Impact of aerobic training on body composition profiles among postmenopausal women with overweight or obesity: a systematic review and dose–response meta-analysis

Abstract

Background

Overweight and obesity are prevalent health concerns among postmenopausal women. However, the effects of aerobic exercise training on multidimensional body composition outcomes in postmenopausal women with overweight or obesity, as well as the potential dose–response relationship, remain unclear.

Methods

This systematic review and meta-analysis followed PRISMA guidelines. PubMed, Cochrane Library, Web of Science, EBSCO, and ProQuest were searched from inception to March 8, 2026. Randomized controlled trials comparing aerobic exercise training with non-exercise control in postmenopausal women with overweight or obesity were included. Three-level meta-analyses were performed to account for dependent effect sizes. Restricted cubic spline models, subgroup analyses, and meta-regression were used to explore dose–response patterns and potential moderators.

Results

Sixteen RCTs involving 1,571 participants were included. Compared with non-exercise control, aerobic exercise training reduced body weight (MD = -2.17 kg, P < 0.01; low certainty), body mass index (MD = -0.73 kg/m², P < 0.01; low certainty), body fat percentage (MD = -1.40%, P < 0.01; moderate certainty), fat mass (MD = -1.83 kg, P < 0.01; moderate certainty), waist circumference (MD = -2.02 cm, P < 0.01; moderate certainty), and hip circumference (MD = -1.39 cm, P = 0.04; low certainty). The pooled estimate for lean body mass was positive but imprecise (MD = 0.72 kg, P = 0.14; low certainty). Baseline age was inversely associated with changes in body weight and BMI. No clear non-linear dose–response association was observed between total aerobic exercise volume and pooled body composition effects, although some fitted curves suggested possible plateau-like patterns.

Conclusion

Aerobic exercise training may improve several adiposity-related body composition outcomes in postmenopausal women with overweight or obesity, although the evidence for hip circumference and lean body mass remains less certain. Future trials should improve reporting of exercise dose, adherence, dietary control, and follow-up outcomes.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13102-026-01778-7.

Introduction

Obesity is recognized as a complex chronic disease influenced by genetic, neurobiological, behavioral, and environmental factors [1]. With rapid population aging, postmenopausal women with overweight or obesity represent an increasingly important public health population [2]. Estrogen deficiency after menopause may disrupt metabolic regulation and contribute to unfavorable body composition changes, including increased whole-body and visceral adiposity and, in some women, reduced metabolically active lean body mass [3, 4]. Nevertheless, the etiology of body composition changes during and after the menopausal transition remains debated. In addition to hormonal changes, aging-related and behavioral factors, such as reduced physical activity, sleep disturbance, changes in appetite or energy intake, and resting energy expenditure, may also contribute to these changes [5, 6]. Therefore, menopause-related body composition changes should not be interpreted as exclusively hormone-driven, but rather as the result of interacting endocrine, aging-related, and lifestyle factors. Given the close association of central adiposity and unfavorable body composition profiles with cardiovascular disease, type 2 diabetes, and mortality risk among postmenopausal women [1, 7, 8], developing effective body composition management strategies for this high-risk population remains clinically important.

Regular aerobic exercise is recommended by public health guidelines and represents a feasible strategy for improving cardiometabolic health and body composition [9–11]. It may improve adiposity-related outcomes through increased energy expenditure and enhanced lipid oxidation. Nevertheless, the effects of aerobic exercise training as the primary exercise modality in postmenopausal women with overweight or obesity remain insufficiently clarified. Compared with resistance-based exercise, aerobic exercise may provide a weaker mechanical stimulus for maintaining or increasing lean body mass [12]. In addition, concerns remain that fat loss, particularly under conditions of energy deficit, may be accompanied by reductions in lean body mass [13, 14]. Therefore, further quantitative evidence is needed to clarify whether aerobic exercise training can reduce adiposity-related outcomes, such as body fat percentage, fat mass, and waist circumference, without unfavorable changes in lean body mass.

Previous meta-analyses have provided valuable evidence regarding exercise interventions in this population, but two methodological issues remain [15, 16]. First, because body mass index (BMI) cannot distinguish fat mass from fat-free mass or adequately capture central adiposity, evaluating exercise-induced changes in postmenopausal women requires multidimensional body composition outcomes [8, 17]. However, many trials report multiple correlated body composition outcomes or include multiple intervention arms sharing a single control group, which may not be adequately handled by conventional two-level meta-analyses when effect sizes are treated as independent [18, 19]. A three-level meta-analysis is therefore particularly appropriate because it allows dependent effect sizes from the same study to be retained while accounting for their clustering within studies. Second, most analyses assume a linear relationship between exercise dose and intervention effects, although exercise-induced changes in body composition may follow non-linear or plateau patterns [20].

To address these gaps, this systematic review and meta-analysis aimed to evaluate the effects of aerobic exercise training on multidimensional body composition outcomes in postmenopausal women with overweight or obesity. Accordingly, we hypothesized that: (1) aerobic exercise training would be associated with improvements in overall and central adiposity-related indicators; (2) participant characteristics and exercise prescription variables may moderate these effects; and (3) exercise dose may be associated with changes in body composition in a potentially non-linear manner.

Methods

This study strictly adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines to ensure the transparency and methodological rigor of the review process [21]. The protocol for this systematic review and meta-analysis was prospectively registered on the PROSPERO platform (Registration number: CRD420261336572).

