Pharmacotherapy for Children and Adolescents With Overweight or Obesity: A Systematic Review and Network Meta‐Analysis of Randomized Controlled Trials

ABSTRACT

Aims

The escalating global prevalence of childhood and adolescent overweight and obesity constitutes a pressing public health issue. This network meta‐analysis aimed to compare the efficacy and safety of pharmacologic interventions as adjuncts to lifestyle therapy in this population.

Materials and Methods

Randomized controlled trials (RCTs) were sourced from systematic searches of PubMed, Embase (using the OVID platform), the Cochrane Library (CENTRAL), the WHO International Clinical Trials Registry Platform (ICTRP), and ClinicalTrials.gov from database inception to July 28, 2025. A frequentist network meta‐analysis was performed using a random‐effects model. We used the GRADE (Grading of Recommendations Assessment, Development, and Evaluation) approach to evaluate the overall certainty of evidence and categorized the interventions.

Results

A total of 41 RCTs (N = 3923) were included. Compared with lifestyle modification alone, semaglutide produced the largest reduction in the 95th BMI percentile (MD −20.40%, 95% CI −24.22 to −16.58), with an additional 500 and 399 patients per 1000 person‐years achieving ≥ 5% and ≥ 10% BMI reduction, respectively. Phentermine‐topiramate showed the next largest reduction (MD −18.35%, 95% CI −22.26 to −14.45), corresponding to 554 and 734 additional responders per 1000 person‐years. Liraglutide and exenatide also significantly reduced the 95th BMI percentile compared with lifestyle modification alone, although the reductions were smaller than those achieved with semaglutide and phentermine‐topiramate.

Conclusions

Both semaglutide and phentermine‐topiramate were closely linked to weight management, and phentermine‐topiramate, with its favourable tolerability, may serve as the optimal adjunct to lifestyle interventions.

1. Introduction

Overweight and obesity have profound implications for numerous children and adolescents, with the data indicating a persistent upward trend in affected populations. Between 1990 and 2021, the global total prevalence of overweight and obesity in children and adolescents more than doubled, while the prevalence of obesity alone tripled [1]. Non‐Communicable Diseases Risk Factor Collaboration (NCD‐RisC) estimated that in 2022, 9.3% of boys and 6.9% of girls aged 5–19 years were classified as obese, corresponding to approximately 159 million school‐age children and adolescents [2]. Beyond the heightened risk of physical health conditions [3, 4, 5], overweight and obesity also significantly impact the societal and emotional well‐being of young individuals [6, 7]. Clearly, such circumstances hinder the optimal growth and development of children and adolescents, emphasizing the urgency of addressing this matter.

Lifestyle intervention serves as the cornerstone of weight management for children and adolescents with overweight or obesity [8, 9], encompassing dietary modification, physical activity promotion, and behavioural therapy [10, 11]. The requirement for sustained, long‐term commitment often leads to suboptimal adherence [12], resulting in frequent treatment failure or rapid weight regain after initial improvement [13]. To overcome these barriers and improve outcomes, evidence‐based pharmacotherapeutic strategies are increasingly recognized as necessary adjuncts [14, 15]. In 2023, the American Academy of Paediatrics (AAP) issued its first clinical practice guideline for paediatric obesity, similarly recommending pharmacotherapy as an adjunct to lifestyle interventions [16].

Although numerous pharmacological interventions for weight reduction have been studied in adults, fewer agents are approved for paediatric use [17]. Existing network meta‐analyses in children and adolescents with overweight or obesity have primarily assessed continuous measures such as BMI and BMI z‐scores, with proportion‐based outcomes receiving limited attention. Thresholds of ≥ 5% and ≥ 10% BMI reduction, which are established benchmarks for clinical success in adult studies, are seldom applied in paediatric research [18]. Furthermore, the BMI z‐score has inherent limitations due to statistical compression at higher BMI values, particularly in youth with severe obesity, potentially underestimating treatment efficacy [19]. To address these gaps, this network meta‐analysis focuses on the proportions of patients achieving ≥ 5% and ≥ 10% BMI reduction, along with the change in the 95th BMI percentile, to supplement current evidence and provide more clinically interpretable guidance.

