Is Hydroxyurea Leukemogenic in Essential Thrombocythemia?

To the Editor:

Hydroxyurea (HU) is a structurally simple antimetabolite that interferes with DNA synthesis by inhibiting ribonucleotide reductase. Although not directly genotoxic, HU may impair the repair of damaged DNA, raising a legitimate concern regarding mutagenecity. Accordingly, Sterkers et al1 reported their experience in 357 patients with essential thrombocythemia (ET) and concluded that treatment with HU may induce acute leukemia characterized by a particular cytogenetic abnormality (17p−). While the study provides valuable clinical information derived from an impressively large collection of patients, several points deserve commentary.

In the particular study, conventional criteria were used to diagnose ET. However, baseline bone marrow and cytogenetic studies were not performed in an unspecified number of patients (the investigators should provide the information). It appears that these studies were performed only when clinical or laboratory features suggested myeloid metaplasia, chronic myelogenous leukemia (CML), or a myelodysplastic syndrome (MDS). However, these measures are not adequate to prevent the inadvertent inclusion of atypical cases of the latter disorders in the study population.2 Biologic heterogeneity in ET is further exemplified by recent studies demonstrating the presence of bcr-abl transcripts3 and polyclonal hematopoiesis4 in a significant number of patients with ET. Therefore, conclusions regarding leukemic risk in ET are more accurate when baseline bone marrow examinations with cytogenetic studies are available in all patients. It might even be argued that molecular studies may be needed in future studies to identify karyotypically occult cases of CML.

The study population included patients who are either untreated (31 patients), or received therapy with single agent pipobroman (12 patients), busulfan (35 patients), HU (201 patients), ³²P (29 patients), or HU and other agents (50 patients). The corresponding leukemic incidences were 0%, 0%, 3%, 3.5%, 7%, and 14%, respectively. It is imperative to note that the study was not randomized and was systematically biased into grouping patients on the basis of treatment requirement. Furthermore, allocation to the different treatment groups was sometimes dictated by age and/or disease refractoriness, both of which could influence leukemic risk. In other words, the patients in the different treatment groups are not biologically comparable. In addition, the number of events occurring in each treatment group are too small to allow statistically valid comparisons. Therefore, it is impossible to draw accurate conclusions regarding drug leukemogenecity from the present study.

Because of the increasing use of HU in nonhematologic disorders,5 any implication regarding drug leukemogenecity should be accurate and based on sound evidence. Data from retrospective cohort studies are often misleading because of the lack of appropriate comparison groups. In general, published reports on the association of HU and acute leukemia have been inconsistent and the strength of the association has been relatively small. In our experience, we did not encounter a single case of acute leukemia in either a population-based study of 39 cases6 or a retrospective study of 74 young women with ET observed for a median of greater than 5 years.7 In both studies, approximately half of the patients were treated with HU. The current availability of nonmutagenic platelet-lowering agents should allow examination of the issue in a prospective randomized setting.8 

To the Editor:

We agree with Dr Tefferi that one should be very cautious before drawing conclusions about leukemogenicity of HU from a retrospective analysis in which treatment was not randomized. A major objective of our work1-1 was to try to understand the relatively high number of cases of progression to MDS and AML with the 17p deletion we had recently observed at our institution in ET and also (unpublished results) in polycythemia vera (PV). These cases appeared to be more frequent after HU or pipobroman than after classical alkylating agents such as busulfan, although the causality between the former drugs and MDS and AML with this 17p⁻ syndrome cannot be determined.

In this study, we also found that 4.5% of all ET patients, including 3.5% of the 201 treated with HU alone progressed to MDS or AML. Only 195 of the patients had been karyotyped at diagnosis, and bone marrow core biopsy is not a usual procedure at our institution when ET is suspected and no clinical or biological features suggesting agnogenic myeloid metaphase (AMM) are present. However, it is improbable that the 17 patients who progressed to MDS or AML, especially the 7 who had received HU alone, initially had another disorder than ET. Indeed, had those cases been CML with thrombocythemic onset, t(9;22) would have been present at leukemic progression. Furthermore, none of the progressions was preceded by a phase of prominent myeloid metaplasia, which would have been observed had the initial disease been incipient AMM. Also of note is that MDS with thrombocythemia mainly include acquired sideroblastic anemia and the 5q⁻ syndrome, which, in our experience, are always complicated by anemia long before a possible (and rare) progression to AML is observed and cannot be mistaken for ET for a long period of time.³

In addition, the incidence of AML and MDS we observed after HU alone (3.5% of 201 cases) was almost similar to that observed in the literature (3.4% of 293 cases).1-2 Although it is true that treatment was not randomized in our study, patients who received HU alone did not seem to represent a population with particularly active disease, which could have been associated with a higher risk of evolution to AML. Indeed, our practice has been to treat most ET (<10% remained untreated in the present series) and, over the last 12 years, first-line treatment has almost always been HU. Only patients who did not achieve permanent normalization of platelet counts with HU (ie, presumably patients with more active disease) received additional drugs.

We believe that our findings confirm other reports that treatment with HU alone is associated with a certain risk of leukemic progression in ET and PV. The absence of progression reported by Tefferi with HU alone in two series rests on a relatively small number of cases and relatively small follow-up. The association between HU and leukemic progression in ET, especially with 17p deletion, does not prove that HU is directly a causative agent. Progression to AML, particularly with 17p deletion, is part of the natural evolution of another myeloproliferative disorder, ie, CML. Therefore, HU may not be leukemogenic in diseases other than MPD. On the other hand, caution over its use and close follow-up of patients treated for nonneoplastic disorders will certainly be required.