Comparative effects of different exercise types on inflammatory markers in type 2 diabetes mellitus patients with overweight and obesity: a systematic review and network meta-analysis

Abstract

Background

For patients with type 2 diabetes mellitus (T2DM) living with overweight or obesity, systemic chronic low-grade inflammation induced by excessive fat accumulation not only exacerbates insulin resistance but also serves as a pivotal driver of disease progression and cardiovascular complications. As a vital component of non-pharmacological therapy, exercise has demonstrated the potential to modulate immune responses and improve metabolic status, holding positive significance for combating this inflammatory process and delaying disease progression.

Objective

This study aims to compare the effects of aerobic exercise (AE), resistance training (RT), and combined aerobic and resistance training (CE) on inflammatory marker levels in T2DM patients living with overweight and obesity through a network meta-analysis (NMA).

Method

Literature searches were conducted up to November 2025 across Cochrane, Embase, PubMed, and Web of Science databases to identify English-language randomized controlled trials (RCTs) meeting the inclusion criteria.

Results

A random-effects NMA was performed within a frequentist framework using Stata 17.0. 71 RCTs (N=4,266) were included. Results indicated that AE exhibits the highest probability of effectiveness in reducing high-sensitivity C-reactive protein (hs-CRP) (SUCRA = 81.7; SMD = -1.18; 95% CI: -1.81, -0.55; P < 0.001) and leptin (SUCRA = 91.5; SMD = -0.71; 95% CI: -1.05, -0.37; P < 0.001) levels as well as increasing adiponectin (SUCRA = 76.7; SMD = 1.13; 95% CI: 0.42, 1.84; P = 0.002) levels. At the same time, CE appears to offer the greatest potential in lowering IL-6 (SUCRA=76.6; SMD=-0.93; 95% CI: -1.63, -0.23; P=0.01) levels. RT did not significantly affect most inflammatory markers. TNF-α levels were not significantly modulated across any of the exercise interventions.

Conclusion

This NMA based on RCTs confirms that different exercise types exhibit varying efficacy on distinct inflammatory markers in T2DM patients with overweight or obesity. These findings provide an evidence-based reference for clinicians when selecting exercise modalities based on a patient’s specific inflammatory profile, thereby more effectively managing their chronic inflammatory state. Given the holistic nature of lifestyle interventions, future research may integrate lifestyle control measures such as calorie restriction or anti-inflammatory diets. The current study was registered with the International Prospective Register of Systematic Reviews (PROSPERO), ID: CRD420251218729.

Clinical trial number

Not applicable.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12902-026-02193-2.

Introduction

According to the latest data from the International Diabetes Federation (IDF), approximately 537 million adults worldwide were living with diabetes in 2021, with type 2 diabetes mellitus (T2DM) accounting for over 90% of cases, primarily driven by obesity [1]. Furthermore, patients with T2DM face a distinctively high risk of comorbidities, constituting a substantial global disease burden. Epidemiological data indicate that cardiovascular disease (CVD) affects approximately 32.2% of individuals with T2DM globally and remains the leading cause of mortality in this population [2]. Regarding microvascular complications, diabetic kidney disease (DKD) develops in approximately 20–40% of patients with diabetes [3]. These complications significantly diminish patients’ quality of life and increase mortality rates. Insulin resistance has long been regarded as the core pathophysiological mechanism of T2DM; however, emerging evidence highlights adipose tissue dysfunction as a critical endocrine driver linking obesity to this metabolic disorder [4, 5]. In the obese state, adipose tissue ceases to function merely as an energy storage depot and instead exhibits maladaptive endocrine behavior. This is characterized by macrophage infiltration and the dysregulated secretion of pro-inflammatory cytokines (e.g., TNF-α and IL-6) and adipokines [4, 6]. These inflammatory mediators act as metabolic endocrine disruptors, directly impairing insulin signaling by interfering with the phosphorylation of insulin receptor substrate-1 (IRS-1), directly inducing systemic insulin resistance [6]. This vicious cycle of chronic inflammation and insulin resistance further triggers endothelial dysfunction and oxidative stress, serving as the core pathological basis for the development of cardiovascular and microvascular complications in T2DM [7]. T2DM patients with overweight or obesity are more prone to various complications, such as cardiovascular diseases, renal lesions, retinopathy, and neuropathy. These complications not only reduce patients’ quality of life but also significantly increase their risk of mortality [8–10]. Therefore, targeting this inflammatory cascade through exercise training is crucial for preventing and managing diabetes-related complications [11].

