Effect of exercise on the human gut microbiota in individuals with overweight and obesity: a systematic review and meta-analysis of randomized controlled trials

Selection and eligibility

This study followed the PRISMA guidelines to ensure a transparent and systematic approach to selecting relevant studies. The eligibility criteria were established based on the PICO framework. Participants included individuals classified as overweight or obese, defined as a BMI ≥ 25 kg/m² or waist circumference (WC) ≥ 90 cm. The intervention of interest was exercise, whereas the comparison group consisted of non-exercise controls, alternative interventions, or varying exercise intensities. The primary outcome was gut microbiota-related changes, focusing on microbial diversity and specific taxonomic alterations. Studies without control groups were considered eligible if they compared different exercise intensities. The review protocol was not registered in PROSPERO or any other public database because of the limited number of eligible studies and the narrow scope of this review.

Search strategies and selection process

A comprehensive literature search was conducted in four databases, PubMed, Embase, Cochrane Library, and Web of Science, covering studies published up to November 5, 2024. The search was limited to randomized controlled trials (RCTs) conducted on human subjects. Keywords were carefully selected to identify relevant studies by combining terms related to metabolic disorders, gut microbiota, and exercise. The primary search terms included: (“diabetes” OR “T2DM” OR “type 2 diabetes” OR “obesity” OR “metabolic syndrome” OR “hypertension” OR “MetS”) AND (“gut microbiota” OR “microbiome” OR “intestinal permeability” OR “gut barrier function” OR “Zonulin”) AND (“physical activity” OR “exercise” OR “HIIT” OR “MICT” OR “aerobic training”).

Studies were excluded if they involved participants who were not overweight or obese, focused on specific diseases, such as celiac disease or coronary artery disease, or lacked a clear exercise intervention. Studies involving pharmacological or nutritional interventions were excluded to prevent potential confounding effects. Abstract-only publications without full-text access were excluded owing to insufficient information. During the eligibility phase, studies that mentioned the gut microbiota but did not provide descriptive outcome data were excluded from the analysis.

Data extraction

Data were extracted by a PhD researcher specializing in exercise and nutrition. The extracted information included participant characteristics (e.g., age and health status), intervention type, and gut microbiota outcomes. The extracted data were systematically organized using EndNote (version 21.4) for reference management. Outcomes related to the gut microbiota were categorized into alpha and beta diversity measures and changes in specific microbial taxa.

Data analysis

Quantitative synthesis was performed for studies that provided sufficient data for the effect size calculation. Comprehensive Meta-Analysis (CMA, version 4) software was used for statistical analysis. The effect sizes were expressed as odds ratios (ORs) with 95% confidence intervals (CIs). Given the expected heterogeneity among the studies, a random-effects model was applied to ensure a more conservative estimate of the intervention effects. Heterogeneity was assessed using the I² statistic with thresholds of 25%, 50%, and 75% representing low, moderate, and high heterogeneity, respectively.

Risk-of-bias assessment

The risk-of-bias of the included studies was assessed using Cochrane’s Risk of Bias 2 (RoB 2) tool. This tool evaluates potential bias across five key domains: randomization process, deviations from intended interventions, missing outcome data, outcome measurement, and selection of the reported result. Each domain was categorized as low-, moderate-, or high-risk. The assessments were performed independently by three reviewers, and any discrepancies were resolved through a consensus process to ensure accuracy and consistency. This multi-reviewer approach minimized subjectivity and improved the reliability of the evaluations. Only studies that reported comparable and quantifiable metrics of microbial beta diversity, specifically Bray–Curtis dissimilarity, were included.

Systematic review of selected studies

A total of 777 records were initially identified through the database search. After removing 212 duplicate records, 565 remained for further screening. Following the screening of titles and abstracts, 295 articles were selected for full-text evaluation based on predefined eligibility criteria. Among these, 13 studies were initially considered eligible; however, two studies were excluded owing to insufficient outcome data and pharmacological treatment, resulting in the final inclusion of 11 RCTs in the systematic review. The detailed study selection process is shown in Figure 1.

