Optimal exercise modality and dose for reducing visceral adipose tissue in overweight or obese adults: a network meta-analysis and dose-response study Authors: Wenjie Xu, Xudong Xie, Huiping Liu Published: 2026 ABSTRACT AbstractObjectiveThis study aimed to evaluate the effects of different exercise interventions on reducing visceral adipose tissue in overweight or obese adults and to explore the exercise dose-response relationship.MethodsWe searched PubMed, Web of Science, Embase, and the Cochrane Library from inception to February 2026 for randomized controlled trials (RCTs). Risk of bias was assessed using the revised Cochrane risk-of-bias tool for randomized trials (RoB 2.0). Standardized mean differences (SMDs) with 95% confidence intervals (95% CI) were used to estimate the relative effects of different exercise modalities on visceral adipose tissue (VAT), and treatments were ranked using the surface under the cumulative ranking curve (SUCRA). Exercise dose was further quantified as weekly metabolic equivalent of task minutes (MET-min/week), and a Bayesian dose-response model was constructed to evaluate the overall dose and the dose-response relationships for different exercise modalities.ResultsA total of 60 RCTs involving 3308 overweight or obese adults were included. The network meta-analysis showed that, compared with the control group, all four exercise interventions significantly reduced VAT in overweight or obese adults. Based on the SUCRA ranking, high-intensity interval training (HIIT) had the highest cumulative probability (SUCRA = 72.5%), followed by combined aerobic and resistance training (COM) (SUCRA = 69.1%), aerobic exercise (AE) (SUCRA = 59.0%), and resistance training (RT) (SUCRA = 49.5%). Dose-response analysis indicated a nonlinear U-shaped association between total exercise dose and VAT outcomes, with an optimal dose of 1100 MET-min/week. The overall effect reached statistical significance when the dose ranged from 430 to 1420 MET-min/week, and the optimal dose and effective dose intervals differed across exercise modalities.ConclusionAll four exercise interventions reduced VAT in overweight or obese adults to varying degrees, with high-intensity interval training showing the most favorable overall efficacy ranking. A U-shaped association was observed between total exercise dose and VAT in overweight or obese adults, suggesting that a moderate exercise load is beneficial for reducing VAT. These findings provide a theoretical basis for developing individualized and feasible exercise prescriptions for this population.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13102-026-01793-8.Keywords: Exercise intervention, Overweight or obese adults, Visceral adipose tissue, Network meta-analysis, Dose-response Introduction Visceral adipose tissue (VAT) is a highly metabolically active fat depot, and excessive accumulation of VAT is closely associated with insulin resistance, dyslipidaemia, chronic low-grade inflammation, and increased cardiometabolic risk [1–3]. Compared with subcutaneous fat, VAT more readily influences hepatic lipid and glucose metabolism via the portal circulation and shows a stronger independent association with multiple adverse outcomes [4]. Therefore, effectively reducing VAT in overweight or obese individuals not only indicates improvements in body composition but may also confer broader benefits to overall metabolic health. Exercise is widely recommended as a core non-pharmacological strategy for reducing VAT because it is relatively low cost, broadly scalable, and provides additional benefits for cardiorespiratory fitness, muscular fitness, and cardiometabolic health [5]. A growing body of systematic review evidence has supported the beneficial role of exercise in VAT reduction. Vissers et al. systematically reviewed exercise interventions in overweight adults and reported that exercise, even without caloric restriction, could reduce VAT, with moderate- to high-intensity aerobic training showing particular potential [6]. Subsequent reviews further refined this evidence by focusing on specific exercise modalities or dose-related questions. Maillard et al. showed that high-intensity interval training (HIIT) may be an effective and time-efficient approach for reducing total, abdominal, and visceral fat mass in adults with different weight statuses [7]. More recently, Recchia et al. reported dose-response evidence supporting the dose-dependent effect of exercise on visceral adiposity in adults with overweight or obesity [8]. In addition, recent network meta-analytic evidence has begun to compare multiple exercise types and suggested that regular exercise can improve VAT in individuals with overweight or obesity [9]. Despite these advances, several important gaps remain. First, many previous syntheses were based on conventional pairwise comparisons or focused on a single exercise modality, which limits