FLT3 mutations are present in ~30% of newly diagnosed AML, most commonly as internal tandem duplications (FLT3-ITD; ~ 20–25%) and less commonly as tyrosine kinase domain (TKD) mutations (FLT3-TKD; ~ 5–10%) [6–15]. FLT3-ITD activates downstream proliferative and survival pathways, and is associated with leukocytosis, increased relapse risk, and inferior survival. Type I inhibitors, such as midostaurin, gilteritinib, and crenolanib, inhibit both FLT3-ITD and a broad range of TKD mutations, whereas type II inhibitors, such as quizartinib and sorafenib, are primarily active against FLT3-ITD and are more vulnerable to TKD-mediated resistance [1, 5, 7, 16]. Because FLT3 inhibitors are now well-established and have been extensively reviewed, this section does not aim to provide a comprehensive review of FLT3-targeted therapies. Instead, it highlights selected recent updates, including MRD-adapted maintenance, resistance-informed combination strategies, and the emerging rationale for FLT3–menin co-inhibition.

Midostaurin established the role of FLT3-directed therapy in newly diagnosed FLT3-mutated AML when added to intensive chemotherapy in the phase III RATIFY trial, improving overall survival (OS) and event-free survival (EFS) compared with chemotherapy alone [13]. Although post-remission FLT3 inhibition has been explored with midostaurin, the role of midostaurin maintenance remains less clearly defined, with post hoc and phase II data showing no definitive survival benefit attributable to maintenance alone [17, 18]. More recently, quizartinib improved OS in newly diagnosed patients with FLT3-ITD AML in the phase III QuANTUM-First trial when added to intensive induction and consolidation therapy followed by continuation therapy [1]. Exploratory analyses have suggested higher MRD conversion rates with quizartinib during consolidation or maintenance, supporting further evaluation of MRD-informed FLT3 inhibitor strategies [19, 20]. Beyond FLT3-mutated disease, the phase II QUIWI study reported the activity of quizartinib in FLT3-wild-type AML, providing a rationale for an ongoing randomized evaluation [21].

In relapsed/refractory (R/R) FLT3-mutated AML, gilteritinib remains a standard targeted option based on the phase III ADMIRAL trial, which showed improved OS and higher complete remission (CR)/complete remission with partial hematologic recovery (CRh) rates compared to salvage chemotherapy [5, 22]. These findings were supported by the phase III COMMODORE trial in a predominantly Asian population, where gilteritinib improved OS, EFS, and composite complete remission (CRc) rates compared with salvage chemotherapy, with fewer grade ≥ 3 adverse events [23]. However, the frontline combination of gilteritinib plus azacitidine did not improve OS despite higher remission rates, underscoring that FLT3-directed combinations require careful partner selection and prospective validation [24].

Post-transplant maintenance remains an important area of FLT3 inhibitor development. Sorafenib maintenance after allo-HCT improved relapse-free survival (RFS) in FLT3-ITD AML in randomized studies, with durable benefit reported on long-term follow-up [25, 26]. In the phase III MORPHO trial, gilteritinib maintenance did not significantly improve RFS in the overall intention-to-treat population. However, benefit was observed among patients with detectable FLT3-ITD MRD before or after transplantation, supporting an MRD-adapted approach to post-HCT FLT3 inhibition [4]. Post-transplant maintenance strategies using other regimens, including azacitidine plus venetoclax, are also under investigation [27]. Taken together, these data highlight the increasing importance of molecular monitoring in selecting patients who are most likely to benefit from maintenance therapy.

Next-generation and combination approaches are becoming increasingly relevant, particularly in older or intensive therapy-ineligible patients and in R/R disease. Crenolanib has activity against FLT3-ITD and multiple TKD resistance mutations, and early-phase studies combining crenolanib or gilteritinib with intensive chemotherapy have shown encouraging activity, supporting ongoing evaluation of more potent or selective FLT3 inhibitor-based regimens [28–30]. Hypomethylating agents (HMA)–venetoclax regimens can induce responses in FLT3-mutated AML; however, resistance may occur through persistent or emergent FLT3 signaling and downstream pathway activation [31–39]. Accordingly, FLT3 inhibitor-based doublets and triplets have been evaluated. Gilteritinib plus venetoclax demonstrated activity in R/R FLT3-mutated AML, with cytopenias managed through dose interruptions and supportive care [35]. Triplet regimens combining HMA, venetoclax, and FLT3 inhibitors have produced high response and MRD-negative rates in early-phase studies, although prolonged myelosuppression and dose modifications remain important limitations [36, 37, 40, 41].

