Fax: (507) 266-4972
Risk factors for leukemic transformation in patients with primary myelofibrosis
Article first published online: 10 APR 2008
Copyright © 2008 American Cancer Society
Volume 112, Issue 12, pages 2726–2732, 15 June 2008
How to Cite
Huang, J., Li, C.-Y., Mesa, R. A., Wu, W., Hanson, C. A., Pardanani, A. and Tefferi, A. (2008), Risk factors for leukemic transformation in patients with primary myelofibrosis. Cancer, 112: 2726–2732. doi: 10.1002/cncr.23505
- Issue published online: 4 JUN 2008
- Article first published online: 10 APR 2008
- Manuscript Accepted: 14 JAN 2008
- Manuscript Revised: 9 JAN 2008
- Manuscript Received: 19 DEC 2007
- myeloproliferative disorder;
- primary myelofibrosis;
Previous prognostic studies in primary myelofibrosis have focused on risk factors for overall survival and have resulted in the establishment of several prognostic scoring systems. However, to the authors' knowledge, information regarding risk factors for leukemic transformation in primary myelofibrosis is limited.
The current retrospective study examined clinical variables at the time of diagnosis and specific treatment modalities for their effect on leukemic transformation in 311 patients with primary myelofibrosis who were seen at the Mayo Clinic.
Univariate analysis of parameters at the time of diagnosis revealed a significant association between inferior leukemia-free survival and a peripheral blood blast percentage ≥3 (P < .0001), a platelet count <100 × 109/L (P = .004), a monocyte count ≥1 × 109/L (P = .02), the presence of hypercatabolic symptoms (P = .03), a low hemoglobin level (P = .04), and a high leukocyte count (P = .04). The first 2 parameters were found to maintain their statistical significance during multivariate analysis. Neither leukemia-free nor overall survival was found to be affected by the presence of <3% peripheral blood blasts or JAK2V617F mutation. The evaluation of treatment effect on leukemic transformation unexpectedly revealed a significant and independent association with previous therapy with either erythropoiesis‒stimulating agents (P = .004) or danazol (P = .007), even when the aforementioned prognostic indicators at the time of diagnosis were added as covariates to the multivariate model. In contrast, leukemia-free survival was not found to be affected by a treatment history with hydroxyurea, thalidomide, or other drugs.
A peripheral blood blast percentage ≥3 and/or a platelet count <100 × 109/L at the time of diagnosis were found to be strong and independent predictors of leukemic transformation in patients with primary myelofibrosis. The unexpected association between leukemic transformation and a history of treatment with erythropoiesis‒stimulating agents or danazol requires validation by prospective studies. Cancer 2008. © 2008 American Cancer Society.
Previous prognostic studies in patients with primary myelofibrosis focused on risk factors for overall survival and have resulted in the establishment of several prognostic scoring systems1-5 based on adverse features deduced from multivariate analysis: a hemoglobin level <10 g/dL,1–15 a leukocyte count <4 or >30 × 109/L,1, 4, 6–8, 11 circulating blasts either ≥1% or ≥3%,1–3, 5 the presence of constitutional symptoms,2, 5 a platelet count <100 × 109/L,4, 6, 7, 11, 13 a monocyte count ≥1 × 109/L,6 circulating immature granulocytes ≥10%,8 advanced age,1-3, 7, 9–11, 14 male sex,1, 3 and cytogenetic abnormalities either in the bone marrow4, 6, 7, 16, 17 or splenic tissue.18 The first 4 variables from the above‒cited list were used to construct what to our knowledge are the 2 most widely used prognostic scoring systems for survival in primary myelofibrosis: the Dupriez et al. (which uses hemoglobin level and leukocyte count)1 and Cervantes et al. (which uses hemoglobin level, circulating blasts, and constitutional symptoms)5 prognostic scoring systems. However, recent studies from the Mayo Clinic have underscored the additional negative prognostic relevance of both a platelet count <100 × 109/L and a monocyte count ≥1 × 109/L19, 20 and the incorporation of these 2 variables into the prognostic scoring system of Dupriez et al.1 has been shown to result in a new “Mayo” prognostic scoring system (which uses hemoglobin level, leukocyte count, platelet count, and monocyte count) that identifies patients who are most suitable for “watchful waiting” (those at low risk), conventional or experimental drug therapy (those at intermediate risk), or allogeneic stem cell transplantation (those at high risk).4, 19, 20
