To evaluate the risk of second primary myeloid leukaemia due to radiotherapy and chemotherapy administered for a first primary cancer, we conducted a population-based case-control study consisting of 97 cases and 194 controls matched on age, date of diagnosis, and site of initial primary cancer among residents of 13 counties in western Washington State. The risk of myeloid leukaemia in patients who received cyclophosphamide as part of their chemotherapy regimen was 7.4 (95% confidence interval 1.3-43.8). This risk was not altered appreciably by the administration of radiotherapy. Compared to patients not receiving any chemotherapy, the relative risk among patients who received prednisone in combination with cyclophosphamide (odds ratio 44.4 95% confidence interval 4.0-496.2) was nearly four times that among patients receiving cyclophosphamide without this steroid (odds ratio 12.6 95% confidence interval 2.4-64.9). The relative risk of second primary myeloid leukaemia in patients who received both prednisone and drugs other than cyclophosphamide (odds ratio 64.2 95% confidence interval 2.6-1582) was 20 times that among patients receiving drugs other than cyclophosphamide and no prednisone (odds ratio 3.2 95% confidence interval 0.6-16.9). These risk estimates were higher when the analysis was restricted to acute myeloid leukaemia. There was no increased risk of second primary myeloid leukaemia associated with radiotherapy. The single unique finding is that the use of prednisone in chemotherapy regimens may enhance the leukaemogenic effect of other chemotherapy drugs.