2. Materials and Methods

2.1. Study Design

This network meta‐analysis was conducted according to the PRISMA 2020 statement (Preferred Reporting Items for Systematic reviews and Meta‐Analyses) and their extension for network meta‐analysis (PRISMA‐NMA) [20, 21]. The review was registered in the International Prospective Register of Systematic Reviews, PROSPERO (CRD42022329226).

2.2. Eligibility Criteria

We included randomized controlled trials (RCTs) of pharmacotherapy for weight reduction in children and adolescents (aged ≤ 18 years) with obesity or overweight, regardless of the presence of weight‐related complications. Trials were excluded if they (1) enrolled patients who had medication‐induced obesity, pregnancy, or normal bodyweight; (2) enrolled patients with malignant tumours or blood diseases; (3) were not published in English; (4) had no available results related to weight loss.

2.3. Search Strategy

PubMed, Embase (using the OVID platform), and the Cochrane Library (CENTRAL) were conducted from inception to July 28, 2025 for the initial search. Furthermore, to supplement the identified citations, we also searched International clinical trials registry platform (ICTRP) and ClinicalTrials.gov. Supporting Information: Appendix 2 shows the detailed search strategy.

2.4. Study Selection

The study selection process followed the PRISMA flowchart. After removing duplicates identified by EndNote 21, two reviewers independently screened records by title and abstract, followed by full‐text assessment. Discrepancies were resolved by consensus or by consulting a senior researcher.

2.5. Outcomes

This study employed both continuous and responder‐based endpoints as primary outcomes. The change in the percentage of the 95th BMI percentile (Δ%BMIp95) was used as a continuous measure, reflecting the mean change in obesity severity among children and adolescents. The proportion of participants achieving at least 5% and at least 10% reductions in BMI from baseline served as binary responder endpoints, representing initial and substantial clinical weight loss benefits, respectively.

Secondary outcomes included change in BMI from baseline and adverse outcomes (such as discontinuation due to adverse events, serious adverse events, total gastrointestinal adverse events, nausea events, vomiting events, and diarrhoea events).

2.6. Data Extraction and Risk of Bias Assessment

Two reviewers independently extracted data into a standardized Excel spreadsheet. The extracted data included study characteristics, participant details, and outcome measures.

Risk of bias (ROB) for all included randomized controlled trials was assessed by two independent assessors with the Cochrane Risk of Bias 2 (ROB‐2) tool [22].

2.7. Publication Bias

Funnel plots were used to assess potential publication bias. Initially, symmetry was evaluated visually. If more than 10 trials were included, statistical testing was performed using Egger's test, the Begg‐Mazumdar test, and the Thompson‐Sharp test.

2.8. Statistical Analysis

Network meta‐analyses were conducted within a frequentist framework, implemented using a graph‐theoretical approach [23]. For dichotomous outcomes, odds ratios (ORs) were calculated, while mean differences (MDs) were utilised for continuous outcomes. Both measures were accompanied by 95% confidence intervals (CIs) to provide a range of estimated effect sizes. A random‐effects model was used to account for variability among the included studies.

Heterogeneity was assessed using Cochran's Q test, with a p value > 0.05 indicating no significant heterogeneity. The transitivity assumption was assessed by comparing the distribution of potential effect modifiers between different intervention comparisons (see Supporting Information: Appendix 6.6). Inconsistency between direct and indirect evidence was examined using the node‐splitting method [24, 25].

To rank interventions by efficacy and safety, P‐scores were calculated for each treatment. Ranging from 0 to 1, a higher P‐score indicates a greater probability of being the most effective option [26]. The absolute risk difference (RD) of the intervention versus lifestyle modification alone was estimated based on the relative effect and baseline risk, where the baseline risk was simulated using the pooled estimate of the control group (lifestyle modification alone) derived from a random‐effects single‐arm meta‐analysis [27] (see Supporting Information: Appendix 3.3).

Seven sensitivity analyses were conducted to ensure the robustness of the main analyses. A detailed description is provided in Supporting Information: Appendix 3.1.

To explore the potential influence of mean age, gender, mean BMI and weight at baseline, as well as the follow‐up duration on the outcomes, a meta‐regression analysis with the restricted maximum likelihood estimator method (REML) was conducted.

