High pulmonary hypertension risk by echocardiogram shortens survival in polycythemia vera

Polycythemia vera (PV) is an insidious and progressive myeloproliferative neoplasm (MPN) characterized by increased red blood cell mass that is virtually always attributed to a Janus kinase 2 (JAK2) driver mutation. Both the inflammatory consequences of JAK2 signaling activation and the hyperviscosity associated with erythrocytosis are mechanisms that predispose patients with PV to thrombosis. The clinical presentation of PV varies, but in addition to laboratory abnormalities; patients may have thrombosis history; cardiovascular disease; splenomegaly; and a multitude of constitutional symptoms including fatigue, early satiety, and pruritis. These symptoms overlap with many other systemic conditions and patients and their physicians often have poor insight into the cause of these symptoms, which can delay workup and treatment.¹ Furthermore, MPNs can be complicated by pulmonary hypertension (PH), although the reported prevalence varies greatly.^2-4 

PH is a progressive, morbid, and deadly condition characterized by elevated pulmonary artery pressures, which, in turn, causes right heart failure, if left unchecked. PH is marked by symptoms such as dyspnea, chest pain, presyncope, lower extremity edema, and profound fatigue, many of which overlap with the clinical presentation of MPN. The World Health Organization (WHO) subclassifies PH according to the underlying pathophysiologic mechanisms. PH due to MPN is defined within WHO group 5: PH due to various chronic diseases but with poorly characterized or understood mechanisms. Furthermore, the mechanisms of PH can hemodynamically vary between precapillary, postcapillary, and combined postcapillary and precapillary PH. Experts agree that the diagnosis of PH in the setting of MPN should follow the defined PH workup algorithms including referral to, and treatment at, an expert center. When initial screening based upon history, physical, vital signs, electrocardiogram, and basic laboratory work such as a β-type natriuretic peptide raises suspicion for PH or cardiac disease, echocardiography (ECHO)-based screening is warranted, possibly followed by cardiopulmonary stress test. Patients with increased risk of PH should be referred to a PH specialized center for comprehensive PH evaluation, often including right heart catheterization (RHC). In practice, this is seldom done because the overlap in presenting syndromes of PV and PH obscures attribution of symptoms and can delay or prevent formal PH evaluation.⁵ 

PV is thought to have a lower prevalence of PH than primary and secondary myelofibrosis (MF).^3,6 Still, some studies report a prevalence of PH in PV as high as 55% as measured by ECHO-derived estimates of right ventricular (RV) systolic pressure of ≥35.⁷ It should be noted that although RHC is the gold standard diagnostic modality, the prevailing literature in group 5 PH uses ECHO derived tricuspid regurgitant jet velocity (TRV) as a surrogate marker of PH risk because of the high correlation between ECHO estimated and invasively measured pulmonary artery pressures, and the relative ease of obtaining an ECHO compared with RHC. The general population of patients with PV have a median survival of at least 14 years but more recent analyses of well-controlled patients receiving available therapy suggest survival of >20 years.⁸ But of course, this means that half of patients die much earlier. In other chronic diseases (eg, sickle cell disease and Sjögren disease) associated with higher risk of PH, survival is dramatically reduced in those who develop PH and this has motivated the development of guidelines for routine screening for PH by ECHO in these populations.^5,9,10 No such recommendations are available for those with PV. It is thus critical that we examine the relationship between mortality and PH in PV to better identify those patients that may benefit from comanagement with pulmonary vascular disease specialists and consideration of advanced pulmonary vascular therapies. Therefore, the goal of this study was to evaluate the longitudinal effects of PH on mortality in patients with PV.

Our cohort was derived from the Weill Cornell Medicine (WCM) research data repository of the Silver MPN Center. All patients with a diagnosis of PV who underwent ECHO between February 2011 and April 2024 were included. PV diagnosis was established by either Polycythemia Vera Study Group (1974-2007), WCM (2008-2016), or WHO (2016-2023) criteria depending on the date of diagnosis. Patients were identified by International Classification of Diseases 9th and 10th Revision (ICD-9 and ICD-10, respectively) codes and confirmed by manual review, as previously reported.¹¹ All echocardiograms performed at WCM were available and included for review. Patients were excluded if echocardiography was only performed before MPN diagnosis, or after progression to MF or leukemia. Data were censored at last follow-up (n = 5) if death status could not be confirmed by either chart review or the National Death Index Registry as of 31 December 2021.

