Leukocytosis in polycythemia vera and splenomegaly in essential thrombocythemia are independent risk factors for hemorrhage

Background: Long‐term outcomes are favorable for patients with polycythemia vera (PV) and for patients with essential thrombocythemia (ET). However, hemorrhage is a significant cause of morbidity and mortality in those patients. Methods: We retrospectively recruited 247 patients who had received a diagnosis of PV (n = 101) or ET (n = 146) during the period 2001–2010. Results: After a median follow‐up period of 36.2 months, the cumulative incidence of hemorrhage was 39.6% in patients with PV (6.2% per person‐year) and 29.7% in patients with ET (5.9% person‐years). Episodes of major bleeding occurred in 9.9% of patients with PV and in 14.4% of patients with ET. Overall survival was significantly shorter among patients with hemorrhage than among those without said complication (P < 0.001 for overall patients; P = 0.002 for patients with PV; P = 0.026 for patients with ET). In the univariate analysis, age ≥60 yr (OR: 4.77, P = 0.046) and WBC ≥ 16 × 109/L (OR: 4.15, P = 0.010) were predictors of hemorrhage in patients with PV, and age ≥60 yr (OR: 3.25, P = 0.040), WBC ≥ 16 × 109/L (OR: 2.89, P = 0.024), albumin <4.0 g/dL (OR: 4.10, P = 0.002), and splenomegaly (OR: 5.19, P = 0.002) were predictors of hemorrhage in patients with ET. Multivariate analysis showed that WBC ≥ 16 × 109/L was the only significant risk factor for hemorrhage in patients with PV (OR: 3.51, P = 0.026) and that splenomegaly was the only risk factor for hemorrhage in patients with ET (OR: 3.00, P = 0.048). Conclusion: Leukocytosis in PV and splenomegaly in ET are independent risk factors for hemorrhage.

polycythemia vera; essential thrombocythemia; hemorrhage; risk factor

Polycythemia vera (PV) and essential thrombocythemia (ET) are myeloproliferative diseases characterized by the increased production of red blood cells in the former and the overproduction of platelets in the latter. Both diseases are associated with a high incidence of JAK2 V617F mutation and constitutive activation of tyrosine kinase and downstream pathways [1] . In addition to thrombosis, hemorrhagic events in the gastrointestinal tract, the central nervous system, and the mucocutaneous membrane are common complications of both diseases [2] , [3] . In several large series, the rate of bleeding episodes ranged from 3% to 8% for patients with PV [4] , [5] and from 3% to 18% for patients with ET [2] , [3] , [6] . The pathogenesis of hemorrhage may be associated with platelet dysfunction, acquired von Willebrand syndrome, and medications such as antiplatelet drugs or anticoagulants used for treatment of arterial or venous thrombosis [7] . Studies have shown that prior history of bleeding, leukocyte count >11 × 109/L, platelet count >1000 × 109/L, splenomegaly, and aspirin use are risk factors for bleeding in patients with ET [2] , [3] and that age, disease duration, and history of bleeding are hemorrhagic risk factors in patients with PV [5] . Nonetheless, few clinical studies on risk factors for hemorrhage among patients with PV or ET have been conducted. In addition, to the best of our knowledge, the effects of hemorrhage on the long‐term outcome of patients with PV and those with ET have not been well characterized. In this study, we evaluated the incidence and general characteristics of bleeding events and the risk factors associated with hemorrhage among patients with PV and those with ET at a medical center in Taiwan during the period 2001–2010.