All statistical analyses were performed using the R software version 4.2.2.

2.9. Certainty of Evidence (GRADE) and Categorization of Interventions

We assessed certainty of the evidence by applying the GRADE approach, thus reflecting the methodological reliability of the network meta‐analysis [28, 29].

The minimal important differences (MID) of the primary outcomes were identified by searching through previous research studies. We classified the interventions into four categories based on whether the effect estimate values were greater than or less than MID [30]. For the outcome of discontinuation due to adverse events and total gastrointestinal events, we set three categories of effect class based on comparisons with other drugs and lifestyle interventions. A detailed description is provided in Supporting Information: Appendix 3.5.

3. Results

3.1. Characteristics of the Included Studies

We screened 2710 records from the databases (PubMed, Embase, the Cochrane Library), ICTRP and ClinicalTrials.gov and 41 eligible trials were identified in this network meta‐analysis, including 39 RCTs published between 2001 and 2024 and 2 RCTs without any publications, only results available on clinical trial websites. The retrieval flow diagram is shown in Figure 1. A total of 3923 participants were included, with a mean age ranging from 8.1 to 16.1 years. The baseline mean BMI ranged from 26.2 to 41.7 kg/m². The follow‐up duration of the included trials ranged from 5 weeks to 2 years. The racial distribution of participants was predominantly White, followed by Black and Asian individuals. In terms of ethnicity, the proportion of Hispanic/Latino participants was relatively low. Supporting Information: Appendix 4 details the study characteristics.

FIGURE 1.

PRISMA flow diagram of the study selection process.

3.2. Risk of Bias Assessment

A summary of the results on the risk of bias is presented in Supporting Information: Appendix 5. According to the RoB 2 tool, three trials were judged to have a high risk of bias, mainly due to issues in the randomization process and missing outcome data. Twenty three trials raised some concerns, primarily related to deviations from intended interventions, the randomization process, and missing outcome data. These trials with risk‐of‐bias concerns involved multiple drugs, including exenatide, liraglutide, orlistat, lixisenatide, and metformin. In the GRADE assessment, we systematically downgraded the certainty of evidence by one level for each relevant outcome based on the specific characteristics of the included studies (see Supporting Information: Appendix 8).

3.3. Network Plots

We established networks involving 10 outcomes. The network plots of the primary outcome are shown in Figure 2. Each circle represents an intervention (including semaglutide, liraglutide, exenatide, orlistat, phentermine‐topiramate, topiramate, and lifestyle modification alone), and the size of the circle is proportional to the number of participants in that intervention. The line indicates that there is a direct comparison between two interventions. The thickness of the line is proportional to the number of trials. Four interventions were included in the network for change in the 95th BMI percentile percentage, and five interventions were included in each of the networks for ≥ 5% and ≥ 10% BMI reduction. The network diagram for the secondary outcomes can be found in Supporting Information: Appendix 6.1.

FIGURE 2.

Network meta‐analysis plots. (A) Percentage of participants achieving BMI reduction of at least 5%, (B) Percentage of participants achieving BMI reduction of at least 10%, (C) Change in the percentage of the 95th BMI percentile. LMA, lifestyle modification alone.

3.4. Primary Outcomes

Seven RCTs with 1369 participants reported the percentage of individuals who achieved a BMI reduction of at least 5% or 10%. As shown in Figure 3, both phentermine‐topiramate and semaglutide were superior to lifestyle modification alone in achieving ≥ 5% BMI reduction, with odds ratios (ORs) of 14.06 (95% CI 2.67 to 73.94; moderate‐quality evidence) and 11.02 (95% CI 2.87 to 42.30; moderate‐quality evidence), respectively. The corresponding ORs for ≥ 10% BMI reduction were 72.21 (95% CI 3.45 to 1510.97; moderate‐quality evidence) for phentermine‐topiramate and 16.35 (95% CI 3.68 to 72.67; moderate‐quality evidence) for semaglutide.

FIGURE 3.

