Exercise-induced anti-inflammatory adaptation in older adults: a Bayesian network meta-analysis and dose-response...

Global population aging is increasing the number of older adults living with chronic diseases and functional limitations. Chronic low-grade inflammation is common in later life and is associated with multisystem decline and age-related disease risk. C-reactive protein (CRP) and interleukin-6 (IL-6) are widely used circulating biomarkers in this context. Exercise may help modulate inflammatory biomarkers, but existing evidence remains heterogeneous across exercise modalities, prescriptions, and outcomes.

This study adhered to the PRISMA-NMA guidelines (PROSPERO: CRD420261397178). Six major databases (PubMed, Embase, Web of Science, Scopus, EBSCOhost, and Cochrane Library) were searched from inception to May 2026. Fifty-four randomized controlled trials (RCTs) enrolling 2,836 participants aged ≥60 years were included. A Bayesian random-effects network meta-analysis (NMA) was conducted using standardized mean differences (SMDs) to compare five exercise modalities: aerobic exercise (AE), resistance training (RT), combined training (CT), mind-body exercise (MBE), and multi-component exercise (MCE). A model-based network meta-analysis (MBNMA) was further constructed with weekly metabolic equivalent minutes (MET-min·week^-1) as a continuous exposure variable and a quadratic function to model non-linear dose-response relationships. Reference coding was applied to isolate baseline drift from natural aging. Spearman rank correlation between CRP and IL-6 effect estimates was assessed strictly at the within-study level.

The evidence network demonstrated satisfactory overall consistency; network meta-regression did not identify baseline disease status as a statistically significant moderator of the dose-response curve (P > 0.05). NMA suggested modality-specific anti-inflammatory patterns: AE retained the most favorable ranking for CRP reduction (SUCRA = 0.89; SMD = -0.83, 95% CrI: -1.42 to -0.24). In the primary IL-6 network, MBE ranked highly; however, influence analysis identified Ali Ismail 2025 as an extreme-effect study. After excluding this study, the MBE 900 MET-min week^-1 node lost its favorable ranking (SUCRA 0.845 to 0.415; probability of being best 0.291 to 0.035), and the estimated effect shifted toward the null (median -5.50 to 0.62). MBNMAdose suggested non-linear dose-response patterns, but Hellinger distances indicated substantial posterior divergence between common-effect and random-effects models. These findings indicate that IL-6 modality rankings, particularly those involving MBE, should be interpreted cautiously.

Exercise interventions are associated with improvements in inflammatory biomarkers among older adults, with effects varying by modality and dose. AE showed the most consistent signal for CRP reduction. The apparent MBE advantage for IL-6 was not robust to exclusion of an influential study and should be treated as exploratory rather than definitive. Dose-response findings suggest a probable moderate-dose benefit region, but credible-interval uncertainty and substantial Bayesian heterogeneity indicate that dose regions should be interpreted cautiously rather than as fixed clinical thresholds.