Impact of acute hyperglycemia on layer-specific left ventricular strain in asymptomatic diabetic patients: an analysis based on two-dimensional speckle tracking echocardiography

A total of 67 patients with diabetes (Group A), 30 with type 1 diabetes (T1D) and 37 with type 2 diabetes (T2D), both gender (45 males), with an average age of 38.3 ± 1.6 years that had acute hyperglycemia (plasma glucose ≥ 11.1 mmol/l) were included in this prospective study which was performed over a period of 4 years (from January 2014 to January 2018). Patients included in this study were carefully selected from the patients admitted to Intensive Care Unit (ICU) according to the following inclusion criteria: (1) previously known diabetes (32 patients, 47.8%) or newly onset diabetes (35 patients, 52.2%) with hyperglycemia (plasma glucose ≥ 11.1 mmol/l) on the admission; (2) free of previously known or newly diagnosed CVD (including hypertension) or on treatment for CVD or hypertension; (3) free of known or newly diagnosed microvascular diabetic complications. Those patients younger than 18 or older than 65 years of age, with unstable angina pectoris, AMI or history of coronary interventions, heart failure, atrial fibrillation or irregular rhythm on ECG, malignant diseases, obstructive pulmonary disease, hepatic or renal failure (eGFR ≤ 60 ml/min/1.73 m²), acute or chronic infections, ketoacidosis, treated with corticosteroids or immunosuppressive agents and with new uncontrolled hypertension (blood pressure higher than 180/110 mmHg) requiring treatment were excluded from the study.

For comparison, there were two control groups: (1) Group B (n = 20, 10 males, 43.2 ± 3.6 years old) consisting of carefully selected patients with diabetes who had optimal metabolic control (HbA1c ≤ 7.0%) with fasting plasma glucose < 7.0 mmol/l at the time of inclusion in the study and without any known or newly diagnosed micro- or macrovascular diabetic complications; (2) Group C consisting of healthy volunteers (n = 20, 8 males, 36.2 ± 2.0 years old).

This study has been approved by the Institutional Ethics Committee and all patients have signed the informed consent form prior to inclusion. All of the investigated patients (Group A) were admitted to the ICU of the Clinic for Endocrinology, Diabetes and Metabolic Diseases, Clinical Center of Serbia, and main reason for admission was hyperglycemia. The echocardiographic examination was exerted by experienced cardiologist in the Clinic for Cardiology, Clinical Center of Serbia, while all biological samples were processed at Center for Medical Biochemistry, Clinical Center of Serbia.

Upon the admission to ICU in all investigated patients (Group A) a full medical history was obtained and physical exam with anthropometric measurements were undertook. Body mass index (BMI) was calculated according to the standard formula (kg/m²) from body weight and height measured with a digital scale. Intravenous cannulas were inserted in antecubital vein of both arms for blood sampling, as well as for saline, glucose and insulin infusion for the next 72 h. First, blood was drowning for basal laboratory analysis and immediately after that all patients underwent a resting standard echocardiographic examination before any antihyperglycemic therapy was started.

All investigated patients were subjected to the same therapeutic protocol until the point when euglycemia (plasma glucose ≤ 6.0 mmol/l) was achieved (by the 72 h). Patients were treated with a solution of rapid acting human insulin (Actrapid, Novo Nordisk, Denmark) dissolved in a physiological saline solution (Hemofarm, Vršac, Serbia) (1 IU of insulin in 1 ml of NaCl 0.9% saline solution) and apply by using continuous insulin infusion pump (Perfusor Space Infusion, BBraun, Germany) starting at a rate of 6 IU/h. Glycemic values were taken every hour (AccuChek glucometer, Roche, Germany) and infusion rate was further adjusted according to the glucose levels using the standard protocol in order to achieve near-normal glycemia. Additional saline salts, 5% glucose solution, electrolyte replacement and antibiotics were also implemented i.v. At the time euglycemia was achieved (after 72 h of i.v. continuous insulin treatment) and after three months ( ± 15 days) following acute hyperglycemia biochemical analyses and echocardiography examinations (done by the same cardiologist) were repeated.