In contrast to the above-mentioned activity in prognostic investigations regarding survival in patients with primary myelofibrosis, to our knowledge the association between clinical and laboratory features at the time of the initial diagnosis and the risk of leukemic transformation (also known as blast phase primary myelofibrosis)21 has not been well studied. In what to our knowledge is 1 of the very few studies to examine that particular issue, 195 patients with primary myelofibrosis were followed for a median of 36 months1; 22 patients developed blast‒phase primary myelofibrosis at a median of 36 months (range, 12-136 months), with a 5‒year risk of death from leukemic transformation of 16% reported.1 In that particular study, univariate analysis identified a leukocyte count >30 × 109/L and abnormal karyotype as risk factors for death from leukemic transformation.1 Because we believe current therapy for blast‒phase primary myelofibrosis is utterly inadequate and does not alter the associated dismal prognosis (median survival of approximately 3 months),22, 23 it is important to accurately identify high-risk patients and consider experimental drug therapy or allogeneic stem cell transplantation while they are still in the chronic phase of their disease. Furthermore, because of a previously reported association between therapeutic splenectomy in patients with primary myelofibrosis and leukemic transformation,24 it is equally important, in both high-risk and low-risk patients, to identify and avoid the use of treatment modalities that might increase the risk of leukemic transformation. Both these issues, risk factors at diagnosis and the contribution of specific therapies to the development of blast‒phase primary myelofibrosis, were examined in the current retrospective study of 311 patients from a single institution.
MATERIALS AND METHODS
Approval was obtained from the Mayo Clinic Institutional Review Board before reviewing the medical records of patients with the diagnosis of primary myelofibrosis during the period from 1976 through 2006. Clinical and laboratory data, including bone marrow histology, were rereviewed and the diagnosis of primary myelofibrosis was confirmed according to the World Health Organization criteria.25 Leukemic transformation was considered in the presence of blast‒phase primary myelofibrosis according to criteria by the International Working Group for Myelofibrosis Research and Treatment: the presence of either 20% blasts in the bone marrow or 30% in the peripheral blood.21
All study patients were required to have documentation of a complete blood count, peripheral blood blast percentage, and bone marrow examination obtained at or within 6 months of diagnosis and before primary myelofibrosis-specific therapy. All bone marrow examinations were either performed or the slides reviewed at our institution. Details regarding the clinical and laboratory presentation at or within 6 months of diagnosis and all primary myelofibrosis-specific therapeutic interventions were abstracted from the medical records. Hypercatabolic symptoms were defined by the presence of ≥1 of the following: unexplained fever (objectively documented by the patient to be >37.8°C at least on 2 occasions), excessive sweats (requiring a change of clothing) persisting for >1 month, and weight loss >10% of the baseline value in the year preceding the initial diagnosis of primary myelofibrosis.5JAK2V617F mutational status was established by either allele-specific polymerase chain reaction or sequencing. Follow-up information on each study patient was obtained by telephone calls or letters sent to patients and/or their physicians.
Statistical analysis was performed using StatView software (SAS Institute Inc, Cary, NC). P values were 2-tailed and statistical significance was set at P < .05. All clinical and laboratory parameters that were evaluated for their effect on leukemia-free survival were those obtained at the time or within 6 months of the initial diagnosis. Continuous variables were summarized as medians and ranges. Categoric variables were described as count and relative frequency (%). The Mann-Whitney U or Kruskal-Wallis tests were used to compare continuous variables between categories. Comparison between categoric variables was performed by chi-square statistics. Variables that were identified as being significant on univariate analysis were subjected to multivariate analysis. Cox regression analysis was used to determine the impact of various clinical and laboratory variables and therapeutic interventions on leukemia-free survival. Kaplan-Meier survival curves were plotted for groups stratified by the presence or absence of independent risk factors for leukemic transformation and compared using the log-rank test. All survival curves were coded from the time of the initial diagnosis and not from the time of referral to the Mayo Clinic. Similarly, the term “treatment exposure” encompasses the entire period between the initial diagnosis and leukemic transformation or, in the absence of leukemic transformation, to last follow-up or death.
