Precision oncology to overcome resistance in R/R AML in children and adults requires combinations of cytotoxic, targeted, and immunological treatments

Abstract

Although outcomes for newly diagnosed acute myeloid leukaemia (AML) have been incrementally improved over the last decades, management of relapsed and refractory (R/R) AML remains a medical challenge. A curative intent for R/R AML usually involves chemotherapy (with or without targeted therapy) with subsequent consolidation, including allogeneic haematopoietic stem cell transplantation. Despite this, long‐term survival rates of R/R AML only reach approximately 10% in adults and 40% in children. Given this great unmet clinical need, this review outlines the current and emerging paradigms for preventing and treating R/R AML. Somatic mutations, gene expression, and functional drug testing are important for the selection of small molecule inhibitors of oncogenic signalling pathways (e.g., FLT3), menin inhibitors that disrupt leukemogenic programmes, inhibitors of isocitrate dehydrogenases to restore oncometabolic homoeostasis, and proapoptotic Bcl‐2 homology 3 (BH3) mimetics, such as venetoclax. Targeting the recently identified resistance factor SAMHD1 promises to overcome resistance to cytarabine and fludarabine. Given the growing number of potential combinatorial drug regimens and the genetic heterogeneity of AML, real‐time ex vivo drug response profiling to guide individualized treatment decisions will become an important complement. We argue that better outcomes for R/R AML critically depend on being guided by precision oncology to define the best combination of chemotherapy, targeted therapy, and immunological therapy informed by phenotypic and genotypic patient‐ and disease‐specific parameters.

Introduction

Acute myeloid leukaemia (AML) is caused by the acquisition of distinct genetic aberrations in myeloid progenitors of the bone marrow, leading to maturation block and uncontrolled proliferation of leukemic blasts. The identification and characterization of distinct genetic and epigenetic aberrations have led to a vastly increased understanding of AML pathobiology in recent years. At the same time, a plethora of novel therapeutic agents have become available. So far, these developments have not been convincingly translated into improved outcomes in patients with relapsed or refractory (R/R) disease. This article aims to provide a comprehensive review of state‐of‐the‐art and biology‐informed emerging strategies for R/R AML management. We argue that future decisions in terms of R/R AML treatment must be approached in a multimodal manner, taking into consideration molecular alterations (e.g., oncogenic mutations), mechanisms of chemotherapy resistance, immunological properties and functional ex vivo drug sensitivity profiling.

Epidemiology and definitions

The incidence of AML is around 4 per 100,000 in adults, with 30% of patients developing AML secondary to a prior myeloid malignancy or to prior cancer treatment [1]. Patients receiving intensive chemotherapy achieve complete remission (CR) or CR with incomplete recovery (CRi) in 50%–90% of cases, with ∼70% being measurable residual disease (MRD) negative after two induction courses [2]. Approximately 65% of unfit adults receiving non‐intensive therapy, primarily azacitidine and venetoclax, achieve CR/CRi, and approximately 15% reach MRD negativity [3]. The probability of obtaining a response to treatment depends on the underlying genetics/disease risk category, treatment intensity and age. The 5‐year overall survival (OS) for intensively treated patients depends on patient characteristics and ranges from around 41% to 62% in patients categorized as low‐risk but only around 8%–25% in those with high‐risk [4].

Refractory AML is defined as the persistence of leukemic blasts in the bone marrow (>5%) after two cycles of intensive induction chemotherapy. Relapsed AML is defined by the recurrence of more than 5% bone marrow blasts or extramedullary growth after achieving an initial remission [5]. However, in many previous and ongoing trials, lower levels of residual leukemic cells as measured by flow cytometry of molecular methods are considered a relapse event. Even fit low‐risk patients treated with high‐intensity chemotherapy have a significant risk of relapse of ∼25% [6], and nearly all high‐risk patients would relapse without subsequent allogeneic haematopoietic cell transplantation (allo‐HSCT).

In children, AML constitutes ∼20% of leukaemias and is characterized by age‐specific features in clinical presentation [7], genetic profile [8, 9] and to some extent outcome [10, 11]. In contrast to adults, AML in children infrequently arises as a continuum of previous myelodysplastic syndrome [12] or after cytotoxic therapy [13]. Instead, germline genetic predisposition is more common in children [14].

Children with de novo AML are treated according to collaborative study group intensive chemotherapy protocols leading to CR in approximately 90% of children [15, 16], of whom 70%–80% may obtain MRD negativity [17]. Despite the immediate efficacy of primary AML therapy, 25%–40% of children experience relapse [17], which remains a significant therapeutic challenge and still portends an unsatisfactorily poor outcome with only 35%–50% long‐term OS [18].

Approximately 5%–10% of children with AML respond poorly to primary therapy with refractory disease (RD) [17, 19]. CR may be achieved through additional induction attempts, but prognosis is generally poor in children with RD, with OS rates below 40% [20, 21].

Current concepts to prevent relapse

Despite the high initial response rates to induction therapy, a significant proportion of patients will eventually relapse, particularly adult patients. To reduce the need for salvage treatment, large efforts both in paediatric and adult AML have been made to optimize the primary treatment to achieve deeper responses and to reduce the risk of relapse. Recently, intensified induction in adults with Fludarabine/Cytarabine/Granulocyte Colony‐Stimulating Factor (G‐CSF)/Idarubicin/Gemtuzumab‐ozogamicin (FLAG‐Ida‐GO) has been shown to result in superior outcome in NPM1‐mutated AML [6].

Importantly, genetic risk stratification has improved substantially, and predictive scores for both fit and unfit AML have recently been published for adults (ELN 2022—fit, ELN 2024—unfit) [5, 22]. The risk scores not only predict response to treatment but also identify patients who are at high risk of disease recurrence. The risk of relapse can be reduced by allo‐HSCT, which is generally considered for non‐favourable risk fit AML patients in first remission. An increasing number of AML patients receive allo‐HSCT thanks to expanding international donor registries, improved protocols for using haploidentical donors, co‐morbidity‐adapted conditioning regimens and improved supportive care.

In the paediatric setting, primary induction therapy is usually more intensive, and further intensification is associated with unacceptable treatment‐related mortality and morbidity [23, 24]. Collaborative study groups have reached a consensus on response features, but risk‐classification systems based on high‐risk genomic features in recent clinical trials still vary [25].

