Core-binding factor acute myeloid leukemia: Heterogeneity, monitoring, and therapy


  • Melhem Solh,

    Corresponding author
    1. Department of Medicine, Florida Center for Cellular Therapy, University of Central Florida, Orlando, Florida
    2. Department of Medicine, University of Central Florida, Orlando, Florida
    • Correspondence to: Melhem Solh, University of Central Florida, Florida Center for Cellular Therapy, 2501 N Orange Ave., Suite 581, Orlando, FL 32804. E-mail:

    Search for more papers by this author
  • Sophia Yohe,

    1. Department of Pathology and Laboratory Medicine, University of Minnesota, Minneapolis, Minnesota
    Search for more papers by this author
  • Daniel Weisdorf,

    1. Department of Medicine, Division of Hematology, Oncology and Transplantation, University of Minnesota, Minneapolis, Minnesota
    Search for more papers by this author
  • Celalettin Ustun

    1. Department of Medicine, Division of Hematology, Oncology and Transplantation, University of Minnesota, Minneapolis, Minnesota
    Search for more papers by this author

  • Conflict of interest: No


Core binding factor acute myelogenous leukemia (CBF AML) constitutes 15% of adult AML and carries an overall good prognosis. CBF AML encodes two recurrent cytogentic abnormalities referred to as t(8;21) and inv (16). The two CBF AML entities are usually grouped together but there is a considerable clinical, pathologic and molecular heterogeneity within this group of diseases. Recent and ongoing studies are addressing the molecular heterogeneity, minimal residual disease and targeted therapies to improve the outcome of CBF AML. In this article, we present a comprehensive review about CBF AML with emphasis on molecular heterogeneity and new therapeutic options. Am. J. Hematol. 89:1121–1131, 2014. © 2014 Wiley Periodicals, Inc.


Most (∼55%) of adult acute myelogenous leukemia (AML) patients harbor nonrandom clonal chromosome aberrations including, but not limited to inversions, deletions, insertions, trisomies, monosomies and reciprocal cytogenetic translocations [1]. Karyotype at the time of diagnosis is one of the important predictors of outcome in AML with a significant better prognosis observed with specific treatments targeting certain cytogenetic groups [2-5]. The chromosome defects t(8;21), inv(16)(p13q22), and t(16;16)(p13;q22) involving the core-binding factors are found in 15% of adult patients with AML and carry an overall favorable prognosis [6].

The core-binding factors (CBF) are a group of DNA-binding transcription factor complexes composed of α and β subunits. The α subunit is the DNA binding element of the complex and is encoded by three mammalian genes. Only one β subunit exists in mammals CBF-β(PEBP2-β) and it functions to stabilize the DNA binding without direct DNA contact [7, 8]. Using standard therapy, adult CBF AML has a higher rate of complete remission (CR), prolonged CR duration, and a better prognosis than patients with AML with normal karyotype or other chromosomal aberrations [9-13]. CBF AMLs have a CR of 87 to 88% with an estimated relapse free survival (RFS) of 42% at 10 years of follow-up [14]. A recent study using the Surveillance, Epidemiology, and End Results (SEER) database showed the median overall survival (OS) of patients with CBF AML increased from 16 months in 2000–2002 to 25 months in 2006–2008 [15]. Despite this improvement in outcomes which is notably better than other subsets of adult AML, survival with CBF AML is still poor, particularly among African Americans and the elderly [15].

The World Health Organization (WHO) recognizes two cytogenetic subtypes of CBF AML within the category of AML with distinct, recurrent cytogenetic abnormalities: t(8;21)(q22;q22) and inv(16)(p13q22) or t(16:16)(p13;q22). These cytogenetic subtypes are commonly referred to as t(8;21) and inv(16) [16]. Due to similarities between the CBF AMLs, they are usually grouped together and managed similarly, however, there is considerable clinical, pathologic, and molecular cytogenetic heterogeneity amongst them [14].

We present a comprehensive review of CBF AML with emphasis on the histopathologic, genetic, molecular, and clinical heterogeneity. In addition, therapeutic options and the role of minimal residual disease (MRD) testing for CBF after therapy to predict relapse will be discussed.

