Liquid biopsy to identify Barrett's oesophagus, dysplasia and oesophageal adenocarcinoma: the EMERALD multicentre study

Abstract

Background: There is no clinically relevant serological marker for the early detection of oesophageal adenocarcinoma (EAC) and its precursor lesion, Barrett's oesophagus (BE).

Objective: To develop and test a blood-based assay for EAC and BE.

Design: Oesophageal MicroRNAs of BaRRett, Adenocarcinoma and Dysplasia (EMERALD) was a large, international, multicentre biomarker cohort study involving 792 patient samples from 4 countries (NCT06381583) to develop and validate a circulating miRNA signature for the early detection of EAC and high-risk BE. Tissue-based miRNA sequencing and microarray datasets (n=134) were used to identify candidate miRNAs of diagnostic potential, followed by validation using 42 pairs of matched cancer and normal tissues. The usefulness of the candidate miRNAs was initially assessed using 108 sera (44 EAC, 34 EAC precursors and 30 non-disease controls). We finally trained a machine learning model (XGBoost+AdaBoost) on RT-qPCR results from circulating miRNAs from a training cohort (n=160) and independently tested it in an external cohort (n=295).

Results: After a strict process of biomarker discovery and selection, we identified six miRNAs that were overexpressed in all sera of patients compared with non-disease controls from three independent cohorts of different nationalities (miR-106b, miR-146a, miR-15a, miR-18a, miR-21 and miR-93). We established a six-miRNA diagnostic signature using the training cohort (area under the receiver operating characteristic curve (AUROC): 97.6%) and tested it in an independent cohort (AUROC: 91.9%). This assay could also identify patients with BE among patients with gastro-oesophageal reflux disease (AUROC: 94.8%, sensitivity: 92.8%, specificity: 85.1%).

Conclusion: Using a comprehensive approach integrating unbiased genome-wide biomarker discovery and several independent experimental validations, we have developed and validated a novel blood test that might complement screening options for BE/EAC.

Trial registration number: NCT06381583.

Keywords: BARRETT'S OESOPHAGUS; DYSPLASIA; OESOPHAGEAL CANCER; TUMOUR MARKERS.

Figure 1

CONSORT diagram for study cohorts allocation. BE, Barrett’s oesophagus; CONSORT, Consolidated Standards of Reporting Trail; EAC, oesophageal adenocarcinoma; EMERALD, Oesophageal MicroRNAs of BaRRett, Adenocarcinoma and Dysplasia; NDC, non-disease control; TCGA, The Cancer Genome Atlas.

Figure 2

In silico discovery and prioritisation of candidate miRNAs. (A) Volcano plot of differentially expressed microRNAs between cases and controls using TCGA miRNA expression dataset. Colour grading follows significance. 22 miRNAs were differentially expressed and 14 miRNAs (identified as fully coloured) were prioritised as our initial candidates for further analysis. (B) Ridgeline plot of the initial pool of 14 microRNAs; (C) A heatmap with unsupervised clustering illustrates the expression levels of the 14 candidate miRNAs in the TCGA miRNA expression dataset. FDR, false discovery rate; TCGA, The Cancer Genome Atlas.

Figure 3

Assessment of the six miRNAs diagnostic potential in the US (development) cohort. Six circulating miRNAs were significantly upregulated in cases (EAC, HGD, LGD and BE) compared with NDCs and demonstrated AUROC values of diagnostic interest, ranging from 62.1% to 80.8%. In subgroup analysis (inserts), all candidate miRNAs demonstrated a significant differential expression between NDCs (blue) versus EAC/HGD (maroon), five for NDCs versus BE/LGD (orange) and two for BE/LGD versus EAC/HGD (*p<0.05, **p<0.01, ***p<0.001; two-sided Student’s t-tests). AUROC, area under the receiver operating characteristic curve; BE, Barrett’s oesophagus; EAC, oesophageal adenocarcinoma; HGD, high-grade dysplasia; LGD, low-grade dysplasia; NDC, non-disease controls; ns, not significant; Sen, sensitivity; Spe, specificity.

