Flatiron Health Cohort of 482 AML Patients Finds Two-Month Real-World Event-Free Survival on FLT3, IDH1 and IDH2 Inhibitors, Whatever the Line of Therapy

How the cohort was built

Manuel R. Espinoza-Gutarra of the O'Neal Comprehensive Cancer Center at UAB wrote the first draft, with Brooke Jarret, Xiaoliang Wang and Anosheh Afghahi of Flatiron Health and Sejong Bae of Augusta University's Georgia Cancer Center and Wellstar MCG Health. The dataset is Flatiron's nationwide electronic health record-derived database, which the paper says drew on around 280 cancer clinics and about 800 sites of care over the study window, most of them community oncology practices. Patients qualified if they were 18 or older with a confirmed AML diagnosis other than acute promyelocytic or mixed-phenotype leukemia and received a FLT3, IDH1 or IDH2 inhibitor, alone or with a hypomethylating agent, between Jan. 1, 2015, and Dec. 31, 2023. The group used the same Flatiron extract for an earlier paper on racial and ethnic disparities in inhibitor outcomes.

An event for real-world event-free survival was induction failure, relapse or death, captured from marrow blast counts above 5% in lab and pathology reports or clinician notes, a documented provider assessment of failure or relapse within 28 days of a marrow sample, or any peripheral blast percentage above zero, a threshold the authors call conservative next to the CIBMTR's 5%. ELN 2017 risk, assigned from available cytogenetic and molecular data, was adverse in 42%, intermediate in 52% and favorable in 6%.

Numbers by line, target and combination

Median time on the inhibitor was 3.7 months in the first line and 3.4 months afterward.

In the multivariable Cox model, starting the inhibitor in the third line (HR 1.67, 95% CI 1.12 to 2.49) or fourth line and beyond (HR 1.83, 1.18 to 2.83) carried a higher hazard of death than starting in the first line, while the second line did not differ significantly. Transplant after the inhibitor lowered the adjusted hazard by 23% but with a confidence interval of 0.54 to 1.41, which the authors attribute to only 49 such patients.

A post hoc analysis restricted to the 2L+ patients without prior transplant, designed to exclude post-transplant maintenance and to resemble relapsed or refractory trial populations, produced median real-world event-free survival of 2.1 months for FLT3, 1.6 for IDH1 and 1.4 for IDH2. Inhibitor plus hypomethylating agent gave median rwEFS of 2.1 months against 1.5 for monotherapy and rwOS of 11.8 against 9.7 months; neither difference reached significance.

Against the registration trials

The comparisons the paper draws against the registration trials are qualitative, with no formal analysis behind them. Gilteritinib's Phase 3 median event-free survival of 2.8 months and overall survival of 9.3 months sit close to the cohort's 2.1 and 9.5 months in FLT3-mutated disease. For IDH1, the 15.4-month real-world overall survival exceeds the 8.8 months of the ivosidenib Phase 1/2 trial and the 12.6 months in its newly diagnosed subset, which the authors attribute to subsequent therapies, above all hypomethylating agents with venetoclax, which was not approved for AML until November 2018, after most of the registration trials had finished accrual. For IDH2, the cohort's 8.7-month overall survival on enasidenib compares with 9.3 and 6.5 months in the early-phase and Phase 3 trials. The pattern across targets is shorter event-free survival than in trials with roughly matched overall survival, which they ascribe to fitter, more adherent trial patients and to differing event definitions.

The discussion ends on regulatory ground, with the authors writing that real-world data and confirmatory Phase 3 studies "shine a light on the actual clinical utility of targeted therapies that receive conditional approval based on response rates in early phase trials," and they cite enasidenib monotherapy, which "was removed from the Canadian market after the confirmatory Phase 3 trial failed to show improvement over best available care." Limitations listed include no response-rate data, inability to apply ELN 2022 risk, too few patients to analyze olutasidenib or mutation-specific combination subgroups, and no way to distinguish post-transplant maintenance from salvage intent. Espinoza-Gutarra reports consulting fees from AbbVie, Stemline, Incyte and Gamida Cell, research support from the BMS Foundation and speaking fees from AAMDSIF and NMDP; the other authors declare none.

Why This Matters to the APO|APE Reader

The 1.4-month median event-free survival for IDH2-mutated patients on later-line enasidenib sits beside the authors' note that the same drug has already lost its Canadian marketing after its confirmatory Phase 3 trial. Because every patient in this cohort had already been molecularly profiled, the Medicaid survival gap cannot be explained by missing testing, a point the paper makes directly against earlier Medicare analyses. The practical reading for laboratories is that the FLT3, IDH1 and IDH2 result that opens access to these drugs is, on this evidence, buying a median of two months before the next event in community practice, and the paper points to an ongoing trial of first- versus second-line targeted therapy in IDH-mutated patients unfit for intensive chemotherapy.