Long-term real-world thrombotic and clinical outcomes in polycythemia vera – a hospital-based i2b2 cohort study

Abstract

Though the thrombotic risk in polycythemia vera (PV) is well known, long-term real-world data remain limited. This retrospective study at Bern University Hospital aimed to evaluate complications in PV patients, focusing on thrombotic events, associated risk factors, and disease outcomes. Among 727,731 screened adults, 107 PV patients were identified (median age 58 years, range 18–92; 51% male). Median follow-up was 115 months (range 1–353). Thrombotic events occurred in 56% of patients, including 58% arterial and 42% venous events. Arterial and venous thrombosis before or at diagnosis was documented in 9 (8.4%) and 36 (33.6%) patients, respectively, and in 13 (12.1%) during follow-up. Splanchnic thrombosis was associated with younger age and splenomegaly. Multivariate analysis showed an increased thrombotic risk linked to dyslipidemia (RR 2.573, 95% CI 1.058–6.257) and older age at diagnosis (> 60 years; RR 2.524, 95% CI 1.111–5.734). Risk of death increased with age (RR 1.052, 95% CI 1.009–1.097) and myelofibrosis transformation (RR 6.856, 95% CI 1.948–24.120). This long-term study confirms a high thrombotic burden in PV, especially arterial events. Splanchnic thrombosis typically presented early on in younger patients without recurrence. The high prevalence of cardiovascular risk factors highlights the critical importance of managing these risks to mitigate thrombotic complications in PV.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00277-026-06768-w.

Introduction

Polycythemia vera (PV) is a Philadelphia-negative myeloproliferative neoplasm (MPN) characterized by overproduction of blood cells, particularly polyglobulia or erythrocytosis [1], as a result of a constitutive activation of the Janus Kinase and Signal Transducer and Activator of Transcription (JAK-STAT) signal transduction pathway [1]. The formal diagnosis of PV requires defined criteria, which were revised by WHO (World Health Organization) [2] and the International Consensus Classification (ICC) of Myeloid Neoplasms and Acute Leukemias in 2022 [2, 3].

Clinical symptoms in PV might affect patients' quality of life, particularly those related to splenomegaly, the presence of fatigue, and pruritus. The most common complications of PV are the increased risk of thrombotic events and less frequently, the occurrence of bleeding complications, both associated with a higher risk of morbidity and mortality [4, 5]. Thrombotic events occur in up to 20% of PV patients at diagnosis, with an annual incidence of 2–3%, increasing to over 7% in patients with a history of prior events [6]. PV has a potential long-term risk of progression to myelofibrosis (MF) and/or acute myeloid leukemia (AML) transformation [7, 8]. To prevent thrombotic complications, the primary goal in PV management is to maintain the hematocrit (Hct) below 45%, along with controlling other cytoses [9, 10]. An additional goal of treatment is to ameliorate PV symptom burden in order to improve quality of life [10, 11]. Risk factors that identify PV patients at higher risk of progression to MF and/or transformation into AML have not been well defined to date [7].

All PV patients benefit from a treatment with phlebotomy targeting a Hct below 45%, in addition to low dose aspirin. Depending on the patient’s risk status and clinical needs, cytoreductive therapies such as hydroxyurea, interferons, including pegylated IFN-alpha or ropeginterferon alpha-2b, and JAK inhibitors may be considered [10]. While the risk of thrombosis in patients with PV is known, real-world data on thrombotic complications and other outcomes in a cohort with a long observation period are scarce. Therefore, we aimed to evaluate real-world data from PV patients focusing on complications, primarily thrombotic events and its associated risk factors and secondarily on disease outcomes.

Methods

The study was performed at the University Hospital Inselspital, Bern, Switzerland. This study has been approved by the local ethics committee, Kantonale Ethikkommission Bern under Basec number 2021 − 00380.