the ability to compare multiple exercise interventions within a unified analytical framework and to generate probabilistic treatment rankings. Although previous network meta-analytic evidence has compared different exercise modalities, some analyses included broader or mixed age populations [9], whereas the present study specifically focused on adults aged ≥ 18 years with overweight or obesity. This focus is clinically relevant because VAT distribution, metabolic risk, and exercise responsiveness may differ across age groups, and adult-specific evidence is needed to inform exercise prescription in this population. Second, exercise dose remains difficult to compare across studies because prescription components, including intensity, frequency, session duration, and intervention period, have been reported and quantified inconsistently. Third, although dose-response evidence has emerged, it remains unclear whether the association between exercise dose and VAT reduction is nonlinear and whether the optimal dose differs across exercise modalities. Therefore, in clinical and public health practice, the key question is not only whether exercise reduces VAT, but also which exercise modality is most effective and what dose range may provide the greatest benefit for adults with overweight or obesity. To address these gaps, the present study combined network meta-analysis with Bayesian dose-response modelling. The network meta-analysis was used to compare, within a unified analytical framework, the relative effects of aerobic exercise (AE), resistance training (RT), combined aerobic and resistance training (COM), and HIIT on VAT in adults aged ≥ 18 years with overweight or obesity and to establish a hierarchy of treatment efficacy. However, treatment ranking alone does not account for the amount of exercise delivered within each modality. Therefore, MET-min/week was adopted as a standardized indicator of exercise dose, and Bayesian dose-response modelling was used to examine whether VAT reduction varied according to exercise dose and whether the optimal dose differed across exercise modalities. By integrating these two approaches, the present study aimed to answer two complementary practice-oriented questions: which exercise modality is most effective and what dose range is likely to provide the greatest benefit. This strategy provides quantitative evidence with direct practical relevance for individualized exercise prescription in adults with overweight or obesity. Materials and methods This systematic review and network meta-analysis was conducted in accordance with the methodological guidelines outlined in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 statement [10] and its extension for network meta-analyses (PRISMA-NMA) [11]. The study protocol was prospectively registered in the International Prospective Register of Systematic Reviews (PROSPERO; registration number: CRD420261328188). Search strategy A comprehensive search strategy was implemented across Web of Science, PubMed, Embase, and the Cochrane Library, covering the period from database inception to February 2026. Only English-language randomized controlled trials were considered for eligibility. Search terms were tailored to the indexing features of each database by combining controlled vocabulary with free-text terms related to exercise interventions, overweight or obesity in adults, visceral adipose tissue, and randomized study design. Detailed search strategies for all databases are provided in Appendix 1. To further improve the completeness of study identification, we also manually screened the reference lists of the included studies and other relevant publications. Eligibility criteria Inclusion criteria The study selection followed the PICOS framework [12] as follows: (P) Participants: Adults (≥18 years) who were overweight or obese, defined as BMI ≥25 kg/m² (overweight) or BMI ≥30 kg/m² (obesity) in European populations, and BMI ≥24 kg/m² (overweight) or BMI ≥28 kg/m² (obesity) in Asian populations. If BMI was not reported, body fat percentage (BF%) was used as the criterion (women BF% ≥30%; men BF% ≥25%). (P) Participants: Adults (≥18 years) who were overweight or obese, defined as BMI ≥25 kg/m² (overweight) or BMI ≥30 kg/m² (obesity) in European populations, and BMI ≥24 kg/m² (overweight) or BMI ≥28 kg/m² (obesity) in Asian populations. If BMI was not reported, body fat percentage (BF%) was used as the criterion (women BF% ≥30%; men BF% ≥25%). (I) Interventions: Exercise modalities including AE, RT, COM, and HIIT (see Appendix 2). (I) Interventions: Exercise modalities including AE, RT, COM, and HIIT (see Appendix 2). (C) Comparators: Health education, no intervention, or an alternative exercise intervention other than that applied in the experimental group. (C) Comparators: Health education, no intervention, or an alternative exercise intervention other than that applied in