This review was conducted following established systematic review methodology for exercise interventions and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement (Impellizzeri and Bizzini, 2012; Page et al., 2021). The protocol was prospectively registered in the PROSPERO database (ID: CRD420250655540).

Searches were conducted in PubMed (MEDLINE), EMBASE, the Cochrane Library, Web of Science, CNKI, and EBSCO. The search was last updated on 11 August 2025, and no date restrictions were applied during database searching. Medical Subject Headings (MeSH) terms utilized included “high-intensity interval training,” “body composition,” “adolescents,” and “children,” along with their respective synonyms. Specifically, these MeSH terms were combined with free-text keywords related to high-intensity interval training (e.g., “high-intensity interval training,” “high-intensity intermittent exercise,” “sprint interval training,” “low-volume high-intensity interval training,” “LV-HIIT,” “HIIT,” “HIIE”) and body-composition or cardiometabolic indices (e.g., “BMI”, “waist circumference”, “hip circumference”, “waist-to-hip ratio”, “body-fat percentage”, “lean body mass”, “blood pressure”, “maximal oxygen uptake”, “physical fitness”, “CRF”, “peak VO₂”, “metabolic-syndrome Z-score”). Inclusion was limited to English-language randomized controlled trials (RCTs) that provided pre- and post-intervention data on body composition or cardiorespiratory variables. We manually searched the reference lists of relevant papers and previous reviews to identify any additional eligible studies. All retrieved records were screened against predetermined inclusion and exclusion criteria.

Eligibility criteria were predefined based on the PICOS framework (Population, Intervention, Comparison, Outcomes, and Study design). To maximize the inclusion of all potentially relevant studies, no time restriction was imposed during the literature search. Subsequent rigorous screening based on predefined eligibility criteria ensured the relevance and methodological quality of the included studies, thereby strengthening the comprehensiveness and validity of the findings. (1) population: According to the WHO 2007 Growth Reference for children and adolescents aged 5–19 years (de Onis et al., 2007), this review included children (5–12 years) and adolescents (13–19 years) with no professional training background (i.e., no structured training for ≥3 months before enrollment) and free of musculoskeletal disorders or other clinical contraindications to exercise (Cole et al., 2000); (2) intervention: LV-HIIT using aerobic modes (e.g., walking, running, cycling), prescribed at a minimum of vigorous intensity (≥77% of maximum heart rate (HRmax) or rating of perceived exertion (RPE) ≥14 on the 6–20 scale) and accumulating <15 min of total time in active high-intensity intervals per session (recovery not counted) (Garber et al., 2011; Sultana et al., 2019; Taylor et al., 2019; Weston K. S. et al., 2014). This threshold was selected to standardize the duration of the true physiological stimulus while ensuring the protocols remain feasible within time-constrained settings, such as school physical education classes, with warm-up and cool-down permitted; MICT was defined as exercising at 64%–76% HRmax, or 46%–63% of maximal oxygen uptake (VO₂max), or 40%–59% of heart rate reserve (HRR), or with RPE of 12–13 (Williams et al., 2019). (3) comparison: a no-exercise control or a MICT comparator using aerobic modalities; (4) outcomes: availability of pre- and post-intervention quantitative data for at least one prespecified endpoint related to body composition (e.g., body mass index, fat mass, body-fat percentage, waist circumference, hip circumference, waist-to-hip ratio) or cardiorespiratory health (e.g., systolic blood pressure, diastolic blood pressure, VO₂ max, CRF, metabolic-syndrome Z-score); and (5) study design: parallel-group RCTs (including randomized crossover trials with pre–post comparisons), conducted in humans, published as full-length articles in English, and implementing supervised exercise interventions (defined as sessions conducted under the direct observation of researchers, physical education teachers, or qualified instructors to ensure protocol fidelity) lasting at least 4 weeks with a minimum frequency of two sessions per week.

Studies were excluded if they: combined LV-HIIT with other training modalities or dietary interventions such that the isolated effect of LV-HIIT could not be determined; used non-aerobic primary modalities; were non-randomized, single-arm, acute, or unsupervised interventions; enrolled adults or normal-weight youths when data were inseparable; lacked sufficient statistics to calculate effect sizes; or were non-English, conference abstracts, protocols, reviews, animal, or in vitro studies.

