Clinical Trial Design

Interpreting the PASHA Phase 3 Results: Gilteritinib Versus Midostaurin in FLT3-Mutated AML

The Phase 1b signal for gilteritinib was encouraging, but the randomized PASHA trial did not demonstrate an overall-survival advantage over midostaurin. Detailed EHA 2026 results show fewer relapses and improved relapse-free survival with gilteritinib, while overall survival remained comparable and post-relapse treatment differed between groups.

1. The Rationale for Combination: Phenomenal Phase 1b Results

FLT3 mutations in acute myeloid leukemia (AML) are notorious markers for high relapse rates and poor prognosis. Gilteritinib (Xospata), a highly potent and selective Type 1 FLT3 inhibitor, initially demonstrated spectacular efficacy. In the Phase 1b study published by Pratz et al. in the Journal of Clinical Oncology (2023), researchers evaluated Gilteritinib in combination with intensive induction and consolidation chemotherapy (7+3 regimen) for newly diagnosed FLT3-mutated AML patients.

Among the 58 participants treated at 120 mg once daily who were evaluable for response, 36 had FLT3-mutated AML. In this subgroup, 32 of 36 participants (89%) achieved a composite complete response (CRc), including 83% with conventional complete remission; median overall survival was 46.1 months. These results are hypothesis-generating rather than evidence of a comparative survival benefit: the phase IB study was small, non-randomized, and designed primarily to assess dose, safety, and tolerability. In addition, 26 of the 36 participants with FLT3-mutated AML (72%) underwent hematopoietic stem-cell transplantation during or after study therapy. The transplant rate and patient selection are important context when interpreting the observed survival estimate.

2. Safety: Interpreting Findings Across Different Trials

A major concern in combination trials is the potential for amplified toxicity. However, Gilteritinib maintained a manageable safety profile. In the Phase 3 ADMIRAL trial—which evaluated Gilteritinib monotherapy in relapsed/refractory (R/R) AML—the most common Grade 3 or higher adverse events were febrile neutropenia (45.9%), anemia (40.7%), and thrombocytopenia (22.8%).

In the Phase Ib combination study, the most common Grade 3/4 treatment-emergent adverse events included febrile neutropenia (66.7%), thrombocytopenia (20.5%), neutropenia (17.9%), and anemia (15.4%) in the safety population. Among participants who achieved complete remission after induction at the 120 mg dose, median time to count recovery was 41 days. These rates should not be interpreted as a direct safety comparison with ADMIRAL: the studies enrolled different populations and used different treatment regimens. The small, non-randomized Phase Ib study characterizes observed tolerability but cannot establish the absence of synergistic toxicity or predict the overall benefit-risk profile in a frontline Phase 3 population.

3. What the Randomized Phase 3 PASHA Trial Showed

The HOVON 156 / AMLSG 28-18 / PASHA trial randomized 768 patients with newly diagnosed FLT3-mutated AML to gilteritinib (n=384) or midostaurin (n=384), each with intensive chemotherapy. The primary endpoint was overall survival (OS). At the EHA 2026 presentation, median follow-up was 43.2 months. Median OS had not been reached in either arm; the hazard ratio was 1.02 (95% CI 0.81–1.28; P=0.864), and 48-month OS was 59% with gilteritinib versus 60% with midostaurin. The primary endpoint was not met.

Other outcomes provide a more nuanced picture. Median event-free survival (EFS) was 51.1 versus 19.9 months (HR 0.83; 95% CI 0.68–1.02; P=0.052), a numerical difference that did not meet the conventional two-sided 0.05 threshold. Morphologic relapse after complete remission occurred in 21% versus 36% (HR 0.68; 95% CI 0.54–0.88; P=0.003), favoring gilteritinib. The EHA presentation also reported a relapse-free survival advantage (HR 0.68; 95% CI 0.54–0.88; P=0.003). These endpoints should not be conflated: the trial did not demonstrate an OS advantage, despite fewer relapses and a favorable RFS result.

4. Why Do the Public Sources Show Different Primary Endpoints?

The public protocol record shows an endpoint change over time. The HOVON 156 protocol synopsis, version 5.0 dated March 13, 2023, lists event-free survival (EFS) as the primary objective and overall survival (OS) as a key secondary objective. HOVON's trial page later records that a new protocol version (v07) was added on April 16, 2024. The current HOVON trial page lists OS as the primary study objective and EFS as a key secondary objective. Therefore, the current OS-primary description is consistent with the sponsor's current trial record; the 2023 EFS-primary synopsis is an earlier protocol snapshot, not necessarily a description of the final analysis plan.

The publicly accessible material reviewed here does not establish the exact amendment in which the primary endpoint changed, the date on which the change was approved, or the sponsor's rationale. It would be speculative to attribute the change to regulatory advice, evolving evidence, or a strategic decision without the relevant protocol amendment or statistical analysis plan. For interpretation of the 2026 primary analysis, OS should be identified as the final/current primary endpoint, while the earlier EFS-primary synopsis should be retained as part of the study's documented design history.

A related denominator point: the HOVON registry describes the overall study population as 777 participants and includes both AML and MDS-EB2 in the trial's eligibility/title. The EHA 2026 AML results are reported for 768 randomized AML patients (384 per arm). These figures refer to differently described populations in the public sources; the available material does not provide a complete participant-level reconciliation, so the nine-person difference should not be assigned to a specific subgroup without further documentation.

5. Post-Relapse Treatment: Relevant, but Not Proof of Crossover Causality

Among patients who relapsed, salvage gilteritinib was used in 50% of those initially assigned to midostaurin and 17% of those initially assigned to gilteritinib. Allogeneic transplantation after relapse was also more frequent in the midostaurin arm (34% versus 22%). Median survival after relapse was 10.2 months in the midostaurin arm and 7.2 months in the gilteritinib arm (HR 1.54; 95% CI 1.06–2.23; P=0.022).

These post-relapse treatment patterns are relevant to OS interpretation, and the investigators discussed greater use of salvage gilteritinib and transplantation in the midostaurin arm as a possible reason why the relapse advantage did not translate into an OS difference. However, this is not evidence that protocol-defined crossover caused the primary endpoint result. Salvage treatment after relapse is not necessarily protocol crossover, and comparisons restricted to patients who relapse are post-randomization analyses subject to selection and prognostic differences. The available conference data support a plausible contribution—not a proven causal explanation.

💡 My Practical Perspective: Choosing Between OS, EFS, and PFS

While Overall Survival (OS) is a direct measure of mortality and an important endpoint, its interpretation can be influenced by post-progression therapies and other subsequent care. These factors should be evaluated using the full trial dataset; the publicly reported PASHA topline result does not establish that crossover occurred at a rate sufficient to explain the primary-endpoint result.

In some settings, Event-Free Survival (EFS) or Progression-Free Survival (PFS) may provide complementary evidence because they focus on earlier disease events. However, endpoint choice must match the clinical question, and EFS/PFS definitions and results can still be affected by trial design, assessment schedules, and intercurrent events. They are not automatically superior to OS or immune to every consequence of subsequent treatment.

The PASHA result is a reminder that promising early-phase outcomes may not translate into a statistically significant advantage in a randomized Phase 3 comparison. Sponsors should prospectively consider subsequent therapies and other intercurrent events in endpoint selection and statistical planning, without assuming in advance that any one factor will explain the result.