Operative treatment is associated with improved 3-year survival after periprosthetic humerus fracture

Background This study examines the survival implications of operative versus nonoperative management of periprosthetic humerus fractures in a large, single-center cohort. Methods Patients at a single tertiary care center with periprosthetic humerus fractures about a shoulder prosthesis between January 2013 and March 2024 were retrospectively identified. Patients were stratified according to operative versus nonoperative treatment. The primary outcome was all-cause mortality. Survival was compared using Kaplan–Meier curves and log-rank tests. Prespecified time horizons were 90 days, 1 year, 2 years, and 3 years. Results Ninety patients with periprosthetic fractures were identified (23 nonoperative, 67 operative). There were 13 total deaths (8/23 [35%] nonoperative, 5/67 [7.5%] operative). Baseline characteristics were largely similar aside from a higher body mass index in the operative group (32 vs. 29 kg/m 2 , P = .046). Age (nonoperative: 74 vs. operative: 71 years, P = .30), sex ( P = .40), hand dominance ( P > .9), race ( P = .07), smoking ( P = .40), comorbidity burden (Elixhauser weighted and count P = .20 and P = .60, respectively), and Charlson distribution ( P = .20) were all similar between groups. On Kaplan–Meier analysis, early mortality was rare: 90-day survival was 95.7% in the nonoperative group versus 100.0% in the operative group ( P = .09). Survival differences widened over time, becoming significant at 3 years after the injury: 1 year (nonoperative: 95.7%; operative: 98.5%; P = .42), 2 years (nonoperative: 90%, operative: 98.5%; P = .09), and 3 years (nonoperative: 72.2%; operative: 96.5%; P = .004). Conclusions Surgical fixation of periprosthetic humerus fractures is associated with lower mortality that becomes apparent in a delayed fashion and grows over time. By 3 years after the injury, the mortality difference approached 25% between operatively and nonoperatively treated patients. These findings should be validated in larger, multicenter studies.