BIOMARKER RISK STRATIFICATION USING THE CAPSULE SPONGE FOR THE SURVEILLANCE OF BARRETT'S ESOPHAGUS: INTERIM RESULTS FROM THE UK REAL-WORLD IMPLEMENTATION PILOTS.

Introduction
Endoscopic surveillance is recommended for Barrett’s esophagus (BE), but the effectiveness is questionable and varies according to expertise. Pan-oesophageal cell collection devices with biomarkers offer a less operator dependent alternative. We previously developed a simple risk score for capsule-sponge surveillance comprising: low (clinical and sponge biomarkers negative), moderate (positive clinical biomarkers, negative sponge biomarkers) and high-risk (positive sponge biomarkers, i.e. atypia or atypia of uncertain significance (AUS), or p53 positive) to prioritize patients for their next endoscopy.¹ Here, we aim to prospectively evaluate the accuracy of the risk stratification tool in patients undergoing capsule-sponge triage for BE surveillance.

Methods
BE patients undergoing capsule-sponge at 5 hospitals within the DELTA (Jan 2020-2022, ISRCTN91655550) and NHS England (Jan 2022-ongoing) implementation studies with endoscopic follow-up were included. All samples were processed in a centralized, accredited laboratory and positive biomarkers were confirmed by 2 pathologists. The responsible clinician determined endoscopic follow-up intervals informed by the risk group – urgent endoscopy in high risk (prevalent dysplasia) versus delayed endoscopy in low and moderate risk (prevalent and incident dysplasia). The primary outcome was the PPV and NPV for detecting dysplasia according to the risk groups. The secondary outcome was the PPV for different sponge biomarker combinations among the high-risk group (atypia, atypia uncertain significance (AUS), p53) for a diagnosis of dysplasia.

Results
Of 326 patients, the risk category breakdown was 172 (52.8%) low risk; 93 (28.5%) moderate risk, and 61 (18.7%) high risk. The PPV of the high-risk category for dysplasia was 44.3% (27/61) (13 LGD or crypt dysplasia, 10 HGD, 3 intramucosal, 1 adenocarcinoma) Within the high-risk category the PPV for dysplasia when p53 and atypia were positive was 92.8% (13/14), compared to 50% when either atypia (1/2) or p53 (4/8), were positive; whereas AUS included inflammatory changes and had less predictive potential with a PPV for dysplasia of 19.4% (6/31). The NPV for dysplasia in low-risk patients was 98.3% (169/172), with 1.7% (3/172) having dysplasia (1 LGD, 1 crypt dysplasia, 1 HGD) after a mean of 351 days of follow-up. The NPV of a moderate risk for dysplasia was 96.8% (90/93) with 3.2% (3/93) having dysplasia (2 LGD and 1 HGD). There were no cases of intramucosal/invasive cancer in the low and moderate risk groups (Table 1).

Conclusion
A clinically applicable risk panel provides accurate stratification in a real-world BE surveillance setting. This simple capsule-sponge procedure has potential as a less operator dependent, cost-effective and patient-friendly triage to help prioritize endoscopy requirements.