In this single-center retrospective cohort study, we included all adult patients who presented to Moffitt Cancer Center between 2001 and 2021 with a diagnosis of primary MF per World Health Organization (WHO) 2016 classification criteria, or post-polycythemia vera or post-essential thrombocythemia MF per International Working Group (IWG) criteria (15, 16). The study protocol was approved by the institutional review board at the University of South Florida. Clinical and genomic data on eligible patients were collected from a well-annotated institutional MF database.

Clinical and genomic variables were collected at time of first presentation to our center. Constitutional symptoms were largely recorded by treating physicians as part of routine clinical care. Anemia was defined as the requirement of red blood cell transfusions and/or a hemoglobin level of <10 g/dL at presentation. Patients who started ruxolitinib ≤ 6 months prior to presentation were excluded from this analysis to better isolate disease-related rather than JAK inhibitor-driven anemia.

Cytogenetic findings were used to categorize patients into ‘favorable’, ‘unfavorable’, and ‘very high risk’ categories (17). Data on relevant somatic mutations was obtained as part of routine clinical care using an in-house targeted DNA sequencing panel for myeloid malignancies. Although this panel has evolved over the years, it has always had coverage for JAK2, MPL, CALR, TET2, DNMT3A, ASXL1, SRSF2, U2AF1, EZH2, IDH1, IDH2, KRAS, NRAS, RUNX1, and TP53. A minimum variant allele frequency of 5% was used to call single nucleotide variants, and a cutoff of 10% was used for insertions or deletions. Pathogenicity of variants was determined by an internally relied on variant classification system, based on joint consensus recommendations of the Association for Molecular Pathology, American Society of Clinical Oncology, and the College of American Pathologists, in conjunction with review of COSMIC and cBioPortal databases as needed.

Categorical variables were summarized in the form of numbers and percentages. Standard descriptive statistics were used for the continuous variables. Categorical variables were compared using the Fisher’s exact test or Chi-squared test, as applicable. Comparisons of continuous variables were performed with the Wilcoxon rank sum test or Kruskal–Wallis test, as appropriate. OS was estimated using the Kaplan-Meier method with survival estimates being compared using log rank test. OS was calculated from the time of presentation. A univariate Cox regression analysis identified variables that correlated with OS. For serum albumin, median level in study cohort was used as cut-off for regression analysis. All variables with a p value of <0.10 were included in the subsequent multivariate analyses. Due to missing genomic data in a significant proportion of patients, two multivariable analyses were performed. The first was performed on the entire cohort and only included clinical variables. The second was only performed on patients with available genomic information. From these findings, a novel prognostic scoring system was proposed, incorporating the independent predictors for OS on multivariate analyses with scores assigned based on magnitude of hazard ratios (HR). All statistical analyses were conducted in IBM SPSS statistics (version 29).

Among total 737 patients with MF, 365 (49.5%) patients were anemic at presentation, including 205 patients who required transfusions. The median age at presentation was 68.5 (range, 18.9-91.4) years. Median time from MF diagnosis to presentation was 2.8 months. Clinical and genomic characteristics of study population are shown in Table 1 . Patients with anemic MF were significantly older (median age, 69.8 vs. 67.2 years; p <0.01). Anemic patients were more likely to be diagnosed with primary MF (72% vs. 65%; p = 0.04). Patients with anemic MF had a lower median WBC (6.9 x 10⁹/L vs. 11 x 10⁹/L; p <0.01) and platelet count (146 x 10⁹/L vs. 279 x 10⁹/L; p <0.01) than non-anemic patients. Serum erythropoietin and ferritin levels were significantly higher in anemic MF (median, 91 vs. 18 mIU/ml, and 567 vs. 136 microgram/L, respectively; p <0.01). Median serum albumin was 4.1 gm/dL in anemic patients vs. 4.4 gm/dL in others (p <0.01). Patients with anemic MF had higher frequency of advanced (MF 2 or 3) marrow fibrosis.

