To investigate the feasibility of a new mode to diagnose and treat intrathoracic gastroesophageal anastomotic leak. From January 2007 to December 2014, fifty-five patients were confirmed intrathoracic gastroesophageal anastomotic leak among those were performed surgical operation due to esophageal or cardiac carcinoma in the First Affiliated Hospital of Soochow University. To retrospectively analyze the clinical data of these patients, thirty-six male and nineteen female were included with the ages from 49 to 81 years (average age of (67±6)years). Among them, forty-two were middle esophageal carcinoma, eleven were lower esophageal carcinoma and two were cardiac carcinoma. According to the differences of diagnosis and treatment methods for anastomotic leak, fifty-five patients were divided into two groups. Thirty-one patients distributed from January 2007 to November 2011 were received conventional management (conventional group): to definitively diagnose by contrast swallow when suspected to be developing anastomotic leaks, to place an esophageal stent when the drainage was sufficient and the infection was controlled. Twenty-four patients distributed from March 2011 to December 2014 were received new-mode management (new-mode group): to perform a anastomotic radioscopy under digital subtraction angiography -guidance instantly when suspected anastomotic leak and find out the fistula, search the shape and size, place a drainage tube into the fistula to drain or lavage the vomica according to the exploration results, pull back the tube gradually and close the leak by clips under endoscope later. The pathoclinical features, the confirmation time (time from clinical signs emergence to leak confirmation), the hospital duration after confirmation, the incidence of severe complications and total mortality were compared between the two groups by t-test and χ(2) test or Fisher's exact test. There was no significant statistical differences in pathoclinical features between two groups (P>0.05). The confirmation time was significantly reduced in new-mode group than that in conventional group ((1.2±0.8) d vs. (3.6±2.2) d, t=5.212, P=0.000), and so was the hospital duration after confirmation ((26±12) d vs. (55±25) d, t=4.992, P=0.000) and the incidence of severe complications (16.7% vs. 48.4%, χ(2)=6.019, P=0.014), although there was no statistical differences in total mortality (4.2% vs. 22.6%, P=0.119). The new mode of early interventional diagnosis, early fistula drainage through nose and clipping under endoscope later is able to shorten diagnosis and treatment period, reduce incidence of severe complications.