Introduction

Patients with type 2 diabetes mellitus can develop heart failure without evidence of cardiac ischemia or presence of systemic hypertension [1]. One of the most important mechanisms responsible for cardiac remodeling in these subjects is cardiac autonomic neuropathy that is present in more than one-third of diabetic patients [2]. However, it is difficult to determine straightforward association between cardiac autonomic dysfunction and diabetic cardiomyopathy due to similar mechanisms that cause both conditions. Furthermore, there are no specific criteria to diagnose these disorders. Heart rate variability (HRV) may be used for non-invasive measurement of cardiac autonomic function, while advanced echocardiographic tools, such as speckle tracking imaging, may provide more detailed insight into cardiac function and mechanics in diabetic patients, and better prediction of adverse events in the growing population of diabetic patients [3]. The introduction of the advanced layer-specific strain imaging provides a new insight into LV mechanics in diabetic patients [4].

A recent study that involved 16,415 non-diabetic Hispanic/Latino adults showed that impairment in glucose homeostasis and abdominal obesity were associated with significantly lower HRV [5]. Similar findings regarding association between reduced HRV and insulin resistance and lower insulin sensitivity were revealed also in Japanese adults who were not taking medication for diabetes [6]. Perciaccante et al. revealed that insulin resistance was associated with sympathetic overactivity, especially during night [7], and Chyun et al. demonstrated that cardiac autonomic neuropathy could predict adverse cardiac outcomes in asymptomatic diabetic patients without known cardiac disease [8].

Our study group previously researched LV remodeling in diabetes and prediabetes [9, 10]. However, investigations that involve evaluation of LV remodeling and cardiac autonomic neuropathy in diabetic patients are scarce [11, 12], and usage of 24-h Holter monitoring for HRV is even more infrequent.

The aim of the present investigation was to evaluate cardiac autonomic function, evaluated with HRV, and LV functional and mechanical changes in diabetic patients. Additionally, we sought to determine the possible relationship between HRV and LV remodeling parameters in the whole study population.

Methodology

This prospective observational cross-sectional study included 50 consecutive normotensive uncomplicated patients with type 2 diabetes and 40 normotensive subjects free of cardiovascular diseases. The diagnosis of diabetes was based on the current recommendations [13]. Diabetes was defined if the blood glucose level was ≥7 mmol/l and HbA1c ≥6.5% or usage of antidiabetic treatment [13].

Exclusion criteria were antihypertensive treatment, heart failure, coronary artery disease, previous cerebrovascular events, atrial fibrillation, congenital heart disease, more than mild valvular heart disease, obesity (BMI ≥35 kg/m2), neoplastic disease, cirrhosis of the liver or kidney failure. Patients with >10% premature supraventricular or ventricular contractions were also excluded from the study due to HRV analysis (two subjects).

Clinic BP values were obtained in two separate visits 3 weeks apart. BP was measured by conventional sphygmomanometer in the morning hours by taking the average value of two consecutive measurements in the sitting position 10 min apart. BP was calculated as average values between all the measurements.

Anthropometric measures (height, weight) and laboratory analyses (level of fasting glucose, blood creatinine and urea, total cholesterol and triglycerides) were obtained from all the subjects included in the study. Body mass index (BMI) and body surface area (BSA) were calculated for each patient. The study was approved by the local Ethics Committee, and informed consent was obtained from all the participants.

24-h Holter monitoring

24-h Holter monitoring was performed with a three-channel digital Schiller Microvit MT-101 system (Schiller AG, Baar, Switzerland) and analyzed by the Schiller software (Schiller AG, Baar, Switzerland). The minimum duration of the recording was 18 h (after exclusion of non-sinusal cardiac cycles). Time-domain HRV parameters were calculated on the 24-h, daytime and nighttime recordings after excluding non-sinusal cardiac cycles, according to the guidelines [14]. SDNN was defined as a standard deviation from all normal RR intervals. SDANN, which reflects long-term HRV and therefore mainly sympathetic activity or sympathovagal balance, was defined as a standard deviation of the averaged normal RR intervals for all 5-min segments. rMSSD was calculated as the root mean square of the difference between the coupling intervals of adjacent RR intervals. pNN50 which reflects short-term beat-to-beat HRV and consequently primarily vagal activity was calculated as the proportion of adjacent RR intervals that varied by >50 ms. After power spectral density estimation, standard frequency-domain HRV measures were calculated for 24-h, daytime and nighttime [14]. Low frequency domain (LF) was defined between 0.04 and 0.15 Hz; high frequency domain (HF) was defined between 0.15 and 0.4 Hz; total spectral power (TP) for all intervals up to 0.4 Hz; and ratio of low to high frequency power (LF/HF).