Search strategy

A systematic and comprehensive literature search was conducted in February 2026 across five major electronic databases: PubMed, Cochrane Library, Web of Science, EBSCO, and ProQuest. The search strategy was constructed using Boolean logic and formulated strictly according to the PICOS framework to ensure maximum sensitivity and precision. The search syntax included the following terms: (“postmenopause” OR “menopause” OR “postmenopaus*” OR “post-menopausal” OR “menopaus*”) AND (“overweight” OR “obesity” OR “obes*” OR “adipos*”) AND (“exercise*” OR “training” OR “aerobic*” OR “high intensity interval” OR “HIIT” OR “walking” OR “cycling” OR “swimming” OR “running” OR “dancing”). An updated search was performed up to March 8, 2026. No additional filters (e.g., language or publication date limits) were applied during the retrieval process. The detailed search strategies for each database are provided in the Supplementary Material. Grey literature, trial registries, and conference abstracts were not included because this review was restricted to peer-reviewed full-text randomized controlled trials (RCTs) with extractable quantitative data.

Selection process

All retrieved records were imported into EndNote 21 software for deduplication. Subsequently, the unique records were exported and independently screened by two researchers. The initial screening evaluated the titles and abstracts based on predefined eligibility criteria. Any discrepancies or disagreements between the two reviewers (KYL and LBAS) were resolved through discussion referencing the established criteria to reach a consensus. If a consensus could not be reached, the corresponding author (SH) was invited to arbitrate the final decision.

Eligibility criteria

The inclusion and exclusion criteria were rigorously defined based on the PICOS framework [22].

Inclusion Criteria: (1) Study design: Publicly available RCTs published in peer-reviewed English-language journals. (2) Participants: Postmenopausal women with overweight or obesity at baseline. Postmenopausal status was defined according to the criteria reported in each included trial, most commonly amenorrhea for at least 12 consecutive months and/or biochemical confirmation [23]. For studies that recruited women aged ≥ 60 years but did not explicitly report menopausal status, postmenopausal status was inferred from age, because natural menopause generally occurs before this age. Overweight or obesity was defined according to WHO criteria (overweight: BMI ≥ 25 kg/m²; obesity: BMI ≥ 30 kg/m²) or population-specific/ethnic-specific diagnostic cut-offs [24–26]. (3) Interventions and comparators: The intervention group received aerobic exercise training (e.g., brisk walking, jogging, swimming, cycling), and the study explicitly reported relevant aerobic exercise dose parameters. The control group received no active exercise intervention (e.g., non-exercise control, maintaining daily habits, or dietary restriction only). (4) Outcomes: The study reported at least one multidimensional body composition outcome of interest—namely, body weight, BMI, body fat percentage, fat mass, lean body mass, waist circumference, or hip circumference—and provided extractable and valid data required for meta-analysis. (5) Comparison structure: A direct comparison between an aerobic exercise intervention arm and a non-exercise control arm was included.

Exclusion Criteria: (1) Non-original research or non-RCT designs (e.g., animal studies, systematic reviews/meta-analyses, case reports, expert consensuses, conference abstracts, cross-sectional studies, cohort studies, and case–control studies). (2) Participants not meeting the inclusion criteria, such as perimenopausal or premenopausal women, individuals with a normal baseline BMI, or populations with comorbidities that might confound body composition changes or preclude exercise (e.g., malignancies, uncontrolled endocrine disorders, severe cardiopulmonary/renal diseases, or musculoskeletal dysfunction). (3) Studies wherein participants received hormone replacement therapy (HRT) or other concurrent interventions known to significantly alter body weight and composition during the trial. (4) Interventions utilizing combined modalities (e.g., aerobic exercise combined with resistance training or pharmacological treatments) where the effect of aerobic exercise could not be extracted, or training modalities not conforming to the definition of aerobic exercise. (5) Duplicate publications (i.e., multiple reports from the same study cohort). Only the article with the most comprehensive data and the longest follow-up period was retained. (6) Studies with severe data missingness where valid data required for meta-analysis could not be obtained via mathematical calculation or by contacting the corresponding authors.

Data extraction

Following a predefined protocol, two independent researchers (KYL and SH) extracted data using an a priori developed and standardized Excel template. The extracted information included: (1) study characteristics (first author, publication year, and country); (2) participant characteristics and sample sizes; (3) aerobic exercise intervention parameters (type, duration, frequency, and intervention period); and (4) body composition outcomes. For datasets that were not reported in the text or were only presented graphically, the corresponding authors were contacted via email, allowing a two-week response window. If no response was received, WebPlotDigitizer v4.8 (https://apps.automeris.io/wpd4/) was utilized to digitally extract data from the published graphs, a tool previously validated for its high reliability and validity [27].

Data conversion

We systematically extracted the mean values, standard deviations (SDs), and sample sizes of the relevant indicators from the original studies to calculate the pre-to-post intervention mean differences. If studies provided only confidence intervals (CIs) or standard errors (SEs), the data were converted in accordance with the Cochrane Handbook for Systematic Reviews of Interventions [22]. For studies that did not directly report SDs of change scores, these were calculated from baseline and post-intervention SDs using an assumed pre–post correlation coefficient. Because most included studies did not report sufficient information to derive study-specific correlation coefficients, r = 0.50 was used in the primary analysis. To examine the robustness of this assumption, sensitivity analyses were further conducted using alternative correlation coefficients of r = 0.30, r = 0.70, and r = 0.90. Given the uniformity of measurement units across the specific body composition outcomes (e.g., kg for body weight, kg/m² for BMI), the Mean Difference (MD) was selected as the effect size. For analyses requiring the overall pooling of all multidimensional body composition outcomes into a unified metric, the Standardized Mean Difference (SMD), corrected using Hedges’ g, was employed [28].

