Quizartinib Resistance Mutations and Treatment Outcomes in Relapsed or Refractory FLT3‐ITD–Positive AML Authors: Yuichiro Semba, Toshihiro Miyamoto, Senji Kasahara, Yoshikane Kikushige, Takahiro Maeda, Goichi Yoshimoto, Shuichi Ota, Akio Saito, Yuho Najima, Hiroki Hosoi, Daigo Niiya, Koji Kato, Koji Nagafuji, Takashi Akasaka, Tomohiko Kamimura, Atsushi Takita, Maiko Narahara, Tadashi Toki, Koichi Iwanaga, Koji Yonemoto, Mine Harada, Koichi Akashi Published: 2026 ABSTRACT ABSTRACTQuizartinib is a FMS‐like tyrosine kinase 3 (FLT3) inhibitor indicated for FLT3 internal tandem duplication (FLT3‐ITD)–positive acute myeloid leukemia (AML). We aimed to evaluate quizartinib resistance mechanisms, in addition to efficacy and safety outcomes, in patients with relapsed or refractory FLT3‐ITD–positive AML. This multicenter, single‐arm study in Japan (jRCTs071200015) enrolled 18 patients between May 2020 and December 2022. Of these, 15 patients received oral quizartinib (up to 53 mg once daily) for up to 12 cycles of 28 days each, then were followed for 12 months. The primary endpoint was to evaluate the type and rate of quizartinib resistance mutations; secondary endpoints included composite complete remission (CRc) rate, overall response rate (ORR), hematopoietic stem cell transplantation (HSCT) rate, relapse‐free survival (RFS), overall survival (OS), and adverse events (AEs). Among seven evaluable patients, acquired mutations were detected in four patients (NF1 [R2616X], CSF3R [Q754X], NRAS [G13R], and FLT3 [D835Y] in one patient each), while loss of FLT3‐ITD was observed in two patients. In efficacy analyses (n = 15), CRc rate was 66.7% (95% confidence interval [CI], 38.4–88.2), ORR was 73.3% (44.9–92.2), and median OS was 13.6 months (5.4–not evaluable). Three patients (20.0%) received HSCT directly after quizartinib; in these patients, median RFS was 8.5 months (95% CI, 6.2–not evaluable). Grade ≥ 3 non‐hematologic AEs and grade 1 QT prolongation were each reported in three patients (20.0%). These data offer additional information on potential resistance mechanisms in patients with relapsed or refractory FLT3‐ITD–positive AML. Trial Registration: Japan Registry of Clinical Trials (jRCTs071200015). Abbreviations adverse event acute myeloid leukemia cytarabine azacitidine cycle confidence interval complete remission composite complete remission complete remission with incomplete neutrophil recovery complete remission with incomplete platelet recovery Common Terminology Criteria for Adverse Events Diagnostic Sequencing of Cancer Associated Variants daunorubicin Eastern Cooperative Oncology Group performance status flow cytometry fludarabine cytarabine granulocyte colony stimulating factor and mitoxantrone FMS‐like tyrosine kinase 3 GATA‐binding protein 2 hematopoietic stem cell transplantation internal tandem duplication juxtamembrane domain juxtamembrane segment juxtamembrane zone leukemic stem cell myelodysplastic syndrome myelodysplasia syndrome 1 and ecotropic viral integration site 1 complex locus mixed phenotype acute leukemia myeloproliferative disorder measurable residual disease next‐generation sequencing not otherwise specified nucleophosmin 1 no response nucleoporin 214 overall response rate overall survival cycles following protocol‐specified quizartinib treatment progressive disease partial remission relapse red blood cell relapse‐free survival months following subsequent treatment single‐cell DNA sequencing stable disease months following hematopoietic stem cell transplantation T‐cell immunoglobulin mucin‐3 tyrosine kinase domain variant allele frequency venetoclax World Health Organization Introduction Acute myeloid leukemia (AML) is a genetically heterogeneous hematologic malignancy characterized by the clonal expansion of undifferentiated myeloid precursor cells, impaired hematopoiesis, and progressive bone marrow failure [1, 2]. Although many patients achieve initial remission with induction chemotherapy, treatment resistance is common, and relapse remains the principal cause of therapeutic failure [3]. Advances in genomic profiling have revealed a wide spectrum of somatically‐acquired driver mutations in AML [1]. Approximately 30% of patients harbor mutations in the gene encoding FMS‐like tyrosine kinase 3 (FLT3), a receptor tyrosine kinase expressed on early hematopoietic progenitors [4]. The majority of FLT3 mutations are internal tandem duplications (ITDs) within the juxtamembrane domain of the receptor, found in approximately 20%–25% of patients [5]. FLT3‐ITD mutations lead to constitutive ligand‐independent activation of the FLT3 receptor, promoting leukemogenesis through enhanced cellular proliferation and inhibition of apoptosis via downstream signaling pathways such as RAS/MAPK, PI3K/AKT, and STAT5 [4, 6, 7]. FLT3‐ITD mutations are associated with poor clinical outcomes in AML, including shorter durations of remission, higher relapse rates, and inferior overall survival (OS) compared with FLT3 wild‐type disease [8, 9]. Quizartinib is a selective type II FLT3 inhibitor that binds to the inactive conformation of the FLT3 receptor, thereby preventing autophosphorylation and downstream signaling through pathways such as MAPK, PI3K/AKT, and STAT5 [10, 11]. Compared with other FLT3 inhibitors, quizartinib and its active metabolite AC886 exhibit high binding affinity and kinase selectivity, with minimal off‐target activity observed across a wide panel of kinases and cellular targets [10]. In Japan, quizartinib was initially approved as monotherapy for patients with relapsed or refractory FLT3‐ITD–positive AML [12]. Its approval in this indication was based on the results