Resistance-containing exercise versus non-exercise control for inflammatory markers and T-cell counts in adults living with HIV: a systematic review and meta-analysis

Background Adults living with HIV may have persistent immune activation and low-grade inflammation despite effective antiretroviral therapy. We assessed whether resistance training or resistance-containing combined exercise alters inflammatory biomarkers or T-cell counts versus non-exercise control. Methods Six databases were searched to 1 June 2026 without language restrictions. Controlled intervention studies lasting at least four weeks were eligible. Screening and extraction were performed independently in duplicate. Inflammatory outcomes were expressed as Hedges’ g; final-value SMDs were primary and change-score SMDs were analyzed separately, with one-trial strata reported individually. CD4+ T-cell count was pooled as a mean difference. Random-effects models used restricted maximum likelihood and Hartung-Knapp-Sidik-Jonkman intervals. RoB 2 or ROBINS-I and GRADE were applied. Results Fifteen trials (18 reports) were included; 13 contributed quantitative data. All primary final-value confidence intervals included no effect: interleukin-6, −0.19 (95% CI −1.14 to 0.76; k = 4); tumor necrosis factor alpha, 0.07 (−1.44 to 1.57; k = 3); C-reactive protein, −0.02 (−1.46 to 1.41; k = 3); interleukin-1 beta, −0.47 (−10.09 to 9.15; k = 2); and interleukin-10, 0.04 (−11.37 to 11.46; k = 2). Interleukin-8 was reported by one final-value trial, −0.63 (−1.39 to 0.12), and was not pooled. Separate change-score analyses were also inconclusive except for one unpooled C-reactive protein trial. The pooled estimate for CD4+ T-cell count favored exercise by 124.2 cells/mm 3 (54.6 to 193.7; k = 7), but the prediction interval crossed no effect. Certainty was very low. Conclusion The effects of resistance-containing exercise on inflammatory biomarkers remain undetermined. The CD4+ estimate suggests a possible increase, but confidence is very limited because the evidence is sparse, heterogeneous, at risk of bias, and its prediction interval includes no effect. Larger trials with well-defined comparators and complete contrast-form reporting are needed. Systematic review registration https://www.crd.york.ac.uk/PROSPERO/view/CRD420261425700 , identifier CRD420261425700.