Patients who were treated for PV or ET at the Taipei Veterans General Hospital during the period July 2001 to July 2010 were retrospectively recruited for analysis. The diagnoses of PV and ET were established after independent review by two hematologists according to the 2008 WHO criteria [1] . Patients with primary myelofibrosis, chronic myelogenous leukemia, acute myeloid leukemia, myelodysplastic syndrome, chronic neutrophilic leukemia, chronic eosinophilic leukemia, mast cell disease, unclassifiable myeloproliferative neoplasms (MPNs), chronic or juvenile myelomonocytic leukemia, and MPNs associated with platelet‐derived growth factor or fibroblast growth factor 1 rearrangements were excluded. Demographic data and clinical characteristics were acquired from chart review and physicians’ records. Detection of JAK2V617F mutation and determination of homozygosity were based on methods involving both allele‐specific PCR and PCR followed by restriction enzyme digestion. Pertinent clinical parameters recorded at initial diagnosis and at the time of hemorrhage were collected for analysis. The local institutional review board approved the study.

Hemorrhagic events were categorized into major bleeding and minor bleeding events. Major bleeding was defined as fatal hemorrhage and as hemorrhage requiring blood transfusions or hospitalization. Upper gastrointestinal (GI) bleeding was defined as symptoms of hematemesis or tarry stool with the diagnosis confirmed by panendoscopic examination of the upper GI tract.

Splenomegaly was diagnosed in patients with a splenic diameter >12 cm as measured on abdominal ultrasound.

Overall survival was determined from the date of diagnosis to the last contact or death. The survival status of all the patients was determined based on the record of medical chart, cancer registry system of our institute, and telephone interview provided that we were able to contact the patients or their family. The average annual incidence and the cumulative incidence of hemorrhage during the 10‐yr study period were also determined. Rates of survival and cumulative incidence were estimated using the Kaplan–Meier method, and the log‐rank test was used for comparison of survival and incidence curves. Binary logistic regression models were applied for univariate and multivariate analysis to determine the prognostic impact of clinicopathologic factors on hemorrhage. Risk factors for statistical significance (P value of <0.1) in the univariate analysis were then included in a multivariate model. Groups of nominal variables were compared using the chi‐square test or Fisher's exact test, as appropriate. A P value <0.05 was considered to indicate statistical significance; all multivariate analyses were two‐tailed. All statistical analyses were performed with the statistical package SPSS for Windows (version 17.00, SPSS, Chicago, IL, USA).

During the 10‐yr period, 247 patients were treated for PV (n = 101, 40.9%) or ET (n = 146, 59.1%). The patient characteristics are summarized in Table [NaN] . In our study, 11 (10.9%) patients with PV and 14 (9.6%) patients with ET were lost to follow‐up before July 2010. The median duration of follow‐up was 36.2 months with a total follow‐up of 878 person‐years. The median overall survival (OS) was 103 months (95% CI: 93–113 months) for patients with PV and not reached for ET (P = 0.534).

General characteristics of patients with polycythemia vera and essential thrombocythemia

1 CHF, congestive heart failure; LDH, lactate dehydrogenase; ET, essential thrombocythemia; Hb, hemoglobin; Hct, hematocrit; PV, polycythemia vera; WBC, white blood cell count.

Relative to patients with ET, patients with PV upon diagnosis had significantly higher WBC counts, higher hemoglobin levels, higher hematocrit (Hct) values, a higher incidence of JAK2 V617F mutation, a higher incidence of homozygous JAK2 V617F mutation, and a higher incidence of splenomegaly but lower platelet count levels and a lower frequency of anagrelide usage. JAK2 V617F mutations were detected in 51 of 57 patients with PV and 65 of 100 patients with ET that have been tested. Homozygosity of the JAK2 V617F mutations was present in 13 patients with PV and 3 patients with ET.

During the study period, there were 53 hemorrhagic events (at least one episode of major or minor bleeding), including 24 events in 18 patients with PV (17.8%) and 29 events in 27 patients with ET (18.6%) (Table [NaN] ). The cumulative overall incidence of hemorrhage at 10 yr was 39.7% (6.0% per person‐year). The incidence of hemorrhagic events was 39.6% (6.2% per person‐year) for patients with PV and 29.7% (5.9% person‐years) for patients with ET (Fig. [NaN] ). In addition, the cumulative overall incidence of major bleeding at 10 yr was 18.0% (4.4% per person‐year). The incidence of major bleeding was 13.0% (4.1% per person‐year) for patients with PV and 22.2% (4.7% per person‐year) for patients with ET (P = 0.177). The incidence of major bleeding was 6.2% at 1 yr and 9.2% at 3 yr for patients with PV and 9.2% at 1 yr and 13.4% at 3 yr for patients with ET (P = 0.177).

