Long-term exercise training and inflammatory biomarkers in healthy subjects: a meta-analysis of randomized controlled trials

Objective This meta-analysis investigated the effect of long-term exercise training (ET) including aerobic, resistance, and multicomponent ET on the levels of inflammatory biomarkers in randomized controlled trials (RCTs) involving healthy subjects. Methods We searched seven databases for articles until May 1st, 2023. A random-effect meta-analysis, subgroup analysis, meta-regressions as well as trim and fill method were conducted using STATA 16.0. Result Thirty-eight studies were included in the meta-analysis, involving 2,557 healthy subjects (mean age varies from 21 to 86 years). Long-term ET induced significantly decreased in the levels of interleukin-6 (IL-6) (SMD -0.16, 95% CI -0.30 to −0.03, p = 0.017), C-reactive protein (CRP) (SMD -0.18, 95% CI -0.31 to −0.06, p = 0.005), as well as tumor necrosis factor alpha (TNFα) (SMD -0.43, 95% CI -0.62 to −0.24, p < 0.001). Subgroup analysis revealed that Long-term ET conducted for more than 12 weeks and exercise of moderate intensity had greater anti-inflammatory effects. Meta-regression analysis showed that the reduction in CRP level induced by long-term ET was weakened by increasing exercise intensity. Conclusion Long-term ET induced significant anti-inflammatory effects in healthy subjects. Long-term ET-induced anti-inflammatory effects were associated with exercise of moderate intensity and training conducted for more than 12 weeks. Systematic Review Registration: https://www.crd.york.ac.uk/PROSPERO/# myprospero, PROSPERO, identifier CRD42022346693.


Introduction
The global physical inactivity of approximately 27.5% of adults (Guthold et al., 2018) and 81% of teenagers (Guthold et al., 2020) increases 6-10% risk of chronic diseases, premature death (Ozemek et al., 2019) and the risk of dementia (Voss et al., 2019).Lack of physical activity is one of the most common causes of chronic low-grade systemic inflammation, which is related

Study selection
Two researchers (YHW and JWT) independently screened titles and abstracts, then reviewed full-text for eligibility.The third researcher (MC) arbitrated any discrepancies to reach consensus.We also conducted a manual search for references to included articles and relevant review articles.Research and review articles were included in the study based on the following inclusion criteria: (Guthold et al., 2018) the volunteers were healthy individuals; (Guthold et al., 2020) the intervention of interest was any type of longterm ET of any intensity, frequency; (Ozemek et al., 2019) the comparisons involved ET versus non-exercise control or ET plus other intervention versus other intervention only; (Voss et al., 2019) the outcomes of interest were inflammatory biomarkers in plasma or serum; (Furman et al., 2019) parallel or crossover randomized controlled trials; and (Nieman and Wentz, 2019) articles published after 1980.The exclusion criteria included (Guthold et al., 2018) studies involving sick individuals; (Guthold et al., 2020) The intervention was the acute effect of one-time ET on inflammatory biomarkers; (Ozemek et al., 2019) studies lacking baseline data or data on the final assessment of outcomes in both intervention group and comparators used to calculate mean changes of treatment ± SD.

