Effect of an mHealth-assisted multifaceted lifestyle intervention on body-mass index, hepatic fat content and stiffness in children with overweight or obesity: a cluster randomised controlled trial

Summary

Background

Lifestyle modification represents the cornerstone of the prevention and early management of obesity, hepatic fat content and stiffness in children. We aimed to evaluate the effectiveness of an mHealth-supported multifaceted lifestyle intervention on body-mass index (BMI), hepatic fat content and stiffness in children with overweight or obesity, and to estimate its potential long-term macroeconomic benefits.

Methods

In this cluster randomised controlled trial, six primary schools in Ningbo, China, were randomly assigned (1:1) to an intervention or control group. Children aged 8–10 years with overweight or obesity were enrolled. The intervention integrated school-based health education and polices, physical activity promotion, dietary guidance delivered by clinical nutritionists, and supportive mHealth components. Primary outcomes were changes in BMI, controlled attenuation parameter (CAP), and liver stiffness measurement (LSM). Intention-to-treat analysis was performed using generalised linear mixed models accounting for school-level clustering and adjusting for baseline outcome values, age, and sex. A post-hoc macroeconomic simulation projected long-term economic effects. This trial is registered with ClinicalTrials.gov, NCT05482191.

Findings

From 6 to 30 September, 2022, 331 children (mean age 8.5 years [SD 0.3]; 36.3% girls) were enrolled. Compared with controls, the intervention group showed significant reductions in BMI (mean difference −0.38 kg/m², 95% CI –0.59 to −0.16; p < 0.0001), CAP (−15.73 dB/m, −22.39 to −9.07; p < 0.0001), and LSM (−0.69 kPa, −0.89 to −0.48; p < 0.0001). Changes in BMI mediated 11.9% and BMI Z score 9.1% of the intervention's effect on CAP (both p < 0.05). No adverse events were reported. Macroeconomic modelling suggested that national implementation could yield cumulative economic gains of approximately RMB 694.96 billion (US$ 97.47 billion) during 2026–2050.

Interpretation

The mHealth-assisted lifestyle intervention significantly improved BMI, hepatic fat content and stiffness in children. The accompanying macroeconomic benefits support its integration into national public health and economic strategies.

Funding

Major Science and Technology Projects for Health of Zhejiang Province; Cyrus Tang Foundation 2022.

Research in context.

Evidence before this study

We searched PubMed for randomised controlled trials and systematic reviews published before Sep 1, 2022, using the search terms “childhood obesity”, “pediatric overweight”, “lifestyle intervention”, “mHealth”, “non-alcoholic fatty liver disease”, “metabolic dysfunction-associated steatotic liver disease (MASLD)”, “hepatic fat content”, “liver stiffness”, and “macroeconomics”. Previous studies indicated that multimodal lifestyle interventions were more effective than single-component interventions, and mHealth-based multimodal interventions also led to significant reductions in body-mass index (BMI). Evidence for improvements in pediatric metabolic dysfunction-associated steatotic liver disease (MASLD) is inconsistent: a recent systematic review examined lifestyle interventions for pediatric MASLD and obesity reported reductions in aspartate aminotransferase levels, but found no significant effects on BMI. Certain exercise studies have demonstrated improvements in hepatic steatosis measured via Magnetic Resonance Imaging. However, whether lifestyle intervention, based on school environments could alleviate the hepatic fat content and stiffness measured by portable vibration-controlled transient elastography (VCTE) is not clear. Meanwhile, the broader macroeconomic implications of pediatric MASLD interventions remain unexplored.

Added value of this study

This study developed an mHealth-assisted multifaceted intervention model, integrating Schools, Clinics, famIlies, and studENTs (SCIENT) to address overweight and obesity in children, as well as related hepatic profile. Unlike previous approaches, the SCIENT study provided structured, professional dietary guidance from clinical nutritionists and used VCTE to assess controlled attenuation parameter (CAP, dB/m) for hepatic fat content and liver stiffness measurement (LSM, kPa) for stiffness. The intervention significantly reduced BMI and improved both CAP and LSM, with mediation analyses indicating that both BMI and BMI Z score changes partially mediated the improvement on CAP. Furthermore, macroeconomic simulation supported the potential societal benefits of nationwide implementation, highlighting both health and economic gains.

Implications of all the available evidence

The SCIENT model provides an effective and practical approach for reducing BMI, improving liver profile in children. Its multi-stakeholder collaborative framework, engaging schools, clinics, families, students, and digital technology, provides a scalable and replicable approach, particularly valuable in large population settings. With the estimated gross domestic product returns of approximately RMB 694.96 billion (US$ 97.47 billion), the intervention represents not only a clinically meaningful advance but also a strategic public health investment. Future studies should focus on the long-term sustainability and broader implementation of this integrated intervention model.

Introduction

Early-onset obesity in childhood is a strong predictor of obesity in adulthood and poses lifelong health risks.¹ Recent data from China show that the prevalence of overweight and obesity in children rose from 5.0% in 1995 to 24.2% in 2019, representing a 4.8-times increase.² This trend is alarming, as pediatric obesity is closely linked to an elevated risk of cardiometabolic diseases, including hepatic steatosis and fibrosis and eventually progressing to metabolic dysfunction–associated steatotic liver disease (MASLD), formerly known as nonalcoholic fatty liver disease (NAFLD).³ MASLD has become the most prevalent chronic liver disease, affecting 6.3% of children and 40.4% of those who are overweight or obese in China.⁴ Early occurrence of hepatic steatosis confers a greater risk of severe health complications in later life, one study reported that each unit increase in body-mass index (BMI) Z score among children aged 7–13 years was associated with a 16% higher risk of adult liver cirrhosis.⁵ Childhood obesity and hepatic steatosis substantially impair quality of life, and have become urgent global public health challenges.

