Update of a prediction model for postoperative shoulder stiffness after arthroscopic rotator cuff repair

• Published in: Communications Medicine
• Main Authors: Thomas Stojanov, Soheila Aghlmandi, Andreas Marc Müller, Philipp Moroder, Alexandre Lädermann, Cornelia Baum, ARCR_Pred Study Group, Laurent Audigé
• Format: Article
• Language: English
• Published: Nature Portfolio 2025-10-01
• Online Access: https://doi.org/10.1038/s43856-025-01125-w

Abstract Background Arthroscopic rotator cuff repair (ARCR) is a common procedure, and postoperative shoulder stiffness (POSS) is one of its most frequent adverse events, potentially necessitating individualized therapy. Our objectives were to update and internally validate a model predicting the occurrence of POSS for patients undergoing an ARCR. Methods We prospectively enrolled 973 patients undergoing primary ARCR included in the ARCR_Pred dataset. A two-round Delphi survey with 53 surgeons established a consensus definition of POSS within 6 months postoperatively and a ranking of candidate prognostic factors. Treating surgeons estimated POSS risk immediately after surgery. We externally validated an existing POSS model and developed updated multivariable logistic regression models using complete-case and multiple imputed datasets. Results We achieved a high consensus (88%) on the POSS definition among 44 responding shoulder surgeons, who also ranked the prognostic relevance of 71 factors for the prediction of POSS. The newly developed ARCR_Pred-POSS included 7 factors (age, acromiohumeral distance, symptom duration, baseline external rotation, active baseline abduction, baseline Oxford Shoulder Score, and surgery duration) and demonstrated superior discrimination (AUC = 0.735) and calibration (slope = 1.022) compared to the original POSS model (AUC = 0.581, slope = 0.508). Surgeons tended to overestimate the risk of POSS in their patients (AUC = 0.563, slope = 1.241). Conclusions These findings support the continued development of prediction models and provide valuable outputs for optimizing surgical timing, indications, and personalized rehabilitation.