This prospective observational study was conducted between June 2017 and November 2023 at a medical center in Taiwan (approval numbers: 201406099RIND and 201810058RIND). Patients with R/R AML harboring FLT3 mutations were eligible for this study. All enrolled patients received gilteritinib as a salvage therapy, followed by allo-HSCT. Gilteritinib was discontinued before initiation of the conditioning regimen. Posttransplantation, gilteritinib maintenance was permitted in patients who demonstrated successful engraftment, which was defined as an absolute neutrophil count of ≥ 500/mm3 and platelet count of ≥ 20,000/mm3, with no transfusion support, no evidence of grade ≥ 2 acute graft-versushost disease (GvHD), and who were between 30 and 90 days posttransplantation. The patients were categorized into maintenance and non-maintenance groups based on whether they continued receiving gilteritinib therapy posttransplant. The maintenance dose of gilteritinib varied (40, 80, or 120 mg daily). In patients who maintained sustained complete bone marrow remission with full donor chimerism, discontinuation after two years of maintenance therapy should be considered. Decisions regarding continuation or discontinuation after two years were individualized and made at the discretion of the treating physician in discussions with the patient, taking into account the MRD status, donor chimerism, treatment tolerance, and patient preference. The mean daily dose was calculated and used for comparison. Notably, three patients participated in the ADMIRAL trial [10] and received gilteritinib prior to formal approval by regulatory authorities. Patient demographics, clinical characteristics, disease status, and treatment patterns were collected through a comprehensive review of institutional electronic medical records. Risk stratification was performed using features at diagnosis that were grouped based on the refined Medical Research Council (MRC) criteria [22] and the European LeukemiaNet (ELN) 2022 classification system [23]. Tapering of immunosuppressive therapy (IST) was performed according to the institutional protocols. In the absence of acute GvHD, calcineurin inhibitors were generally tapered beginning approximately one month after transplantation, with the goal of discontinuation before six months whenever clinically feasible, particularly given the R/R setting. FLT3 mutations were identified via fragment analysis, whereas other myeloid mutations were assessed using the TruSight myeloid sequencing panel and HiSeq platform (Illumina, San Diego, CA, USA) [24]. Library preparation and sequencing were performed according to the manufacturer’s instructions. The details of the variant analysis algorithm used to analyze the diagnostic samples have been described previously [25]. Due to the unavailability of MRD evaluation by next-generation sequencing (NGS), MRD was evaluated by multiparameter flow cytometry and/or quantitative PCR targeting specific genes such as NPM1 and core binding factors, as previously described [26, 27]. The sensitivity of multiparameter flow cytometry was approximately 10⁻4, whereas quantitative PCR assays for gene-specific targets achieved a sensitivity of up to 10⁻5. Pre- and post-HSCT MRD were evaluated within 30 days before conditioning and on day +21 after transplantation, respectively.

The primary endpoint of the study was RFS, which was defined as the length of time from the initiation of allo-HSCT treatment until disease relapse or death from any cause. The secondary endpoint was OS, defined as the length of time from the initiation of allo-HSCT until death or the last follow-up. Efficacy outcomes, including treatment response and relapse, were assessed according to the ELN 2022 guidelines [23]. Adverse events during gilteritinib treatment were evaluated and graded using the Common Terminology Criteria for Adverse Events (CTCAE) v.5.0. Clinically significant cytomegalovirus (csCMV) infection was defined as the development of CMV disease or the initiation of preemptive anti-CMV therapy [28].

Comparisons between continuous variables were performed using the Mann–Whitney U test or Kruskal–Wallis test, whereas differences between categorical variables were assessed using the chi-squared test or Fisher’s exact test. Survival probabilities were estimated using the Kaplan–Meier method, and the log-rank test was used to compare survival curves between groups. Multivariate regression analysis was performed using the Cox proportional hazards model to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). All statistical analyses were performed using the R software (v.4.3.3; https://cran.r-project.org/). Statistical significance was defined as a two-sided P-value of < 0.05.

Between June 2017 and November 2023, 42 patients with R/R AML harboring FLT3 mutations were enrolled in the study. Table 1 summarizes the baseline characteristics, molecular profiles, and treatment outcomes of patients, stratified by post-HSCT maintenance therapy status as follows: no maintenance (n = 14), maintenance with MRD positivity (n = 5), and maintenance with MRD negativity (n = 23). The cohort had a median age of 48 years (range, 10–70 years), with 54.8% female patients. Two patients developed secondary AML from a myelodysplastic syndrome/neoplasm (MDS). The majority (78.6%) of patients had intermediate-risk cytogenetics, and according to the 2022 ELN risk classification, 69.0% were classified as intermediate-risk. The most common molecular alteration at diagnosis was NPM1 mutation (45.2%), followed by DNMT3A and RUNX1 mutations (14.3%). At enrollment, 11 (26.2%) patients were primarily refractory, having received a median of 2.5 prior lines of therapy. In terms of FLT3 mutations, 39 (92.0%) patients harbored the FLT3-ITD mutation, one patient had both FLT3-ITD and FLT3-TKD mutations, and one patient had only the FLT3-TKD mutation. The allelic ratios of FLT3-ITD were uniformly distributed. Notably, five patients acquired FLT3-ITD at relapse despite the absence of FLT3-ITD at the initial diagnosis. Among the patients with FLT3 mutations at diagnosis, 31 (70.3%) were given FLT3 inhibitors (primarily midostaurin, 67.6%) as part of frontline therapy, whereas the remaining patients were not because of financial issues. Menin inhibitors were not administered to patients with persistent NPM1 MRD because these were not available in Taiwan during the study period. The conditioning regimens used were myeloablative (MAC) in 35.7% and reduced-intensity (RIC) in 64.3% of the patients. Donor types included matched siblings (38.1%), haploidentical donors (23.8%), matched unrelated donors (31.0%), and mismatched unrelated donors (7.1%). GvHD prophylaxis primarily involved the use of calcineurin inhibitors and mycophenolate mofetil (69.0%). Among the 26 patients who were alive at the last follow-up, five remained on IST, with a median post-HSCT follow-up of 48.6 months in this subgroup.