To date, no pharmacological therapy for pediatric MASLD has been approved, rendering lifestyle intervention the cornerstone of management, which is also recommended for childhood obesity. However, awareness of pediatric MASLD remains low among both the public and healthcare providers.⁶ Highlighting that weight loss may lead to partial or complete improvement of obesity-related conditions, such as hepatic steatosis and stiffness, strengthens the motivation to maintain a healthy weight. Meanwhile, co-management of these conditions could enhance the efficiency of healthcare resource utilisation. Hepatic steatosis is closely associated with an increased risk of obesity, type 2 diabetes, and cardiovascular diseases in early adulthood, and may eventually lead to long-term consequences such as premature death, increased medical expenditures, and reduced economic productivity.⁷ Despite the growing disease burden, evidence regarding the macroeconomic impact of interventions targeting MASLD prevalence remains limited.

The previous DECIDE-Children study, a cluster randomised controlled trial (RCT), demonstrated that multifaceted lifestyle interventions could significantly attenuate BMI increase.⁸ Several studies in children have demonstrated that exercise can improve hepatic steatosis.⁹ However, as these studies were primarily conducted in clinical settings, their findings may have limited generalisability in school-based intervention. It remains uncertain whether such interventions can also ameliorate the hepatic fat content or stiffness. To enhance intervention efficacy for both obesity and its hepatic profile, the current study integrated clinical professionals, including nutritionists and ultrasound technicians, to develop a collaborative intervention model engaging Schools, Clinics, famIlies, and studENTs (SCIENT). Mobile health (mHealth) technology via a smartphone application was also utilised to strengthen communication among stakeholders. We hypothesised that this multifaceted lifestyle intervention could improve obesity and its hepatic profile in school-aged children. Moreover, this study employed a health-augmented macroeconomic model to estimate the potential gross domestic product (GDP) gains, increase in per capita income, and the proportion of national economic output attributable to reduced MASLD prevalence between 2026 and 2050 in China.

Methods

Study design and participants

The SCIENT study was a cluster RCT aimed at children with overweight or obesity, as defined by age- and sex-specific BMI percentiles based on Chinese reference standards.¹⁰ The trial was conducted across six primary schools in Ningbo, China. Recruitment and baseline assessments commenced in September 2022. The intervention was implemented over one academic year (9 months) and concluded in June 2023. Follow-up assessments were conducted at 3 months (December 2022) and 9 months (June 2023). A detailed description of the SCIENT study design and methodology is available in the published trial protocol.¹¹ This trial is registered at ClinicalTrials.gov, NCT05482191. This study followed the Consolidated Standards of Reporting Trials (CONSORT) 2025 reporting guidelines.

The school recruitment process was conducted in two steps. First, we collaborated with local education authorities to obtain a list of recommended schools along with their basic information. Second, we contacted and visited these schools to assess eligibility, confirm their willingness to participate, and finalise the list of participating schools. The inclusion criteria for schools were as follows: (a) the school principal expressed interest in the project and committed to assigning relevant staff (e.g., school doctors, health teachers, or physical education teachers) to ensure its implementation; (b) the school was not a boarding school or a specialty school focusing on gifted children or specific ethnic minorities; and (c) the school was not currently conducting and had no plans to initiate a similar intervention programme. These criteria ensured the selection of local public primary schools that reflect the general educational and socioeconomic landscape of the region, where educational resources and physical activity standards are highly uniform.

Children aged 8–10 years with overweight or obesity were recruited from third grade based on baseline assessment. The exclusion criteria were as follow: (a) history of heart disease, hypertension, diabetes, asthma, viral hepatitis, or nephritis; (b) obesity attributable to endocrine disorders or medications use; (c) abnormal physical development, such as dwarfism or gigantism; (d) any condition limiting the ability to participate in physical activities; and (e) weight loss induced by vomiting or medications in the past 3 months.

Ethics

Ethic approval was obtained from the Ethics Committee of the First Affiliated Hospital of Ningbo University (approval No. 2021-R168). Written informed consent was obtained from all participants and their guardians before any intervention activities commenced.

Randomisation and masking

Following the completion of baseline assessments, the six participating schools were randomly allocated in a 1:1 ratio to either the intervention or control group using a computer-generated randomisation sequence. The school served as the unit for both randomisation and clustering in our analyses. The randomisation sequence was created by an independent project staff member who was not involved in recruitment, intervention, or outcome assessment. Group assignments were kept concealed until baseline data were collected. Due to the cluster-RCT design, it was not feasible to blind participants, school staff, or project implementers to group assignments. However, to minimise assessment bias, the health professionals responsible for collecting physical measurements and performing vibration-controlled transient elastography (VCTE) were blinded to the group allocations throughout the study.

Procedures

The three control schools continued their standard health education practices as before. The intervention engaged four key stakeholders (schools, clinics, families, and students) and comprised four components; full methodological details for each component can be found in Appendix A (pp 2–6). Building upon our previous multifaceted intervention approach in DECIDE-Children study,⁸ the current trial further intensified the dietary guidance component under the supervision of clinical nutritionists, a detailed comparison is available in Appendix A (p 6). In the first component (implementation of health education courses and distribution of educational materials), trained teachers delivered ten health education sessions to participants based on the “2–2–1” core messages (two “not”: not overeating and not drinking sugar-sweetened beverages; two “less”: less high-energy food and less sedentary time; one “more”: more physical activity⁸). Caregivers attended two educational lectures per semester. Health education material including books, posters, and online sources were distributed. In the second component (dietary guidance delivered by clinical nutritionists), nutritionists provided individualised dietary feedback through the mHealth platform daily during the initial 21 days (1st semester) and 7 days (2nd semester), followed by weekly counseling. School canteen staff received training to align cafeteria options with 2-2-1 principles at the beginning of each semester. In the third (promotion of physical activity), in addition to regular school physical activity, students participated in three 30-min supervised moderate-to-vigorous physical activity (MVPA) sessions per week, which focused on aerobic activities complemented by anaerobic exercises. Parents were encouraged to support children's physical activities outside school, with recommendations reinforced via school communications and mHealth platform. In the fourth (development and enforcement of school-based health policies), school polices were enforced to safeguard time for health education and physical activity, restrict the sale and consumption of unhealthy snacks and beverages on campus, and improve exercise facilities.