Cohort definition

We aimed to identify PV patients among all patients seen at our hospital between October 1, 2008 and July 31, 2019. The hospital database (Microsoft SQL Server Management Studio) was searched to identify patients with polyglobulia, using i2b2 (Informatics for Integrating Biology and the Bedside) research platform coding system implementing search terms erythrocytosis, polycythemia and polyglobulia. This database includes both hospitalized patients and those seen during outpatient follow-up. From all identified patients, only adult patients (≥ 18 years of age) and those who did not refuse the hospital’s informed consent were evaluated. From the cohort of patients with polyglobulia, PV patients were extracted based on the documentation in clinical history and the detection of a JAK2 mutation. All clinical and laboratory data of the included patients were manually acquired from the clinical records of the hospital database (Ipdos). Patients initially diagnosed with PV who later progressed to MF or AML were retained in the analysis. During data collection, the most recent available follow-up was documented.

Definitions

Follow-up time was defined as the period from the initial diagnosis of PV to the date of last follow-up or death. Each treatment attempt with the same class of medication was considered a single line of therapy. Data on splenomegaly were collected either from patients’ medical records or obtained through imaging studies. Bleeding events were classified as severe if they led to a significant drop in hemoglobin, caused hemodynamic instability, involved critical organs, or required hospitalization or surgical intervention.

Statistical analysis

For a univariate analysis, group differences were analyzed using the Mann-Whitney U test for continuous, and the Chi-square test for categorical variables.

The analysis was first focused on thrombotic complications, their type (arterial versus venous), location, and timing in respect to the PV diagnosis. Thereafter, we compared patients with to those without thrombotic complications. We further evaluated the outcome (Overall survival (OS) and event-free survival (EFS)) of the whole cohort as well as of subgroups (type of thrombosis, type of treatment, with or without MF transformation). For EFS analysis, the events considered were death, thrombosis occurring after diagnosis of PV and transformation to MF. The log-rank test with 2-sided significance levels was used for comparison in Kaplan-Meier estimates, and Gray’s test for comparison of cumulative incidence estimates. To calculate the cumulative incidence of transformation to MF, the occurrence of thrombosis, severe bleeding and death were considered as competing risks. OS, EFS, and cumulative incidences for subgroups were estimated at 192 months (16 years) instead of 240 months (20 years) because of the low number of individuals at risk at 240 months.

For multivariate analysis, a multinomial logistic regression was used including variables from univariate analysis with a p-value < 0.10. A backward stepwise procedure was used to eliminate non-significant variables. All p-values are two-sided with type I error rate fixed at 0.05. Statistical analyses were performed with IBM SPSS Statistics, Version 29 (IBM Corp.).

Results

In total, files of 727,731 adult patients seen in the hospital over the study period were screened for polyglobulia. Using i2b2 coding system, we initially identified 4,109 patients with the diagnosis polyglobulia. Of those patients, 1,483 met the laboratory criteria of polyglobulia based on WHO definition [2, 3]. In 401 of these patients, the result of JAK2 mutation was available, and this result was positive in 107 patients. The data from patients with JAK2 unmutated polycythemia have been published elsewhere [12]. The 107 JAK2 positive patients with polyglobulia were the basis of our study (Fig. 1).

Fig. 1.

PV, study population, selection criteria for the study

The median age at PV diagnosis was 58 years (range 18–92), 55 (51%) patients were male. The median time of follow-up was 115 months (range 1–353). The median allele burden of JAK2 at first determination was 31% (range 5.6–89). In 26 patients (24%), the PV was diagnosed before 2006, the year when JAK2 testing became available [13]; therefore, in these patients, the determination was performed at a later point than the initial diagnosis. In 38 patients, a subsequent JAK2 determination including the evaluation of allele burden was performed, the median was 19.8% (range 0–89). A longitudinal follow-up of JAK2 allele burden was available in 15 patients. The follow-up was primarily conducted in the context of treatment, mostly with interferon and, less frequently, under treatment with ruxolitinib. A reduction of the clone by more than 50% was observed in 6 of 15 patients. Notably, one patient exhibited a marked decrease in allele burden during transformation to myelofibrosis; this patient subsequently underwent transplantation and is currently alive.