Assessment of methodological quality and certainty of evidence

Two independent reviewers assessed the methodological quality and reporting completeness of the included RCTs using the revised Cochrane risk-of-bias tool for randomized trials (RoB 2.0). The RoB 2.0 evaluates bias across several domains: randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result, ultimately determining an overall risk of bias [29]. Furthermore, the certainty of the evidence was graded using the GRADE (Grading of Recommendations, Assessment, Development, and Evaluation) framework [30]. Evidence was categorized into “High,” “Moderate,” “Low,” or “Very Low” certainty based on the evaluation of five domains: risk of bias, inconsistency, indirectness, imprecision, and publication bias. Downgrading decisions considered the RoB 2.0 assessment, heterogeneity, directness of evidence, width of the 95% confidence intervals and prediction intervals, whether intervals crossed or approached the null value, sensitivity to data-conversion assumptions, the Optimal Information Size (OIS) criterion for continuous outcomes, and evidence of publication bias [31].

Statistical analysis

Meta-analysis

Given that the included studies frequently reported multiple outcomes, involved multiple follow-up time points, and utilized multiple parallel intervention arms sharing a single control group, a substantial nested effect was generated (i.e., multiple measurements and comparisons clustered within the same primary study). This data structure violates the core assumption of traditional meta-analyses, which strictly requires the absolute statistical independence of effect sizes [18]. To rigorously address these methodological constraints, we implemented a three-level meta-analysis framework, adopting the methodology outlined by Assink and Wibbelink [32]. In this framework, level 1 represents the sampling variance of each effect size, level 2 represents within-study variance among multiple effect sizes from the same study, and level 3 represents between-study variance. This approach allows multiple eligible effect sizes to be retained while accounting for their statistical dependency, rather than excluding or averaging correlated outcomes. The three-level meta-analysis was executed using the metafor package in R software (version 4.5.2). Model parameters were estimated using Restricted Maximum Likelihood (REML) and cross-validated via the Maximum Likelihood (ML) method to ensure the robustness of the results. The pooled main effects were expressed as MD or SMD, accompanied by 95% CIs and 95% Prediction Intervals (PIs), with P < 0.05 indicating statistical significance. Heterogeneity was evaluated using the stratified I² statistic, categorized as low (< 25%), moderate (25–75%), and high (> 75%) [33]. To address within-study dependency and sampling variability, cluster-robust variance estimation with small-sample correction was employed for sensitivity analysis and the identification of significant outliers.

Moderators analysis

Subgroup analyses and meta-regression were conducted to investigate potential sources of heterogeneity and identify modifying factors. Subgroup analysis was applied to categorical variables, including exercise intensity (moderate vs. high), exercise type (walking, walking/jogging, walking/cycling, comprehensive), and exercise frequency (3 sessions/week vs. > 3 sessions/week). Meta-regression was used for continuous variables, including baseline age, baseline BMI, and exercise intensity (%HRmax). Menopause type was not included in subgroup or sensitivity analyses because the original studies did not provide sufficient extractable data to allow separate analyses of natural versus surgical menopause. A P for interaction (P-interaction) < 0.05 indicated a statistically significant moderating effect.

Exercise dose effect analysis

For the dose–response analysis, body composition outcomes were synthesized as a unified metric after considering the direction of favorable change, with reductions in adiposity-related indicators such as body weight, BMI, and body fat percentage representing improvement. To ensure the reliability of the exercise dose analysis, we quantitatively pooled these specific body composition indicators into a single, unified index within the three-level framework, adopting an approach previously validated by Jiang et al. [34]. According to the American College of Sports Medicine (ACSM) guidelines [35], the total exercise dose was calculated using the formula: Session duration × Training frequency × Intervention duration (weeks). Subsequently, linear and various non-linear meta-regression models were compared to analyze the dose–response relationship based on the pooled effect sizes. From a physiological standpoint, a strictly linear assumption—implying that a greater exercise volume invariably yields proportionally better fat-loss results—is implausible; therefore, non-linear models were also explored to examine possible plateau or threshold patterns [36]. In the present study, Restricted Cubic Splines (RCS) demonstrated the optimal fit. Following the open-source RCS fitting R script provided by Ingram et al. [37], we modeled splines with 3, 4, and 5 knots using ML estimation. Model comparisons were performed via the Likelihood Ratio Test (LRT), revealing that the 3-knot spline provided the best model fit. For non-linear models exhibiting a distinct inverted U-shaped trajectory, we mathematically extracted the point of maximum significance (i.e., the zenith of the fitted curve where both bounds of the 95% CI remained negative). A significance test was conducted for this threshold point; a P < 0.05 confirmed that this cut-off point possessed statistical significance.

Publication bias

Potential publication bias was evaluated utilizing a three-level extension of Egger’s regression test combined with the visual inspection of funnel plots. A P > 0.05 was interpreted as showing no clear evidence of small-study effects.

Results

Literature search results

Following a systematic search across five major databases, an initial 4,275 records were identified. After the removal of duplicates, 2,467 records remained. Subsequent screening of titles, abstracts, and full texts for irrelevant literature narrowed the selection to 13 studies. Furthermore, an additional 3 studies were identified through relevant website hand-searching. Ultimately, 16 studies [38–53] were deemed eligible and included in the meta-analysis. The detailed selection process is illustrated in Fig. 1.

Fig. 1.

PRISMA flowchart for inclusion and exclusion of studies

Bias risk assessment results

A risk of bias assessment was conducted for all 16 included studies using the RoB 2.0 tool. Overall, the global risk of bias for all 16 studies was rated as “Some concerns,” with no studies demonstrating a “High risk” of bias. Across the five specific domains, “Measurement of the outcome” and “Missing outcome data” were rated as low risk in all 16 studies. Regarding the “Randomization process,” 12 studies were rated as low risk and 4 as having “Some concerns.” For “Selection of the reported result,” only 4 studies achieved a low-risk rating, while the remaining 12 had “Some concerns.” Notably, in the “Deviations from intended interventions” domain, all 16 studies were rated as having “Some concerns”; this reflects the inherent limitation of exercise intervention trials, wherein the blinding of participants and exercise instructors is practically unfeasible. Details are provided in the Supplementary Material.