Another strategy to reduce the risk of relapse is to administer maintenance therapy after achieving remission, including in some cases maintenance after allo‐HSCT. Successful maintenance strategies that have demonstrated survival benefits in randomized trials in adult AML include the FLT3‐inhibitors midostaurin, quizartinib and gilteritinib [26, 27, 28], the oral hypomethylating agent (HMA) azacitidine [29] and the immune‐stimulatory combination of histamine dihydrochloride and interleukin‐2 [30]. In children with FLT3‐internal tandem duplication (ITD) AML, maintenance with sorafenib has been evaluated [31], and ongoing trials are exploring gilteritinib (NCT04293562) and quizartinib [32]. Several additional maintenance strategies are currently being studied in phase 3 trials in genetic subtypes of AML, including inhibitors of isocitrate dehydrogenase (IDH)1/2 and menin [33].

If a relapse is identified early, at the MRD level, patients are generally in good condition. In this situation, the disease is often easier to treat (absence of cytopenia, coagulopathies, leukostasis and tumour lysis syndrome), and eligible patients can proceed rapidly to allo‐HSCT with or without prior anti‐leukemic treatment. Pre‐emptive treatment of imminent relapse detected by longitudinal molecular MRD monitoring with highly sensitive assays is an area undergoing rapid development and may be an attractive strategy to improve outcome, at least in select AML genotypes [34, 35, 36].

Despite extensive efforts to optimize treatment, R/R AML remains a challenge in many patients. This review focuses on the key aspects of how to achieve a second remission to bridge the patient to a potentially curative allo‐HSCT or to maintain disease control with sustained quality of life.

Challenges in treating R/R AML

Resistance to the first line of intensive therapy may be due to inherent resistance mechanisms in leukemic cells, present either in all AML cells or in a subclone. Drug‐specific resistance mechanisms and strategies to overcome resistance are described below. The importance of disease heterogeneity and clonal evolution of more resistant subclones during the course of the disease has been characterized in depth [37, 38, 39]. As stem cell‐like leukemic cells (leukemic stem cells, LSCs) are less proliferative and possess drug efflux activity [40], they are also less sensitive towards cytotoxic treatments. Accordingly, stem cell disorders are less responsive to chemotherapy [41]. Smouldering MDS‐AML are often true stem cell disorders, where the malignant transformation has occurred near the level of the haematopoietic stem cell. There are several strategies of how to target cells with low proliferation, including apoptosis‐inducers such as the Bcl‐2 homology 3 domain (BH3) mimetic venetoclax or antibodies directed at surface molecules on the LSCs. Another strategy is to use growth factors that increase the proliferation of the LSCs. G‐CSF has been explored greatly, though without much success [42], potentially due to insufficient expression of the G‐CSF receptor on LSCs. Finally, given the low proliferation rate, extended maintenance therapy with HMAs, with or without venetoclax, may gradually wear down the leukemic clones and delay or even prevent relapse [3, 29], a notion most clearly illustrated in acute lymphoblastic leukaemia [43].

Extramedullary AML manifestation (such as skin, lymph nodes, gastrointestinal tract or central nervous system [CNS] [44, 45]) requires special consideration. It is crucial to achieve local disease control because graft‐versus‐leukaemia (GvL) effects seem to be less protective against extramedullary relapse after allo‐HSCT [46]. Depending on the sites and extent of extramedullary involvement, radiotherapy may be useful as an add‐on to systemic treatment to ensure local disease control, particularly in adults. The optimal dose of radiation is unknown, although data suggest that 12 fractions of 2 Gy (24 Gy in total) may be sufficient [47]. Several small case series have indicated that novel AML regimens may be effective also in extramedullary disease, including HMAs, venetoclax and FLT3 inhibitors such as gilteritinib [48, 49]. Some novel drugs are small molecules that penetrate the blood‐brain barrier, which may be important for AML involving the CNS [50]. Additionally, checkpoint inhibitors, such as the anti‐CTLA4 monoclonal antibody ipilimumab, have shown promising efficacy in a small case series [51].

Finally, certain subtypes of AML are exceedingly challenging to cure in the R/R setting, in particular adult patients with double hit TP53 mutations, where a palliative strategy with good supportive care and cytoreductive chemotherapy to control hyperleukocytosis need to be considered [52, 53].

Salvage chemotherapy approaches

Intensive chemotherapy strategies

A prerequisite for long‐term survival is allo‐HSCT, generally after a second CR has been achieved following salvage therapy [18, 20, 54, 55]. Although less intensive regimens based on HMAs with or without the Bcl‐2 inhibitor venetoclax as well as small‐molecule inhibitors of FLT3, IDH1 or IDH2 can achieve CR in a subset of patients [54, 55] (see below), re‐induction with chemotherapy is still mostly used in patients fit‐for‐transplantation (Table 1).

Table 1.

Summary of potential treatment strategies for relapsed and refractory (R/R) acute myeloid leukaemia (AML).

Treatment modality	Treatment specifics	References
Intensive treatment	FLAVIDA FLA(G)‐Ida ± GO FLAMSA CLAM MEC	[56, 57, 58, 59] [6, 55, 60] [61] [62, 63] [64, 65]
Non‐intensive treatment	HMAs + venetoclax CAV (consider in particular for monocytic AML) Immune checkpoint inhibitors Consider early phase clinical trials	[66, 67] [68, 69, 70] [71, 72, 73, 74] [55], this review
Targeted therapy augmentation	Consider adding inhibitors of FLT3 (gilteritinib, quizartinib), IDH1 (ivosidenib), IDH2 (enasidenib), menin (e.g., revumenib) depending on mutations and rearrangements	[31, 66, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84]
Overcoming SAMHD1‐mediated resistance	Potentially add HU for ara‐C or ara‐C + F‐ara‐A ‐containing regimens	[85, 86, 87, 88, 89]
Cardioprotection	Consider dexrazoxane when using anthracyclines/anthraquinones	[90, 91]
Ex vivo drug profiling	May be informative when selecting chemotherapy backbones and targeted inhibitors	[92, 93, 94, 95]
Cure	Allo‐HSCT	[55, 96]
Maintenance post allo‐HSCT	HMAs Venetoclax FLT3i IDH1i IDH2i Menin inhibitors Histamine dihydrochloride + interleukin‐2	[3, 29] [3, 29] [26, 27, 28, 31, 32, 97] [98] [33, 98] [33] [55] [30]

Abbreviations: Allo‐HSCT, allogeneic haematopoietic cell transplantation; CAV, cladribine, ara‐C, and venetoclax; IDH, isocitrate dehydrogenase; HMA, hypomethylating agent; HU, hydroxyurea.

The attempt of re‐induction with intensive chemotherapy must consider treatment resistance. It is therefore surprising that most established salvage chemotherapy regimens contain the same backbone as the primary treatment: intermediate or high doses of cytarabine (ara‐C), with the rationale to overcome ara‐C resistance by increasing the dose. Historically used as monotherapy, some studies reported CR rates of up to 50% [99]. With the exception of HMAs that have a different mode of action (see below), monotherapy with other cytotoxic nucleoside analogues than ara‐C has been less effective, as demonstrated for the purine analogues cladribine [100] and clofarabine [101, 102, 103]. Of note, addition of an anthracycline to cladribine or clofarabine did not improve outcomes either [104, 105].

