OBJECTIVE--To investigate the incidence and aetiology of secondary leukaemia after childhood cancer in Britain. DESIGN--Cohort study and a case-control study. SETTING--Britain and population based National Register of Childhood Tumours. SUBJECTS--Cohort of 16,422 one year survivors of childhood cancer diagnosed in Britain between 1962 and 1983, among whom 22 secondary leukaemias were observed. A case-control study of 26 secondary leukaemias observed among survivors of childhood cancer diagnosed in Britain between 1940 and 1983; 96 controls were selected matched for sex, type of first cancer, age at first cancer, and interval to diagnosis of secondary leukaemia. MAIN OUTCOME MEASURES--Dose of radiation averaged over patients' active bone marrow and total accumulated dose of epipodophyllotoxins, alkylating agents, vinca alkaloids, antimetabolites, and antibiotics (mg/m2) given for the original cancer. RESULTS--Cumulative risk of secondary leukaemia within the cohort did not exceed 0.5% over the initial five years beyond one year survival, except that after non-Hodgkin's lymphomas 1.4% of patients developed secondary leukaemia. Corresponding figure for patients treated for non-Hodgkin's lymphomas in the early 1980s was 4%. The relative risk of secondary leukaemia increased significantly with exposure to epipodophyllotoxins and dose of radiation averaged over patients' active bone marrow. Ten patients developed leukaemia after having an epipodophyllotoxin-teniposide in nine cases, etoposide in one. Chromosomal translocations involving 11q23 were observed relating to two secondary leukaemias from a total of six for which there were successful cytogenetic studies after administration of an epipodophyllotoxin. CONCLUSIONS--Epipodophyllotoxins acting alone or together with alkylating agents or radiation seem to be involved in secondary leukaemia after childhood cancer.