Background

Acute myeloid leukemia (AML) is the most common acute leukemia in adults. AML is thought to arise from somatically acquired mutations, which is a fairly ubiquitous process during human aging [1,2,3,4,5,6,7,8,9,10,11]. However, AML can arise both de novo and secondary to other processes including antecedent hematologic disorders or exposure to immunosuppressive or cytotoxic therapies. The median age of diagnosis of AML is 68 years in the USA, and the incidence continues to increase with age [12]. The median overall survival (OS) from diagnosis for patients under the age of 65 had improved from 8 months between 1975 and 1979 to 46 months between 2010 and 2014; however, survival in patients older than 65 has only marginally improved during this same interval [12, 13]. This is in part because induction with intensive cytarabine- and anthracycline-based chemotherapy (i.e., “7 + 3”) has remained the standard of care for AML for over 40 years, the tolerability of which is limited in older and less fit patients.

While optimization of intensive chemotherapy and better supportive care over the years has improved survival in AML, this benefit is largely confined to younger patients and those without adverse-risk cytogenetics. Moreover, the only potentially curative strategy for those with intermediate- or adverse-risk disease is allogeneic hematopoietic stem cell transplantation (alloHSCT) [14], which is not an option for many patients with AML due to age, frailty, and medical co-morbidities [15]. Accordingly, there has been a considerable interest in de-intensifying induction therapy, guided by an improved understanding of AML pathobiology due to advances in genomic profiling [16,17,18]. This has led to the approval of multiple novel agents and targeted therapies, which are now increasingly employed in the frontline, relapsed/refractory (R/R), and maintenance settings (reviewed in [19,20,21,22]). However, with 10 new Food and Drug Administration (FDA) approvals for AML in the last 5 years and increasing availability of personalized genomic data, important questions arise of how to best personalize the treatment of patients with AML and how to utilize transplantation in the context of targeted therapies (Fig. 1). In this review, we will discuss the recent updates in classification and risk stratification of AML, explore combination therapies in clinical practice and situations in which intensive chemotherapy can potentially be replaced, and highlight ongoing areas of investigation for AML therapeutics.

Fig. 1
figure 1

Approach to frontline treatment of AML with FDA-approved therapies in 2022. Treatment algorithm of AML induction and maintenance therapy is shown. AML Acute myeloid leukemia, GO Gemtuzumab ozogamicin, MIDO Midostaurin, HMA Hypomethylating agent, VEN Venetoclax, ENA Enasidenib; IVO Ivosidenib, LDAC Low-dose cytarabine, alloHSCT Allogeneic stem cell transplantation, MRD Measurable residual disease, CR Complete remission

Full size image

Updates in classification and risk stratification of AML

In 2022, the World Health Organization (WHO) [23] updated their classification of hematolymphoid neoplasms, and a separate International Consensus Classification (ICC) system was formed [24]. While these have led to changes in many hematologic malignancies, we will henceforth focus on changes related to AML (summarized in Table 1). Many disease entities remain the same, and most changes revolve around incorporating cytogenetic and genetic information into diagnostic classification. A significant change is that myelodysplastic syndrome with excess blasts 2 (MDS-EB2) is no longer a recognized MDS subtype, reflecting a spectrum of disease between MDS and AML rather than a discrete transition at the arbitrary cutoff of 20% blasts. The WHO allows for a diagnosis of AML with a blast count below 20% if there are defining genetic abnormalities with the exception of BCR::ABL1 fusions and CEBPA mutations. Cases of AML without a defining genetic alteration are made based on differentiation patterns and still require a blast count of at least 20%. Similarly, the ICC uses a 10% blast cutoff for most molecularly defined subtypes of AML, again with the exception of BCR::ABL1 to avoid confusion with the diagnosis of chronic myeloid leukemia (CML). Moreover, both the WHO and ICC define cytogenetic and genetic changes almost always associated with antecedent MDS (Fig. 2A–B), while the cytogenetic profile has not drastically changed with a few exceptions, mutations in ASXL1, BCOR, EZH2, SF3B1, SRSF2, STAG2, U2AF1, or ZRSR2 now define AML with myelodysplasia-related gene mutations given the high prevalence of these mutations in MDS; RUNX1 mutations are also considered MDS-defining by the ICC but not by the WHO [25,26,27,28]. By relying on molecular rather than morphologic classification of AML, this may allow for nuanced prognostication and definition of targets for measurable residual disease (MRD) monitoring during treatment, though there is still no standard of how to incorporate this into routine care. Two major differences between the updated WHO and ICC systems are how AML with CEBPA and TP53 alterations are defined. Several studies have now shown that basic leucine zipper (bZIP) domain mutations in CEBPA confer a better prognosis with a distinct gene expression profile [29,30,31], so this is now an AML-defining genetic alteration in both classification systems, though the WHO still also includes other bi-allelic CEBPA mutations, while the ICC does not. Furthermore, the ICC defines TP53-mutated AML and MDS/AML as a distinct genetic entity due to the characteristically poor prognosis associated with this mutation [32,33,34]. Any somatic TP53 mutation with a variant allelic frequency (VAF) above 10% now defines this subtype of MDS/AML or AML. The WHO system created a distinct entity for TP53-mutated AML due to the frequent co-occurrence with complex cytogenetics or therapy-related AML (tAML).

Table 1 Updates in WHO/ICC classifications of AML
Full size table
Fig. 2
figure 2

Updates in WHO/ICC MDS-defining genetic alterations in AML. Venn diagrams depict overlapping and distinct MDS-defining cytogenetic alterations A and somatic mutations B determined by the WHO [23] and ICC [24] Green text with arrows denotes overlapping entities with minor differences, black text denotes completely overlapping entities, and red text denotes completely distinct entities. MDS Myelodysplastic syndrome WHO World Health Organization, ICC International Consensus Classification

