Disease-Modification Endpoint Use in Myelofibrosis Clinical Trials: An Analysis of 145 Registered Trials, 2010-2025

Drug approvals in myelofibrosis (MF) rest on two COMFORT-era endpoints, spleen volume reduction (SVR35) and total symptom score reduction (TSS50). Four JAK inhibitors have been approved since 2011, most recently momelotinib for patients with anemia, and agents now in development claim disease modification along three axes: reversal of cytopenia, reduction of the malignant clone, and improvement in marrow fibrosis. Whether MF endpoint practice has moved toward these constructs has not been measured.

We queried a pinned 07/2026 snapshot of the AACT ClinicalTrials.gov database using the ctrialsgov R package. We included interventional treatment trials in phases 1/2 through 3, spanning 2010 to 2025, that included an explicit myelofibrosis condition (primary, post-polycythemia vera, post-essential thrombocythemia); trials identified only by derived MeSH indexing were excluded as non-MF protocols. We classified every protocol-specified endpoint into three disease-modification axes: clinical (anemia response, transfusion independence, hemoglobin improvement), clonal (allele burden, molecular or cytogenetic response, measurable residual disease), and structural (marrow fibrosis grade). The dictionary was validated by independent adjudication in a related cohort (sensitivity 0.81-0.85, PPV 0.85-0.91). Drug mechanisms were assigned from registered intervention names and aliases and modeled using non-exclusive indicators, so that combination trials contributed to each mechanism they contained. We modeled adoption by logistic regression with Firth penalization where strata were sparse, applying a Bonferroni correction with α = 0.01 across the five pre-specified hypotheses reported in order below.

One hundred forty-five trials met criteria, contributing 1,297 endpoints; 89 (61.4%) were industry-led. Seventy trials (48.3%) measured at least one disease-modification endpoint: 53 (36.6%) clinical, 19 (13.1%) clonal, and 19 (13.1%) structural. Spleen endpoints appeared in 57.2% and symptom endpoints in 58.6%. Adoption did not change over 15 years (OR 0.97 per year, 95% CI 0.90-1.05, p=0.48), and no individual axis changed (all p>0.25). The estimate held across sensitivity analyses varying cohort, phase set, classification rule and endpoint-count adjustment (OR 0.96-0.99). Momelotinib approval produced no detectable step in clinical-axis use, though with 21 subsequent trials a substantial change could not be excluded (37.9% before versus 28.6% after; OR 0.80, 95% CI 0.22-2.69). Clonal endpoints appeared in 3 of 8 interferon-containing trials (37.5%) versus 9 of 89 containing a JAK inhibitor (10.1%), but interferon was not independently associated with clonal-endpoint use (Firth OR 2.74, 95% CI 0.54-12.60, p=0.21). The pre-specified mutually exclusive class model gave OR 5.15 (95% CI 1.04-23.52, p=0.045), not meeting α = 0.01; two of the three interferon trials with clonal endpoints were ruxolitinib-interferon combinations. Structural endpoints were more common in ruxolitinib-combination trials from 2018 onward (21.4% versus 12.1%, OR 2.04, 95% CI 0.59-6.94). Clinical-axis adoption was independent of the other axes (OR 1.01), but clonal and structural endpoints co-occurred about fourfold more often than chance (OR 3.96, 95% CI 1.05-13.75); axes per trial did not increase (rate ratio 0.99, p=0.81).

Across 15 years and four drug approvals, MF trials did not increase measurement of disease-modification endpoints. Spleen and symptom endpoints remained in nearly 60% of trials, while clonal and structural endpoints each appeared in roughly one trial in eight. This null was robust to every cohort definition, phase set, and classification rule tested. Clonal and structural endpoints tended to be adopted together, consistent with both requiring a marrow biopsy, but neither tracked the clinical axis, so disease modification is not being measured as a package. We could not demonstrate that momelotinib's approval for MF with anemia increased the use of anemia endpoints, though the post-approval sample is small. Endpoint choice appears to follow the mechanism a sponsor is testing rather than any field-wide shift toward measuring disease modification.