Figure 1.

Flow diagram of process for the systematic review.

The selection process for studies included in this systematic review was conducted in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines.

The 11 studies included collectively involved 319 participants, with individual study sample sizes ranging from 12 to 85. The participants included overweight and obese individuals across various age groups ranging from children to older adults. Except for one study [21], which involved diabetic patients with a mean waist circumference of 109 cm (indicative of central obesity), all other studies reported a mean BMI exceeding 25 kg/m². Exercise interventions were mandatory in all the included studies, whereas the control groups varied across studies and consisted of sedentary groups, dietary interventions, or alternative exercise intensities. One study employed either moderate-intensity continuous training or high-intensity interval training combining aerobic and resistance components (C-MICT and C-HIIT). A detailed summary of the study characteristics is presented in Table 1.

Table 1.

Summary of the effects of different exercise types on gut microbiota diversity and related microbial taxa changes

Meta-analysis of the microbiota diversity

Gut microbiome outcomes were assessed using 16S rRNA gene sequencing or shotgun metagenomics, focusing on changes in alpha diversity (e.g., Shannon index) and beta diversity (e.g., Weighted UniFrac and Bray–Curtis dissimilarity), as well as specific microbial taxa alterations. The included studies were further categorized for quantitative evaluation based on their reporting of alpha and beta diversities. Although various measures of alpha and beta diversities were reported across studies, only Bray–Curtis dissimilarity was presented consistently enough to be included in the meta-analysis. The number of studies that reported alpha diversity and weighted UniFrac values in a comparable manner was insufficient for inclusion in the meta-analysis.

Risk-of-bias assessments were performed for all included studies using the RoB 2 tool. All five domains—D1: bias arising from the randomization process, D2: deviations from intended interventions, D3: missing outcome data, D4: measurement of the outcome, and D5: selection of the reported results—were rated as low-risk across all included studies, based on consensus among three independent reviewers. Specifically, D1 was rated as low-risk because all studies provided a clearly described randomization process. For D2, although blinding is inherently difficult in exercise-based trials and some concerns were initially raised, no major deviations from the intended interventions were identified, and the domain was ultimately judged to be low-risk. D3 was rated as low-risk, as all studies utilized stool-based analyses (16S rRNA or shotgun metagenomics), thus minimizing the possibility of missing data. D4 was also rated low owing to the consistency and comparability of outcome measurement methods across studies. Finally, D5 was rated as low-risk because both significant and non-significant results were transparently reported without evidence of selective outcome reporting. Figures 2 and 3 present the risk-of-bias assessments of the meta-analysis outcomes.

Figure 2.

Weighted bar of risk-of-bias assessment.

This plot summarizes the distribution of risk judgments (low, some concerns, high) across all domains using the ROB 2 tool.

Figure 3.

Traffic light plot of risk-of-bias assessment.

Each row represents an individual study, with color-coded judgments for each domain assessed according to the ROB 2 tool.

The meta-analysis revealed significant improvements in microbiota diversity in participants who underwent exercise interventions compared with those in the control groups. The results based on Bray–Curtis dissimilarity demonstrated consistent patterns of enhanced microbial diversity. The pooled effect size was 4.56, with a moderate level of heterogeneity observed (I² = 48.2%, τ² = 0.444) (Figure 4).

Figure 4.

Meta-analysis of Bray–Curtis dissimilarity.

A significant difference in microbial community composition was observed between the intervention and control groups, with greater Bray–Curtis distances in the exercise intervention group. The pooled effect size was 4.56 (95% CI [1.77, 11.80]), with a Z-value of 3.14 (P = 0.002). Moderate heterogeneity was observed, with an I² value of 48.2%, τ² = 0.444, and a Q-value of 5.793 (P = 0.122).

