Purpose: To compare changes in the prevalence of peri-implantitis when the unit of analysis is the subject and when the unit of analysis is the implant, by means of meta-analysis with subgroup and sensitivity analyses, according to the case definition and cutoffs. Periodontal probing depth (PPD) and bleeding on probing (BOP) were considered the primary variables. Materials and Methods: Electronic and manual searches of observational studies of implants with loading of more than 6 months were conducted. The quality of the studies was evaluated, and finally, a description (qualitative analysis) and a meta-analysis (quantitative analysis) of the available studies were performed. Results: Fifty-five studies were included in this systematic review, 32 of which met the criteria for evaluation of disease based on PPD and BOP. A total of 2,734 subjects and 7,849 implants were evaluated. The prevalence of peri-implantitis, defined by PPD and BOP, was 17% when the unit of analysis was the subject, and 11% when it was the implant. If the clinical criterion was PPD ≥ 4 mm, the prevalence by subject was 34% and by implant 11%. If PPD was ≥ 5 mm, the prevalence by subject was 12% and by implant 10%. Finally, if the clinical criterion was PPD ≥ 6 mm, the prevalence by subject was 18% and by implant 10%. Conclusion: The prevalence of peri-implantitis is influenced by the criteria used for the case definition, and the true prevalence may currently be incorrectly estimated. Keywords: dental implants, meta-analysis, peri-implantitis, prevalence, systematic review