Relative effect sizes of pharmacotherapy and GRADE Rating. (A) Percentage of participants achieving BMI reduction of at least 5% (odd ratios, 95% CI), (B) Percentage of participants achieving BMI reduction of at least 10% (odd ratios, 95% CI). (C) Change in the percentage of the 95th BMI percentile (mean differences, 95% CI). P‐score, a metric used to rank the performance of interventions in terms of weight loss and safety, with a higher p‐score indicating a greater likelihood to be the most effective option.

For the percentage change in the 95th BMI percentile, compared with lifestyle modification alone, semaglutide showed a mean difference of −20.40% (95% CI −24.22 to −16.58; high‐quality evidence), with phentermine‐topiramate showing −18.35% (95% CI −22.26 to −14.45; high‐quality evidence) and liraglutide showing −6.24% (95% CI −6.55 to −5.93; high‐quality evidence). Both semaglutide and phentermine‐topiramate were more effective than the other agents in the network (Figure 3C), with evidence quality ranging from low to high.

3.5. Secondary Outcomes

The relative effects of secondary outcomes are presented in Supporting Information: Appendix 6.2. For the change from baseline in BMI, phentermine‐topiramate, semaglutide, liraglutide, metformin, and orlistat all showed greater reductions than lifestyle modification alone. Among them, phentermine‐topiramate and semaglutide proved more effective than the other agents.

Regarding gastrointestinal adverse events, exenatide, liraglutide, and semaglutide all exhibited higher rates than lifestyle modification alone. Liraglutide, in particular, showed a significantly greater incidence than metformin and phentermine‐topiramate, yet remained below that of orlistat. When focusing on individual symptoms, nausea and vomiting were more frequent with liraglutide, semaglutide, and metformin than with either orlistat or lifestyle modification alone, while diarrhoea occurred more often with orlistat and liraglutide than with lifestyle modification alone.

Compared with lifestyle modification alone, orlistat and liraglutide were more likely to be discontinued due to any adverse event. No significant differences in serious adverse events were observed among any pharmacological therapies or lifestyle intervention. Across all included studies, a total of 84 serious adverse events were reported: 57 in the intervention groups and 27 in the lifestyle intervention groups. Of these, only five were considered by the investigators to be possibly related to the study medication: symptomatic cholelithiasis in the orlistat group, hospitalization due to muscle spasms in the phentermine/topiramate group, and two cases of vomiting and one case of colitis in the liraglutide group (see Supporting Information: Appendix 13).

3.6. Categorization of Interventions

As the minimal important difference (MID) for %BMIp95 in paediatric obesity has not been established, we used a 12‐month change of −2.43% (p = 0.008) in %BMIp95 from a previous observational study as the primary threshold [31]. This value was derived from normoglycemic children with obesity who received lifestyle intervention alone. Given that most participants in our study were nondiabetic, this threshold may be more relevant to our population. To evaluate the robustness of our findings, we performed sensitivity analyses using three alternative thresholds: −2.63% from the prediabetic group in the same study, and −3.0% and −4.03% from two other observational studies that did not exclude participants with diabetes [32, 33]. Phentermine‐topiramate, semaglutide, and liraglutide were classified as “among the best” because the upper limits of their 95% confidence intervals for the effect estimates were smaller than the MID. Sensitivity analyses yielded categorical results for all drugs that were consistent with the primary analysis.

The MID for the proportions of participants achieving ≥ 5% and ≥ 10% BMI reduction was defined as twice the corresponding proportions observed in the lifestyle modification alone group (see Supporting Information: Appendix 3.5). The primary benefit outcomes were assessed according to the MID to determine clinical importance. Additionally, we classified the harm outcomes of interventions into three categories: among the worst, intermediate and among the best. These categorizations were based on the comparisons with lifestyle modification alone and other therapies. Figure 4 illustrates the categorization of interventions.

FIGURE 4.

Relative effects and absolute effects of interventions on benefit and harm outcomes.

We calculated the absolute effects for dichotomous outcomes (see Figure 4 and Supporting Information: Appendix 6.3). With lifestyle modification alone, 138 patients per 1000 person‐years achieved ≥ 5% BMI reduction, and 47 per 1000 person‐years achieved ≥ 10% BMI reduction. For both endpoints, the lower confidence limits of the effect estimates for phentermine‐topiramate and semaglutide exceeded the MID; therefore, both agents were classified as “among the best”. For liraglutide, the 95% confidence interval crossed the MID, but the point estimate fell to the right of the MID, leading to its classification as “possibly better than lifestyle modification alone”.