Overall effect of exercise intervention on microbiota diversity

Among the 10 studies analyzing beta diversity, six [21-26] reported significant changes in beta diversity following exercise intervention between 2 and 6 months. One study [27] reported exercise-induced changes in beta diversity but did not provide statistical significance.

In contrast, three studies [28-30] did not identify any significant differences in beta diversity after exercise interventions. Couvert et al. compared isoenergetic HIIT programs performed at the same intensity and frequency during running and cycling [29]. Therefore, the absence of significant differences between the two groups may result from the matched exercise conditions. A study comparing a group performing HIIT alone with another group performing HIIT combined with intermittent fasting found no relative differences in beta diversity [28]. No differences were observed in the beta diversity clustering based on either Unweighted or Weighted UniFrac analyses [30]. However, functional metabolite analysis based on KEGG pathways revealed significant segregation between HOMA-IR responders and non-responders, as shown in the beta diversity NMDS plot using Bray–Curtis dissimilarity.

Changes in beta diversity were sometimes minimal, depending on the intervention type and study duration. For example, both weight loss strategies (exercise + diet vs. diet only) led to a 10% reduction in body weight with similar directional changes in beta diversity [22]. In the Weighted UniFrac, greater distances were observed in the exercise group than in the non-exercise control group [25]. No significant changes were observed in the Weighted UniFrac and Bray–Curtis indices, whereas a significant difference in beta diversity was detected using the Euclidean distance [26]. In addition, the Bray–Curtis dissimilarity index remained consistently high. Although heterogeneity decreased after the exercise intervention, the index remained higher than that of the control group [25]. These results reinforce the previous findings of a systematic review [31], which found distinct beta diversity separation in animal models, but reported inconsistent patterns in humans depending on disease status. We observed a consistent pattern of a significantly greater Bray–Curtis dissimilarity distance in the exercise intervention group than in the control group among overweight/obese individuals [23-25]. The meta-analytic findings regarding Bray–Curtis dissimilarity are illustrated in Figure 4.

Regarding alpha diversity, the findings have been inconsistent; although most studies reported no significant effect of exercise, two studies reported a significant increase [24,25]. A previous systematic review of healthy individuals reported an association between cardiac fitness and alpha diversity [32]. Similarly, a significant correlation between VO₂ max and Shannon diversity values was observed [23].

Additionally, a positive correlation between the Shannon index and fat mass reduction has been reported, suggesting a potential link between exercise-induced changes in body composition and microbial diversity [29]. Nonetheless, as previously noted, we were unable to identify a consistent relationship between increases in alpha diversity and exercise effects or exercise-mediated factors such as improvements in physical fitness or reductions in body weight. These results may be attributed to the fact that previous studies have primarily focused on comparing the effects relative to a control group. Consequently, in studies that examined differences based on exercise intensity, no significant changes in alpha diversity have been reported between exercise types. However, these studies may have overlooked meaningful changes in the effects of exercise in individual groups.

Overall, these results suggest that exercise induces short-term changes in the beta diversity in overweight and obese individuals. Supporting this, a large-scale meta-analysis of 3,329 samples found that obese individuals exhibited significantly lower beta diversity than healthy controls, with a more homogeneous gut microbiota composition [33]. Our findings align with this, suggesting that exercise interventions can effectively enhance microbial diversity and counteract the monotonous microbial patterns associated with obesity. However, improvements in alpha diversity have not been consistently observed. This finding is inconsistent with human studies that have reported that exercise enhances alpha diversity. One possible explanation is that this observation may be attributable to the fact that our study was limited to overweight and obese populations. Although reduced microbial diversity is often associated with dysbiosis, various indicators have been employed to assess this condition, and no single metric has been consistently recognized as the universal gold standard. Nevertheless, our findings confirmed that exercise interventions serve as a strategy to modulate beta diversity in obese individuals, promoting more phylogenetically distinct and diverse microbial compositions across individuals.