Patients’ characteristics, laboratory data, and cardiovascular risk factors (CVRFs) are summarized in Table 1. Regarding CVRFs, the most frequent factor was the presence of arterial hypertension (63%), followed by dyslipidemia (34%) and diabetes (8%), 67% of the cohort had normal weight. Furthermore, 28 patients (47%) reported smoking. Sixty patients (58%) had splenomegaly at some point during the follow-up, and three patients (3%) had undergone splenectomy. Aquagenic pruritus was a symptom described in 27 patients representing 26% of the cohort.

Table 1.

Patients’ characteristics, laboratory data and cardiovascular risk factors

Number of patients		Missing
Total patients	107	0
Sex
- Male	55 (51.4%)	0
- Female	52 (48.6%)	0
Age at diagnosis (years; median, range)	58 (18–92)	0
Time of follow-up (months; median, range)	115 (1–353)	0
Hemoglobin (g/L)
- Males	179 (125–212)	7
- Females	164 (140–215)	14
Hematocrit (%)
- Males	53.5 (40–67)	10
- Females	51 (41–64)	16
MCV (fL) MCH (pg) MCHC (g/L)	83 (69–97) 27 (20–32) 326 (288–353)	39 40 40
MCV < 80 fL MCH < 27 pg	20/68 27/67
Reticulocytes (G/L)	85.5 (46–151)	68
WBC (G/L)	11.1 (5.1–26.4)	22
Platelets (G/L)	577.5 (192–2271)	21
Number of patients with platelets ≥ 550 G/L	47/86 (55%)	21
Number of patients with WBC ≥ 13.0 G/L	30/85 (35%)	22
LDH U/L	498 (173–1544)	65
Ferritin µg/L	30.5 (6–326)	67
Number of patients with ferritin < 20 µg/L	11/40 (27%)	67
EPO mU/mL	2.57 (2–28)	42
Patients with suppressed EPO (≤ 3.5 mU/mL (4–26))	48/65 (74%)	42
Spleen		4
♣ No splenomegaly	40 (39%)
♣ Splenomegaly (≥ 12 cm)	60 (58%)
♣ Splenectomy	3 (3%)
Diabetes	9 (8%)	0
Hypertension	67 (63%)	0
Weight		20
♣ Normal	58 (67%)
♣ Overweight	25 (28%)
♣ Obesity	4 (5%)
Dyslipidemia	36 (34%)	2
Smoking	28 (47%)	48
Allele burden JAK2 at diagnosis (all values)	31 (5.6–89)	54
Allele burden JAK2 – at late follow-up*	19.8 (0–89)

*Investigated in 38 patients

Across the cohort of 107 patients, 56% (N = 60) experienced thrombosis. Arterial and venous thrombosis before or at diagnosis was documented in 9 (8.4%) and 36 (33.6%) patients, respectively. In the patients who had a thrombosis before PV diagnosis, the median time to diagnosis was 67 months (range 12–265). In 13 (12.1%) patients, thrombosis occurred during PV follow-up with a median time of 66 months (range 2–187) after PV diagnosis. A total of 101 thrombotic events were documented. Arterial thrombosis was more common (58%) than venous thrombosis (42%). Cerebral thrombosis was the most frequent location (35%), followed by peripheral deep vein thrombosis (DVT) or pulmonary embolism (PE) (22%), splanchnic vein or arterial thrombosis (20%), peripheral arterial thrombosis (13%), and coronary artery thrombosis (10%), respectively. The majority of patients (57%) presented a single episode of thrombotic complication, while 20% had two episodes, 12% had three, and 8% had four episodes respectively (Table 2).

Table 2.