Data extraction and quality assessment
Data extraction and quality evaluation were conducted by two trained researchers (YHW and JWT) independently.The data extracted included the first author's surname, publication year, study design, study location, sample size, participants age and gender, baseline body mass index (BMI) of participants, ET intervention (duration, type, intensity, frequency), and outcomes of reported biomarkers.Research articles that contained two or more ET intervention strata (e.g., different type, intensity, frequency or duration of ET), were analyzed as separate trials.All types of ET (i.e., aerobic exercise, resistance exercise and multicomponent exercise) were included in this review.Intensity of ET was classified by maximal heart rate (HRmax), maximal oxygen uptake (VO 2 peak) and repetition maximum (RM) according to the guidelines by American College of Sports Medicine (Haskell et al., 2007).The methodological quality and risk of bias of each included study were evaluated using the Physiotherapy Evidence Database (PEDro) scale (Maher et al., 2003).External validity (item 1: Eligibility criteria and source), internal validity (items 2 to 9: Random allocation; Concealed allocation; Baseline comparability; Blinding of participants; Blinding of therapists; Blinding of assessors; Adequate follow-up (>85%); Intention-to-treat analysis) and statistical reporting (items 10 and 11: Between-group statistical comparisons; Reporting of point measures and measures of variability) are covered by the PEDro scale's eleven items.Items were rated yes or no (1 or 0) based on whether the study obviously meets the criterion.The cumulative PEDro score ranged from 0 to 10 and is calculated by adding the ratings of items 2 to 11.The higher the score, the superior the methodological quality, and the PEDro scale's scores of 4 were considered "poor, " 4 to 5 were considered "fair, " 6 to 8 are considered "good, " and 9 to 10 were considered "excellent" (Cashin and McAuley, 2020).
The Grading of Recommendation, Assessment, Development and Evaluation (GRADE) system was used to assess the evidence level of each outcome (Goldet and Howick, 2013).According to the GRADE guidelines, study design dictates baseline quality of the evidence (RCTs were initially defined as high quality) but other factors could decrease (e.g., unexplained heterogeneity) or increase (e.g., a large magnitude of effect) the quality of the studies (Goldet and Howick, 2013).Discrepancies were resolved through discussion with the third reviewer (MC).

Statistical analysis
There were differences among the included studies in study participants, ET protocols employed and the measurement of biomarkers.As a result, the random effects model was used to pool estimates of net changes (changes of intervention group minus changes of control group) in the concentrations of biomarkers and the results were presented as standardized mean difference (SMD) with 95% confidence intervals (CIs).Advanced data extraction was used for studies that did not directly provide either the baseline or final mean and standard deviation (SD) of outcomes, according to the protocol proposed by Wan et al. (2014).
First, a primary meta-analysis was conducted to establish the overall effect of ET on each biomarker.Then sensitivity analysis was conducted by using a leave-one-out meta-analysis (LOOM) to test the robustness of the primary results and the influence of each report on the effect or heterogeneity.Sources of potential heterogeneity were identified by carrying out subgroup analyzes based on region, gender, age of participants, baseline BMI of participants, duration of intervention, type of ET and intensity of ET (only conducted if more than six trials reported the same outcomes), and subgroup stratification was based on the characteristics of the included studies and also referred to the previous meta-analysis stratification method (Schwingshackl et al., 2014;de Souto Barreto et al., 2019).Difference between groups and sources of heterogeneity were tested using meta-regression analysis.Further meta-regression analyzes were conducted to investigate potential moderators and examine the association between moderators and outcomes, with value of p <0.1 being considered as statistically significant.The heterogeneity among studies was tested using Cochrane's Q test and quantified using I 2 -statistic (Higgins et al., 2003).Presence of heterogeneity was indicated by I 2 > 50% and p value <0.1 for Q test.Begg's and Egger's regression tests as well as funnel plots were utilized to assess publication bias, with a value of p <0.05 suggesting the presence of bias (Egger et al., 1997).If publication bias was encountered, the trim and fill method was used to remove extremely small studies and recalculate the pooled effect iteratively until the funnel plot was symmetrical (Duval and Tweedie, 2000).All analyzes were performed using STATA version 16.0 (Stata Corp, College Station, TX, United States), with double data input to avoid input errors.p < 0.05 was deemed as statistically significant unless specified elsewhere.

Qualities of included studies and outcome measure evidences by GRADE system
The quality of the methodology of three studies (Olesen et al., 2014;Dani et al., 2021;Griffen et al., 2022) was rated as excellent according to the PEDro scale (scores ≥9), while the remaining 35 studies were rated as good quality (6-8 scores) mainly due to lack of blinding (Table 1).According to the GRADE system, the quality of evidence for IL-6, CRP were moderate, and TNFα was low (Supplementary Table S2).