A WeChat official account, named “Population Health Service Platform”, was developed for this study to support delivery of health education via professional articles and animated videos, tracking and feedback of children's weight and height, monitoring of dietary behaviour and feedback, and personalised dietary management by nutritionists. All data were encrypted and stored on a secure server restricted to authorised researchers. All procedures complied with institutional ethics requirements and national regulations on data protection.

Due to the COVID-19 pandemic, the winter vacation was advanced to late-December 2022, approximately 3 weeks earlier than scheduled. Although formal health education sessions were completed before closure, the early break resulted in the suspension of school-organised physical activity. To mitigate this, clinical nutritionists provided continuous online dietary guidance via the mHealth platform, and families were granted access to child-friendly aerobic exercise videos to facilitate indoor activity. No similar disruptions occurred during the remainder of the study period.

All outcome measurements were conducted by trained health professionals following standardised procedures. Height, weight, waist circumference, and body composition were measured by qualified health staff. Overnight fasting venous blood samples were collected by trained nurses in the morning of the physical examination. These samples were analyzed for metabolic parameters, including fasting insulin, fasting blood glucose, alanine aminotransferase (ALT), and serum lipids at the clinical laboratory of the First Affiliated Hospital of Ningbo University. Cardiorespiratory fitness (CRF) was assessed via the 20-m shuttle run test (20mSRT). Child-level data including lifestyle knowledge, weight-loss willingness, physical activity, and eating habits were collected via self-report questionnaires.⁸ Dietary quality, and psychological well-being including depression, social anxiety, sleep quality were evaluated using validated scales. Parents provided ethnicity and sociodemographic data. Detailed descriptions of the instruments and assessment methods were provided in Appendix A (pp 7–9).

The controlled attenuation parameter (CAP) and liver stiffness measurement (LSM) were assessed by a trained professional (J Gu) who completed standardised training using FibroScan Handy device (Echosens, Paris, France), with participants fasted overnight. During the examination, participants were instructed to lie in a dorsal decubitus position with their right arms extended and elevated behind the head to widen the intercostal spaces. The probe of device was positioned perpendicular to the skin surface over the region of the right 7th–9th intercostal spaces, avoiding large vessels and rib shadows. The median and interquartile range (IQR) were calculated from ten valid measurements and results were excluded if the IQR–median ratio exceeded 30%.¹²

A mixed-methods approach was employed to evaluate the implementation of intervention, with a particular emphasis on the fidelity, dose delivered, and dose received. Data were collected through multiple sources: implementation logs maintained by teachers via the DingTalk app (a widely used enterprise communication platform in China that supports messaging, scheduling, and document management); record sheets completed by clinical nutritionists to document dietary guidance sessions; attendance logs for weekly physical activity sessions, recorded by trained physical education teachers; and semi-structured interviews conducted with participants, school staff, parents, and school administrators.

Outcomes

The primary outcomes were changes in BMI, CAP, and LSM from baseline to the end of trial. Secondary outcomes included the prevalence of obesity (defined according to both Chinese¹⁰ and World Health Organisation references¹³), MASLD, and liver fibrosis (LSM ≥7.0 kPa¹²). In accordance with the 2023 multisociety Delphi consensus,¹⁴ MASLD was defined as the presence of liver steatosis (CAP ≥248 dB/m¹⁵), as all participants presented with overweight or obesity, satisfying the cardiometabolic risk criterion. Additional secondary outcomes included obesity-related parameters (BMI Z score, body fat percentage, waist circumference), metabolic biomarkers, CRF, and health-related knowledge and behaviours.

Statistics

Sample sise estimation was based on expected changes in BMI, CAP, and LSM. Based on previous studies, the anticipated effect sises were 0.87 kg/m² (standard deviation [SD] 0.9 for BMI),¹⁶ 72 dB/m (SD 75) for CAP,¹⁷ and 0.30 kPa (SD 0.31) for LSM.¹⁷ An intra-cluster correlation coefficient of 0.05 and an expected 10% attrition rate were taken into account. Assuming an average of 50 students per school, recruitment of 300 students from 6 schools was required to provide 89.1%, 87.6%, and 88.0% power (α = 0.05) to detect the specified changes in BMI, CAP, and LSM, respectively. Calculations were performed using PASS 15.0.5.

The normality of continuous variables was assessed using the Shapiro–Wilk test. Baseline characteristics were presented as means with SDs, medians with IQRs or as numbers with proportions, as appropriate. Generalised linear mixed models (GLMMs) were applied to evaluate intervention effects, incorporating random intercepts at the school level to account for cluster effects. All models were adjusted for the baseline value of the corresponding outcome, age, and sex. Given the influence of BMI on hepatic status, baseline BMI was additionally adjusted in models evaluating CAP and LSM. Primary analyses adhered to the intention-to-treat principle, including all randomly assigned participants. Missing values were imputed using the baseline observation carried forward (BOCF) method. We also performed per-protocol (PP) sensitivity analyses using a complete-case approach for primary outcomes, including only participants who completed the 9-month follow-up assessment. In this study, completion of the final assessment defined protocol adherence, as the intervention was delivered at the school level. Additional sensitivity analyses were conducted by using the last observation carried forward (LOCF) method for primary outcomes at the 3-month follow-up, and by further adjusting the primary BOCF models for family sociodemographic factors.