Clinical characteristics of all polycythemia vera patients during follow-up

Issue	Results
Patients with thrombosis	60 (56%)*
Thrombotic events
Type of first thrombosis event
♣ Venous thrombosis	25 (42%)
♣ Arterial thrombosis	35 (58%)
Timing of first thrombosis (related to diagnosis of PV)
♣ Only before diagnosis of PV	9 (15%)
♣ At diagnosis of PV	36 (60%)
♣ After diagnosis of PV	13 (22%)
♣ No information	2 (3%)
Timing of splanchnic thrombosis
♣ At diagnosis of PV	11/12 (92%)
♣ After diagnosis of PV	1/12 (8%)
♣ Missing	1
Total number of thrombotic events	101
Number of thrombotic events by patient
♣ 1 thrombosis	34 (57%)
♣ 2 thromboses	12 (20%)
♣ 3 thromboses	7 (12%)
♣ 4 thromboses	5 (8%)
♣ Undefined	2 (3%)
Bleeding complications
♣ Total number of patients with bleeding ♣ Severe bleeding	22 (21%) 11 (10%)
Severe bleeding: Total	11
Type of severe bleeding:
♣ GI bleeding	4
♣ Muscular or intraarticular bleeding	3
♣ Intracerebral bleeding	1
♣ Chronic subdural bleeding	1
♣ Intraoperative bleeding	2
Acquired von Willebrand Syndrome	17
Progression myelofibrosis/AML
♣ Progression to myelofibrosis ♣ Progression to AML (this patient had first myelofibrosis progression) ♣ Time to progression to MF (months; median range)	15/106 (14%) 1 (1%) 121.5 (1–242)
Survival data
Survival status
♣ Alive	89 (83.2%)
♣ Dead	18 (16.8%)
Details of the 18 deaths
♣ Males/Females	11/7
♣ Thrombosis (at any time)	12/18 (68%)
♣ Disease state at death:
o Progression in MF (one in AML)	7
o Not well controlled	4
o Stable	7
♣ Median age at diagnosis	66 (51–87)
♣ Median age at death (years)	85 (68–91)
♣ Median time (years) from PV diagnosis to death	13.5 (4–27)

* The percentages are calculated based on the total number of patients with thrombosis

Characteristics of PV patients with and without thrombosis at any time during the course of the disease were compared (Table 3). The most relevant finding of PV patients with thrombosis compared to those without thrombosis was older age. The median age of patients with thrombosis was 63 years (range 26–92) compared to 56 years (range 18–86) of those without thrombosis. In patients with thrombosis, 57% were older than 60 years, compared to 33% of patients without thrombosis (p=0.017). Furthermore, dyslipidemia was more frequent in patients with thrombosis (43% versus 22%; p = 0.024). Of the 60 patients who experienced thrombosis, 34 (57%) had no further events, whereas 26 experienced subsequent episodes (Table 2). At the time of the second thrombotic event, the median age was 69 years (range 36–90), 65% were over 60 years, 50% were over 70, and most (62%) were men. The second events were predominantly arterial (60%), whereas third and fourth events were mostly venous (60%). Notably, in the latter, in 3 out of 5 observed cases with venous events, these were associated with the occurrence of other cancers. CBCs were available for 18 thrombotic events, as these patients were evaluated at the hospital during the episode. The results showed that 60% had a hematocrit < 45%, WBC < 10 G/L, and 70% had platelet counts < 500 G/L, consistent with hematologic disease control. Information on the use of anticoagulants, antiplatelet therapy, as well as cytoreduction treatment at time of the thrombosis was available for only 50% of events. Thus, at the time of the second event, 75% of patients were on aspirin or anticoagulation; by the third and fourth events, all patients were receiving anticoagulation. Regarding cytoreduction therapy, 50% of patients were receiving such treatment (mainly hydroxyurea) at the time of the second event, and 70% were on some form of cytoreduction therapy during the third and fourth events. Treatment compliance was not adequately documented, and no consistent information was available on JAK2 allele burden at the time of the thrombotic events.

Table 3.

Comparison of PV patients with and without thrombosis including all patients with thrombosis: univariate analysis