Characteristics of hemorrhagic and thrombotic complications in 247 patients with PV and ET

• 2 ET, essential thrombocythemia; UGI, upper gastrointestinal bleeding; ICH, intracranial hemorrhage; IHD, ischemic heart disease; IS, ischemic stroke; PV, polycythemia vera; SDH, subdural hemorrhage.
• 3 Number of patients with major or minor bleeding.
• 4 One patients had lower GI bleeding, another 1 patient had hemorrhagic pancreatitis.

Five (4.9%) patients with PV and two (1.4%) patients with ET experienced recurrent hemorrhage. There were no significant differences in major or minor bleeding events between patients with PV and those with ET. The most common complication of major hemorrhage was UGI bleeding followed by intracranial hemorrhage (ICH) and subdural hemorrhage (SDH). The most common complication of minor hemorrhage in both groups was upper GI bleeding followed by hematoma and hemarthrosis. Hemorrhage developed in 15.6% of patients on aspirin and 21.9% of patients without aspirin use (P = 0.208). In the subgroup of patients with platelet count >1000 × 109/L, hemorrhage occurred in 19.2% of patients on aspirin and in 31.0% of patients without aspirin use (P = 0.316).

The OS rate was significantly lower among patients with hemorrhage than among patients without hemorrhage (P < 0.001 overall; P = 0.002 for PV; P = 0.026 for ET; Figs [NaN] , [NaN] , [NaN] ). Furthermore, there was no significant correlation between the occurrence of hemorrhage and thrombosis (P = 0.335).

The main characteristics of patients with recurrent hemorrhage are shown in Table [NaN] . All of the 7 patients were aged >60 yr (median age, 84 yr), and all of the 4 patients who were tested for JAK2 V617F mutations tested positive. The time from diagnosis to hemorrhage ranged from 1.1 to 75.4 months. In addition, at the time of hemorrhage, at least one of the components of the complete blood cell count, namely white cell count (>16 × 109/L), hemoglobin level (>18 g/dL), and platelet count (>800 × 109/L), was poorly controlled despite administration of medications. Six of seven patients had hemorrhagic events in both the central nervous system and GI tract on different occasions. Patient 1 had spontaneous SDH and ICH, and the hematomas were removed by craniotomy. Only patient 1 was treated surgically for his hemorrhage. The other patients received medical treatment and supportive care. Patient 2 developed spontaneous ICH in the left lentiform nucleus, most likely because of underlying hypertension. Patient 3 presented with hematoma in the subdural space within the spinal canal from the level of the 12th thoracic to the first lumbar vertebral body after receiving vertebroplasty for treatment of a compression fracture.

Features of recurrent hemorrhage in seven patients

5 ET, essential thrombocythemia; F, female; Hb, hemoglobin; HTN, hypertension; ICH, intracranial hemorrhage; IHD, ischemic heart disease; IS, ischemic stroke; M, male; NA, not available; OS, overall survival; PV, polycythemia vera; PE, pulmonary embolism; Plt, platelet count; SDH, subdural hemorrhage; UGI, upper gastrointestinal bleeding; WBC, white blood cell count.

Table [NaN] summarizes the results of the univariate analysis of risk factors for hemorrhage.

Univariate analysis of factors associated with hemorrhage in 247 patients

• 6 Alb, albumin; CHF, congestive heart failure; CI, confidence interval; DM, diabetes mellitus; ET, essential thrombocythemia; Hct, hematocrit; NA, not applicable; OR, odds ratio; Plt, platelet; PV, polycythemia vera; WBC, white blood cell count.
• 7 P < 0.05.

