Objectives: The aim of this umbrella review was to determine the effect of physical activity/exercise on improving cognitive and noncognitive outcomes in people with MCI (mild cognitive impairment) and dementia.

Design: Umbrella review of systematic reviews (SR), with or without meta-analyses (MAs), of randomized controlled trials (RCTs) and observational studies. SETTINGS AND

Participants: People with MCI or dementia, confirmed through validated assessment measures. Any form of physical activity/exercise was included. As controls, we included participants not following any prespecified physical activity/exercise intervention or following the same standard protocol with the intervention group.

Methods: The protocol was registered in PROSPERO (CDR 164197). Major databases were searched until December 31, 2019. The certainty of evidence of statistically significant outcomes was evaluated using the Grading of Recommendations Assessment, Development and Evaluation approach. SRs' findings, without a formal MA, were reported descriptively.

Results: Among 1160 articles initially evaluated, 27 SRs (all of RCTs, 9 without MA) for a total of 28,205 participants with MCI/dementia were included. In patients with MCI, mind-body intervention (standardized mean difference [SMD] = 0.36; 95% confidence intervals [CI] 0.20-0.52; low certainty) and mixed physical activity interventions (SMD = 0.30; 95% CI 0.11-0.49; moderate certainty) had a small effect on global cognition, whereas resistance training (SMD = 0.80; 95% CI 0.29-1.31; very low certainty) had a large effect on global cognition. In people affected by dementia, physical activity/exercise was effective in improving global cognition in Alzheimer disease (SMD = 1.10; 95% CI 0.65-1.64; very low certainty) and in all types of dementia (SMD = 0.48; 95% CI 0.22-0.74; low certainty). Finally, physical activity/exercise improved noncognitive outcomes in people with dementia including falls, and neuropsychiatric symptoms.

Conclusions and implications: Supported by very low-to-moderate certainty of evidence, physical activity/exercise has a positive effect on several cognitive and noncognitive outcomes in people with MCI and dementia, but RCTs, with low risk of bias/confounding, are still needed to confirm these relationships.