Post-hoc subgroup analyses were conducted based on sex, baseline BMI category, maternal education and household per-capita disposable income, intervention-by-subgroup interactions were tested. Additional obesity-related indicators were evaluated post hoc, including the percentage of participants above the age- and sex-specific 50th BMI percentile,¹⁸ BMI ≥95th percentile, BMI ≥97th percentile, and severe obesity (defined as BMI of 120% of the 95th percentile¹⁹) were also measured. The potential mediating effects of obesity-related indicators on CAP and LSM were explored using the “mediation” package in R. A two-sided p-value <0.05 was considered statistically significant. All analyses were conducted using R software, version 4.3.1.

We adapted a health-augmented macroeconomic model to quantify the economic gains associated with reduced MASLD prevalence post hoc. This model links improvements in population health to economic growth by treating health capital as a fundamental driver of capital accumulation and aggregate output.²⁰ We projected overall GDP for 2026 to 2050 under two scenarios: a baseline scenario reflecting status quo development trends, and an intervention scenario that incorporating assumed reductions in MASLD prevalence. The model captures two main pathways through which MASLD influences economic outcomes. The first operates through human capital: reduced MASLD-related morbidity and mortality are expected to increase life expectancy and labor force participation, thereby enlarging the effective working-age population and boosting productive capacity. Age-specific human capital was estimated using the Mincer Equation,²¹ which relates educational attainment and work experience to individual earnings, reflecting the marginal benefits of additional schooling or work experience. The second pathway involves physical capital: a lower MASLD prevalence decrease medical expenditures related to disease management, freeing up funds for households and governments to increase savings and investment. This promotes physical capital accumulation and supports long-term economic growth.

We developed a macroeconomic projection model to compare two scenarios from 2026 to 2050: (1) a baseline scenario reflecting current epidemiological trends of MASLD in the absence of policy intervention, and (2) a counterfactual scenario simulating nationwide implementation of evidence-based interventions aimed at reducing the prevalence of MASLD. Key input parameters included MASLD-related disability-adjusted life years (DALYs), mortality rate, health expenditures, GDP forecasts, capital depreciation rates, and labor force projections. A variety of sensitivity analyses were conducted, including policy effect sise, mortality rate assumptions, capital depreciation rates, and a 1000-iteration bootstrap to assess parameter uncertainty. Further details regarding the model structure and parameters were provided in Appendix B (pp 12–21).

Role of the funding source

The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Results

Eight schools were contacted, of which six met the eligibility criteria and agreed to participate in the study. From 6 to 30 September, 2022, 394 students were assessed for eligibility. Of these, 331 met the inclusion criteria and were enrolled in the study (145 in the intervention group and 186 in the control group), 63 children in the intervention group declined before the intervention began, citing the high commitment for additional exercise and the demands of meal photography required for the multifaceted program. No significant differences were observed between included and excluded participants in either the intervention group (Table C1, Appendix p 22) or across all six schools (Table C2, Appendix p 23). Baseline characteristics were comparable between the two groups (Table 1). Among the participants, 120 (36.3%) were female, with a mean age of 8.5 years (SD 0.3). The study population was predominantly Han Chinese (n = 324, 97.9%), with only seven participants (2.1%) identifying as ethnic minorities (one Hui, five Tujia, and one Bouyei). The mean BMI was 20.4 (SD 2.2), corresponding to BMI Z score of 2.0 (SD 0.7). All participating schools completed the trial. No adverse events were reported throughout the trial. Six children (1.9%) were lost to follow-up by the end of the trial (Fig. 1). Missing data for primary outcomes was minimal (1.6–9.1%) and primarily attributed to non-systematic factors such as fear of blood sampling, competing examinations or sports events, and specific test refusals (e.g., VCTE in 15 control-group participants). Statistical comparisons confirmed that baseline characteristics did not differ significantly between participants with and without missing endline assessments (Tables C3–C4, Appendix pp 24–25), supporting the assumption that data were missing at random.

Table 1.

Baseline characteristics of the schools and participants.

Characteristics	Control group (n = 186)	Intervention group (n = 145)
Cluster level
No. of schools	3	3
No. of class	18	21
No. of children/class (Median)	11	7
Individual level
No. of children	186	145
Sex, n (%)
Boys	123 (66.1)	88 (60.7)
Girls	63 (33.9)	57 (39.3)
Household per-capita disposable income, n (%)
Below national average	50 (26.9)	35 (24.1)
At or above national average	136 (73.1)	110 (75.9)
Maternal educational level, n (%)
High school or below	90 (48.4)	32 (22.1)
College graduate and above	96 (51.6)	113 (77.9)
BMI status, n (%)
Overweight	93 (50.0)	70 (48.3)
Obesity	93 (50.0)	75 (51.7)
Severe obesity, n (%)	16 (8.6)	19 (13.1)
Percentage of the WHO 95th BMI percentile value, %	105.3 (11.1)	105.9 (13.2)
Mean (SD)
Age, year	8.5 (0.3)	8.5 (0.3)
Height, cm	135.5 (5.9)	135.3 (5.2)
Weight, kg	37.6 (6.0)	37.6 (6.0)
BMI, kg/cm2	20.4 (2.1)	20.5 (2.4)
BMI Z score	1.9 (0.7)	2.0 (0.8)
Body fat percentage, %	30.1 (5.6)	31.1 (6.4)
Waist circumference, cma	66.2 (6.7)	67.2 (6.5)
CAP, dB/mb	204.5 (42.0)	204.7 (43.1)
LSM, kPab,c	4.1 (1.0)	4.0 (0.8)
20mSRT, lapsd	19.9 (10.1)	21.5 (10.2)

Abbreviations: BMI: body mass index; CAP: controlled attenuation parameter; LSM: liver stiffness measurement; 20mSRT: 20-m shuttle run test.