of a phase 2 bridging trial in Japan [13] and the global phase 3 QuANTUM‐R trial, in which quizartinib offered significant survival benefits versus salvage chemotherapy and displayed a tolerable safety profile [14]. The phase 3 QuANTUM‐First trial has similarly demonstrated the efficacy and safety of quizartinib, administered in combination with chemotherapy and followed by maintenance monotherapy, in newly diagnosed patients with FLT3‐ITD–positive AML [15]. Despite the clinical activity of quizartinib in FLT3‐ITD–positive AML, the mechanisms underlying resistance to quizartinib have not been completely defined. Available evidence suggests that resistance frequently involves the emergence of secondary FLT3‐tyrosine kinase domain (TKD) mutations, particularly at D835 and F691, which impair quizartinib binding [16, 17]. However, these findings are based on limited numbers of patients and focused primarily on FLT3 mutations detected through targeted sequencing. Broader mutational events that may contribute to resistance, either within FLT3 or involving alternative pathways, have not been comprehensively investigated. To this end, the present study was aimed at obtaining further insights into the mechanisms of resistance to quizartinib, in addition to evaluating efficacy and safety outcomes, in patients receiving quizartinib monotherapy for relapsed or refractory FLT3‐ITD–positive AML. Material and Methods Study Design This was an open‐label, single‐arm, interventional study conducted at 11 centers in Japan between May 2020 and March 2025. The study was conducted in accordance with the Declaration of Helsinki, the Clinical Research Act of Japan, and all applicable regulatory guidelines. The study protocol was reviewed and approved by the Kyushu University Certified Institutional Review Board for Clinical Trials, and all patients provided written informed consent to participate. The study was registered with the Japan Registry of Clinical Trials (jRCTs071200015). Study Population Eligible patients were adults aged ≥ 20 years with a diagnosis of relapsed or refractory AML according to the 2017 World Health Organization (WHO) classification, and an Eastern Cooperative Oncology Group performance status (ECOG PS) of 0–2. Patients were required to have FLT3‐ITD positivity confirmed using an approved companion diagnostic after study enrollment, unless FLT3‐ITD positivity had been confirmed following their most recent therapy. Key exclusion criteria were prior treatment with quizartinib, a diagnosis of acute promyelocytic leukemia, and any condition that precluded bone marrow aspiration at the time of enrollment. Patients with multiple malignancies other than AML were generally excluded. Study Treatment Quizartinib was administered orally in accordance with its approved prescribing information [12]. Quizartinib was administered for up to 12 cycles of 28 days each. The treatment period was defined as the time from the first dose of quizartinib until the end of treatment, or until the final dose in those who discontinued quizartinib due to relapse, refractory disease, hematopoietic stem cell transplantation (HSCT), adverse events (AEs), or physician discretion. The follow‐up period was defined as the 12‐month period starting from the day after the last dose of quizartinib. During the follow‐up period, patients were permitted to undergo HSCT at the physician's discretion, or could receive treatment with any anti‐leukemic agent, including continuation of quizartinib. Study Assessments Disease assessments included peripheral blood and bone marrow examinations, which were performed at screening and throughout the treatment period. In particular, bone marrow assessments were performed before cycle 1, cycle 2, cycle 3, every second cycle thereafter, and at the end of treatment. Additional evaluations could be conducted at the investigator's discretion, particularly in cases of suspected disease progression or relapse. Genetic profiling was conducted using the Diagnostic Sequencing of Cancer Associated Variants (DISCAVar) panel, a custom next‐generation sequencing (NGS) panel developed by Kyushu University (Fukuoka, Japan). The detailed methodology has been described previously [18]. Clonal evolution was evaluated by single‐cell DNA sequencing (scDNAseq) using the Tapestri Single‐cell DNA 20‐gene AML panel (Labcorp, Burlington, North Carolina, USA) [19]. Measurable residual disease (MRD) was assessed using NGS and flow cytometry (FCM). NGS assessments of MRD included evaluations of FLT3‐ITD and mutant NPM1 conducted by Invivoscribe Inc. (San Diego, California, USA), with MRD negativity defined as a variant allele frequency (VAF) less than the lower limit of detection. FCM assessments of MRD were performed using previously described methods [20]. Non‐hematological AEs of grade ≥ 3 severity were recorded from the time of quizartinib initiation until 30 days after the last dose. AEs were documented and graded according to the National Cancer Institute's Common Terminology Criteria for Adverse Events (CTCAE version 5.0). Study Endpoints The primary endpoint was to evaluate the type and rate of gene mutations acquired or lost at relapse or refractory disease. Secondary endpoints included the rate of composite complete remission (CRc), time to CRc, duration of CRc, overall response rate (ORR), rate of HSCT, relapse‐free survival (RFS) following HSCT, OS, and the incidence of grade ≥ 3 non‐hematologic AEs and QT prolongation. FLT3‐ITD MRD in patients who achieved CRc was assessed as an exploratory endpoint. Best response was defined as the overall best‐measured response across all time points during the quizartinib