Our primary exposure of interest was the longitudinal effects of PH. Given the paucity of RHC data in patients with PV and the importance of TRV in the most recent European Respiratory Society/European Society of Cardiology guidelines for PH risk stratification, we used ECHO-derived TRV to assess PH risk.⁵ Our primary outcome was overall survival (OS). Additional analysis was then performed based on a stratification of ECHO-based PH risk severity (normal PH risk vs mild PH risk vs high PH risk) and 10-year mortality. Secondary outcomes of interest included incidence of progression to MF and incidence of thrombosis. Lastly, we performed an exploratory analysis to evaluate the optimal age and duration of MPN that would guide screening for PH.

Baseline data including age, sex, date of MPN diagnosis, death status, date of progression (if applicable), comorbid medical conditions, laboratory values, pathology, and medications, were collected in a standardized manner from the electronic medical record and extracted from the research data repository. Echocardiographic data were systemically collected and manually reviewed for accuracy. To examine the longitudinal effect of ECHO PH risk on outcomes in this cohort, the first ECHO after MPN diagnosis was considered the start of follow-up in patients with multiple ECHO studies. Mild PH risk was defined as TRV of >2.8 but ≤3.4 m/s, whereas high PH risk was defined as TRV of >3.4 m/s. In binary analysis, increased PH risk was defined as TRV of >2.8 m/s. If TRV could not be calculated, manual review of 2-dimensional ECHO measurements was undertaken. If both right atrium and RV size were calculated to be normal, the patient was coded as normal PH risk. However, if neither right atrium nor right ventricle was considered abnormal but the TRV was missing, the patient was excluded from the study as unevaluable.¹² 

To ensure that outcomes of patient with an ECHO are not subject to selection bias, we compared their outcomes to a propensity score–matched cohort of patients with PV with similar epidemiologic characteristics who did not undergo an ECHO. We performed a 2:1 nearest neighbor propensity score match with age at PV diagnosis as the covariate to balance and performed follow-up testing of match quality as with the initial imbalance calculations.

The primary outcome was modeled with Kaplan-Meier survival analysis and significance assessed with log-rank test. The secondary outcomes were assessed by competing risk analysis using the Fine-Gray method. To estimate the effect of PH risk on both primary and secondary outcomes, multivariable Cox models were developed using age and history of thrombosis as covariates and PH risk severity as the stratifying variable of interest.¹³ Study duration was calculated as time from ECHO to either last follow-up or death date.

To identify thresholds for age and MPN duration that could predict increased PH risk by ECHO, we performed univariate and multivariate logistic regressions followed by receiver operator curve (ROC) analysis and Youden index analysis for each model. Age and MPN duration were binarized based on identified thresholds in the univariate regressions with predicted probability above the previously identified threshold for PH. Subsequently, a multivariable logistic regression was performed using the 2 binarized variables to predict PH, accompanied by a ROC analysis.

Baseline demographics were compared using Kruskal-Wallis and Wilcoxon rank-sum tests for continuous data, and χ² test and Fisher exact test for categorical data. Analysis was conducted using RStudio version 6.1 build 524 (Posit software, Boston, MA). This study was approved by the WCM institutional review board (no. 19-12021151).

The ECHO cohort was comprised of all 134 patients with PV who underwent ECHO during the study period. A propensity score–matched population of 266 patients with PV who did not receive ECHO cohort was also identified. A median of 2 ECHO studies were performed per patient during the study period. The study included 71 females (53%) and 63 males (47%). There was a history of either arterial or venous thrombosis in 49 patients (37%). Relevant additional medical history included systemic hypertension in 77 (57%), diabetes in 19 (14%), hyperlipidemia in 63 (47%), coronary artery disease in 26 (19%), and obstructive sleep apnea in 13 (10%), and median body mass index was 25. No patients carried a diagnosis of pulmonary parenchymal disease. Treatment for PV at time of ECHO varied substantially but primarily included antiplatelet, anticoagulant, and cytoreductive therapy. Median age at ECHO was 73 years whereas the median duration of MPN disease before ECHO was 4.5 years. TRV was calculated in 99 patients (74%%); 24 (18%) met criteria for PH risk by ECHO, and 8 (6.0%) met criteria for high PH risk.