the experimental group. (O) Outcomes: VAT, including VAT area, volume, mass, or VAT-related indices assessed using computed tomography (CT), magnetic resonance imaging (MRI), dual-energy X-ray absorptiometry (DXA/DEXA), bioelectrical impedance analysis (BIA)-based devices, or validated formula-based indices reported in the original trials. (O) Outcomes: VAT, including VAT area, volume, mass, or VAT-related indices assessed using computed tomography (CT), magnetic resonance imaging (MRI), dual-energy X-ray absorptiometry (DXA/DEXA), bioelectrical impedance analysis (BIA)-based devices, or validated formula-based indices reported in the original trials. (S) Study design: Randomized controlled trials. (S) Study design: Randomized controlled trials. Exclusion criteria Studies were considered ineligible if they were not published in English, lacked an accessible full text, or were available only in the form of conference abstracts or systematic reviews. Trials were also excluded when they did not adopt a randomized controlled design, involved non-exercise interventions, combined exercise with dietary treatment, or used exercise prescriptions that were too complex to permit reliable estimation of dose. In addition, studies were not included in the quantitative synthesis if data on visceral adipose tissue were unavailable and could not be obtained from the study authors. Study selection and data extraction All records retrieved from the four databases were first imported into EndNote 21 for reference management, and duplicate entries were removed before further screening. Study selection was then performed in two stages by applying the predefined PICOS framework. In the initial stage, titles and abstracts were reviewed to exclude clearly irrelevant reports. Articles considered potentially eligible subsequently underwent full-text assessment to determine final inclusion. For each included study, data were collected on publication characteristics, participant information, intervention details, and outcome measures. Publication-related information included the first author, year of publication, and study location. Participant characteristics comprised sample size, age, body mass index (BMI), and body fat percentage (BF%) for each group. Information on the interventions included exercise modality, overall intervention duration, training frequency, and session length. For the quantitative synthesis, change scores from baseline to post-intervention were preferentially used to estimate the intervention effect. When change scores were not directly reported, they were calculated from the available baseline and post-intervention values. When required information was unavailable, the corresponding authors were contacted by email on up to three occasions. Studies for which no response was received within 4 weeks were excluded from the quantitative synthesis. Study screening and data extraction were completed independently by two investigators (WJ-X and XD-X), and any discrepancies were resolved through discussion with a third investigator (HP-L). No AI-assisted tools, large language models, Rayyan, CADIMA, or Python-based automated workflows were used for study screening, eligibility assessment, data extraction, risk-of-bias assessment, or GRADE assessment. Data coding and management To standardize the quantification of exercise exposure across studies, weekly physical activity dose was expressed as MET-min/week in the present study. This metric has been widely adopted in previous research examining dose-response associations between physical activity and a variety of health-related outcomes [13, 14]. In general, it is derived from the combination of three core elements: metabolic intensity in METs, duration of each exercise session, and weekly training frequency. By converting exercise programs that differ in modality, intensity, and prescription characteristics into a common quantitative unit, MET-min/week facilitates cross-study comparison and provides a practical basis for categorizing exercise exposure into graded dose levels. Additional support for this approach is provided by the 2024 Compendium of Adult Physical Activity Metabolic Equivalent Classifications, which offers standardized MET values for a broad range of adult activities and serves as an important reference for the estimation of energy expenditure [15]. Accordingly, MET-min/week was selected as the primary indicator of exercise dose in the present analyses. For the model-based network meta-analysis of dose-response relationships, weekly exercise dose was further classified into seven predefined categories: 0 (CON), 250, 500, 750, 1000, 1250, and 1500 MET-min/week. Risk of bias Risk of bias in the included randomized trials was assessed using the revised Cochrane risk-of-bias tool for randomized trials (RoB 2.0) [16]. Two reviewers (WJ-X and XD-X) independently assessed each eligible trial across