Introduction
Radiofrequency ablation (RFA) is effective to eradicate dysplastic Barrett’s esophagus (BE), however recurrence occurs in up to 25% of cases, typically at the gastro-esophageal junction (GEJ). We have shown that a multi-gene methylation panel (TFPI2, ZNF345 and ZNF569) can differentiate BE from normal GEJ mucosa. This study aims to determine whether this methylation panel correlates with dysplasia and can predict recurrence post-RFA.

Methods
This prospective cohort study (PROBAN) recruited patients with dysplastic BE, who achieved endoscopic remission post-RFA at two centers. Patients were eligible if they had 1) no endoscopic evidence of BE, 2) no suspicious dysplastic lesion at GEJ, and 3) no histological evidence of esophageal intestinal metaplasia (IM) at first post-RFA follow-up (GEJ IM allowed). Patients received endoscopy at baseline, 6, 12 and 24 months. Methylation score (Meth-score) was assessed by Methylight assay on random and narrow-band imaging targeted GEJ biopsies. The primary outcome was the correlation between Meth-score and histology of biopsies. The secondary outcome was the correlation of baseline methylation level with dysplastic recurrence at any time during follow-up.

Results
We included 124 patients who received 261 endoscopies. Mean age was 70.0 years (standard deviation [SD], 7.5 years). Mean pre-RFA maximum BE length was 4.6cm (SD, 3.1). The breakdown of pre-RFA BE histology were low-grade dysplasia (n=24, 19.5%), high-grade dysplasia (n=49, 39.8%) and intramucosal carcinoma (n=50, 40.7%). Overall, 12 patients (9.7%) developed dysplastic recurrence, all of which at the GEJ. On a per-endoscopy level, the median Meth-score of GEJ biopsies comparing patients with gastric metaplasia (GEJ-GM, n=185) intestinal metaplasia (GEJ-IM, n=63) and dysplasia (n=13) was 0%, 18.6%, and 50.0%, respectively, with a significance difference across the three histology groups (Kruskal-Wallis p<0.001) (Figure 1). When evaluating baseline (time 0) Meth-score and dysplastic recurrence at any time during follow-up, patients with dysplastic recurrence exhibited significantly higher median methylation levels compared to patients with no dysplasia (48.4% vs 0.5%, Mann-Whitney p<0.001) (Figure 2). A Meth-score cut-off of 0.73% had a sensitivity and a specificity for dysplastic recurrence of 92% and 55%, respectively.

Conclusion
A multi-gene methylation panel can discriminate patients with GM, IM and dysplasia on GEJ biopsies. Patients with stable histological remission had low baseline methylation levels. This biomarker panel can risk-stratify patients so that approximately half of patients with endoscopic remission post-RFA could benefit from relaxed surveillance intervals or early discharge in the appropriate clinical context, while those with high methylation levels could be offered retreatment of the GEJ before histological recurrence.

<b>Figure 1</b>. Median total methylation of individual endoscopies at any timepoint, and the corresponding histology comparing gastric metaplasia (GM) versus intestinal metaplasia (IM) versus dysplasia.

Figure 1. Median total methylation of individual endoscopies at any timepoint, and the corresponding histology comparing gastric metaplasia (GM) versus intestinal metaplasia (IM) versus dysplasia.

<b>Figure 2. </b>Median baseline methylation levels (time 0) and highest histology recorded at any time during follow-up (up to 24 months) comparing participants without dysplastic vs those with dysplastic recurrence.

Figure 2. Median baseline methylation levels (time 0) and highest histology recorded at any time during follow-up (up to 24 months) comparing participants without dysplastic vs those with dysplastic recurrence.

Ines Modolell

Appearances