Data came from the baseline survey of “Huantai Childhood Cardiovascular Health Cohort Study” conducted in Huantai County, Zibo City, Shandong Province, China, conducted from November 2017 to January 2018. The survey has been described elsewhere (23). In brief, 1,406 children aged 6–11 years from one primary school were included in this study. All children and their parents/guardians provided written informed consent before examinations. Collected data included demographic information (sex and age), anthropometric examinations [height, weight, waist circumference (WC) and blood pressure (BP)], lifestyle factors (sleep duration, screen time, physical activity and intake of vegetables and fruits), fasting blood assays (glucose and lipid profiles) and an echocardiogram examination. The study protocol was approved by the institutional review board at the School of Public Health, Shandong University (approval number: 20160308).

Anthropometric examinations were performed by trained staff according to a standard protocol in the school setting. Height and weight were measured twice in light clothes without shoes using a calibrated scale with stadiometer (HGM-300; Shengyuan Co. Ltd., China). WC was measured twice at 1 cm distance above the horizontal level of umbilicus using a flexible plastic tape. Mean values of height, weight, and WC, as well as body mass index (BMI, kg/m²), respectively, were calculated for data analysis. The BP in a seated position was measured three times consecutively at the heart level at about 20-second intervals, using the validated and calibrated electronic sphygmomanometer (HEM 7012; Omron, Osaka, Japan) (24). The last two BP readings were averaged for data analysis.

Venous blood samples were drawn after an overnight fast (for at least 10 h). Fasting blood glucose (FBG), TG and HDL-C were assayed using an automatic analyzer (Beckman Coulter AU480; Mishima, Shizuoka, Japan).

A same experienced sonographer who was blinded with the study protocol measured the left ventricular structure using a portable color Doppler ultrasound machine (CX30; Royal Philips, Amsterdam, the Netherlands) equipped with an S4-2 convex array transducer (frequency of 2–4 MHz), according to recommendations proposed by the American Society of Echocardiography (25). Using the R-wave apex of the electrocardiogram as the standard phase, the left ventricular end-diastolic internal dimension (LVID) was obtained by measuring the echogenicity of the left ventricular septum from the left ventricular surface to the left ventricular posterior wall. In addition, the vertical distance between the two points on the time axis was measured from the upper edge of the front edge echo line to the upper edge of the rear edge echo line of the measured structure to obtain interventricular septal thickness (IVST) and left ventricular posterior wall thickness (LVPWT). Intra-observer reproducibility of IVST values [intra-class correlation coefficient (ICC) = 0.92] and LVPWT values (ICC = 0.95) was evaluated in the same 20 subjects by the same sonographer who measured twice. Left ventricular mass (LVM, g) was calculated using the Devereux's formula as 0.8 × 1.04 × [(IVST + LVID + LVPWT)³−(LVID)³] + 0.6 (26). Additionally, left ventricular mass index (LVMI, g/m^2.7) was calculated as LVM divided by height to the power of 2.7 (27). RWT was calculated as (LVPWT + IVST)/LVID (28).

CV risk factors including abdominal obesity, elevated BP, high FBG, high TG and low HDL-C were defined separately as recommended by the Society of Pediatrics, Chinese Medical Association (29). Abdominal obesity was defined as WC ≥90th percentile values for sex and age (30, 31). Elevated BP was defined as systolic and/or diastolic BP ≥95th percentile values for sex, age and height (32). High FBG (≥5.60 mmol/L), high TG (≥1.47 mmol/L) and low HDL-C (<1.03 mmol/L) were defined according to recommendations from the Society of Pediatrics, Chinese Medical Association (29). The clustering of CV risk factors was defined as the sum of the number of risk factors (abdominal obesity, elevated BP, high FBG, high TG and low HDL-C) with the range spanning from 0 to 4. As only few children had more than 3 risk factors, we combined those with 3–4 risk factors into a same group as ≥3, and the clustering was further categorized as 0, 1, 2 and ≥3. Additionally, we re-defined these CV risk factors according to the same distribution of the present study population, namely the corresponding 90th, 85th and 80th sex- and age-specific percentile values of WC, systolic/diastolic BP, FBG and TG, and 10th, 15th and 20th percentile values of HDL-C.

LVH was defined as LVMI ≥90th percentile values for sex and age of this population, and high RWT was defined as RWT ≥90th percentile values for sex and age of this population (the sex- and age-specific percentile cutoffs of LVMI and RWT are provided in Supplementary Table S1). LVG patterns were defined as normal geometry (normal LVMI and normal RWT), CR (normal LVMI and high RWT), EH (LVH and normal RWT) and CH (LVH and high RWT).

