[Internal validation of a novel risk classification system for juvenile nasopharyngeal angiofibroma involving the skull base: a comparative analysis of the Fisch-Andrews and UPMC staging systems].
In 143 surgically treated patients with skull-base juvenile nasopharyngeal angiofibroma, the internally validated GVV score predicted several intraoperative bleeding outcomes better than the Fisch-Andrews and UPMC systems and showed comparable recurrence discrimination.
Open original publication →What the AI sees
In 143 surgically treated patients with skull-base juvenile nasopharyngeal angiofibroma, the internally validated GVV score predicted several intraoperative bleeding outcomes better than the Fisch-Andrews and UPMC systems and showed comparable recurrence discrimination.
Research significance
The evidence supports the GVV score as a prognostic risk-stratification tool within the analyzed cohort; it may, by inference, help tailor perioperative preparation, bleeding-mitigation strategies, and recurrence surveillance, but the record does not show that score-guided management improves clinical outcomes.
Source abstract
OBJECTIVES: To develop and internally validate a new prognostic classification system for juvenile nasopharyngeal angiofibroma (JNA) involving the skull base, aimed at improving risk stratification for massive intraoperative blood loss and tumor recurrence. The model's performance was compared with two established systems: Fisch-Andrews and UPMC (University of Pittsburgh Medical Center). MATERIAL AND METHODS: Using multivariable logistic and Cox regression analyses, we created a point-based risk model (GVV score) incorporating three clinical and radiological parameters: (1) Tumor volume, scaled using a coefficient of 0.055; (2) Growth pattern: tumour extension limited to the nasal cavity, nasopharynx, pterygopalatine fossa and/or paranasal sinuses (1 point); involvement of the infratemporal fossa and/or orbit without intracranial extension (2 points); presence of intracranial extension (3 points), destruction of the greater wing of the sphenoid bone (2 points), destruction of the pterygoid process (1 point), and post-pterygopalatine extension (1 point); (3) Vascularization: complete embolization (-1 point), history of prior intervention (+1 point). Patients were stratified into three risk groups based on total score: I stage (<2.0), II stage (2.0-5.79), III stage (≥5.8). For recurrence: low (<5.0), intermediate (5.0-6.99), high risk (≥7.0). Internal validation included AUC analysis, Brier score, Hosmer-Lemeshow test, 10-fold cross-validation, and Net Reclassification Improvement (NRI). Predictive performance was compared with the Fisch-Andrews and UPMC systems. RESULTS: We analyzed 143 patients treated surgically between 2014 and 2024. The new model showed superior performance for predicting massive blood loss (AUC=0.761 vs. 0.597 and 0.574), transfusion (0.739 vs 0.669 and 0.662), and blood loss >15% of circulating volume (0.776 vs 0.726 and 0.675), compared to Fisch-Andrews and UPMC. Recurrence prediction was comparable (AUC=0.713). The model also demonstrated lower Brier scores and more accurate classification of patients without adverse events. High-risk patients had a 6.6-fold greater risk of recurrence (p=0.001). CONCLUSION: This novel classification system provides more accurate and clinically interpretable risk stratification for JNA than current staging systems. External validation in independent cohorts is warranted to support its broader clinical use.