus positions per routine clinical protocol. Two dimensional apical 4 chamber and 2 chamber views, as well as parasternal short axis views, were acquired for strain and strain rate analysis. Left ventricular mass was calculated via Devereux’s formula and by 5/6 area-length methods [17]. LV mass index (LWMI) was calculated by dividing LV mass by height^2.7. Parameters of diastolic function were measured, including mitral valve inflow E & A wave peak velocities, tissue Doppler imaging (TDI) of the mitral valve medial and lateral annulus e′ and a′ wave peak velocities, pulsed Doppler measurements of the right upper pulmonary vein S, D, and A wave peak velocities, and the left atrial volume by biplane Simpson’s method [45, 46]. MV E/A ratio, and lateral and medial mitral E/e′ were also calculated. LAVI > 25 ml/ht^2.7 was considered to be dilated. MV E/e′ ratio > 8 was considered to be abnormal diastolic function. Due to the limited dataset, diastolic dysfunction was defined by an abnormality in any of these parameters. Studies were de-identified and stored on a Siemens SyngoDynamics DICOM system (Siemens Healthcare, Erlangen, Germany).

Ventricular vascular coupling index (VVI) was estimated in a manner similar to that presented by Sanz et al. [18]. Briefly, VVI is the ratio between LV end-systolic elastance (Ees) and effective arterial elasticity (Ea). LV end systolic elastance was approximated as,where ESP is the LV end systolic pressure, ESV is the end systolic volume, and V₀ is the theoretical volume of the unloaded ventricle, determined as the intercept o the linear ESP-ESV relation. End systolic pressure was approximated from systolic blood pressure as, ESP = 0.9 × SBP¹⁸. Effective arterial elasticity was approximated as,where ESP is end systolic pressure and SV is stroke volume. VVI is therefore determined as VVI = Ea/Ees.

Analysis to derive left ventricular longitudinal, circumferential and radial deformation and strain by offline post processing (Image-Arena Version 4.6 Build 4.6.2.12, Unterschleissheim, Tomtec, Germany) was performed on two-dimensional DICOM image datasets, described elsewhere [19, 20]. Briefly, strain is the fractional change in the length of myocardial segment, and strain rate is determined as the rate of change in strain (s⁻¹). S and SR were obtained using standard 2D multiform B-mode grayscale LV images acquired in the apical four, two, and three (3CV) chamber views. S and SR were also obtained in the parasternal short-axis views of the LV at three levels, namely the basal, mid, and apical LV, corresponding to LV short axis sections at the mitral valve (MV), papillary muscles (PM) and LV apex (AP), respectively. A frame rate > 60 Hz was used for all analyses. Five consecutive cardiac cycles at the same heart rate triggered to the R wave of ECG complex were digitally stored in a cine-loop format for offline analysis. Five values were obtained for each strain index (S and SR) and averaged. The timing of aortic valve closure and mitral valve opening with respect to peak S and SR was ascertained from manual inspection of a single PW or CW Doppler from the LV outflow tract. All LV images in the apical and short axis views were obtained at the same heart rate. A region-of-interest width and when necessary was automatically acquired, which was manually adjusted to include the entire myocardial wall. The tracking algorithm followed speckles in the myocardium throughout the cardiac cycle. Results from the tracking algorithm were confirmed by visual inspection, and the number of LV myocardial segments was maximized. By convention, the myocardial strain is obtained from three planes in the LV, namely longitudinal, circumferential, and radial. The corresponding strain indices that were determined include longitudinal strain (LS), longitudinal strain rate (LSR), circumferential strain (CS), circumferential strain rate (CSR), radial strain (RS), and radial strain rate (RSR). All strains were reported as dimensionless percentages (%). LS was calculated as the % change in LV length, which was determined for three individual apical long axis views, namely the 4CV, 3CV, and 2CV. Global longitudinal strain (GLS) was calculated as the average of LS and LSR from all considered views. CS and RS were calculated as the % change in LV circumference and wall thickness, respectively. Both CS and RS were assessed from the LV short axis view at three levels, namely basal (MV), mid ventricular (PM), and distal LV at the apex (AP). Regional CS and RS were obtained from six segments in three parasternal short axis views of the LV, which were averaged to obtain global CS and RS. The software automatically divided the LV short axis views into six segments, namely the anteroseptal, anterior, lateral, posterior, inferior, and sepal segments, as per standard guidelines [23]. SR, the time derivative of S expressed in s⁻¹ was obtained at the peak temporal spread of the strain rate acquisition at systole. The peak systolic SR was obtained for each cardiac segment from the same views as strain, discussed in detail above.

Data were summarized as frequency for categorical variables, mean ± standard deviation (SD) for normally distributed continuous variables, and median and IQR for continuous data with a skewed distribution. Differences between groups were assessed using the unpaired t test or the Wilcoxon rank-sum test for normally and non-normally distributed continuous variables, respectively. Differences between categorical variables were assessed using Fisher’s exact test. Logistic regression was used to assess the univariate associations between presence of strain parameters and study variables. Linear regression was used to illustrate the correlation between average LSR/CSR and VVI (Fig. 1, 2). Multivariable associations between myocardial deformation parameters, namely strain and strain rate, and other variables were investigated in a forward-selection model building procedure, while controlling for BMI, DBP, and SBP. All statistical tests were two-sided, with P < 0.05 considered significant. A commercially available statistical software package (SAS v9.3, SAS Institute Inc., Cary, NC) was used for all analyses.

