Journal of Cardiology 81 (2023) 260–267

Contents lists available at ScienceDirect

Journal of Cardiology

journal homepage: www.elsevier.com/locate/jjcc

Original Article

Predictors of increased risk of adverse cardiovascular outcomes among

patients with myeloproliferative neoplasms and atrial fibrillation
Orly Leiva (MD) a,b, Andrew Jenkins (MD) b, Rachel P. Rosovsky (MD) c, Rebecca Karp Leaf (MD) c,
Katayoon Goodarzi (MD) c, Gabriela Hobbs (MD) c,⁎
a
Division of Cardiovascular Medicine, Department of Medicine, New York University Langone Health, New York, NY, USA
b
Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA
c
Division of Hematology and Oncology, Department of Medicine, Massachusetts General Hospital and Harvard Medical School, Boston, MA, USA

a r t i c l e i n f o a b s t r a c t

Article history: Background: Patients with myeloproliferative neoplasms (MPNs), essential thrombocythemia (ET), polycythemia
Received 30 July 2022 vera (PV), and myelofibrosis (MF), have increased risk of cardiovascular (CV) disease. Atrial fibrillation (AF) is
Received in revised form 4 October 2022 associated with adverse CV outcomes including arterial thrombosis, heart failure (HF), and CV death and coexists
Accepted 10 October 2022 with MPN. Traditional risk scores (CHA2DS2-VASC and HAS-BLED) for estimating risks/benefits of anticoagula-
Available online 29 October 2022
tion to prevent thrombotic events in AF do not include MPN status. Therefore, we aimed to investigate CV out-
comes in patients with MPN and AF and evaluate the predictive ability of traditional risk scores.
Keywords:
Atrial fibrillation
Methods: We conducted a single-center, retrospective cohort study of patients with MPN and AF. Primary out-
Myeloproliferative neoplasms come was composite of CV death and arterial thromboembolism; secondary outcomes were bleeding requiring
Cardio-oncology emergency department visit or hospitalization, hospitalization for HF, and all-cause death. Multivariable
Thrombosis competing-risk and Cox proportional hazards regression models were used to estimate risk of outcomes. Re-
Myelofibrosis ceiver operating characteristic (ROC) curve used to evaluate predictive ability of CHA2DS2-VASC and HAS-
BLED of composite outcome and bleeding, respectively.
Results: A total 142 patients was included (62 ET, 54 PV, 26 MF). Composite outcome, bleeding, HF hospitalization
and all-cause death occurred in 39 %, 30 %, 34 %, and 48 %, of patients respectively. After multivariable modeling,
MF was associated with increased risk of composite outcome (SHR 2.70, 95 % CI 1.38–5.27) and all-cause mortality
(HR 9.77, 95 % CI 4.88–19.54) but not bleeding (SHR 1.19, 95 % CI 0.51–2.80) or HF admissions (SHR 0.57, 95 % CI
0.19–1.72). CHA2DS2-VASC and HAS-BLED were poor predictors of composite outcome (C-statistic 0.52, 95 % CI
0.43–0.62) and bleeding (C-statistic 0.49, 95 % CI 0.40–0.58), respectively.
Conclusion: In patients with MPN and AF, MF is associated with increased risk of CV death and arterial thrombosis
and traditional risk scores do not accurately predict outcomes in this patient population. Further investigation is
needed to refine risk scores in this patient population.
© 2022 Japanese College of Cardiology. Published by Elsevier Ltd. All rights reserved.

Introduction also associated with arterial thromboembolism and HF [3,5]. Risk scores
are used to determine the risk and benefit of anticoagulation in the pre-
Myeloproliferative neoplasm (MPNs), including polycythemia vera vention of thrombotic events in patients with AF. The most commonly
(PV), essential thrombocytosis (ET), and myelofibrosis (MF), are used scores are the CHA2DS2-VASC for thrombotic risk and HAS-BLED
chronic hematopoietic stem cell neoplasms leading to proliferation of for estimation of major bleeding risk [9–12]. The CHA2DS2-VASC in-
mature myeloid cells and are associated with increased risk of cardio- cludes age, sex, heart failure history, hypertension, prior stroke or tran-
vascular (CV) disease including arterial and venous thrombosis and sient ischemic attack (TIA), atherosclerotic vascular disease, and
heart failure (HF) [1–6]. Patients with MPNs are commonly treated diabetes and HAS-BLED includes age, hypertension, liver and renal dis-
with cytoreductive therapy and aspirin to reduce the risk of thrombosis ease, prior bleeding, labile international normalized ratio, medications,
[7,8]. Atrial fibrillation (AF) is the most common atrial arrhythmia and is and alcohol use. Notably, these risk scores do not include the presence
of hematologic disorders or malignancies that may alter thrombotic
⁎ Corresponding author at: Division of Hematology and Oncology, Massachusetts
and bleeding risk [9,12].
General Hospital, 55 Fruit Street, Boston, MA, USA. The coexistence of AF and MPNs is not infrequent; one study of pa-
E-mail address: ghobbs@partners.org (G. Hobbs). tients with ET or PV showed a prevalence of atrial arrhythmias (97 %

https://doi.org/10.1016/j.jjcc.2022.10.007
0914-5087/© 2022 Japanese College of Cardiology. Published by Elsevier Ltd. All rights reserved.
O. Leiva, A. Jenkins, R.P. Rosovsky et al. Journal of Cardiology 81 (2023) 260–267