Overall, age ≥60 yr, WBC count ≥16 × 109/L, albumin <4.0 g/dL, and the presence of splenomegaly were significant predictors of hemorrhage in the univariate analysis but only WBC count ≥16 × 109/L (OR: 2.25, 95% CI: 1.06–4.79, P = 0.035) remained an independent risk factor in the multivariate analysis. Different cutoffs of WBC count were examined as risk factor for hemorrhage in the range of 10–20 × 109/L in univariate and multivariate analysis. Despite the trend exists, only when the WBC level of 16 × 109/L was used as the cutoff did the risk of hemorrhage reach greatest statistical significance. In addition, albumin <4.0 g/dL and splenomegaly failed to reveal statistical significance despite the existence of a trend (for albumin <4.0 g/dL, OR: 2.07, 95% CI: 0.98–4.38, P = 0.058; for splenomegaly, OR: 2.13, 95% CI: 0.99–4.55, P = 0.052).

Among patients with PV, age ≥60 yr and WBC count ≥16 × 109/L were significant predictors in the univariate analysis. Two PV patients (13.3%) with homozygous JAK2 V617F mutation and 4 patients (8.3%) without the homozygosity experienced hemorrhage (P = 0.622). The presence of homozygous JAK2 V617F mutation did not confer an increased risk for hemorrhage (OR: 0.92, 95% CI: 0.09–9.69, P = 0.942). Only WBC count ≥16 × 109/L (OR: 3.51, 95% CI: 1.16–10.58, P = 0.026) remained a significant predictor in the multivariate analysis.

Among patients with ET, age ≥60 yr, WBC count ≥16 × 109/L, albumin <4.0 g/dL, the presence of splenomegaly, and the use of hydroxyurea were significant risk factors for hemorrhage in the univariate analysis. The presence of splenomegaly was the only risk factor for statistical significance in the multivariate analysis (OR: 3.00, 95% CI: 1.01–8.91, P = 0.048). None of the patients with ET who were homozygous for JAK2 V6117F mutations experienced hemorrhage. Patients with JAK2 V617F mutations were not at higher risk of bleeding than patients without said mutations (Table [NaN] ).

We found that leukocytosis in patients with PV and splenomegaly in patients with ET were independent predictors of hemorrhage. In a recent study of 565 patients with ET [3] , splenomegaly, platelet count >1000 × 109/L, and leukocyte count >11 × 109/L were reported to be significant prognosticators of hemorrhagic events. In an analysis of 1104 patients with ET or morphologically confirmed prefibrotic myelofibrosis, Finazzi et al.[2] found that early prefibrotic myelofibrosis, leukocyte count >11 × 109/L, previous bleeding, and aspirin usage were significant risk factors for major bleeding. It is unclear whether leukocytosis contributes to the pathogenesis of hemorrhage in patients with myeloproliferative neoplasms or whether leukocytosis is simply associated with disease severity. Consistent with findings from previous studies, we found that leukocytosis was an independent risk factor for bleeding overall as well as specifically for patients with PV and patients with ET. Leukocytes may directly participate in the interaction between platelet and vascular endothelium [8] , [9] . Evangelista et al.[10] found that neutrophils from MPNs expressed tissue factor when contacting platelets expressing P‐selectin, resulting in increased platelet–leukocyte aggregation and surface expression of adhesion molecules on leukocytes. Cross‐talk between leukocytes and platelets has also been shown to foster thromboinflammation of vessels and contribute to vascular injury or immune‐mediated vasculitis [11] . Moreover, hyperleukocytosis in patients with acute leukemia often manifests as intracranial hemorrhage or pulmonary infiltrates related to leukostasis. Hyperleukocytosis also results in the release of cytokines and chemokines, and the activation of proteases, such as matrix metalloproteinases or elastases, suggesting that hemorrhage is mediated, at least in part, by injury to the endothelium, tissue destruction, and systemic inflammatory reactions [12] . In our study, we examined a series of WBC counts to determine the cutoff value for risk stratification and found that there was a correlation between hemorrhage and leukocytosis when the WBC count was greater than 16 × 109/L. This cutoff is higher than the cutoff value of >11 × 109/L reported in other studies [2] , [3] . Whether the difference is related to race, management, or inherited bleeding tendency is unknown. Further evaluations are needed to clarify the most appropriate cutoff value for WBC count that can be used to classify patients as being at high risk for bleeding.