According to the PICOS criteria, one researcher (CY) independently used EndNote 21 software to remove duplicate records. The deduplicated literature was then exported and provided to two independent researchers (HF and XZ), who screened the titles and abstracts based on the predefined inclusion and exclusion criteria. In case of disagreement, a third independent researcher (ZZ) reviewed the articles to determine their inclusion status. Any discrepancies were subsequently resolved through discussion in accordance with the established protocol. Additionally, potential sources of relevant articles included references from previous systematic reviews on the topic and the professional knowledge of the research team. These sources helped identify articles that might meet the inclusion criteria of this review but were not initially captured in the literature search.

Two reviewers (HF and CY) independently extracted information from each eligible study, recording details such as the first author, publication year, participant profile, study design, and training regimen. We also collected data on adherence, dropout rates, and any reported adverse events. A third reviewer (XZ) verified the extracted data to ensure its accuracy and completeness. All discrepancies between reviewers were resolved through discussion to achieve consensus. The extracted details included: (1) publication information (first author, year, country, and study setting); (2) participant characteristics (sample size, sex, age, and weight status); (3) intervention characteristics (type of exercise, program duration, training frequency, session duration, and other relevant details); and (4) outcome data (mean ± SD) for both the experimental group (EG) and control group (CG) before and after the intervention. If data were missing or presented only graphically, the authors were contacted to request the necessary information. Corresponding authors of three studies were contacted to request missing information; two of them responded and provided usable additional data. Studies for which missing data could not be obtained were excluded from the final analysis.

Two reviewers (HF and CY) independently assessed the risk of bias. Any disagreements were resolved through discussion, and if consensus could not be reached, a third reviewer (XZ) acted as an arbitrator. The risk of bias was assessed using the Cochrane Collaboration’s Risk of Bias Tool 2 (RoB2) (Sterne et al., 2016), which evaluates random sequence generation, random allocation concealment, blinding of outcome assessment, incomplete outcome data, and selective outcome reporting. Additionally, the Physiotherapy Evidence Database (PEDro) scale was used to assess the risk of bias and methodological quality of the included studies (de Morton, 2009). The PEDro scale scores studies on a scale of 0–10, with scores ≥6 considered high quality, 4–5 considered moderate quality, and ≤3 considered low quality.

The risk of bias was considered in the interpretation of the results by applying the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) methodology, which rates the certainty of evidence as “high” (further research is improbable to change our confidence in the estimate of effect), “moderate” (further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate), “low” (further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate) or “very low” (any estimate of effect is very uncertain) (Schünemann et al., 2019). GRADE assessments were completed by one reviewer (CY) and reviewed by a second reviewer (ZZ).

The flow diagram illustrating the trial selection process is depicted in Figure 1. The initial literature search yielded 2,185 potentially eligible articles; after removing 740 duplicates, 1,445 records remained for further screening. Following title and abstract screening, 1,408 articles were excluded, resulting in 37 studies selected for full-text assessment. Ultimately, 12 trials fulfilled the predefined eligibility criteria and were included in the final meta-analysis.

Participant characteristics across the included studies are summarized in Table 1. A comprehensive description of the intervention protocols is presented in Table 2. The studies were published between 2018 and 2025, with seven conducted in Asia, four in Africa, and one in Australia and Europe. The included trials involved a total of 609 overweight or obese participants, whose mean ages ranged from 7 to 16 years. Four studies exclusively recruited girls, four exclusively recruited boys, two included both genders, and the two study did not report the gender distribution of participants (Cao et al., 2022a; Li et al., 2023). Baseline demographic data included BMI (reported in all 12 trials) and body fat percentage (reported in 11 trials; not reported in Su et al., 2024). Ten studies compared the effectiveness of LV-HIIT and MICT, two compared LV-HIIT with non-exercising control groups, and five studies incorporated both MICT and no-exercise control conditions. Intervention durations ranged from 8 to 12 weeks, with most lasting 12 weeks (n = 8). Exercise sessions were performed two to three times weekly. Exercise modalities consisted primarily of running-based or football-based drills. In LV-HIIT and MICT protocols, training intensity was prescribed based on HRmax, HRpeak, HRR, VO₂max, VO₂peak, or Maximum Aerobic Speed (MAS). LV-HIIT sessions involved 4–12 min of high-intensity exercise, preceded by a 5–18-min warm-up at 80%–100% HRmax and followed by a 5–15-min cool-down. Participants in the MICT groups exercised for 12–40 min per session, with each session preceded by a 5–15-min warm-up at 55%–80% HRmax and concluded with a 5–15-min cool-down. Regarding outcome assessments, BMI was derived from height and weight measurements; height was measured using calibrated stadiometers across all included studies, while weight was assessed using calibrated electronic scales in two studies, with the remaining studies utilizing diagnostic scales or body composition analyzers. Body fat percentage was evaluated by bioelectrical impedance analysis (BIA) in six studies, dual-energy X-ray absorptiometry (DXA) in four studies, and skinfold thickness in one study. Cardiorespiratory fitness (VO₂max) was determined using direct gas analysis systems in four studies, while three studies utilized indirect field-based prediction equations (e.g., 20-m shuttle run test) (Leger et al., 1988). Resting blood pressure was monitored using automatic electronic devices in six studies and manual sphygmomanometers in two studies. Detailed device specifications, measurement protocols, and reliability data for each study are provided in Supplementary Table S4.

