Sixty consecutive patients undergoing anterior restorative resection of the rectum for cancer were studied. After full mechanical bowel preparation and intravenous antibiotic prophylaxis, colorectal anastomosis was performed with the EEA (Autosuture) staple gun. The median height of the anastomosis above the anal verge was 9 cm (range 3-15 cm) and no patient had a protecting colostomy fashioned. All 'doughnuts' were checked and following instillation of saline into the pelvis the anastomosis was tested by air insufflation through the rectum. There was a statistically significant association between an incomplete 'doughnut' and peroperative anastomotic leakage (P less than 0.001); however, of the 11 patients shown to have a leak, only seven had deficient 'doughnuts'. All 11 patients had repair of the defect identified by escape of air and none of these patients developed a clinical leak.