Comprehensive management strategies for diabetes include dietary intervention, regular exercise, health education, blood glucose monitoring, and pharmacological therapy, among which exercise occupies a pivotal position in diabetes management [12]. Recent high-quality evidence has further underscored the broader physiological benefits of exercise interventions, extending to improvements in quality of life across various chronic conditions [13]. In T2DM patients specifically, regular exercise can lower patients’ blood glucose, blood pressure, and blood lipid levels, while reducing body weight, improving body composition, and enhancing quality of life [14]. Exercise exerts anti-inflammatory effects through multiple potential pathways, specifically including reducing visceral fat accumulation, inducing skeletal muscle to release anti-inflammatory myokines (e.g., contraction-induced IL-6) [15], and directly regulating immune cell function [16]. However, despite the widespread recognition of exercise’s anti-inflammatory effects, significant controversy and evidence gaps remain regarding the core question: which type of exercise yields the optimal improvement in inflammatory markers among overweight and obese T2DM patients? The main discrepancies in the current research field lie in the lack of a consistent conclusion on the anti-inflammatory effects of different exercise types: some studies have shown that 24 weeks of tai chi exercise or tai chi combined with resistance training (RT) can significantly modulate TNF-α and IL-6 levels in T2DM patients [17]; other studies have indicated that moderate-intensity combined training effectively reduces TNF-α levels in middle-aged T2DM patients but has no significant effect on IL-6 or leptin levels [18]; some research has demonstrated that aerobic exercise (AE) can significantly decrease high-sensitivity C-reactive protein (hs-CRP) levels in T2DM patients with overweight and obesity [19]; yet another study found that 1-year AE improves adiponectin levels in T2DM patients but exerts no significant effect on TNF-α or IL-6 [20]. More critically, most existing studies treat the T2DM population as homogeneous, ignoring the uniqueness of the “overweight/obese” special subgroup. This population has a higher proportion of visceral fat, a more active state of inflammatory pathway activation, and its exercise response may be influenced by factors such as body weight load and metabolic reserve [21], differing from normal-weight T2DM patients [22].

Traditional pairwise meta-analysis can only conduct indirect comparisons between two exercise types and cannot quantify the relative effects and ranking of multiple exercise types within a unified statistical framework, resulting in a lack of comparability among different study conclusions [23]. As a comprehensive analytical method in the field of exercise and health, network meta-analysis (NMA) can directly or indirectly integrate and quantify relevant evidence, simultaneously evaluate the effectiveness of multiple exercise interventions, and incorporate the ranking of treatment strategies into clinical decision-making. Thus, it holds distinct advantages over traditional meta-analysis [24]. Against this background, this study aims to comprehensively search and synthesize randomized controlled trials (RCTs) worldwide that assess the effects of different exercise types (AE, RT, Combined aerobic and resistance training [CE]) on inflammatory markers (hs-CRP, TNF-α, IL-6, leptin, and adiponectin) in T2DM patients with overweight and obesity through a systematic review and NMA. Additionally, surface under the cumulative ranking curve (SUCRA) values will be used to identify potentially optimal exercise regimens for different inflammatory markers. The results of this study are expected to provide high-level evidence-based support for clinicians in formulating individualized exercise anti-inflammatory prescriptions and for public health authorities in optimizing intervention strategies for T2DM patients with obesity, ultimately helping to reduce the burden of inflammation-related complications in this population.

Methods

Registration

This systematic review and NMA were conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement Guidelines [25]. The current study was registered with the International Prospective Register of Systematic Reviews (PROSPERO), ID: CRD420251218729.