The PEDro scores ranged from moderate to high quality (5–9) for the systematic review and meta-analysis. Table 3 provides a detailed summary of the methodological quality assessment, including individual PEDro scores for each study. However, none of the included trials reported test–retest reliability statistics (e.g., intraclass correlation coefficients) for the primary cardiometabolic outcomes such as BMI, body fat, blood pressure, or cardiorespiratory fitness. Most studies described measurement procedures in detail and, in some cases, cited external validation studies for the devices used, but within-study reliability estimates were not provided.

Ten trials including 298 participants assessed the effects of LV-HIIT versus non-exercising control on body composition, CRF and blood pressure (Abassi et al., 2023; Abassi et al., 2025; Abassi et al., 2022; Cao et al., 2022a; 2022b; 2022c; Cao et al., 2024; Cvetkovic et al., 2018; Li et al., 2023; Paahoo et al., 2020). These studies assessed outcomes including body composition, CRF, and cardiometabolic markers. Regarding feasibility, exercise adherence was reported in four studies for the LV-HIIT groups (91%–100%) (Cao et al., 2022a; 2022b; Li et al., 2023; Paahoo et al., 2020). Retention was also high across the included trials, with dropout rates ranging from 0% to 28%. Notably, the majority of studies (8 out of 10) reported a dropout rate of 20% or lower, and no exercise-related adverse events were reported, collectively supporting the feasibility of LV-HIIT for adolescents with obesity.

Ten trials involving a total of 302 participants were included in the direct comparison of LV-HIIT versus MICT (Abassi et al., 2023; Abassi et al., 2022; Cao et al., 2022a; 2022b; Cao et al., 2024; Cvetkovic et al., 2018; Li et al., 2023; Paahoo et al., 2020; Su et al., 2024; Zuo et al., 2023). These studies assessed outcomes including body composition, CRF, and cardiometabolic markers. Regarding feasibility, retention was high across the included trials, with dropout rates ranging from 0% to 29%. Notably, the vast majority of studies (9 out of 10) reported a dropout rate of 20% or lower, and no exercise-related adverse events were reported in either group. Furthermore, exercise adherence was explicitly reported in six studies (Cao, Tang, et al., 2022a; 2022b; Li et al., 2023; Paahoo et al., 2020; Su et al., 2024; Zuo et al., 2023), with comparable high rates observed for both LV-HIIT (91%–100%) and MICT (84%–97%). These findings suggest that LV-HIIT is as feasible and tolerable as traditional moderate-intensity training for adolescents with obesity.

The risk of bias for each study is depicted in Supplementary Figure S2. Overall, the evidence quality for selective reporting and incomplete outcome data was relatively strong: approximately 70%–80% of the studies were assessed as having low risk of bias, indicating that the reporting of primary outcomes and the maintenance of data integrity were generally adequate. Allocation concealment was less consistently implemented, with roughly half or fewer rated as low risk, and the remainder mostly rated as unclear risk. Regarding blinding, participant/investigator and outcome assessor blinding were rated low risk in approximately half of the studies. At the same time, the rest were categorized as unclear risk, suggesting that caution is warranted when interpreting subjective measures or outcomes susceptible to expectation effects. For random sequence generation, most studies were rated as low risk. However, a notable proportion were designated as unclear due to insufficient detail regarding the specific methods or tools used to generate the random sequences and how they were implemented. Other sources of bias were predominantly rated as low risk. At the study level, the majority of trials demonstrated low or unclear risk of bias across all domains; only a few studies (e.g., Li et al., 2023) were assessed as having a high risk of bias in the “incomplete outcome data” domain, likely due to inadequate handling of dropouts or withdrawals, or deviations from the intended protocol.