Echocardiography

Echocardiographic examinations were performed by using a commercially available Vivid 7 (GE Vingmed, Horten, Norway) ultrasound machine equipped with a 2.5 MHz transducer. All echocardiographic data were analyzed off-line.

Aortic root diameter was assessed in all the patients according to the current guidelines. Reported values of all 2DE parameters were obtained as the average value of three consecutive cardiac cycles. LV diameters, posterior wall and septum thickness, were measured according to the current recommendations [15]. Relative wall thickness was calculated according to the formula. LV ejection fraction (EF) was calculated by using the biplane method. LV mass was calculated by using the Devereux formula [16], and indexed for the height powered to 2.7.

Pulsed-wave Doppler assessment of transmitral LV was obtained in the apical four-chamber view according to the guidelines [17]. Tissue Doppler imaging was used to obtain LV myocardial velocities in the apical four-chamber view, with a sample volume placed at the septal and lateral segments of the mitral annulus during early and late diastole (e′ and a′), and systole (s). The average of the peak early diastolic relaxation velocity (e′) of the septal and lateral mitral annulus was calculated, and the E/e′ ratio was computed.

2DE LV strain analysis

2DE strain imaging was performed by using three consecutive cardiac cycles [18], and a commercially available software Q-analysis (EchoPAC 201, GE-Healthcare, Horten, Norway) was used for 2DE strain analysis. The frame rate ranged between 50 and 70 Hz. 2DE speckle tracking analysis was done in three apical (four- and two-chamber, long-axis) views and parasternal short-axis view at the papillary muscle level.

Using the same software we traced short-axis views at basal and apical LV levels at the end-diastolic frame and obtained LV basal and apical rotation values (calculated by software). LV twist represents the summation of absolute values of basal and apical rotation and it was also provided by the software.

Multilayer longitudinal and circumferential strains were determined by modified 2DE strain software (Q-analysis). Multilayer longitudinal was assessed in apical four-chamber, two-chamber and apical long axis view, whereas multilayer circumferential strain was evaluated in short axis at the papillary muscles level. The automatic tracking of the endocardial contour was performed in end-systole. Automatically provided LV myocardium tracking by software was carefully verified and the region of interest was manually corrected to ensure optimal tracking of endocardium and epicardium and inclusion of the entire LV thickness including endocardial, mid-myocardial and epicardial layers in all the observed echocardiographic views. After delineating the region of interest, the software was used to divide the LV into six segments in four-chamber, two-chamber, apical long axis view and allow the investigation of three myocardial layers: endocardial, mid-myocardial and epicardial [18]. Then the global longitudinal and circumferential strain of the endocardial, mid-myocardial and epicardial myocardial layers, and the global radial strain of the LV were calculated [19]. LV global longitudinal and circumferential strains were calculated as the average of all three layer-specific strains (endocardial, mid-myocardial and epicardial), while global radial strain was assessed by the software. Global radial strain was assessed by Q-analysis at the level of papillary muscles.

Statistical analysis

Continuous variables were presented as mean ± standard deviation (SD), and the Student t test was used to detect differences between the two groups for the variables that showed normal distribution. Differences in proportions were compared by using the χ2. Pearson’s correlation coefficients were used for determining the correlation between different demographic and echocardiographic parameters and HRV parameters. Almost all HRV parameters, except SDNN and SDANN, were transformed by natural logarithm, before using the t test or linear regressions because of their high positive skewed distribution. The variables which showed p value ≤0.10 were included into the stepwise multiple regression analyses. The p value <0.05 was considered statistically significant.

Results

There was no significant difference in gender distribution and blood pressure values between the two observed groups (Table 1). BMI and BSA were significantly higher in the diabetic patients. Blood levels of glucose and HbA1c were higher in the diabetic patients, which was not surprising considering the inclusion criteria. Creatinine and lipids levels were similar between the diabetic patients and the controls. Only the triglycerides level was higher in the diabetic patients. The creatinine level was similar between the groups, while the urea level was higher in the diabetic patients (Table 1). CRP and fibrinogen levels were significantly higher in the diabetic subjects.

Table 1 Demographic characteristics and clinical parameters of study population
Full size table

Echocardiographic parameters

There is no difference in aortic root diameter between the observed groups (Table 2). LV diameters and ejection fraction are similar between the groups (Table 2). Interventricular septum thickness and LV mass index were higher in the diabetic patients, whereas relative wall thickness was similar between the groups (Table 2). All the parameters of LV diastolic function (transmitral E/A and DT) and LV diastolic filling were deteriorated in the diabetic patients (Fig. 1).

Table 2 Echocardiographic parameters of study population (adjusted for age, BMI and triglycerides level)
Full size table
Fig. 1
figure 1

Parameters of left ventricular diastolic function in study population

Full size image

Tissue Doppler parameter of LV systolic function (s) was significantly lower in the diabetic individuals (Table 2).

