A model using concomitant markers for predicting outcome in human papillomavirus positive oropharyngeal cancer

Objective: Head-neck cancer therapy has become intensified. With radiotherapy alone, 3-year disease-free survival (DFS) is 80% for HPV-positive TSCC/BOTSCC and better for patients with favorable characteristics, suggesting therapy could be tapered for some, decreasing side-effects. Therefore, we built a model to predict progression-free survival for patients with HPV-positive TSCC and BOTSCC. Material and methods: TSCC/BOTSCC patients treated curatively between 2000 and 2011, with HPV16 DNA/E7 mRNA positive tumors examined for CD8(+) TILs, HPV16 mRNA and HLA class I expression were included. Patients were split randomly 65/35 into training and validation sets, and LASSO regression was used to select a model in the training set, the performance of which was evaluated in the validation set. Results: 258 patients with HPV DNA/E7 mRNA positive tumors could be included, 168 and 90 patients in the respective sets. No treatment improved prognosis compared to radiotherapy alone. CD8(+) TIL counts and young age were the strongest predictors of survival, followed by T-stage <3 and presence of HPV16 E2 mRNA. The model had an area under curve (AUC) of 76%. A model where the presence of three of four of these markers defined good prognosis captured 56% of non-relapsing patients with a positive predictive value of 98% in the validation set. Furthermore, the model identified 35% of our cohort that was over-treated and could safely have received de-escalated therapy. Conclusion: CD8(+) TIL counts, age, T-stage and E2 expression could predict progression-free survival, identifying patients eligible for randomized trials with milder treatment, potentially reducing side effects without worsening prognosis. (C) 2017 Elsevier Ltd. All rights reserved.