The effect of 3 anthracene derivatives, mitoxantrone, ametantrone, bisantrene, on 4 normal human bone marrows, was studied using the myeloid stem cell assay developed by Pike and Robinson, in order to define to what extent this test could be used to predict the relative clinical hematologic toxicity of new anticancer agents. For the 3 drugs, an exponential relationship between colony survival and drug concentration was found, but was much steeper for mitoxantrone (slope = -195.2 +/- 8.8/micrograms/ml) than for ametantrone (slope = 5.1 +/- 1.0/micrograms/ml, p less than or equal to 0.001) and bisantrene (slope = 7.1 +/- 0.3/micrograms/ml, p less than or equal to 0.001). The difference of slope between ametantrone and bisantrene was of borderline significance (p less than or equal to 0.05). The ratios of concentrations inducing a 50% growth inhibition for mitoxantrone versus bisantrene and for ametantrone versus bisantrene were close to the corresponding ratios of concentrations inducing a 90% growth inhibition. The relative in vitro toxicities reproduce very well the relative myelosuppression observed in clinical trials with mitoxantrone versus bisantrene but the results were less satisfactory for the comparison of these 2 agents with ametantrone. In addition, our data suggest that, for these 3 compounds, intrinsic myeloid progenitor sensitivity is a major determinant of leukopenia.