A particularly important emerging area of research involves the combination of FLT3 and menin inhibition. This strategy is biologically relevant because FLT3 mutations frequently co-occur with NPM1 mutations, and menin-dependent transcriptional programs contribute to leukemic self-renewal in NPM1-mutated and KMT2Arearranged (KMT2Ar) AML. Preclinical studies support menin–FLT3 co-inhibition in AML models with FLT3 mutations, concurrent NPM1 mutations, and/or KMT2A rearrangements. In NPM1/FLT3-mutated models, multiagent approaches integrating menin inhibition with BCL-2 and FLT3 inhibition have shown activity against leukemic stem and progenitor compartments, supporting the ongoing clinical development of these combinations [42–44].

Safety considerations with FLT3 inhibitors include cytopenias, infections, QTc prolongation, hepatic enzyme abnormalities, and differentiation syndrome, particularly in combination regimens requiring dose interruptions or marrow-guided adjustments [13, 16, 19, 24, 30, 35, 40].

Neomorphic IDH1/2 mutations in AML confer gain-offunction activity that converts α-ketoglutarate (α-KG) to R-2-hydroxyglutarate (2-HG) [45–47]. Elevated 2-HG competitively inhibits α-KG-dependent dioxygenases, disrupting epigenetic control—particularly TET2-mediated DNA demethylation and related histone demethylases—thereby enforcing a hypermethylated state, a myeloid differentiation block, and leukemogenesis [45–47]. IDH inhibitors suppress the mutant enzymes, reduce 2-HG levels, restore dioxygenase activity, and promote blast differentiation/maturation, a mechanism that underlies the risk of differentiation syndrome [48].

For newly diagnosed patients with IDH1-mutated AML ineligible for intensive chemotherapy, azacitidine plus ivosidenib is an established frontline lower-intensity treatment option [2, 3]. In the phase III AGILE trial (n = 146), azacitidine plus ivosidenib improved longterm survival compared to azacitidine alone, with a median OS of 29.3 vs 7.9 months. The combination also improved CR/CRh rates, transfusion independence rates (53.8% vs 17.1%), and molecular MRD negativity (30.3% of evaluable patients) after a median follow-up of 28.6 months [3, 49].

Ivosidenib monotherapy has demonstrated clinically meaningful activity in IDH1-mutated AML across both newly diagnosed and R/R settings, particularly in patients who are ineligible for intensive therapy, thereby providing a treatment option when combination regimens are not feasible. In a newly diagnosed population, single-agent ivosidenib achieved CR rates of ~30%, with a median OS of ~13 months [50]. In the R/R setting, ivosidenib produced overall response rates (ORR) of ~40%, including CR/CRh rates of ~30% and a median OS of ~8–9 months [50]. Based on these data, ivosidenib is FDA-approved for use as monotherapy or in combination with azacitidine, including as frontline therapy and monotherapy for R/R AML [50].

HMA–venetoclax combinations offer another treatment option alongside the HMA–ivosidenib combination for patients with IDH1-mutated AML who cannot undergo intensive therapy. Without direct head-to-head trials, choosing between ivosidenib- and venetoclaxbased treatments depends on treatment goals, patient and provider preferences, and expected toxicity, rather than direct clinical evidence. The combination of azacitidine and ivosidenib has shown a durable survival benefit in the AGILE trial, with a median OS of ~29 months, whereas OS with venetoclax-based regimens in IDH1-mutated cases was 15.2 months, acknowledging the limitations of crossstudy comparisons [3, 51]. It has been suggested that venetoclax regimens may remain effective as salvage therapy after relapse on IDH inhibitors, supporting a strategy of sequencing IDH inhibitor s first; however, this approach remains under investigation [51]. An ongoing phase III trial (NCT07075016) aims to directly compare the sequencing of these regimens to better determine the optimal first-line treatment for this molecular subgroup [52].

