Three hundred and seventy-one children below 16 years, with newly diagnosed acute myeloid leukaemia, were included in six consecutive GATLA/GLATHEM protocols, from November 1967 to December 1987. The study was divided in three periods: 1967 to 1975, 1976 to 1982, and 1983 to 1987. Three induction schedules were used during the first two periods, and different maintenance schemes alternating with monthly consolidations were explored; the value of immunotherapy with C. Parvum and androgen therapy with stanozolol was also tested. Protocol 3-AML-83, representing the third period, included a four-week induction phase with vincristine, adriamycin, cytosine-arabinoside, prednisone and 6-mercaptopurine, followed by a consolidation phase with cyclophosphamide, cytosine-arabinoside and 6-mercaptopurine for four weeks. Maintenance phase included daily, oral 6-mercaptopurine, and monthly cytosine-arabinoside, both during two years, and adriamycin every eighth week, for one year. Complete remission rates for the first two periods of therapy were 40% and 55%, whereas that of the last period was 74%. The overall results of the period 1967-1982, showed actuarial duration rates of complete remission, event-free survival and survival, at 60 months, between 2% and 6%, their median duration being of 9, 8 and 10 months respectively. No significant difference was observed between the first two periods or protocols. Protocol 3-AML-83, activated in March 1983, achieved actuarial rates of continuous complete remission, event-free survival, and survival of 51%, 37% and 39% respectively, at 48 months. The difference between the first two periods and the last one was highly significant (P less than 0.0001).(ABSTRACT TRUNCATED AT 250 WORDS)