Exercise and Antitumor Immunity: Molecular Signaling, Immune Dynamics, and Remodeling of the Tumor Microenvironment
Epidemiological studies associate higher levels of habitual physical activity with lower cancer risk and improved outcomes in several patient populations.
Recent research has focused on exercise as a host-directed physiological intervention capable of modulating cancer-related immune, metabolic, and vascular processes. Epidemiological data consistently link higher levels of habitual physical activity with reduced cancer risk and better clinical outcomes across diverse patient cohorts.
Preclinical investigations have identified candidate mechanisms involving catecholamines, cytokines, metabolites, extracellular vesicles, immune-cell trafficking, and tumor perfusion. These molecular pathways suggest that exercise may influence the biological environment surrounding tumors through systemic and local signaling networks.
Acute exercise sessions transiently mobilize natural killer cells and cytotoxic T cells, while repeated training regimens appear to modify systemic inflammation, metabolic homeostasis, and vascular function. Evidence from selected animal models further suggests these responses can alter immune-cell infiltration, tumor metabolism, and treatment responsiveness in preclinical settings.
However, direct evidence of sustained remodeling of the human tumor microenvironment remains limited. This review integrates molecular, cellular, and microenvironmental findings while distinguishing established observations from model-dependent mechanisms and proposed cross-tissue pathways. The authors emphasize heterogeneity in exercise regimens and candidate response biomarkers.
Standardized exercise reporting, temporally resolved biosampling, and paired systemic and tumor-level analyses are needed to determine when and in which patients exercise produces biologically meaningful antitumor effects. Future research must prioritize mechanism-informed clinical trials to validate these findings beyond preclinical models.