five domains: bias arising from the randomization process, bias due to deviations from intended interventions, bias due to missing outcome data, bias in measurement of the outcome, and bias in selection of the reported result. Each domain was judged as low risk of bias, some concerns, or high risk of bias according to the RoB 2 signalling questions and algorithm. An overall risk-of-bias judgment was then assigned for each study. In cases of disagreement, a third reviewer (HP-L) was involved to resolve discrepancies and establish a final consensus. Certainty of evidence The certainty of evidence for all intervention comparisons was independently rated by two reviewers (WJ-X and XD-X) using the GRADE approach adapted for network meta-analysis [17]. Five domains were considered: study limitations, inconsistency, indirectness, imprecision, and publication bias. Any disagreements were resolved by a third reviewer (HP-L). Statistical analysis Network meta-analysis To enable cross-study comparison of VAT measured using different methods and units, intervention effects were expressed as standardized mean differences (SMDs). SMDs were calculated in accordance with Chap. 6.5 of the Cochrane Handbook [18]. For each group, the mean change was calculated as the post-intervention mean minus the baseline mean. When the SD of the change score was reported, it was used directly. When the SD of the change score was not directly reported but baseline and post-intervention SDs were available, the SD of the change score was calculated using the following formula: where r represents the within-participant correlation between baseline and post-intervention values. When SDs were reported as standard errors, confidence intervals, or other convertible statistics, they were converted to SDs according to Cochrane Handbook recommendations [18]. A frequentist random-effects network meta-analysis was conducted in Stata 17.0 to generate pooled estimates with 95% confidence intervals in accordance with PRISMA-NMA recommendations; the evidence structure was visualized using a network plot, where node size reflected the number of participants, connecting lines represented direct comparisons, and line thickness indicated the number of contributing studies [19], while network consistency was evaluated using loop-specific, local, and global methods [20], with the consistency model applied when no significant inconsistency was detected (P > 0.05), and intervention rankings were quantified by the surface under the cumulative ranking curve (SUCRA), with funnel plots used to assess publication bias or small-study effects. Bayesian dose-response meta-analysis After confirming adequate network connectivity, transitivity, and consistency for model-based network meta-analysis (Appendix 9), we used a Bayesian random-effects MBNMA to examine the dose-response association between exercise and VAT in adults with overweight or obesity, expressing effects as SMDs with 95% credible intervals because VAT was measured using different scales and units across studies. Among the four candidate functions, including Emax, restricted cubic splines, quadratic, and non-parametric models, the random-effects quadratic model was selected for the primary analysis based on overall model fit, parsimony, interpretability, deviance information criterion, residual deviance, between-study standard deviation, and the number of model parameters (Appendix 10). Results Study selection The study selection process is summarized in Fig. 1. Database searching yielded 1,329 records, including 412 from Embase, 180 from PubMed, 169 from Web of Science, and 568 from the Cochrane Library. An additional 8 records were identified through manual screening of reference lists. After removal of 439 duplicate entries, 898 records remained for title and abstract screening, of which 742 were excluded. The full texts of 156 articles were subsequently reviewed for eligibility. Among these, 96 articles were excluded because of inappropriate study design (n = 19), absence of a control group (n = 24), or unavailable data (n = 53). Ultimately, 60 studies met the eligibility criteria and were included in the meta-analysis. Flow diagram of the study selection process Characteristics of included studies A total of 60 studies published between 2003 and 2025 were included, comprising 3308 participants. Across studies, the reported age of participants ranged from 19.7 to 70.7 years. The trials were conducted in multiple countries and regions, spanning Asia, Europe, North America, South America, Oceania, and Africa. Exercise interventions included AE, RT, COM, and HIIT. Intervention duration ranged from 7 to 64 weeks, session length from 12 to 75 min, and training frequency from 2 to 6 sessions per week. The precisely calculated weekly exercise dose ranged from 315 to 1728 MET-min/week, and the categorized dose levels ranged from 250 to 1500 