Aims. Controversy about the use of an anatomical total shoulder arthroplasty (aTSA) in young arthritic patients relates to which is the ideal form of fixation for the glenoid component: cemented or cementless. This study aimed to evaluate implant survival of aTSA when used in patients aged < 60 years with primary glenohumeral osteoarthritis (OA), and to compare the survival of cemented all-polyethylene and cementless metal-backed glenoid components. Materials and Methods. A total of 69 consecutive aTSAs were performed in 67 patients aged < 60 years with primary glenohumeral OA. Their mean age at the time of surgery was 54 years (35 to 60). Of these aTSAs, 46 were undertaken using a cemented polyethylene component and 23 were undertaken using a cementless metal-backed component. The age, gender, preoperative function, mobility, premorbid glenoid erosion, and length of follow-up were comparable in the two groups. The patients were reviewed clinically and radiographically at a mean of 10.3 years (5 to 12, . sd. 26) postoperatively. Kaplan–Meier survivorship analysis was performed with revision as the endpoint. Results. A total of 26 shoulders (38%) underwent revision surgery: ten (22%) in the polyethylene group and 16 (70%) in the metal-backed group (p < 0.0001). At 12 years’ follow-up, the rate of implant survival was 74% (. sd.  0.09) for polyethylene components and 24% (. sd.  0.10) for metal-backed components (p < 0.0002). Glenoid loosening or failure was the indication for revision in the polyethylene group, whereas polyethylene wear with metal-on-metal contact, instability, and insufficiency of the rotator cuff were the indications for revision in the metal-backed group. Preoperative posterior subluxation of the humeral head with a biconcave/retroverted glenoid (Walch B2) had an adverse effect on the survival of a metal-backed component. Conclusion. The survival of a cemented polyethylene glenoid component is three times higher than that of a cementless metal-backed glenoid component ten years after aTSA in patients aged < 60 years with primary glenohumeral OA. Patients with a biconcave (B2) glenoid have the highest risk of failure. Cite this article: Bone Joint J 2018;100-B:485–92