Abstract
Objectives
People who regularly participate in sports activities and those who work in certain specialized occupations are susceptible to knee injuries and have a substantially increased risk of post-traumatic knee osteoarthritis (KOA). Our aim was to develop a nomogram prediction model for the risk of KOA after sports injury based on imaging features of knee structures and demographic and clinical sport-related variables.
Methods
The modeling group included a total of 1002 patients with a complete history of knee joint sports injury admitted to the General Hospital of the Northern Theater of Surgery from January to December 2023. The patients were divided into KOA and non-KOA groups. Multivariate logistic regression analysis was used to identify risk factors, and a dynamic online nomogram prediction model for the risk of KOA after knee sports injury was constructed. Receiver operating characteristic (ROC) curve analyses, Hosmer-Lemeshow tests, and calibration plots were used to test the goodness of fit and predictive effect of the models. The prediction model was verified in an external validation cohort with a total of 429 patients with knee joint sports injuries, 145 with KOA and 284 with no KOA, admitted to the 962nd Hospital of the People’s Liberation Army (PLA) from October to December 2023.
Results
Among 1002 patients with knee joint sports injuries in the modeling group, 307 (30.64%) had KOA. Multivariate logistic regression analysis identified six factors: age, usual duration of exercise, foot strike pattern, fracture and bone contusion, meniscus injury, and cruciate ligament injury, as independent predictors of KOA after knee joint sports injury (P < 0.05). An online nomogram was constructed based on the six risk factors and the risk of KOA was quantified. The area under the ROC curve (AUC) for KOA after sports injury was 0.746 (95% confidence interval [CI], 0.721–0.768), sensitivity 0.739, and specificity 0.654. The AUC for the validation group was 0.731 (95% CI, 0.712–0.751), sensitivity 0.646, specificity 0.71. For Hosmer-Lemeshow test, P = 0.539 and 0.169, indicating that the model possesses effective discrimination and fitting effects.
Conclusion
The online dynamic nomogram prediction model we established, which includes six risk factors, among them age, exercise duration, and foot strike pattern, can better predict the risk of KOA after knee joint sports injury in a susceptible population and provides a simple quantitative evaluation tool for high-risk patients. It is helpful for the early identification of individual disease risk, timely intervention, and adjustment of training methods to provide a reference for preventive care.