Purpose: To compare two different impression techniques for implants in totally edentulous patients. Materials and methods: A total of 38 patients had impressions taken both using plaster and splinted vinyl polysiloxane (splinted-VPS). Two casts per patient were generated and allocated as test (plaster) and control (splinted-VPS) cast groups according to a randomised cross-over design. One of the two casts from each patient was randomly selected as master cast according to a parallelgroup design and used to fabricate the definitive prosthesis. Outcome measures were implant and prosthetic success rates, complications, marginal bone level (MBL) changes, patient satisfaction, chair time required to take the impressions, inter-implant discrepancy between the casts, sulcus bleeding index (SBI) and plaque score (PS). Results: In total, 76 impressions were taken in 38 patients. Two plaster impressions failed. Furthermore, 38 computer-aided design/computer-assisted manufacturing screw-retained complete-arch prostheses were fabricated onto the master cast (18 from plaster and 20 from splinted-VPS impressions) and the patients were followed up for 3 years after loading. No drop-out occurred and no implants or prostheses failed, accounting for a cumulative implant and prosthesis survival rate of 100% over the 3-year post-loading period. Plaster impressions yielded significantly greater patient satisfaction and shorter chair time. The discrepancy between the casts was 0.055 ± 0.067 mm (P = 0.931). Mixed model analysis revealed a significant main effect from both the implant number and the inter-implant distance, while no difference was found with regard to implant angulation. Five chip-off fractures of the porcelain veneer occurred in 5 of the 38 patients (3 in restorations fabricated onto the plaster cast group and 2 in the splinted-VPS cast group) with no effect from the type of impression on the prosthetic success rate (P = 0.331). However, all of the patients were functionally and aesthetically satisfied with their prostheses. Furthermore, mean MBL, SBI and PS showed no significant differences (P > 0.05) between the groups. Conclusions: The clinical outcome of plaster impressions for completely edentulous patients was found to be the same as that for splinted-VPS impressions. The intraoral pre-scan resin framework try-in can be avoided. Plaster impressions may be less time consuming and thus more comfortable for the patient, but sometimes may have to be repeated due to fractures. Keywords: CAD/CAM, fully edentulous, implant impressions, optical scanning Conflict-of-interest statement: All materials used in this study were purchased by the authors and there were no commercial or institutional interests.