Three primary and two secondary schools were randomly chosen. All pupils from the 2nd grade, or 5th grade, respectively, were invited to participate in this cross-sectional prospective study from April to July 2017. Exclusion criteria included structural or congenital heart disease. All parents and children gave written informed consent before participation. The study was approved by the Institutional Review Board (No. 7290) and performed according to the Declaration of Helsinki.

Height, weight and waist circumference were measured, waist-circumference z-score was calculated according to Sharma et al. [8]. Body surface area (BSA) was calculated according to Du Bois [9]. Body mass index z-score (BMIz) was calculated using WHO reference growth standards [10]. We separated our study group according to children’s BMIz in a group with a BMIz < 1.04 and in an overweight/obese group with a BMIz ≥ 1.04. Age was classified into a group of 7–9 and 10–13 year olds, enabling the comparison of echocardiographic parameters to the published reference values from Eidem et al. [11]. Blood pressure (BP) measurement was performed in accordance with current guidelines [12, 13]. In short, BP was measured in a seated position after 5 min of rest, three times on both arms, with at least a 2-min interval between measurements. We used the oscillometric Dinamap device (Dinamap v150; Fa. GE Medical Systems, Chicago, Illinois, USA). The mean was calculated from the second and third measurements, and normalized for sex, age, and height, expressed as z-score [12, 13]. Physical fitness was assessed by graded exercise tests on an ergometer bicycle (sitting position) according to a modified Godfrey protocol [14]. The workload was increased stepwise by 15 W at 1-min intervals. All subjects were encouraged to exercise until exhaustion (breathlessness and leg muscle pain and/or a heart rate ≥ 85% of maximum (calculated 220—age in years) [15].

Transthoracic echocardiography was performed by two uniformly trained investigators. Both were specialized on pediatric echocardiography and followed a standardized protocol using a PHILIPS CX 50 ultrasound machine (Philips Medical Systems, Bothell, USA) equipped with a 5 MHz transducer. Inter- and intra-observer variability was within the range formerly reported by Colan et al. [16]. A segmental analysis was performed to assure segmental anatomy and to exclude a congenital heart defect in accordance with the Guidelines of the American Society of Echocardiography [17].

Left and right ventricular end-diastolic wall thickness and end-diastolic dimensions were obtained from the parasternal short axis view at the level of the papillary muscles using M-mode. LVMI was defined as LV mass/ht^2.16 as proposed by Chinali et al. [18]. LVM z-score (LVMz) according to height was calculated as described by Foster et al. [19], LVMIz-for-lean body mass (LBM) was calculated as proposed by Foster et al. [20], both methods of normalization were used for further analysis. Left ventricular end-diastolic dimension z-score (LVEDdz) was calculated according to Lopez et al. [21]. Ejection fraction was measured by the biplane modified Simpson method. For diastolic function pulsed Doppler measurements of mitral and tricuspid inflow (E-, A-wave) were performed in the apical four-chamber view, with the sample volume positioned between the tips of the mitral leaflets within ± 15° of the central volume stream, the E/A ratio was calculated. The time intervals of isovolumetric relaxation time (IVRT) and isovolumetric contraction time (IVCT) were measured from the apical five chamber view using pulse-wave (PW) Doppler positioned to record left ventricular inflow and outflow tract simultaneously. To determine the IVRT, the time interval from the ending of the aortic flow to the beginning of the mitral inflow was measured. For the determination of IVCT, the time interval from the ending of the atrioventricular valve inflow to the beginning of the ventricular outflow was measured. For the further assessment of diastolic function the following PW Tissue Doppler parameters were obtained: peak velocities of early (e′) and late diastolic (a′) excursion of the lateral and septal mitral annulus as well as of the tricuspid annulus. IVRT, obtained by tissue Doppler, was measured from the end of the S-wave to beginning of the following e-wave. Pulmonary venous inflow (PV) was measured in an apical four-chamber view, with the PW sample volume placed as far as possible in the right upper pulmonary vein. The recordings comprised peak systolic velocity (S-wave) as well as peak diastolic velocity (D-wave). Five consecutive cardiac cycles were recorded from every approach. All parameters were measured five times, the median value was taken for further analysis. All measurements were performed by a single investigator. For any given structure, measurements were made only if excellent and unambiguous views were available. Thus, not all structures were measured in all patients.