Apart from orlistat and liraglutide, which were associated with 22 and 41 additional discontinuations per 1000 person‐years, respectively, no intervention differed significantly from lifestyle modification alone in terms of discontinuation due to adverse events. Regarding total gastrointestinal adverse events, semaglutide, liraglutide, and orlistat were all found to be more harmful than lifestyle modification alone and than some other interventions.

A two‐dimensional efficacy‐safety plot of the drug treatments is provided in Figure 5. Phentermine‐topiramate was consistently located in the lower‐right region, indicating a preferable balance of efficacy and safety relative to other interventions.

FIGURE 5.

Two‐dimensional graphs of efficacy versus safety of pharmacotherapy. (A) Percentage of participants achieving BMI reduction of at least 5% versus discontinuation due to adverse events, (B) Percentage of participants achieving BMI reduction of at least 5% versus total gastrointestinal adverse events, (C) Percentage of participants achieving BMI reduction of at least 10% versus discontinuation due to adverse events, (D) Percentage of participants achieving BMI reduction of at least 10% versus total gastrointestinal adverse events, (E) Change in the percentage of the 95th BMI percentile versus discontinuation due to adverse events, (F) Change in the percentage of the 95th BMI percentile versus total gastrointestinal adverse events. Colored nodes depict the effect sizes, while coloured lines correspond to the confidence intervals.

3.7. Assessment of Heterogeneity, Transitivity, Inconsistency and Publication Bias

Detailed heterogeneity assessments are provided in Supporting Information: Appendix 6.5. Some degree of heterogeneity was detected for the change in BMI from baseline and for total gastrointestinal adverse events. Transitivity assessment showed that the distributions of key effect modifiers (age, sex, and baseline BMI) were comparable across treatment comparisons (Supporting Information: Appendix 6.6). In all direct comparisons, the changes in outcome measures for lifestyle modification alone group were similarly distributed, further supporting the transitivity assumption. No inconsistency between direct and indirect estimates was found for the change in BMI from baseline, as assessed by the node‐splitting method (Supporting Information: Appendix 6.7). For the primary outcome, each comparison included fewer than 10 studies. Therefore, Egger's, Begg‐Mazumdar, and Thompson‐Sharp tests were not performed, as their statistical power would be low and the results unreliable with such small sample sizes. Although the funnel plots appeared symmetrical, visual assessments can be influenced by between‐study heterogeneity. For the secondary outcomes, all three tests were conducted and yielded p values > 0.05, indicating no significant evidence of publication bias (see Supporting Information: Appendix 6.8).

3.8. Meta‐Regression and Sensitivity Analyses

Meta‐regression results are detailed in Supporting Information: Appendix 9. Follow‐up duration was strongly associated with the occurrence of diarrhoea events (coefficient: 0.0224, 95% CI: 0.0112 to 0.0335, p < 0.0001) and was also identified as a significant moderator of nausea (coefficient: 0.0277, 95% CI: 0.0078 to 0.0476, p = 0.0064). Longer follow‐up duration increases the cumulative risk of such adverse events, complicating comparisons across trials with different durations, as studies with shorter follow‐up may underestimate the burden of adverse events. This highlights the importance of long‐term safety data when evaluating the benefit–risk profile of anti‐obesity medications in children. By contrast, no significant associations were observed between any other outcomes and age, baseline BMI or body weight, gender, or pharmacotherapy duration. Sensitivity analyses demonstrated that the direction and magnitude of mean difference and odds ratio estimates were consistent with the primary analyses, supporting the robustness of the findings (Supporting Information: Appendix 10).