Table 1 summarizes the characteristics of enrolled obese adolescents classified as obese normoglycemia (ONG) and obese dysglycemia (ODG) based on glucose tolerance results from OGTT. Weight and blood pressure were recorded for control subjects, though glucose status was not available. 60% of the cohort was Hispanic. Eight adolescents were found to have prediabetes (IFG and/or IGT) after OGTT. Six adolescents had confirmed T2DM at baseline; thus, a total of 14 adolescents had dysglycemia. There were more males in the ODG group (total = 9; 5 with prediabetes and 2 with T2DM) compared to females (total = 5; 3 with prediabetes and 2 with T2DM). Males were heavier when compared to females (107 kg versus. 93 kg) though they were younger by 6 months when compared to females. T-test P values in Tables 1, 2, 3 reflect differences between ONG and ODG group as control group were matched for age and gender. ODG and ONG groups were matched for BMI (34 ± 7 vs. 37.5 ± 8.8); however, those with dysglycemia were found to have a higher BMI, though not significantly. HbA1c was found to significantly higher (P < 0.001) in the ODG group (7.4 ± 2.3) compared to the ONG group (5.6 ± 0.3), likely because six adolescents were previously diagnosed with T2DM in the ODG group. Significant differences between both ODG and ONG groups were also observed in systolic blood pressure (P = 0.0025).

The Mitral E/A ratio was significantly different (P = 0.0025) between ONG and ODG groups, while no significant differences were observed in index LA volume. E/e′ ratio of lateral and medial mitral valve annulus were not significantly different between the two groups. While only male in ONG had evidence of LA dilatation, five adolescents in ODG had signs of LA dilatation-two (mild); one (moderate) and one adolescents (severe). These results are summarized in Table 2. ESP was observed to be elevated in the ODG group (108 ± 8.1) compared to the ONG group (102 ± 5.9), though not significantly (P = 0.22). No significant differences were observed in LVMI and VVI between both groups.

As previously discussed, ESP was different between both ONG and ODG groups (Table 3). Mean VVI was determined to be 0.7 ± 0.2 for all obese adolescents included in the study, indicating no impairment in ventricular-vascular coupling. VVI was however observed to be negatively correlated with average LS (r = − 0.4; P = 0.025) and circumferential strain rate (CSR) (r = − 0.36; P = 0.04, Figs. 1, 2). The negative correlation between VVI and CSR was also observed in ODG adolescents (r = − 0.7; P = 0.01).

Table 4 compares 2D-STE derived global and average longitudinal strain (LS), circumferential strain (CS), and radial strain (RS) in obese adolescents and lean age matched controls. Briefly, global LS was observed to be significant reduced (P = 0.0005) by approximately 15% in obese adolescents (− 20.98%) compared to lean controls (− 23%). A similar trend was observed in average LS, which was observed to be significantly reduced (P = 0.0027) by approximately 20% in obese adolescents (− 18.8%) compared to lean control (− 20.5%). Results remained significant on multivariate logistic regression analysis for both global LS (P = 0.0045) and average LS (P = 0.014). However, no significant differences were observed in global and average CS and RS between both groups. Global LS were higher in males than females (− 20.1 versus − 19.27) and average LS was lower in males (− 18.2 versus − 19.2) though differences by gender were not significant.

Table 5 compares 2D-STE derived global and average LS, CS and RS between the groups ONG and ODG. Global and average LS was observed to be 10% lower in the ODG group compared to the ONG group (P < 0.05). After multivariate linear regression with adjustment for confounders, namely BMI, SBP, and DBP, average LS, global CS, and average CS retained significance between both groups (P = 0.03).

Subjects in the ODG group were further partitioned into ODG with prediabetes (IFG and/or IGT) and T2DM and compared with ONG subjects. A total of 8 subjects were classified as ODG with prediabetes and 6 subjects as ODG with T2DM. Significant differences (P = 0.01) were observed in global LS between ONG subjects (− 21%), ODG with prediabetes (− 19.71%), and ODG with T2DM (− 19.1%). A similar trend was observed in average LS, in which significant differences (P = 0.04) were observed between ONG subjects (− 19%), ODG with prediabetes (− 18.6%), and ODG with T2DM (− 17.5%). In addition to significant differences in HbA1c, previously discussed, a positive correlation was observed between HbA1c and average CS (r = 0. 23; P = 0.05). Glucose and insulin for time points 0, 3, 5, 10, 30, 60, 90, 120 min during the OGTT were measured. No correlation was observed between time points for both glucose, insulin, mean glucose, mean insulin, and myocardial strain parameters, suggesting that the glycemic status was not related to myocardial deformation independent of obesity.

ICAM was observed to be significantly different between ONG and ODG groups (P = 0.023). Furthermore, a negative correlation was observed between average longitudinal strain rate (LSR) and anti-inflammatory adiponectin (r = − 0.41; P < 0.01) in obese adolescents.

Our study has several limitations. Due to the cross-sectional design, causal relationships cannot be established among the observed associations. Additionally, the predominance of Hispanic subjects, about 60% of the cohort, limits generalizability of the findings. However, the study presented simultaneously provides novel data in for pediatric Hispanic subjects, a population known to be a high risk for diabetes. The reduced sample size in this study reduced power required to detect differences between groups based on OGTT results. Though we had age and gender matched control group, we recognize that the control group was predominantly Caucasian.