This non-randomized, self-controlled longitudinal study was designed to track within-participant changes following a standardized clinical intervention. Randomization was not employed as all eligible participants received the multidisciplinary intervention per hospital protocol. This single-center longitudinal study was conducted at the Pediatric Outpatient Department of Zhejiang Jinhua Municipal People's Hospital from March 2022 to December 2023. The study protocol was approved by the Ethics Committee of Zhejiang Jinhua Municipal People's Hospital (Approval No. 2023068). Given that the study involved only routine clinical interventions and noninvasive ultrasound monitoring, the ethics committee waived informed consent requirements. However, all participants and their guardians signed anonymized data usage authorization forms. The study adhered to the principles of the Declaration of Helsinki, with clinical data collected via a standardized electronic medical record system and patient privacy rigorously protected.

This single-center longitudinal study was conducted at … patient privacy rigorously protected. Given the ethical constraints of withholding clinical intervention from children with obesity and the longitudinal nature of vascular maturation in children, a non-intervention control group was not included. To address potential confounding by natural growth, we compared observed changes with established pediatric reference values for vascular parameters and utilized repeated measurements to mitigate regression-to-mean effects.

To address missing data, multiple imputation (5 imputed datasets) was applied for continuous variables under the missing-at-random assumption, with baseline characteristics and adherence metrics as predictors. All analyses followed intention-to-treat principles.

Inclusion criteria:

Age 7–14 years, meeting the Working Group on Obesity in China criteria (BMI ≥95th percentile) (10);

All enrolled children with obesity received a 12-month multidisciplinary intervention.

Dietary intervention: Registered dietitians developed personalized low-calorie diet plans on the basis of individual basal metabolic rates and dietary habits, reducing total daily caloric intake by 20% while optimizing macronutrient composition (increased dietary fiber, high-quality protein, and unsaturated fats; restricted added sugars and saturated fats). Compliance was monitored weekly via food diaries and nutritional analysis software, with dietary adherence quantified as: (1) proportion of days with completed food diaries (target ≥5 days/week), and (2) percentage of weeks meeting caloric targets (±10% of prescribed intake).

Exercise intervention: Rehabilitation specialists designed moderate-intensity aerobic exercise programs (e.g., brisk walking, swimming, cycling) that were conducted 5 times weekly (60 min/session), targeting heart rates ≥140 bpm. Exercise intensity was progressively adjusted according to baseline fitness levels, with supervision by professional coaches to ensure safety and adherence. Exercise adherence was quantified as session attendance rate (sessions attended/total prescribed sessions), with target adherence set at ≥80%.

Behavioral therapy: Family-involved cognitive behavioral interventions included monthly group sessions covering self-monitoring skills (e.g., diet/exercise logging), goal setting, positive reinforcement strategies, and stress management. Interactive methods (role-playing, simulations) were supplemented with homework (e.g., “Healthy Challenge Week”) to reinforce behavioral changes. Parents participated in all the sessions and adjusted their home environments (e.g., reducing high-calorie snack availability). Additionally, the research team provided personalized guidance via monthly phone calls and instant messaging (WeChat) to address adherence issues (Supplementary Table S1).

Data collection was conducted at three time points: baseline (T0, preintervention), 6 months postintervention (T1), and 12 months postintervention (T2). All measurements were performed by uniformly trained medical staff following standardized protocols to minimize interoperator variability.

