BACKGROUND 9-Methoxypyrazoloacridine (PZA) is an anticancer agent that shows selectivity of action for carcinomas over leukemias. It also has nearly equal potency against cycling and quiescent or hypoxic and normoxic target cells. Phase I trials of PZA in humans are nearing completion. OBJECTIVE This study was conducted to determine (a) if PZA is directly inhibitory to hematopoietic cells and, if it is, to characterize the inhibition pharmacodynamically, (b) whether species-specific differences in direct toxicity could explain differences in myelosuppression in mice, dogs, and humans, and (c) whether in vitro data correlate with in vivo myelosuppression data. METHODS In vitro clonogenic assays of hematopoietic progenitors of myeloid and erythroid lineages from human, canine, and murine femoral marrow were used to measure the direct toxicity of PZA. Results from these assays were compared on an area-under-the-curve (AUC) basis to clinical myelosuppression data. RESULTS On the basis of maximum tolerated concentrations, canine hematopoietic progenitors are most susceptible to PZA, followed by human and then murine progenitors. We found no difference in susceptibility to PZA toxicity between the human progenitors of myeloid and erythroid lineages. Both concentration and duration of exposure contribute to the in vitro toxicity of PZA. In contrast to antimetabolites, the in vitro toxicity of PZA could be minimized at a given AUC by lowering drug concentration and prolonging the period of exposure. On an AUC basis, the in vitro data are consistent with limited in vivo myelosuppression data from preclinical models and correlate with neutropenia data from a phase I trial. CONCLUSIONS PZA directly inhibits hematopoietic progenitors, an action that is responsible for the myelosuppression observed in humans. Human marrow appears able to compensate for the loss of up to 35% of its myeloid progenitors, in that peripheral neutrophil counts remain unchanged at that level of loss. Although in vivo studies show that prolonged infusion reduces myelosuppression at a given total dose in both rodent and canine models, pharmacokinetic differences make it unlikely that this approach will benefit human patients. CONCLUSIONS The in vitro data quantitatively predict the AUCs at maximum tolerated dose in preclinical models and human patients. Thus, in vitro clonogenic assays of myelotoxic agents can provide data that make both preclinical toxicology testing and clinical trial planning and interpretation more efficient and accurate.