Subjects in control groups (Groups B and C) were examined only once when a medical history data was obtained together with anthropometric parameters, laboratory tests and an echocardiogram was performed.

All resting standard echocardiographic examinations were performed using Vivid E9 (General Electric). Data were acquired with a 3.5 MHz transducer in the parasternal (long- and short-axis views) and apical views (two- and four-chamber and apical long-axis views). Echocardiographic methods as M-mode, 2D, color Doppler, pulse Doppler and continue Doppler were performed. All definitions and rule for measurements were in accordance with the recommendations of the European and American Society of Echocardiography [17]. The LV end-diastolic (EDV) and end-systolic volumes (ESV) were measured from the apical two- and four-chamber views, and LV ejection fraction (EF) was calculated using the Simpson’s rule [17]. LV diastolic function was evaluated using early (E-wave) and late (A-wave) transmitral velocities, the E/A ratio and the E-deceleration time (DT) obtained by the spectral pulsed-wave Doppler recordings and by tissue Doppler imaging. The peak early diastolic velocity (E′) was measured at the basal myocardial segments, on the apical four-chamber view, at mitral annulus level (Em′ for the septal and El′ for the lateral) and average value was considered. Finally E/E′ ratio was calculated [17].

The two-dimensional speckle tracking echocardiography (2D-STE), as noninvasive ultrasound imaging technique was used for an objective and quantitative evaluation of global and regional myocardial function. The recordings with a frame rate between 50 and 70 frames/s were performed and analysed offline using General Electric software (EchoPAC software version 113 GE Medical Systems). Standard 2D gray-scale images of the LV were acquired at apical four-, two-chambers and apical long-axis view (Fig. 1), as well as parasternal short-axis view at base (mitral valve), papillary muscles and axis level. Global longitudinal LV myocardial strain (GLS) and systolic longitudinal, circumferential and radial strain were calculated offline. Longitudinal, circumferential and radial strain is analyzed on the 18-segment segmentation model. However, radial strain was not further processed because of a problem using 2D acquisition methods or a low number of high-quality images and suboptimal tracking which influenced the consistency of findings. Multilayer longitudinal and circumferential strains were determined by 2D-STE software which automatically creates a region of interest, which contained subendocardial (Endo), mid-myocardial (Mid) and subepicardial (Epi) layers (Fig. 2). Multilayer longitudinal strain was assessed in apical four-chamber, two-chamber and long-axis views, whereas multilayer circumferential strain was evaluated in the short-axis at the three levels as mentioned [18]. In randomly selected 15 patients, 2–4 weeks after performing the initial measurements, strain analysis was repeated by the same observer (G.K.) to determine intra-observer reproducibility. The absolute difference between two measurements divided by the average value of these two measurements was calculated together with interclass correlation coefficient (ICC).

Plasma glucose concentration (commercial test reagent, Abbott Diagnostic) and HbA1c levels (commercial test reagent, SEBIA, France) were determined by spectrophotometry. Serum lipid levels (total and HDL cholesterol (Ch) and triglycerides) were analyzed enzymatically using commercial kit (ADVIA Chemistry), while LDL-Ch was calculated by standard Friedewald formula. Serum concentrations of troponin (commercial test reagent Roche Diagnostics) and N terminal brain fragment (B-type) for natriuretic peptide (NT-proBNP; commercial test reagent Abbott Diagnostics) were determined by the CLIA test.

All continuous variables are presented as mean ± SE, while categorical data are presented as percentages. Sample size for the study was calculated according to previously published results using standard methods with probability (power) 0.8 and type I error probability 0.05 and we estimated that minimum number of investigated patients (Group A) should be 61. The differences between the groups were tested using a one-way analysis of variance (ANOVA) with Bonferroni post hoc test, while Kruskal–Wallis test was used for categorical variables. The statistical analyses were performed using SPSS software, version 20.0 (SPSS Inc., USA). The level of statistical significance was set at p < 0.05.

