A radiographic model predicting the status of the anterior cruciate ligament in varus knee with osteoarthritis

Abstract

Abstract Purpose The study aims to investigate the accuracy of different radiographic signs for predicting functional deficiency of anterior cruciate ligament (ACL) and test whether the prediction model constructed by integrating multiple radiographic signs can improve the predictive ability. Methods A total number of 122 patients from January 1, 2018, to September 1, 2021, were enrolled in this study. Among them, 96 patients were classified as the ACL-functional (ACLF) group, while 26 patients as the ACL-deficient (ACLD) group after the assessment of magnetic resonance imaging (MRI) and the Lachman’s test. Radiographic measurements, including the maximum wear point of the proximal tibia% (MWPPT%), tibial spine sign (TSS), coronal tibiofemoral subluxation (CTFS), hip–knee–ankle angle (HKA), mechanical proximal tibial angle (mPTA), mechanical lateral distal femoral angle (mLDFA) and posterior tibial slope (PTS) were measured using X-rays and compared between ACLF and ACLD group using univariate analysis. Significant variables (p < 0.05) in univariate analysis were further analyzed using multiple logistic regression analysis and a logistic regression model was also constructed by multivariable regression with generalized estimating models. Receiver-operating-characteristic (ROC) curve and area under the curve (AUC) were used to determine the cut-off value and the diagnostic accuracy of radiographic measurements and the logistic regression model. Results MWPPT% (odds ratio (OR) = 1.383, 95% confidence interval (CI) = 1.193–1.603, p < 0.001), HKA (OR = 1.326, 95%CI = 1.051–1.673, p = 0.017) and PTS (OR = 1.981, 95%CI = 1.207–3.253, p = 0.007) were shown as predictive indicators of ACLD, while age, sex, side, TSS, CTFS, mPTA and mLDFA were not. A predictive model (risk score = -27.147 + [0.342*MWPPT%] + [0.282*HKA] + [0.684*PTS]) of ACLD using the three significant imaging indicators was constructed through multiple logistic regression analysis. The cut-off values of MWPPT%, HKA, PTS and the predictive model were 52.4% (sensitivity:92.3%; specificity:83.3%), 8.5° (sensitivity: 61.5%; specificity: 77.1%), 9.6° (sensitivity: 69.2%; specificity: 78.2%) and 0.1 (sensitivity: 96.2%; specificity: 79.2%) with the AUC (95%CI) values of 0.906 (0.829–0.983), 0.703 (0.574–0.832), 0.740 (0.621–0.860) and 0.949 (0.912–0.986) in the ROC curve. Conclusion MWPPT% (> 52.4%), PTS (> 9.6°), and HKA (> 8.5°) were found to be predictive factors for ACLD, and MWPPT% had the highest sensitivity of the three factors. Therefore, MWPPT% can be used as a screening tool, while the model can be used as a diagnostic tool.