Literature search strategy

This study comprehensively retrieved relevant literature from several authoritative databases, including PubMed, Embase, Cochrane Library, and Web of Science, using a systematic electronic search strategy. For the PubMed/Cochrane Library and Embase databases, the study utilized professional terminology from Medical Subject Headings (MeSH) and Embase Tree Structures (Emtree), respectively, to ensure the comprehensiveness and accuracy of the search results. The search strategy was constructed based on the PICOS framework, with specific elements as follows: (P) Population: T2DM with overweight and obesity; (I) Intervention: Exercise regimens involving different types of exercise; (C) Comparison: Individuals who did not engage in exercise or only performed mild stretching; (O) Outcome: Changes in inflammatory markers; (S) Study design: RCT. In addition, to comprehensively capture potentially relevant literature, the study conducted manual searches of the reference lists of included studies and related reviews to identify research that might have been missed by electronic searches. This study included all eligible English-language RCTs published from database inception up to November 10, 2025. Detailed search strategies for each database are provided in Table S1.

Eligibility criteria

The literature screening and assessment process was conducted in strict accordance with the PRISMA 2020 statement [26]. Two independent reviewers systematically screened the database search results by title and abstract based on predefined inclusion and exclusion criteria to identify potentially relevant studies. Subsequently, full texts of studies that initially met the inclusion criteria were retrieved and independently evaluated by the same two reviewers. Any discrepancies between reviewers regarding study eligibility were resolved through consultation or by involving a third independent researcher to reach a final consensus.

Inclusion criteria were as follows: (1) Study design: RCTs. Non-RCTs, quasi-experimental studies, and single-arm trials were explicitly excluded; (2) Language: English-language literature; (3) Population: T2DM patients with overweight or obesity, with a BMI > 25 kg/m2 (overweight) and/or BMI ≥ 30 kg/m2 (obese) for European participants, BMI ≥ 24 kg/m2 (overweight) and/or BMI ≥ 28 kg/m2 (obese) for Asian participants (if BMI was not provided in a study, body fat percentage [BF%] was used as a criterion [BF% ≥ 30 for females and BF% ≥ 25 for males]), and no restriction on the age in the included studies was applied [27]; (4) Regarding intervention duration, no specific minimum restriction was imposed to maximize the inclusion of relevant evidence and capture both short- and long-term inflammatory responses; (5) Standard Care: The patient maintains their existing, stable medication regimen and daily dietary habits, which remain unchanged throughout the trial period; (6) Outcome measures: Including hs-CRP, TNF-α, IL-6, leptin, and adiponectin. Detailed classifications of exercise types are provided in Table S2.

Exclusion criteria included: (1) Duplicate publications, letters to the editor, theses/dissertations, and animal studies; (2) Non-original studies (e.g., reviews, conference abstracts, case reports); (3) Exercise interventions combined with other interventions (To minimize heterogeneity, lifestyle interventions such as dietary modifications were also excluded.); (4) Studies with incomplete descriptions of participant characteristics or intervention protocols; (5) Studies where full texts were unavailable or studies without extractable post-intervention inflammatory biomarker data.

Data extraction

Data extraction was independently performed by two researchers. Any discrepancies were resolved through joint consultation or by consulting a third researcher to reach a consensus. Specific extracted information included: lead authors, year of publication; baseline characteristics of participants (sample size, age, BMI, and other relevant physiological indicators of the experimental and control groups); details of the intervention protocols (exercise intensity, exercise type, single session duration, weekly frequency, and total intervention period); and primary outcome measures. For studies with incomplete information, corresponding authors of the original studies were contacted via email when necessary to obtain supplementary data and complete the missing information.

Data extraction and quality assessment

Two investigators independently evaluated the risk of bias (ROB) of the included studies according to the Cochrane Risk of Bias Tool [28]. This tool covers seven key areas to ensure a comprehensive and accurate assessment, specifically: (a) allocation generation, (b) concealment of allocation, (c) blinding of participants and personnel, (d) blinding of outcome assessment, (e) incomplete outcome data addressed, (f) freedom from selective reporting bias, and (g) other forms of bias.

