Experimental carcinogenesis is enhanced at colorectal anastomoses, inhibited by proximal faecal diversion and promoted by the closure of a defunctioning stoma. The clinical relevance of these observations was investigated in a retrospective study of curative restorative resection for colorectal carcinoma. The 5-year disease-free survival rate (95 per cent confidence interval) in 122 patients with a temporary stoma (50.4 (41.1-59.7) per cent) was significantly reduced (P < 0.01) compared with that in 218 with no stoma (66.8 (59.4-73.5) per cent). In patients with Dukes B tumours early stoma closure (within 3 months of resection) was associated with a worse survival (P < 0.005) and a higher tumour recurrence rate (P < 0.05) than in those with no stoma. Survival rates after late stoma closure were no different from those in patients with no stoma. Multivariate analysis revealed Dukes stage (P < 0.0001), tumour differentiation (P = 0.02) and timing of stoma closure (P = 0.02) as independent predictors of survival. In curative surgery for colorectal cancer temporary faecal diversion confers a survival disadvantage that can be prevented by delayed closure of the stoma.