Effect of anemia developed after ruxolitinib initiation on adverse outcomes in patients with myelofibrosis: a nationwide cohort study

Introduction

Myelofibrosis is a chronic myeloproliferative neoplasm characterized by progressive bone marrow fibrosis, splenomegaly, and ineffective erythropoiesis [1]. Patients with myelofibrosis are at increased risks of complications including anemia, leukemic transformation, cardiovascular events, and infections [1, 2]. Among these, anemia is particularly common and is strongly associated with shortened overall survival (OS) [1, 3].

Ruxolitinib is a first-line therapy for intermediate- to high-risk myelofibrosis, providing benefits such as spleen size reduction, symptom improvement, and prolonged OS [4, 5]. However, anemia frequently develops or worsens during ruxolitinib treatment due to JAK-STAT pathway inhibition, which suppresses erythropoietin-mediated signaling essential for erythropoiesis [1, 6]. While disease-related anemia is a well-established negative prognostic factor for survival in myelofibrosis [1], whether is associated with shortened OS following ruxolitinib initiation remains uncertain. Post-hoc analyses of COMFORT and JUMP trials suggested no impact on OS [7, 8], whereas a recent observational study reported shortened OS [9]. Moreover, no studies have evaluated the association between anemia after ruxolitinib treatment and other adverse outcomes, such as cardiovascular events, infections, and acute myeloid leukemia (AML) [2, 6].

To address these knowledge gaps, we conducted a nationwide cohort study evaluating the associations between anemia developed after ruxolitinib initiation and risks of all-cause death, cardiovascular disease (CVD) hospitalization, serious infection, and AML in patients with myelofibrosis.

Materials/Subjects and Methods

Using a nationwide claims database of South Korea (2014–2023), we conducted a cohort study among patients with myelofibrosis who received ruxolitinib between March 1, 2015 (the first insurance coverage for ruxolitinib) and September 30, 2023. The index date was defined as 90 days after ruxolitinib initiation. We then excluded following patients: missing age or sex; aged < 20 years; ruxolitinib use prior to MF diagnosis; allogenic hematopoietic stem cell transplantation (HSCT) or AML before the index date; death within 90 days after ruxolitinib initiation; anemia within 90 days prior to ruxolitinib initiation (Suppl. Fig. 1). The study exposure was anemia developed after ruxolitinib initiation, defined as (1) more than two diagnostic codes for anemia, (2) an inpatient medication or outpatient prescription accompanied by anemia diagnosis codes, or (3) a procedure code for red blood cell transfusion during the 90-day anemia assessment period after ruxolitinib initiation [10, 11]. The study outcomes were all-cause death, CVD hospitalization, serious infection, and AML. Patients were followed from the index date until the earliest of outcome, death, allogenic HSCT, or the end of the study period (December 31, 2023).

Incidence rates (IRs) per 1000 person-years with 95% confidence intervals (CIs) were estimated using Poisson distribution. Hazard ratios (HRs) and 95% CIs were estimated using Cox proportional hazards models. To minimize potential confounding, we constructed three sequential models with increasing levels of covariate adjustment. Model 1 was adjusted for age and sex. Model 2 additionally included the number of hospitalizations, duration of myelofibrosis, Charlson Comorbidity Index, and health insurance type. Model 3 was further adjusted for predefined comorbidities and comedications. We plotted Kaplan-Meier curves and performed several subgroup analyses, stratified by timing of anemia development. Several sensitivity analyses were performed to assess the robustness of findings, including a subdistribution hazards regression model for study outcomes (except all-cause death) to account for the competing risk of death. The online supplementary materials provide further details on study design, data source, study population, subgroup and sensitivity analysis, and definitions of exposure, outcomes, and covariates. All statistical analyses were performed using SAS Enterprise Guide version 7.1 (SAS Institute).

Results

We identified 917 eligible patients, and 421 (45.9%) developed anemia after treatment initiation, with a median time of 37.5 days (IQR 13–55). Patients who developed anemia were older and had a higher proportion of females compared to those without anemia. (Suppl. Fig. 2, Suppl. Table 4).

During the mean follow-up of 2.77 years, 286 deaths (31.2%) were observed. Post-treatment anemia was associated with increased risks of all-cause death (aHR 1.35, 95% CI, 1.06–1.73) and serious infection (aHR 1.56, 1.19–2.05) in the fully adjusted model. The risk of CVD hospitalizations was elevated but did not reach statistical significance (aHR 1.20, 0.86–1.66). No significant association with AML was found (aHR 1.07, 0.68–1.70) (Table 1, Fig. 1). When stratified by timing of anemia, early-onset anemia ( < 45 days) was associated with increased risks of all-cause death (aHR 1.63, 1.24–2.16) and serious infection (aHR 1.97, 1.45–2.67) compared with patients without anemia, whereas anemia between 45 and 90 days was not significantly associated with these outcomes The study results were consistent in subgroup and sensitivity analyses, including analyses accounting for death as a competing risk (Table 1).

Table 1.

