Currently, there are no published results of patients in whom a window in the humeral cortex had been created to facilitate revision of a failed shoulder arthroplasty. Therefore, we reviewed the results of patients who underwent creation of a humeral window in the course of a revision shoulder arthroplasty to determine the complications and rate of healing. Twenty patients underwent humeral osteotomy to remove a well-fixed humeral component during the course of revision shoulder arthroplasty at our institution between January 1, 1987, and December 31, 2002. There were 17 cemented humeral components and 3 fully textured, press-fitted components. Of the shoulders, 16 had an anterior window, 3 had a medial window, and 1 had combined anterior and medial windows. There were 4 intraoperative fractures: 3 in the humeral shaft and 1 in the greater tuberosity. At the most recent radiographic follow-up at a mean of 3.3 years (range, 3-176 months), 17 of 20 windows had healed. Among 3 patients with limited radiographic follow-up (1, 3, and 4 months), 2 were healing and 1 did not demonstrate radiographic signs of healing. There were no cases of window malunion. No humeral components have developed clinical loosening. The data from this study suggest that humeral windows may facilitate controlled removal of well-fixed humeral components with a high rate of union. Further evolution of this technique may result in a lower fracture rate.