^a1 missing in the control group.

^b5 missing the Fibroscan test in the control group, and 2 in the intervention group.

^c2 with invalid LSM values.

^d4 missing in the control group, and 2 in the intervention group.

Fig. 1.

Trial profile. ^aOnly complete cases at baseline were used for ITT analysis: in the intervention group, 145 for BMI, 143 for CAP (2 missed due to insufficient fasting time at baseline), 141 for LSM (2 insufficient fasting time, 2 invalid test). ^bIn the control group, 186 for BMI, 181 for CAP and LSM (5 missed due to insufficient fasting time at baseline). ^cPer-protocol (PP) analyses were used as sensitivity analysis for primary outcomes, in the intervention group, 140 students included for BMI as 1 student missed due to competition at the end of trial. ^dIn the control group, 182 students included for BMI as 2 students missed due to competition at the end of trial. Abbreviations: BMI: body mass index; CAP: controlled attenuation parameter; LSM: liver stiffness measurement; PP: per-protocol; VCTE: vibration-controlled transient elastography.

After one academic year, a reduction in mean BMI was observed in the intervention group, whereas an increase was noted in the control group, resulting in a mean difference of −0.38 (95% CI −0.59 to −0.16; p < 0.0001; Fig. 2A). A similar trend was observed for CAP, with a mean difference of −15.73 (−22.39 to −9.07; p < 0.0001; Fig. 2B). LSM increased in the control group but remained stable in the intervention group, leading to a mean difference of −0.69 (−0.89 to −0.48; p < 0.0001; Fig. 2C). Sensitivity analyses using the LOCF and PP approaches yielded results consistent with the primary analysis. Furthermore, additional adjustment for family socioeconomic status did not alter these findings (Table C5, Appendix p 26). At the 3-month follow-up (mid-intervention), favourable although non-significant trends were observed in BMI and CAP in the intervention group compared with the control group (Table C6, Appendix p 27). Subgroup analyses revealed no significant differences in treatment effects according to sex, baseline BMI status, maternal education level or household per-capita disposable income (Figure C1, Appendix p 35). Mediation analyses revealed that changes in BMI and BMI Z score accounted for 11.9% and 9.1%, respectively, of the improvements in CAP (both p < 0.05, Table C7, Appendix p 28). However, no significant mediation effect was observed for LSM.

Fig. 2.

Intervention effect on primary outcomes. (A) Mean BMI changes between intervention and control groups. (B) Mean CAP changes between intervention and control groups. (C) Mean LSM changes between intervention and control groups. Abbreviations: BMI: body mass index; CAP: controlled attenuation parameter; LSM: liver stiffness measurement; MD: mean difference.

Compared with the control group, the intervention group had significantly lower prevalence of MASLD (OR 0.32, 95% CI 0.12–0.87; p = 0.025) and liver fibrosis (0.10, 0.12–0.87; p = 0.032) after the intervention (Table 2). A lower prevalence of obesity was also observed in the intervention group, although this difference was not statistically significant. In addition, the intervention group demonstrated improvements in several secondary outcomes, including reductions in BMI Z score (mean difference −0.13, 95% CI −0.20 to −0.06; p < 0.0001), body fat percentage (−1.30, −2.58 to −0.01; p = 0.047), fasting insulin (−0.36, −0.50 to −0.23; p < 0.0001) and triglyceride levels (−0.27, −0.42 to −0.12; p < 0.0001). Post hoc analyses for other obesity-related indicators also indicated consistent beneficial effects associated with the intervention (Table C8, Appendix p 29). No significant between group differences were observed in waist circumference, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, total cholesterol, ALT levels, or performance on CRF.

Table 2.

Intervention effects on secondary outcomes.

Secondary outcomes	No. in control/intervention group	Control group		Intervention group		Adjusted mean difference/OR (95% CI)	p
		Baseline (September 2022)	End of trial (June 2023)	Baseline (September 2022)	End of trial (June 2023)
Obesity (yes, using Chinese reference), n (%)	186/145	93 (50.0)	76 (40.9)	75 (51.7)	53 (36.6)	0.67 (0.35–1.27)	0.22
Obesity (yes, using WHO reference), n (%)	186/145	78 (41.9)	71 (38.1)	58 (40.0)	45 (31.0)	0.58 (0.29–1.16)	0.13
MASLD (yes), n (%)a	181/143	24 (13.3)	23 (12.7)	19 (13.3)	11 (7.7)	0.32 (0.12–0.87)	0.025
Liver fibrosis (yes), n (%)a,b	181/141	2 (1.1)	10 (5.5)	0 (0)	1 (0.7)	0.10 (0.01–0.82)	0.032
BMI Z score, mean (SD)	186/145	1.95 (0.68)	1.79 (0.71)	1.98 (0.80)	1.68 (0.84)	−0.13 (−0.20 to −0.06)	<0.0001
Body fat percentage, mean (SD), %	186/145	30.1 (5.6)	30.2 (5.9)	31.1 (6.4)	29.9 (6.6)	−1.30 (−2.58 to −0.01)	0.047
Waist circumference, mean (SD), cmc	185/145	66.2 (6.7)	68.3 (7.4)	67.2 (6.5)	68.1 (7.9)	−1.03 (−2.17 to 0.12)	0.080
FINS, median (IQR), pmol/Ld	185/143	65.4 (37.2)	65.7 (48.0)	68.9 (44.6)	45.4 (34.6)	−0.36 (−0.50 to −0.23)f	<0.0001
FPG, mean (SD), mmol/Ld	185/143	5.0 (0.3)	4.9 (0.4)	5.0 (0.4)	5.1 (0.3)	0.20 (0.13–0.28)	<0.0001
TG, median (IQR), mmol/Ld	185/143	0.8 (0.4)	1.1 (0.6)	0.8 (0.5)	0.8 (0.5)	−0.27 (−0.42 to −0.12)f	<0.0001
TC, mean (SD), mmol/Ld	185/143	4.6 (0.8)	4.6 (0.8)	4.7 (0.8)	4.8 (0.8)	0.20 (−0.03 to 0.44)	0.093
LDL-C, mean (SD), mmol/Ld	185/143	2.8 (0.6)	2.6 (0.6)	2.9 (0.6)	3.0 (0.6)	0.23 (−0.04 to 0.48)	0.092
HDL-C, mean (SD), mmol/Ld	185/143	1.5 (0.3)	1.5 (0.2)	1.5 (0.3)	1.5 (0.3)	0.06 (−0.01 to 0.13)	0.084
ALT, median (IQR), U/Ld	185/143	15.0 (9.0)	15.0 (5.0)	15.0 (8.0)	16.0 (10.0)	0.08 (−0.06 to 0.21)f	0.27
20mSRT, mean (SD), lapse	184/141	19.9 (10.1)	27.3 (11.8)	21.5 (10.2)	27.1 (11.7)	−0.93 (−3.17 to 1.31)	0.42

