ESTRO 2024 - Abstract Book

S1392

Clinical - Head & neck

ESTRO 2024

Radiotherapy (RT) and concurrent chemoradiotherapy (CRT) are essential treatments for head and neck cancer (HNC), yet a third of patients undergoing these treatments face unplanned emergency department (ED) visits. These unforeseen visits disrupt treatment and inflate medical costs. Our study aims to establish a personalized risk prediction system to identify those at high risk for such ED visits, enabling preemptive interventions and better patient care.

Material/Methods:

From 2001 to 2019, we analyzed patients aged 20 and above with head and neck cancer from the Chang Gung Research Database who underwent curative RT. We grouped patients based on unplanned ED visits from RT start to three months post-completion and identified risk factors. We used the Kaplan – Meier and log-rank tests for survival rate analysis. Then, patients were randomly assigned into a training set (T-set) for model development and a validation set (V-set), allocated 4:1. A nomogram was created via logistic regression, representing ED visit risks. The model's accuracy was validated by external validation using the V-set and evaluated by the C-index, Hosmer Lemeshow test, and calibration curves.

Results:

Out of 8,904 head and neck cancer patients studied, 39.02% experienced unplanned ED visits during their treatment. Key risk factors included age, primary origin, systemic treatment, comorbidities, RT/CRT intent, pre treatment hemoglobin level, and preventive tube placement. Significantly, unplanned ED visits correlated with poorer overall survival (OS; p < 0.001). For risk prediction, patients were allocated into a training set (T-set; 7,124 patients) and a validation set (V-set; 1,780 patients). A predictive nomogram was developed based on the T-set data. When validated using the V-set, the model showed a satisfactory C-index of 0.634 and a Hosmer-Lemeshow test p value of 0.9805, confirming its reliability.

Conclusion:

Risk factors for HNC with unplanned ED visit during RT/CRT course were identified. A predictive model was built using a large cohort data and validated by external validation with good reliability. The model can be used as individualized risk-based monitoring protocol to prevent such event.

Keywords: Head neck cancer, unplanned emergency, prediction

2315

Mini-Oral

Separating distant recurrences from second primaries in head and neck squamous cell carcinomas

Julie Kjems 1,2 , Jacob K. Lilja-Fischer 1 , Jan Alsner 1 , Jeppe Friborg 2 , Trine Tramm 3 , Jens Overgaard 1