Background: Clinically evident loosening of a glenoid component inserted with cement in total shoulder arthroplasty is infrequent, but radiographic changes that indicate loosening at the implant-bone interface are common and have been associated with functional limitation. We compared the results of total shoulder arthroplasties in which the glenoid implant had been inserted with cement with those of arthroplasties in which a bone-ingrowth glenoid implant had been inserted without cement.Methods: The results of eighty-six consecutive total shoulder arthroplasties, performed by the same surgeon, were retrospectively reviewed between four and seven years after the operation. Fifty-eight shoulders in which the primary glenoid implant was in situ were assessed with use of the Simple Shoulder Test and Short Form-36 questionnaires, clinical examination, and fluoroscopic imaging of the glenoid implant-bone interface. Thirty-two of the glenoid components had been fixed with cement and twenty-six, without cement.Results: Complications occurred in 16 percent (fourteen) of the eighty-six shoulders, and 9 percent (eight) of the shoulders needed a revision operation. None of the revisions were done because of loosening of the glenoid component. Five of the eight revisions involved implants that had been inserted without cement. Three of these implants were revised because of early instability and two, because the polyethylene component had separated from the metal tray of the glenoid implant. With the numbers available, we could not detect any significant differences between the groups with respect to pain, range of motion, function of the shoulder, or general health. Radiographic analysis demonstrated a high level of interobserver agreement (kappa = 0.89). Radiolucent lines were observed after 41 percent (thirteen) of the thirty-two arthroplasties performed with cement compared with 23 percent (six) of the twenty-six arthroplasties performed without cement. The proportion of implants classified as probably loose was approximately three times greater in the group in which cement had been used. Eccentric wear of the posterior rim of the metal tray and focal osteolysis under the metal tray were observed in the group in which the component had been inserted without cement; these findings may indicate a potential for progression of radiographic loosening with increased durations of follow-up.Conclusions: We concluded that, despite the higher rate of early complications, the intermediate-term outcomes of arthroplasties in which the glenoid implant is inserted without cement are comparable with those of arthroplasties with cementing of the glenoid component.