2DE strain analysis

All differences in mechanical parameters between the observed groups were adjusted for age, BMI and the triglycerides level. Multidirectional LV mechanics (global longitudinal, circumferential and radial strain) is significantly deteriorated in the diabetic patients comparing with the controls (Table 3).

Table 3 2DE speckle tracking assessment of the left ventricular function in the study population (adjusted for age, BMI, and triglycerides level)
Full size table

LV basal rotation did not differ between the controls and the diabetics, but apical rotation was significantly higher in the diabetic patients (Table 3). Consequently LV twist was also significantly higher in the diabetic patients (Table 3).

Endocardial, mid-myocardial and epicardial longitudinal and circumferential strains were significantly lower in the diabetic patients than in the controls (Table 3; Fig. 2).

Fig. 2
figure 2

Global and layer-specific longitudinal strain in study population

Full size image

24-h Holter monitoring

Heart rates during 24 h, day and night were higher in the diabetic patients than in the controls (Table 4). SDNN, SDANN, rMSSD and p50NN were significantly lower in the diabetic individuals. Similar results were obtained for 24-h, daytime and nighttime LF, HF and TP, which were reduced in the diabetic group (Table 4). There was no difference in 24-h, daytime and nighttime ratio between LF and HF (LF/HF).

Table 4 Heart rate variability parameters in the study population (adjusted for age, BMI, and triglycerides level)
Full size table

Correlation and multivariate regression analysis

SDNN correlated with HbA1c, E/e′average ratio and 2DE LV endocardial longitudinal and circumferential strain. However, only HbA1c and and 2DE LV endocardial longitudinal strain remained independently associated with SDNN (Table 5).

Table 5 Association between two-dimensional left ventricular parameters and heart rate variability indices
Full size table

rMSSD correlated with HbA1c, LV mass index, E/e′average ratio and 2DE LV endocardial longitudinal strain, but only HbA1c (β = −0.262, p < 0.001) and 2DE LV endocardial longitudinal strain (β = 0.173, p = 0.041) were independently associated with rMSSD (Table 5).

24-h LF correlated with age, HbA1c, LV mass index, E/e′average ratio and 2DE LV endocardial longitudinal and circumferential strain (Fig. 3). However, only HbA1c (β = –0.263, p < 0.001), E/e′average ratio (β = –0.184, p = 0.033) and 2DE LV endocardial longitudinal strain (β = 0.179, p = 0.030) were independent predictors of 24-h LF (Table 5).

Fig. 3
figure 3

Correlation between HRV parameters and longitudinal and circumferential endocardial strain in observed population

Full size image

24-h HF correlated with HbA1c, LV mass index and 2DE LV endocardial longitudinal and circumferential strain (Fig. 3). However, only HbA1c and endocardial longitudinal and circumferential strains were independently associated with 24-h HF (Table 5).

Discussion

The current investigation has provided several important findings which should be discussed: (a) HRV is impaired in the diabetic subjects; (b) LV longitudinal, circumferential and radial deformation are significantly impaired in the diabetic individuals; (c) LV apical rotation was significantly higher in the diabetic patients, which is why LV twist was also significantly increased in this group of patients; (d) all myocardial layers (endocardium, mid-myocardium and epicardium) are impacted by diabetes; (e) parameter of long-term glycemic control (HbA1c) correlated with HRV parameters; (f) LV layer-specific mechanics is associated independently with HRV.

Our study showed that HRV parameters were deteriorated in the diabetic patients, which confirms the previous studies regarding the autonomic imbalance in type 2 diabetes [5–8]. Namely, SDNN, SDANN, HF, LF and TP components were significantly lower in the diabetic subjects, which altogether indicate a significant impairment in autonomic nervous system activity. Earlier investigations revealed that HF component was mainly associated with vagal activity, whereas the significance of LF component was more controversial [14, 20]. However, most authors considered LF as a marker of both sympathetic and vagal influences [14, 21]. Interestingly, VLF could also be influenced by other neurohormonal systems such as the renin-angiotensin-aldosterone system [22]. These sympathovagal disturbances could potentially explain the increased risk of cardiovascular events, even in asymptomatic diabetic patients with normal blood pressure control [8]. Interestingly, due to the reduced powers of both LF and HF components, the LF/HF ratio was similar among the observed groups. Nevertheless, this does not eliminate the presence of impaired innervation in sympathetic and parasympathetic nerves in the diabetic patients. These results are concurrent with previous studies with diabetic patients [5–8, 23].