Data synthesis and statistical analyses

To minimize the impact of baseline differences, this study pooled effect sizes using changes in mean values and standard deviations (SDs) before and after intervention. The calculation of changes in SDs was converted using the formula provided in the Cochrane Handbook for Systematic Reviews of Interventions (Version 6.3) [23]. In accordance with the PRISMA Guidelines for NMA [29], effect size pooling and calculation of 95% confidence intervals (CIs) were performed in Stata 17.0 software using a random-effects model within the frequentist framework [30]. For multi-arm trials, we accounted for the correlation between effect sizes from the same study by adjusting standard errors, treating them as a series of pairwise comparisons within the network structure. Due to inconsistent measurement units of outcome measures, the standardized mean difference (SMD) was adopted as the pooled effect size.

A network plot was generated to describe the relationships between exercise interventions, where lines connecting nodes represent direct comparisons between interventions. The thickness of the lines is proportional to the number of studies, and the size of the nodes is proportional to the total sample size. Assessment of Transitivity and Inconsistency. The transitivity assumption was carefully assessed by comparing the distribution of potential effect modifiers across direct comparisons. We conceptually evaluated whether participants were sufficiently similar to be jointly randomized. Key effect modifiers, including mean age, BMI, baseline inflammatory levels, and intervention duration, were extracted (Table S3) and visually inspected. The consistency of each closed loop was evaluated by calculating the inconsistency factor (IF) and its 95% CI [31]. The consistency model is then utilized to analyze the inconsistency; if P > 0.05, the inconsistency is not significant [32]. The surface under the cumulative ranking curve (SUCRA) was used to rank and compare the intervention effects of different exercise types [33]. SUCRA values range from 0 to 1, with a value of 1 indicating the best intervention effect and 0 indicating the worst [34]. Funnel plot analysis was conducted to assess the presence of publication bias or small-sample effects.

Results

Literature selection

The study selection process is illustrated in Fig. 1. A total of 1432 relevant studies were retrieved from electronic databases, with an additional 2 studies identified through the reference lists. After removing duplicate records, 1228 studies remained. Subsequent screening based on titles and abstracts yielded 219 studies, which were then retrieved in full text for in-depth evaluation. A total of 148 ineligible studies were excluded following the full-text assessment, and 71 studies were finally included in the quantitative synthesis.

Fig. 1.

Flow chart of the study selection process according to PRISMA guidelines, detailing the number of records identified, screened, assessed for eligibility, and included in the final synthesis

Characteristics of the included studies

The characteristics of the included studies are detailed in Table S3. These studies were published between 2003 and 2025. The experimental groups included three exercise modalities: AE, RT, and CE. Specifically, the AE group comprised 1020 participants, the RT group 387 participants, the CE group 966 participants, and the control group 1893 participants. The mean age ranged from 21 to 71 years. Regarding gender distribution, some studies included only females, some included only males, some included both genders, and the gender ratio of the remaining studies was not reported. All participants in the included studies suffer from T2DM and obesity or overweight. 75 intervention groups had a duration exceeding 12 weeks. Regarding exercise frequency, the intervention groups performed 2 to 6 exercise sessions per week, with the most common frequency being three sessions per week. Notably, some studies adopted a progressive load training protocol, which made it difficult to distinguish the exercise intensity among the AE, RT, and CE groups. In this study, participants in the control group only received usual care and did not engage in any form of exercise. The inflammatory markers reported in the included studies are detailed in Table S3. Given that some inflammatory markers (e.g., WBC, PCT) did not meet the requirements of NMA, we decided to focus on in-depth analysis of the following key inflammatory markers: hs-CRP, TNF-α, IL-6, leptin, and adiponectin.

Methodological quality assessment

Figure S1 details the overall results of the risk of bias assessment, visually presenting the risk of bias levels of individual studies in a graphical format. Figure S2 evaluates seven domains: generation of random allocation sequence, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, handling of incomplete outcome data, avoidance of selective reporting bias, and other sources of bias.