Baseline demographic and anthropometric characteristics of the investigated subjects are presented in Table 1. All three groups did not differ in respect to gender, age, body weight and BMI. Proportion of patient with T1D and T2D was similar in group A and group B, as well as average duration of diabetes. The analysis of the previous medication revealed similar pattern the drugs used related to diabetes (oral hypoglycemic agents or insulin), while none of the investigated subjects used any cardiovascular drugs.

The biochemistry analyses performed in the investigated patient groups are shown in Table 2. At baseline, we found significantly higher blood glucose levels in group A (upon hospital admission; first examination) than in both groups B and C (p < 0.001, respectively), while there were no differences when we compared groups B and C. After 72 h hours of insulin treatment in group A (second examination) we achieved normalization of the glycemic values which slightly increase after three months (third examination) (p < 0.001). Similarly, HbA1c level were significantly higher in group A in comparison to both groups B and C (p < 0.001, respectively) and significantly decreased after three months in group A (p < 0.001).

Regarding the lipid parameters, total and LDL-Ch at baseline were similar between the groups, although we found in group A significantly lower levels of both total and LDL-Ch levels at second and third examinations (p < 0.05). In contrast, HDL-Ch was significantly lower at baseline in group A vs both groups B and C (p < 0.001, respectively) and remain unchanged at second and third examinations in group A, while we could not find any differences between groups B and C.

Also, triglycerides value was significantly higher at baseline when we compared group A with both groups B and C (p < 0.05, respectively). In group A, triglycerides significantly decreased at second examination (p < 0.05) and remain unchanged at third examination (after three months).

Simultaneously, the levels of troponin and NT-proBNP were similar across all three groups at baseline, while we registered significant lowering of both parameters at second and third examination in group A (p < 0.05 and p < 0.001, respectively).

LV EDV and ESV, LV EF, as well as left atrium volume index (LAVI) were similar between the groups at baseline. Also, there were no changes in all those parameters in group A at second and third examinations (Table 3).

When we analyzed LV diastolic function we found that both diabetes groups A and B have significantly lower transmitral E wave and E/A ratio compared to healthy controls (group C), while we did not find any differences in transmitral A wave. E/E’ ratio had a trend to be higher in group A compared to both groups B and C at baseline, but did not reach statistical significance. Additionally, in group A, observed lower E and E/A ratio remained unchanged after euglycemia was achieved (second examination) or after 3 months (third examination) (Table 3). In contrast, although there were no significant differences in DT across the groups at baseline, we found that in group A, DT was significantly shorter at both second and third examinations in comparison to baseline (p < 0.001) (Table 3).

We found significantly lower GLS (− 19.6 ± 0.4%) in Group A in comparison to both groups B (− 21.3 ± 0.4%; p < 0.05) and C (− 21.9 ± 0.4%; p < 0.01) at baseline, while we could not detect the differences between groups B and C. Moreover, in group A, after euglycemia was achieved (second examination) we did not find any changes in GLS (− 19.0 ± 0.3%; p = ns) which remain the same at third examination (after 3 months) (− 19.6 ± 0.3%; p = ns) (Fig. 3).

When we analyzed peak systolic longitudinal strain, we detected that Endo, Mid and Epi myocardial layers strain are significantly lower in group A at baseline compared to both groups B and C (Table 4). In addition, observed lower Endo, Mid and Epi peak systolic longitudinal strain in group A remain unchanged at second and third examinations (Table 5).

Similarly, we also found deterioration in peak systolic circumferential strain: at basal LV level, all three layers strain (Endo, Mid and Epi) were significantly lower in group A in comparison to group C, while only Mid layer strain was lower in comparison to both groups B and C (Table 4). However, identified changes in group A have not improved at second (after euglycemia achieved) or third examination (after 3 months) (Table 5). In contrast, at mid-cavity LV level, only in Epi layer we found significantly lower peak systolic circumferential strain in group A vs group C, but not vs group B, which remain unchanged during the follow up period (Tables 4, 5). In apical LV level, there were no difference in peak systolic circumferential strain at any layer (Endo, Mid and Epi) among the investigated groups (Table 4).