Characteristics of included studies

The 16 included studies were published between 2003 and 2026. These trials were conducted globally, with the United States contributing the highest number (6 studies), followed by Iran (2 studies) and Turkey (2 studies). The remaining studies were distributed across Tunisia, the United Kingdom, South Korea, Egypt, China, and Brazil (1 study each). The total sample size across the included literature was 1,571 participants, comprising 631 in the control groups and 940 in the intervention groups. The sample sizes of individual studies ranged from 20 to 439 participants. Regarding participant characteristics, all subjects were overweight or obese postmenopausal women, with the mean baseline age across groups ranging from 53 to 70 years. For the interventions, the experimental groups engaged in various modalities of aerobic exercise, including walking, jogging, cycling, and aquatic exercise. The intervention periods spanned from 8 weeks to 12 months, with training frequencies typically ranging from 3 to 7 sessions per week. The control groups largely maintained their usual lifestyles, received no active exercise intervention, or underwent dietary restriction alone. Intervention adherence was reported in few studies and inconsistently described; therefore, it was not included as a moderator in the quantitative analyses. The reported multidimensional body composition outcomes encompassed body weight, BMI, body fat percentage, fat mass, lean body mass, waist circumference, and hip circumference (See Table 1).

Table 1.

Inclusion of basic characteristics of the literature

Author, Year, Country	Participants	Samplesize (n)	Age (years)	BMI (kg/m2)	Intervention	Training program	Outcomes
Abassi et al., 2026, Tunisia [38]	Postmenopausal women with overweight/obesity	C: 18 T: 18	C: 56.5 ± 3.65 T: 54.8 ± 3.18	C: 35.1 ± 5.21 T: 32.9 ± 4.76	C: Maintain daily habits T: Moderate-intensity intermittent walking	10 weeks, 4 sessions/week, 85 min/session	BW, BMI, Fat%, WC, HC
Tan et al., 2025, UK [39]	Postmenopausal women with overweight/obesity	C: 10 T: 10	C: 54.7 ± 3.3 T: 57.7 ± 4.8	C: 33.1 ± 4.2 T: 29.7 ± 3.5	C: Maintain daily habits T: Home-based interval training	8 weeks, 3 sessions/week, 20 min/session	BW, BMI, FM, LM, WC, HC
Zaravar et al., 2025, Iran [40]	Postmenopausal women with overweight/obesity	C1: 10 T1: 10 C2: 10 T2: 10	C1: 65.40 ± 3.28 T1: 65.20 ± 3.48 C2: 66.00 ± 3.56 T2: 65.10 ± 3.21	C1: 44.92 ± 4.81 T1: 45.15 ± 5.87 C2: 46.81 ± 5.49 T2: 45.64 ± 7.99	C1: Non-intervention T1: Water-based aerobic exercise C2: Vitamin D3 T2: Water-based aerobic exercise + vitamin D3	8 weeks, 3 sessions/week, 60 min/session	BMI
Guzel et al., 2024, Turkey [41]	Postmenopausal women with obesity	C: 12 T: 12	C: 54.42 ± 4.01 T: 55.67 ± 3.44	C: 32.82 ± 3.97 T: 32.89 ± 4.51	C: Maintain daily habits T: Low-volume walking training	10 weeks, 3 sessions/week, progressed to 40 min/session	BW, BMI, Fat%, FM, LM, WC, HC
Son et al., 2023, Korea [42]	Postmenopausal women with overweight/obesity	C: 12 T: 14	C: 69.9 ± 1.14 T: 70.2 ± 1.21	C: 26.06 ± 1.37 T: 26.04 ± 1.94	C: Non-intervention T: Moderate intensity walking exercises	12 weeks, daily, ~ 70–100 min/session	BW, BMI, Fat%
Malandish et al., 2022, Iran [43]	Postmenopausal women with overweight/obesity	C: 13 T: 14	C: 53.00 ± 3.26 T: 53.36 ± 3.98	C: 30.35 ± 6.60 T: 28.61 ± 4.20	C: Maintain daily habits T: Moderate intensity walking exercise	12 weeks, 3 sessions/week, 50–60 min/session	BW, BMI, Fat%
Elsayed et al., 2022, Egypt [44]	Postmenopausal women with overweight/obesity	C: 20 T: 20	C: 65.64 ± 3.03 T: 66.03 ± 2.87	C: 37.53 ± 1.16 T: 37.51 ± 1.39	C: Balanced diet T: Treadmill continuous walking + balanced diet	12 weeks, 3 sessions/week, 50 min/session	BW, BMI
Cao et al., 2019, China [45]	Older women with overweight/obesity; postmenopausal status inferred from age but not explicitly reported	C: 15 T: 13	C: 64.0 ± 4.6 T: 63.8 ± 5.9	C: 26.4 ± 1.4 T: 28.0 ± 2.9	C: Maintain daily habits T: Aerobic walking/jogging	12 weeks, 3 sessions/week, 60 min/session	BW, BMI, Fat%, FM, LM, WC
Azadpour et al., 2016, Turkey [46]	Postmenopausal women with overweight/obesity	C: 12 T: 12	C: 56.58 ± 4.17 T: 57.58 ± 4.29	C: 31.29 ± 1.40 T: 32.15 ± 1.78	C: Maintain daily habits T: Treadmill walking/jogging	10 weeks, 3 sessions/week, 25–40 min/session	BW, BMI, Fat%, FM, LM, WC, HC
Rossi et al., 2016, Brazil [47]	Postmenopausal women with overweight/obesity	C: 18 T: 15	C: 62.6 ± 5.9 T: 60.5 ± 7.3	C: 30.5 ± 4.3 T: 28.4 ± 2.9	C: Maintain daily habits T: Running track based aerobic training	16 weeks, 3 sessions/week, 52 min/session	BW, BMI, Fat%, FM, LM
Ryan et al., 2014, USA [48]	Postmenopausal women with overweight/obesity	C: 40 T: 37	C: 61 ± 1 T: 60 ± 1	C: 33 ± 1 T: 32 ± 1	C: Weight loss diet T: Treadmill and elliptical machine training + weight loss diet	24 weeks, 3 sessions/week, 45 min/session	BW, BMI, Fat%, FM, LM, WC
Foster-Schubert et al., 2012, USA [49]	Postmenopausal women with overweight/obesity	C1: 87 T1: 117 C2: 118 T2: 117	C1: 57.4 ± 4.4 T1: 58.1 ± 5.0 C2: 58.1 ± 6.0 T2: 58.0 ± 4.5	C1: 30.7 ± 3.9 T1: 30.7 ± 3.7 C2: 31.1 ± 3.9 T2: 31.0 ± 4.3	C1: Maintain daily habits T1: Treadmill walking, stationary bicycling, aerobics C2: Reduced-calorie diet T2: Treadmill walking, stationary bicycling, aerobics + Reduced-calorie diet	12 months, 5 sessions/week, 45 min/session	BW, BMI, Fat%, FM, LM, WC
Brinkley et al., 2011, USA [50]	Postmenopausal women with overweight/obesity	C: 22 T1: 22 T2: 17	C: 58.1 ± 1.2 T1: 58.3 ± 1.2 T2: 57.2 ± 1.3	C: 33.3 ± 4.1 T1: 32.9 ± 3.5 T2: 33.6 ± 3.9	C: Reduced-calorie diet T1: Reduced-calorie diet + moderate-intensity aerobic treadmill walking T2: Reduced-calorie diet + vigorous-intensity aerobic treadmill walking	20 weeks, 3 sessions/week, T1: progressed to 55 min/session, T2: progressed to 30 min/session	BW, BMI, Fat%, FM, LM, WC, HC
Nicklas et al., 2011, USA [51]	Postmenopausal women with overweight/obesity	C: 34 T1: 40 T2: 38	C: 58.4 ± 6.0 T1: 57.7 ± 5.5 T2: 59.0 ± 5.0	C: 33.9 ± 4.0 T1: 33.7 ± 3.5 T2: 32.9 ± 3.7	C: Reduced-calorie diet T1: Reduced-calorie diet + moderate-intensity aerobic treadmill walking T2: Reduced-calorie diet + vigorous-intensity aerobic treadmill walking	20 weeks, 3 sessions/week, T1: progressed to 55 min/session, T2: progressed to 30 min/session	BW, BMI, Fat%, FM, LM, WC, HC
Church et al., 2009, USA [52]	Postmenopausal women with overweight/obesity	C: 94 T1: 139 T2: 85 T3: 93	C: 57.2 ± 5.9 T1: 57.9 ± 6.5 T2: 56.7 ± 6.4T3: 56.4 ± 6.3	C: 32.2 ± 3.9 T1: 31.4 ± 3.7 T2: 32.2 ± 4.1 T3: 31.1 ± 3.6	C: Non-exercise T1: Treadmill and cycling training (4 kcal/kg/wk) T2: Treadmill and cycling training (8 kcal/kg/wk) T3: Treadmill and cycling training (12 kcal/kg/wk)	6 months, 3–4 sessions/week, duration based on target energy expenditure (T1: ~ 72 min/week, T2: ~ 136 min/week, T3: ~ 194 min/week)	BW, BMI, Fat%, WC
Irwin et al., 2003, USA [53]	Postmenopausal women with overweight/obesity	C: 86 T: 87	C: 60.6 ± 7.1 T: 61.0 ± 6.9	C: 30.6 ± 3.8 T: 30.5 ± 4.3	C: Maintain daily habits T: Treadmill walking/cycling	12 months, 5 sessions/week, 45 min/session	BW, BMI, Fat%, FM, WC, HC

