A retrospective study was performed of 113 patients who underwent surgical resection of carcinoma of the pancreas from 1970 to 1992. The postoperative mortality rate was 15 per cent (5 per cent in the last 11 years). The actuarial 5-year survival rate was 12 per cent. Survival was significantly influenced by age (P = 0.03), vascular resection (P = 0.02), radicality of operation (P = 0.01), number of transfused blood units (P = 0.01), tumour differentiation (P = 0.002), lymph node status (P = 0.001), perineural invasion (P = 0.01), tumour size (P = 0.008), preoperative diabetes (P = 0.001) and stage (P = 0.0001). Multivariate analysis showed that stage, diabetes, age and grade were independent predictors of long-term survival. The type of pancreatic resection (Whipple, subtotal, total or distal pancreatectomy) did not influence prognosis. The 5-year survival rate was 14 per cent in the period 1970-1981 and 11 per cent in the period 1982-1992, with no statistical difference. These results suggest that patient characteristics and tumour findings rather than operative procedures affect long-term survival after resection for pancreatic carcinoma.