Pathologic complete response after neoadjuvant therapy for resectable esophageal squamous cell carcinoma: Endoscopic...

This study aimed to identify endoscopic characteristics and develop predictive models for detecting a pathologic complete response (pCR) after neoadjuvant therapy in patients with esophageal squamous cell carcinoma (ESCC).

This study enrolled 220 patients including a retrospective cohort (n = 158) and a prospective cohort (n = 62), from May 2018 to March 2023 with ESCC who received neoadjuvant chemoimmunotherapy (nCIT) or neoadjuvant chemotherapy (nCT) followed by surgery. Predictive capability of the endoscopic characteristics for pCR was developed and validated using machine learning.

All patients underwent endoscopic examinations before surgery but after neoadjuvant therapy. Cohort I was divided into a training set (n = 112) and an internal validation set (n = 46) at a 7:3 ratio. Seven endoscopic features were assessed: scarring; intraepithelial papillary capillary loop (IPCL) type B; depressed mucosa post-tumor disappearance; eroding mucosal changes with an uneven surface; nonsuperficial neoplastic lesions; protruded changes; and presence of cancer cells in biopsy specimens. Using these characteristics as predictors, a multivariate logistic regression model was trained to predict pCR. For further validation, data from prospective Cohorts II and III were incorporated. The model achieved 96.43% accuracy (95% confidence interval [CI] 91.11%-99.02%) in the training set, 93.48% (95% CI 82.10%-98.63%) for internal validation of Cohort I, and 96.77% (95% CI 88.83%-99.61%) in the prospective validation set.

Endoscopic characteristics are significant predictors of pCR in patients with ESCC receiving nCIT or nCT. The predictive model demonstrated high accuracy in both derivation and validation cohorts.