Background: Exercise interventions are commonly considered as non-pharmacological approaches to support cognitive and functional outcomes in older adults with dementia. However, the effects reported in the literature remain heterogeneous, and commonly used time-based dose markers may be insufficient to explain variability across trials. Methods: A systematic review and meta-analysis of randomized controlled trials was conducted in accordance with PRISMA 2020 guidelines. Eligible trials described benefits with cognitive, functional or behavioral changes associated with structured exercise interventions in older adults with dementia. Random-effects meta-analysis and meta-regression models were used to derive pooled effects and assess if linear dose indicators (e. g. , duration of intervention, session length, frequency and total cumulative dose) reflected heterogeneity. Results: Twenty-two studies were analyzed. Based on our pooled analyses, a small but statistically significant improvement was observed under the fixed-effects model (g = 0. 106, 95% CI 0. 015–0. 197; p = 0. 023), but this was not significant for random-effects models (g = 0. 117, 95% CI −0. 021–0. 254; p = 0. 097), while suggesting moderate between-study heterogeneity (Q (21) = 43. 530, p = 0. 003; I2 = 51. 757%; τ2 = 0. 052). For the main random-effects meta-regression, standard linear dose indicators did not significantly explain between-study heterogeneity (Qm (3) = 1. 06, p = 0. 7867; R2ₐnalog ≈ 0), while significant residual heterogeneity remained (I2 ≈ 56. 03%). Conclusions: In the literature so far, there are limited and heterogeneous effects of exercise interventions on cognition and functions in older adults with dementia. These findings in all literature suggest that the current evidence does not support a consistent linear dose–response relationship but rather will likely depend to some extent on feasibility and supervision (again, quality of the interventions), thus emphasizing that exercise strategies must be contextually sensitive rather than dose-dependent.
Lu et al. (Mon,) studied this question.