Objective:

To review the current evidence base for NTCP models related to osteoradionecrosis (ORN) in head and neck cancer and propose a framework for clinical application.

Approach:

• Narrative Synthesis: A narrative synthesis of published literature on NTCP models for ORN in head and neck cancer was conducted, considering studies that met specific inclusion criteria.

Key Findings:

• The landmark NTCP model by van Dijk et al. identified D30% and pre-RT dental extraction as key predictors of ORN, achieving AUCs of 0.78 on the training cohort and 0.75 on the validation set.
• The PREDMORN study corroborated these findings, adding V70Gy and smoking status as additional predictors, with AUCs of 0.67–0.69 across various cohorts.
• The WAFT model introduced a time-to-event framework for ORN, identifying D25% as a significant predictor with an Adjusted Time Ratio of 0.88 per Gy increment.
• Deep learning approaches have shown promise but have not yet demonstrated significant advantages over traditional DVH models.

Interpretation:

The clinical implementation of NTCP models for ORN is limited by insufficient cohort diversity, inadequate characterization of RT protocols, and under-representation of oral health variables.

Limitations:

• Insufficient cohort diversity in existing studies.
• Inadequate characterization of radiotherapy protocol heterogeneity.
• Consistent under-representation of oral health variables.

Conclusion:

A strategic framework is proposed to address the limitations of current NTCP models, aiming for improved clinical translation.