The objective of the present randomized clinical trial was to test whether or not the use of screw-retained all ceramic implant-borne reconstructions results in clinical, technical, and biologic outcomes similar to those obtained with cemented all ceramic reconstructions. The hypothesis was that there is no difference in clinical, technical, and biological parameters between the two types of retention. Forty-four patients randomly received 20 cemented reconstruction (CR) and 24 screw-retained (SR) all ceramic single crowns on two-piece dental implants with nonmatching implant-abutment junctions. All patients were recalled after crown insertion, at 6 months, 1 year, as well as at 3 years. At these visits, biological and radiographic evaluations were performed. Technical outcomes were assessed using modified USPHS (United States Public Health Service) criteria. Data were statistically analyzed with Wilcoxon-Mann-Whitney, Wilcoxon and Fisher exact tests. During 3 years of follow-up, eight patients (18.2%) lost the reconstruction due to technical (6 patients, 13.6%, 2 CR and 4 SR group) or biological complications (2 patients, 4.5%, only CR group). Thirty-two subjects with 18 SR and 14 CR reconstructions attended the FU-3Y, whereas four patients (9.1%, 2 SR, 2 CR) were not available (drop-outs). Biological, technical, and radiographic outcomes did not differ significantly between the groups (P > 0.05). One implant (2.3%) was lost in the CR group. One more cemented crown (2.3%) had to be removed because of peri-implant disease. Six patients (13.6%) lost the reconstructions due to a fracture of the zirconia abutment (4 SR, 2 CR). The mean marginal bone level at 3 years was -0.4 mm (-0.5; -0.3) in group SR and - 0.4 mm (-0.6; -0.3) group CR (P = 0.864). At 3 years, CR and SR exhibited similar survival technical, biological and radiographic outcomes. The rate of technical complications was high in both groups.