Full size image

These changes reflect a movement toward molecularly defining AML and related myeloid neoplasms. The natural history of these diseases suggests some overlap in pathogenesis with biological variability perhaps better attributed to distinct genetic driver events rather than morphologic differences. Thus, the European LeukemiaNet (ELN) also updated their risk stratification schema (Fig. 3) to reflect this change [35]. The favorable prognostic impact of CEBPA mutations is driven by bZIP domain mutations [29,30,31], so this is now specified in the favorable-risk category. Moreover, FLT3 internal tandem duplication (FLT3-ITD) mutations had previously been considered a VAF-dependent risk factor in patients who also harbor NPM1 mutations. However, there is variability in the standardization of measuring allelic ratio (AR), and the incorporation of FLT3-targeting multi-tyrosine kinase inhibitor (TKI) midostaurin (MIDO) has demonstrated benefit in patients with FLT3-ITD regardless of VAF [36]. Thus, for epidemiologic and practical reasons, FLT3-ITD is now considered intermediate-risk irrespective of AR or co-occurring NPM1 mutations. Although patients with FLT3-ITD who did not receive MIDO had worse outcomes regardless of ELN 2017 risk category [36], MRD testing could be considered for more dynamic risk stratification after induction therapy and/or prior to alloHSCT given its prognostic relevance even for patients treated with chemotherapy only [37, 38]. In concordance with the expanded list of myelodysplasia-associated genes from the WHO and ICC, mutations in ASXL1, BCOR, EZH2, RUNX1, SF3B1, SRSF2, STAG2, U2AF1, or ZRSR2 are now considered adverse-risk based on updated prognostic studies in MDS, de novo AML, and secondary AML (sAML) [25,26,27,28, 39, 40]. Lastly, new disease-defining cytogenetic changes involving MECOM [41, 42] or KAT6A::CREBBP fusion [43] have been updated in adverse-risk disease, while hyperdiploid karyotype with multiple trisomies/polysomies is no longer considered complex karyotype. The changes to the ELN risk stratification system reflect an evolving understanding of how AML biology impacts clinical phenotype. However, genetic drivers of AML do not always appear in isolation [16, 44], and, as we will discuss later in this review, co-mutation patterns often have conflicting prognostic implications. Accordingly, a more comprehensive incorporation of genetic and cytogenetic alterations along patient and disease characteristics, as has recently been implemented in MDS [40], may improve that personalization of risk stratification.

Fig. 3
figure 3

Updates in ELN risk stratification of AML. A Sankey plot depicts changes in the 2017 [65] and 2022 [35] ELN risk stratification of AML. Prognostic groups are groups by color (favorable—green, intermediate—yellow, adverse—red) and changes are tracked by dashed arrows. bZIP Basic leucine zipper domain, ELN European LeukemiaNet

Full size image

Azacitidine and venetoclax: moving up the ranks

AML in those unfit for intensive chemotherapy

It should be noted that while updates from the WHO, ICC, and ELN emphasize the importance of molecular pathogenesis in AML, there are other patient characteristics that significantly impact prognosis and treatment, namely age and fitness. The National Comprehensive Cancer Network (NCCN) guidelines divide induction treatment algorithms based on age below or above 60 years [45]. While the age of 60 is not uniform in its capacity to discriminate the ability to tolerate high-intensity chemotherapy, clinical trials have traditionally been organized around this age. However, in clinical practice, these historical age cutoffs warrant reconsideration when patients are physiologically fit despite being chronologically older. Historically, non-intensive therapy included the use of single-agent azacitidine (AZA) or decitabine (DAC) with a median OS reaching a dismal 7.7 to 10.4 months [46, 47]. The combination of the BCL2-inhibitor venetoclax (VEN) with AZA or low-dose cytarabine (LDAC) has changed the standard of care for this patient population. The VIALE-A study [48] compared AZA/VEN to AZA/placebo in an elderly population (median age 76 years) and observed an improvement in the rate of complete remission (CR) plus CR with incomplete hematologic recovery (CRi) (66.4% versus 28.3%, respectively) and median OS (14.7 months versus 9.6 months, respectively). Long-term follow-up from this study confirms these findings and notably observed that patients who achieved CR/CRi and MRD negativity with AZA/VEN had a median OS of 34.2 months [49]. Similarly, the VIALE-C study [50] randomized patients with a median age of 76 years to LDAC/VEN or LDAC/placebo, noting an improvement in the rate of CR/CRi from 13 to 48% along with survival benefit at long-term follow-up [51, 52]. Recently, a study by Pollyea et al. found that the use of alloHSCT in patients over 60 years who received AZA/VEN improved median OS compared to those who deferred alloHSCT (not reached versus 17.2 months) [53]. Another retrospective study from the same group [54] compared outcomes of patients undergoing alloHSCT after AZA/VEN to intensive chemotherapy and found no significant difference in recurrence-free survival (RFS) (73.2% versus 66.1%, respectively) or OS (76.3% versus 74.7%, respectively) at 12 months. Age was not a predictive factor of death or relapse in this population, though higher hematopoietic cell transplantation-specific comorbidity index (HCT-CI) and positive pre-alloHSCT MRD were associated with worse outcomes. Theoretically, in older patients with high-risk disease, treatment with AZA/VEN may maintain or improve fitness, allowing more patients to undergo alloHSCT who may not have previously been eligible after receiving induction with 7 + 3. Currently, there is a study (NCT04801797) ongoing to address this question which randomizes patients fit for intensive chemotherapy to receive either 7 + 3 or AZA/VEN.

Secondary AML

As previously discussed, most cases of AML arise de novo, but sAML presents a unique clinical challenge. While the incidence of sAML increases with age [55], sAML frequently harbors adverse-risk cytogenetics and mutational profiles that are often associated with treatment resistance [56,57,58,59,60]. CPX-351, a fixed 5:1 molar ratio of cytarabine/daunorubicin liposome, was developed with the intention of optimizing a synergistic molar ratio of the two chemotherapeutics [61] and was noted to have activity in sAML in early data [62]. A subsequent phase 3 study assessed CPX-351 versus 7 + 3 in patients aged 60–75 with newly diagnosed (ND) tAML, antecedent MDS or chronic myelomonocytic leukemia, or de novo AML with MDS-related cytogenetics (AML-MRC) [63] and observed improved median OS (9.56 months versus 5.95 months) and CR/CRi rates (47.7% versus 33.3%), leading to FDA approval of CPX-351 for patients with newly ND tAML or AML-MRC. Moreover, long-term survival with CPX-351 was attributed to more patients proceeding to alloHSCT, which was performed in 34% of patients [63].

In clinical practice, there is significant overlap in patients who may be eligible for AZA/VEN or CPX-351 in the frontline setting, though these two regimens have not been compared head-to-head. A recent real-world analysis of patients with ND AML who received AZA/VEN or CPX-351 found that those receiving AZA/VEN were more likely to be older (median age 75 years versus 67 years, respectively) [64]. Despite the heterogeneity between cohorts, median OS for all patients was similar between the AZA/VEN and CPX-351 cohorts with no significant difference even when controlling for multiple factors including performance status, MRC, ELN 2017 risk category [65], high-risk mutations, and HCT-CI score. Although fewer patients underwent transplant with AZA/VEN induction, median OS was not significantly affected by the choice of CPX-351 versus AZA/VEN. Thus, while a randomized control trial is lacking, this study provides some clinical equipoise between AZA/VEN and CPX-351 in patients who may be eligible for either therapy.

