We evaluated 17 clinically used anticancer drugs for their effects on total WBC and absolute neutrophil counts in BDF1 mice. These drugs were selected from three categories, based on their myelosuppressive effects in man: myelosuppression is dose-limiting; myelosuppression occurs but is not dose-limiting; or myelosuppression does not occur. The ability of each drug to cause myelosuppression in mice was determined by its effect on the neutrophil count 4 days after single-dose treatment. The neutropenic effect of the maximum tolerated dose (LD0-20) of each drug was characterized as marked (greater than 65% decreases), moderate (35%-65% decreases), or minimal (less than 35% decreases) to correspond with the three clinical categories of myelosuppression. The neutropenic effects in mice correctly predicted (true + or true -) the myelosuppressive effects in man for 13 of the 17 drugs (76%). Marcellomycin and the platinum complexes cisplatin, carboplatin, and spiroplatin did not cause neutropenia at maximum tolerated doses. These represent false-negative predictions, since the drugs are myelosuppressive in man. The results with the platinum complexes indicate that this method is not suitable as a means of screening these agents for myelosuppression. Excluding the platinum complexes, the predictions were correct for 12 of 13 drugs (92%). Therefore, this model is considered a good predictor for the myelosuppressive effects of anticancer drugs in man (except platinum complexes) and can be used effectively to screen drugs for this toxicity. However, it is important that drugs identified by this system as being less myelosuppressive than the reference agent(s) be evaluated further, since all the incorrect predictions were false negatives.