Data availability. The WGS and RNA expression data can be found at the European Genome-phenome Archive (EGA) under accessions EGAD00001004417 and EGAD00001004423, respectively. Code availability. Code associated with the analysis is available upon request. Ethics. The study was registered (UKCRNID 8880), approved by the Institutional Ethics Committees (REC 07/H0305/52 and 10/ H0305/1), and all subjects gave individual informed consent. Reporting summary. Additional information is included in the Life Sciences Reporting Summary, which details exact software and biological materials used and efforts made to ensure reproducibility of results. Author contributions RCF and AMF conceived the overall study. AMF and SJ analyzed the genomic data and performed statistical analyses. RCF, AMF and XL designed the experiments. AMF, XL and JM performed the experiments. GC contributed to the structural variant analysis and data visualization. SK helped compile the clinical data and aided statistical analyses. JP and SA produced and QC'ed the RNA-seq data. EO aided the whole genome sequencing of EAC cell lines. SM and NG coordinated the clinical centres and were responsible for sample collections. ME benchmarked our mutation calling pipelines. MO led the pathological sample QC for sequencing. LB and GD constructed and managed the sequencing alignment and variant calling pipelines. RCF and ST supervised the research. RCF and ST obtained funding. AMF and RCF wrote the manuscript. All authors approved the manuscript.

Background: The incidence of esophageal adenocarcinoma (EAC) is rapidly rising and has a 5-year survival rate of <20%. Beyond TNM (tumorenodeemetastasis) staging, no reliable risk stratification tools exist and no large-scale studies have profiled circulating tumor DNA (ctDNA) at relapse in EAC. Here we analyze the prognostic potential of ctDNA dynamics in EAC, taking into account clonal hematopoiesis with indeterminate potential (CHIP). Patients and methods: A total of 245 samples from 97 patients treated with neoadjuvant chemotherapy and surgery were identified from the prospective national UK Oesophageal Cancer Clinical and Molecular Stratification (OCCAMS) consortium data set. A pan-cancer ctDNA panel comprising 77 genes was used. Plasma and peripheral blood cell samples were sequenced to a mean depth of 7082Â (range 2196-28 524) and ctDNA results correlated with survival. Results: Characteristics of the 97 patients identified were as follows: 83/97 (86%) male, median age 68 years (SD 9.5 years), 100% cT3/T4, 75% cNþ. EAC-specific drivers had higher variant allele fractions than passenger mutations. Using stringent quality criteria 16/79 (20%) were ctDNA positive following resection; recurrence was observed in 12/16 (75%) of these. As much as 78/97 (80%) had CHIP analyses that enabled filtering for CHIP variants, which were found in 18/78 (23%) of cases. When CHIP was excluded, 10/63 (16%) patients were ctDNA positive and 9/10 of these (90%) recurred. With correction for CHIP, median cancer-specific survival for ctDNA-positive patients was 10.0 months versus 29.9 months for ctDNA-negative patients (hazard ratio 5.55, 95% confidence interval 2.42-12.71; P ¼ 0.0003). Similar outcomes were observed for disease-free survival. Conclusions: We demonstrate in a large, national, prospectively collected data set that ctDNA in plasma following surgery for EAC is prognostic for relapse. Inclusion of peripheral blood cell samples can reduce or eliminate false positives from CHIP. In future, post-operative ctDNA could be used to risk stratify patients into high-and low-risk groups for intensification or de-escalation of adjuvant chemotherapy.