The risk of publication bias was investigated using funnel plots combined with Egger’s test for the effects of included studies on body composition and cardiovascular health (Supplementary Figure S3). When ten or more studies were available, funnel plots combined with Egger’s regression test were used to quantitatively assess publication bias. In general, funnel plots showed symmetrical distributions of effect-size points, and Egger’s p-values predominantly exceeded 0.05, indicating no significant publication bias in analyses involving waist circumference, body fat percentage, diastolic blood pressure, or overall LV-HIIT vs. MICT comparisons. It should be noted that the funnel plots for BMI and VO₂max comparisons against the control group showed an asymmetric aggregation pattern. This was corroborated by significant Begg’s and Egger’s test results, suggesting that small-study effects may have inflated these effect sizes. Due to the insufficient number of available studies, a reliable evaluation of publication bias was not feasible for fat mass or systolic blood pressure outcomes. In summary, systematic publication bias was not detected for most outcomes; however, the observed effects on BMI and VO₂max should be interpreted cautiously and require validation through future pre-registered trials with larger sample sizes.

This study integrated the RoB 2 and PEDro tool evaluations to provide a more comprehensive assessment of the methodological quality and reliability of the included studies. The RoB 2 tool focuses on bias control and internal validity, while the PEDro tool emphasizes the structural quality of trial reports. Therefore, it is not uncommon for the scoring results to differ when using these two tools for assessment.

To comprehensively assess the certainty of evidence for each primary outcome indicator, we used the GRADE approach to evaluate the overall quality of the evidence (Supplementary Table S4). In the comparison between LV-HIIT and no training, outcomes such as body fat percentage, blood pressure, and WC were rated as moderate certainty, indicating a reasonable confidence level in the estimated effects. However, specific outcomes, including BMI and VO₂max, were downgraded to low certainty due to inconsistent findings across studies or small sample sizes, leading to wide confidence intervals. In the comparison between LV-HIIT and MICT, VO₂max was assigned a moderate certainty rating, while four outcomes, including BMI and blood pressure, were rated as low certainty primarily due to imprecision. Notably, WC was rated as very low certainty, owing to both limited sample sizes and serious imprecision. These GRADE assessments clarify the robustness of the current evidence and underscore the need for more rigorous, high-quality research in this area.

To assess the robustness of the study results, we applied a two-level random-effects model and the leave-one-out method to evaluate the influence of individual studies on the overall effect estimates. Each study was sequentially removed in the sensitivity analysis, and the pooled effect size was recalculated to determine whether the findings were materially affected. We further examined the impact of studies judged to have a high risk of bias or small sample sizes, based on PEDro scores, RoB 2 assessments, and the characteristics of the included studies. Excluding most of these studies did not substantially alter the combined effect size or its confidence interval. Notably, in the comparisons between LV-HIIT and the no-exercise control group and MICT, the leave-one-out analysis revealed that no single study significantly influenced the overall effect estimates.

In additional sensitivity analyses, we varied the assumed pre–post correlation coefficient used to impute change-score SDs (r = 0.30, 0.50, 0.70) and re-ran all primary random-effects meta-analyses (Supplementary Table S3). For most outcomes, the magnitude and direction of the pooled SMDs, as well as their statistical significance, were largely unchanged across different r values, indicating that the main findings were robust to this assumption. For waist circumference and SBP in the LV-HIIT versus MICT comparison, the point estimates consistently favoured LV-HIIT, but the 95% confidence intervals moved marginally across the line of no effect under different r assumptions, suggesting that these small between-group differences are statistically borderline and should be interpreted with caution. Overall, the convergence of the leave-one-out and correlation-based sensitivity analyses supports the robustness of the primary conclusions of this meta-analysis.

