Influencing factors and prediction model construction of re-tearing risk after arthroscopic rotator cuff repair

LI He; MA Shengshan; SUN Yang; WU Dongze; LI Xiaofei

doi:10.11958/20252352

Abstract：Objective To explore the factors affecting the risk of postoperative re-tearing in patients undergoing arthroscopic rotator cuff repair,and to construct a nomogram prediction model.Methods A total of 587 patients who underwent arthroscopic rotator cuff repair were divided into the training set(470 cases)and the validation set(117 cases).The training set was sub-divided into the re-tearing group and the healing group based on whether re-tearing occurred within 12 months after surgery.Clinical data were compared between the two groups.Multivariate Logistic regression analysis was used to identify the influencing factors of re-tearing,and a nomogram prediction model for re-tearing was constructed.The receiver operating characteristic(ROC)curve and calibration curve were used to evaluate the model.The clinical decision curve was used to evaluate its clinical applicability.Results In the training set,re-tearing occurred in 97 cases(the re-tearing group)and healing was observed in 373 cases(the healing group).There were 23 cases of re-tearing in the validation set.Multivariate Logistic regression analysis results showed that age≥60 years,osteoporosis,length of tearing>3 cm,tendon retraction,fat infiltration≥grade 2 and use of corticosteroids were independent risk factors for re-tearing(P＜0.05).The C-index of the constructed nomogram model was 0.814(95%CI:0.810-0.819).The calibration curve and the ideal curve were close in both the training and validation sets.Internally validation using Bootstrap method found that the mean absolute errors were 0.011 and 0.027,respectively.Hosmer-Lemeshow test showed good goodness of fit(χ2=4.531,P=0.806).The clinical decision curve showed that the model had a significant positive net benefit.The area under the curve(AUC),sensitivity and specificity of the nomogram prediction model for predicting postoperative re-tearing in the training set were 0.818(95%CI:0.771-0.865),78.69%and 79.60%,which in the validation set were 0.807(95%CI:0.730-0.884),77.94%and 76.82%,respectively.Conclusion The nomogram model constructed based on risk factors can better predict the risk of re-tearing after arthroscopic repair.