Background: Human papillomavirus (HPV)-positive oropharyngeal cancer (OPSCC) is associated with improved survival compared with HPV-negative disease. However, a minority of HPV-positive patients have poor prognosis. Currently, there is no generally accepted strategy for identifying these patients.
Methods: We retrospectively analysed 270 consecutively treated OPSCC patients from three centres for effects of clinical, pathological, immunological, and molecular features on disease mortality. We used Cox regression to examine associations between factors and OPSCC death, and developed a prognostic model for 3-year mortality using logistic regression analysis.
Results: Patients with HPV-positive tumours showed improved survival (hazard ratio (HR), 0.33 (0.21-0.53)). High levels of tumour-infiltrating lymphocytes (TILs) stratified HPV-positive patients into high-risk and low-risk groups (3-year survival; HPV-positive/TIL(high)=96%, HPV-positive/TIL(low)=59%). Survival of HPV-positive/TIL(low) patients did not differ from HPV-negative patients (HR, 1.01; P=0.98). We developed a prognostic model for HPV-positive tumours using a 'training' cohort from one centre; the combination of TIL levels, heavy smoking, and T-stage were significant (AUROC=0·87). This model was validated on patients from the other centres (detection rate 67%; false-positive rate 5.6%; AUROC=0·82).
Interpretation: Our data suggest that an immune response, reflected by TIL levels in the primary tumour, has an important role in the improved survival seen in most HPV-positive patients, and is relevant for the clinical evaluation of HPV-positive OPSCC.