Background: Dementia is one of the costliest diseases for health care systems with growing importance for policy makers. Objective: The aim of this study is to systematically review the current literature of excess cost studies for dementia and to analyze excess costs in a meta-analysis. Methods: A systematic literature search was conducted in PubMed, EconLit, NHS-EED, and Cochrane Library. 22 studies were included and assigned to one of three subgroups according to the time period that they analyzed during disease progression: the time of diagnosis, the time between diagnosis and death, and the time prior to death. Excess costs were analyzed using the ratio of means (ROM) and meta-analysis was performed by pooling ROMs in a random effects model. Results: Total costs were significantly higher for demented persons compared to non-demented persons at the time of diagnosis (ROM: 2.08 [1.71, 2.54], p < 0.00001, I 2 = 98%) and in the time period between diagnosis and death (ROM: 2.19 [1.97, 2.44], p < 0.00001, I 2 = 100%). The ROM was highest for professional home care (ROM: 4.96 [2.62, 9.40], p < 0.0001, I 2 = 88%) and for nursing facilities (ROM: 4.02 [2.53, 6.40], p < 0.00001, I 2 = 100%) for the time period between diagnosis and death. Conclusion: This meta-analysis is the first to assess excess costs of dementia by the ROM method on a global scale. We conclude that our findings demonstrate that costs of dementia constitute a substantial economic burden.