Adverse outcomes associated with anemia following ruxolitinib initiation in patients with myelofibrosis.

	Anemic	Non-anemic
All-cause death
No of events/No of patients	165 / 421	121 / 496
IR per 1000 person-years (95% CI)	146.5 (125.8–170.7)	92.6 (77.5–110.7)
Crude HR (95% CI)	1.60 (1.27–2.03)	Reference
Adjusted HR (95% CI)
Model 1a	1.40 (1.11–1.78)	Reference
Model 2b	1.35 (1.06–1.73)	Reference
Model 3c	1.35 (1.06–1.73)	Reference
Time to anemia development after ruxolitinib initiation (days)c
<45	1.63 (1.24–2.16)	Reference
45–90	1.14 (0.81–1.60)	Reference
CVD hospitalization
No of events/No of patientsd	77 / 383	75 / 489
IR per 1000 person-years (95% CI)	82.7 (66.1–103.3)	64.1 (51.1–80.3)
Crude HR (95% CI)	1.30 (0.94–1.78)	Reference
Adjusted HR (95% CI)
Model 1a	1.19 (0.86–1.65)	Reference
Model 2b	1.20 (0.87–1.66)	Reference
Model 3c	1.20 (0.86–1.66)	Reference
SubHR (95% CI)c,e	1.16 (0.84–1.62)	Reference
Time to anemia development after ruxolitinib initiation (days)c
<45	1.34 (0.91–1.98)	Reference
45–90	1.12 (0.73–1.73)	Reference
Serious infection
No of events/No of patientsd	129 / 391	102 / 487
IR per 1000 person-years (95% CI)	147.0 (123.7–174.6)	89.6 (73.8–108.8)
Crude HR (95% CI)	1.64 (1.26–2.12)	Reference
Adjusted HR (95% CI)
Model 1a	1.49 (1.14–1.95)	Reference
Model 2b	1.53 (1.17–2.00)	Reference
Model 3c	1.56 (1.19–2.05)	Reference
SubHR (95% CI)c,e	1.46 (1.11–1.92)	Reference
Time to anemia development after ruxolitinib initiation (days)c
<45	1.97 (1.45–2.67)	Reference
45–90	1.12 (0.77–1.63)	Reference
AML
No of events / No of patients	37 / 421	42 / 496
IR per 1000 person-years (95% CI)	33.6 (24.3–46.3)	32.8 (24.2–44.4)
Crude HR (95% CI)	1.01 (0.65–1.57)	Reference
Adjusted HR (95% CI)
Model 1a	1.14 (0.72–1.79)	Reference
Model 2b	1.13 (0.72–1.79)	Reference
Model 3c	1.07 (0.68–1.70)	Reference
SubHR (95% CI)c,e	1.04 (0.64–1.68)	Reference
Time to anemia development after ruxolitinib initiation (days)c
<45	1.41 (0.84–2.37)	Reference
45–90	0.66 (0.33–1.33)	Reference

IR incidence rate, HR hazard ratio, CI confidence intervals, CVD cardiovascular diseases, AML acute myeloid leukemia.

^aModel 1: Adjusted for age and sex.

^bModel 2: Adjusted for age, sex, number of hospitalizations, duration of myelofibrosis, Charlson comorbidity index, and health insurance type.

^cModel 3: Adjusted for age, sex, number of hospitalizations, duration of myelofibrosis, Charlson comorbidity index, health insurance type, cancer, chronic kidney disease, depression, diabetes, dyslipidemia, hypertension, splenomegaly, and hydroxyurea use.

^dWhen assessing CVD hospitalization, patients with CVD hospitalization within 90 days after ruxolitinib initiation were excluded. Similarly, when assessing serious infection, patients with serious infection within 90 days after ruxolitinib initiation were excluded.

^eSub-hazard ratios were estimated using all-cause death as a competing risk.

Fig. 1. Cumulative incidence curves of all-cause death, CVD hospitalization, serious infections, and AML among patients with myelofibrosis, stratified by ruxolitinib-related anemia.

cHR crude hazard ratio, aHR adjusted hazard ratio, CI confidence interval, CVD cardiovascular disease, AML acute myeloid leukemia.

Discussion

In this nationwide, population-based cohort study, anemia following ruxolitinib initiation was associated with increased risks of all-cause death and serious infection, and a numerically elevated risk of CVD hospitalization, but not with AML. We observed a temporal relationship, indicating that earlier development of anemia after ruxolitinib initiation was associated with poor prognosis.

To our knowledge, no studies have investigated the association between anemia after ruxolitinib initiation and risks of CVD, infection, or AML; previous studies have only focused on OS. Post-hoc analyses of COMFORT and JUMP trials did not observe significant differences in OS between patients with and without anemia after ruxolitinib initiation. (COMFORT trial: HR 1.03, 95% CI, 0.62–1.73; JUMP trial: HR 0.97, 0.61–1.54) [7, 8]. However, in an observational study using electronic health records, newly developed anemia was associated with shorter OS (HR 2.04, 1.21–3.45), consistent with our findings [9]. These discrepancies may reflect differences in follow-up duration, baseline characteristics, and study settings. Notably, limited follow-up in the clinical trials may partly explain the differences when assessing long-term outcomes such as mortality [12].