treatment period (i.e., excluding responses to any subsequent AML therapy or quizartinib continuation). Treatment responses included complete remission (CR), CR with incomplete neutrophil recovery (CRi), CR with incomplete platelet recovery (CRp), partial remission (PR), or no response, based on sponsor‐modified International Working Group criteria (Table S1). The CRc rate was defined as the proportion of patients who achieved a best response of CR, CRi, or CRp; time to CRc was defined as the interval from the first dose of quizartinib to the first achievement of CRc. Duration of CRc was calculated from the first report of CRc to the date of relapse or death from any cause; patients without relapse were censored at the time they were last known to be relapse‐free or when subsequent treatment (excluding HSCT) was initiated. ORR was defined as the proportion of patients who achieved CRc or PR. The rate of HSCT was defined as the proportion of patients who proceeded directly from protocol‐specified quizartinib treatment to allogeneic HSCT without additional anti‐leukemic therapy. RFS was assessed in the subset of patients who underwent HSCT directly after protocol‐specified quizartinib treatment and was defined as the time from transplantation to relapse or death from any cause. OS was defined as the time from the first dose of quizartinib to death from any cause. Statistical Analysis The target sample size was set at 51 patients based on feasibility considerations at the time of protocol development. In previous studies, FLT3‐TKD mutations were observed in six out of eight patients (75%) with acquired resistance to quizartinib [16, 17]. In the QuANTUM‐R trial [14], approximately 65% of patients treated with quizartinib were evaluable for resistance mechanisms due to treatment discontinuation from relapse, refractory disease, or intolerance. Applying this rate to the planned sample size of 51 patients would yield approximately 33 patients evaluable for the primary endpoint, which would allow the prevalence of FLT3‐TKD mutations to be estimated with a two‐sided 95% confidence interval (CI) within ±15%. Safety outcomes were analyzed in the safety analysis population, comprising all enrolled patients who received ≥ 1 dose of quizartinib. The efficacy analysis population included all eligible patients who received quizartinib and had evaluable efficacy data before and after quizartinib treatment. Unless otherwise stated, baseline patient characteristics and study outcomes were summarized using descriptive statistics, including mean, standard deviation, median, and range values for continuous variables, and the number and proportion of patients for categorical variables. The primary endpoint analysis population included all eligible patients who could be evaluated for gene mutations before initiating protocol‐specified treatment and after the discontinuation of quizartinib. For the primary endpoint, the proportion of patients who acquired or lost mutations known to be pathogenic was calculated, with two‐sided 95% CIs estimated using the Clopper–Pearson method. For secondary endpoints, the rate of CRc and ORR was summarized using 95% CIs calculated using the Clopper–Pearson method. Time‐to‐event outcomes were analyzed using the Kaplan–Meier method, with 95% CIs calculated using Greenwood's formula. In the analysis of time to CRc, death was treated as a competing risk. In Kaplan–Meier analyses of RFS and OS, patients were censored at the last date of documented relapse‐free or survival status, respectively. Statistical analyses were conducted using SAS version 9.4, R version 4.2.2 for MRD evaluated by NGS, and Tapestri Insights version 2.2 for scDNAseq data. Analyses were based on observed data, with no imputation for missing values. Results Study Population Between May 2020 and December 2022, 18 patients with relapsed or refractory FLT3‐ITD–positive AML were screened for eligibility (Figure S1). After excluding three patients due to FLT3‐ITD–negative status, 15 patients initiated protocol‐specified quizartinib treatment and comprised the safety and efficacy analysis populations. Quizartinib treatment was discontinued in four, four, and three patients due to refractory disease, relapse, and to proceed to allogeneic HSCT, respectively. However, bone marrow samples at the end of protocol‐specified quizartinib treatment were unavailable for Patients 1 and 2 who proceeded to HSCT, and for Patients 3 and 17 at the time of disease assessment; therefore, seven patients with paired bone marrow samples at baseline and at the end of quizartinib treatment were included in the primary endpoint analysis population. Due to slow patient enrollment, the target sample size was not reached and prespecified analyses to determine the rate of acquired genetic mutations (primary endpoint) were not performed. As a result, the endpoints reported herein should be interpreted as exploratory and hypothesis‐generating. At baseline, the median age of the safety analysis population was 74.0 years (range, 20–88) and most patients (80.0%) were aged ≥ 60 years (Table 1). Eleven patients (73.3%) were male, and 13 (86.7%) had an ECOG PS of 0–1. The median proportion of blasts in bone marrow was 53.7% (range, 3.0–79.6). TABLE 1Baseline patient characteristics.CharacteristicSafety/efficacy analysis population (n = 15)Primary endpoint analysis population (n = 7)Age, yearsMean (standard deviation)68.7 (19.0)66.0 (26.0)Median (range)74.0 (20–88)79.0 (20–85)< 60, n (%)3 (20.0)2 (28.6)≥ 60, n (%)12 (80.0)5 (71.4)Male sex, n (%)11 (73.3)6 (85.7)ECOG PS, n (%)0–113 (86.7)7 (100.0)≥ 22 (13.3)0 (0.0)Proportion of blast cells in bone