The intra-observer variability for longitudinal and circumferential strain was 4.2% and 5.1% (respectively). ICC for intra-observer variability of the corresponding values was 0.899 (95% CI 0.607–0.969; p < 0.001) and 0.721 (95% CI 0.220–0.904; p < 0.01).

Previous investigations have been suggested that in patients with diabetes cardiac dysfunction (such as LV diastolic dysfunction, but also LV systolic dysfunction), defined as separate entity “diabetic cardiomyopathy”, could be detected very early in the clinical course of diabetes, independent of other cardiac risk factors (coronary artery diseases or hypertension) which could progress to heart failure [19, 20]. In that context, echocardiographic studies confirm early diastolic dysfunction in asymptomatic T2D patients [21] which correlate with the level of HbA1c and that a 1% increase in HbA1c was associated with 8% increase of the heart failure risk, independent of other CV risk factors [22]. On the other hands, studies in children and adolescents with T1D found diastolic dysfunction despite intensive insulin treatment and good metabolic control [23‐25] demonstrating the existence of high risk for cardiac dysfunction in those patients, independent of the HbA1c level. Similarly, 4 month of strict glucose control in T2D patients, insulin based, does not improve the parameters of diastolic dysfunction in T2D [26]. Moreover, important echocardiographic changes in diastolic function were observed even in patients with prediabetes [27] suggesting that metabolic milieu in T2D, not only chronic hyperglycemia, could influence the cardiac diastolic function. We found in our study that both groups of diabetic patients, with poor and good metabolic control (both groups A and B), have similar changes in E velocity and E/A ratio implying the presence of diastolic dysfunction (regardless of HbA1c level) in comparison to healthy controls. Interestingly, according to our results, acute hyperglycemia does not additionally impair observed LV diastolic dysfunction in these groups of diabetic patients, except the identified changes in transmitral DT which is slightly, but not significantly increase compared to controls. After euglycemia was achieved, DT significantly decline implying that changes in LV diastolic filling patterns might be one of the consequences of acute hyperglycemia in diabetes.

Strain analysis derived from 2D-STE enables detection of subclinical myocardial systolic dysfunction beyond conventional LV echocardiographic assessment in all three myocardial layers. Studies on layer-specific LV multidirectional strain reported impairments in both longitudinal and circumferential strains in asymptomatic T2D patients which correlated with diabetes duration [28] or overweight [29, 30], while others found longitudinal systolic dysfunction and impairment associated with subendocardial wall thickening, but in poorly controlled asymptomatic and normotensive T2D patients [31]. In addition, STE-derived layer-specific cardiac dysfunctions correlate with the markers of histological remodeling (hypertrophy and fibrosis) and with progression of untreated diabetes in animal models [32, 33]. Using more accurate and sensitive technique, contrast-enhanced cardiovascular magnetic resonance, layer-specific subclinical systolic dysfunction was detected in uncomplicated T2D patients which was associated with impaired coronary microvascular perfusion [34]. Also, it was shown in asymptomatic T2D patients with good glycemic control presence of impairment of LV longitudinal strain but preserved circumferential and radial strain [35], as well as changes in GLS which could be important prognostic factor for the future CV event [36, 37]. In contrast, in 390 patients with chronic ischemic cardiomyopathy (25% with diabetes) endocardial circumferential strain detected by 2D-STE appears to be a better predictor of future cardiac events than total myocardial scar assessed by cardiac magnetic resonance imaging [38]. Very recently, detail analysis of LV mechanics in asymptomatic, hypertensive diabetic patients revealed lower longitudinal strain in Endo layer and lower circumferential strain in Endo and Mid layer compared to healthy controls, which also correlated significantly with HbA1c levels [39]. However, in the study of T2D patients, stratified according to the overall metabolic risk factors, patients with well-control risk factors (including near-normal level of HbA1c), similar to our group B, had only LV diastolic impairment, but there was no evidence of structural or systolic changes of the LV function detected by strain analysis [40].