AML with high-risk mutations

With advances in next-generation sequencing (NGS) technology, the detection of molecular driver mutations in AML has improved. Follow-up analyses of several landmark studies have allowed for the identification of certain mutational subgroups which may derive further benefit from AZA/VEN compared to intensive chemotherapy. Perhaps the most challenging scenario is with TP53-mutant AML. Recent data suggest that patients with TP53 mutations have worse outcomes compared to other adverse-risk features with a 2-year OS of only 12.8% even with intensive treatment [33]. The presence of TP53 mutations confers similarly poor outcomes irrespective of de novo disease, sAML, or blast count [34]. However, there is heterogeneity in TP53-mutant AML based on types of mutations, involved domains, and VAF; a recent report found that patients without abnormal p53 protein expression, intact copy number, and low VAF had a more favorable prognosis in the spectrum of TP53-mutant AML [66]. Nonetheless, even in patients who undergo alloHSCT, relapse rates are very high with median OS often reported as less than 6 months after transplant [67, 68], though this may be longer based on more recent retrospective studies [69]. In the VIALE-A study, CR/CRi for patients with TP53 alterations was drastically improved with AZA/VEN (55.3%) compared to AZA/placebo (0%), though the duration of remission (DOR) was brief, and median OS was only 7.2 months [48, 70].

Notably, AZA/VEN achieved CR/CRi rates of 70% compared to 23% with AZA alone in patients with adverse-risk cytogenetics without TP53 mutations, along with durable remissions (18.4 months versus 8.51 months) and improved median OS (23.4 months versus 11.3 months). However, the benefit in DOR or OS was lost with the addition of TP53 mutations despite improvements in CR/CRi (41% with AZA/VEN versus 17% with AZA) [71]. Unfortunately, several follow-up studies with hypomethylating agents (HMA) and VEN have produced similarly discouraging results [71,72,73]. The lack of durable response may stem from a requirement of intact p53 protein to maintain long-term response to BH3-mimetic drugs such as VEN [74]. While HMA/VEN-based regimens have not yet demonstrably moved the needle forward for patients with TP53-mutant AML, several studies are underway combining these agents in this high-risk disease; these will be discussed later in this review.

Two other high-risk gene mutations of interest in the context of AZA/VEN are ASXL1 and RUNX1. Preclinical data from isogenic leukemic cells harboring ASXL1 mutations demonstrated that genetic correction of this variant slows leukemic cell growth and induces differentiation [75]. CD34 cells with ASXL1 mutations from patients were also shown to have higher BCL2 expression and gene-body methylation, rendering them more sensitive to VEN and AZA, respectively [75]. These findings appear clinically relevant as retrospective studies in patients with R/R AML [76] and MDS-EB2 [77] treated with HMA/VEN observed improved CR/CRi rates in patients with ASXL1 mutations. In the latter study, patients harboring ASXL1 mutations were found to have a better median OS (not reached) compared to those without (10.2 months). However, an independent study could not reproduce improved CR/CRi rates in patients with ASXL1 mutations treated with AZA/VEN compared to intensive chemotherapy [78]. Rather, they observed improved CR/CRi rates in patients with RUNX1 mutations when using AZA/VEN compared to intensive chemotherapy. A significant survival benefit was observed in older patients with RUNX1 mutations receiving AZA/VEN as well. Although retrospective in nature, these data suggest a preferential benefit of HMA/VEN in patients with RUNX1 and ASXL1 mutations and highlight the importance of mutational testing in treatment planning.

Targeted oral therapy combinations

The last 5 years has led to multiple targeted therapy approvals for patients with AML and mutations in IDH1 (ivosidenib [IVO], olutasidenib [OLU]), IDH2 (enasidenib [ENA]), and FLT3 (MIDO, gilteritinib [GILT]). We will briefly review the landmark studies regarding these agents and discuss the emerging roles for approved therapies (Table 2) as well as ongoing areas of investigation into their use.

Table 2 Published studies of approved and investigational uses of FDA-approved and/or NCCN-recommended targeted therapies in AML
Full size table

IDH1 and IDH2-mutant AML

IDH1 and IDH2 mutations are reported at a frequency of 7–14% and 8–19%, respectively, in AML [16, 79]. Mutations in these genes typically occur in the conserved arginine residues (IDH1R132, IDH2R140, and IDH2R172) of the catalytic domain of isocitrate dehydrogenase. The prognostic implications of IDH1/IDH2 mutations are not entirely clear. Some reports have suggested that IDH2 mutations are associated with better outcomes while IDH1 mutations confer worse outcomes [80, 81], though there is significant heterogeneity in the prognostic impact of co-mutations such as NPM1 [16, 81].

In a phase 1/2 study, patients with R/R AML harboring IDH2R140 or IDH2R172 mutations (mean age 67 years) were treated with ENA [82, 83]. Patients with IDH2R140 mutations had 2-HG reductions greater than 90% regardless of response, while 2-HG levels in those with IDH2R172 mutations correlated with response (82.0% reduction from baseline if CR, 44.3% reduction from baseline if non-CR response, and 38.4% reduction from baseline if no response). CR/CRi/CR with incomplete platelet recovery (CRp) was observed in 29% of patients. Median OS for all patients was 8.8 months but extended to 22.9 months in patients achieving CR. While reduction in mutant IDH2 VAF was not required for response, follow-up data demonstrated that clearance of IDH2-mutated clones was associated with 100% CR [83]. Furthermore, co-occurring mutations in NRAS or MAPK pathway were suggested to contribute to treatment resistance [84]. In a phase 3 randomized study from the BEAT AML Master trial, older patients with IDH2-mutant R/R AML were randomized to ENA or conventional care [85]; patients in the ENA arm had a doubling of event-free survival (EFS) and significant improvements in CR/CRi/CRp rates and hematologic response. The presence of DNMT3A co-mutations has been shown to be associated with CR, and no deleterious effects of RAS signaling pathway co-mutations were observed; however, the presence of ≥ 4 co-mutations decreased the overall response rates (ORR) significantly (27.3% compared to 47.1% with < 4 co-mutations) [86]. Thus, although ENA only has an FDA-approval label for R/R AML with IDH2 mutations, the NCCN guidelines provide a recommendation to consider frontline ENA use for patients older than 60 years who are not candidates for intensive remission induction [45].

In a phase 1 study, DiNardo et al. assessed the use of IVO in IDH1-mutant R/R AML [87]. The median age of patients was 67 years, and CR/CRp rates were 30.4% with a median DOR of 8.2 months. Median OS was 8.8 months with an 18-month OS rate of 50.1% in patients with CR/CRp. It was noted that patients with a lower co-mutational burden had improved CR/CRp rates, but no specific predictive co-mutations were identified. As a follow-up to this study, 34 patients with ND AML and IDH1 mutations (median age 76.5 years) received IVO in the frontline setting [88]. The composite CR (CRc) rate was 42.4% with over 60% of patients maintaining CRc at 1 year. Patients who received prior HMA therapy for antecedent hematologic disorder achieved CRc approximately half as frequently as those without prior HMA. Receptor tyrosine kinase (RTK) pathway mutations were observed in 36.8% of patients who did not achieve CRc compared to no patients who achieved CRc. IDH1 mutant clone clearance was reported in 64.3% of patients who achieved CRc and was not observed in any patients without CRc. These two studies ultimately led to the approval of IVO in both R/R AML with IDH1 mutations and ND AML with IDH1 mutations in patients who are ineligible for standard chemotherapy.