BW Body Weight, BMI Body Mass Index, Fat% Body Fat Percentage, FM Body Fat Mass, LM Lean Body Mass, WC Waist Circumference, HC Hip Circumference

Meta-analysis results on the effects on body composition

Body weight

A total of 15 studies (21 effect sizes; N = 920 intervention + 611 control = 1,531 participants) reported that aerobic exercise training reduced body weight in overweight and obese postmenopausal women (MD = −2.17, P < 0.01, 95% CI: [−3.45, −0.89], 95% PI: [−4.95, 0.61]). Heterogeneity testing revealed an intra-study I² = 0% and an inter-study I² = 23.05% (See Fig. 2). Leave-one-out sensitivity analysis indicated that this pooled result was robust. Moderator analysis demonstrated that age significantly and negatively moderated the intervention effect (β = −0.372, 95% CI: [−0.609, −0.134], P = 0.004), indicating that younger participants experienced greater weight-loss benefits compared to older subjects. No other significant moderators were identified (See Supplementary Material).

Fig. 2.

Primary pooled effect sizes for body composition outcomes following aerobic exercise training in overweight and obese postmenopausal women. Note: BW, Body Weight; BMI, Body Mass Index; Fat%, Body Fat Percentage; FM, Body Fat Mass; LM, Lean Body Mass; WC, Waist Circumference; HC, Hip Circumference

Body mass index

Sixteen studies (23 effect sizes; N = 940 intervention + 631 control = 1,571 participants) reported that aerobic exercise training reduced BMI (MD = −0.73, P < 0.01, 95% CI: [−1.13, −0.33], 95% PI: [−1.55, 0.09]). Heterogeneity testing showed an intra-study I² = 0% and an inter-study I² = 18.37% (See Fig. 2). Leave-one-out sensitivity analysis confirmed the robustness of the pooled result. Age was again found to significantly and negatively moderate this effect (β = −0.093, 95% CI: [−0.175, −0.011], P = 0.028), indicating superior benefits for younger participants. No additional significant moderators were found (See Supplementary Material).