Molecular Pathogenesis

The core binding factor complex

The core binding factors are heterodimeric transcriptional regulators that contain one β and one of three possible α subunits [17]. The α subunit is the DNA binding element and is encoded by one of three mammalian genes: RUNX1 (also referred to as AML1 and PEBP2A2), RUNX2 (also referred to as AML3 and PEBP2A1), and RUNX3 (also referred to as AML2 and PEBP2A3) [7, 8, 18]. RUNX2 and RUNX3 do not appear to be significantly involved in leukemia and will not be discussed further. The β subunit, encoded by CBFB (alternate name PEBP2B), stabilizes binding of the α subunit to DNA and also acts to protect RUNX1 from ubiquitin-proteasome-mediated degradation [19]. The CBF complex containing RUNX1 and CBFB regulates many hematopoietic genes involved in myeloid differentiation (Fig. 1).

Figure 1.

Leukomogenesis by core binding factor. (A) binding of α and β subunits leads to transcription of target genes. (B) Leukomogenesis by t(8;21) . (C) Leukomogenesis by inv (16). [Color figure can be viewed in the online issue, which is available at]

The RUNX1 protein can act as a transcription activator or a repressor and this dual function is mediated through interaction with lineage-specific transcription factors and with general coregulators [20]. It is thought that activation of RUNX1 is mediated through binding with the coactivators p300 and CREB-Binding protein (CBP) [21]. These proteins possess histone acetyl transferase (HAT) activity that results in transcription enhancement [22]. The high incidence of RUNX1 and its cofactor CBFB disruption in acute leukemia has suggested a critical role of these genes in normal hematopoiesis [23]. Studies with homologous gene inactivation showed that in RUNX1 or CBFB deficient mice, embryos died between days 12.5 and 13.5 with no evidence of erythropoiesis or myeloid cells in their circulation or fetal liver [24, 25]. These results, when combined with studies showing that RUNX1 or CBFB deficient embryonic stem cells could not contribute to adult hematopoiesis suggest that these genes are required for development of erythroid and myeloid cells [26, 27].

Mutations in RUNX1 have been consistently linked to leukemias including: point mutations in the autosomal dominant disease familial platelet disorder with propensity to AML (FPD/AML) [28]; the minimally differentiated subtype of AML (AML M0) [29, 30]; gene overexpression in pediatric ALL (frequency 1.5–2%) and AML [31-33]; and protein fusion in AML M2 t(8:21) [34-36].

Leukemogenesis by t(8;21)(q22;q22)

The chromosomal translocation t(8;21)(q22;q22) was first described in 1973 by Rowley [37]. Cloning of the t(8;21) provided RUNX1 probes to analyze breakpoint at 21q22 [38]. The translocation t(8;21)(q22;q22) fuses the RUNX1 to RUNX1T1 (formerly known as ETO) generating a novel chimeric gene RUNX1/RUNX1T1 that consists of the proximal DNA-binding region called Runt Homology Domain of RUNX1 and most of RUNX1T1 (ETO) missing only the first 30 amino acids [34]. The resulting fusion lacks the carboxyl terminal transactivation domain of RUNX1, which suggests that RUNX1/RUNX1T1 disrupts hematopoiesis through a dominant-negative mechanism [39] (Fig. 1B). RUNX1T1 (ETO) acts by associating with N-CoR (nuclear receptor corepressor) through recruiting histone deacetylase (HDACs) and sin3A [40]. The active RUNX1T1(ETO) represses transcription by deacetylation of the RUNX1 target genes leading to disruption of normal hematopoiesis and inactivation of tumor suppressor genes needed for neoplastic transformation [41].

The expression of RUNX1/RUNX1T1 alone is not a sufficient to cause leukemia in mice, as other genetic events are needed for leukemia development [42, 43]. Expression of RUNX1/RUNX1T1 reduced definitive hematopoiesis with a decrease in the number of stem cells and their committed progeny, while the viable stem cells showed enhanced self-renewal capacity leading to a preleukemic population [44]. Several in vivo studies suggest that RUNX1/RUNX1T1 increases the self-renewal capacity of hematopoietic precursors and slows their terminal differentiation [20, 45, 46]. RUNX1/RUNX1T1 was shown to trigger heterochromatin silencing of microRNA-193a by binding at RUNX1 binding sites and recruiting chromatin-remodeling enzymes. Suppressing miR-193a enhances the oncogenic activity of RUNX1/RUNX1T1 by directly enhancing expression of DNMT3a, HDAC3, KIT, CCND1 and MMDM2 and indirectly decreasing PTEN [47].