MET-min/week. VAT was assessed using CT, MRI, DXA/DEXA, BIA-based body composition analyzers such as InBody and ACCUNIQ BC300, or formula-based indices reported in the original trials. Further details are provided in Appendix 3. Methodological quality of included studies The methodological quality of the 60 included studies was assessed using the Cochrane risk-of-bias tool, version 2.0 (RoB 2.0) [16]. Overall, the risk of bias was predominantly low, with 35 studies rated as low risk, accounting for 58.3%; 19 studies rated as having some concerns, accounting for 31.7%; and 6 studies rated as high risk, accounting for 10.0%. Across specific domains, the randomization process was generally well controlled, with 57 studies rated as low risk (95.0%) and 3 studies rated as having some concerns (5.0%). For deviations from intended interventions, 51 studies were rated as low risk (85.0%) and 9 studies as having some concerns (15.0%). Regarding missing outcome data, 42 studies were rated as low risk (70.0%), 16 studies as having some concerns (26.7%), and 2 studies as high risk (3.3%). For measurement of the outcome, 47 studies were rated as low risk (78.3%), 9 studies as having some concerns (15.0%), and 4 studies as high risk (6.7%). For selection of the reported result, 57 studies were rated as low risk (95.0%) and 3 studies as having some concerns (5.0%). Further details are provided in Appendix 4. Certainty of evidence According to the GRADE assessment, the overall certainty of evidence was predominantly moderate. Specifically, eight comparisons were rated as moderate certainty and two as low certainty, with the most common reasons for downgrading being study limitations and heterogeneity. For the treatment-ranking evidence based on SUCRA, the certainty was rated as moderate, downgraded due to suspected publication bias (Appendix 8). Network meta-analysis results Network plot Figure 2 illustrates the evidence network for comparisons of AE, RT, COM, and HIIT with respect to their effects on VAT in adults with overweight or obesity. Among these interventions, AE contributed the largest body of evidence, whereas RT was represented by the fewest studies. The detailed contribution matrix is provided in Appendix 5, which summarizes the extent to which each direct comparison informed the corresponding network estimates and the network meta-analysis as a whole. Network plot of evidence on the effects of different exercise interventions on visceral adipose tissue in overweight or obese adults. CON = Control, AE = Aerobic exercise, RT = Resistance training, COM = Combined aerobic and resistance training, HIIT = High-intensity interval training Assessment of inconsistency In the present study, the P values from the loop-specific, global, and local inconsistency tests were all > 0.05, indicating good reliability of the network estimates. Therefore, a consistency model was applied in the subsequent analyses (Appendix 6). Pooled effects and treatment ranking The network meta-analysis showed that, compared with the CON, all four exercise interventions significantly reduced VAT in overweight or obese adults to varying degrees. HIIT demonstrated the greatest reduction (SMD = -0.74, 95% CI: -1.01 to -0.47), followed by COM, AE, and RT, as shown in Fig. 3. Network meta-analysis results. CON = Control, AE = Aerobic exercise, RT = Resistance training, COM = Combined aerobic and resistance training, HIIT = High-intensity interval training According to the SUCRA rankings for VAT outcomes in overweight or obese adults, HIIT had the highest cumulative probability of being the best intervention (SUCRA = 72.5%), followed by COM (SUCRA = 69.1%), AE (SUCRA = 59.0%), and RT (SUCRA = 49.5%), as presented in Fig. 4; Table 1. Surface under the cumulative ranking curve (SUCRA) Table 1SUCRA rankings for each intervention node in the networkTreatmentSUCRA(%)HIIT72.5COM69.1AE59.0RT49.5CON0CON Control, AE Aerobic exercise, RT Resistance training, COM Combined aerobic and resistance training, HIIT High-intensity interval training SUCRA rankings for each intervention node in the network CON Control, AE Aerobic exercise, RT Resistance training, COM Combined aerobic and resistance training, HIIT High-intensity interval training Assessment of publication bias or small-study effects Potential small-study effects and publication bias were explored using funnel plots. Overall, the distribution of study-specific estimates was roughly symmetrical around the null line, and most points fell within the corresponding 95% confidence boundaries, indicating no obvious indication of marked asymmetry. Nevertheless, because the number of studies contributing to certain interventions was relatively small, the results of this assessment should be interpreted cautiously (Fig. 5). Funnel plot for assessing publication bias among the included studies. A: Control, B: Aerobic exercise, C: High-intensity interval training, D: Combined aerobic and resistance training, E: Resistance training Bayesian dose-response meta-analysis results Overall dose-response relationship The overall dose-response association between exercise and VAT reduction in overweight or obese adults followed a nonlinear U-shaped curve (Fig. 6). The greatest reduction in VAT was observed at 1100 MET-min/week (SMD = -0.62, 95% CrI: -0.91 to -0.28; SD = 0.16), after which the intervention effect gradually declined as the dose increased. At 430 MET-min/week, the predicted effect became statistically significant (SMD = -0.33, 95% CrI: -0.60 to -0.01; SD = 0.15). When the dose exceeded 1420 MET-min/week (SMD = -0.56, 95% CrI: -1.06 to -0.04; SD = 0.26), the 95% CrI of the predicted effect gradually crossed 0, indicating that the improvement was no longer statistically significant. This dose-response pattern suggests that exercise dose should, where possible, be maintained within 430–1420 MET-min/week and preferably close to 1100 MET-min/week to achieve more favorable and statistically significant effects. Dose-response relationship between total weekly exercise dose and visceral adipose tissue in overweight or obese adults. Notes: The blue dashed lines indicate the beginning and end points of statistical significance, and the blue numbers denote the corresponding dose values (MET-min/week). The grey shaded area represents the effective dose range. The red dashed line and red numbers indicate the location of the optimal dose-response and the corresponding dose value (MET-min/week). WHO: World Health Organization Dose-response relationships across the four exercise modalities Figure 7 depicts the dose-response relationships for the four exercise modalities. The results indicated that all interventions exhibited nonlinear U-shaped dose-response curves. The optimal dose for COM was 570 MET-min/week (SMD = -1.02; 95% CrI: -1.77 to -0.25; SD = 0.39), and the optimal dose for HIIT was 550 MET-min/week (SMD = -0.70; 95% CrI: -1.13 to -0.25; SD = 0.22). For AE, the minimum effective dose threshold was 580 MET-min/week (SMD = -0.38; 95% CrI: -0.73 to -0.01; SD = 0.18). Owing to the limited number of RT trials and the relatively concentrated dose values, we were unable to estimate an optimal dose or an effective dose interval for RT. Dose-response relationships of the four exercise modalities for reducing visceral adipose tissue in overweight or obese adults. Notes: The blue dashed lines indicate the beginning and end points of statistical significance, and the blue numbers denote the corresponding dose values (MET-min/week). The grey shaded area represents the effective dose range. The red dashed line and red numbers indicate the location of the optimal dose-response and the corresponding dose value (MET-min/week). AE = aerobic exercise, RT = resistance training, COM = combined aerobic and resistance training, HIIT = high-intensity interval training Discussion This network meta-analysis examined the effects of four exercise interventions on VAT in overweight or obese adults. The findings showed that all four exercise modalities significantly improved VAT in this population. Using SUCRA-based ranking, HIIT had the highest cumulative probability of being the most effective intervention (SUCRA = 72.5%), followed by COM (SUCRA = 69.1%), AE (SUCRA = 59.0%), and RT (SUCRA = 49.5%). According to the GRADE assessment, the certainty of evidence was predominantly moderate, with downgrading mainly due to study limitations, heterogeneity, and indirectness. We further observed a nonlinear U-shaped dose-response association between total exercise dose and VAT in overweight or obese adults. The optimal dose for improving VAT was 1100 MET-min/week, which is approximately equivalent to 229 min/week of brisk walking (4.8 METs) or 220 min/week of resistance training (5.0 METs). Comparative effects of different exercise modalities and potential mechanisms Our network comparisons showed that all four exercise modalities reduced VAT compared with control conditions. HIIT had the highest SUCRA ranking, followed by COM; however, the differences between active exercise modalities were relatively small, and the confidence intervals for between-intervention comparisons overlapped. Therefore, based on the current evidence, HIIT may be considered a promising exercise option for reducing VAT. This may be attributable to the capacity of high-intensity intermittent stimuli to elicit higher relative exercise intensity and greater metabolic perturbation within a shorter time period [21, 22], thereby increasing energy expenditure per unit time and inducing more marked excess post-exercise oxygen consumption and enhanced fat oxidation [23]. In addition, high-intensity exercise may produce greater improvements in insulin sensitivity and mitochondrial oxidative capacity [24, 25], promote the restoration of metabolic flexibility [26], and facilitate