The purpose of the present meta-analysis was to test the null hypothesis of no difference in the implant failure rate, postoperative infection, and marginal bone loss for patients being rehabilitated with immediately loaded non-submerged dental implants or delayed loaded submerged implants, against the alternative hypothesis of a difference. An electronic search without time or language restrictions was undertaken in March 2014. Eligibility criteria included clinical human studies, either randomized or not. The search strategy resulted in 28 publications. The inverse variance method was used for a random- or fixed-effects model, depending on the heterogeneity. The estimates of an intervention were expressed as the risk ratio (RR) and mean difference (MD) in millimetres. Twenty-three studies were judged to be at high risk of bias, one at moderate risk of bias, and four studies were considered at low risk of bias. The difference between procedures (submerged vs. non-submerged implants) significantly affected the implant failure rate (P = 0.02), with a RR of 1.78 (95% confidence interval (CI) 1.12-2.83). There was no apparent significant effect of non-submerged dental implants on the occurrence of postoperative infection (P = 0.29; RR 2.13, CI 0.52-8.65) or on marginal bone loss (P = 0.77; MD -0.03, 95% CI -0.23 to 0.17).