Ara‐C is usually combined with a topoisomerase II inhibitor, where daunorubicin is frequently exchanged with another anthracycline (e.g., idarubicin), the anthraquinone mitoxantrone or the non‐anthracycline amsacrine. Shifting the topoisomerase II inhibitor might circumvent daunorubicin resistance [106, 107]. A long‐established two‐drug re‐induction course combines high‐dose ara‐C (considered to be superior to intermediate dose ara‐C [108], see below) with mitoxantrone (HAM) [109]. In a retrospective analysis, the German–Austrian AML Study Group found addition of gemtuzumab‐ozogamicin (together with all‐trans retinoic acid) to the HAM backbone to be more potent in inducing CR/CRi (50% vs. 25%) [110]. Just like with ara‐C, dose‐escalation of anthracyclines has been tested [111, 112]; any potential antileukemic benefit must be balanced against the increased risk of cardiotoxicity. In the primary treatment setting, a large randomized trial did not show a benefit of intensifying daunorubicin doses from 60 to 90 mg/m² [113] (in contrast to a landmark trial comparing 45 to 90 mg/m²) [114]. Another rationale to overcome resistance to first‐line therapy is adding drugs with a different mode of action, such as the canonical topoisomerase‐II inhibitor etoposide, which is most widely used in combinations with ara‐C and mitoxantrone (MEC) and has shown feasibility as a bridge‐to‐transplant approach with success rates of around 50% [64, 65].

More importantly, overcoming resistance to ara‐C remains the central element for re‐induction/salvage courses aiming to increase exposure to the active metabolite ara‐CTP, with levels of ara‐CTP correlating with enhanced anti‐leukemic effect [115, 116, 117]. Further, based on in vitro experiments showing moderate increases of ara‐CTP when pre‐treating leukemic cells with fludarabine (F‐ara‐A) [118], addition of F‐ara‐A to intermediate and high doses of ara‐C with or without (G‐CSF) as a priming strategy was introduced clinically (FLA(G)) [119, 120]. In the paediatric setting, the addition of liposomal daunorubicin to FLAG resulted in better OS of R/R AML; however, only in the subset of core‐binding factor AML [121]. Idarubicin has been incorporated into FLAG‐Ida protocols [60, 122], which improved remission rates. Perplexingly, addition of F‐ara‐A to ara‐C and idarubicin‐based re‐induction therapy only modestly improved CR rates and did not improve overall or relapse‐free survival in a randomized phase 3 trial for R/R AML [122]. This is, however, consistent with a lack of improved survival reported for upfront addition of F‐ara‐A to ara‐C and daunorubicin [123].

Two additional adenosine analogues have been combined with ara‐C‐based intensive reinduction therapies. Addition of cladribine has been shown to achieve an increase of ara‐CTP in cells from R/R AML patients [124]. The Polish Adult Leukemia Group (PALG) showed that addition of cladribine to ara‐C (together with G‐CSF; CLAG) in an anthracycline‐free regimen resulted in a CR rate of 50% in R/R AML, and responding patients had a 1‐year OS of 65% [125]. Contrary to those findings, a paediatric and young adult phase 2 trial at St. Jude's did not show efficacy of cladribine plus (low‐intermediate dose) ara‐C in R/R AML [126]. Another PALG study evaluated the CLAM regimen in R/R AML, that is, addition of cladribine to HAM. with a CR rate of ∼50% and a 1‐year OS in patients achieving CR of 73% [62]. Similar results were obtained in an American trial [63]. Taken together, also considering that addition of cladribine to ara‐C and daunorubicin has been successful in newly diagnosed AML patients [123], salvage regimens based on high‐dose ara‐C, an anthracycline and cladribine appear efficacious.

Clofarabine in combination with ara‐C and mitoxantrone (CLoAM) resulted in overall response rates of 90% in 52 R/R AML patients in a Hongkong phase 2 trial [127]. Addition of clofarabine to ara‐C and idarubicin in MD Anderson R/R AML trials yielded objective response rates of up to 80% [128, 129]. On the other hand, a phase 1B trial in children with R/R AML combining clofarabine, ara‐C and liposomal daunorubicin produced CR rates of 68% with 2‐year OS of 50% [130].

Inclusion of anthracyclines/mitoxantrone increases the risk of cardiotoxicity. Therefore, salvage treatment using amsacrine instead of idarubicin together with F‐ara‐A and ara‐C (FLAMSA) was evaluated and found to be an effective alternative [61]. Despite these modifications, prognosis for R/R AML remains bleak for most patients, warranting further therapy improvements. A standard of care for intensive salvage therapy remains to be established, and the backbone may need to be tailored to patients with different disease characteristics.

Novel concepts of improving the efficacy and tolerability of intensive chemotherapy

Resistance to ara‐C is a hurdle for effective therapy, and salvage courses that can overcome this resistance are aimed at increasing leukemic exposure to the active metabolite ara‐CTP, as its accumulation is critical for its anti‐leukemic effect [117]. The enzyme SAMHD1 is a deoxynucleoside triphosphate triphosphohydrolase that can metabolize ara‐CTP into inactive precursors, leading to ara‐C resistance in leukemic cells [85, 86]. In line with this, patients with AML who have high levels of SAMHD1 fare worse than their SAMHD1‐low counterparts when treated with ara‐C‐based regimens [87]. Hydroxyurea (HU) is a ribonucleotide reductase inhibitor that can indirectly inhibit SAMHD1 by inducing dNTP pool imbalances, thus reducing the enzymatic activity of SAMHD1 towards ara‐CTP [88]. Addition of HU to ara‐C overcomes SAMHD1‐mediated resistance and improves treatment efficacy in vitro, ex vivo and in vivo [88] and has shown the ability to safely increase leukemic ara‐CTP exposure in a phase 1 clinical trial in adults with newly diagnosed AML [89] (Fig. 1).

Fig. 1.

Resistance mechanisms and targets in relapsed and refractory (R/R) acute myeloid leukaemia (AML), part 1. This figure illustrates three resistance mechanisms and molecular targets for the treatment of R/R AML. (a) SAMHD1 is a triphosphohydrolase that detoxifies leukemic cells from cytarabine and fludarabine's active metabolites ara‐CTP and F‐ara‐ATP; SAMHD1 can be indirectly inhibited by hydroxyurea (HU). (b) Mutations in isocitrate dehydrogenase (IDH)1/2 lead to an increase of the oncometabolite 2‐hydroxyglutarate that in turn inhibits histone and DNA demethylases. This process can be inhibited by IDH inhibitors. Hypermethylation can also be targeted by hypomethylating agents (HMAs) and cladribine (2‐CdA). (c) Upregulation of anti‐apoptotic protein Bcl‐2 confers resistance to conventional chemotherapy. The Bcl‐2 inhibitors can alleviate this process, but Bcl‐x_L and MCL‐1 can mediate resistance to Bcl‐2 inhibition.