Olutasidenib, an oral IDH1 inhibitor, provides an additional treatment option as monotherapy in R/R IDH1-mutated AML. A phase I/II trial reported a CR/CRh rate of 51% (with 63% of responders maintaining CR/CRh at 12 months), a median duration of response of 25.9 months, and a median OS of 11.6 months. Notably, patients had no prior IDH1 inhibitor exposure, and only a minority had received venetoclax-based therapy [53]. These data confirm that mutant IDH1 remains therapeutically actionable even after prior therapies; however, responses to IDH inhibitor monotherapy may be limited by incomplete eradication of leukemic clones and adaptive resistance mechanisms. Accumulating clinical experience suggests improved depth and durability of remission with IDH inhibitor–based combinations (with HMA, venetoclax, or intensive chemotherapy) across disease settings [50, 54–56]. One phase I/II study evaluating ivosidenib plus venetoclax, with or without azacitidine, reported CR/complete remission with incomplete hematologic recovery (Cri) rates of ~70–80%, MRD negativity in 63% of MRD-evaluable patients (10/16), and encouraging durability of response in both frontline and R/R cohorts, supporting the biological synergy between IDH inhibition and BCL-2 blockade [53, 57]. These observations support the ongoing development of doublet and triplet strategies with increasing emphasis on time-limited and MRD-guided treatment paradigms [50, 53–56].

Enasidenib is an oral, selective IDH2 inhibitor with established single-agent and combination activity in IDH2-mutated AML. In the pivotal phase I/II AG221-C-001 trial in R/R IDH2-mutated AML, enasidenib 100 mg daily demonstrated an ORR of ~40%, with durable remission and hematologic improvement. Of all patients with baseline RBC transfusion dependence, 43.1% (66/153) attained RBC transfusion independence for 56 days or longer [58]. In the randomized phase III IDHENTIFY trial that enrolled patients 60 years or older with R/R IDH2-mutated AML, enasidenib did not significantly improve OS compared with conventional care (median 6.5 vs 6.2 months; HR 0.86, 95% CI 0.67–1.10; P = 0.23). However, significant improvement in key secondary endpoints, including EFS (median 4.9 vs 2.6 months; HR 0.68, 95% CI 0.52–0.90; P = 0.008), time to treatment failure (median 4.9 vs 1.9 months; HR 0.53, 95% CI 0.41–0.69; P < 0.001), ORR (40.5% vs 9.9%), and RBC transfusion independence (31.7% vs 9.3%) were reported [54]. The lack of an OS benefit may have been confounded by early discontinuation or failure to initiate the assigned therapy in the control arm and subsequent AML therapies, including post-protocol enasidenib exposure, among patients receiving conventional care [54].

In newly diagnosed patients ineligible for intensive therapy, a randomized phase Ib/II trial demonstrated that azacitidine plus enasidenib improved responses compared with azacitidine alone (ORR, 74% vs 36%; P < 0.001), with evidence of deeper molecular responses [59]. Additional cohorts and real-world series have suggested the tolerability and activity of dual (azacitidine plus enasidenib) and triplet (azacitidine plus enasidenib plus venetoclax) combinations [60]. In fit, newly diagnosed patients, the combination of enasidenib with intensive chemotherapy demonstrated CR/CRi/CR with incomplete platelet recovery (CRp) rates of ~63–74% and IDH2 clearance by digital polymerase chain reaction (dPCR) in ~25% of responders, along with MRD negativity by flow cytometry in 63% (10/16) of evaluable responders [61]. These data support enasidenib as a rational standard option for R/R IDH2-mutated AML, particularly when hematologic recovery and transfusion endpoints are prioritized, and as an anchor for evolving doublet and triplet strategies [49, 58–61]. IDH inhibitors require monitoring for differentiation syndromes; ivosidenib may also prolong QTc, whereas enasidenib can cause indirect hyperbilirubinemia [48, 50, 53, 54, 58].

Venetoclax is a highly selective BH3 mimetic that binds the hydrophobic groove of BCL-2, competitively displacing pro-apoptotic BH3-only proteins (e.g. BIM), thereby freeing BAX and BAK to oligomerize and induce mitochondrial outer membrane permeabilization (MOMP), cytochrome-c release, and caspase-dependent apoptosis [62–64]. In AML, BCL-2 dependence is enriched in the leukemic stem and progenitor compartments, providing a biological basis for venetoclax sensitivity. Resistance is commonly associated with adaptive shifts toward MCL-1 or BCL-xL dependence and metabolic reprogramming, supporting rational combination strategies with HMA or other targeted therapies [62–64].

