Predictors of dislocations after reverse shoulder arthroplasty: a study by the ASES complications of RSA multicenter research group

Background: Instability after reverse shoulder arthroplasty (RSA) is one of the most frequent complications and remains a clinical challenge. Current evidence is limited by small sample size, single-center, or single-implant methodologies that limit generalizability. We sought to determine the incidence and patient-related risk factors for dislocation after RSA, using a large, multicenter cohort with varying implants. Methods: A retrospective, multicenter study was performed involving 15 institutions and 24 American Shoulder and Elbow Surgeons members across the United States. Inclusion criteria consisted of patients undergoing primary or revision RSA between January 2013 and June 2019 with minimum 3-month follow-up. All definitions, inclusion criteria, and collected variables were determined using the Delphi method, an iterative survey process involving all primary investigators requiring at least 75% consensus to be considered a final component of the methodology for each study element. Dislocations were defined as complete loss of articulation between the humeral component and the glenosphere and required radiographic confirmation. Binary logistic regression was performed to determine patient predictors of postoperative dislocation after RSA. Results: We identified 6621 patients who met inclusion criteria with a mean follow-up of 19.4 months (range: 3-84 months). The study population was 40% male with an average age of 71.0 years (range: 23-101 years). The rate of dislocation was 2.1% (n = 138) for the whole cohort, 1.6% (n = 99) for primary RSAs, and 6.5% (n = 39) for revision RSAs (P < .001). Dislocations occurred at a median of 7.0 weeks (interquartile range: 3.0-36.0 weeks) after surgery with 23.0% (n = 32) after a trauma. Patients with a primary diagnosis of glenohumeral osteoarthritis with an intact rotator cuff had an overall lower rate of dislocation than patients with other diagnoses (0.8% vs. 2.5%; P < .001). Patient-related factors independently predictive of dislocation, in order of the magnitude of effect, were a history of postoperative subluxations before radiographically confirmed dislocation (odds ratio [OR]: 19.52, P < .001), primary diagnosis of fracture nonunion (OR: 6.53, P < .001), revision arthroplasty (OR: 5.61, P < .001), primary diagnosis of rotator cuff disease (OR: 2.64, P < .001), male sex (OR: 2.21, P < .001), and no subscapularis repair at surgery (OR: 1.95, P = .001). Conclusion: The strongest patient-related factors associated with dislocation were a history of postoperative subluxations and having a primary diagnosis of fracture nonunion. Notably, RSAs for osteoarthritis showed lower rates of dislocations than RSAs for rotator cuff disease. These data can be used to optimize patient counseling before RSA, particularly in male patients undergoing revision RSA.