A consecutive series of 311 patients with primary myelofibrosis were studied. Their presenting clinical and laboratory features are outlined in Table 1. The median age at diagnosis was 57 years and 41% of the patients were women. At the time of diagnosis, 30% of the patients had severe or moderate anemia (a hemoglobin level <10 g/dL), 20% had leukopenia or marked leukocytosis (a leukocyte count <4 or >30 × 109/L, respectively), 18% had severe or moderate thrombocytopenia (a platelet count <100 × 109/L), 12% had absolute monocytosis (a monocyte count ≥1 × 109/L), 28% had hypercatabolic symptoms, 35% had a peripheral blood blast percentage ≥1, and 9% had a peripheral blood blast percentage ≥3 (Table 1). The number of patients with high‒risk, intermediate‒risk, or low‒risk disease according to Dupriez et al.1 was 30 (10%), 94 (30%), and 187 (60%), respectively. Among 182 and 139 evaluable patients, 74 (41%) and 80 (58%) displayed cytogenetic abnormalities and JAK2V617F mutation, respectively. Twenty-seven cases (9%) of leukemic transformation were documented at a median time from the primary myelofibrosis diagnosis to leukemic transformation of 26 months (range, 0.8-266 months).
|Variables||No. of evaluable patients||Values||95% CI||Univariate P||Multivariate P|
|No. of females (%)||311||126 (40.5%)||.59|
|No. of patients age <60 y (%)||311||177 (56.9%)||.74|
|Median age at diagnosis (range) y||311||57.3 (15.3–87.6)||.26|
|No. of patients with JAK2V617F mutation||139||80 (57.6%)||.98|
|No. of patients with hemoglobin <10 g/dL||311||94 (30.2%)||.25|
|Median hemoglobin (range), g/dL||311||11.1 (4.5–16.1)||.04||NS|
|No. of patients with leukocyte count <4 or >30 × 109/L||311||62 (19.9%)||.27|
|Median leukocyte count × 109/L (range)||311||8.3 (1.46–156.7)||.04||NS|
|No. of patients with platelet count <100 × 109/L (%)||311||56 (18.0%)||1.48–8.11||.004||.02|
|Median platelet count × 109/L (range)||311||266 (12–1916)||.17|
|No. of patients with monocyte count ≥1 × 109/L||306||36 (11.8%)||1.18–8.73||.02||NS|
|Median monocyte count × 109/L (range)||306||0.342 (0–5.88)||.03||NS|
|Percent peripheral blood blasts (median and range)||311||0 (0–10.5)||<.0001||<.0001*|
|No. of patients with peripheral blood blasts >0% and ≤1%||311||59 (19.0%)||.93|
|No. of patients with peripheral blood blasts >1% and <3%||311||35 (11.3%)||.49|
|No. of patients with peripheral blood blasts ≥1%||311||110 (35.4%)||1.38–6.57||.006||.02†|
|No. of patients with peripheral blood blasts ≥3% (%)||311||28 (9.0%)||2.86–17.15||<.0001||.0002‡|
|Dupriez prognostic scoring systems: low/intermediate/high||311||187/94/30||.29|
|Mayo prognostic scoring systems: low/intermediate/high||307||151/93/63||0.08–0.88||.02||NS|
|Cervantes prognostic scoring systems: low/high||311||235/76||0.13–0.67||.004||NS|
|No. of patients with hypercatabolic symptoms||311||88 (28.3%)||1.09–5.52||.03||NS|
|No. of patients with cytogenetic abnormalities||182||74 (40.7%)||.09|
Effect of Clinical and Laboratory Variables at the Time of Diagnosis on Leukemia-free Survival
On univariate analysis, leukemia-free survival was found to be negatively affected by the presence of peripheral blood blasts considered as either a continuous variable (P < .0001) or as a categoric variable with cutoff values of both ≥3 (P < .0001) and ≥1 (P = .006), a platelet count <100 × 109/L (P = .004), lower hemoglobin level (P = .04), higher leukocyte count (P = .04), higher monocyte count considered as either a continuous variable (P = .03) or a categoric variable with the cutoff value of ≥1 × 109/L (P = .02), the presence of hypercatabolic symptoms (P = .03), and a higher prognostic score according to either the Mayo19, 20 (P = .02) or Cervantes et al.5 (P = .004) prognostic scoring system (Table 1). Leukemia-free survival was not found to be affected by a peripheral blood blast percentage of between 1 and 3 (P = .49). On multivariate analysis, significance was sustained only for a peripheral blood blast percentage of ≥3 (P = .0002) and a platelet count of <100 × 109/L (P = .02) with hazards ratios (HR) of 5.8 and 2.8 and 95% confidence intervals (95% CI) of 2.3 to 14.6 and 1.2 to 6.6, respectively. The presence of JAK2V617F mutation did not appear to affect leukemia-free survival (P = .98). As indicated earlier, a peripheral blood blast percentage ≥3% was found to have the strongest prognostic significance for leukemic transformation compared with a peripheral blood blast percentage >0% and ≤1% or >1% and <3% (Table 2). The 3% cutoff value was also found to be the most relevant in terms of predicting inferior overall survival (Fig. 1).