Variable	With thrombosis	No thrombosis	p-value	Missing
Number	60	47	n.a.	0
Females (number; %)	30/52 (58%)	22/52 (42%)	0.743	0
Males (number; %)	30/55 (55%)	25/55 (45%)
Median Age at PV (years; median, range)	63 (26–92)	56 (18–86)	0.234	0
Median follow-up (months; median, range)	110 (1–353)	143 (3–348)	0.299	0
Age at diagnosis ≥ 60 years	34/60 (57%)	15/45 (33%)	0.017
Median Hb (median, range)	170 (141–212)	173 (125–215)	0.346	21
Median Hct (range)	51 (41–66)	52 (40–67)	0.422	26
Median WBC (range)	11.0 (6.28–26.4)	11.3 (5.1–22.9)	0.912	20
Median PLT (range)	580 (192–1564)	555 (240–2271)	0.703	21
Patients PLT > 550 G/L	26/47 (55%)	21/39 (45%)	0.891	21
Pat with Hb > 170 g/L	25/47 (53%)	22/39 (56%)	0.765	21
Pat with Hct > 49%	35/46 (76%)	29/35 (83%)	0.459	26
EPO level mU/mL	2.57 (0.6–20)	2.51 (0.2–28)	0.852	42
Splenomegaly	32/57 (56%)	28/43 (65%)	0.364	7
Diabetes	5/60 (8%)	4/45 (9%)	0.947	1
Hypertension	37/58 (64%)	30/47 (64%)	0.997	2
Overweight/Obesity	18/54 (33%)	11/33 (33%)	1.000	20
Dyslipidemia	26/60 (43%)	10/45 (22%)	0.024	2
Smoking	15/35 (43%)	13/24 (54%)	0.393	38
Alive	48/60 (80%)	41/47 (87%)	0.321	0

We focused on 13 PV patients with splanchnic thrombosis, seven were male. Of these, 11 patients (92%) were diagnosed at the time of PV diagnosis, while only in one patient was diagnosed later, in one case this information was missing. Comparison between patients with splanchnic thrombosis to those with other types of thrombosis or no thrombosis revealed significant differences. Patients with splanchnic thrombosis were significantly younger (median age 44 years; range 26–78), compared to patients with another type of thrombosis (median age 65 years; range 29–92; p = 0.003) (Table 1S). All patients with splanchnic thrombosis had splenomegaly compared to 43% of those with other types of thrombosis (p = 0.0003) and 65% of those without thrombosis (p = 0.012).

Bleeding complications were documented in 22 (21%) of all patients. Among them, eleven experienced severe bleeding, with gastrointestinal bleeding being the most commonly reported (Table 2).

We also analyzed the type and the number of changes of PV treatment over the follow-up: 31 patients (29%) received only one-line treatment, 44 patients (42%) a second-line treatment, 20 patients (19%) a third-line treatment and only 11 patients (10%) a fourth-line therapy or more (Table 2S). A total of 31 patients received treatment with ruxolitinib.

Regarding the monitoring of JAK2 allele burden, we observed that 15 patients in this series had longitudinal follow-up. These patients were mainly managed with interferon or ruxolitinib, or the test was performed to assess disease progression. In six patients, a reduction in clone size of more than 50% was observed. Notably, one patient showed a marked decrease in allele burden in the context of transformation to myelofibrosis; this patient subsequently underwent transplantation and is currently alive.

Regarding follow-up, 15 patients (14%) experienced disease progression to MF, with a median time to progression of 121.5 months (range 1–242). Only one patient, who previously progressed to MF, underwent later a leukemic transformation (Table 2). Patients aged 60 years and older at the time of PV diagnosis exhibited a higher incidence of transformation to MF, although the difference was not statistically significant. The overall cumulative incidence of progression to MF at 240 months (20 years) was 25% (95% CI 15–42%) (Fig. 2a). The cumulative incidence of severe bleeding at 192 months was significantly higher in PV patients with MF progression (56%, 95% CI 28–100%), compared to PV patients without progression to MF (15%, 95% CI 7–30%, p = 0.004) (Fig. 2b). PV patients with progression to MF had also higher cumulative incidence of thrombosis, but this difference was not significant.

Fig. 2.

