Prospective validation of ctHPVDNA for detection of minimal residual disease and prediction of recurrence in patients with HPV-associated head and neck cancer treated with surgery.

6010 Background: Prediction of minimal residual disease (MRD) following surgery, and thus the need for adjuvant therapy, is currently based on clinicopathologic risk factors which have poor individual prognostic capacity. We previously reported that MRD detection by ctHPVDNA droplet digital (dd)PCR as early as post-operative day (POD) 1 is predictive of recurrence in stage I-II HPV+HNSCC patients. Here, we applied a significantly more sensitive custom whole genome hybrid-capture-based next generation sequencing assay, termed HPV-DeepSeek, to validate the prognostic capacity of ctHPVDNA detection and explore the optimal timing of MRD testing in HPV+HNSCC patients. We tested the primary hypothesis that patients with MRD detection within 6 weeks of surgery would have inferior PFS at 2 years and the secondary hypothesis that patients with ctHPVDNA positivity detected at any point following treatment completion would have inferior PFS at 2 years. Methods: 98 patients with HPV+HNSCC were prospectively enrolled with a mean follow-up of 712 days. All patients underwent surgery as primary treatment and clinicopathologic adjusted adjuvant treatment. 10ml blood samples were collected before surgery, in the post-operative period (POD 1-42), and serially in follow-up. MRD was defined as a lack of ctHPVDNA clearance during the 6 weeks following surgery. Cell free DNA was extracted from plasma and run on HPV-DeepSeek, and on existing ddPCR assays for head to head comparisons. Results: 96/98 (98%) of patients had detectable ctHPVDNA pre-treatment. 30/98 patients (31%) were MRD positive. Patients who were MRD positive had significantly worse 2 years PFS compared to MRD negative patients (78% vs 98%, P=0.0009). Predictive performance improved by limiting time points to POD 7-42 (2 year PFS 60% vs 97%, P1 week after surgery, suggesting the use of an ultra-sensitive assay such as HPV-DeepSeek requires adjustment of MRD time points. Patients with detectable ctHPVDNA following completion of all treatment had significantly worse 2-year PFS compared to patients without detectable ctHPVDNA (0% vs 97%, P<0.0001). 7/98 patients had cancer recurrence during follow-up that was detected by ctHPVDNA with a mean lead time of 207 days (35-518) to clinical diagnosis. Conclusions: ctHPVDNA detection by HPV-DeepSeek is a highly specific biomarker of MRD in HPV+HNSCC, accurately predicts disease progression and detects recurrence earlier than standard care clinical.