marrow, % a Mean (standard deviation)43.7 (29.1)42.4 (26.2)Median (range)53.7 (3.0–79.6)49.6 (6.0–76.2)AML type, n (%) De novo AML15 (100)7 (100)Secondary AML00Antecedent hematologic disorders, n (%)No12 (80.0)5 (71.4)Yes3 (20.0)2 (28.6)MDS3 (20.0)2 (28.6)Other00Diagnosis (WHO 2017), n (%)AML with t (6;9)(p23;q34.1); DEK‐NUP214 1 (6.7)0AML with inv. (3) (q21.3q26.2) or t (3;3) (q21.3;q26.2); GATA2, MECOM 1 (6.7)1 (14.3)AML with mutated NPM1 4 (26.7)0AML following MDS or MDS/MPD3 (20.0)2 (28.6)AML without maturation1 (6.7)0Acute myelomonocytic leukemia2 (13.3)2 (28.6)Acute monoblastic/monocytic leukemia1 (6.7)1 (14.3)AML, NOS1 (6.7)1 (14.3)MPAL, T/myeloid, NOS1 (6.7)0Relapsed or refractory disease, n (%)Refractory AML11 (73.3)6 (85.7)Relapsed AML4 (26.7)1 (14.3)Period from diagnosis to enrolment, daysMean (standard deviation)128.5 (215.7)168.6 (308.1)Median (range)37 (14–857)36 (20–857) FLT3‐ITD insertion sites, n (%)JMD‐TKD5 (33.3)3 (42.9)JMD3 (20.0)2 (28.6)JMZ1 (6.7)0JMZ‐hinge‐β11 (6.7)1 (14.3)JMS‐JMZ‐hinge1 (6.7)0 (0.0)Unknown4 (26.7)1 (14.3)Abbreviations: AML, acute myeloid leukemia; ECOG PS, Eastern Cooperative Oncology Group performance status; FLT3, FMS‐like tyrosine kinase 3; GATA2, GATA‐binding protein 2; ITD, internal tandem duplication; JMD, juxtamembrane domain; JMS, juxtamembrane segment; JMZ, juxtamembrane zone; MDS, myelodysplastic syndrome; MECOM, myelodysplasia syndrome 1 and ecotropic viral integration site 1 complex locus; MPAL, mixed phenotype acute leukemia; MPD, myeloproliferative disorder; NOS, not otherwise specified; NPM1, nucleophosmin 1; NUP214, nucleoporin 214; TKD, tyrosine kinase domain; WHO, World Health Organization. a Data based on n = 14 in the safety/efficacy analysis population and n = 7 in the primary endpoint analysis population (one patient was excluded due to dry tap). Baseline patient characteristics. Abbreviations: AML, acute myeloid leukemia; ECOG PS, Eastern Cooperative Oncology Group performance status; FLT3, FMS‐like tyrosine kinase 3; GATA2, GATA‐binding protein 2; ITD, internal tandem duplication; JMD, juxtamembrane domain; JMS, juxtamembrane segment; JMZ, juxtamembrane zone; MDS, myelodysplastic syndrome; MECOM, myelodysplasia syndrome 1 and ecotropic viral integration site 1 complex locus; MPAL, mixed phenotype acute leukemia; MPD, myeloproliferative disorder; NOS, not otherwise specified; NPM1, nucleophosmin 1; NUP214, nucleoporin 214; TKD, tyrosine kinase domain; WHO, World Health Organization. Data based on n = 14 in the safety/efficacy analysis population and n = 7 in the primary endpoint analysis population (one patient was excluded due to dry tap). All patients had de novo AML; no cases of secondary AML were reported (Table 1). Three patients (20.0%) had a history of antecedent hematologic disorders, all of which were myelodysplastic syndrome (MDS). Mutation status at study entry (relapse/refractory stage) was assessed by targeted sequencing and is summarized in a supplementary Table S2. Eleven patients (73.3%) were refractory to prior therapies and four (26.7%) had relapsed, and none had previously received gilteritinib. The time from diagnosis to study enrollment varied widely, with a median of 37 days (range, 14–857). FLT3‐ITD insertion site analysis indicated that five patients (33.3%) harbored insertions spanning the juxtamembrane domain and TKD, while three patients (20.0%) had insertions restricted to the juxtamembrane domain (Table 1). Quizartinib and Subsequent Treatment The median duration of quizartinib treatment in the safety analysis population was 75 days (range, 28–336) and the median number of treatment cycles was 4 (range, 1–12; Table S3). The initial quizartinib dose was 17.7 mg in eight patients (53.3%) and 26.5 mg in seven (46.7%). Three patients (20.0%) completed the full 12‐cycle course of quizartinib treatment (Figure 1 and Table S3). Reasons for treatment discontinuation among the 12 patients who prematurely discontinued quizartinib were relapse (n = 4, 33.3%), refractory disease (n = 4, 33.3%), allogeneic HSCT (n = 3, 25.0%), and physician's decision (n = 1, 8.3%). Treatment duration, responses, and follow‐up in individual patients (safety/efficacy analysis population). AraC, cytarabine; AZA, azacitidine; C, cycle; CR, complete remission; CRc, composite complete remission; CRi, complete remission with incomplete neutrophil recovery; CRp, complete remission with incomplete platelet recovery; DNR, daunorubicin; HSCT, hematopoietic stem cell transplantation; FLAGM, fludarabine, cytarabine, granulocyte colony‐stimulating factor, and mitoxantrone; NR, no response; PD, progressive disease; PR, partial remission; SD, stable disease; VEN, venetoclax. After protocol‐specified quizartinib treatment, all 15 patients proceeded to the follow‐up period. The first subsequent therapies received were the continuation of quizartinib (n = 4, 26.7%), gilteritinib (n = 3, 20.0%), venetoclax‐containing regimens (n = 3, 20.0%), allogeneic HSCT (n = 3, 20.0%), chemotherapy (n = 1, 6.7%), and best supportive care (n = 1, 6.7%). Mutation Analysis At the time of quizartinib initiation, all seven patients in the primary endpoint analysis population had several mutations in addition to FLT3 (Figure 2). These included mutations in genes associated with epigenetic regulation (TET2, DNMT3A, BCOR), RNA splicing (SRSF2, SF3B1), and transcriptional control (RUNX1, NFE2, ETNK1). Landscape of gene mutations before and after quizartinib treatment (primary endpoint analysis population). The bottom three rows indicate the reasons for discontinuation of protocol‐specified quizartinib treatment (refractory disease, relapse, or allogeneic