This is a cross-sectional study used the baseline data of “Optimizing Intervention Effects in Children and Adolescents in Ningbo (OptiChild study)” program (28), which is a clustered randomized controlled trial (Registration No. at clinicaltrials.gov: NCT05482165). This program recruited 1,640 third-grade students between 8 to 10 years old from six primary schools in three districts of Ningbo city in September 2022. The current study used the baseline data and included 1,389 students with the flowchart of the selection of the study population showed in Figure 1.

The program received approval from the Ethics Committee of the First Affiliated Hospital of Ningbo University (Approval No. 2021-R168), and obtained written informed consent from all participating students and their primary guardians.

All the anthropometry measurements were measured by trained staff from local community healthcare centers, adhering strictly to standardized procedures at the participants’ schools. Participant height was measured using a mechanical stadiometer, with individuals barefoot to ensure precision. Body weight and fat mass were assessed via a bioimpedance analysis system (Inbody770, California, USA), with participants dressed in lightweight attire and no footwear. Waist circumference (WC) was gauged using a non-elastic tape, placed at the midpoint between the lower rib and iliac crest, taken at the conclusion of a normal exhalation. Hip circumference was s measured at the maximal protrusion of the gluteal muscles using a non-elastic tape, with participants standing upright in lightweight clothing and feet together. All recorded values for height, weight, WC, and hip circumference were precise to within 0.1 units. Diastolic blood pressure (DBP) and systolic blood pressure (SBP) were measured using an Omron digital sphygmomanometer (Omron HEM-7121, Kyoto, Japan) after the participants had rested for at least 5 min.

Blood samples were collected from the elbow vein by certified nurse using a standardized protocol during morning health checkups at school. All samples were obtained following an overnight fast, stored at 4 °C, and analyzed on the same day. Fasting insulin (FINS) were assessed using a chemiluminescent technique on an automated analyzer (Roche Cobas E602 Immunology Analyzer, Basel, Switzerland). Levels of fasting plasma glucose (FPG), triglycerides (TG), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), and total cholesterol (TC) were determined using enzymatic assays with a separate automated system (Beckman AU5800, California, USA).

Body mass index (BMI) was determined by dividing weight (in kilograms) by the square of height (in meters). BMI Z scores were calculated to account for age and sex variations in pediatric growth patterns (29), allowing for more appropriate interpretation of BMI in children. Body fat percentage (BFP) was calculated by dividing fat mass by total body weight and multiplying by 100. The waist-to-hip ratio (WHR) was computed as waist circumference divided by hip circumference, while the waist-to-height ratio (WHtR) was obtained by dividing waist circumference by height. The homeostasis model assessment for insulin resistance (HOMA-IR) was calculated as [FINS (μU/L) * FPG (mg/dL)]/405 (30). Various approaches have been utilized to compute continuous CMR score, most of which include metrics related to adiposity, lipid profiles, metabolic markers, and blood pressure (31). In this study, CMR score were derived by adding the age- and sex-specific Z scores for WC, SBP, TG, the TC/HDL-C ratio, and HOMA-IR, which aligns with established pediatric CMR scoring methods (31, 32). The Z scores for each parameter were calculated using the formula (value − mean)/SD, separately for boys and girls within each one-year age group. Higher CMR score corresponded to an elevated cardiometabolic risks.

All children were asked to recall their food group consumption over the previous day and night, and completed the DQQ with the assistance of a trained project investigator, who used simplified, child-friendly language to ensure understanding. Food intake was coded into 29 food groups following the DQQ tool fitted in Chinese population, which employs binary (yes/no) questions about specific food groups consumed in the past 24 h (33), and the Chinese version DQQ tool been validated for Chinese children aged 7–18 year (19). The GDR-Healthy score and GDR-Limit score have a range from 0 to 9, reflects adherence to global dietary recommendations on healthy components of the diet and components of the diet to limit or avoid, respectively. A higher GDR-Healthy score indicates inclusion of more health-promoting foods in the diet, and the GDR-Limit score is opposite. The overall GDR score was calculated as: GDR scores = GDR-Healthy - GDR-Limit + 9, ensuring a positive scale where higher values indicate better diet quality, it ranged from 0 to 18. The higher the GDR score, the more recommendations are likely to be met. All participants were divided into four quantiles based on their GDR score, the distribution across the quantiles was as follows: 219 in Q1, 397 in Q2, 425 in Q3, and 348 in Q4.

Assessment of physical activity of moderate- or vigorous-intensity physical activity (MVPA) time in school days and weekends was asked by a self-reported Physical Activity Questionnaire, which has been validated in Chinese children (34), with the details of the questionnaire described before elsewhere (35). The daily MVPA time were categorized in to ≥ 60 or < 60 min/day last week according to the WHO recommendations (36).

A qualified physical education instructor administered the CRF evaluation using the 20mSRT, a validated and reliable method for assessing CRF in children and youth (37). During the 20mSRT, participants ran back and forth between two lines set 20 meters apart, synchronizing their pace with audio cues. The test is structured into multiple stages (or levels), each lasting about 1 min and consisting of several 20-meter laps (referred to as shuttles). The initial running speed was 8.5 km/h, increasing by 0.5 km/h each minute (with 1 min corresponding to one stage) (38). The test concluded when a participant could not reach the line in time with the audio signals for two consecutive attempts or when they halted due to exhaustion. The total number of completed laps was then used as the primary indicator to estimate each child’s CRF level.

