An esophagovisceral anastomotic leak is a life-threatening postoperative complication, especially in the mediastinum. Of the 242 patients who underwent intrathoracic esophagogastric anastomosis for esophageal carcinoma (182 patients) and adenocarcinoma of the cardia (60 patients) between January 1980 and June 1985, 14 (5.8%) had esophageal anastomotic leakage and two died (0.8%). Various clinical and biologic parameters and aspects of operative technique were studied prospectively and analyzed statistically to identify possible factors responsible for leaks. Both bivariate and multivariate statistical analysis with logistic regression showed that the following clinical and biologic factors do not influence anastomotic leakage: tumor stage, the curative or palliative purpose of resection, neoplastic permeation of anastomotic margins, total protein concentration below 5 gm/dl, albumin concentration below 3 gm/dl, patient's age, diabetes, high blood pressure, cirrhosis of the liver, and cardiac, respiratory, or renal diseases. Technical factors, on the contrary, were statistically significant and of great clinical importance: manual as opposed to mechanical suturing (chi 2 = 8.8, p = 0.013) and single-layer as opposed to double-layer suturing (chi 2 = 9.9, p = 0.043). The level of the anastomosis was found to be a further statistically significant factor: The incidence of leakage was greater when the anastomosis was located between the azygos vein and the lower pulmonary vein (chi 2 = 15.5, p = 0.004) than above the azygos vein or below the lower pulmonary vein.