Recent reports cite better survival when repeatedly rejecting renal allografts are removed and patients returned to hemodialysis. However, the criteria for graft removal remain undefined; although some reports recommend removing all kidneys undergoing a third rejection. In our series (1968-1973) of 316 patients with technically successful first grafts followed 2 1/2-8 years, graft survival was inversely related to the number of rejection episodes. One hundred per cent of kidneys without rejection are currently functioning or functioned at the time of death compared to 90% with one rejection, 67.4% with two and 21% with three. However, 40% of kidneys having three rejection episodes functioned longer than one year after treatment of the third rejection episode. In an attempt to determine the predictability of one year graft survival or failure following treatment of the third rejection, a formula was developed that correctly predicted in 33 of 38 (87%) patients. The formula was based on information available prior to treatment of the third rejection episode, and represents an index of baseline renal function (serum creatinine after second rejection episode) and two indices of the severity of rejection episodes (serum creatinine change between the first and second rejection episodes; rapidity of sequential rejection). Following its derivation, the formula was applied to a second group (1974) of 19 patients having had three rejection episodes. The formula correctly predicted one year allograft survival or failure following treatment of the third rejection episode in 68% of these patients. A striking finding of our review was a significant difference in current patient survival between those having no rejection episodes (89%) and those having one or more rejection episodes (65%) (p <.00001). There was no significantly greater long-term curtailment in survival if more than one rejection episode was treated. Patients having one rejection episode seemed to die from varying causes and at varying time periods. Patients dying after two or more rejection episodes had an increased incidence of deaths due to bacterial infection.