Characteristics of included studies
Characteristics of studies included in this meta-analysis are shown in Table 1
The result of the Begg's test (p = 0.204), Egger's test (p = 0.112) and funnel plot (Supplementary Figure S2) indicated no significant publication bias in the primary analysis for CRP (Table 2).
10. 3389/fpsyg.2023.1253329Frontiers in Psychology 14 frontiersin.orgmeta-analysis revealed that long-term ET caused a significant decrease in TNFα levels (SMD −0.43, 95% CI −0.62 to −0.24), with a high heterogeneity across the studies (p < 0.001, I 2 = 87.8%)(Table 2 and Figure 4).Subgroup analyzes demonstrated that subjects with significant reduction in TNFα levels were from studies of participants younger than 60 years, those taking aerobic exercise or resistance exercise, frequency of three or more times a week, and exercise of moderate intensity.Funnel plot (Supplementary Figure S3) and Egger's test results for TNFα (p = 0.009) revealed evidence of publication bias (Table 2).

Sensitivity analysis
The sensitivity analysis was carried out by leave-one-out metaanalysis (LOOM).The results of the sensitivity analysis indicated that the effect of long-term ET on IL-6 levels was unstable, suggesting that the findings should be handled with caution (Figure 5A).Long-term ET still had a stable significant impact on CRP (Figure 5B) and TNFα (Figure 5C) levels when a single trial was deleted at a time.

Adjustment of publication bias
As shown in Table 3, results of pooling estimates indicating publication bias based on Begg's and Egger's tests, were recalculated using Duval and Tweedie's trim and fill method.Long-term ET still induced a significant decrease in TNFα levels (SMD −1.02, 95% CI −1.41 to −0.63).

