Background: The repair of chronic, massive rotator cuff tears is associated with a high rate of failure. Prospective studies comparing different repair techniques are difficult to design and carry out because of the many factors that influence structural and clinical outcomes. The objective of this study was to develop a suitable animal model for evaluation of the efficacy of different repair techniques for massive rotator cuff tears and to use this model to compare a new repair technique, tested in vitro, with the conventional technique.Methods: We compared two techniques of rotator cuff repair in vivo using the left shoulders of forty-seven sheep. With the conventional technique, simple stitches were used and both suture ends were passed transosseously and tied over the greater tuberosity of the humerus. With the other technique, the modified Mason-Allen stitch was used and both suture ends were passed transosseously and tied over a cortical-bone-augmentation device. This device consisted of a poly(L/D-lactide) plate that was fifteen millimeters long, ten millimeters wide, and two millimeters thick. Number-3 braided polyester suture material was used in all of the experiments.Results: In pilot studies (without prevention of full weight-bearing), most repairs failed regardless of the technique that was used. The simple stitch always failed by the suture pulling through the tendon or the bone; the suture material did not break or tear. The modified Mason-Allen stitch failed in only two of seventeen shoulders. In ten shoulders, the suture material failed even though the stitches were intact. Thus, we concluded that the modified Mason-Allen stitch is a more secure method of achieving suture purchase in the tendon. In eight of sixteen shoulders, the nonaugmented double transosseous bone-fixation technique failed by the suture pulling through the bone. The cortical-bone-augmentation technique never failed.In definite studies, prevention of full weight-bearing was achieved by fixation of a ten-centimeter-diameter ball under the hoof of the sheep. This led to healing in eight of ten shoulders repaired with the modified Mason-Allen stitch and cortical-bone augmentation.On histological analysis, both the simple-stitch and the modified Mason-Allen technique caused similar degrees of transient localized tissue damage. Mechanical pullout tests of repairs with the new technique showed a failure strength that was approximately 30 percent of that of an intact infraspinatus tendon at six weeks, 52 percent of that of an intact tendon at three months, and 81 percent of that of an intact tendon at six months.Conclusions: The repair technique with a modified Mason-Allen stitch with number-3 braided polyester suture material and cortical-bone augmentation was superior to the conventional repair technique. Use of the modified Mason-Allen stitch and the cortical-bone-augmentation device transferred the weakest point of the repair to the suture material rather than to the bone or the tendon. Failure to protect the rotator cuff postoperatively was associated with an exceedingly high rate of failure, even if optimum repair technique was used.Clinical Relevance: Different techniques for rotator cuff repair substantially influence the rate of failure. A modified Mason-Allen stitch does not cause tendon necrosis, and use of this stitch with cortical-bone augmentation yields a repair that is biologically well tolerated and stronger in vivo than a repair with the conventional technique. Unprotected repairs, however, have an exceedingly high rate of failure even if optimum repair technique is used. Postoperative protection from tension overload, such as with an abduction splint, may be necessary for successful healing of massive rotator cuff tears.