Abstract
METHODS: This study aimed to develop and validate a nomogram for predicting the risk of severe pain in patients with knee osteoarthritis. A total of 150 patients with knee osteoarthritis were enrolled from our hospital, and nomogram was established through a validation cohort (n = 150). An internal validation cohort (n = 64) was applied to validate the model. RESULTS: Eight important variables were identified using the Least absolute shrinkage and selection operator (LASSO) and then a nomogram was developed by Logistics regression analysis. The accuracy of the nomogram was determined based on the C-index, calibration plots, and Receiver Operating Characteristic (ROC) curves. Decision curves were plotted to assess the benefits of the nomogram in clinical decision-making. Several variables were employed to predict severe pain in knee osteoarthritis, including sex, age, height, body mass index (BMI), affected side, Kellgren-Lawrance (K-L) degree, pain during walking, pain going up and down stairs, pain sitting or lying down, pain standing, pain sleeping, cartilage score, Bone marrow lesion (BML) score, synovitis score, patellofemoral synovitis, bone wear score, patellofemoral bone wear, and bone wear scores. The LASSO regression results showed that BMI, affected side, duration of knee osteoarthritis, meniscus score, meniscus displacement, BML score, synovitis score, and bone wear score were the most significant risk factors predicting severe pain. CONCLUSIONS: Based on the eight factors, a nomogram model was developed. The C-index of the model was 0.892 (95% CI: 0.839-0.945), and the C-index of the internal validation was 0.822 (95% CI: 0.722-0.922). Analysis of the ROC curve of the nomogram showed that the nomogram had high accuracy in predicting the occurrence of severe pain [Area Under the Curve (AUC) = 0.892] in patients with knee osteoarthritis (KOA). The calibration curves showed that the prediction model was highly consistent. Decision curve analysis (DCA) showed a higher net benefit for decision-making using the developed nomogram, especially in the >0.1 and <0.86 threshold probability intervals. These findings demonstrate that the nomogram can predict patient prognosis and guide personalized treatment.