The incidence of JAK2 V617F mutation and allele burden in our patients and their irrelevance to hemorrhage were consistent with findings reported in previous studies [3] , [13] , [14] . It has been reported that high WBC count is associated with high JAK2 V617F allele burden in patients with PV, a group of patients who are at risk for clonal myeloproliferation and who are at risk for developing postpolycythemic myelofibrosis [15] . However, in our study, high WBC count was an independent risk factor for hemorrhage and was not associated with JAK2V617F mutation status. The JAK2 V617F mutation is much more useful as a diagnostic tool than as a prognostic or predictive factor [16] .

We found that splenomegaly was an independent risk factor for hemorrhage in patients with ET. The reason for this might be because multiple comorbid factors of splenomegaly, such as thrombocytopenia, portal hypertension, hypersplenism, esophageal and gastric varices, and coagulopathy secondary to depletion of coagulation factors produced by liver induce the generation of the hemorrhagic process, particularly in the GI tract. We also found that lower GI bleeding and hemorrhagic pancreatitis in our study could present as rare complications of hemorrhage. In addition to splenomegaly, the high prevalence of hepatitis B and C virus infections as well as hepatitis‐induced liver cirrhosis in the Taiwanese population may further potentiate the risk of bleeding diathesis. Splenomegaly therefore not only results in the sequestration of blood cells including platelets, but also results in increased extramedullary hematopoiesis, thereby increasing the risk for hemorrhagic events.

Acquired von Willebrand's syndrome is a well‐documented complication of marked thrombocytosis (platelet count greater than 1500 × 109/L), especially in patients who use aspirin [17] . In our study, the risk of bleeding among patients with a platelet count >1000 × 109/L was marginally raised only in patients with ET (OR: 2.12, 95% CI: 0.89–5.04, P = 0.09). Nonetheless, neither overall patients nor the subgroup of patients with a platelet count >1000 × 109/L who used aspirin were at increased risk for bleeding events (P = 0.316), which might be attributed to judicious use of low‐dose aspirin and deliberate avoidance of using potent antiplatelet medications. Platelet dysfunction in MPN may predispose patients to bleeding. The proposed mechanisms of platelet dysfunction include impaired receptor‐mediated granule secretion, content and expression of glycoprotein IIb/IIIa complex [18] , and deficient α‐adrenergic receptors in platelets, thereby leading to defective functional response to epinephrine [19] . Moreover, thrombocytosis alone could be a predisposing factor of hemorrhage. Acquired type 2 von Willebrand syndrome might be caused by increased proteolysis of von Willebrand factor (vWF) multimers by ADAMTS13 cleaving protease, resulting in an inverse relationship between platelet count and large vWF multimers [20] . However, the efficacy of low‐dose aspirin (100 mg) has been shown in the European Collaboration on Low‐Dose Aspirin in Polycythemia Vera (ECLAP) study to reduce the risk of death due to cardiovascular causes, nonfatal myocardial infarction, nonfatal stroke, pulmonary embolism, and major venous thrombosis without increasing the risk of major bleeding [21] . The lack of correlation between hemorrhage and thrombosis in our analysis also suggested that additional factors other than the antithrombotic medications alone result in increased risk of bleeding in these patients. Taken together, lowering both the WBC and platelet counts and judicious administration of low‐dose aspirin should prevent complications of hemorrhage in patients with ET.