HSCT). FLT3, FMS‐like tyrosine kinase 3; ITD, internal tandem duplication; TKD, tyrosine kinase domain. At the end of the treatment period, acquired mutations were detected in four patients (57.1%; Figure 2). These included NF1 (R2616X; VAF 9%) in patient 4, CSF3R (Q754X; VAF 15%) in patient 6, NRAS (G13R; VAF 51%) in patient 10, and FLT3 (D835Y; VAF 8%) in patient 13. Loss of FLT3‐ITD was observed in two patients (4 and 6). No mutations were acquired or lost in the remaining three patients. Additional scDNAseq analyses were performed in eight patients to evaluate clonal dynamics during and after quizartinib treatment (Figure 3). While scDNAseq detected clones with mutations not identified by the DISCAVar panel (e.g., PTPN11 [patient 4], WT1 [patient 12], FLT3‐TKD [patient 15], SF3B1 [patient 16]), other mutations detected by DISCAVar were not loaded on the scDNAseq panel (e.g., NF1 [patient 4], CSF3R [patient 6]). In patients who lost FLT3‐ITD according to DISCAVar (i.e., patients 4 and 6), no clones with FLT3‐ITD were detected by scDNAseq at the end of treatment; in patients with FLT3‐ITD detected by DISCAVar at the start and end of treatment (i.e., patients 10, 12, and 16), scDNAseq found that the proportion of FLT3‐ITD clones by the end of treatment was reduced. These patients showed a relative increase in clones without FLT3‐ITD, and clones with mutations in TET2 (patient 4), TET2 (patient 6), NRAS (patient 10), WT1 (patient 12), and SF3B1 (patient 16). In patients 3 and 15, clones with FLT3‐ITD remained unchanged after treatment. In patient 15, multiple subclones that acquired FLT3‐TKD appeared in clones with or without FLT3‐ITD at the time of relapse. Although it was difficult to detect FLT3‐ITD by scDNAseq in patient 13 due to technical reasons, a clone with FLT3‐TKD was first detected at the time of relapse. Clonal evolution during and after quizartinib treatment assessed using scDNAseq. Fish plots represent relative clonal composition of mutation‐bearing clones; the “no mutation” fraction is not shown. Bone marrow blast percentage is indicated at each sampling timepoint. Days since treatment initiation are shown on the horizontal axis. FLT3, FMS‐like tyrosine kinase 3; ITD, internal tandem duplication; P, cycles following protocol‐specified quizartinib treatment; Ref, refractory disease; Rel, relapse; scDNAseq, single‐cell DNA sequencing; T, months following hematopoietic stem cell transplantation; TKD, tyrosine kinase domain; TT, months following second hematopoietic stem cell transplantation. Efficacy Outcomes In the efficacy analysis population, the CRc rate during the treatment period was 66.7% (95% CI, 38.4–88.2; Table 2). The ORR was 73.3% (95% CI, 44.9–92.2), including three patients (20.0%) with CR, five (33.3%) with CRi, two (13.3%) with CRp, and one (6.7%) with PR. The median time to CRc was 2.0 months (95% CI, 1.0–not evaluable; Figure 4A), and the median duration of CRc was 19.5 months (95% CI, 1.6–not evaluable). At the time of follow‐up, six of 15 patients (40.0%) maintained CRc. TABLE 2Best response during the treatment period.Response, n (%) [95% CI]Efficacy analysis population (n = 15)CR3 (20.0)CRi5 (33.3)CRp2 (13.3)PR1 (6.7)NR4 (26.7)CRc (CR + CRi + CRp)10 (66.7) [38.4–88.2]ORR (CRc + PR)11 (73.3) [44.9–92.2]Abbreviations: CI, confidence interval; CR, complete remission; CRc, composite complete remission; CRi, complete remission with incomplete neutrophil recovery; CRp, complete remission with incomplete platelet recovery; NR, no response; ORR, overall response rate; PR, partial remission. Best response during the treatment period. Abbreviations: CI, confidence interval; CR, complete remission; CRc, composite complete remission; CRi, complete remission with incomplete neutrophil recovery; CRp, complete remission with incomplete platelet recovery; NR, no response; ORR, overall response rate; PR, partial remission. Kaplan–Meier plots of (A) time to CRc and (B) OS (efficacy analysis population). CRc, composite complete remission; OS, overall survival. Three patients (20.0%) in the efficacy analysis population underwent allogeneic HSCT after protocol‐specified quizartinib treatment. Of these three patients (Patients 1, 2, and 12), two achieved CRc and proceeded to HSCT after one and two cycles of quizartinib, respectively, whereas the remaining patient had no response and underwent HSCT after two cycles. The median time from the start of quizartinib treatment to transplantation was 75 days (range, 64–75). In the primary endpoint analysis population, one patient (14.3%) received allogeneic HSCT 64 days after the start of quizartinib treatment. Among the three patients in the efficacy analysis population who received allogeneic HSCT, median RFS was 8.5 months (95% CI, 6.2–not evaluable). Median OS in the efficacy analysis population was 13.6 months (95% CI, 5.4–not evaluable; Figure 4B). The 12‐month OS rate was 53.3% (95% CI, 26.3–74.4). Measurable Residual Disease In NGS analyses of MRD, FLT3‐ITD MRD negativity was achieved in two out of five patients who maintained CRc after quizartinib treatment (Patients 2 and 7; Figure S2A). Among the four patients who achieved CRc but subsequently relapsed, FLT3‐ITD MRD negativity at relapse was observed in one patient (Patient 6; Figure S2B), who also had loss of FLT3‐ITD during treatment confirmed by the DISCAVar panel and scDNAseq. In addition, the course of mutant NPM1 MRD was evaluated in a single patient who harbored this mutation at baseline (Patient 7; Figure S2C). In this patient, mutant NPM1 became undetectable after the tenth cycle of quizartinib and remained below the detection threshold thereafter. In FCM assessments