Granulocyte colony-stimulating factor and severe aplastic anemia

To the editor:

The report by Socié et al¹  on the association between development of myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML) and use of granulocyte colony-stimulating factor (G-CSF) in patients with severe aplastic anemia (AA) treated with immunosuppressive therapy (IST) is thought provoking, and highlights an important unresolved question about the optimal management of this disorder. The authors acknowledge that the retrospective, nonrandomized design of their study “precludes any definitive role, if any, of G-CSF in … the genesis of secondary malignancy.”¹ 

We agree with this caveat, and also hypothesize that the observed association may be the result of “reverse causality” or “confounding by indication.” As reported, G-CSF was used more often in younger patients and in those in whom IST was initiated within the first 2 months after diagnosis. Furthermore, relapse was more frequent, and recovery after relapse less frequent, among patients who received G-CSF. Patients who are younger and those who begin IST more promptly following diagnosis may be more ill, have lower absolute neutrophil counts, and/or have more infections (data related to these parameters were not provided). Since the initiation of G-CSF was at the discretion of the treating physician, it is plausible that its administration may have been a consequence of greater disease severity at presentation. If so, G-CSF might appear to be associated with the risk of MDS/AML because its use is associated with disease severity, which might be the true underlying risk factor for malignant transformation.

This hypothesis was raised previously by Young²  following publication of the study by Kojima et al³  of patients with AA/IST treated with or without G-CSF. That study found a statistically significant association between G-CSF and risk of MDS/AML. Interestingly, in an earlier study by the same group,⁴  11 of 50 children with AA treated with G-CSF plus cyclosporine A (CSA) developed MDS/AML, versus 0 of 32 children treated with G-CSF alone and 0 of 9 children treated with CSA alone. All of the leukemic events occurred in the doubly treated patients, which is also consistent with the hypothesis of confounding by indication, although a G-CSF/CSA interaction cannot be excluded. Similar findings were reported in a review of 48 adults with AA, in which MDS developed in 4 of 13 patients on G-CSF plus CSA and none of those treated with IST alone.⁵ 

Extensively treated patients with AA may have an intrinsically more severe hematopoietic stem cell defect and higher likelihood of developing MDS/AML, a model which may also explain the observed association between G-CSF and risk of MDS/AML in patients with severe congenital neutropenia.⁶  Patients with highly abnormal stem cells may be more likely to relapse, less likely to survive (with or without relapse) and, perhaps, more likely to develop MDS/AML.

As the authors conclude, only a clinical trial such as the ongoing European Group for Blood and Marrow Transplantation (EBMT) trial, which randomizes patients with AA to IST with or without G-CSF, can accurately quantify the risks and benefits of G-CSF in the clinical management of AA, and clarify which hypothesis best explains the statistical association between risk of MDS/AML and use of G-CSF.