Risk factors for major hemorrhage include age, extremely high platelet count, history of bleeding, and disease duration [7] , [5] . Consistent with previous studies, we found that age >60 yr was associated with up to a fourfold increase in the risk of bleeding. In addition, all of the seven patients with recurrent hemorrhage were older than 60 yr of age and all had poorly controlled blood cell counts. Intriguingly, 2 of 3 of the deaths occurred 90 months after the hemorrhagic events and four patients were still alive at the most recent follow‐up, implying that complications of hemorrhage did not lead to death immediately. Nonetheless, compromised OS was observed consistently in all patients in our study (Figs [NaN] , [NaN] , [NaN] ), suggesting that hemorrhage was still associated with negative long‐term outcome. This might be the result of complex factors including impaired quality of life, neurologic, cognitive or motor dysfunction, lower tolerance to diagnostic or invasive procedures, or inadequate dosing of medications that are intended to prevent recurrent hemorrhage or other complications.

Our analysis shares the common limitations of retrospective studies and unavoidable bias may exist, such as patient characteristics, duration of treatments, drug–drug interactions, and chronic comorbidities. Early/prefibrotic PMF can be misdiagnosed as ET, and as pointed out by Finazzi et al. [2] , it is also true that PMF confers greater risk of hemorrhagic complications and might lead to compromised survival after the complications. After reviewing the 27 patients both with ET and hemorrhagic complications, we did suspect one patient to be prefibrotic PMF at initial diagnosis. None of the other 26 patients had evidence of myelofibrosis either at diagnosis or during long period of follow‐up. Some patients with prefibrotic PMF initially diagnosed of ET had been revised of their diagnosis later during follow‐up and excluded in our study. Five patients with ET in our study had the high upper range of WBC count or LDH value. Most of these unusual laboratory data apparently came from transient reaction to acute inflammation and returned to a lower baseline level during follow‐up. According to our analysis, only high WBC count but not high level of LDH (not shown in table) was associated with increased risk of bleeding, suggesting that the association between WBC count and hemorrhage was not due to misdiagnosis of prefibrotic PMF as ET. Last but not least, the results of our study provide valuable data on risk stratification and management of hemorrhage in elderly patients with myeloproliferative neoplasms.

In conclusion, although long‐term survival can be achieved in patients with myeloproliferative neoplasms such as polycythemia vera and essential thrombocythemia, the likelihood of hemorrhage is remarkably high. High WBC count and splenomegaly are risk factors for hemorrhage. The risk of thrombosis and hemorrhage should be balanced by cautiously adjusting the medications for patients who are at high risk of bleeding to avoid unnecessary administration of aspirin. The mechanisms underlying these clinical observations warrant further investigations.

This work was supported in part by grants from the Taipei Veterans General Hospital (V100C‐161 and V102A‐001) and the Taiwan Clinical Oncology Research Foundation.

None declared.

Graph: Cumulative incidence of hemorrhage in polycythemia vera ( PV ) and essential thrombocythemia ( ET ). The cumulative incidence at 10 yr was 39.6% for patients with PV and 29.7% for patients with ET. There is no significant difference in incidence between the two groups.

Graph: Overall survival of the 247 patients with and without hemorrhage. Patients with hemorrhage had significantly worse overall survival ( OS ) compared to patients without hemorrhage (median OS : 94.8 months vs. not reached).

Graph: Overall survival of 101 patients with polycythemia vera ( PV ) and hemorrhage. PV patients with hemorrhage had significantly worse overall survival ( OS ) compared to patients without hemorrhage (median OS : 94.8 months vs. not reached).

Graph: Overall survival of 146 patients with essential thrombocythemia ( ET ) and hemorrhage. ET patients with hemorrhage had significantly worse overall survival ( OS ) compared to patients without hemorrhage (median OS : not reached for both).