Female Children and Adolescents With Type 1 Diabetes Have More Pronounced Early Echocardiographic Signs of Diabetic Cardiomyopathy

Eighty children and adolescents with stable type 1 diabetes were recruited from the patient population attending the Diabetes Clinic for Children and Adolescents of the Antwerp University Hospital for this cross-sectional study approved by the hospital ethical review board. Patient characteristics are given in Table 1. All of the patients were treated with a basal-bolus insulin regimen with four or more subcutaneous injections daily. Exclusion criteria were a significant concomitant illness, medication known to modify cardiac function, obvious clinical signs of cardiac disease, persistent microalbuminuria (albumin excretion >15 μg/min in an overnight, timed urine collection), neuropathy, or proliferative retinopathy, assessed by angiofluorography. The control group consisted of 52 healthy volunteers, matched for age and sex (Table 1). BMI (kg/m²) was calculated for both groups. BMI values were converted to standard deviation scores (BMI-SDS) and adjusted for age and sex using the updated British 1990 reference charts (23).

After recording a standard 12-lead electrocardiogram (Mac 5000; GE, Marquette Medical Systems, Milwaukee, WI) and measuring brachial systolic and diastolic blood pressure in supine and upright position by the cuff method (Dinamap automated vital signs monitor; Criticon, Norderstedt, Germany), a complete echocardiographic study with a simultaneous electrocardiogram (standard lead II) was performed on each patient and control subject using a Hewlett-Packard Sonos 5500, equipped with an S8 (3–8 MHz) transducer. Guided by two-dimensional echocardiography, standard M-mode recordings of the left ventricular dimensions and function were obtained. The left ventricular mass, left ventricular mass index, and relative posterior wall thickness were calculated according to the formulae of Devereux et al. (24). Conventional Doppler tracings of the mitral and the tricuspid valve and pulmonary venous return flow were obtained from an apical four-chamber view; the Doppler cursor line was placed in the inlet or the vein at an angle as parallel to flow as possible and in the position of maximal velocity. Doppler values of the aortic and pulmonary flow were measured from the suprasternal and parasternal short axis, respectively. Duration values were corrected for heart rate. Tissue Doppler values of the septal mitral annulus were obtained from the apical four-chamber view. The ratio of early mitral flow velocity (E wave) to early diastolic velocity of the mitral annulus (E/E′) was calculated. The Doppler-derived index of combined systolic and diastolic myocardial performance (Tei index), defined as the sum of isovolumetric relaxation time (IVRT) and isovolumetric contraction time divided by ejection time, was used for quantification of the global left ventricular function (25) (Fig. 1). All Doppler signals were recorded at a speed of 50 mm/s. For each parameter, the average of three cycles was used. All of the measurements were performed and interpreted by the same reader who was blinded to the clinical status.

Dependent on the distribution, values are given as means ± SD or as median with range. Differences between the diabetic and the control groups were evaluated using Student’s t test or Mann-Whitney U test within the whole group and also separately for girls and boys. The influence of age, BMI-SDS, diabetes duration, and HbA_1c on the different variables was examined by Pearson or Spearman bivariate correlation analysis and the respective importance of influence by multiple linear regression analysis. Univariate ANOVA analysis was performed in the female diabetic group to adjust for HbA_1c. A P value <0.05 was considered as a significant difference. Calculations were performed with the Statistical Package for Social Science version 11 statistical software program.

The patients and control subjects were comparable with respect to age, sex, heart rate, and systolic and diastolic blood pressure; however, diabetic patients had significantly higher z-scores for BMI. HbA_1c values were significantly different between female (8.4 ± 1.0%) and male (7.8 ± 1.1) diabetic patients (P = 0.023) (Table 1).

Left ventricular wall dimensions were higher in the diabetic population but reached only statistical significance for left ventricular posterior wall in diastole and relative posterior wall thickness. Systolic function was comparable in both groups, although wall stress was significantly lower in the diabetic group.

The mitral (V maximum and V means) and tricuspid (V maximum) A-wave velocities were significantly higher in the diabetic group with consequently significant lower early filling velocity–to–atrial contraction velocity (E/A) ratios. Pulmonary venous flow parameters were comparable.

Tissue Doppler-derived indexes of the septal mitral annulus showed generally lower velocities in the diabetic population, although it was only significant for the E′ wave velocity, resulting in a significant higher E/E′ ratio.

The Tei index for the left ventricle was significantly higher in the diabetic subjects. This resulted from significantly longer IVRTs, whereas ejection times and isovolumetric contraction times were comparable in both groups. A representative cumulative frequency distribution curve for IVRT in the diabetic and the control group is given in Fig. 2. The Tei index for the right ventricle was difficult to obtain, because it was often impossible to simultaneously record the interval between cessation and onset of the tricuspid inflow and the right ventricular ejection time.

To eliminate the influence of sex, the parameters were evaluated in boys and girls separately. The increase in left ventricular wall dimensions was only statistically significant in girls and not in boys. A striking difference between diabetic and control girls was noted for mitral and tricuspid atrial contraction waves, with significantly lower E/A ratios. The mitral annulus tissue Doppler imaging E′ velocities and thus E/E′ also differed significantly. IVRT was significantly longer in both female and male diabetic patients (Fig. 2 and Table 2).

Age correlated with almost all the parameters as expected, but the correlation was less pronounced in the diabetic group. We found an expected correlation among BMI-SDS and left ventricular dimensions, left ventricular mass, and aortic and pulmonary Doppler velocities within the control group, but this correlation disappeared almost completely within the diabetic group. Correlations for the significant parameters in Table 2 are given in Table 3. In the first multiple linear regression model, with the control subjects and the diabetic patients as a whole group, diabetes was the only important predictor. In a second model, looking for the respective influence of age, BMI-SDS, diabetes duration, and HbA_1c in the diabetic patient group only, age was the only significant predictor. Adjustment for HbA_1c using univariate ANOVA within the female diabetic group showed no influence on the differences found (Table 3).

