Use of Endoscopic Photographs to Assess for Regression of Barrettʼs Esophagus After Liquid Nitrogen Ablative Therapy: How Reliable Is the Endoscopistʼs Eye?

Abstract

Introduction: Objective endpoints for ablation trials of Barrett's esophagus (BE) are necessary to determine efficacy and dosimetry. While complete eradication of intestinal metaplasia (CEIM) is objective, it may take several sessions, making it time-consuming. We aimed to determine the utility of endoscopic photographs to identify % reversion to squamous epithelium after a single session of spray cryotherapy (SCT). Methods: In this prospective study, patients with ≥3cm of dysplastic BE received a single dose of SCT in a dose escalation protocol (initial 15 person cohort receiving 20 sec; second 15 person cohort receiving 30 sec) at each treatment site. Number of sites varied by BE length. Baseline endoscopy with 1 high-resolution white light (HRWLE) and 1 narrow band (NBI) photo per cm of BE from the top of gastric folds to the top of BE were recorded. At follow-up endoscopy 2 months later, photos were taken at the same locations using the same protocol. Paired before-and-after photos were reviewed by two expert endoscopists, masked to dosage, to estimate % of disease regression. Estimates within 15% between the two reviewers were concordant, and were averaged to determine final % regression. For >15% divergence, a third reviewer resolved discordance. If reviewer 3 was ≤15% of either of the first two, these two estimates were averaged. Cases with discordant estimates from reviewer 3 were excluded. Interobserver agreement was estimated using intraclass correlation coefficients (ICC), weighted kappa, and means of difference in agreement. A Bland-Altman plot showed the relationship between estimates. Results: Reviewers 1 and 2 were concordant in 23/30 (77%) of cases. Three patients had resolution of discordance by reviewer 3, and four were excluded due to discordance. A majority of the images were rated as good quality (Table). Mean difference in estimated BE regression between the two readers was 2.5% (95% CI: -4.2, 9.2). ICC of 0.87 and a weighted kappa of 0.67 reflected a good to high level of agreement between the two reviewers. The highest concordance was when either very little or almost all BE was eradicated (Figure). Overall, both the 20 sec and 30 sec doses of SCT showed similar interobserver agreement. Conclusion: Endoscopic photographs provided good reliability between two independent reviewers to estimate % regression after ablation. This technique may be preferable to CEIM as an outcome for ablation trials seeking to shorten trial duration.Figure 1Table 1: Comparison of Estimates of Percent Disease Regression after Liquid Nitrogen Ablative Therapy between Reviewer 1 and Reviewer 2