Cumulative incidence of progression to myelofibrosis in PV patients at 20 years, stratified by severe bleeding events (a) Overall cumulative incidence of progression to myelofibrosis (MF) in PV patients at 240 months (20 years) (b) Cumulative incidence for progression to myelofibrosis, comparing patients with and without severe bleeding

The OS of PV patients at 240 months (20 years) was 63% (95% CI 46–79%) (Fig. 3a). The EFS at 240 months was 46% (95% CI 31–61%) (Fig. 3b). The OS rates by type of thrombosis at 192 months (16 years) were 90% (95% CI 78–100%) for patients without thrombosis, 92% (95% CI 77–100%) for patients with venous thrombosis, and 57% (95% CI 28–85%) for patients with arterial thrombosis (p = 0.007). The OS rate difference for patients with arterial thrombosis was significant compared to patients with venous thrombosis (p = 0.01) and those without thrombosis (p = 0.007). The difference between patients with arterial and venous thrombosis may be partly explained by a borderline, though not significant, older age in patients with arterial thrombosis (Mann-Whitney U test, p = 0.062).

Fig. 3.

Outcome of PV patients (a) Overall survival (OS) of all 107 PV patients at 240 months (20 years) with 95% CI: 63% (95% CI 46–79%) (b) Event-free survival (EFS) of all PV patients at 240 months (20 years) with 95% CI: 46% (95% CI 31–61%)

The OS at 192 months did not differ for patients with or without bleeding (p = 0.989). There was a non-significant trend (p = 0.078) of worse OS at 192 months for patients with transformation to MF (61%, 95% CI 31–92%) compared to patients without transformation (88%, 95% CI 78–97%). A large number of patients with transformation to MF could possibly lead to a significant difference.

In the multivariate analysis of the relative risk of death, two factors were statistically significant: age as a continuous variable (relative risk (RR) 1.052, 95% CI 1.009–1.097, p = 0.033), and transformation to MF (RR 6.856, 95% CI 1.948–24.120, p = 0.0027). Regarding the relative risk of thrombosis, significant factors included dyslipidemia (RR 2.573, 95% CI 1.058–6.257, p = 0.037) and older age at diagnosis (reference > 60 years) (RR 2.524; 95% CI 1.111–5.734, p = 0.026).

During the observation period, there were 18 deaths reported, 11 were male and the median age at death was 85 years (range 68–95). At the time of death, seven patients had progressed to MF, 3 patients had poorly controlled PV, and the remaining patients had either stable or well-controlled disease. A clear relationship between death and PV was established in only one case: a post PV-MF patient who progressed to AML and died during therapy-induced aplasia. Another case was possibly related to the PV, involving a fatal arterial cardiovascular complication. Eight deaths were likely unrelated to MPN attributed to advanced age, cardiovascular disease, COVID-19, or other factors. For 8 out of 18 deceased patients, no information regarding the cause of death was available (Table 3S).

Discussion

Our real-world study, featuring a well-characterized cohort of JAK2 positive PV patients with a median follow-up of nearly 10 years, provides valuable insights into the long-term thrombotic risk associated with this disease as well as other important outcomes such as disease progression, bleeding complications, and OS. The methodology used, i2b2, identifies patients within a comprehensive hospital database that includes all patients seen at the institution. This approach differs from other retrospective studies, which typically rely on patient-reported data rather than direct patient identification and data utilization. This methodology may partly account for some of the differences observed when compared to other publications. In this cohort, we observed a slight male predominance (51.4%) which is in line with the findings of other studies [9, 14–16]. The median age at diagnosis was 58 years, slightly younger compared to other studies where the median age was 60 years or older [9, 14, 15, 17]. Notably, the median follow-up time was 115 months which was substantially [14, 18, 19] or slightly longer [1] than those reported in other studies allowing for a comprehensive assessment of the long-term risks and complications associated with PV.

