Background: Parkinson's disease (PD) represents the second most common neurodegenerative disease. Objective: To evaluate the effects of dance therapy (DT) aimed at improving non-motor symptoms in PD. Methods: Studies were performed through PubMed, Web of Science, The Cochrane Library, Embase, and Science Direct from inception to October 27, 2021. The data were screened independently by two reviewers, and the quality of the papers was assessed using the Cochrane manual. The included studies were randomized controlled trials and quasi-randomized controlled trials, reporting random-effects standardized mean differences, and 95% confidence intervals as the effect size. I2 statistics were used to assess heterogeneity. The main outcomes included the Montreal Cognitive Assessment Scale (MOCA), Baker Depression Scale (BDI), Parkinson's Fatigue Scale (FPS-16), and Apathy Scale (AS). RevMan 5.3 software was integrated for meta-analysis. Results: Nine literatures were analyzed for the meta-analysis with a total of 307 patients. Random effects showed that DT significantly improved cognitive of PD (MD = 1.50, 95% CI [0.52, 2.48], P = 0.0003; I2 = 51%). However, this meta-analysis demonstrated that dance therapy had no significance for improving depression (MD = - 1.33, 95% CI [- 4.11, 1.45], P = 0.35; I2 = 79%), fatigue (MD = 0.26, 95% CI [- 0.31, 0.83], P = 0.37; I2 = 0%), and apathy (MD = 0.07, 95% CI [- 2.55, 2.69], P = 0.96; I2 = 50%). Conclusion: The meta-analysis suggests that dance can improve cognitive function in PD.