The optimal choice for the treatment of end-stage primary glenohumeral osteoarthritis remains controversial, with alternatives including total shoulder replacement (TSR) and humeral head replacement (HHR). The objective of this review was to analyze the effect of TSR compared with HHR on rates of pain relief, range of motion, patient satisfaction, and revision surgery in patients with primary glenohumeral osteoarthritis. We searched computerized databases for clinical studies published between 1966 and 2004 that reported on shoulder replacement for primary glenohumeral osteoarthritis. Pain data were converted to a 100-point score. Outcome assessment data were pooled when possible, and analyses via normal test statistics were performed. We identified 23 studies, with a total of 1952 patients and mean follow-up of 43.4 months (range, 30-116.4 months). The mean level of evidence was 3.73. Among the 23 studies, 7 different outcome instruments were used. Of the 23 studies, 14 (n = 1185) reported pain relief, 15 (n = 1080) reported range of motion, 12 (n = 969) reported patient satisfaction, and 14 (n = 1474) reported revision surgery. Compared with HHR, TSR provided significantly greater pain relief (P < .0001), forward elevation (P < .0001), gain in forward elevation (P < .0001), gain in external rotation (P = .0002), and patient satisfaction (P < .0001). Furthermore, only 6.5% of all TSRs required revision surgery, which was significantly lower than the percentage for all patients undergoing HHR (10.2%) (P < .025). Only 1.7% of all-polyethylene glenoid components required revision. On the basis of this review and analysis, in comparison with HHR, TSR for the treatment of primary glenohumeral osteoarthritis significantly improves pain relief, range of motion, and satisfaction and has a significantly lower rate of revision surgery. Inconsistent outcome reporting and poor study design may warrant standardization of outcome instruments and improved study design in the future.