more effective mobilization and redistribution of visceral fat [27, 28]. For individuals who are overweight or obese, HIIT also offers a “time-efficiency” advantage [29], which may reduce the opportunity costs of long-term adherence and, consequently, improve adherence and the sustainability of total training volume in real-world settings [23, 30]. COM also showed a relatively high ranking and consistent benefits in this study, and its advantage is more likely attributable to the complementary effects of “aerobic + resistance” training. Aerobic exercise primarily promotes VAT reduction by increasing energy expenditure and fat oxidation and by improving cardiorespiratory fitness and peripheral metabolic efficiency [31, 32]. In contrast, resistance training may indirectly influence fat distribution and metabolic risk by increasing or preserving lean mass, elevating resting metabolic rate, and improving skeletal muscle glucose uptake and insulin signaling pathways [33–35]. Their combination may not only yield a more favorable structural change in body composition, namely “fat loss without loss of muscle mass,” but also improve post-exercise energy balance and the inflammatory–metabolic milieu [36], thereby supporting a more sustainable pathway for VAT reduction. In addition, COM typically allows for more flexible prescription design, facilitating adjustments based on baseline fitness and joint tolerance and offering good clinical feasibility [37]. AE ranked in the intermediate tier for VAT reduction, with a relatively clear physiological pathway. Regular moderate-intensity aerobic exercise can enhance fat oxidation capacity, improve cardiorespiratory fitness and vascular function, and promote VAT reduction by lowering the inflammatory and metabolic burden associated with visceral fat [37, 38]. Notably, our dose-response findings suggested a minimum effective dose threshold for AE, implying that insufficient training intensity and/or volume in practical prescriptions may fail to induce an adequate energy deficit and metabolic adaptations, thereby limiting the magnitude of VAT improvement. Therefore, for individuals in whom VAT reduction is a primary therapeutic target, AE should aim to achieve a sufficient cumulative dose while ensuring safety and should be accompanied by behavioral support to enhance long-term adherence. RT ranked relatively lower in this study, and an unambiguous optimal dose point could not be consistently identified, which likely reflects limitations in the evidence structure. First, the number of included RT trials was relatively small and the dose distribution was concentrated, limiting the ability of dose–response models to discriminate effects across different load gradients. Second, the effects of RT on VAT may rely more on indirect pathways mediated by gains in muscle mass and improvements in metabolic function [39, 40]. When training intensity, progression, and supervision are inadequate, or when intervention duration is relatively short, sufficient structural and metabolic adaptations may not occur, thereby attenuating the overall effect [41]. Nevertheless, RT has unique value in obesity-related health management by preserving lean mass, improving physical function, and reducing skeletal muscle-related metabolic risk [42, 43]. Future well-designed RCTs with comprehensive reporting of prescription components are still needed to clarify the true effect of RT on VAT and to determine optimal combination strategies with other exercise modalities. Dose-response relationship analysis and recommendations for exercise prescription Bayesian model-based network meta-analysis showed that, in adults with overweight or obesity, the association between exercise dose and VAT reduction followed an overall nonlinear U-shaped pattern, with the greatest estimated benefit at 1100 MET-min/week and statistically significant improvements across 430–1420 MET-min/week, while modality-specific analyses further indicated similar U-shaped dose-response relationships for AE, RT, COM, and HIIT, identifying optimal doses of 570 MET-min/week for COM and 550 MET-min/week for HIIT, a minimum effective threshold of 580 MET-min/week for AE, but no clearly defined optimal or effective dose interval for RT; corresponding exercise recommendations are summarized in Table 2. Table 2Exercise recommendations for reducing visceral adipose tissueExercise modalityOptimal dose(MET-min/week)IntensityExercise intensity(MET)Exercise duration(min/week)Exercise prescription recommendations(sessions × min/per week)HIIT550High8.8 (code 01305)~ 63~ 3 × 20~ 4 × 15COM570Moderate5.0 (code 02035)~ 114~ 4 × 30~ 5 × 20Intensity coding was extracted from the Compendium of Metabolic Equivalent Values for Physical Activities in Adults: code 01305: Bicycling, high intensity interval training; Code 02035: Circuit training, moderate effortCOM Combined aerobic and resistance training, HIIT