Analysis of risk factors and development of a nomogram-based prediction model for defective bony non-union.

Objective: To explore risk factors for defective non-union of bone and develop a nomogram-based prediction model for such an outcome.

Methods: This retrospective study analysed the case data of patients with defective bony non-unions who were treated at the authors' hospital between January 2010 and December 2020. Patients were divided into the union and non-union groups according to their Radiographic Union Score for Tibia scores 1 year after surgery. Univariate analysis was performed to assess factors related to demographic characteristics, laboratory investigations, surgery, and trauma in both groups. Subsequently, statistically significant factors were included in the multivariate logistic regression analysis to identify independent risk factors. A nomogram-based prediction model was established using statistically significant variables in the multivariate analysis. The accuracy and stability of the model were evaluated using receiver operating characteristic (ROC) and calibration curves. The clinical applicability of the nomogram model was evaluated using decision curve analysis.

Results: In total, 204 patients (171 male, 33 female; mean [±SD] age, 39.75 ± 13.00 years) were included. The mean body mass index was 22.95 ± 3.64 kg/m. Among the included patients, 29 were smokers, 18 were alcohol drinkers, and 21 had a previous comorbid systemic disease (PCSD). Univariate analysis revealed that age, occupation, PCSD, smoking, drinking, interleukin-6, C-reactive protein (CRP), procalcitonin, alkaline phosphatase, glucose, and uric acid levels; blood calcium ion concentration; and bone defect size (BDS) were correlated with defective bone union (all P < 0.05). Multivariate logistic regression analysis revealed that PCSD, smoking, interleukin-6, CRP, and glucose levels; and BDS were associated with defective bone union (all P < 0.05), and the variables in the multivariate analysis were included in the nomogram-based prediction model. The value of the area under the ROC curve for the predictive model for bone defects was 0.95.

Conclusion: PCSD, smoking, interleukin-6, CRP, and glucose levels; and BDS were independent risk factors for defective bony non-union, and the incidence of such non-union was predicted using the nomogram. These findings are important for clinical interventions and decision-making.