Our endpoints comprised different echocardiographic parameters (LVMI, LVMIz, mitral E/A ratio, mitral annular E/e′, septal annular E/e′, septal annular IVRT, tricuspid E/A ratio). Data are given as mean ± standard deviations (SD) or absolute and relative frequencies. Echocardiographic parameters were compared between age, sex and BMI categories. Differences in echocardiographic parameters between boys and girls were assessed using t test. Potential covariates were selected for each outcome variable based on prior knowledge. Backward linear regression modeling with a p value of 0.2 or less as selection criteria were performed. Linear regression models for LVMI, LVMz, LVMIz, septal annular IVRT and PV flow D-wave velocity including interaction term between BMI and sex as well as stratification by sex were performed to investigate the varying extent of risk factors on diastolic function and LVMI in boys and girls. A p-value of 0.05 was considered statistically significant. Statistical analyses were performed using SAS 9.4 (Statistical Analysis Software, Cary, North Carolina, USA).

Of the 356 children initially examined, 351 (187 boys; 53%) were included in further analysis. Four children were excluded due to preexisting heart conditions. Two-hundred-two were 2nd grade (8.21 ± 0.52 years of age) and 149 were 5th grade pupils (11.41 ± 0.55 years of age). Both sexes were equally represented in both groups. Anthropometrics, demographical characteristics and clinical details are shown in Table 1. One-hundred-two children (29%) had a BMIz ≥ 1.04 and 55 (16%) had a BMIz ≥ 1.64. The prevalence of overweight children was similar between 2nd and 5th graders (27% vs. 32%; p = 0.2814), while the prevalence of obesity tended to be lower in 2nd grade children (13% vs. 19%; p = 0.0987). The mean BMIz was elevated in both age groups (2nd grade: 0.28; 5th grade: 0.45). The mean waist-circumference z-score increased with age (0.69 vs. 0.91).

We present our data according to children’s BMIz (< 1.04 and ≥ 1.04) and in comparison to the current reference values published by Eidem et al. [11] (Table 2; Fig. 1). Statistically significant differences between children and adolescents with a BMIz ≥ 1.04 compared to those with BMIz below 1.04 were found for several structural and functional parameters. Left ventricular end-diastolic diameter z-score (LVEDdz) and LVMI [13] were significantly higher in overweight and obese children (p =  < 0.0001 and < 0.0001). Furthermore, several diastolic parameters were out of the normal range. While we cannot compare our values with those of the reference studies statistically [11], one can appreciate that for most values the exclusion of overweight and obese children resulted in more favorable values regarding function and morphology. According to the published reference data from Eidem et al., the number of out of range values are presented in Table 2. We only report those cases, in which the measurement pointed toward a potential unfavorable abnormality concerning diastolic function. The comparison of girls and boys revealed significant differences for the mitral and tricuspid E/A ratio as well as for the systolic PV flow (Supplementary Table S1).

In the following, we will highlight particular morphological and functional aspects of the cohort.

As indicated above, higher BMI z-scores were associated with increases in LVMI (p < 0.001, Fig. 2). To explore which other factors contributed to a higher LVMI, we performed a multivariable linear regression analysis (Table 3). In addition to BMI, we found that sex, age and physical fitness were independent predictors of a higher LVMI. Importantly, we could not find an association between LVMI and systolic or diastolic BP in our cohort of healthy school children.