of which were AF) of 13.5 %, rising to as high as 31 % in patients aged 80 MF), and were diagnosed with AF at any time between January 1,
years or greater [13]. In this cohort, atrial arrhythmias were associated 2000 and January 1, 2020 (Fig. 1) [15]. Patients without complete infor-
with increased risk of thrombosis with an odds ratio (OR) of 2.47 (95 % mation on MPN diagnosis or AF diagnosis (including date of diagnoses,
confidence interval 1.12 to 5.45) [13]. Another study showed that patients N = 7) and patients who had AF after bone marrow transplantation
with PV and AF treated with oral anticoagulation had similar thrombotic (BMT) or acute leukemia transformation (N = 4) were excluded. Pa-
and bleeding events compared with patients with AF without PV [14]. tients were followed until death or date of last clinical encounter in
However, CV outcomes (including thrombosis, bleeding, and HF) after our electronic medical health record. The diagnosis of AF was made on
AF diagnosis in MPN patients has not been thorough investigated. Addi- electrocardiogram (ECG) or portable ECG monitoring. Patient demo-
tionally, the accuracy of commonly used risk scores such as CHA2DS2- graphics, baseline characteristics, and co-morbidities at the time of AF
VASC and HAS-BLED in patients with MPN in predicting adverse CV and MPN diagnosis, MPN treatment and cytoreduction, aspirin use, an-
events has not been well characterized. The identification of disease- ticoagulation use, and driver mutations (if known) were gathered.
specific risk factors may help stratify patients with MPN and AF. There- CHA2DS2-VASC and HAS-BLED scores were calculated at the time of AF
fore, we investigated whether CHA2DS2-VASC and HAS-BLED scores are and MPN co-diagnosis. Use of anti-arrhythmic therapy, including amio-
associated with risk of adverse CV outcomes (including thrombosis, darone, flecainide, sotalol, and dofetilide, at any time after AF diag-
bleeding, and HF) and if patients with MF are at increased risk of adverse nosis was captured. Patients were analyzed by type of MPN at the
cardiovascular outcomes compared with patients with ET or PV. time of AF diagnosis, and if patients had ET or PV but progressed to
myelofibrosis (MF) before diagnosis of AF then patients were labeled
Methods as MF for our analysis.
Outcomes of interest were a composite outcome of CV death and
This was a single-center retrospective observational cohort study non-fatal arterial thrombotic event. CV death was defined as death
using data gathered through the electronic health record (EHR) within due to myocardial infarction (MI), stroke, arrhythmia or sudden death,
Massachusetts General Hospital. The study was approved by the Institu- pulmonary embolisms (PE), or other cardiovascular causes including
tional Review Board of Massachusetts General Hospital. The require- HF or cardiogenic shock. Arterial thrombotic event was defined as either
ment of informed consent was waived by the Institutional Review MI, stroke or TIA, peripheral arterial thromboembolism or acute limb is-
Board. chemia. Acute ischemic stroke was defined as new focal neurologic def-
icit of sudden onset not due to non-vascular cause as confirmed by
Study population neurologist or imaging. Acute MI was defined per fourth universal defi-
nition of MI [16]. Other outcomes of interest include bleeding requiring
We identified 153 consecutive patients who were 18 years or older, hospitalization or emergency department visit, all-cause death, and
met World Health Organization diagnostic criteria for MPNs (ET, PV, hospitalization for HF.

Fig. 1. Inclusion and exclusion flow chart. A total of 153 patients of at least 18 years of age with MPN and AF diagnosis, of those 7 were excluded for incomplete information on MPN or AF
diagnosis and 4 were excluded for having initial AF diagnosis after bone marrow transplantation or progression to acute leukemia. A total of 142 patients were used in our analyses. Ab-
breviations: AF, atrial fibrillation; BMT, bone marrow transplantation; MPN, myeloproliferative neoplasm.