4. Discussion

Based on 41 trials involving 3923 participants, a frequentist network meta‐analysis was conducted to evaluate the efficacy and safety of pharmacological interventions for the treatment of overweight or obesity in children and adolescents. Phentermine‐topiramate, semaglutide, and liraglutide outperformed lifestyle modification alone in achieving clinically meaningful BMI reductions (≥ 5% and ≥ 10%) and in reducing the 95th BMI percentile percentage. An additional assessment of absolute BMI change (32 studies, 3257 participants) produced aligned results, supporting these conclusions (Supporting Information: Appendix 6.2). Regarding safety, semaglutide and liraglutide carried a higher burden of gastrointestinal side effects than lifestyle intervention, while phentermine‐topiramate was not associated with increased gastrointestinal adverse events or discontinuations due to adverse events.

However, some degree of heterogeneity was observed for the change from baseline in BMI when comparing orlistat, metformin, and liraglutide with lifestyle modification alone; similarly, heterogeneity was also noted for gastrointestinal adverse events when comparing orlistat with lifestyle modification alone. The relatively small trial samples may have constrained the ability to detect meaningful between‐group differences. Several trials involving orlistat, metformin, and liraglutide also raised concerns regarding randomization, deviations from intended interventions, and missing outcome data, which could further explain the observed heterogeneity. Despite the presence of heterogeneity, assessment of the transitivity assumption yielded reasonable results in this network meta‐analysis. To ensure the reliability of the findings, the certainty of evidence was downgraded in the GRADE assessments for outcomes affected by heterogeneity and risk of bias. Sensitivity analyses that excluded trials from different perspectives further confirmed the robustness of the results.

Phentermine‐topiramate is a novel combination agent that works through complementary pathways to support weight management. Phentermine reduces appetite through noradrenaline reuptake inhibition [34], while topiramate is thought to reduce caloric intake by enhancing the activity of GABA (γ‐aminobutyric acid) neurotransmitter and inhibiting carbonic anhydrase [35]. When used in combination at lower doses (e.g., phentermine 7.5 mg/topiramate 46 mg), the two agents exert synergistic effects via complementary central pathways, leading to greater weight reduction than either drug alone at equivalent or higher doses [36]. This synergy allows for reduced exposure to individual agents, potentially mitigating dose‐dependent adverse effects and improving tolerability [37, 38]. Phentermine‐topiramate was initially approved by the U.S. Food and Drug Administration (FDA) in July 2012 for long‐term weight management in adults. On July 20, 2022, the FDA granted its first approval for the treatment of obesity for paediatric patients aged ≥ 12 years [39, 40], based primarily on a 56‐week double‐blind, placebo‐controlled trial [41], which was included in our study. Prior meta‐analyses have seldom included phentermine‐topiramate, but our analysis demonstrated that this agent provides meaningful weight‐management benefits. A plausible reason for its limited recognition is that it lacks authorization from major agencies such as the European Medicines Agency (EMA), largely due to unanswered questions surrounding its long‐term cardiovascular and central nervous system safety. Both phentermine and topiramate carry central nervous system‐related psychiatric risks, including worsening of depression and anxiety [37, 42]. Cardiovascular concerns arise mainly from phentermine‐induced heart rate elevation. Adult trials showed mild, dose‐related increases in mean heart rate [37, 43], but no such differences were observed in adolescents aged 12–17 years [41]. Although no serious cardiovascular events have been reported, all studies had short follow‐up periods. Long‐term cardiovascular safety data in children are lacking, and addressing this gap is essential for clinical decision‐making.

GLP‐1 receptor agonists, which stimulate postprandial insulin secretion and inhibit glucagon secretion in a glucose‐dependent manner, were originally used to treat type 2 diabetes mellitus (T2DM) [44]. Subsequent studies demonstrated their beneficial effects on weight reduction [45, 46, 47, 48]. To date, two GLP‐1 receptor agonists, liraglutide and semaglutide, have been approved by the FDA and EMA for weight management in adolescents [49, 50, 51]. Their mechanism involves targeting receptors in the hypothalamus to suppress appetite, controlling body weight through both a direct effect of reducing food intake and an indirect effect of slowing gastric emptying [52]. However, the long‐term effects of GLP‐1 receptor agonists on nutritional status and growth in children remain unclear. Sustained appetite suppression may reduce the intake of essential nutrients, which could in turn impair bone mineralization, pubertal development, and final adult height. Chronic delayed gastric emptying may also alter the absorption patterns of nutrients. Although existing paediatric trials have not reported significant growth retardation over short‐term follow‐up [53], long‐term studies are warranted to evaluate their impact on growth trajectories, particularly among younger children with remaining growth potential. Furthermore, the developing hypothalamic–pituitary axis may exhibit distinct sensitivity to centrally acting appetite inhibitors. Drug absorption, distribution, metabolism, and excretion vary with age, body composition, and organ maturation, yet these parameters are rarely fully characterized in paediatric trials. Further pharmacodynamic and long‐term safety studies specifically designed for this population are therefore essential.