Leukemogenesis by inv(16)(p13q22) and t(16;16)(p13;q22)

The inv(16) or t(16;16)(p13;q22) occur in 5% of AML cases [2] and result in the fusion of the CBFB gene from chromosome 16q22 with the MYH11 gene from chromosome 16p13 forming two novel genes; CBFB/MYH11 and MYH11/CBFB. Only CBFB/MYH11 has leukemogenic potential [48, 49] resulting from the fusion of the majority of CBFB with the tail domain of MYH11 [50] (Fig. 1C). CBFB/MYH11 acts in a similar manner to RUNX1/RUNX1T1 by associating with co-repressor complexes leading to recruitment of HDAC activity and subsequent silencing of gene function [51]. The breakpoints in MYH11 gene incorporate a myosin long tail functional domain that mediates homodimerization of the CBFB/MYH11 protein into high molecular weight structures. These filamentous structures also incorporates RUNX1 upon co-transfection, preventing RUNX1 localizing in the nucleus [49]. These results suggest that CBFB/MYH11 disrupts hematopoiesis by sequestering the RUNX1 protein pool.

Knockout models showed that mice lacking CBFB had an early embryonic death. These results were reproduced by knock in models where mice heterozygous for CBFB/MYH11 or RUNX1/RUNX1T1 had the same phenotypic expression suggesting a common pathway for all these mutations [52]. KIT D816 mutations were shown to cooperate with CBFB/MYH11 in contributing to leukemogenesis in CBFB/MYH11 mice [53].

Gene mutations

CBF AML and KIT mutations

The KIT gene is located on chromosome 4q11-12 and encodes a 145-KD transmembrane glycoprotein (member of the type III receptor tyrosine kinase family) [54, 55]. KIT activates signaling pathways involved in proliferation, differentiation and survival [56]. Gain of function mutations leading to ligand independent receptor activation have been reported in CBF AML [57-61] and other malignancies such as gastrointestinal stromal tumors [62] and germ cell tumors [63].

Most of the KIT mutations in CBF AML are within exon 17 (KIT17) and exon 8 (KIT8) [57, 64, 65]. The incidence of the KIT gene mutations in patients with CBF AML varies between 6.6% up to 46.1% [60, 65, 66]. A recent United Kingdom Medical Research Council (UK MRC) study showed the incidence was 28% in 354 younger (<60 years) adults with t(8;21) (n = 199) or inv(16) (n = 155) [67]. Mutations were distributed equally between inv(16) and t(8;21) [57], but mutations involving exon 8 of the KIT was more common in patients with inv(16) compared to those who had t(8;21) [68, 69]. Gene expression profiling was analyzed among 83 CBF-AML patients with KIT mutations. These were characterized by dysregulation of genes belonging to the NFКB signaling complex indicating impairment of apoptosis; hence they may benefit from tyrosine kinase inhibitor (TKI) therapy [70].

Some CBF AML patients have associated systemic mastocytosis (SM) as a part of an entity systemic mastocytosis (SM) with an associated clonal hematological non-mast cell lineage disease (SM-AHNMD) [71]. The presence of SM component can be occult at diagnosis and uncovered only after induction therapy [72, 73]. SM-AHNMD was found in approximately 10% of CBF AML patients [72, 73]. Exon 17 KIT mutations (most commonly KIT D816V) are detected in >90% of patients with SM and constitute one of the diagnostic criteria of SM [74].

Most, but not all, studies show an adverse outcome in CBF AML with KIT mutations [67-69]. For example, mutations involving codon 816 of the tyrosine kinase domain TKD in t(8;21) AML [57, 69] and KIT17 in inv(16) have a higher 5 year relapse rate of 80% versus 26% in non-KIT mutations and t(8;21) alone [69]. The UK MRC study also showed similar CR rates but increased relapse rates [67]. Exon 17 mutations may have more adverse impact than the others [67, 69]. The presence of KIT mutations at relapse was evaluated in 31 patients [67]. Nearly half of the KIT mutant-positive cases either lost or changed their mutation at relapse. This is a clinically important finding because it indicates that, in at least some cases, KIT mutations are secondary and not responsible for relapse.