Ultrasound parameters: All ultrasound examinations were performed by a senior radiologist with over 10 years of experience in pediatric vascular ultrasonography. The examiner was certified by the Chinese Medical Doctor Association and the Chinese Society of Ultrasound Medicine, trained per standardized protocols (12). Measurements followed technical guidelines: (1) Fasting >3 h; (2) Supine position with right arm maximally abducted; (3) Probe perpendicular to liver capsule with light pressure; (4) Sampling depth 3–6 cm below Glisson's capsule, avoiding vessels >3 mm. A Siemens S2000 color Doppler ultrasound system (linear probe frequency: 10 MHz; convex probe frequency: 4.5 MHz) was used to measure the following parameters: carotid IMT: participants were placed in the supine position, and the IMT was measured at 12 sites (bilateral common carotid artery, bifurcation, and proximal/distal internal carotid artery), with the mean value calculated. FMD: The brachial artery was selected as the target vessel. The baseline diameter (Bd) and postreactive hyperemia diameter (Fd) were measured, and FMD was calculated as (Fd−Bd)/Bd × 100%. Arterial elasticity parameters: Systolic pressure (Ps), diastolic pressure (Pd), and carotid artery diameter changes (Ds, Dd) were synchronously recorded to calculate the elasticity coefficient (Ep), stiffness β, distensibility coefficient (DC), and arterial compliance (AC). Liver stiffness measurement: Shear wave velocity (SWV) was measured using acoustic radiation force impulse (ARFI) imaging. Five measurements were taken in the right anterior liver lobe (while avoiding vascular structures), and the mean value was calculated and reported in m/s. Young's modulus (E) was calculated from SWV using the formula represents SWV (12), with results expressed in kilopascals (kPa). The representative image of liver elasticity measured by ARFI technology is shown in Figure 1. Carotid artery ultrasound measurements for vascular elasticity assessment were performed as shown in Supplementary Figure S1. The methodology for obtaining carotid artery parameters, including intima-media thickness and blood flow velocity measurements, is illustrated in Supplementary Figure S2.

To assess measurement reliability, intra-observer and inter-observer reproducibility were evaluated for critical ultrasound parameters in a subset of 30 randomly selected participants who underwent duplicate scans by the same operator and a second blinded operator within 48 h.

Clinical parameters: Weight and height: measured via a calibrated electronic stadiometer (model: SECA 284), with BMI calculated as weight (kg)/height (m²); blood pressure: measured three times via a standard cuff sphygmomanometer (Omron HEM-7121), with the mean value recorded; and lipid profile: fasting venous blood samples were analyzed for total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), and high-density lipoprotein cholesterol (HDL-C) via the hospital's central laboratory.

Data management and quality control: All the data were entered in real time via an electronic medical record (EMR) system, with dual independent verification to ensure accuracy. Ultrasound images and raw data were archived for review. Lipid measurements were validated through internal laboratory quality controls (coefficient of variation <5%).

Blinding procedures: Ultrasound operators were blinded to (1) participant identities (assigned anonymized IDs), (2) intervention time points (scans labeled only with study codes), and (3) clinical parameters (blood pressure/lipid results). Image analyses were performed on de-identified datasets using standardized protocols.

Data analysis was performed via SPSS 26.0. Missing data were handled via multiple imputation using chained equations (MICE) with 5 iterations, incorporating baseline covariates and observed follow-up data to preserve statistical power and reduce bias. Continuous variables are expressed as the mean ± standard deviation (SD). The primary analysis employed paired t tests to compare changes in ultrasound parameters and clinical indicators pre- and postintervention, with a significance threshold of P < 0.05. The secondary analysis utilized multiple linear regression models, where ΔFMD and ΔSWV served as dependent variables and where ΔBMI and Δsystolic/diastolic blood pressure were independent variables. Adherence-outcome associations were assessed via linear regression models, with dietary/exercise adherence rates as covariates in sensitivity analyses of ΔFMD and ΔSWV. Adjustments were made for age, sex, and other potential confounders to evaluate independent associations (β coefficients and P values). Multicollinearity among independent variables (ΔBMI, ΔSBP, ΔDBP, ΔLDL-C) was assessed using variance inflation factors (VIF). The sensitivity analysis excluded 15 participants with postintervention weight rebound (BMI increase ≥5%) to verify the robustness of the results. All the statistical tests were two-tailed. Data normality was assessed via the Shapiro‒Wilk test, with nonnormal data analyzed via nonparametric tests (Mann‒Whitney U).

During the 12-month intervention period, 12 participants (10.0%) were lost to follow-up due to relocation (n = 7), withdrawal of consent (n = 3), or inability to comply with scheduled visits (n = 2). The study enrolled 120 children with obesity, including 68 males (56.7%) and 52 females (43.3%). Baseline clinical and ultrasound parameters revealed a mean BMI of 28.4 ± 3.1 kg/m² and systolic and diastolic blood pressures of 125 ± 8 mmHg and 78 ± 6 mmHg, respectively, with total cholesterol (TC) and LDL-C levels exceeding normal pediatric reference ranges. Ultrasound measurements revealed an IMT of 0.58 ± 0.05 mm, FMD of 7.1 ± 2.3%, and SWV of 1.36 ± 0.23 m/s. Gender comparisons revealed no statistically significant differences in BMI, blood pressure, or ultrasound parameters between males and females (P > 0.05), confirming baseline homogeneity (Table 1).