261
O. Leiva, A. Jenkins, R.P. Rosovsky et al. Journal of Cardiology 81 (2023) 260–267

Statistical analysis 46 %). Seven patients (5 %) with MPN underwent AF ablation (3 patients
with ET, 4 patients with PV, 0 patients with MF). Thirty-three (23 %) pa-
Continuous and categorical variables were compared using non- tients with MPN were treated with anti-arrhythmic medication at any
parametric Mann-Whitney test and Fisher's Exact Test, respectively. A time (13 patients with ET, 13 patients with PV, 7 patients with MF). The
two-sided p-value of <0.05 was considered significant. Cases with missing median LVEF at time of AF diagnosis was 63 % (IQR 55, 69). Seventeen per-
values were omitted from analyses requiring the missing variables. Al- cent of MPN patients had prior ischemic strokes or TIA (ET 18 %, PV 20 %,
though the MPNs share genetic similarities, MF is more aggressive and as- MF 8 %) and 18 % had prior bleeding events (ET 11 %, PV 19 %, MF 31 %).
sociated with worse prognosis and more frequent cytopenias compared to Eighty percent of patients with MPN had a driver mutation in the JAK2
ET and PV. Thus, we compared ET and PV with MF. To compare the overall gene (ET 69 %, PV 91 %, MF 81 %), 1 had a myeloproliferative leukemia
discriminative power of CHA2DS2-VASC score on composite outcome (ar- (MPL) mutation (1 MF patient), 5 had a calreticulin (CALR) mutation (all
terial thrombosis and CV death) and HF hospitalization and HAS-BLED ET), and 23 patients had unknown driver mutations (ET 14, PV 5, MF 4).
score on bleeding outcome we calculated the C-statistics of the receiver Ninety-four percent of MPN patients received cytoreduction treatment
operating characteristics (ROC) curve using non-parametric method. for MPN (ET 90 %, PV 98 %, MF 92 %). Hydroxyurea was commonly used
To assess the risk of composite outcome, bleeding, and HF hospitaliza- in MPN patients with 82 % of patients receiving it (ET 89 %, PV 76 %, MF
tion, we employed a time-to-event analysis using multivariable Cox pro- 77 %). MF patients (54 %) had high rates of ruxolitinib use compared
portional hazards regression and competing-risk regression model (Fine with ET (11 %) and PV (8 %) patients. Patient and MPN characteristics
and Gray method, with non-CV death as competing risk for composite out- are summarized in Table 1.
come analysis and all-cause death as competing risk for arterial thrombo-
sis, bleeding, and HF hospitalization). Cox proportional hazards results Outcomes of patients with MPN and atrial fibrillation
were reported as hazards ratio (HR) and competing risk regression results
were reported as subdistribution hazard ratio (SHR). For composite After a median follow-up of 49 months (IQR 18 to 86), all-cause
outcome we used age, Janus Kinase 2 (JAK2) mutation status, CHA2DS2- mortality occurred in 68 (48 %) of patients. Composite outcome of CV
VASC score, MPN type (ET or PV vs MF), treatment for MPN, death or arterial thrombosis occurred in 55 (39 %) of patients (36 % of
anticoagulation (AC) use, and aspirin use as co-variables. For arterial ET/PV patients and 50 % of MF patients, p = 0.27). CV death occurred
thrombosis outcome we used CHA2DS2-VASC score, MPN treatment, in 26 (18 %) of all patients (16 % of ET/PV vs 27 % of MF, p = 0.26)
MPN type (ET or PV vs MF), JAK2 mutations status, AC use, and aspirin and arterial thrombosis in 35 (25 %) of all MPN patients (25 % ET/PV
use as co-variables. For bleeding risk, we used age, HAS-BLED score, MPN vs 23 % MF, p > 0.99). Bleeding occurred in 42 (30 %) MPN patients
type, AC and aspirin use as co-variables. For risk of HF hospitalization, we (29 % ET/PV vs 31 % MF, p > 0.99). Hospitalization for HF occurred in
used age, sex, MPN type, MI after AF diagnosis, hospitalization for HF prior 48 (34 %) of MPN patients (37 % ET/PV vs 19 % MF, p = 0.11). MF had
to AF diagnosis, chronic kidney disease (CKD), left ventricular ejection frac- higher rates of all-cause mortality compared with ET or PV patients
tion (LVEF), and CHA2DS2-VASC score as co-variables. We also assessed (85 % vs 40 %, p < 0.0001). Results of outcomes are summarized in
the risk of all-cause mortality by using a Cox proportional hazards model Table 2.
using MPN type, age, CHA2DS2-VASC score, JAK2 mutation status, arterial
thrombosis after AF, bleeding after AF, HF hospitalization after AF, CKD, Time-to-event analyses
LVEF, aspirin use, and AC use as co-variables. Patients that were lost to
follow-up were censored at the time of their last clinical encounter. The Cox proportional hazards model demonstrated that MF (aHR 4.66,
As survival analysis only considers a single failure event and arterial 95 % CI 2.35–9.22) but not CHA2DS2-VASC (aHR 1.06, 95 % CI 0.84–1.34)
thrombosis, bleeding events and hospitalizations for HF may recur, a was associated with increased risk of composite outcome of arterial
multivariable negative binomial regression was used to model arterial thrombosis or CV death. Similarly, in the competing risk model, MF
thrombosis, bleeding, and HF hospitalization rate. These models con- (SHR 2.70, 95 % CI 1.38–5.27) was associated with increased risk of the
trolled for the same variables used in the arterial thrombosis, bleeding, composite outcome and CHA2DS2-VASC score was not (SHR 1.01, 95 %
and HF hospitalization Cox and competing risk models. The total dura- CI 0.80–1.29). For arterial thrombosis event, Cox proportional hazards
tion at-risk (from AF diagnosis to death or last follow-up) was the expo- model showed that MF (aHR 2.88, 95 % CI 1.14–7.25) was associated
sure variable in all three models. Results of negative binomial regression with increased risk arterial thrombosis, but CHA2DS2-VASC was not
modeling were reported as incident-rate ratios (IRR). Statistical analysis (aHR 1.09, 95 % CI 0.86–1.39). In the competing risk regression model,
was performed using Stata version 15.1 (StataCorp LLC, College Station, MF (SHR 1.45, 95 % CI 0.57–3.70) and CHA2DS2-VASC (SHR 1.00, 95 % CI
TX, USA). 0.80–1.25) were not associated with increased risk of arterial thrombosis.
The Cox proportional hazards model demonstrated increased risk of
Results bleeding in patients with MF (aHR 2.34, 95 % CI 1.05–5.22), however
this increased risk was not seen after competing risk regression model-
Patient characteristics ing (SHR 1.19, 95 % CI 0.51–2.80). A HAS-BLED score ≥ 2 was not associ-
ated with increased risk of bleeding in both Cox and competing risk
After excluding 11 patients, a total of 142 patients were included models (aHR 1.34, 95 % CI 0.55–3.29, and SHR 1.25, 95 % CI 0.50–3.12,
who had AF and MPN (62 with ET, 54 with PV, 26 with MF). The median respectively).
age at time of MPN and AF diagnosis was 78 (IQR 71, 83) years for all pa- There was no increased risk of HF hospitalization with MF in either
tients, 80 (IQR 73, 86) years for ET patients, 78 (IQR 70, 84) years for PV Cox proportional (aHR 1.25, 95 % CI 0.45–3.51) or competing risk regres-
patients, and 74 (IQR 69, 80) years for MF patients. Fifty-four percent of sion model (SHR 0.57, 95 % CI 0.19–1.72). However, after Cox and com-
all MPN patients were male (ET 45 %, PV 54 %, MF 73 %). Median follow- peting risk regression models, CHA2DS2-VASC score was associated with
up time (IQR) was 49 months for all MPN patients (18, 86), 61 months increased risk of HF hospitalization after AF diagnosis (aHR 1.35, 95 % CI
(28, 69) for ET, 58 months (20, 90) for PV, and 14 months (3, 29) for MF. 1.06–1.73, and SHR 1.34, 95 % CI 1.08–1.65).
Thirty-five patients were lost to follow-up. After Cox proportional hazards model, MF (aHR 9.77, 95 % CI 4.88–
Eighteen percent of all MPN patients had AF prior to MPN diagnosis 19.54) but not CHA2DS2-VASC score (aHR 1.21 95 % CI 0.95–1.53) was
(19 % ET and PV, 15 % MF). Median CHA2DS2-VASC score for all MPN pa- associated with increased risk of all-cause mortality. Cumulative inci-
tients was 4 (ET 4, PV 4, MF 3) and median HAS-BLED score was 2 (ET, PV, dence curves of CV death or arterial thrombosis, arterial thrombosis,
and MF 2). AC was used in 69 % of patients with MPN (ET 68 %, PV 80 %, bleeding, and HF hospitalization by MPN type, CHA2DS2-VASC, and
MF 50 %) and aspirin was used in 68 % of patients (ET 73 %, PV 72 %, MF HAS-BLED scores are shown in Fig. 2.

