The 1997 acute lymphoblastic leukaemia (ALL) trial (ALL97) was a randomised comparison of prednisolone versus dexamethasone and of 6-mercaptopurine versus 6-thioguanine. During the first 2 years of the trial, review of survival data showed the preceding trial, UKALL XI, was no better than its predecessor and that survival for childhood ALL in the UK had not improved in the fashion witnessed by other cooperative treatment groups. The therapy template was therefore altered to an American Children's Cancer Group (CCG) style regimen, including stratification by age, white cell count and early response to therapy by assessment of the bone marrow. This phase of the trial was designated ALL97/99. Comparison of the two phases showed that the event-free survival (EFS) for both ALL97 and ALL97/99 was better than previous UKALL trials, as was overall survival (OS) for ALL97/99. Both EFS and OS were significantly better in ALL97/99 than in ALL97 (at five years, 80.0% vs. 74.0%, P = 0.002; and 88.0% vs. 83.5%, P = 0.005, respectively). Isolated central nervous system (CNS) relapse for patients in ALL97/99 was half that in ALL97 (3.0% vs. 4.9%), P = 0.03) and the overall CNS relapse rate was halved in ALL97/99 (4.4% vs. 9.6%, P < 0.00005). There were no significant differences for non-CNS relapse, induction deaths or deaths in remission between the two phases of the trial.