In our group of asymptomatic, normotensive diabetic patients with strict metabolic control (mean HbA1c 6.1 ± 1.0%; Group B) we did not find any statistically significant differences in GLS, peak longitudinal or circumferential strain in comparison to healthy controls (Table 4, Fig. 3). In contrast, in diabetic patients with really poor metabolic control and acute hyperglycemia (Group A, mean HbA1c 11.5 ± 0.3% and glycaemia 22.5 ± 1.1 mmol/l) we found significantly lower GLS, peak longitudinal strain in Endo, Mid and Epi layer, peak circumferential strain in all three layers at basal LV level and in Epi layer in mid-cavity LV level, compared to healthy controls and diabetic patients with good glycemic control (Table 4, Fig. 3). Thus, these results suggest that acute hyperglycemia in asymptomatic diabetic patients have negative impact dominantly on LV systolic function at all three layers (longitudinal and circumferential strain).

To the best of our knowledge, this is the first study investigating the effect of acute hyperglycemia on LV diastolic and systolic function using 2D-STE in diabetic patients without CVD or hypertension. Our results shown that acute hyperglycemia in asymptomatic diabetic patients induces significant LV systolic changes in multilayer myocardial strain, particularly in GLS, peak longitudinal and circumferential strain which remain unchanged after short-term (after 72 h) and long-term (after 3 months) euglycemia was achieved.

Assessment of the effects of acute hyperglycemia on the LV contractile function in diabetic patients has only been previously the subject of a few research papers. In one study, Nielsen R et al. [41] unexpectedly revealed that short-term hyperglycemia (9–12 h) induced by insulin discontinuation, increased LV contractile function (detected as strain rate) in relatively small number (20) of T2D patients with or without heart failure. This result was contrary to previously findings that hyperglycemia is considered harmful in patients with heart failure, particularly when heart failure coexists with diabetes [42]. However, in this mentioned study [41], after insulin discontinuation the level of blood glucose rise till mean 9.9 ± 2.1 mmol/l in diabetic patients without heart failure, while in our group of asymptomatic diabetic patients hyperglycemia was significantly higher and we analyzed not only global strain (like in Nielsen study) [41], but all three LV myocardial layers which could explain different results.

Some of the previously conducted studies have shown that hyperglycemia following a glucose infusion in diabetic patients with normal LVEF, had no effect on the global LV function [43] detected by standard echocardiography, while we used more sophisticated method (strain analysis) for detection of LV systolic function. Studies in animal models indicated that acute hyperglycemia could suppress LV diastolic function and influences mitochondrial energy signaling [44], or induces changes in polyol metabolic pathway with consequently increase in oxidative stress leading to myocardial contractile dysfunction [45]. Compelling evidence is accumulating which suggests a role of oxidative stress as a pathogenic factor underlying the negative effect of acute hyperglycemia on cardiovascular system [9] which could partly explain our results.

Finally, according to our result, the observed changes in LV systolic function during acute hyperglycemia are not reversible neither after short-term (72 h) nor after longer (3 months) period of good glycemic control, although some studies have shown that amelioration in glycemic control over a 1 year period led to improvements in LV systolic function [46]. Several studies suggest the concept that negative early effect of hyperglycemia on cells in diabetes is remembered, the phenomenon named “metabolic memory”, primarily based on mitochondrial damages, overproduction of reactive oxygen species and oxidative stress [47]. Considering that phenomenon, it could be possible that the risk of hyperglycemia-related vascular complications in diabetes persist even when hyperglycemia is normalized, as previously shown in long-term (several years) trials [48, 49]. However, in our study, the follow-up period with good glycemic control was only three months which possibly would not be long enough to see improvement in the observed LV systolic dysfunction.