Abstract

Objective: To identify the comparative efficacy of exercise for reducing neuropsychiatric symptoms (NPS) among older adults with mild cognitive impairment (MCI) or dementia.

Methods: Ten databases were systematically searched from their inception to April 29, 2025, with the latest update in July 13, 2025. Randomized controlled trials (RCTs) and quasi-experimental studies evaluating the effectiveness of exercise on NPS in older adults with MCI or dementia were included. Risk of bias was assessed using the Risk of Bias tool (RoB 2) tool for RCTs and the Joanna Briggs Institute (JBI) critical appraisal checklists. A random-effects network meta-analysis model was employed to synthesize all available evidence. The registration number of this study is CRD420251087869.

Results: A total of 34 studies involving 3655 participants were included. Among them, 29 RCTs showed a low to high risk of bias, while 5 quasi-experimental studies had moderate risk. A network meta-analysis revealed that for NPS, resistance exercise (SMD = -1.70, 95 % CI: -3.12 to -0.29) ranked first with 91.8 %; for cognition function, multi-component exercise (SMD = 1.45, 95 % CI: -0.56 to -3.47) ranked first with 80.2 %; for quality of daily life (QoL), aerobic exercise (SMD = 0.37, 95 % CI: -0.26 to -0.99) ranked first with 77.5 %; and for activities of daily living (ADL), finger exercise (SMD = 0.86, 95 % CI: 0.02 to -1.71) ranked first with 89.6 %.

Conclusions: This network meta-analysis suggests that resistance exercise is most likely the optimal intervention for improving NPS among older adults with MCI or dementia, while finger exercise appears most effective for enhancing ADL. However, due to the limited number of studies on resistance exercise, these findings should be interpreted with caution, and further high-quality research is needed to confirm its efficacy.