On December 1, 2022, the FDA approved another IDH1 inhibitor, OLU [89], for use in R/R AML based on a phase 1/2 trial of patients with IDH1-mutant R/R AML who were naïve to IDH1 inhibitors [90]. Patients with a median age of 71 years were treated with OLU until progression. Notably, CR/CRh rates were 35% and were achieved at a median of 1.9 months. Responses appeared durable with a median DOR of 25.9 months in patients who achieved CR/CRh. Prior VEN exposure did not appear to decrease response efficacy. Survival data will require further maturation. Another phase 1/2 study examining OLU with or without AZA in ND and R/R AML demonstrated similar CR/CRh rates in R/R AML patients treated with monotherapy while responses were surprisingly worse when combined with AZA [91]. While the role of OLU is not clear considering experience with using IVO, one compelling scenario could be in the setting of IVO resistance through IDH1 mutations, though this is solely based on preclinical data [92] with no current data on response rates after prior IVO exposure.

Recent studies have evaluated the use of IVO or ENA in combination with other frontline therapies in ND AML. A phase 1 study assessed the use of IVO or ENA in combination with 7 + 3 (or bioequivalent dose of idarubicin) in patients with IDH1/IDH2 mutations [93]. Patients received IVO or ENA throughout induction, consolidation, and maintenance, though IVO or ENA were discontinued in patients who underwent alloHSCT. CR/CRi/CRp rates were 72% for IVO and 63% for ENA at the end of induction, which is slightly better compared to historical controls with IDH1/2 mutations [94]. In an updated analysis [95], the authors reported CR/CRi/CRp rates of 78.3% in the IVO subgroup and 73.6% in the ENA subgroup. Responses for the sAML subgroup were improved if there was no prior HMA exposure, consistent with previously reported data [87]. Co-mutations did not impact response rates in the IVO cohort, but in the ENA cohort, co-mutations with ASXL1, NRAS, U2AF1, and TP53 were associated with worse response rates, while DNMT3A co-mutations were associated with marginally improved CR/CRi/CRp. When assayed by digital polymerase chain reaction (dPCR), 39% of patients treated with IVO cleared IDH1-mutant clones and 23% of patients treated with ENA cleared IDH2-mutant clones. Approximately half of the patients receiving IVO or ENA ultimately proceeded to alloHSCT. The use of IVO or ENA in combination with induction therapy was tolerable and ultimately did not significantly impact the time to recovery of the absolute neutrophil count or platelet count. An unanswered question that will require further investigation is the role of maintenance IVO or ENA after alloHSCT, particularly in patients who are unable to clear their mutant clone prior to transplantation. In patients eligible for intensive chemotherapy, the improved response rates compared to historical controls with the addition of IVO or ENA [93, 95] provides a compelling argument for this practice. However, in the absence of randomized head-to-head comparison, this combination is neither FDA-approved nor recommended by the NCCN guidelines [45].

The majority of IDH1/2 mutations have been shown to be exquisitely sensitive to the combination of AZA/VEN with response rates similar to or higher than those achieved with standard induction. Therefore, it is reasonable to consider low-intensity therapy in this patient population without compromising outcomes. In a pooled analysis of from VIALE-A [48] and a phase 1b HMA/VEN study [70], the patients with IDH1/2 mutations achieved CR/CRi rates of 79% with AZA/VEN compared with 11% with AZA alone [96]. Median DOR was 29.5 months and 9.5 months, respectively, and median OS was 24.5 months and 6.2 months, respectively. CR/CRi and OS were relatively better with IDH2 mutations compared to IDH1.

The combination of AZA with IVO has recently emerged as an alternative non-intensive treatment option [97]. The AGILE study was a phase 3 trial randomizing patients with ND IDH1-mutant AML ineligible for intensive chemotherapy to receive AZA/IVO or AZA/placebo [98]. At a median follow-up of 12.4 months, EFS was significantly longer in the AZA/IVO group compared to AZA/placebo with an estimated 12-month EFS of 37% and 12%, respectively. As a secondary endpoint, the median OS was 24 months for AZA/IVO and 7.9 months for AZA/placebo. CR/CRp rates were 53% with AZA/IVO compared to 18% with AZA/placebo, and DOR was longer for AZA/IVO compared to AZA/placebo (22.1 months versus 9.2 months). Patients with RTK pathway mutations (FLT3, KIT, NRAS, KRAS, PTPN11) and TP53 mutations were more likely to respond to AZA/IVO, and follow-up data suggest that relapse appears to preferentially occur with the acquisition of secondary high-risk mutations, independent of IDH1 [99]. The findings of this study led to the recent FDA approval of AZA/IVO for the frontline treatment of patients with ND IDH1-mutant AML. While the rate of differentiation syndrome in these patients approaches 20%, the rate of cytopenias compared to AZA/VEN is significantly less. Therefore, when thinking about the various options for patients with IDH1 mutations, toxicity, quality of life, and sequencing of treatment should be considered.

The combination of AZA/ENA has also been studied in patients with R/R [100] and ND AML with IDH2 mutations ineligible for intensive chemotherapy [100, 101]. In the phase 2 analysis of patients with ND AML, patients were randomized to AZA/ENA or AZA monotherapy [101]. The median age of patients was 75 years with CR/CRi/CRp rates of 63% with the AZA/ENA group compared to 30% with AZA alone; similar ORR were seen regardless of R140 or R172 mutations. Both ORR and CR were more durable with AZA/ENA compared to AZA (24.1 months versus 9.9 months and not reached versus 12.7 months, respectively). In a 2-year post hoc analysis, median EFS with AZA/ENA was 15.7 months compared to 11.9 months with AZA alone and OS was 22 months with AZA/ENA compared to 18.6 months with AZA alone; while the survival differences were not statistically significant, this study was not powered to detect significant differences in survival outcomes.

