On consultation, percutaneous umbilical blood sampling (PUBS) was offered to women with immune thrombocytopenic purpura (ITP) to determine the mode and site of delivery prior to labour. Between January 1989 and December 1993, 41 pregnant women underwent PUBS. All women had a history of ITP, a platelet count less than 90 K, (+) antiplatelet antibody, and/or thrombocytopenia diagnosed early in pregnancy. PUBS was performed at term except in one patient with preterm labour, who underwent PUBS at 31 weeks' gestation. Patients with fetal platelet counts greater than 50 K returned to their referring physician for delivery. Records were complete in 39 pregnancies. Fetal blood was successfully obtained in 37 of 39 cases (95 per cent). Fetal platelet counts correlated with neonatal platelet counts in 36 of 37 cases (97 per cent). The interval between PUBS and delivery ranged from 0 to 31 days. Six of 37 fetuses (16 per cent) had significant fetal thrombocytopenia (< 50 K). These six patients underwent Caesarean section. Vaginal delivery was recommended in all others. There were two procedure-related complications. There were no cases of intraventricular haemorrhage in any of the neonates. In conclusion, there is a high incidence of fetal thrombocytopenia in women with ITP. PUBS reliably detects fetal thrombocytopenia and is therefore useful in the perinatal planning of the mode and site of delivery.