Abbreviations: ALT: alanine transaminase; BMI: body mass index; FBG: fasting blood glucose; FINS: fasting insulin; HDL-C: high-density lipoprotein cholesterol; LDL-C: low-density lipoprotein cholesterol; LSM: liver stiffness measurement; MASLD: metabolic dysfunction–associated steatotic liver disease; TC: total cholesterol; TG: triglyceride; 20mSRT: 20-m shuttle run test.

^a1 missing at baseline in the control group.

^b5 missing the Fibroscan test at baseline in the control group, and 2 in the intervention group.

^c1 missing at baseline in the control group.

^d1 missing the blood test at baseline in the control group, and 2 in the intervention group.

^e4 missing at baseline in the control group, and 2 in the intervention group.

^fThe dependent variable was log-transformed when included in the model.

In terms of knowledge and behaviours (Table C9, Appendix p 30), the intervention group showed significant improvements in healthy lifestyle knowledge scores (mean difference 0.49, 95% CI 0.18–0.79; p = 0.0020) and Global Dietary Recommendation scores (0.63, 0.12–1.14; p = 0.015). More students in the intervention group reached the action stage of behavioural change (OR 2.41, 95% CI 1.25–4.66; p = 0.091), ate out less frequently (0.52, 0.30–0.90; p = 0.018) and had more days of MVPA (>1 h/day; mean difference 0.90, 95% CI 0.40–1.41; p < 0.0001). No significant differences were observed between two groups in depression, social anxiety, or sleep quality scores.

In three intervention schools, the intervention was delivered with high fidelity (>88%) for health education activities for students, parents and school staff. Regarding mHealth engagement, 139 of 145 caregivers (95.9%) accessed the platform, and 99 (71.2%) uploaded dietary photos. Adherence rates were 78.6% for weight monitoring and 66.2% for dietary behaviour monitoring. Student attendance at weekly extracurricular physical activities ranged from 77% to 82%. At the school level, all intervention schools allocated curriculum time for health education and physical activity, implemented restrictive food policies, while two schools expanded exercise spaces and facilities (Table C10, Appendix p 31).

In the baseline scenario in post-hoc macroeconomic analyses, which assumes no intervention and a continuation of current MASLD epidemiological trends, the projected total economic burden of MASLD is Chinese RMB (¥) 2004.98 billion (US$ 281.20 billion) at constant 2025 prices (based on the 2025 exchange rate of US $1 = RMB ¥7.13; Table 3). This corresponds to approximately 0.038% of the cumulative GDP between 2026 and 2050. In contrast, under a steady-state policy scenario simulating a nationwide MASLD intervention in primary school with full population coverage, the projected cumulative macroeconomic gains reach RMB 694.96 billion (US$ 97.47 billion). This is equivalent to RMB 493.48 (US$ 69.21) per capita, accounting 0.013% of the cumulative GDP over the same period.

Table 3.

Estimated macroeconomic gains from MASLD reduction, 2026–50.

	Total GDP (billion CNY)	Gains per capita (CNY)	GDP share (${\times \mathbf{10}}^{-\mathbf{3}}\%$)
Panel A: Baseline results
Total burden of MASLD	2004.98	1423.71	37.66
Baseline effects (from OR)	694.96	493.48	13.05
Panel B: Variation in policy effects
Lower bound of estimated OR	132.87	94.35	2.50
Upper bound of estimated OR	899.34	638.61	16.89
Taking effects gradually	293.86	208.66	5.52
Panel C: Mortality variation
Lower bound of mortality rate	630.93	448.02	11.85
Upper bound of mortality rate	782.86	555.90	14.70
Panel D: Alternative depreciation rates
Depreciation rate = 0	556.96	395.49	10.46
Depreciation rate = 3%	637.03	452.34	11.97
Depreciation rate = 10%	846.19	600.87	15.89
Panel E: Permutation tests of parameters
Random bootstrap (50%–150% of initial value)	677.98 (149.07, 1108.87)	481.42 (105.85, 787.40)	12.73 (2.80, 20.83)

Notes: Total economic gains and gains per capita between 2026 and 2050 were adjusted to the price level in 2025 through the GDP deflator. The specific specifications used in Panels B to D are described in Appendix. In Panel E, a permutation test was performed 1000 times, with the resulting 95% confidence intervals (CIs) displayed in parentheses.

Abbreviations: CNY: China Yuan; GDP: gross domestic product.

