Evaluation of a tissue-free genome-wide methylome enrichment assay for detecting molecular residual disease (MRD) in patients with head and neck squamous cell carcinoma (HNSCC).

Abstract

6084 Background: Squamous cell carcinomas comprise ~90% of head and neck cancers (HNCs). Although ~80% of patients with HNSCC are eligible for curative therapy, an estimated 20% to 35% will develop recurrence, typically within 2 years. Recurrence is associated with poor survival, but early detection may enable salvage therapy. We previously reported clinical validation of a tissue-free, genome-wide methylome enrichment MRD assay for recurrence detection in HNC. Here we present performance of an updated classifier in a subset of patients with HNSCC. Methods: Test performance was evaluated using blood samples collected at ~3 (landmark), 12, and 24 mo after curative-intent treatment from patients with stage I-IVb HNSCC. MRD testing was performed using an assay based on cell-free methylated DNA immunoprecipitation and high throughput sequencing (cfMeDIP-seq) as previously described (Liu et al. Ann Oncol . 2025;36(1):108-117). To improve test performance and clinical actionability, we evaluated the effect of defining an indeterminate test category near the classifier threshold. An updated classifier defining 5% of training samples as indeterminate was analyzed in a training set, followed by a pre-specified clinical validation analysis. The primary endpoint was recurrence-free survival (RFS), comparing patients with MRD detected vs not detected. RFS was estimated using the Kaplan-Meier method, with differences assessed by the 2-sided log-rank test. Hazard ratios (HR) were estimated using the Cox proportional hazards model. Results: Among 117 patients in the validation set, median follow-up was 65.3 mo. MRD detection during surveillance was strongly associated with inferior RFS (HR, 54.5 [95% CI, 18.4-161.2]), an improvement over the previously reported HR of 35.7. Lead time between MRD detection and clinical recurrence was up to 10.8 mo (mean, 3.3 mo). MRD detection was also prognostic for OS (HR, 89.3 [11.8-676.4]). The updated classifier showed improved performance, with a sensitivity of 86.7% (26/30) and a specificity of 97.7% (85/87) for recurrence detection. Positive predictive value was 92.9% (26/28); negative predictive value was 95.5% (85/89). Performance was similar ( P =0.24) among patients with HPV-positive oropharyngeal carcinoma (n=52) vs HPV-negative disease (n=65), patients with stage I-II (n=72) vs stage III-IVb (n=45) disease ( P =0.20), and patients with surgery (n=34) vs non-surgical (n=83) definitive treatment ( P =0.37). Conclusions: This study demonstrates that tissue-free, blood-based MRD testing using a genome-wide methylome enrichment platform is prognostic for recurrence in patients with HNSCC treated with curative intent, regardless of stage, HPV status, or treatment modality. The platform offers a robust, sensitive approach for early detection of recurrence in a broad patient population.