Similarly, SAMHD1 has been shown to confer resistance to F‐ara‐A in vitro [85]. We have gathered unpublished evidence that resistance to F‐ara‐A‐containing regimens, in particular FLA, against AML is mediated by SAMHD1 and can be alleviated by addition of HU (Lilienthal et al., personal communcation).

While increasing the anthracycline dose has been shown to improve survival in primary treatment [114], dose‐escalation of anthracyclines is limited by the dose‐dependent risk of cardiotoxicity [131]. The cardioprotectant dexrazoxane effectively reduces risk of cardiotoxicity by protecting cardiac topoisomerase 2β from anthracyclines [90]. In osteosarcoma, addition of dexrazoxane allowed a safe increase in doxorubicin cumulative doses from 450 to 600 mg/m² without compromising treatment efficacy or safety [91]. Hence, adding dexrazoxane appears a feasible strategy to allow for anthracycline‐intensified treatment of R/R AML. A different strategy to increase daunorubicin efficacy while reducing cardiotoxicity is to use liposomal formulations that have been shown to decrease cardiotoxicity in vitro [132]. CPX‐351 is a liposomal ara‐C/daunorubicin formulation that has shown similar efficacy as compared to FLA‐Ida in newly diagnosed AML [133]. In a non‐randomized paediatric R/R AML study, CPX‐351 nevertheless was associated with a significant deterioration of left‐ventricular ejection fraction and cardiac dysfunction [134]. A randomized phase 2 trial in adults with first relapse of AML comparing CPX‐351 with investigator's choice of salvage resulted in a CR rate of ∼40% [135]. In children with de novo AML, CPX‐351 + GO resulted in inferior efficacy as compared to DA + GO [136] (Fig. 2).

Fig. 2.

Resistance mechanisms and targets in relapsed and refractory (R/R) acute myeloid leukaemia (AML), part 2. This figure illustrates three resistance mechanisms and molecular targets for the treatment of R/R AML. (a) Activating mutations in the receptor tyrosine kinase FLT3, in particular internal tandem duplications, can be targeted FLT3 inhibitors. (b) Surface expression of CD33 makes AML vulnerable to the antibody‐drug conjugate Gemtuzumab‐ozogamicin. (c) Interaction of KTM2A (and NUP98) chimera with menin leads to transactivation of HOXA9 and myeloid ecotropic virus insertion site 1 (MEIS1). This interaction can be inhibited by menin inhibitors. They are also effective in NPM1‐mutated and upstream binding transcription factor tandem duplicated (UBTF‐TD) AML.

Targeted therapies

Venetoclax and other BH3 mimetics

BH3‐only proteins constitute an important part of a pro‐apoptotic signalling pathway, resulting in activation of Bax and Bak, which triggers apoptosis via cytochrome c release from mitochondria. BH3‐only proteins are partially sequestered by anti‐apoptotic Bcl‐2 family proteins Bcl‐xL, MCL1 and Bcl‐2‐related protein A1, which often are upregulated in cancer cells, thus inhibiting apoptosis [137].

BH3‐mimetics are a class of drugs that bind the anti‐apoptotic Bcl‐2 family proteins, which unleashes the BH3‐only proteins to trigger apoptosis via activation of Bax and Bak [138]. BH3‐mimetics may also interfere with the oxidative metabolism in LSCs, which starves them to a greater degree than normal haematopoietic stem cells [139]. Importantly, preclinical and clinical data have demonstrated synergy between BH3‐mimetics and numerous other drugs with different modes of action, making them some of the most promising novel targeted drugs in AML.

Venetoclax specifically inhibits Bcl‐2 (Fig. 1) and is at present the only BH3‐mimetic approved for AML by the European Medicines Agency (EMA). The EMA approval is based on phase 2 and phase 3 data demonstrating markedly improved outcomes when combining venetoclax with the HMAs azacitidine or decitabine [3, 140]. Promising results were also observed in newly diagnosed AML when venetoclax was combined with low‐dose cytarabine in a randomized trial; however, OS was not significantly improved, which contributed to the EMA's decision not to approve the combination for this indication [59]. Positive predictors for response to HMA and venetoclax include mutations in IDH2 and NPM1, and negative predictors include mutations in TP53, FLT3‐ITD or K / NRAS [3, 141]. The main side effect is haematotoxicity, and some patients experience extended cytopenias already after the first treatment cycle [3].

Resistance to venetoclax can be mediated via various mechanisms, including activation of tyrosine kinase signalling (e.g., via mutations in genes such as FLT3, NRAS, KRAS), inactivation of TP53, mutations in BCL2 that impair the binding of venetoclax and inactivating mutations in BAX [141, 142, 143, 144, 145]. Moreover, leukemic cells may be dependent on other anti‐apoptotic Bcl‐2‐family proteins, exemplified by a relative dependence on MCL1 in monocytic leukemic progenitors and Bcl‐xL in leukemic cells with erythroid/megakaryocytic features [146, 147]. These resistance factors may be present at diagnosis, resulting in primary resistance, or may be acquired or enriched during therapy, resulting in relapse. Compounds inhibiting MCL1 and Bcl‐xL have also been explored, with and without venetoclax, with promising efficacy. Some safety concerns have been observed, including potential cardiac toxicity when using MCL1 inhibitors, and optimal drug combinations and dosage of these novel BH3‐mimetics remain to be defined [138].

In R/R AML, venetoclax combinations have been extensively explored and will be covered in other sections of this review. Promising combinations include HMAs and venetoclax, particularly in HMA‐naïve patients, and combinations with intensive chemotherapy or kinase inhibitors and other targeted drugs.

Several novel BH3 mimetics are under development in AML, including the next generation Bcl‐2 inhibitor sonrotoclax [148]. Careful assessment of toxicity will be important, and it remains to be seen if they eventually will replace venetoclax in up‐front therapy or be reserved for R/R AML.