262
O. Leiva, A. Jenkins, R.P. Rosovsky et al. Journal of Cardiology 81 (2023) 260–267

Table 1 Table 2
Patient characteristics and co-morbidities. Outcomes by MPN type.

All ET or PV MF p-Value All ET or PV MF p-Value

N = 142 N = 116 N = 26 N = 142 N = 116 N = 26

Age at MPN diagnosis, median (IQR) 73 (63, 81) 74 (64, 82) 70 (63, 75) 0.080 Composite outcome, N (%) 55 (39) 42 (36) 13 (50) 0.27
Age at MPN and AF diagnosis, median 78 (71, 83) 79 (71, 84) 74 (69, 80) 0.061 CV death 26 (18) 19 (16) 7 (27) 0.26
(IQR) Arterial thrombosis 35 (25) 29 (25) 6 (23) 1.00
Median follow-up, months (IQR) 49 (18, 86) 60 (25, 89) 14 (3, 29) <0.0001 Myocardial infarction 14 12 2 –
BMI, median (IQR)a 24 (21, 27) 24 (22, 27) Stroke or TIA 14 12 2 –
Male 76 (54) 57 (49) 19 (73) 0.031 ATE 12 9 3 –
Female 66 (46) 59 (51) 7 (27) Median time to composite, 23 (5, 64) 28 (8, 74) 10 (0, 24) 0.014
AF diagnosed before MPN, N (%) 26 (18) 22 (19) 4 (15) 0.78 months (IQR)
CHA2DS2-VASC, median (IQR) 4 (3, 5) 4 (3, 5) 3 (2, 4) 0.001 Bleed, N (%) 42 (30) 34 (29) 8 (31) 1.00
HAS-BLED, median (IQR) 2 (2,2) 2 (2, 2) 2 (2, 2) 0.19 Gastrointestinal 19 14 5 –
Anticoagulation, N (%) 98 (69) 85 (73) 13 (50) 0.132 Intracranial 7 5 2 –
Warfarin, N (% AC) 72 (73) 61 (72) 11 (85) Musculoskeletal 10 10 0 –
DOACb, N (% of AC) 23 (23) 21 (25) 2 (15) Other 6 5 1 –
LMWH, N (% of AC) 3 (3) 3 (3) 0 Median time to first bleeding 28 (9, 64) 31 (11, 72) 16 (1, 36) 0.14
Ablation for AF, N (%) 7 (5) 7 (6) 0 0.35 event, months (IQR)
Anti-arrhythmic use at any time, N (%) 33 (23) 26 (22) 7 (27) 0.61 HF hospitalization, N (%) 48 (34) 43 (37) 5 (19) 0.11
LVEF, median % (IQR)a 63 (55, 69) 64 (55, 70) 62 (56, 66) 0.90 All-cause mortality, N (%) 68 (48) 46 (40) 22 (85) < 0.0001
LVEF <40 % 12 (6) 11 (10) 1 (4) 0.46
ATE, arterial thromboembolism; CV, cardiovascular; ET, essential thrombocythemia; HF,
Aspirin use, N (%) 96 (68) 84 (72) 12 (46) 0.019
heart failure; IQR, interquartile range; MF, myelofibrosis; PV, polycythemia vera; TIA, tran-
Co-morbidities, N (%)
sient ischemic attack.
Hypertension 97 (68) 84 (72) 13 (50) 0.036
Heart failure 50 (35) 42 (36) 8 (31) 0.66
Prior HF hospitalization 16 (11) 12 (10) 4 (15) 0.49
Hyperlipidemia 66 (46) 53 (46) 13 (50) 0.83 patients with MF (IRR 2.76, 95 % CI 1.01–7.53). Additionally, CHA2DS2-
Diabetes 15 (11) 13 (11) 2 (8) 1.00 VASC score was associated with higher rates of arterial thrombotic
Prior stroke or TIA 24 (17) 22 (19) 2 (8) 0.25
events (IRR 1.28, 95 % CI 1.00–1.63).
CAD 46 (32) 39 (34) 7 (27) 0.64
Prior VTE 24 (17) 21 (18) 3 (12) 0.57
A total of 34 bleeding events (14 gastrointestinal, 5 intracranial, 10
Prior bleed 25 (18) 17 (15) 8 (31) 0.083 musculoskeletal, 5 other sites) occurred in 34 patients (29 %) with ET
Chronic kidney disease 18 (13) 13 (11) 5 (19) 0.32 or PV compared with 8 events (5 gastrointestinal, 2 intracranial, 0 mus-
Smoking history culoskeletal, 1 other site) in 7 patients (27 %) with MF. The multivari-
Never smoker 59 (42) 52 (45) 7 (27) 0.17
able, negative binomial regression model revealed a significantly
Former or current smoker 81 (57) 63 (54) 18 (69)
Unknown smoking 2 (1) 1 (1) 1 (4) higher rate of bleeding events (including recurrent events) in patients
MPN characteristics with MF (IRR 2.45, 95 % CI 1.09–5.48). However, a HAS-BLED score of
Driver mutation 2 or greater was not associated with higher rates of bleeding events
JAK2 113 (80) 92 (79) 21 (81) 0.27
(IRR 0.89, 95 % CI 0.41–1.93).
MPL 1 (1) 0 1 (4)
CALR 5 (4) 5 (4) 0
