Genome-wide and size-based cell-free DNA indices as predictive biomarkers for locally advanced esophageal squamous cell carcinoma treated with preoperative or definitive chemoradiotherapy

Abstract

For locally advanced esophageal cancer, concurrent chemoradiotherapy (CRT) followed by surgery has been a standard treatment, while clinical studies showed comparable survival outcomes between definitive CRT and neoadjuvant CRT followed by surgery in patients responding to CRT. Thus, biomarkers are required to predict treatment outcomes and benefit of adding surgery after CRT. This prospective biomarker study examined the role of cell-free DNA (cfDNA) fragmentation profiles and genomic copy number variations (CNVs) in predicting treatment outcomes in esophageal squamous cell carcinoma patients treated with neoadjuvant or definitive CRT. The clinical response was evaluated after induction chemotherapy and after CRT. Fragment Ratio (FR)-score and I-score were calculated from plasma cfDNA reflecting fragment lengths and CNV of cfDNA, respectively. The association between indices of cfDNA (cfDNA concentration, FR-score, and I-score) and treatment outcomes (clinical response, time to progression [TTP], and overall survival [OS]) were evaluated. Sixty-one patients were included. Thirty patients received neoadjuvant CRT followed by surgery, whereas 31 received definitive CRT. Low baseline, post-induction chemotherapy, and post-CRT FR-scores and low post-induction I-score were significantly associated with improved treatment response (P < 0.05). Additionally, patients with surgery after CRT showed significantly longer survival than patients without surgery in the FR-score-high group (median TTP, 12.7 vs 3.4 months; P = 0.011; OS, not reached vs 12.9 months; P = 0.02), while there was no survival benefit with surgery in the FR-score-low group. FR-score may be a new biomarker to predict treatment response, residual tumor burden after CRT, and consequently, survival benefit of adding morbid surgery after CRT. FR-score has strength in a relatively simple and inexpensive methodology compared to deep sequencing, resulting in high availability and accessibility, despite limited sensitivity.