Clinical characteristics and epidemiologic data stratified based on the severity of PH risk are reported in Table 1. Patients with increased PH risk were older than those with normal PH risk (median age, 78 vs 70 years; P < .001) and had significantly longer duration of PV before ECHO (median duration, 11 vs 3 years; P = .002). There was no statistically significant difference between high PH risk, mild PH risk, and normal PH risk groups with respect to sex, history of thrombosis, and other historical medical factors (Table 1). There was a nonsignificant trend toward higher New York Heart Associate functional classification in patients with increased PH risk. There were differences across pharmacologic therapy groups including an increased prevalence of anagrelide (n = 5) among those with increased PH risk by ECHO and decreased prevalence of interferon use (n = 26) in the PH risk group. Increased PH risk by ECHO was associated with signs of diastolic and RV dysfunction, particularly in those at high PH risk (Table 1). Death during study period was twice as common in patients with increased PH risk than in those with normal PH risk (67% vs 33%, P = .001). Kaplan-Meier estimates of OS were identical for the ECHO cohort and the cohort that did not receive ECHO (10-year OS of 84% vs 81%, respectively; P = .51), indicating that the ECHO cohort was not comprised of patients who were intrinsically sicker than the general population of patients with PV at WCM (supplemental Figure 1). Aggregate data for the 16 patients who underwent RHC during the study period are reported in Table 2. Median mean pulmonary pressure was 25 mmHg, median pulmonary capillary wedge pressure was 14 mmHg, and median pulmonary vascular resistance was 2.2 Wood units.

When Kaplan-Meier survival curves were stratified by binary PH risk (increased vs normal), 10-year OS was higher for patients with normal PH risk than those with increased PH risk (57% vs 18%; P = .0024; Figure 1). When stratified by PH risk tertiles, 10-year survival for patients with normal PH risk, with mild PH risk, and high PH risk was 57%, <21%, and <25% respectively (P = .003; Figure 2). Median survival was 3.7 years for those with mild PH risk, 1.8 years for those with high PH risk, and not yet reached for patients with normal PH risk.

Cox proportional hazards analysis showed that increased PH Risk of any degree was significantly associated with mortality in univariate analysis (hazard ratio [HR], 2.5; 95% confidence interval [CI], 1.4-4.6) and in a multivariable model that included the standard PV risk variables of age and thrombosis history (PH HR, 1.9; CI, 1.0-3.5), indicating that ECHO-based PH risk assessment independently predicts survival in PV. Multivariate Cox-model analysis considering age at ECHO, thrombosis history, and tertiles of PH risk (normal, mild, and high risk), identified high PH risk as independently associated with mortality (HR, 3.3; CI, 1.4-8.0; Table 3).

The proportion of patients who progressed to MF during the study follow-up was higher for those with increased PH risk than those with normal PH risk (Table 1). Although median cumulative incidence of progression to MF was not reached in any group, cumulative incidence at 10 years was significantly higher in patients with mild PH risk than those with normal PH risk (25% vs 3.6%, P = .005, Figure 3). In a multivariable Cox-model, mild PH risk was independently associated with progression to MF during the follow-up period (HR, 6.9; CI, 1.3-36). There was a significant association between high PH risk and progression, but this association was characterized by high standard error and a wide CI, likely owing to the significantly shortened OS in this group. Logistic regression was used to calculate the odds ratio of progression to MF by PH risk tertile and there was a nonsignificant trend toward greater likelihood of progression to MF in the high PH risk group (supplemental Figure 2). There was no difference in cumulative incidence of thrombosis among PH risk tertiles (Figure 4). Multivariable Cox-model analysis showed prior history of thrombosis was not significantly associated with incident thrombosis during the follow-up period (HR, 2.2; CI, 0.72-6.5).

We identified significant differences among epidemiologic data between patients with normal and elevated PH risk by ECHO. Age at ECHO and duration of PV before ECHO were statistically significantly higher in the elevated PH risk group. Univariate logistic regression followed by ROC analysis on each individual variable identified optimal probability of PH risk thresholds of 0.17 for age at ECHO and 0.18 for duration of PV. From these threshold probabilities, we calculated the optimal age and PV duration cutoffs for PH screening to be 72 years old and 8.7 years, respectively. Both variables were then binarized based on simplified thresholds of 70 years of age and 8 years duration of PV, and subsequent multivariable logistic regression was performed to compare the ability of these 2 variables to predict PH. Area under the ROC based on this multivariable logistic regression was calculated to be 0.79 (Figure 5). The resultant odds ratios from exponentiated regression coefficients were 6.9 for age at ECHO and 5.4 for duration of PV.

Our study finds that presence of increased PH risk by ECHO was associated with shortened OS in patients with PV. Identification of increased PH risk in this population was linked to duration of PV diagnosis, and less strongly to patient age, in concordance with previous metadata analysis.³ Approximately one-fifth of patients in our cohort had some degree of increased PH risk by ECHO and 6.0% had high PH risk. We found that high PH risk was associated with significantly shortened OS in patients with PV. Although mild PH risk was associated with shortened OS in univariate analysis, this relationship was not statistically significant when considering factors predisposing to mortality in MPN such as older age and thrombosis history.