A comprehensive rating was assigned to each study based on the risk assessment results for each item, with the rating method referring to existing studies: studies were classified as high quality if there was no high risk in any of the above domains and the number of unclear risks was ≤ 3; classified as moderate quality if there was one high risk, or no high risk but the number of unclear risks was ≥ 4; and classified as low quality in all other cases [35]. It is worth noting that, given the particular nature of exercise intervention studies, both participants and staff face a high risk of unblinding. Drawing upon prior research experience, this aspect has therefore been excluded from the comprehensive rating in this paper [27]. There were 57 high-quality studies, 4 moderate-quality studies, and 10 low-quality studies. Regarding the robustness of these findings, the influence of risk of bias must be considered. A notable proportion of included studies were rated as ‘high risk’ or ‘unclear’ for performance bias (due to lack of blinding), which may theoretically inflate the pooled effect estimates. However, due to the limited number of studies available for each direct comparison, sensitivity analyses excluding high-risk studies were not feasible. Thus, findings should be interpreted with this limitation in mind.

Network meta-analysis

The NMA comparison network plot is presented in Fig. 2. Contribution plots of direct and indirect comparisons among all groups, along with relevant information, are shown in Figure S3. Forest plots of eligible comparisons can be found in Figure S4. Results of the global inconsistency test are reported in Table S4, loop inconsistency results are detailed in Table S5, and node-splitting analysis results are displayed in Table S6. Global inconsistency tests revealed no significant differences for any outcome (P > 0.05, Table S4). Furthermore, local inconsistency assessments via node-splitting models showed no significant discrepancies between direct and indirect evidence (P > 0.05, Table S5, Table S6). Figure 3 presents the complete NMA results matrix, which includes relative efficacy data for comparisons between all groups. The funnel plots in Figure S5 show that the distribution of each indicator is basically symmetric, indicating a low risk of publication bias. SUCRA values and mean rankings of interventions are included in Figure S6.

Fig. 2.

Diagram of network meta-analysis comparisons. Each circle (node) represents an intervention type, with the size proportional to the total number of participants. The lines (edges) connecting the nodes represent direct comparisons, with the thickness proportional to the number of studies. hs-CRP: high-sensitivity C-reactive protein; TNF-α: tumor necrosis factor-alpha; IL-6: interleukin-6; CON: control; AE: aerobic exercise; RT: resistance training; CE: combined aerobic and resistance training

Fig. 3.

Network meta-analysis matrix of results. This inverted triangle plot shows the relative efficacy of each pair of comparisons, showing the effect of the type of intervention in each row compared to the type of intervention in each column. Data are presented as SMD (95% CI). hs-CRP: high-sensitivity C-reactive protein; TNF-α: tumor necrosis factor-alpha; IL-6: interleukin-6; CON: control; AE: aerobic exercise; RT: resistance training; CE: combined aerobic and resistance training. *The comparative groups with significant effects are marked in bold

hs-CRP

This study pooled data from 29 studies to statistically analyze hs-CRP levels in 1200 participants in the experimental groups and 921 in the control group. Statistical results showed that compared with the control group, both the AE group (SMD = -1.18; 95% CI: -1.81, -0.55; P < 0.001) and the CE group (SMD = -1.14; 95% CI: -1.88, -0.40; P = 0.003) resulted in a significant reduction in hs-CRP levels (Fig. 3). The effect of RT did not reach significant levels (SMD = − 0.50; 95% CI: -1.38 to 0.38; P = 0.269). AE emerged as the exercise intervention with the highest probability of reducing hs-CRP levels, achieving a SUCRA value of 81.7 (Figure S6).

TNF-α

This study pooled data from 27 studies to statistically analyze TNF-αlevels. Results showed that compared with the control group, AE (SMD = -0.79; 95% CI: -1.65, 0.07; P = 0.072), RT (SMD = -0.30; 95% CI: -1.47, 0.88; P = 0.619), and CE (SMD = -0.41; 95% CI: -1.22, 0.40; P = 0.322) all exerted no significant effect on TNF-α levels.

IL-6

This study pooled data from 26 studies to statistically analyze IL-6 levels in 663 participants in the experimental groups and 580 in the control group. Results showed that compared with the control group, both AE (SMD = -0.85; 95% CI: -1.54, -0.15; P = 0.017) and CE (SMD = -0.93; 95% CI: -1.63, -0.23; P = 0.01) resulted in a significant reduction in IL-6 levels. In contrast, RT did not show a reduction effect (SMD = -0.54; 95% CI: -1.62, 0.54; P = 0.328) (Fig. 3). With a SUCRA value of 76.6, CE may represent the exercise intervention with the highest probability of reducing IL-6 levels (Figure S6).

