To assess accuracy of multidetector computed tomography (MDCT) and individual radiological signs in the diagnosis of anastomotic leaks.Patients undergoing anterior resection with a stapled anastomosis over a 2 year period were identified. Electronic and clinical records of these patients were reviewed. Unenhanced and/or enhanced MDCT was performed with intravenous and/or per-rectal contrast medium and read by a radiologist blinded to the patients' clinical details to determine the sensitivity and specificity of specific findings at MDCT for identifying leaks.Seventeen percent (30/170) of the anterior resections were suspected to have an anastomotic leak. Ninety-three percent (28/30) of patients underwent MDCT. Seven point six percent (11+2/170) had a confirmed leak. Two patients underwent surgery without MDCT. A leak was confirmed by MDCT in 91% (10/11) of patients. The sensitivity, specificity, and positive and negative predictive values of MDCT in diagnosing a leak was 0.91, 1, 1, and 0.95, respectively. The sensitivity of peri-anastomotic air, peri-anastomotic collection, extravasation of rectal contrast medium, and staple line integrity was 0.81, 0.63, 0.54, and 0.72, respectively. Use of rectal contrast medium (8/11 cases) increased the subjective ease of diagnosis and was the only sign in one patient.Presence of peri-anastomotic air is a reliable marker of anastomotic leaks at MDCT. Leakage of rectal contrast medium is highly accurate and increases confidence of diagnosis. The appearance of the staple line itself is not accurate in assessing anastomotic integrity.