of MRD using T‐cell immunoglobulin mucin‐3 (TIM‐3) as a marker of leukemic stem cell (LSC) burden, patients who achieved CRc showed a reduction in LSCs, regardless of whether the denominator was the number of CD34+CD38−/low or CD45dim cells (Figure S3). An initial decline in LSC levels was generally followed by re‐expansion at relapse; this observation was true for patient 6, who had FLT3‐ITD MRD negativity in NGS analyses. Conversely, sustained clinical response was typically accompanied by continued suppression of LSCs. Safety Outcomes Non‐hematologic AEs of grade ≥ 3 severity were reported in three patients (20.0%) in the safety analysis population. These included one patient (6.7%) with organizing pneumonia and COVID‐19 (both grade 3), one patient (6.7%) with grade 3 febrile neutropenia, and one patient (6.7%) with grade 4 transaminase elevation. QT prolongation was observed in three patients (20.0%) during protocol‐specified quizartinib treatment, all cases of which were grade 1 in severity. No serious AEs were observed, and there were no AEs leading to treatment discontinuation or death. Discussion This multicenter study investigated mechanisms of quizartinib resistance in patients with relapsed or refractory FLT3‐ITD–positive AML. Although the small sample size precluded an estimation of the rate of acquired genetic mutations (primary endpoint), mutation analyses, using both targeted panel sequencing and scDNAseq, nevertheless provided hypothesis‐generating insights into clonal evolution over the course of quizartinib treatment. Targeted panel sequencing revealed that one patient acquired the FLT3 (D835Y) mutation after quizartinib treatment, which has been previously observed in a small subset of patients at relapse [21, 22]. In this patient, scDNAseq revealed expansion of a non‐mutational clone, while the FLT3 (D835Y) mutation represented a minor clone. In contrast, the phase 2 QUIWI trial of quizartinib in newly‐diagnosed FLT3‐ITD–negative AML demonstrated clinical benefit in patients harboring FLT3‐TKD mutations [23]. Considering the limited sample sizes across studies, further investigation is needed. One patient each acquired NRAS (G13R) and NF1 (R2616X) mutations. An NRAS mutation at G13 is a hotspot mutation known to constitutively activate NRAS, while loss‐of‐function mutations in NF1 lead to constitutive activation of RAS signaling [24]. RAS pathway mutations are often associated with resistance to FLT3 inhibitors [25], as they can drive parallel signaling that diminishes reliance on FLT3 [26]. One patient acquired a CSF3R mutation, which is commonly associated with chronic neutrophilic leukemia or secondary AML [27]. Molecular profiling of gilteritinib resistance revealed CSF3R mutations in a subset of patients, although their functional contribution to resistance is unclear [28]. Two patients lost FLT3‐ITD clones at the end of quizartinib treatment, which is a phenomenon previously reported with other FLT3 inhibitors. In the ADMIRAL trial of gilteritinib, approximately 22.9% of relapsed patients had no detectable FLT3 mutation at relapse despite FLT3‐ITD–positivity determined initially [29]. Similarly, an analysis of patients who relapsed after midostaurin showed loss of FLT3‐ITD in 46% of cases [30]. This is believed to occur because FLT3 inhibitors apply strong selective pressure, and subclones that lack FLT3‐ITD have a survival advantage under drug treatment [5]. In our study, the two patients who lost FLT3‐ITD both acquired alternative driver mutations (NF1 and CSF3R). One of these patients exhibited FLT3‐ITD MRD negativity at relapse in NGS analyses. In the present study, mutational changes appeared to differ by clinical course. The two patients who acquired RAS pathway–related mutations (NF1 or NRAS) did not achieve a response, whereas the two patients who acquired other mutations (FLT3‐TKD or CSF3R) initially achieved CRc but subsequently relapsed. The duration of CRc was 173 days in the patient with a CSF3R mutation (Patient 6) and 84 days in the patient with a FLT3‐TKD mutation (Patient 13). However, given the limited sample size, these observations remain hypothesis‐generating and warrant validation in larger cohorts. Previous studies have reported a correlation between FLT3‐ITD MRD depth and prognosis, suggesting that prognosis can be stratified by 1 × 10−4 borderline or undetectable MRD [31, 32, 33]. In this study, MRD was not detected or < 1 × 10−4 in patients who achieved sustained remission. FLT3‐ITD is a genetic alteration acquired in the later stages of AML development and is known to be clonally eliminated by treatment [34]. On the other hand, it has been reported that MRD, which is an indicator of gene mutations such as NPM1 that occur in the early stages of disease onset, correlates with prognosis and increases before relapse [35]. In this study, only one patient with an NPM1 mutation had MRD measured, so it is not possible to confirm the relationship between FLT3‐ITD and NPM1 MRD. In addition, time points for FLT3‐ITD MRD assessment were limited in this study, and further studies are needed to determine the optimal frequency of MRD assessments in order to predict recurrence. In this study, the proportion of NPM1 positivity at registration was 27% (four of 15 patients), which was lower than previous reports showing that approximately half of patients with FLT3‐ITD–positive AML are NPM1‐positive [1, 15, 36]. Because NPM1 mutation status at initial diagnosis was not assessed in this study and prior therapies before enrollment were heterogeneous, the observed mutation profile at study entry may reflect clonal selection driven by pre‐enrollment