Incidence, characteristics and risk factors of thromboembolic events in East Asian patients with BCR-ABL1 negative myeloproliferative neoplasms

The vascular complications have been a major cause of morbidity and mortality among all subtypes of BCR-ABL1 negative myeloproliferative neoplasms (MPN), but the ethnicity-specific data was limited. We therefore conducted a multi-center retrospective, longitudinal cohort study to evaluate the incidence, characteristics and risk factors of thromboembolic events of MPN patients. Of 256 patients, 27.3% experienced thromboembolic events, majority of which occurred before or within 12 months of MPN diagnosis. The multivariable Cox proportional analyses identified leukocytosis (HR 2.67, 95% CI 1.36–5.24, q = 0.004) and history of thrombosis (HR 9.68, 95% CI 2.00–46.88, q = 0.005) as the risk factors for thromboembolism. In subgroup analysis of polycythemia vera and hemoglobin concentration (HR 1.97, 95% CI 1.28–3.04, q = 0.002) appeared to be a significant risk factor of thrombosis, along with age and thrombosis history. In essential thrombocythemia, severity of the established IPSET score was closely correlated with the frequency of thromboembolic events. In primary myelofibrosis, history of thrombosis was associated with thrombosis events (HR 13.85, 95% CI 1.2–159.5, q = 0.035). Overall survival was worse in patients who experienced thromboembolic events. Our study highlighted the importance of recognizing high risk patients and implementing personalized intervention.

Definitions. Cardiovascular disease (CVD) risk factors were obesity with BMI over 25, smoking, hypertension, diabetes, and dyslipidemia. Thrombosis was categorized into arterial or venous thromboembolisms. Arterial thrombosis included acute coronary syndrome (ACS), stroke and peripheral artery disease (PAD). Venous thromboembolism included deep vein thrombosis (DVT), pulmonary embolism (PE) and splanchnic vein thrombosis (SVT). Other minor occlusive events, including stable angina and superficial thrombophlebitis were not included.
The variables clinically relevant to thrombosis were defined per NCCN Guidelines® 17 . Clinically relevant risk factors included age at diagnosis, sex, previous thrombotic event, hepatomegaly, splenomegaly, WBC > 15 × 10 9 /L, hemoglobin count, platelet count, JAK2 V617F mutation status, and two or more of the cardiovascular risk factors mentioned above.
The presence of thromboembolic event was confirmed with an imaging modality, and the time of event was recorded as the date of imaging, when thromboembolic event was subjectively confirmed.
Cytogenetic studies were performed onsite, whose satisfactory performance was monitored by a national external quality assurance scheme. Bone marrow cells were cultured for 24 h then karyotype was analyzed using the standard G-banding technique. The karyotypes were constructed and chromosomal abnormalities were reported in accordance with the 2016 International System for Human Cytogenetic Nomenclature. Statistical analysis. The primary objective of this study was to investigate the incidence of thromboembolic events in homogeneous East Asian MPN patients per MPN subtype. MPN was divided into PV, ET, PMF and MPN-unclassifiable (MPN-U). The secondary objectives included the characteristics of thromboembolism, the effect of thromboembolism on survival, and indirect comparison with other ethnicities.
The incidence of a thromboembolic event was calculated at any time (within 12 months or at the time of MPN diagnosis, and during the follow-up). Categorical variables were summarized with the frequencies in number and rates in percentages. Continuous variables were represented with the median values and ranges. Differences were assessed using Mann-Whitney test for continuous variables and Pearson's χ2 or Fisher's exact test for categorical variables. The multiple imputation approach was applied to the missing values of BMI in 82 patients. Of multiple sets of data imputed, the set that had lowest standard error was chosen. Then obesity and the number of cardiovascular risk factors were recalculated upon the newly imputed data. Statistical analyses with the multiple regression model were reconstructed from the newly imputed data. While all patients had information on JAK2V617F mutations, 196 patients lacked MPL data and 192 patients lacked CALR data. Because more than one third of the patients lacked these data, MPL and CALR variables were not considered in this study. Multivariable Cox proportional-hazard models were constructed to find risk factors for thromboembolic events and correlation among covariates. False Discovery Rate approach with Benjamini-Hochberg procedures was used, and q-value < 0.05 was considered statistically significant. Since cardiovascular risk factors have been well defined in previous study 5 , instead of examining each of the cardiovascular risk factors as separate variables in the regression model, 'two or more cardiovascular risks' was considered to be an important risk factor. Since some variables are well-known risk factors with established threshold for the thrombotic events, they were dichotomized for multiple regression model. For example, age over 60 years old is a well-known risk factor for PV 2 . BMI was specifically defined as categorical variable because obesity criteria for Korean population was defined as BMI over 25. Leukocytosis is another widely studied risk factor. ECLAP study had shown the association between leukocytosis over 15 × 10 9 /L with the increase of venous thrombosis in PV patients 18 . Thus, these variables were converted into categorical variables. Other continuous variables such as hemoglobin and platelet counts were handled as quantitative values in the multiple regression model. www.nature.com/scientificreports/ Overall survival (OS) was defined as the time from MPN diagnosis to death of any cause. The OS curves were estimated using the Kaplan-Meier method. If patients survived without death, the survival was censored at the latest date of follow-up when no death was confirmed. p-values of < 0.05 were considered statistically significant for Kaplan-Meier curves. These data were analyzed using the Statistical Package for the Social Sciences software (IBM SPSS Statistics, Version 22.0, New York, NY, USA).
Cumulative incidence of thromboembolic events of the MPN disease and subgroups was estimated taking into account the first event and considering death as a competitor and compared by Gray's test. For this part of the analyses, the statistical software R (www.r-proje ct. org) was used.
The risk stratification of PV and ET is based on the age (older than 60 years) and history of previous thrombosis 19,20 . When PV patients were classified accordingly, thrombosis occurred in 31.7% of high-risk group, in contrast to 3.2% in low-risk group, as shown in Fig. 2a. In the subgroup analysis of PV patients, the history of thrombosis (HR 9.68, 95% CI 2.00-46.88, p = 0.005), and hemoglobin count (HR 1.97, 95% CI 1.28-3.04, p = 0.002) were recognized as risk factors for thrombosis in multivariable analyses, as shown in Table 2.
When ET were classified according to traditional risk stratification, thrombosis occurred in 23.1% of highrisk group, in comparison to 1.9% in low-risk group as shown in Fig. 2a. When the ET patients were analyzed by International Prognostic Score of Thrombosis for Essential Thrombocythemia (IPSET) 19 , thrombosis rate increased from 1.9% in low risk to 20.4% in high risk as shown in Fig. 2b. It is also notable that thrombosis prior to ET diagnosis only occurred in the high-risk groups. The multivariable Cox proportional model for ET patients showed that the history of thrombosis was associated with increased thrombosis events, as in Table 2. www.nature.com/scientificreports/ For PMF patients, history of thrombosis was associated with increased thrombosis events in multivariable analysis, as shown in Table 2.
Survival. The follow-up period for each MPN subtypes were 49 months, 64 months, 35 months, and 33 months for PV, ET, PMF, and MPN-U, respectively. The number of the patients followed up at 1 year, 5 years and 10 years were 136, 77, and 21, respectively. During the median follow-up of 46 months (range 12-161), 5 patients experienced secondary transformation: 2 to acute myeloid leukemia (1 ET, time to progression 43 months; 1 PMF, time to progression 7 months) and 3 to secondary myelofibrosis (1 PV, time to progression 127 months; 2 ET, median time to progression 63 months). The estimated overall survival for the entire cohort was 96.2%. While PV patients were associated with best survival and PMF patients with worst survival, the difference did not reach statistical difference. As shown in Fig. 3, patients with thromboembolic events had shorter overall survival compared to those who did not (p = 0.009).

