Treatment failure prediction for head-and-neck cancer radiation therapy.

Objective: Treatment outcome prediction is an important emerging topic in oncologic care. To support radiation oncologists on their decisions, with individualized, tailored treatment regimens increasingly becoming the standard of care, accurate tools to predict tumour response to treatment are needed. The goal of this work is to identify the most determinant factor(s) for treatment response aiming to develop prediction models that robustly estimate tumour response to radiation therapy in patients with head-and-neck cancer.

Methods: A population-based cohort study was performed on 92 patients with head-and-neck cancer treated with radiation from 2007 until 2014 at the Portuguese Institute of Oncology of Coimbra (IPOCFG). Correlation analysis and multivariate binary logistic regression analysis were conducted in order to explore the predictive power of the considered predictors. Performance of the models is expressed as the area under the curve (AUC) of the receiver operating characteristics (ROC) curve. A nomogram to predict treatment failure was developed.

Results: Significant prognostic factors for treatment failure, after multivariate regression, were older age, non-concomitant radiation therapy and larger primary tumour volume. A regression model with these predictors revealed an AUC of .78 for an independent data set.

Conclusions: For patients with head-and-neck cancer treated with definitive radiation, we have developed a prediction nomogram based on models that presented good discriminative ability in making predictions of tumour response to treatment. The probability of treatment failure is higher for older patients with larger tumours treated with non-concomitant radiation.