As a prelude to a nationwide randomized trial of thioguanine (TG) versus mercaptopurine (MP) for childhood lymphoblastic leukaemia we compared a pilot group of 23 children taking TG with a matched group taking MP. We assessed drug tolerance based on haematological toxicity and measured erythrocyte (RBC) concentrations of thioguanine nucleotides (TGN). The median tolerated dose of TG was 30 mg/m2 compared to 55 mg/m2 for MP. There was no difference in the pattern of anaemia or neutropenia between the two groups, but dose-limiting thrombocytopenia was more evident in the TG children (P< 0.001), four of whom had a decrease in platelet count to <20 x 10(9)/l compared to only one on MP. The median RBC TGN concentration for those on 40 mg/m2 TG was 1726 pmol/8 x 10(8) RBCs compared with 308 pmol/8 x 10(8) RBCs for those on 75 mg/m2 MP (P< 0.0001). There was an inverse correlation between RBC TGNs and neutrophil count in the MP group but not in those on TG. No correlation between metabolite concentration and thrombocytopenia was found in either group. These results provide further evidence that TG has a selective effect on platelets. They also showed that RBC TGN were, on average, 5-fold higher in those taking TG but did not obviously relate to myelotoxicity as found in children on MP. The higher concentrations seen may partly reflect the erythrocyte's ability to metabolize TG directly to TGN by pathways not open to MP.