Of 565 patients with thrombocytopenia admitted to Duke University Hospital between 1975 and 1985, 100 had splenectomy. Ninety-eight patients had failed chronic immunosuppressive therapy and three patients had acute intracranial bleeding or total absence of platelets in the peripheral blood smear, and had urgent splenectomy. At primary splenectomy, accessory spleens were identified and resected in 18% of patients. There was no operative mortality. Fifty-eight patients had an excellent response to splenectomy and their steroids were tapered off within 3 weeks. Thirteen patients had a poor response to primary splenectomy of whom eight remitted spontaneously and five required accessory splenectomy resulting in complete remission in three patients. Twenty-nine patients were considered nonresponders, 25 of whom had radionuclide scanning for accessory spleens. Seven of these patients had accessory spleens identified but only four consented to accessory splenectomy. In three of the four patients, a complete remission was achieved. Neither platelet antibody titers nor measurements of platelet survival or turnover predicted platelet response to splenectomy. However, immune thrombocytopenic purpura (ITP) in older patients was significantly less likely to respond to splenectomy. These data support continuing use of splenectomy in selected patients with ITP and an aggressive search for accessory spleens in patients who relapse since they are easily localized at operation by hand-held isotope detector probe.