Body fat percentage

Thirteen studies (18 effect sizes; N = 890 intervention + 581 control = 1,471 participants) reported that aerobic exercise training reduced body fat percentage (MD = −1.40, P < 0.01, 95% CI: [−1.84, −0.96], 95% PI: [−1.84, −0.96]). Heterogeneity testing showed an intra-study I² = 0% and an inter-study I² = 0% (See Fig. 2). Leave-one-out sensitivity analysis indicated that this pooled result was robust, and no significant modifying factors were identified in the moderator analysis (See Supplementary Material).

Body fat mass

Ten studies (13 effect sizes; N = 537 intervention + 454 control = 991 participants) reported that aerobic exercise reduced body fat mass (MD = −1.83, P < 0.01, 95% CI: [−2.80, −0.85], 95% PI: [−2.80, −0.85]). Heterogeneity testing indicated intra-study I² = 0% and inter-study I² = 0% (See Fig. 2). Leave-one-out sensitivity analysis indicated that this pooled result was robust. No significant moderators were identified (See Supplementary Material).

Lean body mass

A total of 9 studies (12 effect sizes; N = 450 intervention + 368 control = 818 participants) reported the effect of aerobic exercise on lean body mass. Notably, the pooled estimate for lean body mass was positive but imprecise, with the confidence interval crossing the null value (MD = 0.72, P = 0.14, 95% CI: [−0.24, 1.68], 95% PI: [−1.28, 2.72]). Heterogeneity testing showed an intra-study I² = 0% and an inter-study I² = 38.65% (See Fig. 2). Leave-one-out sensitivity analysis confirmed the robustness of this finding, and moderator analysis revealed no significant moderating factors (See Supplementary Material).

Waist circumference

Eleven studies (17 effect sizes; N = 857 intervention + 548 control = 1,405 participants) demonstrated that aerobic exercise training reduced waist circumference (MD = −2.02, P < 0.01, 95% CI: [−3.06, −0.98], 95% PI: [−3.38, −0.66]). Heterogeneity testing showed an intra-study I² = 0% and an inter-study I² = 4.34% (See Fig. 2). Leave-one-out sensitivity analysis indicated that this pooled result was robust, with no significant moderators detected (See Supplementary Material).

Hip circumference

Seven studies (10 effect sizes; N = 256 intervention + 194 control = 450 participants) reported hip circumference. Aerobic exercise training was associated with a borderline reduction in hip circumference (MD = −1.39, P = 0.04, 95% CI: [−2.75, −0.02], 95% PI: [−2.75, −0.02]). Heterogeneity testing revealed an intra-study I² = 0% and an inter-study I² = 0% (See Fig. 2). Leave-one-out sensitivity analysis revealed that the pooled hip circumference estimate was sensitive to the exclusion of individual studies; in 6 of 10 iterations, the 95% confidence interval crossed the null value (See Supplementary Material). Combined with its borderline primary significance and sensitivity to the assumed correlation coefficient (Sect. 3.4.8), this finding should be interpreted with caution.

Sensitivity analysis for the assumed pre–post correlation coefficient

Sensitivity analyses using alternative assumed pre–post correlation coefficients (r = 0.30, 0.70, and 0.90) showed that the pooled results for body weight, BMI, body fat percentage, and fat mass remained robust. The result for waist circumference was generally stable under the primary and higher r assumptions. The estimate for lean body mass remained imprecise across different r assumptions. In contrast, the statistical interpretation for hip circumference varied across r assumptions, indicating that this finding should be interpreted cautiously. Detailed results are provided in the Supplementary Material.

Dose–response effects on body composition

The dose–response patterns between aerobic exercise training dose and body composition outcomes are illustrated in Fig. 3. The individual dose–response fitting results for the seven body composition outcomes (Fig. 3A) showed different fitted trajectories across total aerobic exercise volume. The fitted curves for body weight, waist circumference, and hip circumference initially decreased and then tended to plateau. The curves for BMI and fat mass showed an initial decline followed by a slight rebound. The curve for body fat percentage initially increased and then decreased, whereas the curve for lean body mass showed an initial decline before stabilizing. However, the scatter distribution for each indicator was relatively sparse, and the non-linear dose–response tests for individual outcomes did not reach statistical significance.

Fig. 3.

Dose–response and exercise frequency subgroup analysis results. A Dose–response results for all body composition parameters; B Dose–response results for comprehensive body composition; C Exercise frequency subgroup analysis results for comprehensive body composition. Note: BW, Body Weight; BMI, Body Mass Index; Fat%, Body Fat Percentage; FM, Body Fat Mass; LM, Lean Body Mass; WC, Waist Circumference; HC, Hip Circumference

The overall dose–response analysis, which pooled all multidimensional body composition outcomes (Fig. 3B), showed that the non-linear term fitted by the 3-knot RCS model was not statistically significant (QM (df = 2) = 0.53, P = 0.76). No clear enhancing or attenuating trend in the pooled body composition effect was observed across different levels of total aerobic exercise volume. Furthermore, the exercise frequency subgroup analysis based on the pooled results (Fig. 3C) showed that the pooled estimates for both moderate (3 sessions/week) and higher (> 3 sessions/week) exercise frequencies favored aerobic exercise training. The between-group difference test did not indicate a statistically significant difference in the pooled effect between these frequency categories (QM (df = 1) = 0.02, P = 0.88).