HIIT is a time-efficient training strategy that rapidly improves physiological function and exercise performance, achieving enhancements comparable to those of traditional endurance training (Gibala et al., 2006). An increasing body of high-quality evidence indicates that HIIT significantly improves cardiorespiratory health in youth populations, especially among obese adolescents (Cao et al., 2019; Ortega et al., 2011). These improvements are equivalent to traditional aerobic training but require less time investment, offering greater efficiency (Corte de Araujo et al., 2012). From a physiological perspective, the brief vigorous bouts used in LV-HIIT fall within the high-intensity domain defined in current exercise physiology terminology, providing a substantial cardiometabolic load despite their short duration (Chamari and Padulo, 2015). Similar cardiorespiratory benefits have been consistently reported among adult populations (Wang et al., 2024). These observations align closely with the results of our meta-analysis. Specifically, our analysis revealed substantial improvements in VO₂max among obese adolescents following LV-HIIT interventions. Martin-Smith et al. reported moderate improvements in adolescent CRF following HIIT (Men et al., 2023). Similarly, Martin-Smith et al. documented significant increases in VO₂max following HIIT compared to inactivity (Martin-Smith et al., 2020). These outcome variations may be attributed to differences in intervention protocols and participant characteristics. Including highly intense, low-volume sessions in our analysis might have contributed disproportionately larger improvements among less-fit obese adolescents. Additionally, the relatively small number of included studies might have inflated effect sizes, introducing uncertainty into these findings.

In discussions concerning CRF enhancement, HIIT and MICT have consistently been central themes (Milanović et al., 2015). It is noteworthy that this prior meta-analysis predominantly included higher-volume HIIT protocols (typically >20 min per session) and healthy adult populations, whereas our study specifically isolated the effects of time-efficient LV-HIIT in adolescents with obesity. Our findings further substantiate the effectiveness of low-volume, time-efficient HIIT in enhancing CRF. Specifically, our results indicate that LV-HIIT may be superior to traditional aerobic training methods for improving CRF. Russomando et al. also reported significant increases in VO₂max in the HIIT group compared with MICT (Russomando et al., 2020). The relatively greater magnitude of improvements observed in our meta-analysis suggests that obese adolescents might respond more sensitively to intense interval stimuli; notably, both higher-volume, high-intensity HIIT and lower-volume, shorter-duration HIIT protocols can yield comparable cardiorespiratory enhancements. This finding holds practical significance: brief HIIT protocols can meaningfully enhance the health-related outcomes of school-based physical education, effectively improving CRF within limited class durations (Liu et al., 2020). Notably, when data from adult studies are pooled, specific investigations have reported no significant differences between HIIT and MICT regarding improvements in VO₂max (Khodadadi et al., 2023). Therefore, our findings confirm that LV-HIIT protocols tailored explicitly for obese adolescents may enhance cardiorespiratory health more effectively than traditional endurance training methods. This potential advantage may depend on population-specific characteristics and specific training protocols; thus, further targeted research is warranted to confirm these observations.

Our comparison with a non-exercise control group demonstrated that a period of LV-HIIT significantly reduced resting blood pressure (BP) among obese adolescents. SBP decreased markedly relative to controls, and DBP fell in parallel. This finding aligns with earlier evidence; a meta-analysis of overweight adolescents reported significant reductions in SBP (SMD = −0.71, 95% CI [−1.33,−0.09], p = 0.02) and DBP (SMD = −0.88, 95% CI [−1.50, −0.25], p < 0.01) compared with no-exercise controls (Tozo et al., 2025). Similar antihypertensive effects have been observed in hypertensive adolescents, where HIIT induced a mean SBP reduction of approximately 12.8 mmHg—substantially greater than that observed in control subjects (Popowczak et al., 2022). This finding is further supported by recent evidence in obese adolescents, where short-term HIIT significantly improved cardiovascular profiles, including reductions in systolic blood pressure and optimization of lipid metabolism (Racil et al., 2023). Notably, the BP-lowering effect of brief high-intensity bouts appears larger in individuals with elevated baseline BP, potentially due to greater vascular tone stimulation followed by a pronounced post-exercise hypotensive response (Li et al., 2022). These mechanisms may account for the marked BP reductions observed with low-volume HIIT in obese youths; high-intensity intervals enhance endothelial function and reduce peripheral resistance, conferring benefits comparable to those of longer, conventional exercise sessions (K et al., 2024).