Discussion
The importance of our study lies in that (1) we present rare data on the incidence and characteristics of East Asian patients, which seem to be similar to Western data albeit the traditional belief that Asians are less prone to thromboembolism; (2) we report risk factors for thromboembolism; and that (3) management of thromboembolism can lead not only to better quality of life but also better survival.
As shown in Table 3, our study provides evidence that Korean patients had similar frequency of thrombosis compared to that of Western patients, ranging 16 ~ 41% 19,23,24,28,31-33 . More specifically, recent meta-analysis of 29 cohort studies including populations of Europe, North America, Asia, and Australia reported that the pooled Table 2. Risk factors for thromboembolic events. * Patients who had two or more cardiovascular disease risk factors; CVD risks were defined as obesity (BMI > 25), smoking, hypertension, diabetes and dyslipidemia. ** Leukocytosis was defined as WBC > 15 × 10 9 /L. HR, Hazard ratio; CI, confidence interval; MPN, myeloproliferative neoplasms; PV, polycythemia vera; ET, essential thrombocythemia; PMF, primary myelofibrosis; dx, diagnosis; BMI, body mass index; WBC, white blood cell; CVD, cardiovascular disease.  12 . In our study, 27.3% of the patients experienced overall thrombosis, within close range of other ethnic populations. Also in subgroup analysis, despite the diverging rates of the thromboembolic events among multiple ethnic groups, our data provided the evidence that the Asians were not of lower risk of vascular complications. It is also notable that over one third of MPN-U patients experienced thromboembolism among all subgroups, albeit the small number of populations and the scarcity of the preceding data. When compared by the site of the thrombosis, stroke and ACS were the most prevalent but thromboses in uncommon sites like SVT were also reported, shown in Table 4. Markedly high prevalence of stroke and its decrease during the follow-up period indicated the role of the treatment with antiplatelet agents in reducing vascular complications in Asian MPN patients. Considerable number of patients who were diagnosed with MPNs within a month of thrombotic events also suggested the need for the early diagnosis and treatment to prevent vascular complications, as observed in the ECLAP study and the German MPN registry 13,18 . Risk factors for thromboembolism varied across subtypes of MPN, emphasizing the importance of individualized treatment depending on the subtype, presentation and comorbidities. PV was associated with the highest incidence of thromboembolism among MPN subgroups, comparable to previous observations that reported 26-39% 19,22,23 . In addition, that hemoglobin concentration was markedly associated with thrombosis events implied that effective control of hyperviscosity with phlebotomy is required to control the disease and reduce the vascular complications 23 . In ET, the severity of IPSET score was predictive of thromboembolic events. Therefore, ET patients should be treated more comprehensively including management of cardiovascular comorbidities to prevent thromboembolism in ET patients. In PMF, age over 60 years, JAK2V617F mutational status, and previous thrombosis had been identified as risk factors of thrombosis 26,27 . In our study, however, history of thrombotic event was the only predictive variable on multivariable analysis that included age, sex, organomegaly, JAK2V617F mutation status, blood counts, and cardiovascular risk factors as covariates (p = 0.035). The shorter www.nature.com/scientificreports/ overall survival of the patients with thromboembolic events underscored the importance of recognizing high risk patients and implementing personalized intervention. One of the most obvious pitfall of the study is its retrospective nature. Also, there is a possibility of referral bias, since all participating hospitals are academic centers. However, considering the longevity of MPN patients, prospective data is very difficult to accrue. The small number of MPN-U patients may have overestimated the prevalence of thromboembolism which was comparable to that of the PV, and the rate of splanchnic vein thrombosis (7.1%, 1/14). However, preliminary data showed that MPN-U patients with normal blood