This retrospective observational study included pediatric participants aged <18 years who were newly diagnosed with severe obesity at Seoul National University Bundang Hospital, Soonchunhyang University Bucheon Hospital, and Inje University Ilsan Paik Hospital between June 2017 and February 2024.

This study was approved by the institutional review boards (IRBs) of all participating hospitals—Seoul National University Bundang Hospital (IRB No. B-2506-978-102), Soonchunhyang University Bucheon Hospital (IRB No. SCHBC 2025-09-011), and Ilsan Paik Hospital (IRB No. ISPAIK 2025-10-032)—and was conducted in accordance with the principles of the Declaration of Helsinki. Informed consent was not required because this study involved retrospective analysis of de-identified data, in accordance with institutional policies.

Participants with at least one year follow-up after the diagnosis of severe obesity were enrolled. Exclusion criteria were (1) prior use of medications that could affect growth or weight, such as growth hormone, gonadotropin-releasing hormone agonists, or corticosteroids; (2) diagnosis of diabetes mellitus, thyroid disease, or congenital or chronic systemic conditions such as leukemia, osteosarcoma, chronic kidney disease, nephrotic syndrome, or Down syndrome; (3) prior use of anti-obesity pharmacotherapy, (4) history of bariatric surgery, or (5) presence of psychiatric disorders. After applying these criteria, 124 participants were included in the final analysis (Supplementary Fig. 1).

Participants received routine lifestyle and dietary interventions during follow-up visits every 2–6 months. Dietary counseling by a dietitian was generally provided at diagnosis, with 1–2 additional sessions offered as required. At each subsequent visit, pediatric endocrinologists provided consistent lifestyle education. The counseling focused on age- and weight-appropriate nutritional guidance and specific lifestyle modifications, including limiting screen time to less than 2 hours per day, engaging in at least 30 minutes of daily physical activity, and ensuring adequate sleep.

Anthropometric measurements including height, weight, and waist circumference (WC), and pubertal stage; were measured at each outpatient visit by the attending pediatric endocrinologists. Height was measured to the nearest 0.1 cm using a stadiometer (Seca 213, Seca 216; Seca, Germany), and weight was measured to the nearest 0.1 kg using an electronic balance (GL-310B, GL-6000S; Gtech, Korea). WC was measured to the nearest 0.1 cm using a tape measure (Rollfix; Hoechstmass Balzer GmbH, Germany) at the midpoint between the lowest margin of the rib and the highest margin of the iliac crest during expiration. Pubertal stage was assessed according to the Tanner staging system based on breast development in girls and genital development in boys [22,23].

Observations were categorized as baseline (initial), final (last follow-up), and one or more intermediate visits conducted approximately every 2–6 months. These intermediate data points were retained for longitudinal analysis of anthropometric trajectories over time. BMI and WC z scores were calculated using reference data derived from Korean children and adolescents [21,24].

Blood pressure (systolic and diastolic, measured as the average of 3 seated readings after at least 5 minutes of rest) and metabolic parameters—including fasting blood glucose, total cholesterol, triglycerides, high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), aspartate aminotransferase (AST), alanine aminotransferase (ALT), and uric acid—were evaluated at the baseline visit for all participants with available data to characterize the initial metabolic status of the cohort. To assess longitudinal changes in metabolic profiles, data were analyzed in a subset of participants who had complete metabolic syndrome-related indicators available from no less than 2 time points at least 6 months apart; specifically, this analysis was restricted to participants aged ≥10 years, for whom the criteria for metabolic syndrome were applicable. For these participants, the earliest and latest available data during the follow-up period were analyzed. Similarly, body composition indices—including body fat percentage and skeletal muscle mass—were measured using bioelectrical impedance analysis (InBody 570; InBody Co., Korea) in participants with ≥2 measurements obtained at least 6 months apart. For these subsets, the first and last available results were included in the analysis. These metabolic and body composition assessments were performed based on data availability in a real-world clinical setting, where initial and repeat laboratory tests and bioelectrical impedance analyses were conducted at the discretion of the attending physician and subject to parental consent. Metabolic syndrome was defined according to the International Diabetes Federation criteria [25].