Low intensity chemotherapy combinations with venetoclax

Hypomethylating agents and venetoclax

Treatment with HMAs and venetoclax has revolutionized treatment of newly diagnosed AML, whereas the response rate and duration in R/R AML is markedly lower. The probability of response largely depends on the underlying disease biology. Recently, a large retrospective analysis compared patients with R/R AML who had failed intensive chemotherapy and received azacitidine with or without venetoclax. The composite CR, CRc, (CR + Cri + CR with incomplete platelet recovery [CRp]) rate was 55 versus 22%, with and without venetoclax, respectively, and the OS was significantly longer both in patients receiving subsequent allo‐HSCT and in those who did not [66]. Combinations of azacitidine and venetoclax may also induce durable responses in up to 40% of heavily pre‐treated children with R/R AML [67]. Positive predictors for better response to the combination were, as expected, mutations in IDH1, IDH2 or NPM1 [66].

Cladribine, low‐dose cytarabine and venetoclax

A lower intensity regimen for newly diagnosed AML aimed at older or unfit patients, combining cladribine, low‐dose cytarabine and venetoclax (CAV) alternated with azacitidine‐venetoclax, has shown an impressive CR/CRi rate of 93% in phase 2 data [68]. In the relapsed setting, this would be an attractive option in patients failing azacitidine‐venetoclax who are ineligible for intensive treatment or who are lacking druggable targets.

However, in a retrospective analysis of 39 elderly patients (median age 65 years) with R/R AML and intermediate or high‐risk genetics, of which the majority had received azacitidine‐venetoclax as prior therapy, the CAV regimen only resulted in a 28% CR/CRi rate [69]. Similar results were observed in prospective non‐randomized study in R/R AML, including 30 patients who were considerably younger (median age 40 years) and with a large proportion of favourable genetics (47%), with a CR/CRi rate of 27% [70]. However, the ORR was high (70%), and the majority (60%) could be transitioned to allo‐HSCT.

Importantly, monocytic leukaemia appears to respond better to CAV, which is in line with preclinical evidence that the addition of cladribine to azacitidine‐venetoclax targets the monocytic leukemic stem cell [149]. AML with monocytic phenotype is associated with mutations in the mitogen‐activated protein kinase (MAPK) pathway, and MAPK pathway mutations are known to respond less favourably to azacitidine‐venetoclax [149, 150]. In the trials above evaluating CAV in R/R AML, the response rates for patients with RAS mutations were unfortunately inconsistent, with 0 of 8 K/NRAS and 7 of 9 NRAS mutated patients achieving CR/CRi in the two trials, respectively [69, 70].

In summary, cladribine, low‐dose cytarabine and venetoclax is a tolerable option to consider in R/R AML with promising responses in particular in monocytic AML, whereas molecular predictors warrant further exploration.

High‐intensity chemotherapy combination with venetoclax

FLAG‐Ida with venetoclax (FLAVIDA)

Addition of venetoclax to FLAG‐Ida (FLAVIDA) has been attempted in a phase‐2 trial in R/R AML with improved CR rates of 47%, MRD negativity of 74% and 1‐year survival of around 50% [56, 57]. Similar results were obtained from registry data even though during the short follow‐up and with a very small cohort of 13 patients, no differences in outcome were seen as compared to a control cohort of FLAG‐Ida patients (n = 81) [58].

Other intensive regimens with venetoclax

Venetoclax is also being explored together with several other intensive regimens, both in the up‐front and R/R setting. In newly diagnosed AML, the prospective CAVEAT trial—combining venetoclax with intensive chemotherapy—showed that this combination was efficacious and well tolerated even in elderly patients [151]. In R/R AML, a retrospective report demonstrated that venetoclax added to cladribine, high‐dose cytarabine and idarubicin or mitoxantrone resulted in ORR of 77% versus 50% with or without venetoclax, respectively [152]. These intensive combinations need further exploration, with particular focus on toxicity, until they can be considered for broader use.

In early phase trials in children, venetoclax in combination with intensive chemotherapy appears well tolerated, and promising efficacy has been reported [59].

FLT3 inhibitors including combinations

Mutations in FMS‐related tyrosine kinase 3 (FLT3) occurs in 20%–30% of patients with AML and is often associated with rapidly proliferating disease. Two main categories of mutations occur; ITD or activating mutations in the tyrosine kinase domain (TKD). Available FLT3 inhibitors differ in their kinase specificity and in their mode of action. Type 1 inhibitors (midostaurin, gilteritinib, crenolanib) target both ITD and TKD mutations, whereas type 2 inhibitors (sorafenib, quizartinib) target only ITD. Mechanisms of resistance include acquisition of other downstream kinase mutations or mutations in FLT3 that are resistant to the utilized inhibitor, including the known gatekeeper mutation F691L, which confers resistance to all currently approved FLT3 inhibitors [153].

In newly diagnosed AML, midostaurin and quizartinib are approved by the EMA as an add‐on to intensive chemotherapy, based on confirmed survival benefit in randomized trials [26, 27]. Quizartinib in combination with chemotherapy has an acceptable toxicity profile in children and is currently tested by the NOPHO‐DB‐SHIP consortium [32]. Sorafenib lacks formal approval in AML by the EMA but is recommended in several guidelines, including the Swedish National Guidelines for maintenance treatment post allo‐HSCT, also based on results from randomized trials [97]. Sorafenib administered in between courses of primary therapy is well tolerated in children and showed encouraging signals of activity in FLT3‐ITD AML with high allelic ratio [31]. Gilteritinib used as post‐transplantation maintenance improves survival in patients who have detectable MRD for FLT3‐ITD pre‐ or post‐transplantation; however, at present, gilteritinib is not approved in this setting by the EMA [28].

In R/R AML, the only drug currently approved by the EMA is gilteritinib (Fig. 2). Used as a single agent in R/R AML with FLT3‐ITD or TKD, the CRc rate was 54.3% versus 21.8% for conventional salvage therapy, with a significant improvement in median OS of 9.3 versus 5.6 months (HR 0.64, p < 0.001). In children, only limited data on gilteritinib is available. Overall, gilteritinib is well tolerated, with moderate haematotoxicity, although other significant side effects may occur, including differentiation syndrome [75].

Preclinical data indicate a synergy between FLT3 inhibitors and venetoclax [154]. So‐called doublet therapies (FLT3 inhibitor + venetoclax) and triplet therapies (FLT3 inhibitor + HMA + venetoclax) have been explored in phase 1 and 2 trials in R/R AML, with impressive response rates. The mCRc (CRc + MLFS, morphologic leukaemia‐free state) for gilteritinib combined with venetoclax was 75%, and 44% of evaluated patients achieved at least a 3‐log reduction of the FLT3‐ITD MRD level [76]. Corresponding figures for the triplet of gilteritinib, azacitidine and venetoclax were an mCRc rate of 73%, with MRD negativity in 43% of evaluable patients [77]. Somewhat disappointingly, both the gilteritinib doublet and triplet therapies yielded similar median OS as gilteritinib alone (9.3, 10.0, 5.8 months, respectively, for singlet, doublet, triplet therapy). Despite high initial efficacy, relapse remains a significant challenge even in combination trials. Another caveat is that both doublet and triplet therapies are highly myelotoxic, with frequent extended cytopenias in addition to other adverse events.