A major question remains as to whether therapies such as AZA/IVO or AZA/ENA would outperform AZA/VEN for ND IDH1/2-mutant AML. The lack of survival advantage with AZA/ENA compared to AZA monotherapy [101] would suggest that AZA/VEN is superior in IDH2 mutations with the caveat that the median OS of patients with AZA in this study was significantly longer than reported for patients with IDH2 mutations in the AZA/VEN studies [96]. For patients with IDH1 mutations, CR rates appear to be better with AZA/VEN [96] than with AZA/IVO [97], though median OS was essentially the same. At present, the widespread availability of AZA/VEN and its ability to bridge to alloHSCT favors its use in the frontline setting for patients with IDH1/2 mutations who are ineligible for intensive chemotherapy, thereby preserving IVO, OLU, or ENA in the case of R/R disease. However, AZA/IVO could be considered in patients who are at high risk of complications with the myelosuppression of AZA/VEN. While ENA and IVO monotherapy are both NCCN-recommended options for frontline therapy [45], only IVO is approved in this setting, and these should only be considered in patients with a very poor performance status. An area for future study will be triplet therapies. In an exploratory study of patients with treatment-naïve and R/R AML with IDH1 mutations [102], patients who received AZA/VEN/IVO had a CRc rate of 85–100% depending on dose intensity; however, this compares similarly to CRc rates of 67–100% with IVO/VEN alone. Nevertheless, the use of a triplet regimen improved MRD to 86% from 25% with doublet therapy. Studies combining different permutations of ENA, OLU, or IVO with HMA and/or VEN are currently underway (NCT04092179, NCT03471260, NCT04774393, NCT02719574) as are studies of these agents in the maintenance setting (NCT05010772, NCT03728335, NCT03564821, NCT03515512, NCT04522895).

FLT3-mutant AML

FLT3 encodes a type 3 RTK (FMS-like tyrosine kinase 3) and is widely expressed on AML blasts [103]. Mutations in FLT3 are seen in about 25–32% of cases of ND AML with 25% harboring ITDs and 7–10% harboring TKD mutations [104, 105]. Previous reports prior to the era of FLT3-targeting TKIs (henceforth referred to as FLT3i) have suggested that AR of FLT3-ITD and the presence of NPM1 co-mutations variably affect outcomes [106,107,108]. Nevertheless, recent data suggest that relapse risk is higher in patients with FLT3-ITD AML irrespective of AR or presence of NPM1 mutation, and these patients should be considered for alloHSCT in first remission (CR1) if eligible [109, 110]; this is reflected in the updated ELN recommendations [24]. FLT3i can be divided into Type I and Type II inhibitors [111], which are active against both ITD and TKD mutations or ITD only, respectively. The first approved FLT3i in AML was the Type I staurosporine-derived inhibitor MIDO [112, 113] with early reports of its synergy with chemotherapy in patients with ND FLT3-mutant AML [114]. Newer and more selective FLT3i, GILT (Type I) and quizartinib (QUIZ) (Type II), have demonstrated promising responses in patients with R/R AML [115,116,117].

The RATIFY trial [118] was a randomized, placebo-controlled phase 3 trial investigating the addition of MIDO to standard induction chemotherapy and high-dose cytarabine (HiDAC) consolidation in adult patients under the age of 60 with ND AML and FLT3 mutations (TKD or ITD). CR rates were similar between both groups, though median EFS and OS were significantly improved with MIDO (8.2 months versus 3 months and 74.7 months versus 25.6 months, respectively). OS was durable with 51.4% of patients in the MIDO group surviving at 4 years. More patients underwent alloHSCT in CR1 with MIDO compared to placebo (28% versus 23%, respectively); notably, follow-up studies after the publication of the RATIFY trial demonstrated deeper molecular remission with the addition of FLT3i to induction therapy [119, 120], perhaps explaining in part the durable differences in OS despite similar CR and EFS rates.

Recent findings from a phase 2 study have also established the efficacy of adding to MIDO to induction, HiDAC consolidation, and maintenance in patients up to the age of 70 with ND FLT3-mutant AML [121, 122]. It should be noted that at present, MIDO is not currently approved as monotherapy and therefore is not recommended for post-consolidation maintenance given minimal benefit demonstrated after alloHSCT [36, 118, 121,122,123]. Preliminary data from a phase 1 study (NCT02236013) evaluating GILT in combination with 7 + 3, consolidation, and maintenance in ND AML [124] noted a median OS of 35.8 months with CRc achieved by 81.8% of all patients. AlloHSCT was performed in 30.4% of all patients. These data have led to ongoing clinical trials of GILT versus MIDO in addition to induction chemotherapy and consolidation (NCT04027309, NCT03836209).

GILT is the only FDA-approved FLT3i for use in R/R AML with FLT3 mutations. The ADMIRAL study was a phase 3 randomized control trial of patients with R/R AML and FLT3-mutations who received GILT or salvage chemotherapy [117]. Similar rates of prior FLT3i exposure were noted in both arms, and approximately 20% of patients in either group had previously undergone alloHSCT. Median OS for patients receiving GILT was 9.3 months versus 5.6 months for those receiving salvage chemotherapy. [125]. CRc rates were 54.3% with GILT and 24.8% with chemotherapy, and the median DOR was 11 months in the GILT group. Median OS for the FLT3-ITD and FLT3-TKD groups that received GILT were 9.3 months and 8 months, respectively. An important aspect of this study was the efficacy in both FLT3-ITD and FLT3-TKD populations, as the latter has been demonstrated to confer secondary resistance to type II FLT3i [126].

Given the increasing use of low-intensity regimens in AML, pooled data from VIALE-A [48] and a phase 1b HMA/VEN study [70] showed that patients with FLT3-ITD had a CR/CRi rate of 63% with AZA/VEN and a median OS of 9.9 months, while those with FLT3-TKD had a CR/CRi rate of 77% and a median OS of 19.2 months [127]. Of patients with FLT3 mutations, approximately 36% had NPM1 mutations in each of the AZA/VEN and AZA groups. Of those with concurrent FLT3-ITD and NPM1 mutation, AZA/VEN conferred a CR/CRi rate of 70% and a median OS of 9.1 months; patients with FLT3-ITD and wild-type NPM1 had a median OS of 10.6 months. This study has two important takeaways for older patients with mutated FLT3. First, the rate of FLT3 mutations in this population was lower, and the patients were older than typical FLT3-driven AML seen in younger patients, possibly suggesting different disease kinetics and biology. Second, while CR/CRi rates were worse for patients with FLT3-ITD mutations with wild-type NPM1 compared to those with mutated NPM1, the overall survival did not differ significantly, suggesting that NPM1 status has an unclear prognostic value for patients treated with HMA/VEN.

Trials combining GILT [128] or MIDO [129] with HMA have not yielded encouraging results to date, though early data suggest that AZA/sorafenib (SORA) may be effective in patients with R/R AML and FLT3-ITD [130]. Data demonstrating the efficacy and tolerability of DAC/SORA in patients with R/R AML with FLT3-ITD [131] have led to NCCN recommendations for the use of AZA/SORA or DAC/SORA as low-intensity therapy in elderly patients with FLT3-ITD AML or in R/R AML with FLT3-ITD [45], though it does not carry FDA approval for these indications. Despite the modest benefit with HMA, there seems to be synergy between FLT3i and VEN [132,133,134]. A phase 1b study for VEN/GILT enrolled patients with FLT3-wild-type or FLT3-mutant (dose escalation) and FLT3-mutant (dose expansion) R/R AML [135]. The median age of patients was 63 years, 31% of whom had received prior alloHSCT and 16% of whom received prior VEN. No patients had previously received GILT, though 64% of patients with FLT3 mutations had received other prior FLT3i. Patients with FLT3-ITD had CR/CRi/CRp rates of 43%, while those with FLT3-TKD has rates of 33%, and response rates were slightly better in those who were FLT3i-naïve. Median OS was 10 months for all FLT3-mutated patients, though there was a significant improvement in those who had undergone alloHSCT after VEN/GILT (not reached) compared to those who did not receive alloHSCT (6.3 months).

