Transfusion independence with momelotinib regardless of baseline erythropoietin levels in the Phase 3 SIMPLIFY-1 trial

Abstract Introduction: Traditional approaches to anemia management in myelofibrosis (MF) include red blood cell transfusions and supportive care such as erythropoiesis-stimulating agents (ESAs), often added to JAK inhibitors (JAKis) such as ruxolitinib (RUX). The efficacy of ESAs is dependent on erythropoietin (EPO), with levels <125 U/L associated with response (Hernández-Boluda J, et al. Eur J Haematol. 2017); guidelines recommend use of ESAs in patients with EPO <500 U/L. Even in patients with EPO below these thresholds, less than half respond, and most responses are <1 year in duration. The JAK1/JAK2/ACVR1 inhibitor momelotinib (MMB) is approved for the treatment of patients with MF who have anemia and demonstrated increased transfusion independence (TI) rates at week 24 vs comparators across 3 phase 3 trials in both JAKi-naive and -experienced patients with MF, including vs RUX in JAKi-naive patients in SIMPLIFY-1. The relative anemia-related benefits of MMB vs comparators as a function of EPO level has not been explored. Methods:SIMPLIFY-1 (NCT01969838) was a randomized phase 3 trial of MMB vs RUX in JAKi-naive patients with MF. Baseline EPO subgroups were defined post hoc using clinically suggested thresholds: <500 vs ≥500 U/L, or <125 vs 125-500 vs >500 U/L. A logistic regression model fit the key secondary endpoint of TI at week 24 (defined as no transfusions and no hemoglobin <8 g/dL in the 12 weeks immediately preceding week 24), estimating the effect of treatment and its interaction with baseline EPO subgroup while adjusting for baseline TI status; the overall interaction between treatment and baseline EPO subgroup was assessed via odds ratios (ORs) and 95% CIs. An additional analysis relaxed the clinically suggested subgroup thresholds to evaluate treatment and baseline EPO in its entirety using all available patient information. A similar logistic regression model fit TI as a function of treatment and baseline continuous EPO with restricted cubic splines; a joint test assessed interaction across all spline terms of treatment and baseline EPO. Postmodel prediction of TI generated an OR and 95% CI for further assessment of the impact of treatment as a function of baseline EPO across nearly the entire sample (1st-99th percentiles). Results: In SIMPLIFY-1, 201/215 MMB-randomized and 206/217 RUX-randomized patients were evaluable for baseline EPO; 196 patients in each arm had EPO <500 U/L. Within the <500 vs ≥500 U/L subgroup set, TI rates with MMB vs RUX were 69% vs 53% (OR, 2.37; 95% CI, 1.50-3.73) and 40% vs 10% (OR, 8.38; 95% CI, 0.51-138.2) in patients with EPO <500 and ≥500 U/L, respectively; there was no difference in treatment benefit between EPO (<500 vs ≥500 U/L) subgroups (OR, 0.28; 95% CI, 0.02-4.77), although the EPO ≥500 U/L sample was small. For assessment in the 3-category subgroup set (<125 vs 125-500 vs ≥500 U/L), within the <500 U/L subset, TI rates were 72% vs 55% in patients with EPO <125 U/L (MMB, n=179; RUX, n=183; OR, 2.27; 95% CI,1.41-3.64) and 35% vs 15% in those with EPO 125-500 U/L (MMB, n=17; RUX, n=13; OR, 7.13; 95% CI,1.06-48.23); assessment of the treatment effect between subgroups again showed no evidence of between subgroup–specific differences (<125 vs 125-500 U/L: OR, 0.32; 95% CI, 0.04-2.26; 125-500 vs ≥500 U/L: OR, 0.87; 95%CI, 0.03-25.22; <125 vs ≥500 U/L: OR, 0.28; 95% CI, 0.02-4.66). In an additional analysis assessing baseline EPO as a continuous variable, a joint test across all treatment and EPO spline terms showed no evidence of interaction (P=.11). Additional prediction of the effect of treatment and baseline, continuous EPO consistently displayed ORs that favored MMB vs RUX (ORs >1). While the majority of 95% CIs also resided above the null for the lower region of baseline EPO (20-550 U/L), the main proportion of the patient sample also resided in this region (EPO<500 U/L). Conclusions: In JAKi-naive patients with MF, rates of maintaining or achieving TI at week 24 were higher with MMB vs RUX regardless of baseline EPO level. There was no significant effect between baseline EPO subgroups on TI rates with MMB, suggesting no comparative EPO-based subgroup advantage on MMB treatment benefit. These results highlight the capacity of single-agent MMB to offer comprehensive spleen, symptom, and anemia-related benefits across a broad JAKi-naive patient population irrespective of EPO levels.