Latent class analysis–derived prediction model of shoulder function after arthroscopic repair for rotator cuff tear: a large retrospective cohort study

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Objectives Considering the potential probabilities of shoulder function following arthroscopic rotator cuff repair (ARCR) was crucial for clinical decision-making and postoperative managements. This study aimed to identify distinct patterns of postoperative shoulder function and their prognostic factors. Methods Records of adult patients who underwent ARCR for the treatment of rotator cuff tear were reviewed from the electronic medical records. Constant–Murley scores were utilized to measure shoulder function at baseline and 3- and 12-month follow-ups. A latent class analysis was utilized to identify distinct classes of Constant–Murley trajectories. A multiple logistic regression analysis was conducted to examine risk factors. Results A total of 553 patients were included in the study. Changes in postoperative shoulder function were best categorized into three latent classes: high function (Class 1, 65.7%), low function (Class 2, 24.1%), and high retear risk (Class 3, 10.2%, accounting for 71.7% of all retears). Patients in Class 1 experienced significantly greater improvements in shoulder function compared with those in both Class 2 and Class 3, achieving the highest Constant–Murley scores over 12 months following surgery. Using Class 1 as the reference, a regression analysis was performed, which identified risk factors for Class 2, including having at least two full-thickness tears, longer symptom duration, undergoing acromioplasty combined with capsulotomy, undergoing acromioplasty combined with acromioclavicular resection, lower nutritional risk index scores, having two or more elevated inflammatory cytokines, and using weak opioids for more than one postoperative month (all p < 0.05). For Class 3, risk factors included a massive tear, a global fatty degeneration index between 1 and 2 or >2, an intraoperative repair tension of 8 Ibf or more, and having at least two elevated inflammatory cytokines (all p < 0.05). The area under the receiver operating characteristic curves was 0.903 and 0.931 for the two models. Conclusions Patients exhibited heterogeneous patterns of shoulder function following ARCR, each associated with different risk factors. These clinically distinct, trajectory-specific risk profiles provided actionable evidence to implement targeted, risk-stratified postoperative management protocols early after surgery. These findings emphasized the importance of individualized assessments and managements in the clinical setting, to reduce unfavorable outcomes and maximize postoperative shoulder function recovery for patients undergoing ARCR.