Furthermore, a phase 2 trial evaluated the use of triplet therapy (DAC/VEN/FLT3i) in older patients with ND FLT3-mutant AML and all adult patients with R/R FLT3-mutant AML. In ND AML, the CRc rate was 92% with high rates of 91% MRD negativity in responders by PCR. In patients with R/R AML, CRc rates were 63% with MRD negativity by PCR in all patients who responded. At a median follow-up of 14.5 months, the median OS was not reached in ND patients (2-year OS estimated at 80%); the median OS in R/R patients was 6.8 months. Approximately one-third of patients underwent alloHSCT in either group. These results compare favorably to other reports of FLT3i/HMA in the ND setting [128, 131, 136], though CRc rates appear to be higher with VEN/GILT in patients with R/R AML [135].

Although not approved, two other FLT3i deserve mention given recent reports of their efficacy in AML. QUIZ is a second-generation type I FLT3i that can achieve significant marrow remissions in R/R FLT3-mutant AML [116, 137,138,139], though survival advantage was minimal compared to salvage chemotherapy in the phase 3 QuANTUM-R study [138]. Due to these underwhelming results and concerns about cardiotoxicity and increased myelosuppression compared to other FLT3i, QUIZ has not been approved in the USA or Europe, though it is approved for use in Japan. In the frontline setting, the phase 3 QuANTUM-FIRST (NCT02668653) trial [140] enrolled patients up to age 75 with ND AML and FLT3-ITD and randomized them to QUIZ or placebo in addition to induction therapy with 7 + 3. Patients who achieved CR/CRi received up to 4 cycles of HiDAC with QUIZ or placebo and/or alloHSCT followed by up to 3 years of maintenance therapy with QUIZ or placebo. CR/CRi rates were 71.6% and 64.9% in the QUIZ and placebo arms, respectively, with DOR of 38.6 months and 12.4 months, respectively. Median OS and RFS were 31.9 months versus 15.1 months and 39.3 months versus 13.6 months in the QUIZ and placebo arms, respectively. AlloHSCT was performed in CR1 at similar rates between both arms; when censored for alloHSCT, OS trended toward a benefit with QUIZ over placebo. Moreover, an updated report from the study found that QUIZ conferred a deeper molecular remission compared to the placebo arm, perhaps underscoring the durability of benefit [141]. Although RATIFY [118] had already demonstrated a benefit to the addition of MIDO to induction chemotherapy, the QuANTUM-FIRST study is unique in that it evaluates the addition of an FLT3i for the higher-risk FLT3-ITD mutation.

Lastly, a type I FLT3i emerging in clinical discussion is crenolanib (CREN). Long-term data were recently reported regarding the use of CREN in combination with 7 + 3 in adult patients with FLT-mutant ND AML [142]. CREN maintenance was offered up to 1 year after HiDAC or alloHSCT. The median age of patients enrolled was 57 years, 34% of which were over the age of 60 years. FLT3 mutations were 75% ITD, 18% TKD, and 7% both ITD and TKD. CR/CRi rates above 80% were reported across several subgroups including those with FLT3-ITD mutations or concomitant FLT3/DNMT3A/NPM1 mutations. MRD-negative CR/CRi was achieved in 94% of evaluable patients, and 50% of patients underwent alloHSCT. Median OS has not been reached at a median follow-up of 45 months. Furthermore, translational studies found that no FLT3 mutant clones were found at relapse in patients who completed protocol therapy.

In considering the role of FLT3i in ND FLT3-mutant AML, intensive induction chemotherapy plus MIDO remains a standard of care for eligible patients. However, the formal release of data from QuANTUM-FIRST is awaiting, and ongoing trials will assess other frontline combinations with QUIZ (NCT04209725, NCT04047641), CREN (NCT03258931), and GILT (NCT04027309, NCT03836209), including head-to-head comparisons against MIDO. If QUIZ is approved for ND AML, it should be emphasized that its use would be limited to patients with FLT3-ITD, while those with TKD mutations should still receive MIDO. For patients who are ineligible for intensive chemotherapy, AZA/VEN is effective for those with FLT3-TKD mutations; unfortunately, better frontline options for those with FLT3-ITD are currently limited. Nonetheless, GILT remains a very active FLT3i in the relapsed setting, and early data from doublet and triplet FLT3i combinations are encouraging in ND and R/R AML [143]. Several trials exploring triplet combinations with DAC/VEN/QUIZ (NCT03661307) and AZA/VEN/GILT (NCT04140487) are currently enrolling with results highly anticipated.

Maintenance therapy and consideration of treatment-free remissions

While the historical focus of induction therapy in AML is to achieve remission and proceed with transplantation in eligible patients, the emergence of tolerable oral therapies posits the role of maintenance therapy, particularly in patients with a higher risk of relapse, such as those with pre-transplantation MRD positivity. There are limited data to support the use of targeted therapies as maintenance after transplant except for SORA. Although SORA has limited efficacy in the frontline setting [144, 145], the SORMAIN trial evaluated SORA maintenance in patients with FLT3-ITD after alloHSCT [146]. The investigators noted a 25–30% absolute improvement in 2-year RFS and OS compared to the placebo. Patients treated with SORA had higher rates of GVHD and skin toxicity, consistent with previous reports about its immunogenicity [147]. These results have since been corroborated by another phase 3 study of SORA maintenance post-alloHSCT in CR1 for patients with FLT3-ITD [148]; thus, while SORA does not have an FDA label indication for use in the treatment of ND or R/R FLT3-mutant AML, it is recommended for use in patients with FLT3-ITD who achieve remission after alloHSCT [45]. While no other agents are currently recommended for post-alloHSCT maintenance, the MORPHO trial (NCT02997202) is evaluating the efficacy of GILT versus placebo in this setting for patients with AML and FLT3 mutations and has completed enrollment.