All analyses were conducted using R ver. 4.5.0 (R Foundation for Statistical Computing, Austria). Baseline characteristics were compared between sexes and groups defined by BMI z score change (Improved vs. not improved) using the independent t test or Mann-Whitney U test for continuous variables, depending on normality, and Fisher exact test for categorical variables. To account for the potential confounding effect of age, analysis of covariance (ANCOVA) with age at diagnosis as a covariate was additionally performed for these comparisons to assess metabolic parameters. The Improved group was defined as participants with any decrease in BMI z score between the initial and last visits, while the Not improved group included those with maintained or increased BMI z scores.

BMI z score trajectories were plotted across all visits, with smoothed curves generated using locally estimated scatterplot smoothing (LOESS) regression. Similar LOESS plots for WC and WtHR z scores were stratified by sex. Transitions in obesity classes between the initial and last visits were visualized using a Sankey diagram, illustrating individual-level changes in obesity severity. Changes in BMI z score were analyzed using the paired Wilcoxon signed-rank test, stratified by sex, baseline BMI z score category, age group, and pubertal stage at diagnosis. Multivariable linear regression and linear mixed-effects model with participant-level random intercepts were used to identify predictors of BMI z score change over time, with sex, age, pubertal stage, baseline BMI z score, follow-up duration, and visit frequency as covariates.

Changes in body composition parameters were analyzed using the paired Wilcoxon signed-rank test among participants with available data. Comparisons between participants with and without body composition data were performed using the Welch t test. Changes in metabolic syndrome-related indicators—including WC z score, blood pressure, fasting blood glucose, glycated hemoglobin (HbA1c), total cholesterol, triglycerides, HDL-C, LDL-C, AST, ALT, and uric acid—were analyzed using paired t test or Wilcoxon signed-rank test among participants with available data, and categorical variables were compared using the McNemar test. Log-transformation was applied to skewed variables (total cholesterol, triglycerides, HDL-C, LDL-C, AST, and ALT) before comparison.

A total of 124 participants (boys, 61.3%) were included in the study. The mean age at diagnosis was 9.69±3.39 years, and girls were significantly younger than boys (P=0.044); however, pubertal stage distribution did not differ significantly by sex. The mean baseline BMI was 29.14±5.05 kg/ m², corresponding to a mean BMI z score of 3.68±0.75. Boys had a higher mean BMI than girls (P=0.001), whereas BMI z scores were comparable between the sexes. Among participants with available waist measurements (n=69), the mean WC was 88.22±13.92 cm, with a mean WC z score of 3.03±0.47. The mean WtHR was 0.63±0.05, with a corresponding WtHR z score of 3.11±0.52. WC and WtHR z scores did not differ significantly by sex. Regarding obesity severity, 80.6% of participants had class 2 obesity and 19.4% had class 3 obesity, with no significant sex difference. The mean follow-up duration was 2.64±1.51 years, with an average visit frequency of 2.33±0.92 visits per year, both comparable between sexes.

The mean systolic and diastolic pressures were 122.43±14.85 and 72.26±10.07 mmHg, respectively, with no significant differences observed between sexes. Laboratory evaluations, available for the majority of the cohort, revealed several sex-specific differences. The mean fasting glucose level was significantly higher in boys than in girls (P<0.001). Boys also exhibited significantly higher levels of liver transaminases, including AST (P=0.001) and ALT (P<0.001), as well as serum uric acid (P=0.014). When adjusted for age at diagnosis using ANCOVA, the sex difference in fasting glucose, AST, and ALT remained significant (all P<0.05). However, the difference in serum uric acid was no longer significant after age adjustment. Other metabolic parameters, including HbA1c and lipid profiles, were comparable between the sexes (Table 1).

When participants were grouped by their BMI z score change, 85 (68.5%) were in the improved group and 39 (31.5%) were in the Not improved group. Participants in the improved group were significantly younger at diagnosis than those in the not improved group (P<0.001). Consistently, the improved group exhibited a significantly less advanced pubertal status at diagnosis (P=0.039). Although the mean baseline BMI and WC were significantly lower in the improved group (P=0.001 and P=0.046, respectively), their corresponding z scores showed no significant differences between the 2 groups.