The high prevalence of CVRFs in our cohort, particularly arterial hypertension (63%), followed by dyslipidemia and diabetes underscores the critical importance of the need of their management in PV patients. The predominance of hypertension aligns with findings from other studies, where it has been identified as the most common CVRF in PV patients [9, 15, 20, 21]. Other authors have assessed the impact of arterial hypertension on thrombosis in low-risk PV patients and found that arterial hypertension plays a significant prognostic role in the incidence of arterial thrombosis among patients conventionally classified as low-risk for future vascular complications [22]. As in other published series, dyslipidemia was a common finding in our cohort and was directly associated with the occurrence of thrombosis in PV [9, 20, 21]. Similar to other series, we found that PV patients with thrombosis were older (57% over 60 years vs. 33%, p=0.017), and had a higher prevalence of dyslipidemia compared to those without thrombosis (43% vs. 22%, p=0.024) [23]. Studies from Japan and Brazil have demonstrated that CVRFs such as diabetes, hypertension, and hyperlipidemia significantly elevate the risk of thromboembolic events in patients with PV [24, 25]. A recent study evaluating 739 PV patients found that 78.1% had at least one CVRF. Multivariate analysis showed that relevant CVRFs, along with inadequate Hct control, progressive splenomegaly, and previous thrombosis, are significant independent predictors of thrombotic risk [19]. Although there are no data on the clinical effect of early interventions to correct CVRFs in PV patients, various strategies have been discussed in the literature [26]. Addressing both age-related risks and CVRFs is critical for reducing the thrombotic burden in patients with PV.

The overall thrombosis rate in our cohort was 56%, notably higher than reported in the Swedish registry (16%) [27], by Barbui et al. in 2014 (21%) [28] the ECLAP study (38%) [29] and Cyto-PV study (29%) [9]. The higher thrombosis rate and consequently, a high proportion of high-risk patients may reflect our longer follow-up and the implemented methodology in this study, which captures cases directly from the database rather than reporting as in formal registries. Furthermore, being a tertiary center likely introduces a bias toward high-risk patients, as uncomplicated PV is usually managed in the community. Additionally, the relatively small size of our cohort is a limitation. The Spanish registry analyzed data from PV patients with low-risk, and this particularly younger population, managed with phlebotomies, showed thrombosis in 4.19% of patients, demonstrating that even thrombotic risk in low-risk patients should not be neglected [30].

Furthermore, the timing and clinical patterns of thrombotic events in our study provide important insights into the course of PV. Thromboses were most frequently observed before or at the time of diagnosis affecting 42% of our cohort. Data from a large cohort of 1,545 patients with PV, published by the International Working Group for Myeloproliferative Neoplasms Research and Treatment (IWG-MRT), showed that thrombosis before or at diagnosis occurred in 360 (23.4%) patients [31]. Barbui et al. reported a comparable thrombotic event rate at 56%, however, during follow-up they showed a thrombotic rate of 44%, indeed notably higher compared with our findings [17]. The observed differences may be attributed that our cohort was evaluated more recently, during a period of increased awareness regarding disease management to prevent complications.

Our study demonstrated a predominance of arterial thromboses (58%) over venous thromboses (42%), a finding consistent with the results of Landolfi et al. and Griesshammer et al. [32, 33]. The higher prevalence of arterial events highlights the intrinsic predisposition of PV patients to arterial thrombosis, potentially driven by erythrocytosis-induced hyperviscosity, platelets activation, endothelial dysfunction, leukocytosis and the detrimental synergism of the association with CVRFs [34–36].

The proportion of splanchnic thrombosis among all venous thromboembolic events is disproportionately higher in patients with MPN compared to the general population, accounting for an estimated 7.5% of total venous thromboembolism (VTE) cases [37]. This elevated incidence was also observed in our cohort, where 20% of patients who experienced thrombotic events presented with this specific type of complication. Importantly, these patients were significantly younger than those suffering from other thrombotic manifestations. Of the 13 patients with splanchnic thrombosis, 92% had it as the initial presenting event leading to the diagnosis of PV. This observation aligns with findings from Kiladjian et al., who also noted that PV patients experiencing splanchnic thromboses were younger than those with other thrombotic complications [38]. This finding highlights the importance of recognizing this complication early, especially in younger patients, as it may be the first clinical indication of underlying PV.

A large prospective study showed that combining aspirin with anticoagulants significantly increases the risk of bleeding in patients with PV compared to aspirin alone. Additionally, high platelet counts were associated with a greater bleeding risk. In that study, 69 patients experienced bleeding events, with 25 classified as severe, representing 4.3% and 1.5% of the cohort, respectively, over a median follow-up of 2.4 years [39]. The higher rates of severe bleeding observed in our cohort are likely related to the longer observation period. 10% of patients receiving combined treatment with anticoagulation and aspirin experienced bleeding complications.