When appraising effectiveness, a simultaneous and balanced consideration of safety is essential. A key consideration in youth is gastrointestinal adverse events, which may influence nutritional status and growth trajectories. Our findings revealed a distinct efficacy benefit for phentermine‐topiramate versus lifestyle modification, with no marked disadvantage in safety. Based on this integrated evaluation, phentermine‐topiramate represents a more favourable option for weight management in children and adolescents. However, it should be noted that the confidence intervals were relatively wide, primarily attributable to limited sample sizes. Randomized controlled trials of pharmacotherapies for weight management remain scarce in paediatric and adolescent populations. Current evidence is largely derived from short‐ to intermediate‐term studies, and long‐term efficacy, safety, and post‐discontinuation outcomes remain unclear. Although our findings may provide clinical reference, caution is warranted in long‐term clinical practice.

The cost of pharmacotherapy represents an important practical consideration. Novel anti‐obesity agents, particularly GLP‐1 receptor agonists such as liraglutide and semaglutide, are considerably more expensive than lifestyle intervention alone. Previous pharmacoeconomic studies have shown that lifestyle intervention alone is optimal for short‐term treatment (≤ 2 years) [54], whereas phentermine‐topiramate combined with lifestyle intervention is more cost‐effective in the long term [54, 55]. Although semaglutide yields the highest QALY gains, it incurs considerable monthly costs.

To our knowledge, this network meta‐analysis provides the most comprehensive evidence base to date on pharmacological treatment in children and adolescents with overweight or obesity. Previous studies have primarily assessed the efficacy of pharmacotherapies through changes in BMI and BMI z‐score [56]. In the present study, BMI change was analysed as a secondary outcome. The estimated effect sizes for phentermine‐topiramate (−4.83 kg/m²) and semaglutide (−5.90 kg/m²) were generally consistent with those reported in previous studies [56, 57, 58]. Compared with the most comparable meta‐analysis [56], our study incorporated the latest RCT [59] evidence published up to July 2025. Furthermore, by not excluding participants with weight‐related comorbidities, we included a larger number of liraglutide trials. In addition to BMI reduction, the present study further focused on more clinically meaningful efficacy endpoints, including the proportion of participants achieving ≥ 5% and ≥ 10% BMI reduction, as well as the change in the 95th BMI percentile. By integrating these efficacy outcomes with safety data, this analysis aimed to provide practical guidance for clinical practice and to contribute more comprehensive evidence for future guideline updates. We employed a comprehensive set of statistical methods to evaluate each outcome, incorporating assessments of heterogeneity, inconsistency, and transitivity. Multi‐faceted sensitivity analyses and GRADE assessments further confirmed the robustness of our results. To enhance clinical interpretability, we also estimated absolute effects and classified interventions according to predefined minimal important differences.