A total of 62 HF hospitalizations occurred in 43 patients (37 %) with
Unknown 23 (16) 19 (16) 4 (15) ET or PV compared with 5 HF hospitalizations in 5 patients (19 %) with
MPN treatment MF. After multivariable negative binomial regression model, MF was not
Any 133 (94) 109 (94) 24 (92) 0.67 associated with rates of HF hospitalization (IRR 1.27, 95 % CI 0.49–3.33)
Ruxolitinib 26 (18) 12 (10) 14 (54) <0.0001
although CHA2DS2-VASC score was associated with increased rates of
Hydroxyurea 116 (82) 96 (83) 20 (77) 0.57
Anagrelide 17 (12) 15 (13) 2 (8) 0.74 HF hospitalization (IRR 1.30, 95 % CI 1.05–1.60). The results of these
Spleen size, cm (IQR) a 14 (11, 17) 13 (11, 15) 19 (16, 23) <0.0001 multivariable models and univariable analyses are summarized in
Progression to MF or leukemic 16 (11) 12 (10) 4 (15) 0.49 Table 3.
transformation, N (%)
Labs at MPN diagnosis, mean ± SD
WBC (K/μL) 13.5 ± 10.0 11.8 ± 7.1 20.9 ± 16.2 0.005
Predictive performance of CHA2DS2-VASC and HAS-BLED scores
Hematocrit (%) 44 ± 8 45 ± 8 38 ± 7 <0.0001
Platelets (K/μL) 587 ± 276 621 ± 255 434 ± 320 0.001 CHA2DS2-VASC score did not predict composite outcome (C statistic
LDH (IU/L)a 353 ± 270 433 ± 320 585 ± 437 0.0001 0.52, 95 % CI 0.43–0.62) or arterial thrombosis (C statistic 0.49, 95 % CI
AF, atrial fibrillation; BMI, body mass index; CAD, coronary artery disease; CALR, 0.38–0.59). However, CHA2DS2-VASC score had moderate ability to pre-
calrecticulin; DOAC, direct-acting oral anticoagulant; ET, essential thrombocythemia; HF, dict HF hospitalization (C statistic 0.66, 95 % CI 0.57–0.76). HAS-BLED
heart failure; IQR, interquartile range; JAK2, Janus kinase 2; LDH, lactate dehydrogenase;
score did not predict bleeding event (C statistic 0.49, 95 % CI 0.40–0.59).
LMWH, low-molecular weight heparin; LVEF, left ventricular ejection fraction; MF, myelo-
fibrosis; MPL, myeloproliferative leukemia protein; MPN, myeloproliferative neoplasm;
Receiver-operating characteristic (ROC) curves are illustrated in Fig. 3.
PV, polycythemia vera; TIA, transient ischemic attack; VTE, venous thromboembolism;
WBC, white blood cell. Discussion
a
Patients with missing information: BMI 5 patients, LVEF 8 patients, spleen size 23
patients, LDH 63 patients.
b MPN and AF are both risk factors for arterial thromboembolic events
DOACs include apixaban, rivaroxaban, dabigatran, and edoxaban.
and often coexist. Our study suggested that patients with MPN have
high rates of CV death or arterial thrombotic events (39 % of all MPN pa-
Negative binomial modeling of recurrent events tients) and arterial thrombosis (25 % for all MPN patients) despite high
rates of AC, aspirin use, and cytoreductive therapy (69 %, 68 %, and 92 %
A total of 33 arterial thrombotic events (12 MIs, 12 strokes, 9 periph- respectively). After multivariable time-to-event analyses, our study
eral arterial thromboembolism) occurred in 29 patients (25 %) with ET suggests that among patients with MPN, MF patients are associated
or PV compared with 7 events (2 MIs, 2 strokes, 3 peripheral arterial with increased risk of CV death or arterial thrombotic events, bleeding,
thromboembolic events) in 6 patients (23 %) with MF. The multivari- and all-cause mortality compared with ET and PV patients. Additionally,
able, negative binomial regression model revealed a significantly higher our study suggests that the traditional risk stratification scores
rate of arterial thrombosis events (including recurrent events) in CHA2DS2-VASC and HAS-BLED for arterial thrombotic events and