Publication bias

The results of the publication bias assessments and the corresponding funnel plots for all indicators are presented in Fig. 2. The P-values for Egger’s regression test across all outcomes were consistently greater than 0.05, suggesting no clear evidence of small-study effects or publication bias.

Evidence certainty

The certainty of evidence ranged from low to moderate. Moderate-certainty evidence suggested reductions in body fat percentage, body fat mass, and waist circumference, whereas low-certainty evidence suggested reductions in body weight, BMI, and hip circumference. For lean body mass, the estimate was positive but imprecise. Downgrading was mainly due to risk of bias and imprecision. In the imprecision assessment, body weight, BMI, body fat percentage, body fat mass, and waist circumference met or approached the approximate OIS threshold, whereas lean body mass only approached this threshold and hip circumference clearly did not meet it. The low certainty rating for hip circumference was therefore further supported by insufficient OIS, borderline primary evidence, leave-one-out non-robustness, and sensitivity to the assumed pre-post correlation coefficient. No outcome was downgraded for inconsistency, indirectness, or publication bias. Detailed GRADE assessments are provided in Fig. 2 and Supplementary Material.

Discussion

The present study aimed to evaluate the effects of aerobic exercise training on multidimensional body composition outcomes in postmenopausal women with overweight or obesity through a systematic review and three-level dose–response meta-analysis. The pooled findings indicated that aerobic exercise training was associated with reductions in body weight (MD = −2.17 kg), BMI (MD = −0.73 kg/m²), body fat percentage (MD = −1.40%), fat mass (MD = −1.83 kg), waist circumference (MD = −2.02 cm), and hip circumference (MD = −1.39 cm) compared with non-exercise control. The certainty of evidence was moderate for body fat percentage, fat mass, and waist circumference, and low for body weight, BMI, hip circumference, and lean body mass. For lean body mass, the pooled estimate was positive but imprecise, with the confidence interval crossing the null value; therefore, the available evidence does not support a clear reduction in lean body mass after aerobic exercise training. The hip circumference finding should also be interpreted cautiously because the confidence interval was close to the null value, the leave-one-out analysis showed non-robust statistical evidence, the result was sensitive to assumptions used in data conversion, and the accumulated sample size did not meet the OIS criterion. Furthermore, moderator analysis suggested that younger baseline age was associated with larger reductions in body weight and BMI; however, this should be interpreted as an associative moderator effect rather than evidence of causality. In exploring the dose–response relationship and subgroup analyses, no statistically significant non-linear association was observed between total exercise dose and pooled body composition effects, and higher training frequency (> 3 sessions/week) did not show a clear advantage over 3 sessions/week.

Summary of evidence

This study supports the potential benefits of aerobic exercise training for improving adiposity-related body composition outcomes. This conclusion is broadly consistent with recent similar meta-analyses, while also reflecting the methodological focus of the present review. Khalafi et al. [15] demonstrated that exercise training improved body composition in postmenopausal women, noting that aerobic exercise was superior to resistance training in reducing fat mass and waist circumference. This is generally consistent with our findings, though differences in pooled estimates may be partly related to our focus on postmenopausal women with overweight or obesity at baseline [54, 55]. Another meta-analysis by Dupuit et al. [16] explored the impact of high-intensity interval training (HIIT) on premenopausal and postmenopausal women, concluding that HIIT reduced whole-body and abdominal/visceral fat mass in overweight/obese cohorts. Our findings extend this evidence by suggesting that aerobic exercise training, including moderate-to-vigorous continuous aerobic exercise, may also improve several adiposity-related indicators in postmenopausal women with overweight or obesity [56].

From a systems physiology perspective, the capacity of aerobic exercise to promote reductions in adiposity-related outcomes may be partly explained by several plausible mechanisms. First, aerobic exercise can increase energy expenditure and enhance lipid mobilization through catecholamine-related pathways [57, 58]. Second, repeated skeletal muscle contraction during aerobic exercise may improve mitochondrial function and metabolic flexibility, partly through pathways involving AMP-activated protein kinase (AMPK) and peroxisome proliferator-activated receptor-gamma coactivator-1 alpha (PGC-1α) [59–62]. These adaptations could facilitate greater reliance on fatty acid oxidation over repeated training sessions and contribute to reductions in fat-related outcomes [63]. However, these mechanisms should be interpreted as explanatory hypotheses rather than direct evidence from the present meta-analysis, because the included trials were not designed to test these molecular pathways.

On the other hand, during the energy deficit state required for fat loss, aerobic exercise lacks the high-load mechanical tension typically provided by resistance training [64]. Traditional perspectives express concern that it might fail to counteract muscle atrophy induced by aging and postmenopausal estrogen depletion, and could potentially accelerate skeletal muscle protein breakdown due to gluconeogenic demands, leading to unfavorable changes in lean body mass [65]. However, our meta-analytic results showed that the pooled estimate for lean body mass was positive but imprecise, with the confidence interval compatible with both a small reduction and an increase. Therefore, it should not be concluded that aerobic exercise increases or preserves lean body mass with certainty. Rather, the available evidence suggests that reductions in adiposity-related outcomes were not accompanied by a clear loss of lean body mass in the included aerobic interventions. This phenomenon may partly stem from the specific baseline characteristics of our cohort: as these overweight or obese women engaged in weight-bearing aerobic interventions (e.g., walking, jogging, or aquatic exercises), their own substantial body mass may have provided some low-level mechanical loading stimulus to the lower extremities and core musculature. As noted by Cava et al. [13] in their research on preserving healthy muscle mass during weight loss, adequate mechanical loading is important for attenuating catabolism. For obese women, the mechanical tension generated by such weight-bearing aerobic exercise may contribute to maintaining lean body mass, even within a postmenopausal endocrine environment. However, this interpretation remains speculative and requires confirmation in trials directly assessing muscle mass, strength, protein turnover, and exercise adherence.