Leptin

This study pooled data from 18 studies to statistically analyze leptin levels, including 500 participants in the experimental groups and 339 in the control group. Results showed that compared with the control group, both AE (SMD = -0.71; 95% CI: -1.05, -0.37; P < 0.001) and CE (SMD = -0.55; 95% CI: -0.98, -0.12; P = 0.012) resulted in a significant reduction in leptin levels. In contrast, RT exerted no significant effect on reducing leptin levels (SMD = 0.27; 95% CI: -0.39, 0.92; P = 0.427) (Fig. 3). The SUCRA probability ranking results indicate that AE (SUCRA = 91.5) is the exercise intervention with the highest probability of reducing leptin levels (Figure S6).

Adiponectin

This study pooled data from 26 studies to statistically analyze adiponectin levels in 692 participants in the experimental groups and 582 in the control group. Results showed that compared with the control group, AE (SMD = 1.13; 95% CI: 0.42, 1.84; P = 0.002) resulted in a significant increase in adiponectin levels. In contrast, neither RT (SMD = 0.78; 95% CI: -0.88, 2.44; P = 0.357) nor CE (SMD = 0.91; 95% CI: -0.13, 1.96; P = 0.087) exerted a significant effect on increasing adiponectin levels (Fig. 3). The SUCRA probability ranking results indicate that AE (SUCRA = 76.7) is the exercise intervention with the highest probability of increasing adiponectin levels (Figure S6).

Discussion

The NMA synthesized data from 71 RCTs involving 4,266 participants to investigate the specific effects of different exercise types on inflammatory markers (including hs-CRP, TNF-α, IL-6, leptin, and adiponectin) in overweight and obese individuals with T2DM. The findings indicated that AE demonstrated the most favorable effects in reducing hs-CRP and leptin levels while increasing adiponectin. CE showed the greatest potential for lowering IL-6 levels. RT failed to achieve significant effects for most inflammatory markers, and TNF-α levels remained unchanged across all exercise interventions.

The results of this study reveal the specificity of different exercise types in regulating inflammatory markers. The superiority of AE in reducing hs-CRP and leptin, as well as increasing adiponectin, may stem from its ability to effectively promote negative energy balance and reduce visceral fat accumulation. From an endocrine perspective, this reduction in visceral adiposity alleviates adipose tissue dysfunction, thereby downregulating the secretion of adipose-derived inflammatory factors and helping to restore systemic metabolic endocrine pathways [36, 37]. Specifically, hs-CRP, as a systemic inflammatory marker synthesized by the liver, is sensitive to overall metabolic improvements [38]; A significant reduction in hs-CRP levels is directly associated with a decreased risk of cardiovascular events (such as myocardial infarction and stroke) and improved all-cause mortality in T2DM [39]. The regulatory effect of exercise on hs-CRP is highly significant. However, it is important to note that the majority of included trials had a high risk of performance bias due to the lack of blinding, which may influence the certainty of these findings. This may be attributable to the fact that, compared to the general population, this cohort exhibits higher baseline inflammatory levels, thereby possessing greater potential for improvement through exercise-induced fat reduction and anti-inflammatory interventions. Leptin and adiponectin, as key adipokines secreted by white adipose tissue (WAT), serve as pivotal endocrine signals linking energy stores to metabolic and immune homeostasis [40]. Leptin is positively correlated with BMI [41], and leptin levels are significantly elevated in obese individuals. Reduced levels of leptin are typically accompanied by a decrease in body fat mass, which helps to improve the obesity-associated state of “leptin resistance”, restoring the hypothalamus’s normal regulation of appetite and metabolic homeostasis in peripheral tissues [42]. Elevated adiponectin levels exert distinct insulin-sensitising, anti-inflammatory, and anti-atherosclerotic effects, constituting a significant protective factor in preventing diabetic vascular complications [43]. Adiponectin can improve insulin sensitivity and prevent elevated blood glucose and fat accumulation. AE-induced fat reduction and enhanced insulin sensitivity can directly inhibit hs-CRP synthesis [44, 45]. The significant leptin-lowering effect of AE may be associated with reduced fat mass and improved leptin signaling pathways [46–48], while the elevation of adiponectin levels reflects AE-induced enhancements in fatty acid oxidation and anti-inflammatory metabolic adaptation via the AMPK pathway [49–51].