CBF AML and RAS mutations

NRAS or KRAS mutations are more common in patients with inv(16) than in t(8;21) (36% vs. 8%, P = 0.001); however may not have an effect on prognosis [68]. Kuchenbauer et al. found a low incidence of NRAS (8.9%), but did not detect KRAS mutation in 99 patients with RUNX1/RUNX1T1 (91 had de novo AML and 8 had therapy-related (t-AML)) [75]. The large UK MRC study found the incidence of RAS mutations was 27% [67].

CBF AML and FLT3 mutations

FLT3 mutations that involve that internal tandem duplication (ITD) within the juxtamembrane domain of the FLT3 gene are infrequent in CBF leukemia; detected in 2-9% of t(8;21) and 0 to 7% of inv(16) patients [60, 68, 76, 77]. FLT3 tyrosine kinase domain (TKD) point mutation (most commonly D835) are more frequent with inv(16) (6-24%) than t(8;21) (2–7%) [68, 78]. The negative impact of FLT3 mutations reported among AML patients with normal karyotype is not established in the CBF-AML group.

CBF AML and JAK2 mutations

JAK2 mutations were evaluated in 21 t(8;21) patients with de novo and 3 t-AML [79] and were positive in two patients (8%) who had received anthracyclines and topoisomerase inhibitors. In a larger study (n = 135) the incidence of JAK2 was 3.6% in patients with CBF AML, which was significantly higher when compared to other AML patients (0.6%) (n = 811), P < 0.01 [80]. The majority (80%) of patients with JAK2 positive AML relapsed. In another study, JAK2 mutations were identified in 4 of 64 (6%) t (8;21) and none of 99 inv(16) patients [81].

CBF AML and CBL mutations

UK MRC trial showed 6% of patients with CBF AML from various MRC AML trials had CBL mutations [67]. Patients with CBL-HIGH (>25% expression) had a better OS (CI95%: 0.05–0.85, P = 0.02).

Clinical and diagnostic features of CBF AML

Although inv(16) and t(8;21) are usually grouped together, recent studies showed differences in their clinical presentation as well as response to treatment [14]. Thus, we will review each of these separately, pointing out the similarities and differences.


CBF leukemia with t(8;21) are usually subclassified as French-American-British (FAB) M2 in 80-90% of the cases and M1 in about 10% [14, 82]. Patients with t(8;21) often present at a younger age (median 36 years) [14, 83, 84]. t(8;21) is more common in blacks and when compared to inv(16) usually have a lower WBC (10.6 vs. 17 × 109/L) at diagnosis [14, 85]. Splenomegaly and extramedullary involvement are seen in 20 to 25% of patients with t (8;21). Morphologically, myeloblasts in t(8;21) usually have an indented nuclei, a perinuclear hof, and basophilic cytoplasm that has a variable number of azurophilic granules and occasional large salmon colored granules. The blasts frequently contain a single Auer rod and Auer rods may also be seen in maturing neutrophils. Dysplasia in the maturing neutrophils is often present as well. Bone marrow eosinophilia is common, but they do not have the abnormal granulation that in seen in CBF leukemia with inv(16) [86]. The blasts have a myeloid immunophenotype but often also express B-cell markers, including CD19, PAX5, and CD79a [87]. TdT expression may be present but is usually weak and CD56 is present in a subset of cases.

