Comprehensive Summary
This study by Takada, Tambas, and colleagues evaluated existing models predicting complications after radiation therapy for head and neck cancers. Researchers reviewed 143 studies covering 592 models across 140,767 patients using databases such as Ovid MedLine, Embase, and the WHO Clinical Trials Registry. Model quality was assessed with the PROBAST tool, and performance summarized with c-statistics. Results showed that 81% of models lacked external validation, and only 9 models had been validated in two or more independent studies. Many validations had high risk of bias due to small sample sizes and methodological limitations. The authors conclude that current models for radiation toxicity prediction have significant gaps and require rigorous reassessment before clinical application.
Outcomes and Implications
This study highlights the limitations of current normal tissue complication probability (NTCP) models in predicting side effects of radiotherapy for head and neck cancers. While some centers already use NTCP models to guide proton therapy eligibility, most models lack external validation. The implication is that large, rigorous validation studies are needed before NTCP models can be reliably implemented in clinical decision-making and broadly adopted into practice.