Currently, HMA–venetoclax is the standard of care for most patients with AML who are ineligible for intensive therapy. In the phase III VIALE-A study (n = 431), compared with azacitidine alone, the combination of azacitidine and venetoclax improved median OS from 9.6 to 14.7 months (HR 0.66, 95% CI 0.52–0.85) and CR/CRi rate from 28 to 66% in patients with newly diagnosed AML who were ineligible for intensive chemotherapy. This combination also achieved higher MRD-negative rates (CRc with flow MRD < 10⁻3: 23.4% vs 7.6%), with MRD < 10⁻3 by multiparameter flow cytometry achieved in 41% (67/164) of CRc responders, and more durable transfusion independence [2, 65]. The combination of decitabine and venetoclax also showed similarly high CR/CRi rates and OS in a phase II study and real-world cohorts. Although no phase III clinical trial has been conducted with the decitabine plus venetoclax combination, both azacitidine- and decitabine-based regimens are considered equally effective in routine clinical practice, and the choice between the two HMAs remains patient-, center-, and physician-dependent [66]. The phase III VIALE-C trial of low-dose cytarabine (LDAC) plus venetoclax (n = 211) reported increased CR/CRi rates (48% vs 13%) and numerically improved median OS (~7.2 vs 4.1 months) compared with LDAC alone [67]. The primary OS endpoint was not met at the planned analysis (median OS 7.2 vs 4.1 months; HR 0.75, 95% CI 0.52–1.07; P = 0.11) [67], but with an additional 6 months of follow-up the OS benefit became statistically significant (median OS 8.4 vs 4.1 months; HR 0.70, 95% CI 0.50–0.98; P = 0.04) and was most evident in patients without prior HMA exposure [68]. In practice, LDAC plus venetoclax is generally reserved for patients who cannot receive an HMA regimen.

Venetoclax was also evaluated in a phase II study of patients aged 60 or older, or deemed ineligible for intensive therapy with newly diagnosed AML. In this study, the treatment alternated between venetoclax combined with cladribine and LDAC and venetoclax with azacitidine. Ninety-three percent of patients achieved a composite remission (CR/CRi), with 84% MRD negativity among assessed responders [69]. At a median follow-up of 22.1 months, the median OS and DFS were not reached, with a 2% early mortality rate within 4 weeks [69].

Multiple clinical trials, mostly single-arm studies, have evaluated venetoclax in combination with intensive chemotherapy in fit patients and have achieved high remission rates. Fludarabine, cytarabine, granulocyte colony-stimulating factor, and idarubicin (FLAG-IDA) plus venetoclax demonstrated an MRD-negative CRc of ~96% in newly diagnosed patients and ~69% in R/R cohorts. A substantial proportion of these patients eventually proceeded to allo-HCT (~69% in newly diagnosed and ~46% in R/R disease) [70, 71]. Cladribine, idarubicin, and cytarabine (CLIA) plus venetoclax is another intensive regimen with a CRc rate of ~96% and MRD negativity of ~90% in newly diagnosed AML [72]. In secondary AML, CPX-351 plus venetoclax has shown encouraging activity in the R/R setting, particularly in first salvage and in AML without TP53 mutations, with a CRc rate of ~40–50%, supporting this combination as a potential option in secondary or therapy-related AML [73, 74].

Azacitidine plus venetoclax was also compared with intensive chemotherapy in fit patients with newly diagnosed AML in the phase II PARADIGM trial. The trial demonstrated better EFS with azacitidine and venetoclax combination compared to intensive chemotherapy (median EFS of 14.6 months vs 6.15 months, HR 0.61, 95% CI 0.42–0.61) with similar OS (median OS of 21.5 months vs 18.6 months, P = 0.187). Early mortality favored azacitidine and venetoclax over intensive regimens (60-day mortality 0% vs 4.7%), and patient-reported outcomes also showed better quality of life and symptom control, supporting azacitidine and venetoclax as a tolerable frontline alternative treatment for fit patients with adverse-risk AML. Importantly, a higher proportion of patients in the azacitidine and venetoclax arm proceeded to allo-HCT, supporting its role as an effective bridge to transplantation in eligible patients [75].

Venetoclax is increasingly being combined with targeted and immunotherapies, including menin inhibitors, for which preclinical and early clinical synergy has been observed. These combinations are discussed in detail in the sections on menin inhibitors and immunotherapy below [76–78].

Venetoclax-based maintenance strategies are being explored after intensive induction/consolidation [79] and in the post-allo-HCT setting (e.g., azacitidine–venetoclax), with early feasibility signals [80]. However, definitive randomized efficacy data are lacking to date.