Novel trials exploring other FLT3 inhibitor combinations aim to optimize dosing and treatment duration of the doublet or triplet drugs during each treatment cycle, to improve both efficacy and tolerance. Moreover, implementing doublet or triplet therapy as an early salvage therapy likely leads to deeper responses and potentially improved long‐term outcome, particularly for patients bridged to allo‐HSCT.

Finally, available FLT3 inhibitors are not fully specific for FLT3, and inhibition of other kinases may play a role in their anti‐leukemic effect. Clinical and pre‐clinical data suggest that FLT3 inhibitors may be efficacious both as single agents and in combination with chemotherapy or venetoclax, even in subsets of AML patients without FLT3 mutations [154]. Intriguingly, a randomized trial with sorafenib as an add‐on to conventional chemotherapy demonstrated an improved event‐free but not OS, indicating an added efficacy but also increased toxicity [155]. Moreover, preliminary results from a randomized trial of standard chemotherapy with or without quizartinib demonstrated and increased OS in FLT3‐negative patients (median not reached vs. 29.3 months, hazard ratio 0.63) [78]. Quizartinib is being further explored in this setting in a recently started prospective randomized phase 3 trial (QuANTUM‐Wild) [156].

IDH inhibitors including combinations

Mutations in IDH 1 or IDH 2 occur in around 20% of patients with AML, with IDH2 being somewhat more frequent. Several inhibitors have been developed, and the most extensively investigated are ivosidenib (IDH1) and enasidenib (IDH2). When the mutant IDH protein is inhibited, the oncometabolite 2‐hydroxyglutarate is formed, which interferes with the cellular metabolism — resulting in DNA hypermethylation by impairing the function of TET2 — and also induces Bcl‐2 dependency to evade apoptosis [157, 158]. These mechanisms of action provide rationales for combining IDH inhibitors with HMAs and Bcl‐2 inhibitors. In line with this, clinical data have demonstrated high activity of venetoclax‐based regimens in patients with IDH mutations [3].

In R/R AML treated with ivosidenib, the CRc rate was 30.4% and the median response duration 8.2 months [79]. In IDH2 mutated R/R AML, enasidenib resulted in a CRc rate of 27.8%, with a median response duration of 5.6 months. IDH inhibitors are well tolerated, and a class effect is that they may result in differentiation syndrome.

To improve efficacy, IDH inhibitors have been combined both with low‐ and high‐intensity chemotherapy. In newly diagnosed AML with IDH1 mutation, a randomized trial of azacitidine with or without ivosidenib demonstrated a marked improvement in OS for the combination, which has led to market approval by tEMA [159]. Moreover, in newly diagnosed AML, ivosidenib or enasidenib have been combined with standard intensive chemotherapy, with promising activity, and the concept is now being explored in a randomized setting [80]. Interestingly, a retrospective analysis of 398 newly diagnosed AML patients treated in a randomized fashion with ara‐C and daunorubicin plus/minus cladribine suggested that IDH2 mutations conferred a better survival in patients receiving cladribine, possibly through a reversal of IDH2‐induced hypermethylation [160] (Fig. 1). IDH inhibition as maintenance therapy after induction and consolidation, as well as after allo‐HSCT, is an important area currently under investigation, where early data are promising [98].

The triplet of ivosidenib, venetoclax and azacitidine has shown promising activity in newly diagnosed AML, whereas the results were less impressive in R/R AML, with a CRc rate of 63% and an OS of 9 months [81]. Moving forward, novel IDH inhibitors may prove to have even higher efficacy, and a key issue will be how best to combine the targeted inhibitors with other available drugs.

Menin inhibitors including combinations

The complex of the nuclear scaffold protein menin and the histone‐lysine N‐methyltransferase 2A (KMT2A) protein constitutes a molecular dependency for pro‐leukemic gene expression in acute leukaemia characterized by enforced transcription of HOX homeobox genes and the co‐factor myeloid ecotropic virus insertion site 1 (MEIS1) [161]. These entities include AML with KMT2A‐rearrangements (KMT2A‐r) [162, 163] as the most prominent example but also AML with rearranged nucleoporin 98 (NUP98‐r) [164], mutated nucleophosmin (NPM1) [165] and tandem duplications in the upstream binding transcription factor (UBTF‐TD) [166]. KMT2A‐r occur in up to 10% of all acute leukaemias in adults and children, are particularly frequent in infants with ALL or AML [8], and are generally associated with high risk of relapse [167]. During the last decade, extensive preclinical research using models of acute leukaemia with aberrant HOX/MEIS1 expression has demonstrated how menin inhibitors selectively block the menin‐KMT2A interaction, leading to restoration of myeloid differentiation, leukaemia cell apoptosis and loss of the leukemic transcription signature [165, 166, 168, 169, 170]. These results established the foundation of recently initiated early‐phase clinical trials of menin inhibitors in the setting of R/R AML (Fig. 2).

The first‐in‐human study with a menin inhibitor was AUGMENT‐101 (NCT04065399), a phase 1/2 trial of revumenib in children and adults with R/R acute leukaemias. The toxicity profile was acceptable, mainly with mild to moderate cytopenias and a moderate risk of differentiation syndrome and QTc prolongation [82]. In the biomarker‐driven phase 2 portion of AUGMENT‐101, NPM1‐mutated patients are still enrolling while efficacy data on 57 patients with KMTA2‐r acute leukaemias showed an overall response rate of 63% of whom 15/22 responders (68%) achieved MRD negativity, and 14 patients (39%) were bridged to allo‐HSCT [83].

In the KOMET‐001 phase 1/2 trial of ziftomenib in adults with R/R AML, the safety profile was manageable (cytopenias and differentiation syndrome). Twelve of 38 patients (32%) with KMT2A‐r or NPM1‐mutated AML receiving ziftomenib at the highest dose level of the dose‐validation phase achieved an overall response [84].

Early phase clinical trials of bleximenib (NCT04811560) and enzomenib (NCT04988555) are currently recruiting adult acute leukaemia patients, and preliminary results indicate comparable safety and efficacy profiles across the different menin inhibitors [171, 172].

Genome‐wide sequencing of bone‐marrow samples from patients with only temporary responses to revumenib in the AUGMENT‐101 identified mutations in the MEN1 gene conferring structural changes of the KMT2A‐binding pocket that prohibit interaction with the inhibitor but without affecting menin‐KMT2A complex formation. Longitudinal tracking by digital PCR demonstrated emergence of these MEN1 mutations in approximately 40% of patients continuously exposed to revumenib [173].