Development of PH microthrombosis and progression of thrombosis to chronic thromboembolic PH (CTEPH) would seem, at first glance, to represent the most logical pathophysiologic mechanism for development of PH in patients with PV. However, direct testing identified no association between thrombosis history and development of increased PH risk. Similarly, neither PH risk tertile, nor patient age was linked to the cumulative incidence of thrombosis during the follow-up period. Despite its importance in assessing future thrombosis risk in PV, we did not find thrombosis history predictive of OS in our cohort.¹⁴ Similarly, multivariable models identified no association between thrombosis history and OS when considering age at ECHO and PH risk tertile (normal PH risk, mild PH risk, and high PH risk). A significant number of those with increased PH risk in our cohort also had ECHO findings consistent with postcapillary pathophysiology (increased left atrial [LA] dimension and left ventricular [LV] diastolic dysfunction). Of 8 ventilation-perfusion scans performed among this cohort, none were consistent with CTEPH. These results suggest that ECHO-based PH risk in PV is largely driven by mechanisms other than thrombosis, such as increased blood viscosity, diastolic dysfunction, and inflammatory changes, and that the excess mortality because of PH risk in PV is not from thrombotic events and CTEPH.

Our study uncovered an unexpected link between PH risk by ECHO and progression to MF. Multivariate analysis confirmed this association, suggesting that factors driving PH risk may overlap those that contribute to MF progression (supplemental Figure 2). Although primary MF is known to have a higher prevalence of PH than PV, our findings point to the possibility that inflammatory mediators or other factors associated marrow fibrosis may also play a role in the vascular and cardiac remodeling responsible for development of PH.^15,16 PH is also associated with nonmalignant marrow fibrosis, further suggesting a shared biology of PH and marrow fibrosis.^17,18 Certainly, the inflammatory milieu of patients with PV, MF and post-PV MF points to a possible root cause of this complex association. Although it would be premature to consider PH an early marker of MF transformation, our findings highlight the need for further research to identify the mechanisms linking PH and MF and to determine whether intervention could mitigate risk of both PH and MF progression in PV.

The link between PH and shortened survival is well established in virtually all settings. Even among patients with group 5 PH--causes of PH that are the least well understood such as end-stage renal disease (ESRD), sickle cell disease, and sarcoidosis--PH is known to shorten survival.^19,20 Still, prior studies disagree on the true effect of PH on survival in MPNs.^3,21-23 One driver of this controversy is the heterogeneity of cohorts and methods used to assess risk in these prior reports and lack of consideration of unique MPN subtypes and their pathophysiologic consequences.²⁴ Our study helps fill the gap by studying a deeply annotated cohort of patients with PV, allowing direct comparison of outcomes in patients with PV with and without elevated PH risk.

Despite the high prevalence of PH in PV, the pathophysiology is not understood. Poor understanding of outcomes related to PH highlights the need for better diagnostic algorithms to drive appropriate therapy. Treatment of PH due to PV with pulmonary vasodilator medications is not standard, although it has been reported in the literature to be efficacious in some patients.²⁵ Both diastolic and high output heart failure can passively increase pulmonary artery pressures and are collectively known as postcapillary causes of PH.^4,26 Patients with PV in our cohort had a high prevalence of systemic hypertension, which can lead to both diastolic and systolic heart failure, left atrial hypertension, and accompanying postcapillary PH.²⁷ Excess thrombotic risk is at the center of PV risk stratification, so it is not surprising that pulmonary thrombus formation has been reported in patients with PV and PH.^28,29 However, in our well-controlled study, we could not identify a link between thrombosis and increased PH risk in our cohort of patients with otherwise well controlled PV.^24,25 Although less common than postcapillary pathophysiology in our cohort, a significant subset of patients had features of precapillary PH, several confirmed by RHC. It is possible that PV related mechanisms driven by JAK2^V61F signaling result in hyperproliferative and hyperinflammatory consequences may increase pulmonary vascular resistance and contribute to development this process.^30,31 

