Development and validation of a predictive nomogram for vascular crises in oral and maxillofacial cancer patients undergoing free flap surgery

Fig 1. Nomogram model for predicting vascular crises after free flap repair of oral and maxillofacial tumors. Nomogram for predicting vascular crises following free tissue flap repair of oral and maxillofacial tumors. The model incorporates key predictors and displays a favorable fit, achieving a cutoff value of 0.069 and explaining 16.9% of the total variation. Internal and external validation results support its effectiveness in clinical applications. Note: ot:express other; pa:express palate; md:express mandible; mf: express mouth floor; ck: express cheek; tn: express tongue; gu: express upper gum.

Fig 2. Calibration curve for internal validation in the training group. The curve demonstrates a close alignment with the 45° reference line, indicating effective predictive accuracy and calibration of the model for vascular crises following free tissue flap repair of oral and maxillofacial tumors.

Fig 3. Training group internal validation. Calibration curve for internal validation using the Bootstrap method (1000 samples). The average absolute error of 0.01 demonstrates excellent fit against the ideal curve.

Fig 4. ROC curve of training group. ROC curve for the training group, demonstrating the nomogram model’s predictive performance for vascular crises after free tissue flap repair in oral and maxillofacial tumors. The AUC is 0.780, with sensitivity of 0.708 and specificity of 0.764, indicating good overall accuracy. The calibration curve aligns closely with the ideal reference line, confirming the model’s strong fit.

Fig 5. ROC curve of validation group. ROC curve for the validation group, illustrating the model’s performance in predicting vascular crises post-free tissue flap repair. The AUC is 0.701 (95% CI: 0.586–0.817), indicating moderate predictive accuracy, with a specificity of 0.990 and sensitivity of 0.074. The calibration curve aligns closely with the ideal reference, reinforcing the nomogram’s robust external validation.

Fig 6. Calibration curve for external validation in the validation group. Calibration curve for the validation group. The curve closely resembles that of the training group, affirming the model’s robustness and predictive accuracy.

Fig 7. DCA curve of training group. The DCA curve illustrates the net benefit of the nomogram model compared to the "treat-all" and "treat-none" strategies across a range of threshold probabilities (0.000–0.670).

Fig 8. DCA curve of validation group. The DCA curve confirms the model’s clinical applicability, showing net benefits over a similar range of threshold probabilities.