Clinical predictors of early reflux and symptom severity after esophagectomy: a risk prediction tool

BACKGROUND: Reflux after curative esophagectomy is common and may persist, disrupting eating, sleep, and recovery, yet most prediction tools estimate risk only at a single late time point. METHODS: We retrospectively analyzed an esophagectomy derivation cohort (2018-2021; n = 488) and a temporal validation cohort (2023-2024; n = 143). Clinically significant reflux was defined as an EORTC QLQ-OES18 item 14 or 15 score ≥ 2. Time to first event within 12 months was modeled using a time-varying coefficient Cox model, and longitudinal symptom severity (1-4) was modeled using an ordinal mixed-effects model. Predictors were prespecified from routine pre-discharge variables. Model performance was assessed using time-dependent AUC and calibration. RESULTS: Cox analysis linked distal/esophagogastric junction tumors and higher BMI to increased early reflux risk, while neoadjuvant therapy, older age, and a longer oral-to-solid diet interval were protective. The ordinal mixed model showed greater symptom severity in women and in patients with distal/esophagogastric junction tumors, but lower severity with neoadjuvant therapy. Interactions suggested slower early symptom escalation with older age and steeper escalation with higher BMI or shorter oral-to-intake intervals. The time-varying Cox model showed good discrimination, with time-dependent AUCs at 1, 3, and 6 months of 0.702, 0.766, and 0.794 in the development cohort, and 0.708, 0.767, and 0.716 in the external validation cohort, with good calibration. CONCLUSIONS: A pre-discharge, time-sensitive risk model with a web-based calculator may support risk-stratified early screening, education, and targeted supportive care after discharge to reduce reflux-related morbidity and protect postoperative health-related quality of life.