Regarding metabolic parameters, the improved group exhibited significantly lower mean systolic blood pressure and baseline triglyceride levels (P=0.034 and P=0.011, respectively). However, after adjusting for age at diagnosis, these differences in systolic blood pressure and triglycerides were no longer significant. Other parameters showed no significant differences between the 2 groups. Furthermore, other baseline characteristics, including sex distribution and follow-up frequency, were comparable between the groups (Supplementary Table 1).

BMI z score trajectories demonstrated a distinct age-and sex-dependent pattern (Fig. 1). Among boys, BMI z scores gradually decreased until approximately 11–12 years of age, reaching a nadir before increasing thereafter. In girls, a similar decline was observed until approximately 9–10 years of age, followed by a subsequent rise. When comparing baseline and last-visit values, boys showed an overall trend toward lower BMI z scores, whereas girls did not show a similar pattern. In contrast, WC and WtHR z scores remained largely stable throughout the observation period, showing no clear directional trend (Supplementary Figs. 2 and 3).

Transitions in obesity class between the initial and last visits are visualized using a Sankey diagram (Fig. 2). Most participants remained in the same obesity class during follow-up; however, downward transitions indicating improvement were observed in a subset of participants, primarily from class 3 to class 2 and from class 2 to class 1. Only 2 participants improved to the overweight range at the last visit. In addition, 2 participants developed diabetes mellitus and 3 required pharmacotherapies for weight control after at least 1 year of follow-up.

Overall, BMI z scores significantly decreased from baseline to the last visit (mean ΔZ=–0.27±0.79, P<0.001) (Table 2). Boys showed a significant reduction in BMI z scores (P<0.001), whereas the change in girls was not significant. Younger participants tended to experience greater improvement. The largest decrease occurred in the 3- to 6-year group (P=0.003), followed by the 6- to 9-year group (P<0.001). In contrast, an increase in BMI z score was observed among participants aged 12 to 15 years (ΔBMI z score=+0.35±0.60, P=0.024), whereas minimal or no significant change was evident in the other older groups. When stratified by pubertal stage at diagnosis, BMI z scores significantly decreased among prepubertal participants (P<0.001), with no significant change in early or late puberty groups. When stratified by baseline BMI z score category, the greatest reduction occurred in participants with initial scores of 3.5–4.0 (P<0.001), followed by those with scores of 3.0–3.5 (P=0.022). In contrast, participants with baseline BMI z scores ≤3.0 or >4.0 showed no significant change over time.

In the multivariable linear regression model assessing factors associated with changes in BMI z score between the initial and last visits, younger age at diagnosis (P<0.001) and male sex (P=0.015) were independently associated with a greater reduction in BMI z score. Baseline BMI z score, pubertal stage, follow-up duration, and visit frequency were not significantly associated with BMI z score change (Table 3).

To further evaluate the longitudinal changes, a linear mixed model was employed using all visit-level data. Across all visits, increasing age at each visit was significantly associated with lower BMI z scores over time (P<0.001), indicating a gradual decline during follow-up. Similarly, younger age at diagnosis (P<0.001) and male sex (P=0.015) were associated with lower BMI z scores. In contrast, higher baseline BMI z scores (P<0.001) and longer follow-up duration (P=0.036) were linked to higher BMI z scores; pubertal status and visit frequency were not significant predictors (Table 4).

Among the 40 participants (27 boys) with available body composition data, the mean age at earliest evaluation was 9.63±3.19 years, and the mean follow-up duration was 2.44±1.52 years. BMI z scores significantly decreased between the earliest and latest measurements (P=0.012), whereas no significant changes were observed in body fat percentage or the muscle-to-fat ratio (Supplementary Table 2). Participants with and without body composition data did not differ significantly in baseline characteristics or BMI z score changes.

Among the 32 participants (20 boys) with available metabolic data, the mean age at the initial evaluation was 11.69±2.13 years, and the mean follow-up duration was 1.80±0.95 years. No significant changes were observed in BMI z score, WC z score, blood pressure, or most biochemical indices, including fasting glucose, HbA1c, and lipid profile. In contrast, liver transaminase levels decreased modestly (P=0.013 for AST; P=0.021 for ALT), whereas serum uric acid levels increased significantly during follow-up (P<0.001). Within this subgroup, the number of participants meeting the criteria for metabolic syndrome decreased slightly from 10 (31.2%) to 7 (21.9%), but this change was not significant (Supplementary Table 3).