Risk stratification models in human papillomavirus-associated oropharyngeal squamous cell carcinoma: the Nova Scotia...

The incidence of oropharyngeal squamous cell carcinoma is increasing with a growing proportion of diagnoses associated with human papillomavirus (p16 + OSCC), which generally confers a favorable prognosis. For these reasons, novel risk stratification models specific to the p16 + OSCC population have recently been proposed to guide future research on treatment de-intensification for appropriate patients. This study aimed to quantify patient risk distribution using multiple published risk models and investigate the hypothesis that the local p16 + OSCC population includes a smaller proportion of low-risk patients due to a high prevalence of concurrent tobacco exposure.

A retrospective cohort study was performed including patients diagnosed with p16 + OSCC in Nova Scotia between 2011 and 2015. Patient identification was obtained through the CCNS registry and an institutional database. Exclusion criteria included HPV negative status, second primary cases, incomplete data availability, and local recurrence cases.

Following exclusion, 117 patients met study criteria. The majority had small primary tumors (70.9% ≤ T2) and advanced nodal status on presentation (60.7% ≥ N2b). Most patients had a positive smoking history (62.4%), with 53.0% of patients having a pack-year history greater than 10 pack-years. In four of the five risk stratification models, the majority of the study population fell into the lowest risk category. The risk stratification distribution of our local population was similar to the populations used to validate the published models, with the largest single category difference being 13.3% (range - 12.3 to + 13.3%).

This is the first study to compare multiple currently published risk stratification models to a local population and address the uncertainty of risk stratification in the Nova Scotian p16 + OSCC population. Despite a high prevalence of concurrent tobacco exposure, the study population was found to be overall low risk, with similar risk compared to model validation populations.