Navitoclax, a BCL-2/BCL-xL inhibitor, is also being studied to address BCL-xL-mediated resistance in AML. In an early-phase study of navitoclax + venetoclax + decitabine in venetoclax-exposed R/R AML, the CRi/CRh rate was ~20%, supporting further study despite expected on-target thrombocytopenia [81, 82].

The major clinical limitation of venetoclax-based therapy is prolonged myelosuppression with infection risk, often requiring cycle delays, shortened venetoclax duration, and marrow-guided dose adjustment [2, 65–69, 71].

Menin functions as a critical scaffold for KMT2A-associated transcriptional complexes that maintain aberrant HOX/MEIS1 programs and leukemic self-renewal in KMT2A-rearranged and NPM1-mutated AML cells. Menin inhibitors disrupt menin–KMT2A binding, downregulate HOX/MEIS1 expression, and promote leukemic differentiation, providing a rationale for targeted therapy in these molecular subsets [83–87].

Revumenib is the most extensively studied oral menin inhibitor and is FDA-approved for R/R acute leukemia with KMT2A rearrangements and for R/R NPM1-mutated AML [88, 89]. In the phase I/II AUGMENT-101 study, revumenib demonstrated meaningful singleagent activity in heavily pretreated patients, including KMT2A-rearranged AML with an ORR of 60% and CR/CRh of 20–25%, and NPM1-mutated AML with an ORR of 45–50% and CR/CRh of 20–25%. Among patients achieving CR/CRh, MRD negativity by nextgeneration sequencing (NGS) was observed in 50–60%, and responses were associated with prolonged remission duration and OS, supporting the clinical relevance of menin inhibition in these molecular subsets [86, 90–92]. QTc prolongation is a key safety consideration for revumenib reported in ~30–40% of patients (Grade ≥ 1–20%). Patients also need close monitoring for differentiation syndrome, an on-target class toxicity requiring prompt recognition, and steroid-based management (reported in ~15% of patients treated with revumenib) [93].

Ziftomenib, another oral menin inhibitor, is FDAapproved for adults with R/R NPM1-mutated AML based on KOMET-001, in which the pivotal 600 mg daily dose achieved CR/CRh rates of 20–25% in a heavily pretreated population. Among the evaluable CR/CRh responders who underwent central MRD testing, ~ 60% were MRD negative. Responses occurred early, with a median time to response of 1.9 months. Durability was clinically meaningful in this setting, with a median duration of CR/CRh of 3.7 months and a median duration of response of 4.6 months among responders [57, 94–96]. Ziftomenib and revumenib share on-target differentiation syndrome risk with similar overall incidence reported in the prescribing information (26% with ziftomenib vs 29% with revumenib; grade ≥ 3 ≈13% for both), whereas QTc prolongation is reported less frequently with ziftomenib (12% vs 29% with revumenib), making QTc monitoring a particularly important safety consideration for revumenib [93].

Several additional menin–KMT2A inhibitors are being evaluated for the treatment of KMT2A-rearranged and NPM1-mutated acute leukemia, including bleximenib (JNJ-75276617), enzomenib (DSP-5336), and BN104. In the phase I first-in-human study (n = 56) of R/R KMT2Ar or NPM1-mutated acute leukemia, bleximenib showed acceptable tolerability (grade ≥ 3 AEs 29%, largely cytopenias; differentiation syndrome reported) and early efficacy, with an ORR of 50% at 90 mg twice daily, a mean time to response of 1.77 months, and ≥ 50% marrow blast reduction in 39% of evaluable patients [97]. Ongoing trials are testing bleximenib with venetoclax and with intensive “7 + 3” (NCT05453903) [98], with preliminary activity reported [99, 100]. In an ongoing phase I/II study, bleximenib [101] in early combination cohorts with venetoclax + azacitidine reached the Recommended Phase II Dose (RP2D), achieving an ORR of 82% with a CRc of 59% in R/R and an ORR of 90% with a CRc of 75% in newly diagnosed chemo-ineligible patients, with low rates of differentiation syndrome observed at the RP2D [102]. Enzomenib (DSP-5336), an oral menin–KMT2A inhibitor, is being evaluated in the treatment of R/R acute leukemia (NCT04988555) [103]. In a phase I monotherapy program, clinically significant activity was observed in molecularly selected cohorts, including those with menin inhibitor–naive KMT2Ar disease (ORR 73.3%; CR/CRh 40% at RP2D) [104]. In patients with KMT2Ar acute leukemia treated with enzomenib alongside concomitant azoles, ORR increased with dose, reaching 75% at 400 mg twice daily, with a median OS of 11.4 months among those receiving ≥ 200 mg twice daily [105]. In NPM1-mutated AML, the ORR was 58.8% (CR/CRh 47%), with a median OS of 8.5 months across the 200–400 mg twice daily dose range [105]. Early combination data with venetoclax/azacitidine further supported robust activity (ORR ~83–85%; CRc ~56–62%), particularly in venetoclax- and menin-inhibitor-naive patients [106]. BN104 is being evaluated in a first-in-human phase I/II trial for R/R acute leukemia (NCT06052813) and is also being explored as post-HSCT maintenance (NCT07101497), extending menin inhibition beyond salvage settings; in an interim phase I analysis, 11 patients completed post-baseline assessment and, among 9 efficacy-evaluable patients with target alterations, ORR was 88.9% (8/9) with CR/CRh 33.3% (3/9), supporting continued clinical investigation [107–109]. Key targeted-therapy trials discussed in this review are summarized in Table 1.