When LV-HIIT was compared with MICT, both interventions effectively reduced BP in obese adolescents, producing comparable moderate decreases in SBP and DBP. In controlled total exercise volume, LV-HIIT did not compromise the antihypertensive effect. The magnitude of BP reduction was similar to that achieved with MICT. A meta-analysis of RCTs involving 309 obese children found that SBP decreased by about 4 mmHg in the HIIT group and 3 mmHg in the MICT group relative to baseline, with no significant difference between groups. Likewise, both groups exhibited a mean DBP reduction of around three mmHg, with similar degrees of improvement (Liu et al., 2020). Variability among study findings may reflect differences in intervention protocols and participant characteristics, as training intensity, interval structure, and recovery duration can modulate outcomes. Indeed, physiological responses to interval training are strongly dependent on the precise manipulation of work-to-rest ratios and recovery intervals (Padulo et al., 2015). Consistent with this, experimental work has shown that differences in interval configuration and intensity prescription can significantly influence adaptations such as VO₂max and SBP (Cao et al., 2021). Baseline BP status also acts as a key effect modifier. Pre-hypertensive individuals typically demonstrate greater reductions in blood pressure following high-intensity stimuli, whereas normotensive individuals often exhibit responses comparable to those elicited by MICT (Clark et al., 2020). Thus, even at lower exercise volumes, high-intensity intervals remain effective for BP improvement in obese adolescents, underscoring the pivotal role of structured physical activity in reducing cardiovascular risk.

We observed significant reductions in BMI, body fat percentage, and waist circumference in adolescents following the LV-HIIT intervention. HIIT induces pronounced excess post-exercise oxygen consumption (EPOC) and elevates catecholamine secretion, collectively increasing energy expenditure and fat oxidation, thereby promoting lipolysis and fat reduction (Jiang et al., 2024; Kolnes et al., 2021). These physiological responses are dose-dependent. Insufficient total training load may result in negligible fat-loss effects. Indeed, studies employing LV-HIIT protocols of ≤500 MET-min·week⁻¹ in obese adults found no significant differences in fat reduction compared with control groups (Boer and Moss, 2016; Higgins et al., 2015; Macpherson et al., 2011). These findings suggest that achieving meaningful fat-loss benefits requires HIIT to meet specific training intensity and frequency thresholds. Our study defined LV-HIIT as sessions involving less than 15 min of active high-intensity intervals, effectively maintaining an appropriate intensity–dose balance.

LV-HIIT elicited improvements in BMI, fat mass, body fat percentage, and waist circumference that were comparable to those of MICT. Despite a lower total workload, LV-HIIT achieved similar body-composition improvements as MICT, requiring only one-third to one-half of the weekly time commitment. Comparable results have been reported in adult populations, with HIIT and continuous aerobic training leading to significant weight loss, although differences between methods are minimal. Wewege et al. reported that both HIIT and MICT reduced total fat mass by approximately 1–2 kg and decreased waist circumference among overweight or obese adults, with neither approach demonstrating superior efficacy (Wewege et al., 2017). Similarly, studies involving obese adolescents report no significant differences between HIIT and MICT regarding improvements in body fat percentage, BMI, or related anthropometric indices (Liu et al., 2020; Cao et al., 2022b). Such time-efficient yet effective training modalities are especially attractive to students who must manage demanding academic schedules (Yin H. et al., 2025). Some researchers have mentioned that if HIIT volume is excessively low, fat-loss effects may be less pronounced than energy-matched MICT. This suggests that insufficient energy expenditure can limit the cumulative fat-burning efficacy of brief, high-intensity exercise sessions (Dias et al., 2018). However, our meta-regression analysis did not find a significant dose-response relationship between training volume parameters and fat loss (p > 0.05 for all; Supplementary Figure S1), although the small number of studies limits the statistical power to detect such effects. With matched total exercise volumes, HIIT may have a more favorable effect on BMI in young adults than MICT. Nevertheless, outcomes are often influenced by training frequency, session duration, and participant demographics (Song et al., 2024; Zhu et al., 2021). Given that insufficient time is frequently cited as a primary barrier to adopting active lifestyles among obese individuals (Andersen and Jakicic, 2009), our key finding that even reduced-volume HIIT protocols can significantly improve body composition in obese adolescents has important practical implications. The specific exercise modality employed also critically influences outcomes, as weight-bearing aerobic activities like running generally involve greater muscle recruitment and energy expenditure than non-weight-bearing activities such as cycling. Some studies have shown that running-based HIIT protocols reduce body fat more effectively than cycling-based protocols (Khodadadi et al., 2023). Although fat loss between HIIT and MICT may be comparable, HIIT generally confers additional metabolic benefits. Indeed, our LV-HIIT protocol yielded additional improvements in CRF and cardiovascular health indicators beyond those attained with MICT.

