A multicentre retrospective study was carried out to analyse short- and long-term results of 787 pancreatic resections performed for cancer between 1982 and 1988. The postoperative mortality rate was 10 per cent and the morbidity rate 35 per cent. Age above 70 years and systemic organ failure independently influenced operative mortality. In patients surviving more than 30 days the median survival was 12.3 months and the actuarial survival rate at 5 years 12 per cent. The 5-year survival rate was lower for patients with lymph node involvement than for those without (4 versus 20 per cent, P = 0.001). The operative mortality rate was higher after total pancreatectomy than pancreatoduodenectomy (17 versus 8 per cent, P = 0.015). The median survival time and 5-year survival rate after total pancreatectomy and pancreatoduodenectomy were 11 versus 14 months and 3 versus 15 per cent respectively. Of the clinical and pathological factors studied, location of the tumour in the left pancreas was most strongly related to survival, with no survivors at 4 years. These results suggest that resection should be avoided in patients over 70 years old with systemic organ failure. Pancreatoduodenectomy remains the best procedure for resection, total pancreatectomy being performed only in patients with multifocal carcinoma or those in whom a safe pancreatic anastomosis cannot be constructed.