Development and Internal Evaluation of a Biomarker-Based Model for Preoperative Diagnosis of Infectious Nonunion.

OBJECTIVE

Preoperative identification of infectious nonunion remains challenging, particularly in patients without overt clinical signs of infection. Conventional inflammatory markers alone often provide insufficient diagnostic accuracy, creating uncertainty in surgical planning. This study aimed to develop and internally evaluate a predictive model combining novel and conventional biomarkers for preoperative diagnosis of infectious nonunion in long bone fractures. We hypothesized that integrating procalcitonin (PCT) with classical inflammatory markers would enhance diagnostic accuracy.

METHODS

We retrospectively analyzed 276 patients treated for femoral or tibial nonunion after internal fixation (2018-2023). Patients were randomly divided into a model-development cohort (n = 221) and a hold-out test cohort (n = 55). A multivariable logistic regression model was built using serum levels of PCT, interleukin-6 (IL-6), C-reactive protein (CRP), and erythrocyte sedimentation rate (ESR). Model performance was assessed by AUC, calibration plots, and decision curve analysis.

RESULTS

In the model-development cohort, 110 patients had infectious nonunion and 111 had aseptic nonunion. Infected cases had significantly higher biomarker levels (p < 0.001). Among single markers, ESR had the best performance (AUC ~0.79), followed by PCT (~0.71), IL-6 (~0.62), and CRP (~0.61). The best multivariable model included PCT, CRP, and ESR, achieving an AUC of ~0.85 (95% CI ~0.80-0.90). Bootstrapping confirmed minimal overfitting (corrected AUC ~0.84). In the hold-out test cohort, this model maintained strong performance (AUC ~0.85, 95% CI ~0.72-0.96), with ~75% sensitivity and 85% specificity-superior to any single biomarker. Calibration plots showed good agreement, and decision analysis demonstrated higher clinical benefit than using individual markers or treating all patients.

CONCLUSION

The PCT + CRP + ESR model reliably distinguishes infectious from aseptic nonunion, outperforming traditional markers alone. It offers a practical, accurate preoperative tool for infection screening, supporting more informed surgical decision-making. In clinical use, interpretation should be based on the predicted probability generated by the combined model rather than isolated elevation of individual biomarkers. Further prospective and external validation is warranted.