Abstract
BACKGROUND: Conventional two-dimensional (2D) imaging measurements of femoral neck fractures (FNFs) merely reflect angular displacement. We developed a novel measurement method for measuring the three-dimensional (3D) quantitative displacements of FNF site via CT. This study aimed to introduce this technique, assess its reliability, determine the correlation of 3D displacement with femoral head avascular necrosis (AVN), and develop a prediction model. METHODS: A retrospective cohort study was conducted on 200 FNF patients who underwent closed reduction and internal fixation from January 2018 to January 2022 and were followed for at least 3 years. The vertical, horizontal, and separation displacements of the fracture site were measured using computed tomography (CT) 3D reconstruction technique. One attending orthopedic surgeon, one orthopedic resident, and one radiology resident independently performed measurements using both 2D and 3D imaging techniques on all 200 patients; the radiology resident repeated the measurements after a 3-month interval. Inter- and intra-observer reliability were assessed using intraclass correlation coefficients (ICCs). Bivariate correlation identified variables associated with AVN, followed by binary logistic regression to evaluate the relationship between displacement parameters and AVN. Patients were divided into training and validation sets (7:3 ratio) for development and validation of a prediction model. RESULTS: The 3D measurement method showed higher ICCs [inter-observer: 0.963, 95% confidence interval (CI): 0.949-0.974; intra-observer: 0.962, 95% CI: 0.945-0.975] than the 2D measurement method [inter-observer: 0.723 (95% CI: 0.639-0.794); intra-observer: 0.776 (95% CI: 0.685-0.844)]. Bivariate correlation showed that both 2D and 3D displacements significantly correlated with AVN [all P<0.05; OR: 0.82-9.83 (95% CI: 0.73-36.56)]. Logistic regression identified separation and comprehensive displacements as independent risk factors for AVN [odds ratio: 1.12 (95% CI: 1.01-1.25), and 0.787 (95% CI: 0.70-0.86), respectively]. In receiver operating characteristic (ROC) analysis, the comprehensive displacement exhibited the greatest area under the curve (AUC), followed by the separation displacement, (both P<0.05; 95% CI: 0.700-0.912), whereas the 2D displacement exhibited the least AUC (P>0.05; 95% CI: 0.484-0.655). The predication nomogram indicated substantial clinical net benefit when the AVN risk threshold was set between 0.1 and 0.9. The ROC and calibration curves, and confusion matrix of the prediction model showed high consistency for the training and validation sets, with high sensitivity (97%), specificity (71%) and accuracy (76%). Decision curve analysis confirmed clinical utility within a risk threshold range of 10% to 90%. CONCLUSIONS: CT-measured 3D displacements offer superior reliability, stronger correlation with AVN, and better predictive potential compared to traditional 2D methods. The impact of displacement in different directions on AVN varies, but further research is needed.