Exercise-induced reduction in IL-6 levels is clinically meaningful as it contributes to alleviating systemic low-grade inflammation, potentially reducing the risk of chronic complications, thereby potentially mitigating the risk of insulin resistance progression and protecting pancreatic β-cell function [7]. The top ranking of CE in reducing IL-6 levels suggested the potential role of a synergistic mechanism. IL-6 is a multifunctional cytokine acting through complex autocrine, paracrine, and endocrine pathways. It is secreted primarily by immune cells, adipocytes, myocytes, endocrine cells, and the liver [43]. It plays a crucial role in transmitting information, activating, and regulating immune cells [52], and its expression is upregulated in obese individuals [53]. IL-6 plays a dual role in exercise: it can act as a pro-inflammatory cytokine and also be released by muscle contraction as an anti-inflammatory signal. By combining the metabolic benefits of AE and the myokine-inducing capacity of RT, CE may increase muscle mass and optimize the transient release of myokine IL-6, thereby achieving a net reduction in circulating IL-6 levels during long-term interventions. However, the insignificant effect of RT alone is inconsistent with the findings of some previous studies [54]. This may be due to its weak direct effect on reducing visceral fat [55] or an insufficient intervention dose to activate systemic anti-inflammatory pathways [56]. Some studies suggested that IL-6 released during a single RT session is mainly confined to local muscles, and the amount entering the circulatory system may not be sufficient to cause significant changes in systemic inflammatory markers [57]. Additionally, research indicated that RT intensity needs to be ≥ 75% one-repetition maximum (1RM) to effectively activate systemic anti-inflammatory pathways, and low-intensity RT may not be sufficient to trigger a significant anti-inflammatory response [58].

TNF-α, as a pro-inflammatory cytokine, is involved in inflammatory reactions and immune responses [59]. Increased TNF-αlevels reduce insulin sensitivity and promote fat accumulation [60]. Excessive accumulation of adipose tissue, which is extensively infiltrated by macrophages [61], leads to the release of free fatty acids (FFA) [22, 62] and induces the phenotypic switch of macrophages from the anti-inflammatory M2 phenotype to the pro-inflammatory M1 phenotype [63]. Elevated numbers of M1 macrophages also contribute to excessive TNF-α accumulation in obese individuals [22]; thus, TNF-α levels in obese individuals are typically higher than those in individuals with normal weight [64]. TNF-αis involved in the regulation of insulin resistance and dyslipidemia, and is one of the risk factors for T2DM. Despite being a core mediator of obesity-related inflammation, the response of TNF-αto exercise interventions may be modulated by exercise intensity and individual baseline inflammatory status [65, 66]. Some studies suggested that high-intensity exercise may temporarily reduce TNF-α [67], while most studies included in this analysis adopted moderate-to-low intensity protocols. Furthermore, the differential responses of IL-6 and TNF-α highlight the complexity of the inflammatory network: as a “myokine,” IL-6 may be more susceptible to regulation by exercise modalities [68, 69], whereas TNF-α is relatively stable and requires longer-term or combined interventions to induce changes [70].

The findings of this study shared similarities with some previous research [71–74], all supporting the important role of AE in improving metabolic inflammation in T2DM patients. However, discrepancies exist, such as the effect of RT on inflammatory markers [75]. This may be because the impact of exercise interventions on inflammatory markers is influenced by multiple factors, including gender, age, health status, characteristics of the intervention population, type, intensity, and duration of exercise intervention, as well as the characteristics of various inflammatory markers [66, 70, 76, 77].