For group 5 PH, the mantra among cardiologists and pulmonologists is to treat PH by treating the underlying disease process, but this has not always yielded positive clinical outcomes in MPNs.² The lack of routine referral algorithms for cardiopulmonary signs and symptoms may be to blame for the dearth of outcomes data to inform PH therapy. Routine screening for PH by ECHO is recommended in other diseases with known risk for PH such as sickle cell disease, systemic sclerosis, and sarcoidosis.⁵ The prevalence of PH in PV appears to be at least as high as it is in these other disorders, and the significantly reduced OS in those with PH suggests that routine screening for PH should be considered for patients with PV. Our study identified age and duration of PV as important covariates associated with PH, and ROC–area under the curve analysis identified age of >72 years and PV duration of >8.7 years as the optimal cutoffs above which the yield of ECHO diagnosis of PH is maximized (Figure 5; supplemental Figures 2-6). We believe that these data justify and provide a rational basis for recommending ECHO-based screening for patients with PV if any of the following are present: cardiopulmonary symptoms (unexplained dyspnea, chest/abdominal pain, presyncope/syncope, palpitations, cough, exercise intolerance, lower extremity edema, or profound fatigue; also see supplemental Table 1), age of >70 years, or duration of PV of >8 years. These simplified cutoffs showed acceptable performance based on an area under the ROC of 0.79, and we recommend patients with PV presenting with suggestive symptoms or exceeding these temporal cutoffs be screened by ECHO for PH and, when risk is elevated, appropriate referral is made for specialized PH evaluation and care.

A high index of suspicion is necessary to ensure early detection of PH. It is our hope that more ubiquitous and standardized screening will enable progress in altering the grim prognosis for these patients. In parallel with this, new tools are needed to identify subsets of patients with PV who have a molecular phenotype or pathophysiology amenable to targeted PH therapies such as sotatercept whose therapeutic pathway interacts with those of PV.³² Although ECHO is an invaluable tool for the preliminary assessment of PH and we have found high concordance between ECHO-estimated pulmonary artery systolic pressure (PASP) and PASP measured by RHC in our cohort, patients with increased PH risk by ECHO should be referred for evaluation including possible RHC to establish diagnosis, pathophysiology, and responsiveness to PH therapy. We recommend that patients with ECHO findings suggestive of PH be referred to PH specialists, as outcomes are improved with specialized care.

To our knowledge, this is the largest PV-specific study examining the association between diagnosis of secondary PH risk by ECHO and mortality. Our study was limited by the diversity of therapies used for treatment of PV, which reduced the power to draw conclusions regarding the effects of particular drugs on likelihood of increased PH risk. Gold-standard RHC measurements were performed only in a subset of our patients because of older age, the invasive risks of the procedure, and variation in interpractitioner attention to PH. Future development of screening, referral, and evaluation guidelines, based upon those already present for other types of PH, is greatly needed. Greater protocolization of PH screening will lead to more definitive diagnostics with RHC and thus a greater number of patients eligible for pulmonary vasodilators, small molecule inhibitors, or future disease modifying therapies. RHC is helpful both to define the pathophysiologic mechanisms of PH as well as to test and guide therapy for PH. Further studies, perhaps using metabolomic and proteomic studies of serum and pulmonary artery catheter samples will be useful to further establish a common mechanistic pathway for the development of a PAH-like precapillary PH endotype in PV. Such studies are required to improve the overall understanding of the mechanisms for precapillary/postcapillary PH in patients with PV and for development/selection of targeted treatments for PH in PV. Therefore, by enhancing early detection and awareness of PV-related PH Risk, we can promote timely intervention with current therapies, while also stimulating research into new treatments that target both pulmonary and hematologic complications.

Although most patients with PV achieve favorable long-term survival with standard therapies, secondary complications such as PH can adversely affect morbidity and mortality. Our study highlights the high prevalence of increased PH risk by ECHO in patients with PV and demonstrates that patients with high PH risk have significantly shortened survival. These findings underscore the importance of early detection and awareness of PH risk in PV, as well as the need for structured screening and referral protocols. ECHO-based PH risk assessment provides a practical tool for identifying patients at increased risk, enabling timely referral for specialized care. Further research into the mechanisms linking PH and MF progression, along with the development of targeted therapies, will be critical to improving outcomes for patients with PV and increased PH risk.

The authors acknowledge the Cancer Research & Treatment Fund (CR&T), the Myeloproliferative Neoplasms Research Foundation (MPN-RF) and MPN Peoria (JMS); and the National Center for Advancing Translational Sciences (NCATS) grant UL1 TR002384 to the Clinical and Translational Science Center (CTSC) of the Weill Medical College of Cornell University.

Contribution: A.J.G. performed manual review of data, compiled the finalized data set, analyzed data, and wrote the manuscript; D.P. helped design the research study, participated in data analysis, and offered key revisions to the manuscript; K.E. facilitated database query and data set compilation; G.A.-Z. and E.M.H. helped design the research study and offered key revisions to the manuscript; J.M.S conceived of the study, helped design research and analysis and wrote the manuscript, and A.R. offered key revisions to the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Joseph M. Scandura, Department of Medicine, Weill Cornell Medical College, 1300 York Ave, Box 113, New York, NY 10065; email: jms2003@med.cornell.edu.