A 20-patient venetoclax induction pilot in secondary AML stopped at 10, and the interim result was never posted

What was tried

The University of Michigan Rogel Cancer Center, with Genentech, ran UMCC 2021.145 (NCT05780879), an open-label single-arm Phase 2 pilot of venetoclax added to FLAG or CLAG remission induction as frontline therapy for secondary acute myeloid leukaemia. Enrolment opened 3 June 2024 at one United States site. The record lists an actual primary completion date of 16 August 2025, an actual enrolment of 10 participants, and a status of Terminated. The posted reason is "Interim Analysis per Protocol". No results are posted.

FLAG was fludarabine 30 mg per square metre daily and cytarabine 2 g per square metre daily for five days, with tbo-filgrastim continued until neutrophil recovery above 1000. Venetoclax was given orally once daily on days 3 to 16. The single primary outcome was morphologic complete remission assessed to day 43.

Eligibility used the 2016 World Health Organization definition of secondary acute myeloid leukaemia: disease arising after an antecedent haematologic neoplasm including aplastic anaemia, myelodysplastic syndrome, myeloproliferative neoplasms and overlap syndromes, disease carrying myelodysplasia-defining cytogenetic changes, and therapy-related disease. Prior treatment of the antecedent disorder was permitted, prior treatment of the leukaemia itself was not. The stated design was a proof of concept pilot to determine complete remission rate based on 20 patients.

The biological hypothesis

BCL2 suppresses apoptosis by controlling mitochondrial outer membrane permeability, preventing cytochrome c release and binding APAF-1 to restrain caspase activation. Venetoclax (CHEMBL3137309) is a selective BH3 mimetic that displaces pro-apoptotic partners from BCL2.

The pilot rested on two separate claims. The first is that adding venetoclax to intensive induction deepens remissions. That claim had strong support. FLAG with idarubicin plus venetoclax produced a composite complete remission rate of 90 percent in newly diagnosed acute myeloid leukaemia in the original Phase 1b/2 (DOI 10.1200/JCO.20.03736), and 95 percent with 90 percent undetectable measurable residual disease in the 77-patient long-term series, where three-year overall survival was 66 percent and outcomes were similar across European LeukemiaNet 2022 risk groups (DOI 10.1038/s41375-025-02531-8).

The second claim is that this benefit carries into secondary disease, and that is where the evidence thins. Mutations in SRSF2, SF3B1, U2AF1, ZRSR2, ASXL1, EZH2, BCOR or STAG2 are more than 95 percent specific for secondary disease, arise early, persist through clonal remission, and mark patients with lower complete remission rates and shorter event-free survival (DOI 10.1182/blood-2014-11-610543). The benchmark here is not the 90 percent seen in largely de novo cohorts. It is the 47.7 percent overall remission rate CPX-351 achieved against 33.3 percent for 7+3 in 309 older patients with high-risk or secondary disease (DOI 10.1200/JCO.2017.77.6112).

Open Targets scores BCL2 to acute myeloid leukaemia at 0.51388, with literature at 0.983, clinical evidence at 0.691, somatic mutation at 0.4559 and affected pathway at 0.2523. Genetic association contributes zero. The target is well drugged rather than genetically validated, and Open Targets lists five distinct programmes against it.

What actually happened

The registry states only that a protocol-specified interim analysis ended the study, with 10 of a planned 20 participants enrolled and no results posted. It does not name which boundary was crossed, and no publication has appeared.

What the record establishes is the shape of the decision. This was a single-arm pilot whose only primary endpoint was complete remission rate, stopped at a prespecified look at half of planned accrual, with no funding, sponsor or enrolment reason given. That architecture stops early for insufficient activity. The termination is classified here as an efficacy failure on that basis, and the classification would change if the investigators publish an interim saying otherwise.

Failure mechanism, best guess

The most likely mechanism is that a venetoclax benefit demonstrated largely in de novo disease did not reproduce in an ontogeny-defined population selected for resistance. Secondary-type mutations mark a biology with intrinsically lower remission rates independent of what is added to induction. Venetoclax sensitivity also tracks differentiation state: monocytic leukaemia resists venetoclax-based therapy, loses BCL2 expression, and relies instead on MCL1 for oxidative phosphorylation and survival (DOI 10.1158/2159-8290.CD-19-0710). Secondary leukaemia arising from myelodysplastic and myeloproliferative overlap is enriched for the phenotypes in which the drug's target is least available.

How to prevent this next time

No endpoint-level data were released, so no posterior, hazard ratio or subgroup effect can be computed. Two levers are available, and one precision calculation is.

Set the base rate from the matched population, not the flagship trial. A pilot benchmarked against 90 to 95 percent composite remission in de novo cohorts uses the wrong number. The relevant published comparator in older secondary disease is CPX-351 at 47.7 percent.

Enrich on differentiation state, not on ontogeny alone. Secondary leukaemia is a clinical category. Venetoclax resistance is a monocytic phenotype with a measurable BCL2 and MCL1 signature. Screening for it at entry separates a testable population from an untestable mixture.

The precision problem is arithmetic. An exact 95 percent interval around 7 responses in 10 patients runs from 34.8 to 93.3 percent, computed by Clopper-Pearson from those two inputs alone. A 10-patient stage cannot distinguish a regimen performing at the CPX-351 benchmark from one performing at the FLAG-IDA benchmark, so an interim at that size can only detect activity far below either.

The single highest leverage change would have been to power the pilot against the published secondary-disease remission rate of 47.7 percent rather than the de novo rate near 90 percent, which would have set a first-stage size capable of separating those two hypotheses instead of one that could only stop.

What this means for similar programs

BCL2 now carries two recorded terminations in the Claidex graph and a mechanistic risk score of 51, in the orange band and up from 34. The other is [[venetoclax-azacitidine-bcl2-aml-post-transplant-maintenance-phase3-strategic-termination]], the Phase 3 VIALE-T maintenance trial after transplant. The two sit at opposite ends of the treatment course, and both test whether venetoclax adds to a setting where the backbone already works. Neither constrains the approved venetoclax plus hypomethylating agent indication, which remains the strongest evidence for the target. The score is driven by phase burden and saturation across five programmes, not by any safety finding, and the genetic deficit term is the lowest component at 7.29.

Open questions

How many of the 10 participants achieved morphologic complete remission by day 43, and what was the interim boundary? Were participants sequenced for secondary-type mutations, and did responders differ by differentiation state? Was the stop driven by remission rate, by induction mortality, or both? A published interim would answer all three.

Sources

1. • ClinicalTrials.gov, NCT05780879, accessed 30 August 2026. https://clinicaltrials.gov/study/NCT05780879 - Venetoclax with FLAG-IDA induction and consolidation in AML. J Clin Oncol, 2021. https://- Long term results of venetoclax with FLAG-IDA induction and consolidation. Leukemia, 2025. https://- CPX-351 versus 7+3 in older patients with secondary AML. J Clin Oncol, 2018. https://- Acute myeloid leukemia ontogeny is defined by distinct somatic mutations. Blood, 2015. https://- Monocytic subclones confer resistance to venetoclax-based therapy in AML. Cancer Discov, 2020. https://- Diagnosis and management of AML in adults: 2022 ELN recommendations. Blood, 2022. https://- Open Targets Platform, BCL2 and acute myeloid leukemia. https://platform.opentargets.org/evidence/ENSG00000171791/MONDO_0018874 - ChEMBL, CHEMBL3137309. https://www.ebi.ac.uk/chembl/compound_report_card/CHEMBL3137309/ - openFDA FAERS drug event endpoint, 30 August 2026. https://api.fda.gov/drug/event.json.