Panels B–E in Table 3 reported the results of sensitivity analyses for economic gains. Panel B, showed that under the lower and upper estimates of policy effectiveness, cumulative gains ranged from 0.003% to 0.017% of GDP. When a gradual effect onset was assumed, the estimated gain decreased to 0.006% of GDP. Panel C reported the sensitivity to the 95% CIs of MASLD-related mortality rates, with projected GDP gains ranging between 0.012% and 0.015% of GDP. Panel D examined sensitivity to capital depreciation rate, yielding gains between 0.010% and 0.016% of GDP. In Panel E, a bootstrap simulation with 1000 iterations was conducted by resampling key parameters between 50% and 150% of their original values, and the mean gain was 0.013% (95% CI 0.003%–0.021%) of total GDP.

Discussion

This mHealth-assisted multifaceted lifestyle intervention effectively reduced BMI, CAP, and LSM in 8- to 10-year-old children with overweight or obesity after one academic year. The intervention also demonstrated significant reductions in metabolic outcomes, including body fat percentage, fasting insulin, and triglycerides, alongside improvements in several diets and physical activity knowledge and behaviours. Furthermore, the post-hoc, health-augmented macroeconomic model projects that these health improvements could translate into substantial long-term economic dividends, estimated at approximately 0.013% of cumulative GDP by 2050. To our knowledge, this study represents the first cluster RCT addressing both pediatric obesity, MASLD and their economic implications through a multifaceted intervention approach.

Dietary habits play an essential role in obesity and hepatic steatosis pathogenesis. In this study, clinical nutritionists guided dietary changes both at school and home, resulting in improved dietary quality and a reduction in eating out frequency among students. Previous research indicated that nutrition literacy alone may be insufficient to drive behaviour changes in patients with MASLD.²² However, with professional guidance, our intervention group presented improved Global Dietary Recommendation scores and reduced rates of eating out. Such professional support is crucial to translating knowledge into action, which aligns with US Preventive Services Task Force recommendations for intensive behavioural interventions in children aged 6 years or older with high BMI (>95th percentile, children with obesity).²³ Given China's vast territory and diverse culinary cultures, nutritionists' familiarity with regional dietary habits allows them to adapt recommendations to local, seasonal foods, making the guidance both practical and culturally relevant.²⁴ Moreover, the mobile application facilitated communication between parents and nutritionists, enabling parents to receive timely, practical guidance on healthy food substitutions.

The effect of the intervention on BMI in this study (mean difference −0.38, 95% CI −0.59 to −0.16) was smaller than that observed among children with overweight or obesity in DECIDE-Children study (−0.76, −1.02 to −0.50),⁸ although the 95% CIs overlapped. Several factors may explain this disparity. The COVID-19-related early winter break reduced the duration of school-based moderate-to-vigorous physical activity. This temporary reduction in intervention intensity likely led to an underestimation of the true intervention effect sise. Participants in the present study were also younger (Grade 3) than those in the DECIDE-Children study (Grade 4). Although biological maturation can influence adiposity measures, recent data from the Chinese National Surveys on Students' Constitution and Health indicate that the median ages at menarche and spermarche are 12.0 and 13.9 years, respectively.²⁵ Given our participants’ age and the observed annualised growth velocity of ∼5.4 cm/year, which aligns with prepubertal norms, it is unlikely that a significant proportion of participants had entered the rapid pubertal growth phase. This limits the potential confounding effect of maturation. Furthermore, consistent intervention effects were observed across supplementary outcomes (BMI Z score, body fat percentage, and waist circumference), all of which were adjusted for age and sex, suggesting that the results are robust to maturational variations.

Although the reduction in BMI is modest, evidence indicates that even small improvements can bring meaningful clinical benefits in school-aged children. In the study, the decrease in BMI Z score exceeded thresholds (0.1-unit) previously linked to cardiometabolic improvements and favorable changes in liver enzymes, and was accompanied by favorable changes in body fat percentage, fasting insulin, triglycerides, and hepatic steatosis.²⁶ Importantly, intervening at 8–10 years of age, before pubertal metabolic changes, may help redirect long-term obesity trajectories and potentially reduce risks of future chronic diseases such as type 2 diabetes and premature mortality.²⁷ Previous evidence demonstrated that such interventions were likely to be more effective for girls.¹⁶ Nonetheless, we did not find any difference in sex subgroup analyses, which is in line with the findings of the DECIDE-Children study.⁸ The possible reason is the higher level of family involvement and relatively high fidelity of the intervention delivery. As parents playing a central role in executing dietary and physical activity strategies, potential sex-related differences in behavioural responses may have been attenuated.

Despite the significant burden of pediatric MASLD, its awareness and treatments remain limited.⁶ A recent review of 10 RCTs (one school-based, 9 hospital-based) on childhood MASLD, which focused on diet or exercise, showed promising results but were limited by small sample sises and a lack of cluster RCTs.²⁸ In this study, while the absolute mean reductions in CAP and LSM may appear modest compared to intensive therapeutic trials in severe disease,¹⁷ their clinical relevance is underscored by the “prevention of progression”. Most importantly, this shift in continuous variables translated into tangible clinical benefits: a significant reduction in MASLD risk (OR 0.32) and a marked decrease in the proportion of children exceeding pediatric liver stiffness thresholds (5.1 kPa, Table C11, Appendix p 32). Although primarily focused on a childhood obesity intervention, this study provides exploratory insights into the primary prevention of hepatic health. Clinically, variations within this low CAP range (<248 dB/m) do not represent current pathology; however, evidence suggests that rising CAP levels increase diabetes risk,²⁹ and elevated LSM serves as a predictor for metabolic syndrome.³⁰ Ectopic fat deposition driven by sustained adiposity is a continuous pathophysiological process.³¹ While the categorical prevalence of overweight and obesity in our cohort did not dramatically reverse, the intervention successfully induced a concurrent reduction in both continuous adiposity metrics (BMI and BMI Z scores) and hepatic fat content. This demonstrates that even before children completely transition out of overweight/obesity categories, lifestyle modifications can effectively intercept the upward trend of early subclinical fat accumulation, essentially shifting the population's long-term risk curve to the left before pathological thresholds are reached. Moreover, compared with magnetic resonance imaging (MRI), VCTE has limited sensitivity for quantifying subtle changes in hepatic steatosis, particularly when steatosis is below 20%,³² but its low cost and portability are essential for translating public health research into large-scale practice.

