Two hundred and fifty patients with acute myeloid leukaemia (AML) were randomized between 2 regimens of chemotherapy: TRAP and BARTS III. Overall, patients randomized to TRAP, which was the more intensive of the 2 regimens, fared slightly better (P = 0.06) than those on BARTS III. However, the improvement in survival associated with more intensive chemotherapy was substantial only for patients who had favourable prognostic features at presentation, such as a normal total leucocyte count, or absence of palpable liver, or, especially, age under 40. Indeed, for patients under 40, those allocated to the more intensive regimen (TRAP) lived considerably longer than those allocated to BARTS III (P less than 0.002) while for patients over 40 there was no material difference in survival between patients on the 2 protocols. It thus appears that intensive chemotherapy is likely to be more effective when favourable prognostic features are recorded.