GO is an established anti-CD33 antibody–drug conjugate linked to the cytotoxic payload calicheamicin. After CD33-mediated internalization, calicheamicin induces DNA double-strand breaks and apoptosis. Because GO has been extensively studied and incorporated into selected frontline regimens, it is discussed briefly here in the clinical context of newer antibody–drug conjugates and antigen-directed immunotherapies for AML.

The clearest benefit of GO has been observed in selected favorable-risk AML, particularly core-binding factor AML, when added to intensive chemotherapy. In the phase III ALFA-0701 trial, fractionated GO added to standard induction improved EFS and RFS with a trend toward improved OS [110, 111]. An individual-patientlevel meta-analysis of five randomized trials confirmed that GO reduced relapse risk and improved survival, with the greatest absolute benefit in favorable-risk AML and no clear benefit in adverse-risk cytogenetic AML [112]. Additional studies support GO-containing regimens in core-binding factor AML, and contemporary treatment approaches support the incorporation of GO into frontline intensive therapy for selected fit patients with favorable-risk disease [113, 114]. GO has also been evaluated in NPM1-mutated AML, where relapse reduction was observed without a clear OS benefit [115].

The principal toxicities of GO include prolonged thrombocytopenia and hepatic toxicity, including sinusoidal obstruction syndrome (SOS) and veno-occlusive disease, particularly when GO is followed closely by myeloablative allo-HCT [111, 116]. Accordingly, GO is generally avoided in patients with significant baseline hepatic dysfunction, and an adequate washout interval before transplantation is recommended. Fractionated dosing has been associated with a lower risk of SOS and has supported safer incorporation of GO into selected intensive regimens [117, 118].

CD47 expressed on leukemic blasts engages SIRPα on macrophages to deliver a “don’t-eat-me” signal, thereby promoting macrophage-mediated phagocytosis and augmenting cross-priming of adaptive immunity [119, 120]. Magrolimab, an anti-CD47 monoclonal antibody, was evaluated with azacitidine in the phase III ENHANCE-2 trial in previously untreated TP53-mutated AML versus physician’s choice therapy and did not improve survival (IC-ineligible cohort: median OS 4.4 vs 6.6 months; HR 1.13, 95% CI 0.78–1.64; IC-eligible cohort: median OS 7.3 vs 11.1 months; HR 1.43, 95% CI 0.64–3.24) [121]. Similarly, ENHANCE-3 tested magrolimab versus placebo added to azacitidine/venetoclax in untreated AML ineligible for intensive therapy and was stopped early for futility and excess mortality; final analysis showed no survival benefit with inferior OS (median OS 10.7 vs 14.1 months; HR 1.18, 95% CI 0.85–1.64) and comparable early remission (CR within six cycles 41.3% vs 46.0%) [122]. The sponsor’s February 7, 2024 disclosure of increased mortality risk—largely driven by infection and respiratory failure—followed by an FDA clinical hold, led to discontinuation of the study [123]. Newer, CD47-axis agents are being re-engineered using Fc-silent antibodies and SIRPα fusion formats to improve therapeutic index [119, 120].

AML is associated with an immunosuppressive marrow microenvironment, including PD-1/PD-L1 pathway activation on leukemic blasts, antigen-presenting cells, and exhausted effector T cells [124, 125]. HMAs may further modulate antigen presentation and checkpoint expression, providing a biological rationale for combining HMAs with PD-1/PD-L1 blockade [126]. However, despite this rationale, the clinical activity of checkpoint inhibition in AML has been inconsistent and has not translated into routine practice.

