Background: There is a growing need to offer appropriate services to persons with mild cognitive impairment (MCI) and dementia who are faced with depression and anxiety distresses beyond traditional pharmacological treatment. Dance-based interventions as multi-dimensional interventions address persons' physical, emotional, social, and spiritual aspects of well-being. However, no meta-analysis of randomized controlled treatment trials (RCTs) has examined the effectiveness of dance-based interventions on depression and anxiety among persons with MCI and dementia, and the results of RCTs are inconsistent. The study aimed to examine the effectiveness of dance-based interventions on depression (a primary outcome) and anxiety (a secondary outcome) among persons with MCI and dementia.Methods: A systematic review with meta-analysis was conducted. The inclusion criteria were: population: people of all ages with MCI and dementia; intervention: dance-based interventions; control group: no treatment, usual care, or waiting list group; outcome: depression and anxiety; study design: published or unpublished RCTs. Seven electronic databases (Cochrane, PsycINFO, Web of Science, PubMed, EBSCO, CNKI, WanFang) were searched from 1970 to March 2021. Grey literature and reference lists from relevant articles were also searched and reviewed. The Cochrane “Risk of Bias” tool was used to assess study quality. RevMan 5.4 was used for meta-analysis and heterogeneity was investigated by subgroup and sensitivity analysis. GRADE was applied to assess the evidence quality of depression and anxiety outcomes.Results: Five randomized controlled trials were identified. Sample sizes ranged from 21 to 204. The risk of bias was low, except for being rated as high or unclear for most included studies in two domains: allocation concealment, blinding participants and personnel. Meta-analysis of depression outcome showed no heterogeneity (I2 = 0%), indicating that the variation in study outcomes did not influence the interpretation of results. There were significant differences in decreasing depression in favor of dance-based interventions compared with controls [SMD = −0.42, 95% CI (−0.60, −0.23), p < 0.0001] with a small effect size (Cohen's d = 0.3669); Compared with the post-intervention data, the follow-up data indicated diminishing effects (Cohen's d = 0.1355). Dance-based interventions were more effective in reducing depression for persons with dementia than with those having MCI, and were more effective with the delivery frequency of 1 h twice a week than 35 min 2–3 times a week. Also, one included RCT study showed no significant benefit on anxiety rating scores, which demonstrated small effect sizes at 6 weeks and 12 weeks (Cohen's d = 0.1378, 0.1675, respectively). GRADE analysis indicated the evidence quality of depression was moderate, and the evidence quality of anxiety was low.Conclusions: Dance-based interventions are beneficial to alleviate depression among persons with MCI and dementia. More trials of high quality, large sample sizes are needed to gain more profound insight into dance-based interventions, such as their effects of alleviating anxiety, and the best approaches to perform dance-based interventions.