We acknowledge several limitations of this review. First, for the proportions of participants achieving ≥ 5% and ≥ 10% BMI reduction, the confidence intervals around the effect estimates for phentermine‐topiramate, semaglutide, and liraglutide were wide, likely due to the small sample sizes of the included trials. This imprecision introduced greater uncertainty into the effect estimates, thereby reducing the certainty of the evidence and the strength of the associated clinical recommendations. In addition, the available data were insufficient to conduct dose–response analyses for the various GLP‐1 receptor agonists or phentermine‐topiramate. Future paediatric trials should focus on high‐quality head‐to‐head comparisons and dose–response differences. Second, our analysis focused exclusively on gastrointestinal adverse events, as there were insufficient data to examine adverse effects in other organ systems. Third, the MID for paediatric populations likely varies by age due to differences in ideal body size during growth. Applying a single MID across a broad age range such as 8.1 to 16.1 years could lead to overestimation or underestimation of treatment effects in certain age subgroups. Ideally, age‐specific MID values would be explored, but this was not feasible in the current study. Furthermore, MID values may also differ across disease states. Although we tested the robustness of our findings through sensitivity analyses, external validation across multiple age groups and different geographic populations remains lacking. Fourth, disease status heterogeneity may have influenced the findings. While our study did not exclude participants with type 2 diabetes or metabolic syndrome, nearly half of the included trials explicitly did so, limiting the generalizability of our results to this comorbid population. Fifth, data from weight‐loss trials in younger children (ages 5–8) were lacking (minimum mean age: 8.1 years), limiting applicability to early‐onset obesity. The included trials were geographically dominated by North America and Europe, with relatively few studies conducted in Asia and other continents. Regarding racial distribution, approximately 70% of participants were White, suggesting underrepresentation of other racial and ethnic groups. Sixth, none of the included studies reported household socioeconomic status, and only a few documented parental educational level. Accordingly, it remains unclear whether our findings are generalizable to different socioeconomic strata. Seventh, the number of interventions included varied across different BMI‐related outcomes, leading to incomplete connectivity in some evidence networks and potentially reducing the reliability of indirect comparisons. Moreover, owing to the current lack of RCTs between interventions, the comparisons of efficacy and safety in this study were primarily based on indirect evidence, and the results may therefore differ from those of direct comparisons. Future RCTs may focus on head‐to‐head comparisons among different interventions. Finally, publication bias assessment in network meta‐analysis is inherently constrained by multiple comparisons and low statistical power when few studies are available per comparison. Paediatric obesity trials may be susceptible to selective reporting of favourable outcomes, and small‐scale studies often lack representativeness.

In conclusion, this network meta‐analysis summarizes the current evidence on pharmacological interventions for overweight and obesity in children and adolescents, providing information for clinical decision‐making. Phentermine‐topiramate showed the largest BMI reduction in this network meta‐analysis, although confidence intervals were wide and long‐term safety data are limited.

Author Contributions

L.L., T.H. and Z.Y. contributed equally. S.Z., L.L., T.H. and Z.Y. conceived and designed the study. L.L., T.H. and H.W. screened articles, L.L., Z.Y. and Y.Q. extracted the data. Q.W., Z.Y., N.L., C.X. assessed ROB. L.L., T.H., Z.Y., Y.Q., Y.X., N.L. and Y.C. contributed to the statistical analysis. S.Z. obtained funding for this study and supervised the work. L.L., T.H., R.H. and J.Z. rated the certainty of evidence. L.L., T.H., Y.Q. and Z.Y. drafted the manuscript. T.H., L.L., J.Z., H.W. and S.Z. contributed to the revision and discussed the final edition. All authors read and approved the final manuscript.

Funding

This study is sponsored by The Second Group of Tianshan Talent Training Program (Grant Number 2023TSYCQNTJ0029), and Central Hospital of Karamay.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Appendix 1. PRISMA checklist.

Appendix 2. Search strategy.

Appendix 3. Supplementary of methods.

Appendix 4. Characteristics of the included studies.

Appendix 5. Risk of bias assessments.

Appendix 6. Other outcomes.

Appendix 7. Contribution matrices.

Appendix 8. GRADE assessments.

Appendix 9. Meta‐regression analyses.

Appendix 10. Results of sensitivity analyses.

Appendix 11. Sensitivity analysis of baseline risk.

Appendix 12. P‐scores for each outcomes.

Appendix 13. The number and types of serious adverse events.

Appendix 14. Summary of previous meta‐analyses.

Appendix 15. Reference list for included studies.

Luo L., Huang T., Yan Z., et al., “Pharmacotherapy for Children and Adolescents With Overweight or Obesity: A Systematic Review and Network Meta‐Analysis of Randomized Controlled Trials,” Diabetes, Obesity and Metabolism 28, no. 9 (2026): 8145–8156, 10.1111/dom.70986.

Data Availability Statement

The data can be obtained by contacting the corresponding author.