Moderating effect of age

Our moderator analysis indicated that the participants’ baseline age was negatively associated with changes in body weight and BMI. Based on the characteristics of the included subjects, this suggests that younger baseline age was associated with larger reductions in weight and BMI, but it should not be interpreted as evidence that age causally determines responsiveness to aerobic exercise. This finding is broadly consistent with Sims et al. [66], who reported that high-intensity habitual physical activity was more strongly associated with weight management in a younger cohort (50–59 years) than in an older cohort (70–79 years).

Physiologically, several mechanisms may partly explain this association. First, older women may exhibit lower maximal fat oxidation rates and reduced lipid mobilization capacities than younger women [67, 68]. Second, age-related reductions in sympathetic nervous system activity, brown adipose tissue thermogenesis, and resting metabolic rate may contribute to smaller changes in weight-related outcomes [69]. Third, age-related chronic low-grade inflammation and impaired insulin or AMPK-related signaling may influence metabolic responsiveness to exercise [70]. However, these explanations remain hypothetical in the context of the present study, because the moderator analysis was conducted at the study level and may be affected by residual confounding, including baseline health status, dietary intake, habitual physical activity, medication use, and adherence. Therefore, future trials and individual participant data meta-analyses are needed to determine whether age independently modifies the body composition response to aerobic exercise.

Exercise dose–response effects

In exploring the association between aerobic exercise dose and fat-loss efficacy, we did not observe a statistically significant non-linear association between total exercise volume and pooled body composition effects. For individual indicators, the fitted curves suggested possible plateau-like patterns for some outcomes, but the scatter distribution was sparse and the number of effect sizes was limited. Further subgroup analysis did not show a clear difference between higher-frequency aerobic exercise (> 3 sessions/week) and the standard frequency of 3 sessions/week.

The mechanisms driving this phenomenon are uncertain and should be interpreted cautiously. One possible explanation is that, over longer intervention periods, improvements in movement economy may reduce the energy cost of the same exercise workload [71, 72]. Another possibility is behavioral compensation, whereby higher exercise frequency may increase fatigue or reduce non-exercise activity thermogenesis (NEAT), such as daily walking or household chores [73, 74]. However, because adherence, dietary intake, and habitual physical activity were not consistently reported across studies, the present meta-analysis cannot directly test these mechanisms. Importantly, the absence of statistically significant non-linearity should not be interpreted as evidence that dose–response relationships are absent; rather, the available data may have had limited statistical power to detect non-linear patterns. From a practical perspective, these findings suggest that aerobic exercise prescriptions for postmenopausal women with overweight or obesity should prioritize sustainability, adherence, and individual tolerance rather than simply increasing weekly frequency or total exercise volume.

Limitations and future perspectives

This study is not without limitations. First, all 16 included RCTs faced the inherent methodological challenge of being unable to blind participants and exercise instructors. This unblinded status may have, to some extent, induced placebo effects or unanticipated subjective lifestyle modifications among participants (e.g., consciously adopting healthier dietary habits simply because they were assigned to an exercise group). Second, the relatively small number of studies included in the moderator analysis may have reduced statistical power. Furthermore, the results of the meta-regression are observational in nature rather than establishing causality; thus, interpretations should be made with caution. In this context, one included trial, Zaravar et al. [40], had a substantially higher baseline BMI than the other trials, which may have introduced clinical heterogeneity into the BMI analysis. Third, over intervention periods spanning several weeks to months (or even up to a year), the subtle yet continuous spontaneous lifestyle changes (e.g., habitual diet, daily physical activity) among participants could not be absolutely quantified or isolated. Fourth, intervention adherence and post-intervention follow-up were insufficiently and inconsistently reported across the included trials, limiting our ability to evaluate dose completion, long-term sustainability, and whether body composition changes were maintained after the intervention. Fifth, although most pooled outcomes met or approached the OIS threshold, the evidence for hip circumference was based on a relatively small accumulated sample size and should therefore be interpreted with particular caution. In addition, for body fat percentage, body fat mass, and hip circumference, the identical 95% confidence intervals and 95% prediction intervals resulted from zero estimated heterogeneity variance components in the fitted models; therefore, these prediction intervals should not be overinterpreted as evidence of no variability in future studies. Finally, although the search was updated to March 2026, potential publication lag bias cannot be fully excluded, particularly for recent or ongoing trials that may not yet have been published.

To address these limitations, future large-scale, multi-center, high-quality RCTs or prospective cohort studies should consider incorporating more rigorous daily energy and behavioral monitoring systems. We recommend integrating high-precision wearable devices into the study design to accurately track participants’ real-world energy expenditure trajectories and shifts in daily activity patterns. Future studies should also standardize the reporting of exercise dose, adherence, adverse events, and follow-up outcomes. This would not only help comprehensively isolate the potential dietary and behavioral compensatory effects during the intervention but also provide invaluable empirical data to further delineate the applicable boundaries of the “constrained total energy expenditure” model within the high-risk population of overweight and obese postmenopausal women.

Conclusion

In conclusion, aerobic exercise training may be beneficial for improving adiposity-related and anthropometric outcomes in postmenopausal women with overweight or obesity. However, the effects on lean body mass remain uncertain, and the evidence for hip circumference should be interpreted cautiously. Considering the variability in trial design, co-intervention control, and reporting quality across studies, further well-designed trials are needed to clarify the independent and sustained effects of aerobic exercise on body composition.

Authors’ contributions

Conception and design of study: KYL and SH; Retrieve articles and acquisition of data: KYL, SH and LBAS; Analysis of data: KYL and SH; Interpretation of data: KYL and SH; Drafting the manuscript: KYL; All authors read and approved the final manuscript.

Funding

Not applicable.

Data availability

All other data are available upon request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.