Clinicians should adopt a stratified strategy when formulating exercise prescriptions for overweight and obese T2DM patients. The necessity of comparing different exercise modalities lies in the distinct physiological, mechanical, and functional adaptations they induce. AE exerts anti-inflammatory effects primarily by enhancing mitochondrial biogenesis, increasing fatty acid oxidation, and regulating systemic metabolic homeostasis [78]. In contrast, RT focuses on improving inflammatory status by increasing skeletal muscle mass and inducing the release of anti-inflammatory myokines [44]. CE seeks to integrate the metabolic and mechanical advantages of both modalities [79]. The preferred option is AE (e.g., brisk walking, cycling) to optimally improve hs-CRP, leptin, and adiponectin levels; the alternative option may consider CE, especially for individuals with elevated IL-6 or at risk of sarcopenia; RT is recommended as an adjunctive measure rather than a core anti-inflammatory intervention. A profound understanding of these differential benefits not only enables clinicians to develop more targeted anti-inflammatory recommendations for T2DM patients with overweight or obesity but also caters to diverse functional needs and psychological preferences. Given the low levels of physical activity participation globally, highlighting the specific benefits of various exercise modalities is of profound public health significance for enhancing long-term engagement and exercise adherence among this high-risk population [80].

Strengths and limitations

This study represents the first NMA evaluating exercise interventions on inflammatory markers in overweight and obese individuals with T2DM. It strictly adhered to the PRISMA-NMA guidelines and employed SUCRA probability ranking to quantify intervention effects, thereby significantly elevating the level of evidence. Several inherent limitations necessitate a cautious interpretation of these findings. First, constrained by the limited number of original studies, this research was unable to conduct stratified analyses according to gender, age group, exercise intensity, or disease severity. Should such stratified analyses have been performed, the excessively small sample sizes within each subgroup would likely have increased heterogeneity, thereby resulting in unstable estimates. Second, regarding the network geometry, although we assessed the transitivity conceptually, the variability in baseline characteristics (e.g., disease duration, baseline inflammatory status) across comparisons implies that potential threats to the transitivity assumption cannot be entirely ruled out. Third, methodological heterogeneity remains a concern. This includes differences in exercise protocols, inconsistencies in intervention adherence, and the inherent variability of inflammatory marker assays across laboratories. Moreover, the predominance of short- to medium-term trials limits our ability to predict long-term endocrine outcomes and sustained metabolic benefits. Fourth, regarding the certainty of evidence, most included trials were rated as ‘high risk’ for performance bias due to the lack of blinding. Collectively, these factors, along with potential publication bias, may partially inflate the observed efficacy. Consequently, SUCRA values should be regarded solely as a reference for probabilistic ranking rather than definitive evidence of clinical superiority, and effect sizes must be considered. Finally, this study did not incorporate lifestyle interventions encompassing dietary changes. Numerous existing studies have reported that combined dietary and exercise interventions yield more pronounced improvements in the inflammatory status of T2DM patients with overweight and obesity compared to exercise interventions alone. Future studies should conduct in-depth research in this field to explore optimal, comprehensive lifestyle prescriptions.

Conclusions

This study demonstrates that AE exhibits the highest probability of effectiveness in reducing hs-CRP and leptin levels as well as increasing adiponectin levels. At the same time, CE appears to offer the greatest potential in lowering IL-6 levels. RT fails to achieve significant efficacy in regulating most inflammatory markers, and TNF-α levels show no significant changes following all exercise interventions. Future RCTs should incorporate a broader range of participants, stratified according to gender, age, and disease severity or degree of obesity. Given the holistic nature of lifestyle interventions, future research may integrate lifestyle control measures such as calorie restriction or anti-inflammatory diets.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Mengjing Dong: conceptualization, methodology, formal analysis, writing - original draft; Xu Zhang: conceptualization, supervision, writing - original draft, writing - review and editing; Simeng Lu: conceptualization, methodology.

Funding

None.

Data availability

Data is provided within the manuscript or supplementary information files.

Declarations

Ethics approval and consent to participate

Not applicable. This study is a systematic review and network meta-analysis based on previously published literature and does not involve the recruitment of human participants or animal subjects. Therefore, the study was conducted in accordance with the ethical standards of the Declaration of Helsinki. As no original or individual-level data were collected, formal approval from an Institutional Review Board (IRB) or an Ethics Committee was not required for this research. The study protocol was registered with PROSPERO (ID: CRD420251218729) and strictly followed the PRISMA guidelines.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.