|Blast count (%)||No. of patients (% of total)||No. of patients with Leukemic transformation||HR* (95% CI)||P|
|0||190 (61%)||13 (6.8%)||0.13 (0.05–0.35)||<.0001|
|>0 and ≤1||59 (19%)||6 (10.2%)||0.19 (0.06–0.59)||.004|
|>1 and <3||35 (11%)||1 (2.9%)||0.10 (0.01–0.82)||.03|
|≥3||27 (9%)||7 (25.9%)|
Effect of Specific Treatment on Leukemia-free Survival
On univariate analysis, leukemia-free survival was found to be significantly inferior in patients with a history of splenectomy (P = .01) or treatment with erythropoiesis‒stimulating agents including erythropoietin‒α or darbepoetin‒α (P = .004), danazol (P = .007), or androgens (P = .03) (Table 3). Leukemia-free survival did not appear to be affected by a history of treatment with hydroxyurea (P = .17), interferon‒α (P = .95), thalidomide (P = .26), lenalidomide (P = .89), or corticosteroids (P = .07) (Table 3). On multivariate analysis, statistical significance was sustained for both erythropoiesis‒stimulating agents (P = .005; HR of 3.1 [95% CI, 1.4-6.8]) and danazol (P = 0.01; HR of 3.4 [95% CI, 1.3-8.5]), even when the aforementioned prognostic indicators at the time of diagnosis were added as covariates to the multivariable model.
|Treatment||No. of evaluable patients||Values||95% CI||Univariate P||Multivariate P|
|Erythropoiesis- stimulating agent||311||96 (30.9%)||1.44–6.89||.004||.005|
|Interferon-α (SC)||311||22 (7.1%)||.26|
Combined Value of Risk Factors at Diagnosis for Leukemia-free Survival in Patients With Primary Myelofibrosis
Using the 2 aforementioned independent prognostic factors for leukemia-free survival, obtained at the time of diagnosis (ie, the presence of peripheral blood blasts ≥3% and a platelet count <100 × 109/L), patients can be categorized into 3 groups with 0, 1, or 2 risk factors for leukemia-free survival. In the current series of 311 patients, 240 patients (77.2%) presented with none of the 2 risk factors whereas 60 patients (19.3%) and 11 patients (3.5%) patients presented with either 1 or both risk factors, respectively. The corresponding frequencies of leukemic transformation were 5.8%, 18.3%, and 18.2% respectively (P = .003). The likelihood of leukemic transformation in the absence of both risk factors was significantly less than that noted in the presence of either 1 (P < .0001; HR of 0.18 [95% CI, 0.08-0.41]) or both (P = 0.003; HR of 0.10 [95% CI, 0.02-0.46]) risk factors. Accordingly, one can construct leukemia-free survival curves for patients with no risk factors (ie, those at low risk) versus those with either 1 or 2 risk factors (ie, those at high risk). In addition to a significant difference in overall leukemia-free survival (Fig. 2), the 2 groups (low risk vs high risk) displayed significantly different incidences of leukemic transformation at 1 year (1% vs 10%; P = .002), 3 years (5% vs 35%; P < .0001), and 5 years (18% vs 65%; P < .0001) from diagnosis, when considering only the at-risk population at these different time points (180 patients vs 50 patients at 1 year, 111 patients vs 29 patients at 3 years, and 78 patients vs 20 patients at 5 years).