Application of contemporary chemotherapy to menin inhibitors may be an appealing strategy to obtain therapeutic synergy and prevent expansion of MEN1‐mutated leukemic clones. A number of trials, including both intensive and less‐intensive chemotherapy, with or without venetoclax, in combination with menin inhibitors, are currently recruiting patients with newly diagnosed and R/R acute leukaemias. Given the acquisition of resistance to menin inhibitors over the course of treatment reported from the AUGMENT‐101 trial [173], it will be important to monitor resistance in these trials and to evaluate how combination therapies may be used to circumvent resistance.

Targeting the p53 pathway

Up to 20% of patients with newly diagnosed AML harbour TP53 aberrations [174], and TP53‐mutated AML is strongly associated with resistance to traditional therapy options and higher rates of relapse after allo‐HSCT [174, 175, 176]. One study found that 15% of patients with R/R AML had TP53 mutations that were not detectable in their primary disease, suggesting that these mutations can emerge under selective pressure of treatment [177]. Given the inferior outcomes of TP53‐mutated AML with traditional regimens, targeting the p53 pathway and harnessing the immune system (see below) have been evaluated as alternative strategies.

APR‐246 (eprenetapopt) is a p53 activator, albeit with many off‐target effects [178]. Two parallel phase 2 trials in patients with newly diagnosed TP53‐mutated MDS or AML demonstrated that APR‐246 in combination with azacitidine yielded high CR rates (40%–50%) and notably improved OS in patients who were bridged to allo‐HSCT [179, 180]. One trial combining APR‐246 and azacitidine as maintenance therapy following allo‐HSCT in TP53‐mutated MDS/AML showed promising OS [181]. Although trials using APR‐246 in the R/R setting are lacking, results indicate it may be beneficial in bridging TP53‐mutated patients to transplant, and its use in the setting of R/R AML warrants further investigation.

Leukemic cells with wild‐type TP53 can be sensitized to p53‐mediated apoptosis by MDM2 inhibition. One phase 1b trial tested a combination of venetoclax and the MDM2 inhibitor idasanutlin in a cohort of patients with R/R AML with modest results [182]. Another study in the R/R AML setting with a different MDM2 inhibitor (milademetan) showed similar results [183]. This effect was not enhanced when combining milademetan with venetoclax and low‐dose cytarabine in R/R AML [184].

Ex vivo drug screening to guide therapeutic choices

Evaluating laboratory responses to anti‐cancer therapies prior to clinical treatment—inspired by the success of antibiotic sensitivity testing in microbial diseases—has been a longstanding goal in leukaemia and cancer research [185]. Multiple trials have investigated the feasibility of ex vivo drug sensitivity testing for AML in a prospective setting.

The Helsinki Functional Precision Medicine Tumour Board combined clinical, molecular and functional data to guide AML treatment decisions [92]. In 39 of 186 R/R AML patients, recommendations based on drug response data resulted in a 59% response rate for tailored therapies. This trial primarily relied on bulk‐sensitivity assays. The EXALT trial profiled 143 patients with advanced haematologic malignancies using a microscopy‐based single‐cell readout, with 56 participants treated [93]. Among these, 54% achieved clinical benefit, demonstrating more than a 1.3‐fold increase in PFS compared to prior therapies, although data specific to AML patients were limited. Similarly, the SMART trial provided drug‐response profiling reports to 91% of participants within 7 days, linking ex vivo chemotherapy resistance to in vivo treatment failure. However, this study mainly focused on newly diagnosed AML cases [186].

From these studies, it is clear that ex vivo drug sensitivity testing for AML is technically achievable. Due to the nature of these studies, they did not systematically compare ex vivo drug responses with clinical outcomes in a prospective manner.

Translating ex vivo drug testing into clinical care requires careful consideration of methodological factors, including variations in cell culture media and viability testing techniques. Previous studies [187, 188] have shown that evaluating response using bulk bone marrow samples can lead to false predictions of resistance. Flow cytometry analysis demonstrated that distinct cell populations, such as monocytic and granulocytic cells, have different sensitivities to venetoclax, potentially masking blast‐specific responses [94, 188].

The prospective VenEx trial [94, 95] (NCT04267081) evaluated the correlation between ex vivo venetoclax sensitivity and clinical outcomes with venetoclax and azacitidine in patients with AML. To systematically assess the variables, the trial's initial phase, involving 39 participants, tested three cell culture media and two viability methods. The highest correlation with clinical outcomes was achieved using conditioned culture medium and blast‐specific response evaluation by flow cytometry. Final results from VenEx demonstrated the practicality of integrating ex vivo drug sensitivity testing into clinical workflows and highlighted its value in identifying AML patients likely to benefit from venetoclax, particularly in the R/R setting.

Ex vivo venetoclax sensitivity testing was successfully performed in 98% of participants, with results available within 3 days of sampling. In R/R and secondary AML patients, where treatment options are often limited, ex vivo sensitivity corresponded to a 62% CR/CRi rate and a median OS of 9.7 months compared to 3.3 months in resistant patients. In untreated AML, sensitivity was associated with an 85% CR/CRi rate and a median OS of 28.7 months. Additionally, ex vivo sensitivity correlated with MRD negativity and emerged as the most significant predictor of favourable outcomes in univariate and multivariate analyses [94, 95].

These findings underscore the transformative potential of ex vivo drug sensitivity testing for tailoring AML treatments. By addressing technical challenges and enhancing predictive accuracy, this approach holds promise for broader clinical implementation and improved patient outcomes.

In recent years, efforts have been undertaken to establish patient‐derived AML reconstitution in PDX models to study drug sensitivity. Following the initial challenges in engrafting immunodeficient mice with primary AML cells, protocols were developed to allow for such engraftment [189]. Interestingly, in such a model, lower EFS rates in patients were associated with more successful reconstitution of their primary AML cells in PDX animals. Furthermore, enrichment was observed for R/R clones upon serial transfer in PDX animals that mimicked clonal evolution upon R/R disease in corresponding patients [190]. Although several groups have reported the use of PDX models to test specific drugs, the models still lack robustness to effectively engraft AML blasts from all patients; hence, implementation in clinical practice has yet to happen.