263
O. Leiva, A. Jenkins, R.P. Rosovsky et al. Journal of Cardiology 81 (2023) 260–267

Fig. 2. Cumulative incidence of composite outcome, arterial thrombosis, bleeding, and heart failure. Incidence curve of composite by MPN (A), arterial thrombosis by CHA2DS2-VASC score
(B), bleeding by HAS-BLED scores (C), and heart failure hospitalization by CHA2DS2-VASC score (D). Abbreviations: AF, atrial fibrillation; CV, cardiovascular; ET, essential
thrombocythemia; MF, myelofibrosis; PV, polycythemia vera.

bleeding in AF, respectively, may have limitations in predicting events compared with patients with ET or PV. This is concordant with a meta-
in MPN patients. However, CHA2DS2-VASC score was associated with analysis of MPN patients which showed patients with MF having higher
HF hospitalizations in patients with MPN and AF. prevalence of hemorrhagic complications compared with ET and PV (8.9
Arterial thrombotic events are common in patients with MPNs. A re- % vs 7.3 % and 6.9 %, respectively) [17]. Similar to our results on the effect
cent meta-analysis of over 13,000 patients showed a pooled prevalence of MF on arterial thrombosis, despite discordant findings of no increased
of overall thrombosis of 20 % and arterial thrombosis of 16.2 % among all risk on competing-risk regression of bleeding, patients with MF had in-
MPN patients [17]. In this study, patients with MF had lower prevalence creased risk of bleeding on Cox proportional hazards regression and in-
of overall thrombosis compared with patients with ET or PV (9.5 % for creased rates of bleeding on negative binomial regression. Further
MF, 28.5 % for PV, and 20.7 % for ET). Indeed, other papers have shown research is needed to identify risk factors for bleeding in patients with
lesser numbers of thrombotic events in patients with MF compared with MF and AF. Our results also suggest that HAS-BLED score may not be accu-
ET or PV patients [18]. However, these studies did not specifically investi- rate in patients with MPN. Given the high risk of bleeding with anticoagu-
gate patients with AF. Additionally, patients with MF have worse long- lation in patients with MPN, further investigation is needed [20,21]. HF is
term survival than patients with ET or PV and thus their time at-risk of also commonly associated with AF and contributes to morbidity and mor-
thrombosis is shorter. In fact, after adjusting for non-CV death, patients tality in patients with this condition [3]. Patients with MPN are at increased
with MF had similar rates of thrombosis compared with ET in one study risk of HF although HF outcomes in patients with AF and MPN have not
[19]. Our study employed a competing risk regression model and our find- been thoroughly investigated [22]. Our study showed a high rate of HF
ings suggest that patients with MF and AF are at increased risk of compos- hospitalization after AF diagnosis with more than one third of patients
ite outcome of arterial thrombosis and CV death compared with ET and PV. with MPN and AF having at least one HF hospitalization.
Additionally, while competing risk regression did not suggest an increased Patients with MF have a comparable risk of arterial thrombosis and
risk in arterial thrombosis in patients with MF compared with ET or PV increased risk of bleeding compared with patients with ET or PV [23,
(likely owing to increased all-cause death), our Cox proportional hazards 24]. This is of particular importance in patients with AF, where bleeding
and negative binomial regression results suggest that arterial thrombosis and thrombosis are competing risks. Potential mechanisms of increased
may occur more frequently in patients with MF. Further studies on miti- risk of thrombosis in patients with MF include abnormal platelet activa-
gating this risk in patients with MF are warranted. tion and excreted pro-thrombotic factors [6]. Patients with MF have el-
Bleeding has also been described to be a complication of MPN, par- evated serum levels and platelet expression of lysyl oxidase (LOX).
ticularly in patients receiving anticoagulation [17]. Our study suggests While LOX is implicated in collagen crosslinking, it has also been
that in patients with AF, MF is associated with increased bleeding risk shown to increase platelet adhesion to collagen and platelet activation.

264
 O. Leiva, A. Jenkins, R.P. Rosovsky et al.

Table 3
Risk (HR, SHR, or IRR) of adverse cardiovascular events.

Outcome Cox model, Cox model, Competing risk model, Competing risk model, Negative binomial model, Negative binomial model, C statistic (95 % CI)
unadjusted adjusted unadjusted adjusted unadjusted adjusted
HR (95 % CI) HR (95 % CI) SHR (95 % CI) SHR (95 % CI) IRR (95 % CI) IRR (95 % CI)

MF
Compositea 3.72 (1.92–7.22) 4.66 (2.35–9.22) 2.15 (1.10–4.22) 2.70 (1.38–5.27) – – –
Arterial 2.38 (0.95–5.97) 2.88 (1.14–7.25) 0.95 (0.38–2.36) 1.45 (0.57–3.70) 2.15 (0.83–5.55) 2.76 (1.01–7.53) –
thrombosisb
Bleedingc 2.26 (1.02–5.00) 2.34 (1.05–5.22) 1.06 (0.47–2.39) 1.19 (0.51–2.80) 2.73 (1.13–6.60) 2.45 (1.09–5.48) –
HF hospitalizationd 1.14 (0.44–2.92) 1.25 (0.45–3.51) 0.47 (0.18–1.22) 0.57 (0.19–1.72) 0.76 (0.24–2.46) 1.27 (0.49–3.33) –
All-cause mortalitye 5.02 (2.96–8.53) 9.77 (4.89–19.54) – – – – –

CHADS2VASC or HAS-BLEDf
Compositea 1.11 (0.93–1.34) 1.06 (0.84–1.34) 1.07 (0.90–1.28) 1.01 (0.80–1.29) – – 0.52 (0.43–0.62)

265
Arterial 1.10 (0.87–1.38) 1.09 (0.86–1.39) 1.06 (0.86–1.30) 1.00 (0.80–1.25) 1.25 (0.99–1.57) 1.28 (1.00–1.63) 0.49 (0.38–0.59)
thrombosisb
Bleedingf,c 1.98 (0.85–4.57) 1.34 (0.55–3.29) 1.52 (0.69–3.35) 1.25 (0.50–3.12) 1.34 (0.61–2.93) 0.89 (0.41–1.93) 0.49 (0.40–0.58)
HF hospitalizationd 1.51 (1.23–1.84) 1.35 (1.06–1.73) 1.43 (1.20–1.71) 1.34 (1.08–1.65) 1.63 (1.30–2.05) 1.30 (1.05–1.60) 0.66 (0.57–0.76)
e
All-cause mortality 1.15 (0.98–1.36) 1.21 (0.95–1.53) – – – – –

CI, confidence interval; HF, heart failure; HR, hazards ratio; IRR, incidence rate ratio; MF, myelofibrosis; SHR, subhazard ratio.
a
Adjusted for age, JAK2 mutation status, CHA2DS2-VASC score, MPN type (ET or PV vs MF), treatment for MPN, anticoagulation (AC) use, and aspirin use.
b
Adjusted for CHA2DS2-VASC score, MPN treatment, MPN type (ET or PV vs MF), JAK2 mutations status, anticoagulation (AC) use, and aspirin use.
c
Adjusted for age, HAS-BLED score ≥ 2, MPN type, AC, and aspirin use.
d
Adjusted for age, sex, MPN type, MI after AF diagnosis, hospitalization for HF prior to AF diagnosis, CHA2DS2-VASC score, LVEF, and CKD.
e
Adjusted for MPN type, age, CHA2DS2-VASC score, JAK2 mutation status, arterial thrombosis after AF, bleeding after AF, HF hospitalization after AF, LVEF, and CKD aspirin use and AC use.
f
HAS-BLED for bleeding outcome only.
Journal of Cardiology 81 (2023) 260–267
O. Leiva, A. Jenkins, R.P. Rosovsky et al. Journal of Cardiology 81 (2023) 260–267

Fig. 3. Receiver-operating characteristics (ROC) curves for outcome detection. ROC curves for CHA2DS2-VASC score for composite outcome (A), arterial thrombosis (B), heart failure
hospitalization (C) detection, and HAS-BLED score for bleeding detection (D). Abbreviations: AUC, area under curve; ROC, receiver-operating characteristics.

