The relations between incidence and prognosis of postoperative fistulas after gastrectomy and some different variables were analysed in the present retrospective study. Thirteen digestive fistulas of 113 patients (11.9%) submitted to gastrectomy during the period 1989-1994 represent the study population. The incidence of postoperative fistulas was compared to the kind of gastric pathology, to the extension of gastrectomy, to different nutritional (serum haemoglobin, albumin and transferrin level, weight loss) and immunological factors (serum lymphocytes) and, for oncological patients, to the stage of the disease. Incidence was directly related to the extension of gastrectomy, to serum albumin and haemoglobin level, and to weight loss rate. The results were not statistically significant at Kruskal-Wallis and ANOVA tests. No relation was found between incidence of fistulas and serum transferrin level, number of lymphocytes and adoption of early postoperative enteral nutrition. Six patients had spontaneous closure of the fistula with conservative therapy. Seven patients required reoperation because of abdominal sepsis (53.8%). Three patients died (23%). Although spontaneous closure, reoperation and mortality were related to nutritional and immunological state, no examined variables showed a statistically significative relation. The adoption of early postoperative enteral nutrition was not related to the prognosis, unlike the stage of the disease: patients submitted to reoperation had a TNM III or IV stage; dead patients had a TNM IV stage. Treatment of metabolic-nutritional unbalance can prevent anastomotic failure and fistula after gastrectomy and improve the prognosis. The relation between early postoperative enteral nutrition and incidence and prognosis of postoperative fistulas remains unclear.