The UK MRC ALL97 trial included a randomised comparison of the efficacy and toxicity of dexamethasone and thioguanine (TG, experimental arms) with prednisolone and mercaptopurine (MP). The cytotoxic effects of MP and TG are mediated, in part, by thioguanine nucleotide (TGN) metabolites. TGNs are formed directly from TG whereas MP produces intermediate metabolites which are substrates for the enzyme thiopurine methyltransferase (TPMT). TPMT is under the control of a common genetic polymorphism. Both drugs are good substrates for TPMT. Both drugs undergo oxidative metabolism, but TG requires prior deamination. Consecutive children with ALL, diagnosed in the United Kingdom and Ireland between April 1997 and June 2002, were randomised to receive either TG or MP in maintenance courses. All patients received TG in intensification courses. Toxicity data was collected by an adverse event reporting system with follow-up questionnaires to seek detailed information on specific toxicities. Red blood cell (RBC) TGN concentrations and TPMT activities were measured by standard techniques in specific sub-groups. Of 1498 randomised patients, 750 were allocated TG and 748 MP. The trial produced a 5 year event-free survival (EFS) of 80%, with no difference in EFS or overall survival between thiopurine arms. However, TG was associated with an acute hepatitis syndrome with features of veno-occlusive disease (VOD) in 95 patients. Eighty-two of these were in the randomised TG arm (11% of TG recipients), the remaining were patients allocated MP who developed VOD during TG intensification courses. On long term follow-up, 43 (5.7%) TG recipients developed persistent splenomegaly and mild to moderate thrombocytopenia due to portal hypertension. TPMT activities were measured in 73 of the 95 VOD children at a standard protocol dose of thiopurine. For comparison, TPMT activities were measured in children who did not develop VOD or splenomegaly; there was no difference in the range of TPMT activities measured at 40mg/m2 TG (n = 83, median 15.4 units/ml RBCs) compared to 75mg/m2 MP (n = 78, median 15.2 units). Median TPMT activities in these 161 children, taking thiopurines at the protocol standard dose, were 15.2 units (range 5.8 to 23) compared to a median of 13.4 units (range 5.8 to 23) in the 73 VOD children (median difference 1.8 units, 95% CI 0.9 to 2.7, p < 0.001). A similar difference in TPMT activity was also seen when comparing persistent splenomegaly cases with controls. There was no difference in accumulation of TGNs at comparable doses. When used for maintenance therapy, TG causes liver toxicity in a significant minority of children with ALL without improving long term survival. An association with polymorphic variations in TPMT levels, a critical enzyme in thiopurine metabolism, implicates inter-individual variations in metabolism of the drug as a mechanism for liver damage, and points to a possible means of identifying at-risk patients. MP should remain the standard thiopurine for maintenance therapy unless longer follow-up reveals a survival advantage in favour of TG, and the mechanisms of liver toxicity are understood to avoid exposure of at risk patients.