Real-world duration of therapy with erythropoietic stimulating agents is predictive of overall survival in Myelodysplastic Syndromes (MDS)

Abstract BACKGROUND: Erythropoiesis-stimulating agents (ESAs) are a cornerstone of supportive care for anemia in lower-risk MDS, aimed at reducing transfusion dependence and improving quality of life. While ESA response has been linked to improved overall survival (OS), most evidence comes from post-hoc comparisons between responders and non-responders, limiting its prognostic value. We hypothesize that duration of ESA therapy (DOT) may serve as a real-time surrogate for underlying disease biology, with longer DOT reflecting more indolent disease and potentially favorable disease course. This study explores the association between ESA DOT and OS in a large retrospective cohort of MDS patients treated in the US community oncology setting. METHODS: We conducted a retrospective analysis of MDS patients treated with ESAs using data from the IKnowMed electronic health records (EHR). DOT was defined from ESA initiation to the start of a non-ESA regimen, death, or last ESA exposure, whichever occurred first. Patients were grouped into DOT quartiles, and OS was assessed using Kaplan-Meier methods across these groups. Baseline characteristics, including serum erythropoietin (EPO), R-IPSS score, absolute neutrophil count (ANC), platelet count (PLT), hemoglobin (HGB), age, and sex - were evaluated for confounding. The association between DOT and OS was evaluated using Kendall's tau and landmark analyses at 9, 12, 18, and 24 months. RESULTS: A total of 3,419 MDS patients treated with ESAs between July 1, 2020, and June 30, 2025, were included. Baseline characteristics were generally balanced across DOT quartiles, with similar median age, ANC, time from diagnosis to ESA initiation, and gender distribution. The median (interquartile range [IQR]) of the four ESA DOT quartiles was: Q1: 0.20 (0.03–0.56), Q2: 2.50 (1.64–3.71), Q3: 9.56 (7.33–13.07), and Q4: 29.53 (23.72–39.36) months. Median time from MDS diagnosis to ESA start was similar across ESA quartiles ranging from 1.15 to 1.64 months. The IPSS-R scores at diagnosis of very low, low or intermediate risk were 81.5% for Q4 and 45.4% for Q1. Baseline EPO levels were lower among patients with longer ESA exposure: Q1: 72.35 (31.82–226.22), Q4: 39.10 (20.90–76.62) mU/mL. Baseline platelet levels favored the quartiles with successively longer ESA DOT, Q1: 111.00 (52.00–199.25) versus Q4: 203.50 (137.00–281.25). The median OS was longer for quartile 4: Q1: 9.86 (3.40-21.70), Q2: 12.00 (5.81-25.34), Q3: 18.1 (12.39-28.61), Q4: 35.78 (28.48-44.25) months. Kendall's Tau statistic was 0.374 (p<0.001) for testing the relationship between DOT and OS. This trend suggests a strong correlation between prolonged ESA therapy and survival outcomes. Landmark analysis showed that longer duration of ESA use had a significant lower hazard ratio (0.38, 0.36, 0.34, 0.36 at 9, 12, 18 and 24 months, respectively) as compared with shorter duration of ESA use (P<0.001). The percentages of patients achieving at least a 1.5 g/dL increase in hemoglobin from baseline were successively higher for quartiles with longer median ESA DOT, Q1: 48.3%, Q4: 74.7%. CONCLUSIONS: These findings suggest that ESA treatment duration may serve as a real-time prognostic marker in MDS. We hypothesize that ESA DOT may supplement the use of traditional risk factors. Our observations suggest a strong correlation between ESA duration of therapy and improved survival outcomes, offering clinicians a dynamic tool to assess disease trajectory. Patients with shorter ESA responses may warrant earlier consideration of disease-modifying therapies, including hypomethylating agents or allogeneic stem cell transplantation. These findings may support the integration of ESA DOT into clinical decision-making and risk stratification frameworks. Prospective validation is warranted to confirm these associations and to explore the biological underpinnings of ESA responsiveness in MDS. Future research includes performing multivariate regression models to assess the relative impact of known MDS risk factors along with ESA DOT.