This may be partially responsible for increased thrombosis risk in MF and selection bias. Additionally, CHA2DS2-VASC and HAS-BLED scores are
further studies are needed to elucidate its clinical implications [25–27]. tools best validated in patients who have not been started on AC to as-
Patients with MF are at increased risk of thrombocytopenia due to bone sess benefits and risks of AC. Most of the patients in our cohort were
marrow replacement with fibrosis which can increase risk of bleeding started on AC and scores were calculated at the time when both AF
and may limit anticoagulation use [28]. Additionally, patients with MF and MPN were diagnosed, thus care should be taken in interpreting
have worse outcomes in general compared with ET and PV patients our results. However, further studies are necessary to investigate the ac-
given their increased risk of leukemic transformation, infections due to curacy of risk scores in patients with concurrent MPN and AF. Addition-
cytopenias, and constitutional symptoms (fevers, weight loss, and malnu- ally, our small sample size limits our power to detect smaller magnitude
trition) [29]. Further investigation is needed to better assess risk of throm- of association between CHA2DS2-VASC and HAS-BLED scores and out-
bosis and bleeding in patients with MF and AF. comes and thus larger studies are warranted.
Our study also suggests that CHA2DS2-VASC and HAS-BLED risk In conclusion, our study suggests that patients with MPN and AF
scores, which do not include hematologic or oncologic conditions, may have high rates of thrombotic and bleeding events. Patients with MF
have limitations in predicting thrombotic and hemorrhagic events in may have increased risk of CV death and arterial thrombosis compared
this population [30]. In one study, CHA2DS2-VASC score did not predict with patients with ET or PV. Additionally, further investigation into the
thromboembolic risk in patients with malignancy and new AF after can- validation and development of risk scores may improve patient selec-
cer diagnosis [31]. HAS-BLED score, on the other hand, has been shown tion for AC in this patient population and the identification of patients
to correlate with bleeding risk in patients with cancer [32]. However, at high risk of CV events (including HF).
these risk scores have not been thoroughly studied in patients with
MPN. Although the CHA2DS2-VASC score is routinely used in risk-
stratification for thromboembolism in AF, it includes several cardiovas- Acknowledgments
cular risk factors and has been shown to have prognostic implications in
other cardiovascular conditions including acute MI and HF [33,34]. In- O.L. and G.H. received the American Society of Hematology (ASH)
deed, our study suggests that while CHA2DS2-VASC score was not asso- MHRAP grant. G.H. received grants from ASH-AMFDP and K12 Paul
ciated with our composite outcome or arterial thrombosis, it was Calabresi award.
associated with increased risk of HF hospitalization and overall rate of
arterial thrombosis. Further investigation on refining cardiovascular References
risk scores is needed in this patient population.
[1] Barbui T, Finazzi G, Falanga A. Myeloproliferative neoplasms and thrombosis. Blood
Our study is limited by its single-center and retrospective nature.
2013;122:2176–84.
The selection of patients seen at only one tertiary referral center and [2] Falanga A, Marchetti M. Thrombosis in myeloproliferative neoplasms. Semin
the exclusion of patients with incomplete information may introduce Thromb Hemost 2014;40:348–58.