Discussion
To the best of our knowledge, this is the first meta-analysis from 38 RCTs to estimate the effect of long-term ET on individual biomarkers of inflammation among healthy subjects.Overall, our results revealed that long-term ET induced significant decrease in IL-6, CRP, and TNFα levels.Long-term ET-induced reduction of IL-6 was more evident in participants with a BMI >25, engaged in exercise for less than 12 weeks and engaged in type of multicomponent exercise.On the other hand, Long-term ET-induced decrease in CRP levels was associated with participants involved in exercise for more than 12 weeks and involved in aerobic exercise.Long-term ET-induced decrease in TNFα levels was associated with participants of younger than 60 years of age, involved in exercise for more than 12 weeks, and involved in exercise of moderate intensity.Elevated levels of circulating inflammatory markers such as IL-6, TNFα and CRP, induces the development of chronic low-grade inflammation which has been identified as a risk predictor for several diseases such as T2DM (Pradhan et al., 2001) and dementia (Leonard, 2007).Our findings showed that engaging in ET for over 12 weeks effectively reduced the levels of CRP and TNFα.This is consistent with previous studies which showed that ET interventions conducted over longer durations can minimize inflammation.A meta-analysis based on elderly participants showed that resistance training alone can reduce CRP and TNFα when conducted for more than 12 weeks (Sardeli et al., 2018).It is well known that one-time exercise interferes with cell homeostasis leading to inflammation, while repeated exercise training improves immunocompetence (Scheffer and Latini, 2020).Skeletal muscle is an endocrine organ.During muscle contraction, it can produce cytokines and release them into the blood, which can systematically reduce inflammation (Furman et al., 2019).It was reported that several myokines, such as CRP, peaked at the end of exercise and returned to baseline levels within several hours (Keller et al., 2005;Fischer, 2006), which could mediate metabolic changes followed by increase anti-inflammatory cytokine (Febbraio and Pedersen, 2002), thus activate anti-inflammatory immune response.Therefore, long-term exercise may contribute to lower basal levels of inflammatory biomarkers.
Consistent with the previous studies (Scheffer and Latini, 2020), we also found that exercise of moderate intensity was more beneficial for the reduction of CRP and TNFα levels.The intensity of training The forest plot of the effect of long-term ET on TNFα levels.has been shown to affect inflammatory markers in a dose-dependent manner (Fatouros et al., 2009).Several studies have shown that ET of moderate intensity promotes anti-inflammatory response, while ET of high intensity promotes an inflammatory response (Pinho et al., 2010;Lindsay et al., 2015).However, a recent meta-analysis based on 27 studies, including 17 studies with patients, revealed that intensity of exercise did not influence the chronic inflammatory response (Rose et al., 2021).Nevertheless, the study by Rose et al.only included studies that compared more than two different intensities of exercise, with significant heterogeneity in the exercise prescription among the studies.As a result, total exercise volume may have confounded the outcomes.Overall, our meta-analysis suggested that long-term three or more times a week of moderate intensity ET could significantly reduce the level of inflammatory biomarkers in healthy adults, and that such exercise frequency and intensity were consistent with the guidelines of the American College of Sports Medicine (Garber et al., 2011).
Aging is inherently associated with chronic increase in cytokine concentration, a condition termed as inflammaging (Franceschi and Campisi, 2014).Indeed, a meta-analysis from observational studies demonstrated that the frailty and pre-frailty in the elderly were associated with higher inflammatory parameters, especially CRP and IL-6 (Soysal et al., 2016).Therefore, the elderly are more likely to benefit from the anti-inflammatory effect of ET.Meanwhile, further studies are needed to establish the relationship between age and the anti-inflammatory effect of PE, as well as the underlying mechanisms.There are several limitations in this study.First, there was no blinding in 34 of the 38 included studies, which may have introduced some bias.The high heterogeneity observed across studies might lead to degradation of the credibility of the results.Thus, there is need to excess caution when interpreting the results of our study.The high heterogeneity may have been due to the use of different experimental designs, particularly study samples, interventions and measures of outcome, among the different studies.Although we explored the source of heterogeneity via subgroup analysis and meta-regression, some of the results were also limited by the differences in sample size between subgroups.Furthermore, some of our results from subgroup analysis showed publication bias, which may affect the validity of the effects observed.However, most of these results remained unchanged or showed greater effect size after adjustment using the trim and fill method, thus making the results credible.

Conclusion
Taken together, long-term ET induced significant decrease in IL-6, CRP, and TNFα levels in healthy subjects compared to non-active interventions.Long-term ET of moderate intensity and conducted for more than 12 weeks induced more pronounced anti-inflammatory effects in healthy subjects.There is need for future RCTs to explore the optimum long-term ET protocol beneficial for people of different ages.The results of leave-one-out meta-analysis on IL-6 (A), CRP (B), and TNFα (C).

FIGURE 1 A
FIGURE 1A flowchart showing the procedure used for study selection.

FIGURE 3
FIGURE 3The forest plot of the effect of long-term ET on CRP levels.

TABLE 1
Characteristics of included studies in this meta-analysis (38 studies).
a Alghadir et al.

TABLE 2
Results of sensitivity analysis, subgroup analysis and publication bias stratified by study characteristics.
P 1 value for net change; P 2 value for heterogeneity in the subgroup; P 3 value for heterogeneity between groups with meta-regression, analyzed as categorical variables; P 4 value for publication bias (conducted only when trials > 5); significant p-values are highlighted in bold prints.BMI, body mass index; CI, confidence interval; CRP, C-reactive protein; IL-6, interleukin 6; RC, randomized crossover; RP, randomized-parallel; TNFα, tumor necrosis factor alpha.
involving 32 interventions evaluated the effect of long-term ET on TNFα levels.Results of the

TABLE 3
Trim and fill adjusted analysis for outcomes with publication bias.Significant p-values are highlighted in bold prints.CI, confidence interval; TNFα, tumor necrosis factor alpha.