The aim of this study was to compare implant failure and radiographic bone level changes with different loading protocols for unsplinted two-implant-supported mandibular overdentures. An electronic search of two databases (PubMed, Cochrane Library) was performed, without language restriction, to identify randomized controlled trials (RCTs) comparing immediate or early versus conventional dental implant loading protocols for unsplinted two-implant-supported mandibular overdentures. Data were extracted independently by two reviewers. The Cochrane tool was used to assess the quality of included studies. A meta-analysis was performed. Eight RCTs were identified, seven of which were included; one trial was excluded because related outcomes were not measured. Four of the seven studies were considered to have a high risk of bias and three an unclear risk. Meta-analysis revealed no difference between immediate versus conventional or early versus conventional implant loading protocols regarding implant failure (risk difference (RD) -0.02, 95% confidence interval (CI) -0.13 to 0.10; RD 0.09, 95% CI -0.03 to 0.20) or marginal bone loss (mean difference (MD) 0.09, 95% CI -0.10 to 0.28; MD -0.05, 95% CI -0.12 to 0.02) for implants supporting mandibular overdentures. These findings should be interpreted with great caution given the serious numerical limitations of the studies included.