Objective To use a meta-analytic technique to estimate the survival of implants between immediate loading/early loading and delayed loading. Methods We carried out a systematic search of electronic databases for all prospective trials comparing conventional delayed implant loading with early or immediate implant loading, reported between 1997 and 2007. The outcome of interest was implant failure rate. Quality assessment was performed for prospective trials that met the eligibility criteria and the data were then extracted and analyzed. Results Sixteen articles were found to meet the eligibility criteria, but two studies were reported in four articles, so that 14 articles were analyzed. There were five randomized controlled trials (RCTs). Compared to delayed loading, implant failure occurred sl ightly, but not statistically significant,less often with early implant loading (OR=0.54, 95%CI 0.22 to 1.33, P =0.18). We combined all cohort studies and this analysis was consistent with this result. Immediate implant loading was associated with slightly, but not statistically significant, worse outcomes (OR=1.51, 95%CI 0.53 to 4.25, P =0.44). We only pooled the RCTs and results showed better implant success rate with immediate implant loading, but with no significant difference. When sensitivity analyses were performed by the sequential dropping of a single study, no significant differences were observed except when the study of Jo et al was excluded. Conclusion Early implant loading was associated with better outcomes compared to delayed loading when the implants were placed into good quality bone. There was no significant difference of the implant success rate between immediate loading and delayed loading. Further evaluations in adequately powered large RCTs are needed to confirm these findings.