Rotator cuff tears are a common cause of shoulder pain and dysfunction. After surgical repair, there is a significant re-tear rate (25%-90%). The aim of this study was to determine the primary mode of mechanical failure for rotator cuffs repaired with suture anchors at the time of revision rotator cuff repair. We prospectively followed 342 consecutive torn rotator cuffs, repaired by a single surgeon using suture anchors and a mattress-suturing configuration. Of those shoulders, 21 (6%) subsequently underwent a revision rotator cuff repair by the original surgeon, and 1 underwent a second revision repair. Intraoperative findings, including the mode of failure, were systematically recorded at revision surgery and compared with the findings at the primary repair. In addition, 81 primary rotator cuff repairs had a radiographic and fluoroscopic evaluation at a mean of 37 weeks after repair to assess for any loosening or migration of the anchors. At revision rotator cuff repair, the predominant mode of failure was tendon pulling through sutures (19/22 shoulders) (P <.001). Two recurrent tears occurred in a new location adjacent to the previous repair, and one anchor was found loose in the supraspinatus tendon. The mean size of the rotator cuff tear was larger at the revision surgery (P =.043), the tendon quality ranked poorer (P =.013), and the tendon mobility decreased (P =.002), as compared with the index procedure. The radiographs and fluoroscopic examination showed that all 335 anchors in 81 patients were in bone. Rotator cuff repairs with suture anchors that underwent revision surgery failed mechanically by three mechanisms, the most common of which was tendon pulling through sutures. This suggests that the weak link in rotator cuff repairs with suture anchors and horizontal mattress sutures, as determined at revision surgery, is the tendon-suture interface.