Conventional cytogenetics is usually sufficient to detect t(8;21) and can also detect secondary abnormalities which occur in about 70% of these patients. Cryptic rearrangements of t(8;21) are uncommon. Patients with t(8;21) are more likely to harbor additional chromosomal abnormalities than inv(16), in particular del(9q) (present in 15%), LOS including either –Y (30% of cases) and –X (16%). As a presenting feature, LOS almost exclusively presents with t(8;21) [14, 83, 84]. Few patients may have a normal karyotype or contain only del(9q) or –Y with morphological and clinical features suggestive of t(8;21). Fluorescence in situ hybridization or RT-PCR can further demonstrate the cryptic RUNX1-RUNX1T1 rearrangement [88]. CBF AML with t(8;21) is also associated with a lower percentage of peripheral blasts and bone marrow blasts than inv(16) [14, 83, 84]. The blasts in KIT mutation positive t(8;21) show less frequent CD19 and CD56 expression on leukemic blasts [89] compared to KIT negative cases. This association with KIT mutations may explain the poor prognosis that has been reported with CD56 expression [14, 83, 84, 90-92]. Additional poor prognostic factors in t(8;21) include granulocytic sarcoma [93], complex cytogenetic abnormalities except for del(9q) [13], and specific gene expression profiling (GEP) signatures.

In the Bullinger et al. study, anti-apoptotic mechanisms and deregulated mTOR signaling were involved in better prognostic CBF (median OS not reached) while aberrant MAPK signaling and chemotherapy-resistance mechanisms were involved in poor prognostic CBF (median OS ∼1,500 days) [94]. Single nucleotide polymorphism genomic arrays (SNP-chip) analysis on 48 newly diagnosed AML patients with t(8;21) showed that 16 (33%) had one or more genomic abnormalities including copy number changes or copy number neutral loss of heterozygosity. OS and DFS were inferior in these 16 compared to the other 32 patients (OS: median 24 months vs. not reached; DFS: median 12 to 14 months vs. not reached) [95].

inv(16)(p13q22) and t(16;16)(p13;q22)

Most patients with inv(16) present as FAB M4 with dysplastic eosinophils in the bone marrow (M4Eos) [14, 83, 84]. Association with other FAB groups have been reported mainly AML-M4 with excess or abnormal marrow eosinophils (M4Eo), M2, and blast crisis of chronic myelogenous leukemia (CML) [96]. Inv(16) is reported in older ages (median 41 years), higher WBC and a higher peripheral and marrow blast percentage, lymphadenopathy, hepatosplenomegaly and more CNS involvement at relapse when compared to t(8;21) [14, 96]. Lymphadenopathy was seen in 5 of 27 patients with inv(16) and was associated with a shorter duration of CR1 [97]. This leukemia is commonly referred to as M4Eo because of the myelomonocytic differentiation of the blasts and the abnormal and usually increased eosinophils in the bone marrow. The abnormal eosinophils, especially the immature eosinophils, contain variable numbers of large basophilic granules. Hyposegmented nuclei are occasional seen in mature eosinophils. Immunophenotyping often reveals multiple blast populations including an immature population expressing CD34 and/or CD117 and populations with monocytic differentiation and granulocytic differentiation. Expression of CD2 is often seen but is not specific for this leukemia. The M4Eo syndrome associated with CBFB-MYH11 has eosinophils at all stages of maturation with immature basophilic granules in the earlier phases of cell development [98]. The abnormal eosinophils stain positive with PAS, Sudan Black and nonspecific esterase, typical finding in granulocytic but not in eosinophilic precursors. The eosinophils usually express CD 2 T cell antigen [98].

Inv(16)/t(16;16) can be a subtle finding by conventional cytogenetics. Therefore if the morphology is suggestive of this process, FISH or molecular methods should be pursued to demonstrate the CBFB/MYH11 rearrangement. Conventional cytogenetics is still important in these cases as approximately 40% of patients with inv(16) will have secondary abnormalitiesSome of the secondary cytogenetic abnormalities differ from those seen with t(8;21). These include +22 (15–20% in inv(16)), and +21(6% in inv(16)) both observed more often in inv(16) than t(8;21) [14, 83, 84]. Some authors have shown that AML M4 with +22 as the only aberration, usually has an undetected inv(16) [99]. Other abnormalities are similar in inv(16) and t(8;21) including complex cytogenetic abnormalities (in 15%) [14], +8 (6–8%) and -7/7q- (5%) [14, 83, 84].There are few patients reported with CML who had 16q22 (t(16;16)(p13;q22) and inv(16)(p13;q22) at the time of blast crisis [100]. These patients had extramedullary disease and poor outcome.