High-intensity interval training Exercise recommendations for reducing visceral adipose tissue Intensity coding was extracted from the Compendium of Metabolic Equivalent Values for Physical Activities in Adults: code 01305: Bicycling, high intensity interval training; Code 02035: Circuit training, moderate effort COM Combined aerobic and resistance training, HIIT High-intensity interval training Innovations and limitations This study provides several methodological advances over earlier meta-analyses of exercise and VAT reduction in adults with overweight or obesity. Whereas previous syntheses have relied largely on conventional pairwise meta-analysis, we combined network meta-analysis with Bayesian dose-response modelling to evaluate four exercise modalities within a unified analytical framework. This design not only enables a broader comparison of relative efficacy but also incorporates exercise dose, thereby offering more clinically relevant evidence regarding both the optimal exercise modality and the tailoring of training load at the individual level. Several limitations should also be considered. First, clinical and methodological heterogeneity existed across the included studies, with potential sources including VAT measurement method; demographic and adiposity-related characteristics such as age, sex, race/ethnicity, and obesity classification; and intervention-related factors such as intervention duration and MET-min/week. Because the number of studies available for further stratified analyses was limited, and race/ethnicity and obesity-class-specific outcomes were not consistently reported, stratification or covariate adjustment by these factors may have resulted in sparse networks and unstable estimates. Therefore, these variables were not statistically adjusted for in the primary network meta-analysis or SUCRA-based ranking, and caution is warranted when interpreting the pooled estimates and treatment rankings. As more primary studies become available, sufficiently powered subgroup analyses and meta-regression will be needed to clarify whether these factors modify the effects of exercise interventions on VAT. Second, although MET-min/week is a practical standardized metric for exercise dose, it integrates exercise intensity, session duration, and weekly frequency and therefore cannot fully disentangle the independent contribution of each intervention component. In addition, differences in progression, adherence, and supervision were not fully captured. Third, to isolate the independent effect of exercise, studies combining exercise with dietary interventions were excluded. However, in real-world weight management and VAT reduction programs, exercise is often implemented together with dietary modification. Therefore, the exclusion of combined exercise and dietary interventions may limit the external validity and generalizability of the findings to comprehensive lifestyle programs. Conclusion The present study provides clinically relevant evidence for individualized exercise prescription in adults with overweight or obesity by combining network meta-analysis with Bayesian dose-response modelling to determine both the most effective exercise modality for reducing VAT and the most appropriate training load. Although all four exercise modalities were associated with significant reductions in VAT, their effects were not equivalent, with high-intensity interval training demonstrating the greatest benefit. Moreover, the dose-response analysis revealed a nonlinear U-shaped association between exercise volume and VAT reduction, indicating that moderate-to-high training doses may maximize improvement, whereas both inadequate and excessive exercise loads may attenuate the effect. Taken together, these findings offer quantitative support for optimizing exercise strategies aimed at VAT reduction in this population. Supplementary Information Supplementary Material 1. Acknowledgements The authors would like to express their sincere gratitude to all researchers who provided assistance with this study. Clinical trial number Not applicable. Abbreviations Aerobic exercise Combined aerobic and resistance training High-intensity interval training Model-based network meta-analysis PRISMA extension for network meta-analyses Randomized controlled trials Resistance training Standardized mean differences Surface under the cumulative ranking curve Visceral adipose tissue World Health Organization 95% confidence intervals 95% credible intervals Authors' contributions Wenjie Xu: Data curation, Formal analysis, Methodology, Supervision, Writing-original draft, Writing-review & editing. Xudong Xie: Data curation, Formal analysis, Writing-review & editing. Huiping Liu: Formal analysis, Supervision, Validation, Writing-review & editing. Funding Not applicable. Data availability The datasets used and analyzed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.