We’re all familiar with the need to better identify which patients with p16+ oropharyngeal squamous cell carcinoma (oSCC) can safely omit cisplatin concurrent with definitive radiation. Deep learning models were retrospectively used to develop and validate an imaging risk score to predict who most stands to benefit from chemo. As you might imagine, it ultimately comes down to: bigger disease is riskier disease. Across 811 patients (in both training and validation cohorts) with initial staging CT neck imaging for p16+ oSCC, the artificial intelligence-generated imaging risk score–primarily based on sum total volume of primary tumor and nodal disease volume–was indeed significantly associated with disease-free and overall survival. Most interesting, when the imaging risk score was used to dichotomize patients with stage I-II disease into low versus high risk, receipt of concurrent chemo significantly improved disease-free survival for the high-risk cohort with no discernible benefit seen in the low-risk cohort. There is much left to learn, but if incorporated into affordable post-processing CT imaging software, this could prove an accessible adjunct to circulating tumor HPV-DNA risk stratification. | Song, JAMA Netw Open 2025