The spectrum of CBFB/MYH11 fusion transcripts with inv(16)/t(16;16) is heterogeneous, depending on the exons of the CBFB and MYH11 genes that are fused; type A fusion transcript is the most common. Rare CBFB/MYH11 fusion genes (D, E, Avar, Bvar, F, G, H, J, S/L) were detected in 34 out of 162 cases (21.0%). These rare types were more frequent in t-AML (P = 0.0106), were associated with lower WBC (P < 0.0001) and with lower levels of CD2, CD13, CD33 and CD90 compared to type A [101]. These rare types did not influence outcome compared to common type A. In a study by Schwind et al., KIT mutation was not detected in any of the 26 non-type A patients, but was seen in 26% of the type A group [102].

Current Therapy and Outcomes

The CR rate in CBF AML has consistently been reported at 86 to 88% [14, 83, 84]. Although more patients with t(8;21) needed two or more cycles of induction (30% vs. 20% for inv(16)), this did not affect RFS [14]. In a Cancer and Leukemia group B (CALGB) study, higher bone marrow blasts, older age, lower platelet counts at diagnosis and non-white race were each (in multivariate analysis) associated with a lower CR rates in all CBF leukemias [84]. In the German AML intergroup, a high WBC count and old age were associated with lower CR and a lower OS in inv(16), but not t(8;21) patients [83]. The French AML group reported a lower CR associated with higher WBC counts and low platelet counts in inv(16), but reported no data on t(8;21) [90, 103]. A recent report by Appelbaum et al. supported the association of older age, inv(16) with uncommon cytogenetic abnormalities and higher WBC with a lower CR, but found no association with low platelet count or race [14]. This report also noted resistant disease following induction in 8% of patients having -7/7q- and complex cytogenetic abnormalities [82]. Patients with LOS had a slightly higher CR than the rest of CBF patients [14].

The estimated OS for patients with CBF AML is 48% at 5 years and 44% at 10 years [14]. Several factors affect the OS. The CALGB study reported shorter survival with old age, t(8;21) and low platelet count on diagnosis [84]. The German AML group reported lower OS in t(8;21) patients associated with higher WBC, low platelet counts, presence of LOS, but found no single factor predictive of OS in inv(16) [83]. Appelbaum et al. reported a lower OS associated with increasing age, higher peripheral blast percentage, complex abnormalities and t(8;21). The age, blast percentage and complex abnormalities had a similar negative OS impact in both t(8;21) and inv(16). After adjusting for these effects, t(8;21) by itself had a poorer outcome (45% 5 year survival vs. 50% for inv(16)) [14].

The RFS in CBF AML was influenced by age and peripheral blast percentage [14]. The German AML group reported a shorter RFS in t(8;21) patients with high WBC and low platelet counts and improved RFS in inv(16) with +22 [83]. Despite minor differences in the reported studies, it can be generalized that old age, high WBC, high peripheral blast percentage and perhaps low platelet count at diagnosis are associated with a worse OS and RFS.

Patients with del(9q) with t(8;21) were reported to have unfavorable prognosis by Schoch et al. [104], but this was refuted in several other reports with no significant effect in del(9q) on OS, CR or RFS [14, 88, 105].

KIT mutations with CBF AML has been reported recently in several studies [59, 60, 106]. KIT17 mutation at codon 816 in patients with t(8;21) has been associated with a high WBC count at diagnosis, higher relapse rate and a poorer OS [57, 58, 60, 69]. Paschka et al. recently reported that in patient with inv(16), KIT mutations were associated with a higher relapse incidence (56% at 5-year vs. 29% for non-KIT mutants), similar CR and a worse overall survival when compared to wild type-KIT patients [69]. In patients with t(8;21), KIT mutations had no effect on OS and CR, but affected the relapse rate (70% @ 5 years vs. 36% for wildtype KIT, P = 0.004). In a pediatric population, KIT mutations were detected in 38% of 203 AML patients with CBF [107]. Relapse rate, DFS, and OS were similar at 5-year between the patients with KIT mutations and those without. Cairoli et al. showed worse prognosis and higher relapse rate associated with KIT 816 in t(8;21) patients, but no prognostic impact in inv(16) patients [57]. Care et al. reported that KITD816V affected the relapse rate in inv(16), but not the OS [60]. In Summary, most but not all studies show that KIT mutations are associated with worse outcomes in the setting of CBF leukemia. These patients are considered for clinical trials or allogeneic HCT aiming at improving remission rates.

