Session Item

Experiencing difficulties swallowing, or dysphagia, is a common side effect of head and neck radiotherapy.  It is experienced by more than 80% of the patients at the end of treatment, with a significant proportion of patients having permanent impairment of swallowing sufficient to cause aspiration. For these patients, modification of diet, with potential requirement for enteral feeding, is generally recommended with consequent negative effect on long term quality of life. There is an unmet clinical need to better predict the long-term aspiration risk during initial discussions with patients of treatment side effects.  In this study, we present a simple model, including baseline variables to predict aspiration risk at 1 year, assessed via 100 ml water test.

Data from 108 oropharyngeal cancer patients were collected, including baseline clinical and tumour factors and various pre-treatment dysphagia scores: M.D. Anderson Dysphagia Inventory (MDADI), water test (WT), performance status scale for normalcy of diet (HN-PSS), table 1. A logistic regression predicting WT failure (defined as the patient coughing, having a wet voice quality post swallow, or not being able to finish the task) at 1 year was conducted. The final model was found by performing backward variable selection with bootstrapping (keeping top 3 selected variables, n=1000) and adjusting for internal overfitting using uniform shrinkage (based on optimism-adjusted calibration slope, n=1000). Model performance was assessed using area under the curve (AUC) of the receiver operating characteristic curve.

30.6% of patients failed the water test at 1 year. Water test at baseline, HPV status (assessed via P16), and performance status (PS) were selected for the final model (table 1 and figure 1). Performance of the model, assessed via AUC and adjusted for optimism, was 0.76. The strongest predictor for failing water test at 1 year is the test at base line (OR 6.8, p<0.001). Additionally, HPV negative patients have approximately 3 times the chance of failing the test at 1 year (OR 2.9, p=0.04), indicating possible differences in treatment toxicity trajectory. Conversely, poorer general status (PS 1+ vs PS 0) is also associated with a higher risk of failing the test (OR 2.18, p=0.09).

We propose a simple model using only baseline variables, and corrected for overfitting, which predicts dysphagia at 1 year (assessed via water test). Although the major factor for failing  WT at 1 year is the pre-treatment test result, it is important to consider other factors when initially discussing treatment as suggested by our model. We find a novel association with HPV status independent of disease stage or location, suggesting a possible underlying variation in response to radiotherapy related to HPV infection. After external validation, this simple model could be used during patient consultations to personalise discussion of possible side effects.