The demographic and physical characteristics of the participants were summarized using descriptive statistics. Normality of continuous variables was assessed using Shapiro–Wilk tests and visual inspection of Q-Q plots. Group differences across GDR score quantiles were examined using one-way analysis of variance (ANOVA) for normally distributed variables (reported as mean ± SD) or the Kruskal-Wallis test for skewed variables (reported as median [IQR]). Categorical variables were expressed as counts (n) with percentages (%), and their distribution across GDR score quantile groups was assessed using the chi-square test. For continuous variables with significant overall differences across GDR score quantiles, pairwise comparisons were performed using Tukey’s post-hoc test for normally distributed variables and Dunn’s test with Bonferroni correction for non-normally distributed variables. For categorical variables with significant overall differences, pairwise comparisons were performed using Bonferroni-corrected chi-square tests.

The relationships between GDR score and CMR indicators and the interaction terms (GDR score ×sex) were explored by a generalized linear mixed-effects model (GLMM), and the FINS, HOMA-IR and TG levels were Ln-transformed due to skewed distributions. This model incorporated school-level random intercepts to account for the clustering of participants within schools, as data collection was conducted at the school level. Two models were constructed: Model 1 was unadjusted, while Model 2 included additional adjustments for age, sex, and MVPA based on Model 1. Stratified analyses were conducted to evaluate the association of GDR score and CMR indicators among children with different gender.

To assess the mediating role of CRF in the association between the GDR score and CMR indicators, the mediation package in R was utilized. Bootstrapping with 5,000 resamples was performed to estimate the indirect effects. Statistical significance was determined using a two-tailed test with a p-value threshold of <0.05. All statistical analyses were conducted using R version 4.3.0 (R Core Team).

A total of 1,389 participants were included in the analysis, we compared the characteristics of the children included and 251 children excluded (Supplementary Table S1), no differences were detected between the students included and excluded. The histogram of the GDR score showed a normal distribution, with values ranging from 3 to 18 (Supplementary Figure S1).

The overall sample consisted of 52.7% boys and 47.3% girls, with a mean age of 8.48 ± 0.29 years. The clinical and demographic characteristics stratified by GDR score quartiles are presented in Table 1. Significant differences were observed in sex distribution (p < 0.001), WHR (p = 0.029), FPG (p = 0.022), and CRF (laps, p = 0.020). Participants in higher GDR quantiles (Q4) exhibited greater proportions of girls, higher CRF, and lower FPG. For example, FPG levels were significantly higher in Q1 compared with Q4, whereas Q2 and Q3 were intermediate and did not differ significantly from either group. Time spent in MVPA also differed significantly across quantiles (p < 0.001), with Q4 showing the highest proportion of participants engaging in >1 h/day of MVPA (60.6%).

Dietary characteristics (Table 2) revealed marked variations in food consumption patterns. The consumption of healthy food items increased significantly with higher GDR scores. For example, the proportion of children consuming whole grains rose from 20.1% in Q1 to 64.9% in Q4, with significant pairwise differences across most quantiles (p < 0.05). Similar increasing trends were observed for pulses, vitamin A-rich vegetables, dark green leafy vegetables, other vegetables, citrus fruits, and other fruits (all p < 0.001). Conversely, the consumption of unhealthy food items declined markedly with higher GDR scores. For instance, the prevalence of soft drink intake decreased from 41.6% in Q1 to just 3.7% in Q4, and significant differences were also observed for baked sweets, processed meats, deep-fried foods, fast food, and packaged salty snacks (all p < 0.001). Overall, the GDR score itself increased stepwise across quantiles (7.1 ± 1.2 in Q1 vs. 13.9 ± 1.1 in Q4, p < 0.001).

The associations between GDR score and cardiometabolic indicators were assessed using two models (as presented in Table 3): an unadjusted model (Model 1) and a model adjusted for age, sex and MVPA (Model 2). In Model 1, higher GDR scores were inversely associated with BMI Z-score (β = −0.029, p = 0.044), WC (β = −0.161, p = 0.035), WHR (β = −0.002, p = 0.003), WHtR (β = −0.001, p = 0.024), FPG (β = −0.007, p = 0.047), FINS (β = −0.011, p = 0.048) and HOMA-IR (β = −0.013, p = 0.028), while positively linked to CRF (β = 0.405, p = 0.006). After adjustment for covariates in Model 2, GDR score remained significantly associated with FINS (β = −0.013, p = 0.023), HOMA-IR (β = −0.014, p = 0.019), and CRF (β = 0.441, p = 0.003). Additionally, GDR score was negatively associated with CMR score (β = −0.074, p = 0.030) in Model 2.

The associations between GDR score and cardiometabolic indicators were further stratified by sex (Table 4). The interaction terms between sex and GDR score were significant for BMI (p = 0.020), BMI Z-score (p = 0.037), and BFP (p = 0.025), indicating that the associations between GDR score and these indicators differed by sex. Among boys, GDR score was not significantly associated with most cardiometabolic indicators, except for a positive association with 20mSRT (β = 0.451, p = 0.042). In contrast, among girls, GDR score was significantly associated with BMI (β = −0.108, p = 0.005), BMI Z-score (β = −0.049, p = 0.008), BFP (β = −0.322, p = 0.008), WC (β = −0.225, p = 0.023), WHtR (β = −0.002, p = 0.014), FINS (β = −0.021, p = 0.012), HOMA-IR (β = −0.022, p = 0.010), and CMR score (β = −0.126, p = 0.014).

We explored the mediating effect of CRF between GDR score and cardiometabolic indictors (Figure 2). CRF mediated 26.1% (p = 0.036, Figure 2A) and 25.3% (p = 0.028, Figure 2B), of the association between GDR score and FINS and HOMA-IR, and we found the largest mediating proportion of the association between GDR score and CMR score (mediation effect: 32.9%, p = 0.034, Figure 2C).