To the Editor:
Agnogenic myeloid metaplasia (AMM) is a clonal myeloproliferative disorder characterized by progressive bone marrow fibrosis and extramedullary hematopoiesis.1 Because of the lack of curative options, treatment is mostly palliative and is essentially aimed at improving severe cytopenia and at relieving symptomatic organomegaly.2 In a substantial number of patients this can only be achieved with splenectomy.3 It was therefore with the greatest interest that we read the report by Barosi et al4 reporting the association of splenectomy with a higher than expected incidence of acute leukemic transformation in patients with AMM. Although the observations and the conclusions reported in the study are interesting, we believe that alternative explanations should also be considered. This has important clinical implications, because the hypothesis proposed by Barosi et al,4 by potentially discouraging the use of splenectomy, may in fact be detrimental for the optimal management of some patients with AMM. Indeed, the hypothesis that splenectomy directly affects the rate or the frequency of leukemic transformation in AMM remains speculative and, in our view, should not influence at this time the physician’s decision to consider splenectomy in selected cases. Drawing from other case-control studies of secondary malignancies after splenectomy in Hodgkin’s disease5 and aplastic anemia,6 Barosi et al4 imply that the effect of splenectomy may result from a disruption of the host-tumor relationship. However, convincing experimental or clinicopathological evidence is lacking. Furthermore, despite a sophisticated statistical analysis aimed at retrospectively adjusting for selection bias, risk-time dependency, and bias related to splenectomy indications, the confounding effect of many of these factors cannot be entirely eliminated. Specifically, the possibility that the need for splenectomy identifies a group of AMM patients who are undergoing subclinical blastic transformation, either in the bone marrow or in sites of extramedullary hematopoiesis (EMH), was not considered by Barosi et al,4 and potentially informative data on the histopathology of bone marrow and spleen in these patients were not provided.
We recently reported preliminary clinicopathological observations in a small number of patients with primary myelofibrosis and prominent splenic EMH, correlating the histopathological findings in the spleen with clinical course and survival.7 We believe that our data not only are of general relevance to the issue raised by Barosi et al,4 but also may provide an alternative explanation for the high incidence of leukemic transformation after splenectomy. We studied surgically removed spleens from 8 patients with advanced stage AMM. For all patients, a bone marrow biopsy was available demonstrating no evidence of acute leukemia around the time (±2 weeks) of splenectomy. Splenectomy was performed in an attempt to ameliorate severe pancytopenia (n = 2), to palliate severe abdominal pain (n = 2), or for both reasons (n = 4). Median time from diagnosis to splenectomy was 4.5 years (range, 2 to 7 years). Seven of eight patients developed overt acute myeloid leukemia (>20% circulating blasts or >30% marrow blasts) within 1 to 10 months (median, 4 months) from the time of splenectomy, and 1 patient developed refractory anemia with excess of blasts in transformation (RAEB-t) 4 months after splenectomy. Median survival from the time of blastic transformation was 13 months (range, 8 to 25 months). At the time of last follow-up, 4 patients were dead of disease at a median of 9 months (range, 8 to 15 months) from transformation. In all cases, histological examination of the spleen sections showed trilineage hematopoiesis with a significant predominance of myeloid cells, clusters of immature mononuclear cells, and a large number of cells staining for myeloperoxidase (MPO) and lysozime (LY). As controls, we examined the spleens of 5 patients with early stage AMM who underwent splenectomy for reasons unrelated to their primary hematological condition (eg, motor vehicle accidents). The median time from diagnosis to splenectomy for these patients was 12 months (range, 3 to 14 months). Control spleens showed a predominance of erythroid cells; less numerous MPO+, LY+ cells; and the absence of clustering of immature mononuclear cells. When the proliferation rate of splenic hematopoietic cells was examined by staining for proliferating cell nuclear antigen (PCNA), significant differences were found between the splenectomy specimens of early stage AMM and late stage AMM (32.3% v 59.7%, respectively;P < .001). These data suggest that late stage AMM patients who require splenectomy for symptom control may be undergoing acute leukemic transformation in areas of splenic EMH, often without overt marrow transformation. This accelerated phase of extramedullary myeloid proliferation may be responsible for the symptom escalation that eventually leads to splenectomy and, ultimately, for the development of acute leukemic transformation. Splenectomy, therefore, may not promote blastic transformation, as suggested by Barosi et al,4 but the need for splenectomy may identify patients who are already developing it. Conversely, it is not surprising that the removal of spleens harboring EMH in blastic transformation does not prevent the development of acute leukemia, because it is likely that a significant number of circulating malignant myeloid progenitor cells remain unaffected by the splenectomy. Although based on very preliminary data, our model of a predominant extramedullary (splenic) blastic transformation in AMM, predating overt acute leukemia, provides a plausible explanation for the observations of Barosi et al4and deserves further investigation. A histopathological review of a large number of splenectomy specimens could show whether leukemic evolution occurs only or predominantly in patients who have areas of splenic myeloid metaplasia in transformation. Specifically, the relationship between the number of blasts or the expression of proliferative markers and the time to develop acute leukemia should be studied. Furthermore, sensitive cytogenetic and molecular data simultaneously obtained on the spleen, peripheral blood, and bone marrow could indicate if any other anatomic compartment is affected at the time of splenectomy.
We read with interest the comments of Porcu et al to our study about the influence splenectomy has on the development of blast transformation in MMM. The interest derives from the data they offered as a possible alternative explanation of our results. First, they documented a progressive shift of splenic metaplasia from mostly erythroid to mostly myeloid and from mostly mature to mostly immature cells during the progression of the disease. These data are consistent with the previously reported findings of a progressively increasing number of promegakaryocytes in the spleen as opposed to the bone marrow1-1 and support a model of the disease in which splenic hematopoiesis originates by displacement of the most immature precursor myeloid cells from the bone marrow. Second, they reported a high number of splenic mononuclear cells positive for lisozyme staining. This stands for the monocytic nature of the cells and confirms the hypothesis1-2 that in MMM, a progressively increasing population of monocyte-macrophages both produces fibrogenic growth factors (as transforming growth factor-β and bFGF), and participates in myeloid metaplasia under the regulation of macrophage colony-stimulating factor.1-3
However, these data disclose only small hints about the influence of splenectomy on blast transformation. Porcu et al suggested that the development of acute leukemia in patients who had been splenectomized could be coincidental to the propensity to develop acute leukemia in those patients for whom splenectomy was advocated. However, this was the fundamental issue of our report, ie, to distinguish the factors that induce blast transformation from confounding factors. For this purpose, we used two different and complementary approaches: multivariate analysis (Cox model) and propensity score. When corrected for other covariates, splenectomy remained an independent factor for blast transformation. The hypothesis that an occult splenic blast transformation could modify the picture of the disease to advocate splenectomy might justify our results only by assuming that our study did not retrieve all the clinical features leading to splenectomy. In particular, the propensity score we identified included age, appearance of anemia, and increase in spleen volume.
Histological examination of the removed spleens was performed in our collection of cases, and no accumulation of blasts in this organ was reported by our pathologists. However, our study did not specifically address the accumulation of blasts, and, in absence of any standardized method of research, it might have been overlooked.
Splenectomy increases blast transformation in MMM in every setting of care, as also supported by a recent report from the Mayo Clinic.1-4 Discerning the reasons of this phenomenon is, in our opinion, of primary importance for understanding the biology of the disease.