To our knowledge, the current study represents the first systematic investigation of prognostic factors for leukemic transformation in patients with primary myelofibrosis and identifies peripheral blood blast blasts ≥3% and a platelet count <100 × 109/L at the time of diagnosis as independent risk factors for the development of blast‒phase primary myelofibrosis. The study also provides risk-adapted estimations of leukemic transformation risk at 1 year, 3 years, and 5 years after the initial diagnosis. Such information complements survival-based risk stratification in identifying patients who warrant special consideration for close monitoring, stem cell transplantation, or experimental drug therapy.
It is to be noted that both a platelet count <100 × 109/L and the presence of circulating blasts are also independent predictors of inferior survival in patients with primary myelofibrosis; the former is incorporated in the Mayo prognostic scoring system19, 20 and the latter in the Cervantes prognostic scoring systems.5 However, the cutoff blast percentage level used in the Cervantes prognostic scoring system is ≥1%, whereas the optimal cutoff value in the current study was ≥3% in terms of predicting both leukemia-free and overall survival. In this regard, the findings of the current study are consistent with those of a large Japanese study of 336 patients, in which the presence of ≥3% circulating blasts was found to be an independent predictor of inferior survival.3
As illustrated in Figure 2, patients with either 1 of the 2 aforementioned risk factors for leukemic transformation have a significantly worse outcome but the additional risk conferred by the presence of 2, as opposed to 1, risk factors could not be accurately assessed because of the small number of patients with both risk factors (n = 11). Sample size and heterogeneity in cytogenetic abnormalities might also underlie the lack of significance attached to abnormal karyotype.1 It should be remembered that the prognostic significance, for survival, of karyotype in patients with primary myelofibrosis might be aligned with certain specific lesions rather than the mere presence or absence of cytogenetic abnormalities.16, 17 Conversely, in the current study, there was not even a hint of an association between the presence of JAK2V617F mutation and leukemic transformation, which is consistent with our previous studies in that regard26, 27 and different from the observations made by others.28
The most surprising finding in the current study was the unexpected association between leukemic transformation and a history of treatment with either an erythropoiesis‒stimulating agent or danazol, which was not accounted for by the presence of anemia at the time of diagnosis or the aforementioned risk factors for either inferior survival or leukemic transformation. Because our retrospective study was not designed to monitor the development of anemia or red blood cell transfusion dependency after initial diagnosis, it is possible that a history of treatment with an erythropoiesis‒stimulating agent or danazol was a surrogate for a biologically more aggressive disease. However, an association between other drugs used for anemia (eg, thalidomide) and leukemic transformation was not evident and univariate analysis-based associations between leukemic transformation and splenectomy or androgen therapy were not sustained during multivariate analysis.
The association between leukemic transformation in patients with primary myelofibrosis and the use of erythropoiesis‒stimulating agents was particularly intriguing considering the finding that erythropoiesis‒stimulating agents have recently been implicated in tumor progression and inferior survival in certain solid tumors including head and neck,29, 30 metastatic breast,31, 32 and nonsmall cell lung33 cancers. The putative mechanism of detrimental effect, in this context, is tumor and endothelial cell expression of erythropoietin receptors with the respective augmentation of neoplastic growth and tumor-associated angiogenesis. Such a scenario is also possible in primary myelofibrosis because myeloid progenitors readily express erythropoietin receptor and bone marrow angiogenesis is considered to be an adverse prognostic factor in primary myelofibrosis.34 Regardless, the observation from the current retrospective study must be validated by properly designed prospective studies before its application to current management.
- 10Chronic idiopathic myelofibrosis: prognostic impact of myelofibrosis and clinical parameters on event-free survival in 122 patients who presented in prefibrotic and fibrotic stages. A retrospective study identifying subgroups of different prognoses by using the RECPAM method. Ann Hematol. 2003; 82: 605–611., , , et al.
- 21Primary myelofibrosis (PMF), post polycythemia vera myelofibrosis (post-PV MF), post essential thrombocythemia myelofibrosis (post-ET MF), blast phase PMF (PMF-BP): consensus on terminology by the international working group for myelofibrosis research and treatment (IWG-MRT). Leuk Res. 2007; 31: 737–740., , , et al.
- 25Chronic idiopathic myelofibrosis. In: JaffeES,HarrisNL,SteinH,VardimanJW, eds. World Health Organization classification of tumors: tumours of the haematopoietic and lymphoid tissues. Lyon, France: International Agency for Research on Cancer (IARC) Press; 2001: 35–38., , , , , .