Figure 4

Establishment and external testing of the six circulating-miRNA signature (A) AUROC and performance metrics of the two first-level classifiers (XGB: XGBoost, orange; ADA, AdaBoost, black) and the resulting stacked model (EMERALD) in the Italian (training) cohort. (B) AUROC and performance metrics of the two first-level classifiers (XGB: XGBoost, orange; ADA, AdaBoost, black) and the resulting stacked model (EMERALD) in the UK/Irish (testing) cohort; (C, D) Density-scatter plots: the background colour gradient is a density plot and highlights the areas where most observations are encountered, without knowledge of them being cases and controls. The dots (green for cases and pink for NDCs) represent the scatter plot. The same patterns can be seen in both the Italian (training, C) and UK/Irish (testing, D) cohorts, with two clusters where cases and controls appear to separate (top right and bottom left, respectively). AUROC, area under the receiver operating characteristic curve; EMERALD, Oesophageal MicroRNAs of BaRRett, Adenocarcinoma and Dysplasia.

Figure 5

Evaluation of the diagnostic performance of the six-circulating-miRNA signature. (A, B) Raincloud plots with super-imposed box and whisker plots demonstrating the distribution of EMERALD values between cases and controls in the Italian (training, A) and UK/Irish (testing, B) cohorts; (C, D) ORs for the presence of a compound endpoint of EAC, HGD, LGD or BE with restricted cubic splines in the Italian (training, C) and UK/Irish (testing, D) cohorts. BE, Barrett’s oesophagus; EAC, oesophageal adenocarcinoma; EC, Oesophageal cases; HGD, high-grade dysplasia; LGD, low-grade dysplasia; NDC, Non-disease controls.

Figure 6

Evaluation of the EMERALD assay among symptomatic controls with GERD. (A) Waterfall plot of the EMERALD values in the UK/Irish (testing) cohort, with controls sublabelled by the presence of symptoms; (B) Violin plot of EMERALD values according to the presence of GERD-related symptoms, with and without reflux oesophagitis versus cases with BE or LGE and cases with EAC or HGD (****p<0.0001; two-sided Student’s t-tests). (C, D) AUROCs of the two first-level classifiers (XGB: thin line; ADA, dashed line) and the stacked model (thick line with 95% CIs) in a subanalysis of the UK/Irish (testing) cohort which only included symptomatic controls versus BE cases, with or without LGD (C) and cases with either HGD or adenocarcinoma (D).AUROC, area under the receiver operating characteristic curve; BE, Barrett’s oesophagus; EAC, oesophageal adenocarcinoma; EMERALD, Oesophageal MicroRNAs of BaRRett, Adenocarcinoma, and Dysplasia; GERD, gastro-oesophageal reflux disease; HGD, high-grade dysplasia; LGD, low-grade dysplasia; RE, reflux oesophagitis.

Figure 7

Markov simulation. (A) Markov model structure: for each strategy, the Markov chain assumes a progression through pre-malignant disease, malignant disease, cancer recovery, recurrence, and, eventually death. Mortality can be from any cause (cardiovascular, other cancers, etc) or stage-specific mortality rates. Early detection of Barrett’s oesophagus with dysplasia is assumed to trigger its eradication. EAC diagnosis, whether due to symptoms, endoscopic screening uptake, or non-invasive screening uptake, is assumed to trigger treatment. Costs and utility values are derived from the literature or, if unavailable, from internal estimates. The starting condition of each individual in the Markov chain is a 45-year-old with chronic GERD who is followed up for 30 years or until death. Five screening strategies are tested: non-invasive, EMERALD-based screening at 45% compliance every 5, 3 or 1 year(s), endoscopy-based screening at 10% compliance every 10–15 years, or no screening. (B) Number of stage I EAC diagnoses per screening strategy. (C) Number of stage II EAC diagnoses per screening strategy. (D) Number of stage III EAC diagnoses per screening strategy. (E) Number of stage IV EAC diagnoses per screening strategy. (F) Number of deaths from all causes (both EAC and non-EAC related) per screening strategy. (G) Annual costs associated with each screening strategy. (H) Quality-adjusted life-years per year by screening strategy, with a call-out box to demonstrate the effects of different screening strategies. (I) Incremental costs and effectiveness of each screening strategy against the most cost-effective screening strategy and interval. EAC, oesophageal adenocarcinoma; GERD, gastro-oesophageal reflux disease.