Changes in microbiota

The overall changes in microbiota following exercise intervention are presented in Table 1. High-intensity exercises exceeding 80% HR and sprint-based exhaustive exercise interventions commonly result in an increase in bacteria of the order Lachnospirales, particularly within the Clostridium citroniae, Lachnospiraceae family, and Lachnospira genus [21,30,34].

In addition, moderate-intensity exercise training led to an increase in Lachnospira eligens [21]. At the species level, Roseburia, which belongs to the Lachnospirales order, also increased following exercise [26,27]. Blautia, a butyrate-producing genus, has been associated with increased visceral fat when its abundance is reduced [35], whereas it has been associated with obesity and metabolic disorder alleviation [36]. However, its response to exercise remains inconsistent. The relative abundance of children with obesity increased following an exercise intervention [27]. In contrast, a decrease was observed after moderate-intensity sprint interval training [34]. Considering the short intervention period (2 weeks), these findings suggest that long-term dietary patterns can influence changes in SCFA levels within subjects and that these changes may vary depending on factors such as the subject’s growth phase. Given that these factors may lead to different outcomes in children with obesity, distinguishing growing adolescents from separate populations is essential.

Regarding the order Oscillospirales, classified under Firmicutes, the changes observed following the exercise interventions were primarily driven by members of the Ruminococcaceae family. Specifically, after exercise intervention, higher abundances of Faecalibacterium prausnitzii, Clostridium leptum, and Ruminococcus bromii have been reported [21,34].

Differences in the exercise intensity were observed. The relative abundance of Ruminococcus bromii increased following C-HIIT, whereas after C-MICT, the relative abundance of Faecalibacterium prausnitzii within the Ruminococcaceae family increased [21].

Similarly, only MICT showed a trend toward an increased relative abundance of the genus Faecalibacterium [34]. Faecalibacterium is a butyrate-producing bacterium with anti-inflammatory properties that decreases pro-inflammatory cytokine levels, enhances intestinal barrier integrity, and potentially alleviates inflammatory bowel disease [37,38]. Furthermore, its beneficial effects extend to metabolic disorders including obesity and T2DM [37,39]. These findings highlight that the gut microbiota composition may exhibit differential responses to exercise intensity at a finer taxonomic resolution within the same family. Although variations in the relative abundance of microbes were detected between C-MICT and C-HIIT, no significant differences were found in the overall production of SCFA [21].

Based on a comparison of Carbohydrate-Active Enzymes (CAZy), both Ruminococcaceae and Lachnospiraceae harbor a greater number of glycoside hydrolases (GH) than Clostridiaceae, facilitating the more efficient degradation of complex carbohydrates [40]. Notably, Lachnospiraceae exhibited a significantly higher abundance of ATP-binding cassette (ABC) transporter genes.

Firmicutes was the most dominant phylum in the gut bacterial composition, with a highly variable species level composition. Our findings indicate that, whereas some studies reported an increase in Lachnospiraceae and Ruminococcaceae, which tended to increase within Firmicutes after exercise, Lachnospiraceae consistently showed a greater and more frequent increase across studies. Exceptionally, an increase in members of the Lachnospiraceae family and a reduction in species belonging to the Ruminococcaceae family were observed at the 1-month time point of weight loss, both in exercise- and diet-induced interventions, as well as in the diet-only group [22]. These results suggest that although exercise intervention is generally beneficial for host energy metabolism, as it can promote an increase or decrease in SCFA-producing bacteria within a broader framework, its effects do not always follow a consistent pattern.

Several studies have reported changes in the Bacteroidetes phylum after exercise. A relative increase in the abundance of Coprobacter, Rikenellaceae, and Bacteroidetes was observed after exercise [29,30,34]. A significant increase in Bacteroidetes was detected in the group that underwent intermittent fasting alone, whereas a similar increasing trend was noted in the group that underwent combined intermittent fasting and exercise, although the difference was not statistically significant (P > 0.05) [28]. In contrast, in a study in which Alistipes shahii showed a relative decrease, HIIT in patients with T2DM led to significant improvements in whole-body insulin sensitivity (Matsuda Index) among those with improved insulin resistance (HOMA-IR) [30]. However, except for the resting heart rate, no changes were observed in BMI or other clinical markers.

