Review of data from 79 children with the haemolytic uraemic syndrome (HUS) showed that the polymorphonuclear leucocyte (PMN) count at presentation in childhood HUS predicts outcome. Logistic regression analysis of several features at presentation identified only the PMN count and the presence of a diarrhoeal prodrome as having a significant effect on the outcome (P less than 0.01 and P less than 0.001 respectively). The geometric mean PMN count was significantly raised in 70 children who had typical HUS following a diarrhoeal prodrome (D+ cases) compared with that of 9 children who had atypical disease without diarrhoea (D- cases) (t-test on log-transformed data, P less than 0.005). Fifty-seven children with D+ HUS who recovered completely had a significantly lower geometric mean PMN count than D+ cases with a bad outcome (P less than 0.001). Four of these patients, who died in the acute stage of the disease, had a significantly higher mean count than the rest of the D+ patients (P less than 0.001). Multiple regression analysis demonstrated that the PMN count in D+ cases was not significantly influenced by haemoglobin concentration, platelet count, length of the prodrome, or the administration of antibiotics in the prodromal period. A high PMN count at presentation in D+ HUS indicates a poor prognosis. The data emphasise the heterogeneity of HUS and suggest that PMN participate in the pathogenesis of the disorder in typical D+ cases but not in atypical D- cases.