Kojima and colleagues (page 786) report a correlation between G-CSF treatment of children with aplastic anemia and evolution to cytogenetic abnormalities and myelodysplasia. How strong is this relationship and how generalizable the conclusions?

The Japanese investigators administered high G-CSF doses (close to the adult equivalent of 10 μg/kg/d, later thrice weekly) for very long periods (in all patients for at least 3 months and in nonresponders beyond, in some cases for years) and largely independent of blood count values. This is far from the usual practice. G-CSF is not recommended as first therapy in marrow failure; therapeutic trials aim at elevating the neutrophil counts; and in those few patients who respond, the dose is adjusted to maintain granulocytes above a safe level. The Japanese protocol likely was based on earlier favorable results from Europe, where G-CSF combined with immunosuppression in producing excellent survival in aplastic anemia patients, but neither later analysis of these data nor a formal randomized comparison could confirm the specific value of the cytokine, even in protecting patients from serious infection.

The results of Kojima et al must be balanced by other evidence. G-CSF has passed long-term testing in many animal species. Its standard use in cyclic neutropenia has been free of complications. In aplastic anemia, G-CSF in the European trials has not been a risk factor for evolution, and the overall rates of evolution to myelodysplasia there and in our NIH experience, where G-CSF is used sparingly, are comparable to those of Kojima et al. Most important: the protocol used by Kojima et al dictated prolonged G-CSF exposure for nonresponders, and poor clinical status is the major confounder of their study, because evolution to monosomy 7 most frequently is seen after failure of immunosuppressive therapy.

Chronic G-CSF therapy also has been associated with monosomy 7 in children with Kostmann syndrome. In both constitutional neutropenia and aplastic anemia, improved survival secondary to better supportive care may allow manifestation of a latent malignant potential. That small numbers of abnormal cells may reside quietly, even innocuously, in failed bone marrows has become apparent only with the development of increasingly sensitive techniques, as flow cytometry for paroxysmal nocturnal hemoglobinuria and fluorescence in situ hybridization for aneuploidy. Whether and how G-CSF might promote expansion of stem cells lacking chromosome 7 is amenable to laboratory experimentation; clarification of this relationship will have serious clinical implications for marrow failure patients.