The CR rate was 76.2%, partial response (PR) rate was 11.9%, with progressive disease occurring in 11.9% of patients. With a median follow-up of 46.2 months (range, 1.8–93.1 months), the OS of the entire cohort was not reached (NR), and the RFS was 62.0 months. The median exposure of post-HSCT gilteritinib maintenance was 21.2 months (range, 0.8–85.5 months). Among the 28 patients who received post-HSCT gilteritinib maintenance, 11 (39.2%) electively discontinued therapy while remaining in complete remission, with a median treatment duration of 23.3 months (range, 6.3–34.4 months) in this subgroup. Among these patients, two subsequently experienced relapse at 14.3 and 38.2 months after discontinuation of gilteritinib, whereas the remaining nine (81.8%) patients remained relapse-free at last follow-up. In the univariate analysis, patients with acquired FLT3 mutations had significantly worse OS and RFS than those without (NR vs. 15.4 months, P < 0.001; 62.9 vs. 8.8 months, P = 0.002). Patients with a history of MDS had worse OS than those without (NR vs. 11.0 months, P = 0.054). However, a higher FLT3 allelic ratio (cutoff 0.5), cytogenetic risk based on the refined MRC classification, genomic risk based on the ELN 2022 guidelines, initial hyperleukocytosis, and co-occurring mutations were not significantly associated with clinical outcomes. Notably, primary refractory disease was no longer a poor prognostic factor in patients who underwent allo-HSCT followed by gilteritinib maintenance therapy.

With respect to MRD, 7 (43.8%) patients with detectable pre-HSCT MRD achieved MRD negativity after HSCT. OS and RFS were significantly worse among patients with pre-HSCT MRD compared with those without (OS: 18.1 months vs. NR, P < 0.001; RFS: 9.7 months vs. NR, P < 0.001) (Supplementary Fig. 1). Similar findings were observed for post-HSCT MRD findings (OS: 18.1 months vs. NR, P < 0.001; RFS: 12.1 months vs. NR, P = 0.002) (Supplementary Fig. 2). Interestingly, patients who were MRD-positive before HSCT but MRD-negative after HSCT had outcomes comparable to those who were MRD-negative before and after HSCT (P = 0.317 for OS).

The presence of chronic GvHD was also associated with better OS and RFS (OS: NR vs. 32.8 months, P = 0.008; RFS: NR vs. 13.1 months, P = 0.002) (Supplementary Fig. 3). Gilteritinib maintenance therapy was a favorable prognostic factor for both OS (NR vs. 17.0 months, P < 0.001) and RFS (NR vs. 10.9 months, P < 0.001) (Fig. 1A, B). The 1-year cumulative incidence of relapse (CIR) was significantly lower in the gilteritinib maintenance group (10.7% ± 5.8% vs. 14.3% ± 9.4%, P < 0.001). Additionally, we included post-HSCT MRD status, presence of chronic GvHD, and gilteritinib maintenance as covariables in the multivariate Cox proportional hazards regression analysis model, which remained independent favorable prognostic factors for both OS and RFS (OS: HR 0.086, P = 0.002; RFS: HR 0.148, P = 0.002). With respect to HSCT-related complications, the use of gilteritinib did not increase the risk of csCMV infection (P > 0.9) and no cases of posttransplant lymphoproliferative disorders were observed. To explore whether the effect of gilteritinib maintenance varied according to MRD status, we performed exploratory subgroup analyses stratified by pre- and post-HSCT MRD status, as assessed by flow cytometry. Among patients with detectable pre-HSCT MRD, no statistically significant differences in OS or RFS were observed between the gilteritinib and non-gilteritinib groups. However, any interpretation of this subgroup is limited by the small sample size and the high frequency of alternative posttransplant interventions in the non-gilteritinib group (Table 1), including venetoclax-based therapy and donor lymphocyte infusion, both of which may have mitigated the relapse risk.

In contrast, among patients with no detectable pre-HSCT MRD, gilteritinib maintenance was associated with significantly improved OS and RFS (OS: NR vs. 20.6 months, P < 0.001; RFS: NR vs. 13.8 months, P < 0.001) (Supplementary Fig. 4). Similar patterns were observed when stratified by post-HSCT MRD status (OS: NR vs. 22.1 months, P < 0.001; RFS: NR vs. 13.1 months, P < 0.001)(Supplementary Fig. 5). Importantly, these subgroup analyses were not powered to detect treatment– MRD interactions and were based on flow cytometry rather than FLT3-targeted NGS MRD, which is now considered a more sensitive and specific modality.

Among patients receiving gilteritinib maintenance, no active GvHD greater than grade 2 was observed, and no patient died from GvHD or GvHD-related complications. Two patients experienced grade II transaminitis, which either improved or stabilized with no dose modification. One patient presented with increased creatine kinase levels despite dose reduction, leading to discontinuation of gilteritinib.