There are several mechanisms that could explain the relationship between autonomic disbalance and diabetes. Most of them refer to biohumoral systems such as renin-angiotensin-aldosterone and sympathetic nervous system or inflammatory processes. Lampert et al. showed that inflammatory parameters CRP and IL-6 were inversely associated with HRV variables [24]. These findings suggest that inflammatory and autonomic processes may be connected, and that many factors for coronary artery disease development may be the consequence of autonomic dysregulation. Endothelial dysfunction could also be one of the important steps that relates HRV and glucose metabolism in the diabetic patients [25].

Previous studies showed that LV structure, diastolic and systolic function, as well as mechanical function, were impaired in diabetic patients [25, 26]. Habek et al. previously indicated the relationship between LV diastolic function and HRV in diabetics [11]. This is entirely concurrent with our present results. Furthermore, our results for the first time showed significant decrease in layer-specific longitudinal and circumferential strain in diabetic patients. Additionally, our results for the first time demonstrated the association between HRV indices and layer-specific longitudinal and circumferential strain. Recently Enomoto et al. revealed decrease in endocardial radial strain, but not endocardial circumferential strain in diabetic patients [4]. Latest studies have shown that E/e′ ratio and LV longitudinal strain are the best negative predictive value for cardiovascular events in diabetic population free of cardiovascular disease [3], which increases the potential value of our present results that connect autonomic disbalance and LV remodeling in diabetic patients. To our knowledge, this is the first study that brings into relationship HRV and LV longitudinal strain in diabetic population. The association between endocardial longitudinal strain with HRV indices is particularly important because it confirms that endocardium is the most susceptible and vulnerable myocardial layer which is the first to show abnormalities, even in asymptomatic patients. Additionally, this is also significant from the technical point of view because some vendors and software packages that we use in echocardiography determine only endocardial strain. Therefore, our findings revealed that this kind of software could promptly determine subtle changes in LV myocardium.

Our findings showed that that LV apical rotation and LV twist were significantly increased in diabetic patients. The LV twist reflects impaired LV relaxation and diastolic function [27], as well as declined systolic function [28] as it was previously reported [29, 30]. In diabetic patients increased twist could serve as a compensatory mechanism that should maintain adequate diastolic filling in the condition of reduced longitudinal strain. This is supported by the fact that diabetic patients in our study in general did not show typical echocardiographic signs of increased LV filling pressure. Parameters of LV diastolic function were definitely worse in diabetic patients in our study, but still not in the range of abnormal LV diastolic function.

There are several important clinical implications of the current study. Assessment of LV global and multilayer longitudinal and circumferential strain, as well as estimation of HRV, may represent effective and sensitive methods in evaluation of preclinical cardiac damage in diabetic patients. LV strain and HRV can be used in the follow-up of diabetic patients, which represents an important clinical implication. Additionally, the tight glycemic control over a longer time period, evaluated by HbA1c level, is related with better cardiac autonomic function, which suggests that HbA1c can be a surrogate marker of autonomic nervous system dysfunction in clinical routine. The inverse relationship between LV mechanics and HRV indices shows that the assessment performed by one of these techniques can provide valuable information about the remaining one. This is very important for everyday clinical practice that frequently allows using only one of these methods.

The autonomic imbalance is associated with increased cardiovascular morbidity and mortality [31], as well as LV structural and functional changes [32] in global population, not only in diabetics [3, 33]. The most important clinical application of our study is the fact that our findings indicate the relationship between glucose metabolism, autonomic nervous system and LV structural, functional and mechanical remodeling that can potentially serve as an explanation for higher cardiovascular and overall morbidity and mortality in diabetic patients. It is important to emphasize that autonomic dysregulation can be efficiently and accurately assessed by HRV obtained from 24-h Holter monitoring or even in a shorter period of time, which represent additional clinical implications of the present investigation.

Considering the fact that diabetes represents one of the most important factors in the development of heart failure with preserved ejection fraction and ischemic heart disease and the fact that its negative influence cannot be detected by simple estimation of LV ejection fraction, we strongly recommend determination of LV longitudinal strain in all diabetic patients.

Limitations

The present investigation has several limitations. First, all the patients with comorbidities have been excluded, which decreases potential generalization of our results. Second, echocardiographic assessment of LV mechanics could be significantly influenced by the quality of ultrasound images. Third, HRV components are indirect measurements of autonomic nervous system and might be weaker predictors of the sympathovagal imbalance than direct measurements such as muscle sympathetic nerve activity. Fourth, this is the cross-sectional study, which does not allow determination of causal relationship between diabetes, HRV and LV remodeling.

Conclusion

The present research reveals that autonomic nervous function and LV structure, function and mechanics are significantly deteriorated in uncomplicated diabetic patients. The results demonstrate a significant relationship between HRV indices, parameter of long-term glucose control (HbA1c) and LV layer-specific deformation. Further follow-up studies with a larger population of diabetic patients are necessary to evaluate a long-term prognostic importance of the interaction between HRV and LV remodeling on cardiovascular morbidity and mortality in diabetic population.