Our study has several important clinical implications. Our findings suggest that individual risk profiles for all-cause death and serious infections should be considered when determining the need for additional monitoring, and that earlier intervention for anemia (i.e., erythropoiesis-stimulating agents, transfusion support, or switching to anemia-sparing JAK inhibitors) may be warranted. Furthermore, as earlier development of anemia was associated with higher risks of adverse outcomes, patients with anemia following ruxolitinib initiation may require more comprehensive care and closer monitoring for both myelofibrosis and related complications. Several mechanisms may explain the elevated risk of all-cause death, CVD hospitalization, and serious infections associated with post-treatment anemia. First, ruxolitinib may trigger the development of anemia in patients with myelofibrosis, who are already susceptible to anemia due to low erythropoietic capacity. Given that anemia is a well-established prognostic factor for OS in myelofibrosis [3], our findings suggest that anemia developed after ruxolitinib initiation may be associated with a poorer prognosis. However, as it is difficult to completely differentiate whether post-ruxolitinib anemia was attributable to myelofibrosis, ruxolitinib, or their interaction [8], caution is warranted when interpreting these findings, given that anemia may reflect either a treatment-related adverse effect or underlying disease severity. Second, reduced oxygen transport capacity and tissue hypoxia caused by anemia can increase the risk of CVD by inducing myocardial overload [13–15]. Third, an increased risk of infection may be partially explained by increased bioavailable iron, which is potentially related to reduced erythroid maturation through ruxolitinib-mediated suppression of EPO-EPOR-JAK2-STAT5 signaling [16, 17]. Despite these possible mechanisms, whether improvement or treatment of anemia from ruxolitinib reduces the risk of these adverse outcomes is unclear; therefore, further studies addressing them should be warranted.

Our study has several strengths. First, to our knowledge, this is the first study assessing comprehensive risks of CVD hospitalization, serious infections, and AML associated with anemia developed after ruxolitinib initiation. Second, we used a nationwide healthcare database covering the entire South Korean population, thereby allowing for representative study findings. However, this study has several limitations. First, residual confounding due to the absence of key prognostic variables, such as Dynamic International Prognostic Scoring System (DIPSS) risk categories, cytogenetic abnormalities, and molecular mutations, cannot be ruled out. Second, as ruxolitinib was the only first-line JAK inhibitor available for myelofibrosis in the database, further studies including newer agents such as fedratinib and momelotinib are needed. Third, the results of some subgroup analyses were underpowered and thus should not be overinterpreted. Fourth, exposure underestimation cannot be entirely excluded, as most anemia cases were identified through transfusion (Supplementary Table 3), likely reflecting more severe anemia than those reported in the COMFORT and JUMP trials. However, we expect that such misclassification is unlikely to be substantial, as guidelines recommend regular complete blood count monitoring in patients with myelofibrosis [18], and transfusion may be considered even at higher hemoglobin levels than those recommended by guidelines [19]. Fifth, the generalizability of our study findings to other countries, races, or ethnicities may be limited. In addition, because patients with anemia within 90 days prior to ruxolitinib initiation were excluded to define newly developed anemia, our findings may not be directly applicable to this subgroup.

Author contributions

MJ, JW, JHK, and JYS designed the study and developed the protocol and statistical analysis plan. MJ analyzed the data. MJ and YL drafted the manuscript. JHK and JYS supervised the research. All authors contributed to the study design and revised the manuscript. The corresponding authors attest that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.

Data availability

Data generated and/or analyzed during the current study cannot be shared publicly due to the data-sharing policy of the National Health Insurance Service (NHIS) of Korea, governed by Article 18 of the Personal Information Protection Act (“Limitation to Out-of-Purpose Use and Provision of Personal Information” available at https://elaw.klri.re.kr/kor_service/lawView.do?hseq=53044&lang=ENG). However, the data are available from the NHIS (study identifier: NHIS-2024-2-171) on reasonable request for researchers who meet the criteria for access to confidential data (https://www.data.go.kr/en/tcs/eds/selectCoreDataView.do?coreDataInsttCode=B551182&coreDataSn=1&searchCondition2=coreDataNmEn&searchKeyword2=).

Competing interests

Dr. Shin received grants from the Ministry of Food and Drug Safety, the National Research Foundation of Korea, and pharmaceutical companies, including Pfizer, Celltrion, SK Bioscience, LG Chemical, Union Chimique Belge, and GlaxoSmithKline. Dr. Kim received grants from the Ministry of Health & Welfare. No other relationships or activities have influenced the submitted work.

Ethics approval

This study was approved by the Institutional Review Board of Sungkyunkwan University, South Korea (No. 2025-08-022).

Supplementary information

The online version contains supplementary material available at 10.1038/s41408-026-01557-0.