Oesophageal endoscopic submucosal dissection (ESD) is increasingly used for the treatment of early oesophageal neoplasia. Post-ESD stricture remains a frequent adverse event, often requiring repeated therapeutic endoscopy. Although several factors have been proposed to influence stricture risk, reported associations vary and the relative importance of individual factors remains uncertain. We aimed to identify predictors of stricture requiring therapeutic endoscopic intervention (SRT) following oesophageal ESD in a real-world cohort.

We performed a retrospective analysis of all oesophageal ESDs with accessible follow-up undertaken at our centre between July 2023 and September 2025. Collected variables included demographic, lesion-related (location, length, circumferential extent, histologic subtype, depth of invasion, histology-based risk category); and procedural characteristics (knife type, endoscopist, procedural complication).

The primary outcome was SRT. Secondary outcomes included the number of endoscopic dilatations at predefined intervals (2 months, 4 months, 6 months, last follow-up) and change in dysphagia grade from pre-procedure to peak post-procedure and last follow-up. Clinically important events including oesophageal stenting, non-oral feeding, and stricture-related perforation were also recorded. All variables underwent univariate testing, and the strongest univariate predictors were entered into an exploratory multivariable model.

70 oesophageal ESDs were included in the analysis (median age 71.5 years [IQR 62–77], 71% male). Lesions were located in the upper (6%), mid (26%), lower (47%) oesophagus or at the gastro-oesophageal junction (19%). Median lesion length was 40 mm (IQR 30–60). Median circumferential extent was 50% (IQR 26.3–78.8), with 22/70 (31.4%) involving ≥75% of the circumference. Histology comprised glandular neoplasia in 61.4%, squamous neoplasia in 32.9% and other subtypes. Depth of invasion included no invasion (24.3%), T1a (50.0%) and T1b (22.9%).

28 patients (40%) developed SRT. On univariate analysis, circumferential extent showed a strong association with SRT (median 87.5% [IQR 46.5–100%] vs 31.5% [IQR 25–50%]; p<0.0001), while all other variables were non-significant. In the multivariable model containing the strongest univariate predictors, circumferential extent remained the only independent predictor of SRT (OR 1.04 per 1%, 95% CI 1.014–1.061; p=0.0016). Histologic subtype (OR 2.92, 95% CI 0.91–9.41; p=0.073), lesion length (OR 0.996, 95% CI 0.98–1.02; p=0.70) and age (OR 1.009, 95% CI 0.95–1.07; p=0.78) were not significant.

Circumferential extent also demonstrated moderate to strong associations with dilatation burden across all intervals (Spearman ρ=0.55–0.59, p<0.0001), was strongly associated with peak dysphagia worsening (ρ=0.60, p<0.0001), and moderately associated with dysphagia at last follow-up (ρ=0.40, p=0.001). Lesion length showed weak associations with dilatation number and peak dysphagia (ρ≈0.25, p<0.04), and histologic subtype showed borderline associations with dilatation burden (p≈0.05–0.09) but no association with dysphagia. Neither variable remained significant after adjusting for circumferential extent. All other variables showed no meaningful association with any secondary outcome.

All clinically important events were confined to 100% circumferential ESDs, including oesophageal stenting (n=5), need for non-oral feeding (n=2), and dilatation-related perforation (n=3).

Circumferential extent was the sole independent predictor of SRT following oesophageal ESD and the only factor consistently associated with stricture severity across dysphagia and dilatation outcomes. These findings highlight circumferential involvement as the dominant determinant of clinically significant post-ESD stricture and support its use as the primary basis for risk stratification and preventive strategies.