treatment. Additionally, the limited number of patients in this exploratory study precluded evaluation of any association between NPM1 mutation status and the efficacy of quizartinib. We performed a longitudinal observational analysis in which mutation profiling by targeted panel sequencing and scDNAseq was conducted alongside NGS‐ and FCM‐based MRD assessments in patient samples obtained during quizartinib treatment. This design allowed for the descriptive evaluation of mutational changes and MRD‐related parameters at both genetic and immunophenotypic levels over time. The longitudinal mutation profiles and parallel MRD measurements provided contextual information on the temporal relationship between clonal changes, resistance‐associated mutations, and disease status. The potential implications of such observations for MRD monitoring strategies, including target selection for NGS‐based panels and the combined use of NGS and FCM for MRD assessment, remain to be clarified. Sakoda and colleagues previously reported an association between TIM‐3–positive LSCs and prognosis after HSCT [20]. Although mutational analysis of TIM‐3–positive cells was not performed in the present study, we and others have previously demonstrated that TIM‐3 serves as a functional marker of leukemic stem cells, and that pathogenic variants detected in bulk leukemic cells originate from TIM‐3–positive leukemic stem cells [20, 37]. Based on these findings, it is reasonable to infer that the pathogenic variants identified in the present study were also acquired at the leukemic stem cell level. In this study, variations in TIM‐3–positive LSC burden were observed over the course of remission and relapse. The clinical relevance of integrating longitudinal mutational data with TIM‐3–guided FCM assessment for risk stratification and MRD evaluation will require further investigation in larger independent cohorts. In this study, quizartinib demonstrated encouraging efficacy in a heavily pretreated, predominantly older population with relapsed or refractory FLT3‐ITD–positive AML. This is consistent with the efficacy reported in earlier trials of quizartinib. In the phase 3 QuANTUM‐R trial, single‐agent quizartinib achieved a CRc rate of 47% and a median response duration of 4 months in relapsed or refractory FLT3‐ITD AML [14]. In the present study, median OS (13.6 months) was longer than that reported in QuANTUM‐R (6.2 months). This difference may reflect differences in patient populations, supportive care practices, and the availability of subsequent therapies. In particular, compared with QuANTUM‐R, our cohort may have included patients with a relatively more favorable disease trajectory, and patients in our study had access to venetoclax‐ and gilteritinib‐based salvage regimens after quizartinib, which may have contributed to longer post‐progression survival. In a Japanese phase 2 study of quizartinib, the CRc rate was 54% and the median CRc duration was 16 weeks [13]. Despite the older age and lower transplant rate in our cohort, efficacy outcomes were comparable, supporting the utility of quizartinib monotherapy in real‐world clinical practice. The safety profile of quizartinib in the present study was consistent with prior reports. Grade ≥ 3 non‐hematologic AEs were infrequent, and there were no treatment‐emergent deaths or unexpected toxicities. QT prolongation was reported in three patients (20%) in this study; however, all were grade 1 in severity. This finding is consistent with the safety results of the QuANTUM‐R trial, in which the incidence of grade 1–2 QT prolongation with quizartinib was 22%, and grade ≥ 3 QT prolongation was uncommon (4%) [14]. A Japanese phase 2 study of quizartinib similarly reported that while mild QT prolongation occurred in approximately one‐third of patients, no torsade de pointes or fatal arrhythmias were observed [13]. The main limitation of the present study was the low number of included patients, particularly those evaluable for the primary endpoint. Although it was originally planned to enroll 51 patients, the study encountered slow accrual and closed with 15 treated patients, seven of whom had paired samples for sequencing. One major barrier to patient enrollment may have been the availability of gilteritinib, which was approved in Japan for relapsed or refractory FLT3‐mutant AML one year prior to quizartinib. Moreover, this study was conducted in Japan, included patients who could undergo bone marrow collection, and mostly enrolled patients with good performance status which may limit the generalizability of our results to other regions and patient populations. Some mutations detected by targeted panel sequencing were outside the AML panel target regions captured by the scDNAseq platform, whereas the DISCAVar panel may be limited by its detection sensitivity. The NGS‐based and FCM‐based MRD assays differ in their requirements for cell numbers and in the timing of sample collection. Consequently, particularly during remission when only a limited number of cells are available, the patients and timepoints for which results could actually be obtained varied between the two modalities. This heterogeneity represents a substantial barrier to a rigorous comparison between the two MRD approaches. Pharmacokinetic and pharmacodynamic data were not collected in this study; therefore, it remains unclear whether quizartinib plasma concentrations reached therapeutic levels in patients who experienced relapse or refractory disease. Finally, the heterogeneity of prior treatments among enrolled patients may have influenced treatment outcomes and the emergence of resistance mutations, representing