To define better the risk of epipodophyllotoxin-related acute myeloid leukemia (AML) after extended follow-up and to assess responses to intensive salvage therapy, all patients who developed this complication after treatment for acute lymphoblastic leukemia (ALL) or non-Hodgkin lymphoma (NHL) in consecutive clinical trials at St Jude Children's Research Hospital from 1979 to 1994 were studied. Cases with 'lineage switch' or 'clonal selection' were excluded. Epipodophyllotoxin-related AML developed in 32 of 1140 patients treated for ALL and in three of 332 treated for NHL; it was a first adverse event in 25 and two cases, respectively. The complication was diagnosed at 12-130 months (median 34 months) after the initiation of treatment with epipodophyllotoxins; all but one of the cases occurred within 73 months, indicating that the risk is negligible after 6 years. The predominant karyotypic feature was 11q23 translocations (71% of cases); 21q22 rearrangements were rare. In a stepwise Cox regression analysis, two factors increased the risk of this complication: weekly or twice weekly administration of epipodophyllotoxins (P < 0.001); and the administration of asparaginase immediately before epipodophyllotoxin therapy (P < 0.001). Initial responses to salvage therapy were comparable to those reported for de novo AML: 92% of the evaluable patients entered complete remission after combination treatment. Single-agent therapy with 2-chlorodeoxyadenosine induced complete or partial remissions in one-half of the patients treated. The long-term survival rate was dismal. Of the 17 evaluable patients treated exclusively with chemotherapy, only one is alive at 84 months, compared to three of 16 patients who underwent bone marrow transplantation (alive at 10, 23 and 73 months). Cases of epipodophyllotoxin-related AML constitute a unique clinical syndrome that will require innovative strategies for cure.