Several methods can be used to re-establish gastrointestinal continuity after the standard Whipple's Billroth II procedure or Longmire pylorus-preserving pancreatoduodenectomy (PPPD). Standard pancreatoduodenectomy abolishes the postprandial gastrin response whereas the response after Longmire PPPD is similar to that in controls. A novel reconstruction after PPPD has been designed to separate the biliary and pancreatic secretions by restoring continuity in the Billroth I manner, retaining the gastric antrum and with the pancreatic remnant anastomosed to a separate defunctioned Roux loop. Gastrin and cholecystokinin (CCK) responses were measured in the fasting state and after a standard liquid meal of 250 ml containing 15 per cent protein, 30 per cent fat and 55 per cent carbohydrate with an energy value of 525 kJ. Basal gastrin and CCK concentrations were similar in five healthy unoperated controls, and after standard pancreatoduodenectomy (five patients) and isolated Roux loop PPPD (six patients). In controls and after the new PPPD operation, but not after standard pancreatoduodenectomy, mean(s.e.m.) postprandial plasma gastrin values rose from 8.66(0.33) to 13.00(0.33) pmol l-1 at 10 min (controls) and from 7.66(1.56) to 15.00(1.36) pmol l-1 at 10 min (defunctioned Roux loop PPPD); the concentration remained raised for 20 min then fell transiently; a second peak was maintained for 40 min (controls) to 60 min (modified PPPD). In controls, mean(s.e.m.) postprandial plasma CCK concentrations rose from 0.43(0.12) basal to 10.50(1.03) pmol l-1 and returned to basal after 60 min. After standard pancreatoduodenectomy postprandial CCK concentrations rose at a similar rate and remained raised for 80 min. After isolated Roux loop PPPD plasma CCK concentrations remained low (mean(s.e.m.) 0.38(0.29) pmol l-1) after the test meal.