Similarly, we hypothesized that excess energy produced by Firmicutes could be alleviated through exercise. However, our findings on the Firmicutes/Bacteroidetes (F/B) ratio did not always align with our expectations. Although both the F/B ratio and its relationship with obesity are multifactorial indicators [41,42], only one study [34] reported this ratio in the context of exercise. This inconsistency may result from the absence of dramatic weight loss following interventions in the selected studies. Except for one study that reported a body weight reduction of approximately 10% [22], the interventions in the other reviewed studies did not lead to dramatic weight loss, despite some achieving statistically significant but modest weight reductions.

Furthermore, exercise is a promising strategy for increasing SCFA production, although its effects can be superimposed by well-mediated interventions such as the Mediterranean diet or pharmacological treatments for diabetes. The Mediterranean diet increases beneficial bacteria to appropriate amounts [43], whereas drug-based interventions improve the F/B ratio, eliciting similar effects on the gut microbiota [44]. These overlapping effects complicate the attribution of changes in microbiota to exercise alone. Nevertheless, both studies reported that exercise and weight loss significantly improved physical fitness, promoted SCFA-producing bacteria, and contributed to the favorable modulation of the F/B ratio. Collectively, these findings suggest that exercise, as a broad-acting intervention, may contribute to improvements in both host metabolic health and gut microbiota composition. However, its independent effects should be interpreted with caution because of the potential confounding influence of diet and medication (e.g., metformin and GLP-1 receptor agonists).

Excluding the changes in Firmicutes and Bacteroidetes, our studies consistently emphasized the benefits associated with the expression of SCFA-producing bacteria. However, the expected microbial shifts in the specific target taxa were not consistently observed. A representative example of this is Akkermansia. Akkermansia, a beneficial gut bacterium belonging to the phylum Verrucomicrobia, enhances the intestinal mucosal barrier and exerts anti-inflammatory effects [45,46]. For instance, its abundance increased significantly following moderate-intensity exercise compared to high-intensity exercise [21]. Although these levels remained higher than those in the non-responder group, a reduction was observed [30]. Similarly, no significant changes in the relative abundance have been reported [34]. Changes in the relative abundance of Akkermansia were positively correlated with changes in visceral fat mass [29]. Nevertheless, a causal relationship indicating that high-intensity training directly increased Akkermansia abundance was not observed. Such inconsistencies may stem from methodological heterogeneity across studies, including differences in statistical approaches, a focus on between-group versus within-subject comparisons, and variations in control group characteristics. Taken together, these results indicate that in the case of microbial taxa, particularly species such as Akkermansia, relative changes can vary in direction depending on the baseline abundance levels.

Although gut microbiota cannot be definitively identified as the primary cause of obesity, modulation of its composition continues to be regarded as a key therapeutic strategy. Whether beneficial bacteria should be targeted at the species level or administered as part of a multi-strain cocktail remains an open question [47]. Therefore, when aiming to modulate the gut microbiota through natural interventions such as exercise, rather than targeted treatments such as prebiotics, it is necessary to account for the heterogeneity of microbial responses and individual health characteristics.

Given the variability in the F/B ratio across obese populations, which is largely influenced by individual characteristics and environmental factors, improved taxonomic markers are needed for a more accurate assessment of obesity-related dysbiosis [41]. Reliance solely on F/B ratio may be insufficient for guiding obesity management strategies. Therefore, multiple microbial markers should be evaluated. Although weight loss-induced changes in F/B ratio and SCFA expression are generally associated with improvements in gut health, these effects vary significantly among individuals. The question remains as to whether gut microbiota-related benefits stem primarily from fasting-induced weight loss or exercise-mediated metabolic improvements. Therefore, individual characteristics, such as baseline fitness level, abdominal adiposity, and total fat mass, should be carefully considered as they may significantly influence the gut microbiota composition and its metabolic consequences.