Our meta-regression analysis further revealed a marginal moderating effect of baseline obesity status on BMI outcomes. Participants with a lower baseline BMI (<30 kg/m²) achieved a greater reduction in BMI from the LV-HIIT intervention, while those with severe obesity showed a smaller decrease. This trend did not reach statistical significance, indicating that a lower baseline BMI may be associated with a greater BMI reduction following LV-HIIT intervention. From a practical perspective, our findings suggest that LV-HIIT may be particularly beneficial for children and adolescents with relatively mild obesity. This specificity aligns with the call for personalized exercise prescriptions based on initial obesity levels. The boundary effect observed in this regression analysis is not isolated; it is consistent with findings from the subgroup analysis of this study regarding baseline BMI. This finding is consistent with high-quality research on adolescent weight management through exercise interventions, especially HIIT. Costigan et al. pointed out that HIIT can improve health-related fitness in overweight and obese adolescents, and baseline characteristics may influence outcomes (Costigan et al., 2015). Poon et al. emphasized that HIIT has universal benefits in improving cardiopulmonary health in children and adolescents, but individual baseline conditions may have a greater influence (Poon et al., 2024). Therefore, promoting LV-HIIT programs among moderately overweight adolescents may lead to significant improvements in BMI and represent a promising, resource-efficient weight management strategy for this subgroup.

The borderline trend (p = 0.05) suggesting that lower baseline BMI may predict greater BMI reductions should be interpreted with caution. Our analysis was likely underpowered given the limited number of trials, and unmeasured confounders (e.g., diet, pubertal status) may partly explain this pattern. Notably, previous studies have reported opposite findings, associating higher baseline BMI with greater reduction (Imperiali et al., 2025; Su et al., 2025). Given these inconsistencies, this finding should be considered hypothesis-generating rather than conclusive, and confirmation in adequately powered trials or individual participant data meta-analyses is required.

This study has several limitations that warrant further investigation. First, this meta-analysis incorporated only 12 peer-reviewed studies; this relatively small sample may limit the robustness and generalizability of our conclusions. Second, considerable heterogeneity was observed among studies for specific outcomes. Although random-effects models, sensitivity analyses, and subgroup analyses were systematically conducted to identify sources of heterogeneity, underlying confounding variables—such as differences in regional or cultural contexts and experimental designs—may not have been entirely accounted for. Third, measurement reliability for primary outcomes was inadequately reported in the included trials, which may have introduced additional measurement error and potentially attenuated the true intervention effects. Fourth, meta-regression analyses with a limited number of studies have low statistical power to detect true moderator effects and are prone to false negatives. The borderline finding for baseline BMI (p = 0.05) should be interpreted with particular caution given only 10 studies contributed data. While this meets the minimum of 10 studies per covariate typically recommended for meta-regression, the analysis remains susceptible to limited statistical power. Fifth, standard deviations for change scores were imputed assuming a correlation coefficient of 0.5; although sensitivity analyses confirmed the robustness of our results, actual correlations were not available. Moreover, the lack of individual participant data (IPD) limited our ability to explore participant-level moderators. Specifically, although BMI Z-score is the preferred metric for pediatric obesity, most included studies only reported raw BMI values. Consequently, we were unable to conduct analyses based on Z-scores, and the use of raw BMI (with a cut-off of 30 kg/m² for subgrouping) may not fully capture age- and sex-specific adiposity status, particularly in diverse ethnic populations. Additionally, funnel plot asymmetry detected for BMI and VO₂max outcomes suggests possible publication bias favoring positive findings, meaning the true effects may be smaller than estimated. However, alternative explanations including small-study effects and genuine heterogeneity cannot be excluded. Furthermore, we acknowledge that certain other obesity-related indicators, such as metabolic markers (e.g., blood lipids, glycemic traits), were not included in this meta-analysis; future research should aim to evaluate these diverse profiles as more standardized trials become available. Lastly, most studies included in this meta-analysis employed short-term interventions (≤12 weeks), limiting our capacity to evaluate the long-term effectiveness of LV-HIIT on cardiovascular disease risk and trajectories of weight regain.