We further explored sex differences and found higher LVMI in boys than in girls for children with a BMIz < 1.04 (p < 0.0001), but this difference was no longer observed in overweight and obese children (p = 0.096, Fig. 3). We performed a sex stratified analysis to investigate how changes in BMI affect LVMI in each sex (Table 4). Our results support the observation that girls and boys with higher BMI values have comparable absolute LVMI values. We show that, e.g., at a BMI value of 30 kg/m² girls are expected to have an estimated LVMI of 37.9 g/m² whereas boys will have 39.5 g/m². This convergence at higher BMI values is the result of a greater increase in LVMI for a given BMI change. Per 1 kg/m² BMI increase LVMI increases by 0.5272 g/m^2.16 in boys, whereas by 0.7336 g/m^2.16 in girls.

We also calculated LVMIz [20], expressing LV mass relative to LBM, meant as sensitivity analysis. Similar to our calculations based on LVMI by Chinali et al., we only saw very few children with LV hypertrophy. Using LBM normalized values we did not expect to see an effect of BMI. However, we confirmed our previous observation that higher BMI values were associated with smaller increases in LVM in boys (boys: β =  − 0.033, p = 0.045; girls: β =  − 0.005, p = 0.72) (Supplementary Table S2).

As indicated above overweight and obese children also tended to have more unfavorable echocardiographic diastolic parameters than children with BMIz < 1.04. Early mitral and tricuspid inflow velocities (E) were similar between the groups, but both mitral and tricuspid A-wave velocities were increased, resulting in a decreased E/A ratio. The values for the E/e′ ratio (septal and mitral annular) were higher in overweight and obese children and the IVRT measured at the septal annulus was significant longer. The effect of increasing BMIz is also highlighted in Fig. 1.

Multivariable linear regression analysis was used to further explore contributing factors for the selected diastolic parameters (Table 3). The E/A ratio of the mitral valve was independently associated with sex, BMI and diastolic BP. Girls showed a significant lower E/A ratio compared to boys while age had no effect. Interestingly higher BMI was associated with an increase in the A-wave, while the E-wave essentially was left unchanged. E/e′ ratio (either septal or mitral) were only predicted by BMI without a contribution of sex and age. IVRT was independently associated with age, BMI and diastolic BP, while heart rate was inversely related.

We assumed again an interaction with sex. By creating an interaction term (Table 4; Fig. 4a), we demonstrated a significantly higher increase for IVRT in girls for a given BMI. Sex stratified regression models showed how BMI affect IVRT in boys and girls differently. We found a prolonged IVRT of 0.5525 ms for a given BMI in girls compared to boys. The stratified regression analysis demonstrated that a BMI increase of 1 kg/m² was associated with an IVRT increase of 0.4507 ms in boys (p = 0.028), while in girls IVRT extended significantly more with 1.0609 ms (p < 0.001).

Interestingly this phenomenon was also found in the association between reduction of diastolic PV flow and BMI, in which girls showed a reduction of 0.54 m/s with a BMI increase of 1 kg/m² compared to boys. The sex stratified regression model confirmed a reduced diastolic PV flow with increasing BMI in girls (Fig. 4b; Table 4).

Sex differences in LVMI as an important structural parameter have been repeatedly reported [22]. However, information about sex-related differences in diastolic echocardiographic parameters in children is scarce most likely due to the limited number of cases available in earlier reference studies [23, 24]. Values for several echocardiographic parameters measured in girls from our study were found in the respective upper or lower ranges of normality, indicating a real biological difference between the sexes. Apart from those physiological differences, we demonstrate that in girls the same BMI increase resulted in a greater increase in LVMI as well as a greater increase in IVRT and diastolic PV flow compared to boys. These findings suggest that there are sex-related negative effects of being overweight or obese on mass as well as compliance of the left ventricle with increasing BMI. We therefore assume a higher susceptibility toward cardiac changes in presence of increasing BMI as an important cardiovascular risk factor in girls compared to boys. Our findings in these otherwise healthy but overweight children are in accordance with data from a large cohort of children with chronic kidney disease, in whom a stronger association of obesity with LVMIz and LVH was demonstrated among girls compared to boys [7]. A greater vulnerability of girls upon cardiovascular stresses had been demonstrated in older girls after renal transplantation that displayed higher BP values when exposed to an additional risk, i.e., higher trough levels for cyclosporine A [25]. Further support comes from recent adult data. The protective effect of female sex on CV health is lost in presence of additional comorbidities and stresses (such as type 2 diabetes, hypertension, hypercholesterinaemia, sedentary lifestyle, mental stress) [26‐28].

