354 Early Postoperative Risk Prediction of Major Complications Following Robotic Oesophagectomy: A Multivariate Model

Abstract Aim To develop a risk prediction model for major complications following robotic assisted minimally invasive oesophagectomy (RAMIO), using early postoperative physiological markers. Method A retrospective cohort study was conducted on patients undergoing RAMIO (Ivor-Lewis or McKeown approach) between May 2022 and March 2025. Clinical, inflammatory, and haemodynamic variables within the first 72 hours post-operatively were extracted. The primary outcome was the development of major complications (Clavien-Dindo grade ≥IIIb). Univariate analysis identified candidate predictors, which were entered into multivariable logistic regression. Model performance was assessed using area under the receiver operating characteristic curve (AUC), Hosmer-Lemeshow goodness-of-fit, and McFadden’s Pseudo R². Results Among 168 RAMIO patients, 19 (11.3%) developed complications of Clavien-Dindo grade ≥IIIb. In multivariate analysis of early (<72h post-operatively) variables, peak WCC >17.1 (OR 12.65, p < 0.001), combined drain output >639 ml/24h (OR 5.45, p = 0.008), and noradrenaline use (OR 3.32, p = 0.037) were independently associated with major complications. The model demonstrated strong discrimination (AUC = 0.84, 95% CI 0.75–0.93) and good calibration (Hosmer-Lemeshow p = 0.76). Conclusions Early postoperative variables, particularly WCC, drain output, and noradrenaline use, independently predict major complications following RAMIO. This model provides a practical, real-time tool to support early risk stratification and targeted intervention. Integration into enhanced recovery protocols and validation in external cohorts are warranted to refine its clinical utility.