266
O. Leiva, A. Jenkins, R.P. Rosovsky et al. Journal of Cardiology 81 (2023) 260–267

[3] Noseworthy PA, Kaufman ES, Chen LY, Chung MK, Elkind MSV, Joglar JA, et al. Sub- [19] Barbui T, Carobbio A, Cervantes F, Vannucchi AM, Guglielmelli P, Antonioli E, et al.
clinical and device-detected atrial fibrillation: pondering the knowledge gap: a sci- Thrombosis in primary myelofibrosis: incidence and risk factors. Blood 2010;115:
entific statement from the American Heart Association. Circulation 2019;140: 778–82.
e944–63. [20] Hamulyak EN, Daams JG, Leebeek FWG, Biemond BJ, Te Boekhorst PAW, Middeldorp
[4] Vermond RA, Crijns HJ, Tijssen JG, Alings AM, Van den Berg MP, Hillege HL, et al. S, et al. A systematic review of antithrombotic treatment of venous thromboembo-
Symptom severity is associated with cardiovascular outcome in patients with per- lism in patients with myeloproliferative neoplasms. Blood Adv 2021;5:113–21.
manent atrial fibrillation in the RACE II study. Europace 2014;16:1417–25. [21] How C-J, Story CM, Ren S, Neuberg DS, Rosovsky RP, Hobbs G, et al. High incidence of
[5] Soliman EZ, Safford MM, Muntner P, Khodneva Y, Dawood FZ, Zakai NA, et al. Atrial bleeding found with direct oral anticoagulant use in myeloproliferative neoplasm
fibrillation and the risk of myocardial infarction. JAMA Intern Med 2014;174: patients. Blood 2021;138(Suppl. 1):3632.
107–14. [22] Wu J, Zhang L, Vaze A, Lin S, Juhaeri J. Risk of Wernicke's encephalopathy and cardiac
[6] Leiva O, Hobbs G, Ravid K, Libby P. Cardiovascular disease in myeloproliferative neo- disorders in patients with myeloproliferative neoplasm. Cancer Epidemiol 2015;39:
plasms. JACC CardioOncol 2022;4:166–82. 242–9.
[7] Carobbio A, Thiele J, Passamonti F, Rumi E, Ruggeri M, Rodeghiero F, et al. Risk factors for [23] Kc D, Falchi L, Verstovsek S. The underappreciated risk of thrombosis and bleeding in
arterial and venous thrombosis in WHO-defined essential thrombocythemia: an inter- patients with myelofibrosis: a review. Ann Hematol 2017;96:1595–604.
national study of 891 patients. Blood 2011;117:5857–9. [24] Elliott MA, Pardanani A, Lasho TL, Schwager SM, Tefferi A. Thrombosis in myelofibro-
[8] Cerquozzi S, Barraco D, Lasho T, Finke C, Hanson CA, Ketterling RP, et al. Risk factors sis: prior thrombosis is the only predictive factor and most venous events are pro-
for arterial versus venous thrombosis in polycythemia vera: a single center experi- voked. Haematologica 2010;95:1788–91.
ence in 587 patients. Blood Cancer J 2017;7:662. [25] Abbonante V, Chitalia V, Rosti V, Leiva O, Matsuura S, Balduini A, et al. Upregulation
[9] Lip GY, Nieuwlaat R, Pisters R, Lane DA, Crijns HJ. Refining clinical risk stratification of lysyl oxidase and adhesion to collagen of human megakaryocytes and platelets in
for predicting stroke and thromboembolism in atrial fibrillation using a novel risk primary myelofibrosis. Blood 2017;130:829–31.
factor-based approach: the euro heart survey on atrial fibrillation. Chest 2010;137:
[26] Matsuura S, Mi R, Koupenova M, Eliades A, Patterson S, Toselli P, et al. Lysyl oxidase
263–72.
is associated with increased thrombosis and platelet reactivity. Blood 2016;127:
[10] Friberg L, Rosenqvist M, Lip GY. Evaluation of risk stratification schemes for ischae-
1493–501.
mic stroke and bleeding in 182 678 patients with atrial fibrillation: the Swedish
[27] Tadmor T, Bejar J, Attias D, Mischenko E, Sabo E, Neufeld G, et al. The expression of
atrial fibrillation cohort study. Eur Heart J 2012;33:1500–10.
lysyl-oxidase gene family members in myeloproliferative neoplasms. Am J Hematol
[11] Ntaios G, Lip GY, Makaritsis K, Papavasileiou V, Vemmou A, Koroboki E, et al. CHADS
2013;88:355–8.
(2), CHA(2)S(2)DS(2)-VASc, and long-term stroke outcome in patients without
[28] Didisheim P, Bunting D. Abnormal platelet function in myelofibrosis. Am J Clin
atrial fibrillation. Neurology 2013;80:1009–17.
Pathol 1966;45:566–73.
[12] Pisters R, Lane DA, Nieuwlaat R, de Vos CB, Crijns HJ, Lip GY. A novel user-friendly
score (HAS-BLED) to assess 1-year risk of major bleeding in patients with atrial fi- [29] Mughal TI, Vaddi K, Sarlis NJ, Verstovsek S. Myelofibrosis-associated complications:
brillation: the euro heart survey. Chest 2010;138:1093–100. pathogenesis, clinical manifestations, and effects on outcomes. Int J Gen Med 2014;
[13] Mahe K, Delluc A, Chauveau A, Castellant P, Mottier D, Dalbies F, et al. Incidence and 7:89–101.
impact of atrial arrhythmias on thrombotic events in MPNs. Ann Hematol 2018;97: [30] Leiva O, AbdelHameid D, Connors JM, Cannon CP, Bhatt DL. Common pathophysiol-
101–7. ogy in cancer, atrial fibrillation, atherosclerosis, and thrombosis. JACC CardioOncol
[14] de Freitas AS, Alvarez-Larran A. Risk of thrombosis and hemorrhage in patients with 2021;3:619–34.
polycythemia vera and atrial fibrillation treated with prophylactic oral anticoagu- [31] Hu YF, Liu CJ, Chang PM, Tsao HM, Lin YJ, Chang SL, et al. Incident thromboembolism
lants. Ann Hematol 2016;95:1903–4. and heart failure associated with new-onset atrial fibrillation in cancer patients. Int J
[15] Arber DA, Orazi A, Hasserjian R, Thiele J, Borowitz MJ, Le Beau MM, et al. The 2016 Cardiol 2013;165:355–7.
revision to the World Health Organization classification of myeloid neoplasms and [32] Pastori D, Marang A, Bisson A, Menichelli D, Herbert J, Lip GYH, et al. Thromboembo-
acute leukemia. Blood 2016;127:2391–405. lism, mortality, and bleeding in 2,435,541 atrial fibrillation patients with and with-
[16] Thygesen K, Alpert JS, Jaffe AS, Simoons ML, Chaitman BR, White HD, et al. Third uni- out cancer: a nationwide cohort study. Cancer 2021;127:2122–9.
versal definition of myocardial infarction. Glob Heart 2012;7:275–95. [33] Kim KH, Kim W, Hwang SH, Kang WY, Cho SC, Kim W, et al. The CHA2DS2VASc score
[17] Rungjirajittranon T, Owattanapanich W, Ungprasert P, Siritanaratkul N, can be used to stratify the prognosis of acute myocardial infarction patients irrespec-
Ruchutrakool T. A systematic review and meta-analysis of the prevalence of throm- tive of presence of atrial fibrillation. J Cardiol 2015;65:121–7.
bosis and bleeding at diagnosis of Philadelphia-negative myeloproliferative neo- [34] Sonaglioni A, Lonati C, Rigamonti E, Vigano M, Nicolosi GL, Proietti M, et al.
plasms. BMC Cancer 2019;19:184. CHA2DS2-VASc score stratifies mortality risk in heart failure patients aged 75
[18] Song IC, Yeon SH, Lee MW, Ryu H, Lee HJ, Yun HJ, et al. Thrombotic and hemorrhagic years and older with and without atrial fibrillation. Aging Clin Exp Res 2022;34:
events in 2016 World Health Organization-defined Philadelphia-negative myelopro- 1707–20.
liferative neoplasm. Korean J Intern Med 2021;36:1190–203.

267