Randomized and post-remission studies have not demonstrated clear practice-changing benefit. In older or unfit patients, azacitidine with or without durvalumab showed no significant improvement in response or survival outcomes [127]. Post-remission checkpoint inhibitor strategies have also produced limited benefit and may increase treatment-related toxicity [128, 129]. Checkpoint blockade after allo-HCT remains particularly challenging because of immune-related toxicity and the risk of precipitating or worsening graft-versus-host disease. Ipilimumab after allo-HCT demonstrated some activity at higher doses but with increased immune-related adverse events, illustrating a narrow therapeutic window in this setting [130]. In R/R AML, azacitidine plus nivolumab showed modest activity in a phase II study, but without definitive evidence that PD-1 blockade adds benefit beyond the HMA backbone [131].

Overall, PD-1/PD-L1 inhibitors remain investigational in AML and are most appropriately evaluated in biomarker-driven clinical trials or carefully selected postremission strategies [127–131].

Tagraxofusp (SL-401) is a CD123-directed cytotoxin consisting of recombinant interleukin-3 fused to a truncated diphtheria toxin [132]. The drug binds to CD123 (IL-3Rα), is internalized, and then blocks protein synthesis, leading to cell death in CD123-expressing cells [132]. Tagraxofusp is established in blastic plasmacytoid dendritic cell neoplasms (BPDCN), but its role in AML remains unclear. Interest in AML comes from the frequent CD123 expression on blasts and stem/progenitor-enriched compartments. In a phase Ib study of tagraxofusp combined with azacitidine, with or without venetoclax, patients with adverse-risk AML showed a CR/CRi/morphologic leukemia-free state (MLFS) rate of 69%, with MRD negativity by multiparameter flow in 12/17 (71%) assessed responders (median time to MRD negativity: cycle 2 [range, 1–4]) [133, 134].

Pivekimab sunirine is a CD123-targeting humanized antibody drug conjugate (ADC) linking the G4723A antibody to a DNA monoalkylating indolinobenzodiazepine pseudodimer payload (DGN549) via a protease-cleavable linker, enabling potent DNA damage in CD123-high AML blasts and progenitors with relative sparing of normal hematopoietic cells [135]. In a first-in-human phase I/II trial in R/R AML, pivekimab sunirine achieved an ORR of ~21% and a CRc rate of ~17% in a heavily pretreated cohort. Dose-limiting toxicities primarily consisted of reversible SOS at higher dose levels and transient neutropenia, with grade ≥ 3 treatment-related events largely limited to febrile neutropenia, infusion reactions, and anemia [136].

Early phase Ib/II studies of pivekimab sunirine with azacitidine, with or without venetoclax, showed encouraging activity in CD123-positive AML, with CR/CRi ~30–35% in R/R disease and ~65–70% in newly diagnosed adverse-risk AML, and deep remissions among responders, including MRD negativity (< 0.1% by multiparameter flow) in 76% (22/29) of MRD-evaluable patients with composite CR [136, 137]. Updated analyses confirmed durable depth of response without a clear increase in SOS or capillary-leak syndrome and with myelosuppression largely comparable to azacitidine/venetoclax alone, supporting the continued evaluation of pivekimab-based combinations to deepen remission and bridge high-risk patients to allo-HCT [135, 137].

Vibecotamab, a bispecific antibody targeting both CD123 and CD3, redirects cytotoxic T-cells to CD123-expressing blasts. In a phase I study, AML responses were observed in ~9% of patients (MLFS or better), particularly in those with lower baseline blast counts and favorable T-cell fitness biomarkers. [138]. In low-blast or MRD-positive cohorts, vibecotamab demonstrated response rates of ~25–27% with limited additional myelosuppression, supporting its development as a strategy for deepening remission [138]. Cytokine release syndrome (CRS) was observed in ~59% of patients treated with vibecotamab, predominantly at grade ≤ 2.

AFM28 is a natural killer (NK)-cell-redirecting bispecific CD123/CD16A innate cell engager. It has shown strong preclinical activity against CD123-positive leukemic stem cells and progenitor cells [139]. In the first-in-human phase I dose-escalation trial in R/R AML, CR/CRi rates of up to ~40% have been reported with AFM28 at higher dose levels [140]. Treatment-related toxicity has been dominated by infusion-related reactions (mostly grades 1–2), with no clear neurotoxicity signal in the reported datasets [140]. These early data support the evaluation of AFM28 in venetoclax-based combinations to deepen remissions and improve molecular response rates [139, 140].