While peripubertal and perimenopausal differences point toward a hormonal cause for CVD risk [29], our data suggest that risk factors earlier in life must be taken into account, as most of our study population had not reached puberty yet. A possible explanation for our findings is a significantly higher adipose mass in females than in males for each BMI category, as described by De Simone et al. [30] and for pediatric age by Taylor et al. [31]. Adipose tissue is considered to be a metabolic highly active tissue and a large endocrine organ [32] expressing several hormones, growth factors and cytokines [33, 34]. It has been shown that especially in women adiposity and inflammation pathways are highly relevant in the development of CVD [35]. Lau et al. [36] demonstrated sex differences in circulating biomarkers, with significant higher levels of leptin and ceruloplasmin in women than in men. For leptin levels, there were found associations with diastolic function (E/e′) in adults, after adjusting for age, sex and BMI [37]. Higher ceruloplasmin levels were associated with heart failure and were weakly associated with CVD [38]. Unfortunately we do not have information on leptin, ceruloplasmin or other factors derived from adipose tissue of our cohort. Notably, a correlation of high-sensitive C-reactive protein with echocardiographic parameters was not found.

The importance of obesity for the different echocardiographic parameters has been controversially discussed especially concerning echocardiographic parameters of diastolic function. But in nearly all studies, indices of LV mass were greater in overweight and obese than in normal weight children or adolescents [39‐42]. It should be noted that, the prevalence of left ventricular hypertrophy in overweight children varies dependent of the method used for normalization [43]. Children in our study showed a higher LVMI and LVMz with increasing BMI, which is in agreement with previous pediatric studies, describing a correlation between BMI and LVMI [4, 5, 39, 44‐47]. In agreement with recent results from larger cohorts of children [41] the influence of BP, often discussed as possible reason for a higher LVMI in overweight children [4], could not be confirmed in our cohort.

In adulthood early subclinical changes of diastolic parameters are very sensitive indicators for disturbances of myocardial energy metabolism [48]. It is known that, overweight and obesity lead to energetic abnormalities and insufficient energy supply of the myocardium [49], but the evaluation of diastolic function is still a very complex field of science. Our data on lower LV compliance in case of overweight and obesity resemble data from larger studies [5]. Di Salvo et al. investigated 150 obese children and adolescents and showed a significant higher E/e′ ratio as well as a longer IVRT in these subjects in comparison to a control group [50]. Dhuper et al. presented a higher E/e′ ratio and lower mitral E/A ratio in 213 obese children, while subjects with hypertension were not excluded and the effect of obesity solely could not be rated [46]. While some smaller studies could not show any differences between overweight and normal weight [44], others did but with heterogeneous results for the different parameters describing diastolic function [4, 5, 40]. Especially a difference in E/e′ and E/A ratio could be observed in several studies [4, 39, 46, 51]. A possible explanation for these partly contradicting results may be that in presence of chronic volume overload—as assumed in overweight and obesity—only minimal changes in tissue Doppler velocities can be expected [52]. Studies with rather smaller sample sizes might have had not enough power to show significant differences in all the parameters.

Our findings suggest that changes in cardiac structure start early in life and occur at much lower BMIz level than expected. This implicates an urgent need for an effective obesity prevention in children and adolescents, with a special focus on obesity-related cardiac alterations in girls.

Potential limitation of the current study is the lack of information about the pubertal status of our study participants. Since sex-specific hormone levels play a significant role for cardiovascular health in adults, a closer investigation of pubertal status would have been favorable. This cross-sectional study design does not provide information about the potential progression of the described cardiac alterations in adulthood or potential reversibility. So the clinical significance of these findings remains unknown and will require longitudinal follow-up over several years to determine their predictive value. LVMI and left ventricular diastolic function were probably also influenced by other factors that could not be examined, e.g. hyperinsulinemia or leptin levels.