The performed multivariate analysis in our study revealed that advancing age and progression to MF significantly increase the risk of death in PV patients. Additionally, dyslipidemia and being diagnosed after age 60 were associated with a higher risk of thrombosis, underscoring the importance of managing CVRFs alongside disease control. These findings align with those reported by other authors describing similar observations. A large meta-analysis demonstrated the influence of age on the occurrence of thrombosis during the follow-up of PV patients, even under cytoreductive treatment [40]. Schafer et al. emphasize that 10–20% of PV patients are at risk of transitioning to MF over time, with the process driven by molecular mechanisms, particularly the presence and burden of JAK2 mutations. The same study highlights that the progression of PV to MF is influenced by chronic accumulation of genetic alterations and their effects on hematopoietic stem cells [41]. Mirroring these results, the median time to MF progression of 121.5 months observed in our cohort indicates a slow disease evolution towards MF. This supports the understanding that transformation is not an early event, but rather a consequence of progressive molecular and cellular stress over time, reinforcing the need for long-term disease. Additionally, a recent Swedish registry study reported significantly increased thrombotic events and all-cause mortality in PV patients, particularly those over 50 years of age, emphasizing the heightened vulnerability of this group [27]. Similarly, the observed association between arterial thrombosis and poorer survival outcomes in our study underscores the serious impact of arterial events on prognosis. However, our findings may be interpreted with a certain degree of caution due to potential confounding factors such as differences in patient age and comorbidities, which were not fully controlled for in the analysis. The REVEAL study, the largest prospective cohort of PV patients in the US, revealed a mortality rate exceeding 10% within four years. Patients who died were more likely to have high-risk disease features, including older age and history of thrombosis, as well as higher rates of comorbidities such as cardiac and respiratory disorders. Thrombotic complications accounted for approximately one-third of deaths, and many patients exhibited signs of uncontrolled myeloproliferation shortly before death [42].

Our study has several limitations and strengths. Its retrospective design and the relatively small number of patients in some subgroups limit the generalizability of certain findings. In addition, treatment heterogeneity across the cohort precluded a detailed analysis of therapeutic impact on outcomes. Diagnostic decisions and treatment initiation were made at the discretion of the treating physician, introducing potential variability. Although this cohort contains data on JAK2 allele burden, it was insufficient to establish a correlation with thrombosis risk, as half of the patients have follow-up exceeding 10 years, and this information was therefore unavailable. However, the study’s strengths include a large number of screened patients and comprehensive access to laboratory, imaging, and clinical data from medical records. The use of standardized coding systems such as i2b2 can transform heterogeneous clinical data into AI-ready formats, potentially enhancing the reliability and reproducibility of studies such as ours. This approach allowed for an in-depth, individualized assessment and helped ensure consistency while minimizing underreporting. Notably, 50% of the cohort had over 10 years of follow-up, providing substantial long-term data relevant to the study’s objectives.

In conclusion, our real-world study provides valuable long-term data that enable a comprehensive assessment of the clinical course and complications of PV. The findings emphasize the importance of managing CVRFs, the relevance of early recognition of thrombotic events, particularly splanchnic thrombosis, and the need for monitoring for disease progression. However, given the limited sample size, the results should be interpreted with caution. Despite its limitations, the study highlights key prognostic indicators and reinforces the need for individualized, long-term management strategies to reduce complications and improve outcomes in PV patients.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Natalie Neumark, Katarzyna Aleksandra Jalowiec, and Alicia Rovó conceived and designed the study. Natalie Neumark and Alicia Rovó had full access to all data, performed data curation, analyzed the data, and wrote the original draft of the manuscript. André Tichelli conducted the statistical analysis. Katarzyna Aleksandra Jalowiec, André Tichelli, Naomi Porret, Linet Njue, Kristina Vrotniakaite-Bajarciene, Raja Prince-Eladnani, Anne Angelillo-Scherrer, and Sara C. Meyer analyzed the data, and reviewed and edited the manuscript. All authors have read and approved the final version of the manuscript.

Funding sources

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Competing interests

The authors declare no competing interests.