Choice of therapy

The choice of treatment for CBF AML has a significant impact on CR, OS, and RFS (Table 1). In 1994, it was observed that AML patients had a better outcome with high dose cytarabine than with low or intermediate doses [108]. Later it was shown that the greatest benefit of high dose cytarabine is seen in patients with inv(16) or t(8;21) [109]. Cytarabine was further studied and it was shown that in addition to its dose effect, the number of courses given had an impact on survival. Byrd et al. reported three or four induction and consolidation courses to be superior to one course of high dose cytarabine [3, 4]. The foundation of treatment for CBF AML is recommended to be at least three courses of high dose cytarabine [16]. In a recent multicenter retrospective study comparing the impact of treatment regimens on outcome, FA (fludarabine with intermediate dose cytarabine) and HIDAC (high dose cytarabine) induction regimens were compared with alternatives. For patients who achieved CR with induction, FA, HIDAC, hematopoietic cell transplantation (HCT) and other post remission protocols were compared [14]. The CR was somewhat, but not significantly higher with FA than with other induction regimens. There was better OS and RFS in patients receiving FA or HIDAC after induction when compared to other chemotherapies or low dose cytarabine. Patients who received FA, HIDAC or HCT during consolidation had superior survival when compared with conventional or low dose therapy (5-year survival estimate: 61% for FA, 61% for HIDAC, 50% for HCT, 31% for other, P = 0.007) [14]. The French group reported a better survival with intensive therapy when compared to one cycle of intermediate dose cytarabine (IDAC) in patients with t(8;21), but no similar effect in inv(16) [90, 103]. The German AML group reported no significant prognostic impact of the cytarabine dose on OS and RFS [83]. MD Anderson reported that fludarabine-based chemotherapy regimens might be superior than anthracycline and cytarabine regimens [110]. The median DFS was not reached in patients receiving fludarabine and cytarabine (FLAG), but approximately 80 weeks in patients receiving no fludarabine, P = 0.02.

Table 1. Outcomes after Different Cytarabine Based Consolidation Therapies for CBF AML
 CBF-AML cytogeneticsNTreatment planLeukemia free survivalCumulative incidence of relapseOverall survival
  1. a

    HDAC defined as cytarabine administered by 2.0 g/m2 or more bolus intravenous infusions for a total dose of 16 g/m2 or more, while intermediate-dose cytarabine (IDAC) was defined as cytarabine administered either by bolus or continuous intravenous infusions for a total dose of between 1.5 and 16 g/m2.

  2. b

    Intermediate-high dose cytarabine defined as intravenous bolus of cytarabine 500mg/m2/day for ≥2 days.

  3. CBF = core binding factor; AML = acute myelogenous leukemia; Ara-C = cytarabine; HDAC = high dose Ara-C; HDAC-2 = 2 intensive postremission cycles including 1 HDAC cycle.

Byrd et al., 1999 [3]t(8;21)50Ara-C 3g/m2 every 12 hr on days 1,3,5; 1 vs. 3 to 4 courses@5 Years: 38% for single vs. 71% for 3 to 4 courses @5 Years: 44% vs. 76% for single vs. multiple courses
Nguyen et al., 2002a [103]t(8;21)114HDAC-2 IDAC-2 IDAC-1@ 3 Years 65% 59% 35% @3 Years 72% 72% 39%
Byrd et al., 2004 [4]inv (16) or t(16,16)48Ara-C 3g/m2 every 12 hr on days 1,3,5; 1 vs. 3 to 4 courses@5 Years: 30% for 1 vs. 57% for 3 to 4 courses@5 Years: 70% vs. 43%@5 Years: 70% vs. 75% for 1 vs. multiple courses
Marcucci et al., 2005 [85]inv (16)101Ara-C Multiple courses a: 100 mg/m2/d for 5 days × 4 courses; b: 400 mg/m2d for 5 days × 4 courses; c: 3 g/m2 every 12 hr days 1,3,5 for 3-4 courses Single course 3 g/m2 every 12 hr days 1, 3, 5 followed by etoposide/cyclophosphamide for one course and diaziquinone/mitoxantrone for one course @5 Years: 70% for 1 course courses vs. 43% for multiple@5 Years: 70% vs. 69%
t(8;21)96 @5 Years: 38% vs. 64%@5 Years 56% vs. 43%
Prebet et al., 2009b [114]inv (16)73Low dose Ara-C vs. intermediate-high dose Ara-cMedian 23 months for both  
t(8;21)56@10 months for low dose vs. not reached  