Key immunotherapy trials discussed in this review are summarized in Table 2.

CAR-engineered T and NK cells target AML surface antigens (e.g., CD33, CD123, CLL-1, and FLT3) in an HLA-independent manner, mediating direct cytotoxicity against blasts and leukemic progenitors [141, 142]. CAR-T approaches in AML remain early-phase, limited by antigen overlap with normal myeloid progenitors (on-target myelotoxicity), disease heterogeneity, and impaired persistence/trafficking in an immunosuppressive marrow niche. Early signals of activity have been observed with CLL-1 (CLEC12A) targeting. In a first-in-human phase I study in R/R AML (n = 10), CR/CRi was achieved in 70% of patients; CRS occurred in all patients (including 6/10 high-grade) without CAR-T cell-related encephalopathy syndrome (CRES). However, all patients had severe pancytopenia, and two infection-related deaths occurred secondary to chronic agranulocytosis [141–143]. In contrast, a phase I autologous CD33 CAR-T experience highlighted feasibility and efficacy constraints (10 enrolled; product meeting prespecified release criteria in 4/10; 3 patients had CRS/immune effector cell-associated neurotoxicity syndrome (ICANS) signals, but there were no objective responses, and all died from progression), reinforcing the need for optimized targets (e.g., CLL-1, FLT3), dual/logicgated designs, and post-cytoreduction or post-transplant sequencing [141, 142, 144]. CAR-NK platforms offer potential advantages, including lower rates of severe CRS and ICANS and the feasibility of allogeneic, off-the-shelf manufacturing. In a phase I study of CD33 CAR NK cells, MRD-negative CR was achieved in 6 of 10 patients by day 28, with only low-grade CRS reported [145–147]. Newer approaches use dual-target or logic-gated CARs, optimized lymphodepletion, and built-in “off-switches” to improve safety and efficacy. Many trials now focus on MRD-based endpoints and evaluate integration with allo-HCT to restore hematopoiesis [141, 142, 147, 148]. In the phase I donor CIML-NK study for posthaplo-HCT relapse, MRD negativity by next-generation sequencing was achieved in 2 of 6 patients (33%) at day +28, including patients attaining CR/marrow CR, suggesting deep molecular responses despite the small cohort size [149].

The BiTE and DART formats simultaneously bind CD3 on T cells and a myeloid antigen (e.g., CD123 or CD33) on AML blasts, inducing immune synapse formation and HLA-independent T-cell cytotoxicity. In a phase I/II study of R/R AML (median 4 prior lines; N = 88), flotetuzumab (CD123 × CD3 DART) achieved a CR/CRh of ~18%, with a median OS of ~10.2 months among responders. CRS was a common adverse event noted in this study (any-grade CRS ~96%), supporting step-up dosing and standardized mitigation strategies [150]. In another phase I study of R/R AML (N = 77; 60 evaluable for efficacy), AMG 330 (CD33 × CD3 BiTE) demonstrated eight responses (CR 3, CRi 4, and MLFS 1) with CRS observed in ~78% of patients, underscoring a therapeutic class in which clinical activity must be balanced against cytokine-driven toxicity through proactive management [151].

Key cellular therapy studies discussed in this review are summarized in Table 3.

AML treatment selection is increasingly guided by patient fitness, ELN risk, actionable genomic alterations, MRD status, and transplantation eligibility. In fit patients, intensive chemotherapy remains the backbone treatment, with FLT3 inhibitors for FLT3-mutated AML and gemtuzumab ozogamicin for selected favorable risk/corebinding factor AML [1, 13, 110–112]. In older or intensive therapy-ineligible patients, lower-intensity options include azacitidine–venetoclax for many patients and azacitidine–ivosidenib for IDH1-mutated AML [2, 3, 49, 65]. In R/R AML, targeted options include gilteritinib for FLT3-mutated disease, IDH inhibitors for IDH1/2-mutated disease, and menin inhibitors for KMT2Arearranged or NPM1-mutated AML [5, 50, 53, 54, 58, 86, 87, 90]. MRD status and transplant fitness further guide consolidation, allo-HCT, maintenance, and clinical-trial strategies [4, 25, 26, 65, 79, 80]. A simplified geneticsinformed framework integrating patient fitness, genomic risk, actionable mutations, MRD status, and transplant eligibility is summarized in Fig. 1.