ObjectiveTo investigate the characteristics and predictors for anastomotic leakage after oesophagectomy for oesophageal carcinoma from the perspective of anastomotic level.DesignRetrospective cohort study.SettingsA single tertiary medical centre in China.ParticipantsFrom January 2010 to December 2016, all patients with oesophageal cancer of the distal oesophagus or gastro-oesophageal junction undergoing elective oesophagectomy with a curative intent for oesophageal carcinoma with intrathoracic oesophagogastric anastomosis (IOA) versus cervical oesophagogastric anastomosis (COA) were included. We investigated anastomotic level and perioperative confounding factors as potential risk factors for postoperative leakage by univariate and multivariate logistic regression.Primary outcome measuresThe primary outcome was the odds of anastomotic leakage by different confounding factors. Secondary outcome was the association of IOA versus COA with other postoperative outcomes.ResultsOf 458 patients included, 126 underwent cervical anastomosis and 332 underwent intrathoracic anastomosis. Anastomotic leakage developed in 55 patients (12.0%), with no statistical differences between COA and IOA (16.6% vs 10.2%; p=0.058). Multivariable analysis identified active diabetes mellitus (OR 2.001, p=0.047), surgical procedure (open: reference; minimally invasive: OR 1.770, p=0.049) and anastomotic method (semimechanical: reference; stapled: OR 1.821; handsewn: OR 2.271, p=0.048) rather than anastomotic level (IOA: reference; COA: OR 1.622, p=0.110) were independent predictors of leakage.ConclusionsSurgical and anastomotic techniques rather than the level of anastomotic site were independent predictors of postoperative anastomotic leakage in patients undergoing oesophageal cancer surgery.