In this study, M-mode echocardiography disclosed a tendency to ventricular hypertrophy early in the course of type 1 diabetes in the absence of hypertension. Indexes of left ventricular dimensions were increased in young diabetic patients, especially in the female subjects. BMI-SDS was significantly higher in the diabetic group, but this was not correlated with the larger left ventricular dimensions. Several studies have already suggested that diabetes has direct adverse effects on the heart; the Framingham Heart Study identified an association between diabetes and increased left ventricular wall thickness and mass, independent of conventional risk factors, especially in women (1,2). The left ventricular end-systolic wall stress, an indicator of left ventricular afterload, was significantly reduced in our young diabetic population. This could only be explained by the augmented left ventricular posterior wall dimension, assuming that systemic resistance was equal or even higher, as already stated in earlier reports (26). Further on, our diabetic patients had signs of significant left and partially right ventricular filling abnormalities on conventional and tissue Doppler echocardiography, including a decreased peak filling rate and a greater dependence on atrial contraction for ventricular filling (increased peak A velocity with decreased E/A ratio), together with significant smaller tissue Doppler E′ velocities. This resulted in a higher E/E′ ratio. The combination of transmitral flow velocity with annular velocity (E/E′), which integrates transmitral driving pressure and myocardial relaxation, has been proposed and proven useful as a tool for assessing left ventricular filling pressures (27,28). Additionally, the isovolumetric relaxation time of the left ventricle was significantly longer and the Tei index higher in diabetic patients. The pulmonary venous flow parameters did not reveal obvious differences. All of these abnormalities are compatible with early changes in myocardial relaxation; the described differences in flow patterns and time intervals could fit well in a disturbance of the “active” relaxation, an early diastolic event, as described in a recent overview by Zile and Brutsaert (29). However, this can only be established without doubt if these abnormalities are independent of variables known to alter left ventricular filling, especially ventricular load. Interpretation of load is difficult during relaxation, but can be roughly approached by certain noninvasive measurements, including systolic blood pressure and aortic pulse wave velocity, which are related to left ventricular afterload, whereas normalized pulmonary acceleration time reflects pulmonary arterial pressure and, indirectly, left atrial pressure. In our study, these parameters were comparable between the diabetic and control groups. Abnormalities in ventricular relaxation or compliance were already described earlier, certainly in adults but also in young diabetic populations, although these studies dealt with smaller groups or used fewer echocardiographic parameters (16–20,22). Additionally, tissue Doppler findings were already described in an adult diabetic population but not in children and young adolescents (30). Our study confirms the additional value of tissue Doppler in the evaluation of ventricular filling.

We found almost no influence of HbA_1c and diabetes duration on the measured parameters. This relation remains controversial in many studies. Recently, Holzmann et al. (21) reported that left ventricular diastolic function is related to glucose load and glycated hemoglobin in a middle-aged population. Ventricular relaxation could be altered in diabetes, because calcium transport in the sarcoplasmic reticulum may be abnormal (31). Recently, levels of advanced glycation end products, formed by nonenzymatic glycation of proteins or lipids, were reported to be associated with left ventricular diastolic dysfunction in patients with type 1 diabetes (10). Most advanced glycation end products can increase the cross-linking of proteins like collagen and elastin, causing reduced tissue elasticity and decreased protein turnover (8,9).

ACE inhibition and nonselective β-blockers have been proposed to prevent or treat diabetic cardiomyopathy (32,33), but more studies are needed to determine the timing of such an intervention.

It is not clear which mechanism(s) could explain the striking sex difference we have found: could it be the significantly higher BMI-SDS or HbA_1c in girls or are hormonal changes playing a role? Some reports from the literature showed that diastolic function and BMI are inversely correlated in adults; in our study the correlation of BMI-SDS with the measured parameters within the control group almost disappeared in the diabetic group. Therefore, it is highly unlikely that the higher BMI-SDS and HbA_1c explain why girls are more affected than boys. In addition, we found no significant correlation between HbA_1c and the measured parameters in contrast to most of the studies in adult diabetic patients (21,30). Finally, in the multiple linear regression model, taking age, BMI-SDS, diabetes duration, and HbA_1c as variables, only age remained as a significant determinant. Although the data are intriguing and potentially important, it is difficult to know how these changes determine risk or how these early abnormalities translate into overt dysfunction. However, our findings resemble the hypertrophy and the altered myocardial relaxation described in adult diabetic patients. Therefore, we hypothesize that perhaps already a small increase in blood glucose is sufficient to initiate changes in the cardiovascular system and that the long-term metabolic control probably determines the rate of progression of these changes (initiation-progression model). A prospective follow-up of this cohort, with regular echocardiographies, could help to clarify the mechanism(s) and to identify the parameters that are predictive for the development of a diabetic cardiomyopathy.

Young, predominantly female, diabetic patients have significant changes in left ventricular dimensions and myocardial relaxation properties as compared with control subjects and seem to be at risk for developing a “diabetic cardiomyopathy.” Because we found almost no correlations with HbA_1c and diabetes duration, it remains unclear whether these alterations could be modulated by improving the metabolic control.

This study was supported by the Belgian National Foundation for Research in Pediatric Cardiology.

We gratefully thank Professor Guido Van Nooten for critically reviewing the manuscript, and we thank the nurses Anne Gotemans, Maria Van Assche, and Sandra Collet for their practical help.