MF patients with anemia more frequently harbored unfavorable or very high-risk cytogenetics than patients without disease-driven anemia (36% vs. 28%; p =0.03). Mutations in JAK2 were less common in anemic than non-anemic subgroup (57% vs. 65%; p =0.02), whereas mutations involving U2AF1 (20% vs. 3%; p <0.01) and EZH2 (11% vs. 6) were more prevalent in patients with anemic MF.

In our study population, the treatment modalities across all lines of therapy included erythropoiesis stimulating agents (ESA) (32%), hydroxyurea (34%), thalidomide or lenalidomide (24%), ruxolitinib (38%), androgen (8%), and interferon (5%). Use of ESA, thalidomide or lenalidomide, and androgen was more common, and treatment with hydroxyurea and ruxolitinib was less common in the anemic MF subgroup ( Table 1 ). Overall, 14% of patients underwent allogeneic hematopoietic stem cell transplant (allo-HSCT), with this intervention being more commonly used in anemic patients (17% vs. 11%; p = 0.03).

Rate of blast phase transformation was significantly higher in patients with anemic MF than those without anemia (14% vs. 7%; p <0.01). After a median follow up duration of 34.5 months, median OS was 45.4 (95% confidence interval [CI]: 39.9-50.8) months for the entire study population. Anemia at presentation was associated with significantly worse OS in patients with MF (30.2 [95% CI: 24-36.4] vs. 73.9 [95% CI: 54.6-93.3] months; p <0.01).

In the univariate analysis ( Table 2 ), variables associated with an inferior OS in the anemic MF cohort (n=365) included age > 65 years (HR 2.07; p <0.01), constitutional symptoms (HR 1.59; p <0.01), WBC count >25 x 10⁹/L (HR 3.23; p <0.01), platelet count <100 x 10⁹/L (HR 1.57; p <0.01), peripheral blasts ≥ 1% (HR 1.66; p <0.01), serum albumin <4.1 gm/dL (HR 1.86; p <0.01), and the presence of an ASXL1 (HR 1.23; p <0.01), EZH2 (HR 1.22; p <0.01), SRSF2 (HR 1.21; p =0.01), or TP53 (HR 1.22; p <0.01) mutation.

Two separate multivariate analyses were performed. In the first multivariate analysis, we used data from entire anemic MF cohort (n=365) and included all clinical and laboratory variables with borderline significance (p <0.10) on univariate analysis. Age >65 years (HR 1.90; p <0.01), WBC count >25 x 10⁹/L (HR 2.43; p <0.01), platelet count <100 x 10⁹/L (HR 1.54; p <0.01), peripheral blasts ≥ 1% (HR 1.41; p =0.02), and serum albumin <4.1 gm/dL (HR 1.47; p =0.01) were the independent prognostic factors for OS.

The second multivariate analysis included only the anemic MF cases with available baseline molecular information (n=211). In this analysis, clinical, laboratory and genomic variables were included. WBC count >25 x 10⁹/L (HR 2.19, 95% CI: 1.16-4.14; p =0.01), platelet count <100 x 10⁹/L (HR 2.15, 95% CI: 1.31-3.53; p <0.01), serum albumin <4.1 gm/dL (HR 1.71, 95% CI: 1.06-2.75; p =0.02), SRSF2 mutation (HR 2.79, 95% CI: 1.47-5.28; p <0.01), and TP53 mutation (HR 2.49, 95% CI: 1.10-5.61; p =0.02) were the independent predictors of OS in patients with MF and anemia ( Table 2 , Figure 1 ).

Based upon these findings, we developed a prognostic scoring system comprised of the five significant variables in the multivariate analysis on molecularly annotated cohort of anemic MF. This score included serum albumin <4.1 gm/dL (1 point), WBC count >25 x 10⁹/L (1.5 points), platelet count <100 x 10⁹/L (1.5 points), TP53 mutation (1.5 points), and SRSF2 mutation (2 points). Patients in the low risk (no risk factor; aggregate score 0), intermediate risk (1 risk factor; aggregate score 1 to 2), and high risk (>1 risk factor; aggregate score >2) categories had distinct median OS estimates of 69, 37.7, and 11.6 months, respectively (p <0.01) ( Figure 2 ).