a potential confounding factor. In conclusion, this study provides exploratory, hypothesis‐generating insights into mutational changes during quizartinib therapy in patients with FLT3‐ITD–positive relapsed or refractory AML. The combined use of targeted panel sequencing and scDNAseq allowed exploratory assessments of both on‐target (FLT3‐D835Y) and off‐target (NRAS, CSF3R, NF1) mutations in serial bone marrow samples obtained throughout the course of quizartinib treatment. Although the clinical implications remain to be fully established, longitudinal analyses of gene mutation profiles may help to inform future investigations aimed at refining MRD assessment approaches, including the potential selection of target genes for combined NGS‐ and FCM‐based MRD monitoring. Safety and efficacy findings in this study were consistent with previous clinical trials and support the continued use of quizartinib in selected patients. Further studies with larger cohorts are warranted to better characterize resistance patterns and to guide personalized treatment strategies for patients with FLT3‐mutated AML. Author Contributions Yuichiro Semba: investigation, methodology, writing – original draft, writing – review and editing. Toshihiro Miyamoto: conceptualization, funding acquisition, resources, supervision, writing – original draft, writing – review and editing. Senji Kasahara: resources, writing – review and editing. Yoshikane Kikushige: conceptualization, investigation, methodology, supervision, writing – review and editing. Takahiro Maeda: conceptualization, supervision, writing – review and editing. Goichi Yoshimoto: resources, writing – review and editing. Shuichi Ota: resources, writing – review and editing. Akio Saito: resources, writing – review and editing. Yuho Najima: resources, writing – review and editing. Hiroki Hosoi: resources, writing – review and editing. Daigo Niiya: resources, writing – review and editing. Koji Kato: resources, writing – review and editing. Koji Nagafuji: resources, writing – review and editing. Takashi Akasaka: resources, writing – review and editing. Tomohiko Kamimura: resources, writing – review and editing. Atsushi Takita: methodology, writing – review and editing. Maiko Narahara: formal analysis, visualization, writing – review and editing. Tadashi Toki: formal analysis, visualization, writing – review and editing. Koichi Iwanaga: methodology, writing – original draft, writing – review and editing. Koji Yonemoto: formal analysis, resources, writing – review and editing. Mine Harada: conceptualization, funding acquisition, supervision, writing – review and editing. Koichi Akashi: conceptualization, funding acquisition, supervision, writing – review and editing. Funding This study was funded by Daiichi Sankyo Co. Ltd. The sponsor participated in the design of the study, analysis and interpretation of biomarker data, and writing the report. Daiichi Sankyo did not participate in the analysis or interpretation of clinical data. Ethics Statement The research protocol for this study was reviewed and approved by the Kyushu University Certified Institutional Review Board for Clinical Trials, and all patients provided written informed consent to participate. This study was registered with the Japan Registry of Clinical Trials (jRCTs071200015). Animal Studies, N/A. Conflicts of Interest T. Miyamoto reports honoraria from AbbVie Inc., Amgen Inc., Asahi Kasei Pharma Co., Astellas Pharma Inc., AstraZeneca K.K., Bristol‐Myers Squibb Co., Daiichi Sankyo Co. Ltd., Janssen Pharmaceutical K.K., Kyowa Kirin Co., Nippon Shinyaku Co., Novartis Pharmaceuticals K.K., and Otsuka Pharmaceutical Co.; and donations from Chugai Pharmaceutical Co. and Kyowa Kirin Co. K.A. reports honoraria from AbbVie Inc., Asahi Kasei Pharma Co., Bristol‐Myers Squibb Co., Chugai Pharmaceutical Co., Janssen Pharmaceutical K.K., Kyowa Kirin Co., and Nippon Shinyaku Co.; and speaker fees from AbbVie Inc., Kyowa Kirin Co., and LSI Medience Co. S.K. reports honoraria from AbbVie Inc. and Daiichi Sankyo Co. Ltd.; and research fees from Astellas Pharma Inc. and Novartis Pharmaceuticals K.K. Y.K. reports honoraria from Astellas Pharma Inc., Janssen Pharmaceutical K.K., and Kyowa Kirin Co.; and research fees from Astellas Pharma Inc. T. Maeda reports honoraria from Astellas Pharma Inc. and Otsuka Pharmaceutical Co.; and research fees from Takeda Pharmaceutical Co. and Xcoo Inc. G.Y. reports honoraria from AbbVie Inc., Amgen Inc., Astellas Pharma Inc., and Daiichi Sankyo Co. Ltd. A.S. reports honoraria from Eisai Co.; and research fees from Argenx Japan K.K. and Novartis Pharmaceuticals K.K. K.K. reports honoraria from AbbVie Inc., Chugai Pharmaceutical Co., and Gilead Sciences Inc.; and research fees from AbbVie Inc., Bristol‐Myers Squibb Co., Chugai Pharmaceutical Co., Daiichi Sankyo Co. Ltd., Gilead Sciences Inc., Novartis Pharmaceuticals K.K., and Ono Pharmaceutical Co. K.N. reports honoraria from Amgen Inc. and Kyowa Kirin Co. T.K. reports honoraria from AbbVie Inc., Janssen Pharmaceutical K.K., and Ono Pharmaceutical Co. A.T., M.N., T.T., and K.I. are employees of Daiichi Sankyo Co. Ltd. All other authors declare no conflicts of interest. Supporting information Figure S1: Patient disposition. Figure S2: Longitudinal FLT3‐ITD and NPM1 MRD dynamics assessed using NGS: (A) FLT3‐ITD MRD in patients who maintained CRc; (B) FLT3‐ITD MRD in patients with relapse; (C) NPM1 mutation MRD in a patient who achieved sustained remission. Figure S3: Longitudinal MRD assessed using FCM. Table S1: Definitions of treatment response. Table S2: Mutation status at study entry. Table S3: Quizartinib exposure and discontinuation.