All statistical analyses were performed using SAS software version 9.4 (SAS Institute, Cary, NC, USA), and a two-sided P < 0.05 was considered statistically significant. Continuous variables were presented as means ± standard deviations and categorical variables as numbers (percentages). Students' t test and chi-square test were used to examine differences in characteristics by the LVH status (presence vs. absence) or the RWT status (high vs. normal). Pearson correlation analysis was performed between single CV risk factors and LVMI and RWT. Potential covariates used for adjustment included sex, age, sleep duration (<9 vs. ≥9 h per day), screen time (≤2 vs. >2 h per day), physical activity (<1 vs. ≥1 h per day), and intake of vegetables and fruits (<5 vs. ≥5 servings per day). Adjusted LVMI and RWT levels between each CV risk factor status (abnormal vs. normal) and across the number of CV risk factors (0, 1, 2 and ≥3) were compared using covariance analysis. We also conducted trend analysis for LVMI and RWT levels according to the number of CV risk factors using multivariate linear regression analysis. Binary logistic regression analysis was used to calculate adjusted odds ratios (ORs) and 95% confidence intervals (CIs) of LVH and high RWT, and multi-class logistic regression analysis for LVG remodeling with the normal geometry as the reference category.

A total of 1,406 children (boys: 52.8%) aged 6–11 years were included in this study. Characteristics of children by the LVH status are shown in Table 1. Children with LVH (n = 133) had higher levels of BMI, WC, systolic and diastolic BP, and TG, and a lower level of HDL-C than those without LVH (n = 1,273). Similar characteristics were found between children with high RWT and those with normal RWT (Supplementary Table S2).

Abdominal obesity, high TG and low HDL-C were associated with the risk of LVH irrespective of definitions of CV risk factors (Table 2). Elevated BP was associated with LVH when using percentile values of this population (Table 2). High FBG was associated with LVH when using recommended guideline (Table 2). Similarly, regardless of definitions of CV risk factors, abdominal obesity and high TG were associated with increased risk of high RWT, while associations of elevated BP, high FBG and low HDL-C with high RWT varied slightly based on different definitions of CV risk factors (Table 2). Besides, there were positive correlations between specific CV risk factors (including WC, SBP, DBP, FBG, and TG; Supplementary Table S3) and LVMI and RWT, and HDL-C was related inversely with LVMI (r = −0.12, P < 0.001; Supplementary Table S3). Additionally, covariance analysis showed significant associations of each CV risk factor with LVMI and RWT levels (Supplementary Table S4).

Abdominal obesity (OR = 2.34, 95% CI: 1.53–3.59) and low HDL-C (OR = 2.36, 95% CI: 1.12–4.97) were associated with CR. Elevated BP (90th or 85th) and high TG (85th) defined using the percentiles were also associated with CR (Supplementary Table S5). Abdominal obesity, elevated BP, high TG, and low HDL-C increased the risk of EH with ORs (95% CIs) of 5.26 (3.34–8.29), 1.72 (1.03–2.89), 5.92 (2.92–11.98), and 2.58 (1.21–5.47), respectively; these associations were mostly similar when defining CV risk factors based on the percentiles (Supplementary Table S5). Abdominal obesity, high FBG, high TG, and low HDL-C also increased the risk of CH, with ORs (95% CIs) of 20.96 (8.13–54.03), 3.33 (1.38–8.06), 4.82 (1.86–12.52), and 4.58 (1.92–10.92), respectively; similar associations were found when CV risk factors were defined using the percentiles for definitions (Supplementary Table S5).

There were 787 (55.97%) children without CV risk factor, 413 (29.37%) with one, 156 (11.10%) with two, and 50 (3.56%) with at least three. LVMI and RWT levels were higher along increasing number of CV risk factors (P for linear trend <0.001), after adjustment for sex, age, sleep duration, screen time, physical activity, and intake of vegetables and fruits (Figures 1, 2).

The prevalence of LVH and high RWT increased progressively across the clustering of CV risk factors (Table 3). Compared with children without CV risk factor, those with 1, 2 and ≥3 CV risk factors were associated with an increased risk of LVH, with ORs (95% CIs) of 3.49 (2.19–5.56), 5.53 (3.20–9.55), and 19.19 (9.67–38.08), respectively (Table 3). Similarly, the more number of CV risk factors was associated with a higher risk of high RWT [2.47 (1.63–3.74) for 1, 3.76 (2.25–6.27) for 2, and 5.47 (2.65–11.28) for ≥3, Table 3]. Similar associations were found when CV risk factors were defined based on the percentiles (Table 3).

The clustering of CV risk factors was associated with LVG remodeling including CR, EH and CH (Table 4). The corresponding ORs (95% CIs) of CR were 1.71 (1.06–2.76), 2.83 (1.54–5.18), and 3.82 (1.37–10.62), respectively, for children with 1, 2 and ≥3 CV risk factors; those of EH were 2.42 (1.42–4.11), 4.23 (2.24–7.96), and 16.86 (7.70–36.92), respectively; those of CH were 14.92 (4.41–50.47), 23.15 (6.32–84.83), and 71.19 (17.09–296.56), respectively. Similar associations of the clustering of CV risk factors with CR, EH and CH were found when using the percentiles for definitions.