Immunotherapy and allo‐HSCT

Allo‐HSCT plays a critical role in the management of R/R AML and is one of the few potentially curative treatment options for these high‐risk patients (Table 1). In cases in which AML is refractory to initial therapy or has relapsed following a period of remission, outcomes with conventional salvage chemotherapy are generally poor. Allo‐HSCT offers the benefit of both intensive cytoreduction and a GvL effect mediated by donor immune cells, which can target residual leukemic cells and reduce the risk of relapse. Patients with R/R AML who achieve a second CR or even proceed to transplantation with residual disease may benefit from transplantation [5]. Novel strategies — such as the use of reduced‐intensity conditioning, haploidentical donors and post‐transplant maintenance therapies — make allo‐HSCT more accessible and may improve outcomes in this challenging patient population [96]. Despite these advances, relapse remains a significant cause of post‐transplant mortality.

The monoclonal antibody drug conjugate GO that has been described as an augmentation to conventional chemotherapy for CD33+ disease above (Fig. 2) does not in itself elicit direct immunotherapeutic effects such as antibody‐dependent cellular cytotoxicity [191]. However, GO simultaneously targets CD33+ myeloid‐derived suppressor cells [192] and has therefore been suggested for a clinical trial in patients with solid tumours (EUDRACT: 2020‐002428‐36).

Immune checkpoint inhibitors have been suggested in particular for patients who do not respond to HMAs, as this was correlated with higher expression of PD‐1/PD‐L1 and therefore suggested to be a resistance mechanism (reviewed in [71]). Accordingly, anti‐PD‐1 (nivolumab or pembrolizumab) antibodies have been evaluated in combination with azacitidine in non‐randomized phase 2 clinical trials for R/R AML with an ORR of 18%–33% [72, 73]. Pembrolizumab has also been given in a phase 2 trial as a maintenance treatment following high‐dose ara‐C with a CR rate of 38% in R/R AML [74].

Targeting immune evasion, CD47‐SIRPα interaction has been explored extensively. CD47 is a cell surface protein expressed on malignant cells often referred to as a ‘don't eat me’‐signal binding to SIRPα on macrophages, inhibiting phagocytosis. Blocking the interaction with anti‐CD47/SIRPα‐antibodies or Fc fusion proteins promote phagocytosis of leukaemia cells by macrophages and other immune cells. The first‐in‐class monoclonal antibody against CD47, magrolimab, showed promising early‐phase efficacy in combination with azacitidine in previously untreated AML [193]. The initially high expectations of the anti‐CD47/SIRPα concept have been dampened when FDA placed all studies of magrolimab in haematologic malignancies on hold due to futility and increased risk of death. However, several other drugs targeting CD47/SIRPα combined with azacitidine ± venetoclax are currently being investigated in phase 1/2 clinical trials in both newly diagnosed and R/R AML [194].

Following the success of anti‐CD19 chimeric antigen receptor (CAR)‐T cells in lymphoid disease, efforts have been made to introduce CAR‐T cell treatment for R/R myeloid disease. CAR‐T cells have been developed to target, for instance, CD33, CD123 or CD135 [195]. A complicating factor in these approaches is that many of these targets are expressed on haematopoietic stem and/or progenitor cells, resulting in severe and prolonged cytopenia. In addition, downregulation of antigens (immune escape) is frequently observed in relapsed AML [196], which hinders antigen recognition by CAR‐T cells. The development of newer generations of CAR‐T cells is aimed at overcoming these challenges.

Finally, a large number of other immunotherapies are being explored that are not covered here due to a paucity of clinical data and space limitations. This field is developing rapidly, as recently reviewed by Bawek et al. [71]. Given the impressive results in other haematological malignancies, we trust that, with further optimization, immunotherapies will play a key role in the future management of R/R AML, in particular for patients who are ineligible for curative allo‐HSCT.

Conclusions and future directions

Despite the rapidly expanding portfolio of active anti‐leukemic agents and the broader availability of genetic and phenotypic diagnostic tools, including MRD monitoring, the treatment of R/R AML remains a clinical challenge, and the long‐term overall outcome is poor. Novel combinations, including targeted drugs, hold promise to improve outcomes in specific subtypes of AML. Conversely, extreme high‐risk diseases, such as double‐hit TP53 mutated AML, require particular efforts and entirely new approaches, ideally directed at overcoming the intrinsic apoptosis resistance.

Conventional clinical trials, even when biomarker‐driven, are hampered by the inherent clinical and biological heterogeneity of the patient population; in children, a limited number of patients, unequal access to new drugs and limited health economy resources are further complications. Multimodal precision oncology approaches appear to be a rational strategy for guiding future practice, but they require prospective evaluation (Fig. 3). It is important to consider the genetic and phenotypic architecture at high resolution (e.g., massive parallel sequencing allowing for detection of minor allele frequencies) as well as the intra‐patient disease heterogeneity (e.g., single‐cell based approaches) in combination with informed ex vivo testing of drugs and drug combinations with flow cytometry‐based readouts. Regarding maintenance therapies, proper ex vivo simulation remains difficult. We believe that a combination of genetic and phenotypic data together with ex vivo drug profiling may allow better individualized treatment decisions. In addition, new methods need to be developed to allow prediction of the efficacy of immunomodulating and immunotherapeutic drugs.

Fig. 3.

Schematic of treatment modalities for treatment of relapsed and refractory (R/R) acute myeloid leukaemia (AML) with intention‐to‐cure. Upper panel: To cure R/R AML, remission induction followed by consolidation and maintenance therapy is advised. Lower panel: Remission induction heavily relies on intensive chemotherapy that might be informed by ex vivo drug testing. Intensive chemotherapy should be complemented by targeted therapy (as outlined in Figs. 1 and 2), and immunotherapy might be considered for individual cases. Consolidation is achieved by allogeneic haematopoietic cell transplantation (allo‐HSCT). Maintenance therapy is usually reserved for targeted agents, but also hypomethylating agents (HMAs) and immunotherapy might be considered (see Table 1).

We trust that ongoing efforts to further understand the complex biology of AML and improve diagnostic tools and predictive capabilities will serve as a foundation to develop novel and more efficient patient‐tailored therapies, including a combination of conventional chemotherapy, targeted drugs and immunotherapeutic approaches.

Conflict of interest statement

M.J. has been part of AdBoards (receiving institutional support) for AbbVie and Cancell Therapeutics. He has also been involved in educational events (receiving institutional support) for AbbVie, Pfizer, Lab Delbert, and Astra Zeneca. M.K. has been part of AdBoards for Faron and Proteina. He has also been an advisor for BMS and Astellas, served as a consultant for Ferring and Faron, and been a member of the speaker's bureau for AbbVie, Servier, Faron and Jazz. The remaining authors declare no conflicts of interest.

Jädersten M, Lilienthal I, Nilsson C, Fredrikson L, Pronk CJ, Juul‐Dam KL, et al. Precision oncology to overcome resistance in R/R AML in children and adults requires combinations of cytotoxic, targeted, and immunological treatments. J Intern Med. 2025;298:297–318.