A weight reduction (>5%) has been shown to lead to steatosis remission in adults, with 10% needed for fibrosis improvement.³ Our study observed a 6.5% reduction in BMI Z score, correlating with liver steatosis and fibrosis improvement. However, BMI explained a limited proportion of the improvement in CAP, suggesting that other factors might have contribution to the intervention effect, such as improved dietary quality, increased physical activity, and metabolic changes including better insulin sensitivity and lower triglycerides.³³ Although the intervention showed some effect on LSM, no mediation effect for BMI was found, possibly due to the limited BMI change or short intervention duration. In contrast, ALT did not show significant inter-group differences. Given the relatively low baseline ALT levels across the intervention, there was minimal room for further reduction. Furthermore, reduction in hepatic fat content can occur independently of significant alterations in ALT, particularly when baseline ALT levels are not markedly elevated.³⁴

Building upon the observed reduction of prevalence of MASLD, we translated these clinical improvements into long-term societal impacts using a health-augmented macroeconomic model. Our projections estimate cumulative economic gains of approximately 0.013% of GDP (¥694.96 billion) by 2050. Notably, this magnitude aligns with other population-level public health policies, falling between the projected gains of sugar-sweetened beverage taxation (∼0.0016% of GDP) and alcohol excise tax increases (∼0.034% of GDP), underscoring the substantial economic potential of early MASLD prevention.³⁵ However, these health-economic findings must be interpreted as a scenario-based exploration rather than a definitive macroeconomic forecast. The model used national-level demographic and epidemiological parameters combined with effect sizes from a specific regional sample (Ningbo, China). Given the limited sample size and regional focus of the trial, extrapolating these results requires caution regarding real-world scalability. These estimates serve as a proof-of-concept for future cost-effectiveness studies. Second, regarding sustainability, although the one-year timeframe inherently captured the impact of the winter break, potential “behavioural wash-out” during the longer summer hiatus remains unassessed. To address these uncertainties, we conducted extensive sensitivity analyses adjusting for sex, income, and applying a conservative 62.5% disease persistence rate from childhood to adulthood (Table C12, Appendix p 33). Results indicated that even under the most conservative assumptions, the intervention consistently generated positive economic returns (ranging from 0.003% to 0.017% of GDP). Thus, while local adaptation is necessary, the current evidence suggests that a nationwide rollout would likely generate meaningful and robust macroeconomic benefits.

This study has several strengths. First, clinical nutritionists were involved in delivering dietary guidance, which helped translate weight management knowledge into practical skills, benefiting not only the students’ health but also dietary practices among parents and school cafeterias. Second, the purposely designed mobile application effectively facilitated timely and convenient communication among schools, nutritionists, and parents, aligning with the communication preferences of modern families. Third, the adequate sample sise and cluster-RCT design strengthen the reliability of our findings. A post hoc power analysis based on the observed effect sises suggested moderate to adequate statistical power (Table C13, Appendix p 34). Fourth, the economic modelling suggests that interventions targeting childhood MASLD may bring considerable economic returns.

Nonetheless, several limitations should be noted. First, this study focused on one city, Ningbo, and included six public primary schools in eastern China, the findings are likely generalisable to similar urban contexts, but should be cautiously extended to regions with different economic or medical conditions. Second, COVID-19 pandemic restrictions and an early winter break confined children to their homes, significantly reduced extra school-based MVPA in the intervention groups, which likely underestimate the effect sise in the intervention group. Third, due to the nature of the lifestyle intervention, blinding of participants was infeasible, which may introduce performance bias. However, this risk was mitigated by the use of objective primary outcomes and blinded outcome assessment. Finally, the intervention covered only one academic year, and longer-term sustainability of the observed effects could not be evaluated.

This cluster-randomised trial demonstrated that an mHealth-assisted, multifaceted lifestyle intervention effectively improved in BMI, liver health, and metabolic outcomes among children with overweight or obesity. These findings support the incorporation of such interventions into national child health strategies, particularly for mitigating obesity and preventing non-communicable disease. Ultimately, this study highlights the potential of multifaced interventions to yield both immediate clinical benefits and substantial long-term macroeconomic gains.

Contributors

All authors contributed to data interpretation, critically reviewed drafts, and read and approved the final version of the manuscript. Conceptualisation: LL, QS, HW, HJW. Data curation: PZ, YW. Formal analysis: PZ, DX. Funding acquisition: LL, HW. Investigation; PZ, EN, JG, MX, YZ, FS, HJ, XL, YX, SJ, JL. Methodology: PZ, LL, HF, HW, HJW. Project administration: JL, HW. Resources: LL, HL, HJW. Supervision: MX, HW, HJW. Validation: YW, EN, HF. Writing—original draft: PZ, LL. Writing—review & editing: HF, LW, HW, HJW. All authors had full access to all the data.

Data sharing statement

De-identified datasets are available upon reasonable request to the corresponding author. To ensure alignment with the original ethical framework and participant privacy, data requests must be accompanied by a detailed research protocol and a rigorous analysis plan. All requests are subject to approval by the Institutional Review Board (IRB).

Declaration of interests

All authors declare no competing interests.