For patients in remission after induction therapy but unfit for transplant, maintenance options have been limited in the absence of FLT3-ITD. The QUAZAR AML-001 study [149] randomized patients who had achieved CR/CRi after intensive chemotherapy but were not fit for alloHSCT to receive oral azacitidine (CC-486) maintenance or placebo. Median OS after randomization was longer in patients receiving CC-486 compared to placebo (24.7 months versus 14.8 months), though it should be noted that only 14% of patients had adverse-risk cytogenetics. Indeed, studies have demonstrated disease-free survival (DFS) but no OS benefit with subcutaneous AZA maintenance in older patients in CR/CRi after induction [150] or with CC-486 in patients with adverse-risk cytogenetics after alloHSCT [151]. Combining VEN with AZA may improve these outcomes based on preliminary data from a phase 2 study (NCT04062266), particularly in those who received VEN-based induction [152]. Furthermore, the VIALE-M study (NCT04102020) is also investigating the role of CC-486 in combination with VEN for patients with CR1 after induction, and the VIALE-T study (NCT04161885) is assessing AZA/VEN maintenance after alloHSCT.

With several options for low-intensity maintenance therapies, the question remains as to whether treatment needs to be indefinite. A retrospective study reported their experience with discontinuing HMA/VEN or LDAC/VEN in transplant-ineligible patients older than 65 years who achieved CRc after receiving either combination for at least 12 months in the frontline setting [153]. Patients who stopped therapy experienced a median treatment-free remission (TFR) of 45.8 months with over half still in remission at end of data collection. No significant differences in RFS or OS were noted between cohorts. One caveat is that patients were highly selected; the vast majority of patients achieved true CR with MRD negativity at the time of discontinuation. Moreover, of patients who sustained treatment-free remission, 86% had a prior NPM1 or IDH2 mutation with MRD negativity at cessation. Consequently, in this specific population, MRD negativity may be a reasonable impetus to interrupt treatment, though prospective studies formally evaluating this question are warranted.

Investigational agents and future directions

The aforementioned studies emphasize an increasingly nuanced approach to treatment decision-making in AML as molecular data such as types of mutations (i.e., FLT3-ITD versus FLT3-TKD) and co-mutational patterns (i.e., IDH1/2 with DNMT3A or RTK pathway mutations) have important prognostic and treatment implications. Although TP53-mutant myeloid neoplasms remain one of the largest unmet needs in care, it is encouraging that detailed mechanistic studies may open the door to the development of further targeted therapies. Moreover, with the adoption of genomic data into routine care for patients with AML [18, 154], larger population-based studies will hopefully improve the personalization of treatments. Below we briefly highlight investigational agents (Fig. 4, Table 3), with a focus on those in later clinical development.

Fig. 4
figure 4

Mechanisms of novel targeted therapies in AML in later-stage clinical development. A Mutations in TP53 lead to altered conformation of p53 leading to a subtype of AML characterized by treatment resistance, high relapse rates, and poor overall survival. Novel agents such as eprenetapopt are metabolized into MQ which covalently modifies the mutant p53 protein leading to a wild-type-like conformational change and restoration of normal p53 activity [155]. Recent reports have demonstrated that MQ can drive p53-independent cell death through ROS accumulation [156, 157] and ferroptosis [158]. B Leukemic cells can evade immune surveillance by upregulation of CD47, which binds SIRPa; this emits a “don’t eat me signal” to macrophages [163,164,165]. Antibodies targeting CD47 can block this inhibitory signal and allow for phagocytosis of leukemic cells [165,166,167]. C Certain types of AML are characterized by mutated NPM1c or oncogenic fusion partners associated with MLL [173, 174, 179]. These lead to complex formation with menin and LEDGF, ultimately resulting in transcriptional activation of leukemia stem cell promoting genes [175,176,177]. Blocking this pathway with menin inhibitors such as KO-539 or SNDX-5613 can repress this transcriptional program allowing for differentiation of granulocytes [178, 180, 181]. AML Acute myeloid leukemia, MQ Methylene quinuclidinone, ROS Reactive oxidative species, SIRPa Signal regulatory protein alpha, NPM1c Cytoplasmic NPM1 (mutant NPM1), MLL Histone lysine N-methyltransferase 2A (KMT2A), LEDGF Lens epithelium-derived growth factor, MI Menin inhibitor

Full size image
Table 3 Investigational agents in clinical development for AML and/or MDS
Full size table

P53 reactivation

Eprenetapopt (APR-246) is a small-molecular inhibitor originally thought to work through covalent modification of mutated p53, which restores wild-type-like p53 conformation, thus functionally reactivating it [155]. However, its mechanism may also target other synergistic pathways which can drive p53-independent cell death including modulation of cellular redox [156, 157] and increasing glutathione turnover, leading to ferroptosis [158]. Two phase 2 studies are evaluating the use of this agent in combination with AZA in TP53-mutant MDS and AML [159, 160]. In these studies, CR rates of patients with MDS were 47–50% with durable responses. In patients with AML, however, CR was only 17%. Notably, responding patients in both studies had significant reductions in TP53 VAF. Median OS was 10.8–12.1 months for MDS patients and 13.9 months for AML patients. Notable toxicities were febrile neutropenia and neurologic toxicity. Recent data from a phase 2 study (NCT03931291) was presented, which assessed eprenetapopt/AZA as post-alloHSCT maintenance in patients with TP53-mutant MDS and AML [161]. Patients received a median of 7 cycles of treatment with a median RFS of 12.5 months and a median OS of 20.6 months. No 30-day mortalities were noted from the first dose. These results, while modest, are encouraging in this very high-risk population. Preclinical studies suggest that XPO1 upregulation may contribute to eprenetapopt resistance in AML and can be overcome with agents like selinexor, though this will require in vivo validation [162]. An additional trial combining eprenetapopt with AZA/VEN has completed enrollment and awaiting data release (NCT04214860). A phase 1 study of APR-548, which is a next-generation molecule, in combination with AZA (NCT04638309) had opened but was terminated by the sponsor.