Liver SWV and derived Young's modulus (E) significantly decreased after intervention. The baseline SWV (1.36 ± 0.23 m/s) and E (5.54 ± 1.84 kPa) indicated elevated liver stiffness. After 6 and 12 months of intervention, SWV decreased to 1.12 ± 0.18 m/s (E: 3.76 ± 1.21 kPa) and 0.98 ± 0.12 m/s (E: 2.88 ± 0.71 kPa), respectively, with significant time main effects [SWV: F₍₂₁₁₈₎ = 48.15, P < 0.001; E: F₍₂₁₁₈₎ = 46.32, P < 0.001], confirming time-dependent improvements in liver stiffness (Table 4).

Correlations between changes in vascular function (ΔFMD, ΔEp, Δβ, ΔCD, and ΔAC) and liver stiffness (ΔSWV) were analyzed. The ΔFMD was significantly negatively correlated with the ΔSWV (r = −0.46, P < 0.001), suggesting that vascular endothelial improvement is associated with a reduction in liver stiffness. Additionally, ΔEp (r = 0.38, P < 0.001) and Δβ (r = 0.32, P = 0.001) were positively correlated with ΔSWV. To account for potential confounding by weight change, partial correlation analyses adjusting for ΔBMI were performed. After adjustment, ΔFMD remained negatively correlated with ΔSWV (r_partial = −0.35, P = 0.001), while ΔEp (r_partial = 0.28, P = 0.007) and Δβ (r_partial = 0.24, P = 0.02) retained positive correlations with ΔSWV. The associations for ΔCD (r_partial = −0.15, P = 0.15) and ΔAC (r_partial = −0.14, P = 0.18) became nonsignificant. These findings suggest that the deterioration of arterial elasticity parallels the progression of liver fibrosis (Table 5).

Multivariate models adjusted for multicollinearity (VIF <5 for all predictors) were used to assess the independent effects of weight change (ΔBMI) and clinical indicators on vascular and hepatic improvements (Table 6). ΔBMI independently predicted ΔFMD (β = 0.32, P = 0.003) and ΔSWV (β = −0.41, P < 0.001), indicating that weight loss directly enhances vascular repair and liver softening. ΔSBP independently influenced ΔEp (β = 0.18, P = 0.04) and Δβ (β = 0.15, P = 0.03), suggesting that blood pressure control partially mitigated arterial stiffening (Table 7).

Dietary adherence averaged 85.4% (SD, 8.2%) for diary completion and 78.2% (SD, 9.5%) for caloric targets. Exercise attendance was 89.6% (SD, 7.3%). Regression models demonstrated that higher dietary adherence predicted greater ΔFMD (β = 0.24, P = 0.01) and ΔSWV reduction (β = −0.19, P = 0.03), while exercise adherence independently associated with ΔCD improvement (β = 0.18, P = 0.04) (Table 8).

After 15 participants who experienced weight rebound (BMI increase ≥ 5%) after intervention were excluded, the improvement in the ΔSWV further increased (−0.45 ± 0.07 m/s vs. −0.38 ± 0.09 m/s, t = 2.98, P = 0.004). The significant increase in the ΔFMD (2.9 ± 0.8% vs. 2.5 ± 0.9%, t = 2.11, P = 0.04) and ΔCD (4.9 ± 0.4 kPa vs. 4.5 ± 0.5 kPa, t = 2.34, P = 0.02) indicated that the main results were robust to weight rebound.

The intra-observer and inter-observer reliability for key ultrasound parameters were excellent. For carotid IMT, the intra-observer ICC was 0.92 (95% CI: 0.87–0.95) with a CV of 4.8%, and the inter-observer ICC was 0.88 (95% CI: 0.81–0.93) with a CV of 6.3%. For liver SWV, the intra-observer ICC was 0.94 (95% CI: 0.90–0.97) with a CV of 5.1%, and the inter-observer ICC was 0.90 (95% CI: 0.84–0.94) with a CV of 7.2%.