Alpha and beta Diversity: interpretation

Our meta-analysis demonstrated a significant increase in beta diversity (Bray–Curtis dissimilarity) following exercise intervention, indicating that physical activity may lead to greater microbial differentiation among individuals. Although several individual studies have reported changes in alpha diversity, including increases in Shannon and Simpson indices, inconsistent reporting formats and methodological variability have prevented quantitative synthesis. Consequently, our meta-analysis focused exclusively on beta diversity using Bray–Curtis dissimilarity. The observed increase in beta diversity suggests greater inter-individual microbial differentiation following exercise. However, the absence of a meta-analysis of alpha diversity limited our ability to assess within-individual richness or evenness, thereby moderating the overall strength of our conclusions.

Limitations and further research

One limitation of this study is that it relied solely on data from published articles and supplementary materials. To minimize the risk of overinterpretation, we reported effect sizes based exclusively on between-group comparisons with the control groups. Although some studies reported significant within-subject changes, they were not included in the meta-analysis to maintain methodological rigor and ensure a more conservative interpretation of the findings. In line with this limitation, the small number of eligible studies precluded meta-regression analyses based on intervention frequency, exercise modality, or participant characteristics. Only four studies have provided comparable and quantifiable data on microbial beta diversity.

Lachnospirales and members of the Ruminococcaceae family belong to the phylum Firmicutes, and although their abundance increases following exercise, their roles are not always beneficial, particularly when considered in the context of obesity-related elevations in the Firmicutes/Bacteroidetes (F/B) ratio and energy surplus. Nonetheless, their potential advantages are often highlighted owing to their involvement in SCFA production and associated mechanisms.

In our review, we sought to identify consistent patterns, such as sustained associations between specific microbial taxa and improved metabolic outcomes (e.g., fat reduction, insulin sensitivity, and KEGG pathway activation); however, no such convergent responses were observed. This limitation may be attributed to the heterogeneity in intervention duration, exercise modality, and the presence or absence of underlying metabolic conditions, despite the consistent inclusion of overweight or obese individuals across studies. In addition, some key variables, including functional outputs such as KEGG pathway data, were not assessed in several studies, further limiting cross-study comparisons.

Although many existing reviews emphasize the general benefits of exercise, to the best of our knowledge, very few have specifically focused on overweight or obese individuals. Most previous studies have either combined human and animal data or included heterogeneous populations without stratification by weight status. In contrast, the strength of this review lies in its exclusive focus on overweight and obese individuals. However, one limitation is that many of the included studies did not report significant weight loss following the intervention, which somewhat constrained our ability to draw distinct conclusions regarding the weight-independent effects of exercise.

The human gut microbiome is highly individualized and is influenced by various factors, including physiology, environment, and genetics [48,49]. Although all studies included overweight or obese participants, this does not necessarily confirm that all measures of the gut microbiome composition reflect a consistent state of dysbiosis.

In addition to exercise, prebiotics and fecal microbiota transplantation (FMT) have been proposed as promising strategies for modulating gut microbiota in obesity management. However, the underlying mechanisms by which these interventions influence host metabolism remain poorly understood, and there is a lack of conclusive evidence linking specific microbial species to the direct regulation of obesity [4]. Future studies should focus on monitoring key microbial markers related to F/B ratio control and conducting functional analyses of metabolic pathways (e.g., Kyoto Encyclopedia of Genes and Genomes pathway) to better understand these interactions.

To advance this field, future research should explore the combined effects of physical fitness improvement, weight loss, and body fat reduction on the expression and metabolic activity of key gut microorganisms. Personalized intervention strategies that integrate exercise, prebiotics, and other dietary approaches tailored to individual microbiota profiles could serve as a foundation for precision medicine. Such an approach could help address the significant variability in gut microbiota responses and provide more targeted solutions for obesity management.