CD47—targeting the “don’t eat me” signal

CD47 is a heavily glycosylated cell surface protein and is expressed by virtually all cells in the body, including those that do not express integrin, such as red blood cells [163]. It provides an anti-phagocytic signal in healthy cells but was discovered as an adverse prognostic factor in AML as it is overexpressed on leukemic stem cells compared to non-leukemic stem cells; preclinical murine models demonstrated that blockade of CD47 with monoclonal antibodies could enable phagocytosis of leukemic stem cells and prevent in vivo engraftment [164, 165]. These findings led to the development of a humanized anti-CD47 antibody known as Hu5F9-G4 or magrolimab [166]. A phase 1b study explored the tolerability of magrolimab in combination with AZA (NCT03248479) for patients with untreated intermediate- to very high-risk MDS and with untreated AML unfit for intensive chemotherapy [167]. TP53 mutations were noted in 27% of patients. Common adverse events included anemia, neutropenia, thrombocytopenia, and infusion reactions. In transfusion-dependent MDS and AML patients, 58% and 64% were able to achieve transfusion independence; moreover, CR/CRi rate was 56% in AML patients. The median duration of response was not reached in MDS, AML, or TP53-mutant AML subpopulations. Preliminary data from a phase 1b/2 study evaluating the combination of magrolimab/AZA/VEN (NCT04435691) noted CR/CRi rates of 63% and 86% in patients with ND AML with or without TP53 mutations, respectively, conferring 1-year OS rates of 53% and 83%, respectively [168]. In patients with R/R AML, median OS was only 7.4 months with responses especially limited in those with prior VEN exposure. Given these findings, particularly in the very high-risk TP53-mutant group, three phase 3 trials opened which are comparing magrolimab/AZA versus AZA in patients with untreated intermediate- to very high-risk MDS (NCT04313881), magrolimab/AZA versus AZA/VEN or intensive chemotherapy in patients with untreated TP53-mutant AML (NCT04778397), and magrolimab/AZA/VEN versus AZA/VEN in patients with untreated AML who are ineligible for standard intensive chemotherapy (NCT05079230). A phase 2 study of magrolimab with various anti-leukemic therapies in patients with untreated AML is enrolling as well (NCT04778410). An additional combination study of another anti-CD47 antibody, lemzoparlimab, had opened for patients with a higher-risk MDS or AML ineligible for intensive chemotherapy (NCT04912063) but was recently stopped.

Menin inhibition

Several types of AML have overexpression of HOXA/B cluster genes and MEIS1, which are critical regulators of hematopoietic stem cell self-renewal and differentiation [169,170,171]. These can be dysregulated with alterations of histone lysine N-methyltransferase 2A (KMT2A or MLL) and/or NPM1 mutations [172,173,174]. KMT2A binds menin as part of a histone methyltransferase complex, and when it is involved in an oncogenic fusion, it may lead to aberrant transactivation of leukemia-promoting genes [175,176,177]. While mechanistically unclear, cytoplasmic localization of the mutant NPM1 is associated with a similar phenotype and genetic signature as KMT2A-driven AML [173, 178, 179]. Consequently, several preclinical models examined the role of menin inhibition in these subtypes of AML and observed efficacy [178, 180, 181]. This has led to the opening of multiple phase 1/2 studies in patients with R/R AML with KMT2A-rearrangement or NPM1-mutation using the menin-MLL inhibitor KO-539 (NCT04067336) or menin inhibitor SNDX-5613 (NCT04065399, NCT05326516, NCT05406817, NCT05360160, NCT03013998) which are currently enrolling patients or opening soon. Early data from NCT04067336 and NCT04065399 suggest similar CR/CRh rates (25–30%) and MRD negativity rates in responding patients (75–78%) with either KO-539 or SNDX-5613, respectively [182,183,184]. Toxicities include QTc prolongation with SNDX-5613 and differentiation syndrome, particularly with KO-539. Of note, however, the activity of KO-539 appears largely restricted to patients with NPM1-mutated AML, in whom differentiation syndrome was not observed.

Other therapies in clinical development

In addition to the aforementioned therapies, several other pathways have been identified as possible therapeutic vulnerabilities in AML and are emerging in early clinical development [185, 186]. Some of these agents target epigenetic and oncogenic signaling pathways and may allow for the augmentation of available therapies. The ALICE study is evaluating the use of iadademstat, a lysine-specific demethylase 1 inhibitor, in combination with AZA for the frontline treatment of AML in patients unfit for intensive chemotherapy. Preliminary data suggest CR/CRi rates of 64% with median OS extending to 14.3 months in responding patients. Strikingly, 75% of patients with TP53 alterations responded [187]. There has also been a considerable interest in exploiting advances in immunotherapy and cellular therapy for the treatment of AML. Unfortunately, findings from early studies on these agents have been underwhelming thus far. Flotetuzumab is a dual affinity retargeting antibody that targets CD123 and CD3. A phase 1/2 study reported flotetuzumab could achieve CR/CRh/CRi rates of 30% with 12-month OS of 75% in patients with R/R AML [188] with recent data suggesting similar outcomes in pediatric and adolescent/young adult patients [189]. The majority of patients who benefit from flotetuzumab are primary refractory and TP53-mutated, which is related to a distinct immune microenvironment compared to non-TP53-mutated AML. An ongoing phase 1b/2 study is also evaluating the combination of pivekimab sunirine (IMGN632), a CD123-targeting antibody–drug conjugate, with AZA/VEN [190]. Patients with R/R AML treated with this were reported to have CRc rates of 31%, including 26% in those with ELN adverse disease and 64% of those with FLT3-ITD, with minimal additive myelosuppression beyond that of AZA/VEN. In regard to checkpoint inhibition, findings from the unpublished REMAIN trial did not demonstrate a PFS or OS benefit with the use of nivolumab maintenance in patients with CR/CRi who are ineligible for alloHSCT [191]. Nevertheless, immune dysregulation is important in AML, though it is not yet clear how to effectively target the tumor microenvironment or which factors can be used to predict response to immune-based therapies.

Conclusion

AML is a very complex and heterogeneous disease as evidenced by the expansion of genetic and cytogenetic qualifiers in the updated WHO [23] and ICC [24] classification systems. While outcomes for AML continually improve by decade, a refined understanding of patient and tumor characteristics is needed to continue this upward trend. This is especially true when selecting patients for intensive chemotherapy and/or alloHSCT, as a standardized and validated metric of physiologic age or fitness would greatly improve our ability to personalize treatments and design clinical trials more representative of the actual patient population. However, we also now have several active, low-intensity therapies approved or in the developmental pipeline. Perhaps, the most transformative has been the combination of AZA/VEN, which has supplanted conventional cytotoxic chemotherapy in many cases. Unfortunately, outcomes in patients who progress on HMA/VEN are poor, particularly in patients who harbor TP53 or RAS-pathway mutations [192], and clinical development of additional therapeutic options is critical. Moreover, several groups will soon be reporting data from trials investigating the addition of VEN to induction chemotherapy in both the ND and R/R AML setting to answer the question if there is any benefit to further intensification of therapy [193,194,195,196].

We are entering a unique era of precision oncology whereby molecularly informed data can be exploited to tailor treatments based on disease pathobiology. But while options have rapidly increased for patients IDH1/2 and FLT3 mutations, progress has unfortunately been slow for those with the highest risk forms of AML, such as TP53-mutant disease. Nevertheless, the incorporation of data such as co-mutational burden and MRD analysis will hopefully allow us to better define patients at the highest risk of relapse [197,198,199] and who may benefit from early relapse intervention [200,201,202]. Pertinent questions in this regard will be whether induction of deeper molecular remissions with combination therapies would improve outcomes compared to sequencing therapies, how we can minimize toxicity associated with combinations of novel agents, and if we can use MRD negativity to interrupt treatment.

The multitude of active studies addressing these questions will lend nuance to clinical practice and contribute to improved outcomes for patients.