Based on superior outcomes seen with high dose cytarabine, we recommend consolidation with three to four cycles of Hi-DAC for most CBF AML patients . On the other hand, the dose of HIDAC (mainly because of effects and toxicity) andthe exact number of cycles vary between different institutions (Table 1) [111-113].

CBF AML in the elderly

A French study evaluated the significance of t(8;21) or inv(16) in 147 elderly patients (>60 years) [114]. CR rates were encouraging at 88%. Treatment-related mortality was also acceptable ∼15%. However, long-term outcome was poor with 5-year OS and DFS of 31% and 27%, respectively. The presence of del (9q) carried a poor prognosis. This study indicated that anthracyline and cytarabine based induction is effective, but postremission therapy needs to be improved in elderly CBF AML. Similar findings regarding a worse OS was seen among CBF AML patients >50 years [115]. CBF AML was rare (1.7%, 5 of 293) in elderly patients (≥60 years), but associated with the best outcome compared normal cytogenetics or other cytogenetic aberrations [116]. A recent SEER database analysis showed that patients with CBF AML aged 75 to 84 had worse survival than patients aged 15 to 44 years (HR 5.61, P = 0.0002) [15]. Our institutional preference is to consider clinical trials or allogeneic HCT for elderly (>60 years) CBF AML patients.

Secondary CBF AML

Although primary CBF AMLs have a better prognosis than other AML subgroups, this is not true for secondary CBF AMLs [110, 117, 118]. In a single center report from MD Anderson, secondary CBF AML was found in 9% of 188 CBF AML patients. The definition of secondary CBF AML required a prior malignancy, chemotherapy or radiation. The median OS was approximately 400 weeks for de novo CBF AML compared to 150 weeks in secondary CBF AML, P = 0.001 [110]. Event-free survival (EFS) was also inferior in the secondary CBF AML group, P = 0.04. From the same center, Gustafson et al compared 13 cases of t(8;21) treatment-induced AML (t-AML) with 38 adult cases of de novo t(8;21) AML [119]. t-AML patients were older, (P = 0.001), had a lower WBC (P = 0.039), and had substantial morphologic dysplasia compared to the de novo AML patients. Additional abnormalities in eight patients included loss of a sex chromosome and trisomy 4. However, cytogenetic findings did not differ between the two groups. OS was worse in t-AML patients compared to de novo AML (19 months vs. not reached; P = 0.002). Another study confirmed that in t-AML with t(8;21), most of 44 patients (75%) had another cytogenetic abnormality, including loss of a sex chromosome, del(9q), abnormalities of chromosome 7 or chromosome 8 [118]. OS was not significantly affected by additional chromosomal aberrations. Interestingly, when this study compared t(8;21) t-AML patients with other 21q22 subsets, OS was significantly better in those with t(8;21), P = 0.014. In a report of 32 patients with secondary CBF AML (26 t(8;21) and 6 inv(16)), CR rates were around 80% and DFS at 2-year was around 50% [117, 120].

In most of the poor prognostic CBF-AMLs (KIT mutation+, secondary CBF-AMLs), induction therapy followed by HiDAC remains the standard therapy. Clinical trials or allo HCT is a consideration especially in the KIT positive CBF AMLs. As for the older fit AML patients, we recommend a clinical trial or allo HCT in first complete remission

Hematopoietic cell transplantation (HCT) in CBF AML


A Korean study reported on postremission therapy for 138 CBF AML patients [121]. DFS and OS at 5-year were similar between patients undergoing allogeneic HCT and HiDAC. However, when outcome analysis was examined before and after 2003, 3-year DFS and OS was significantly superior in the HCT group after 2003 (DFS: 95.2% with HCT vs. 59.3% HiDAC, P = 0.008; OS: 95.2% with HCT vs. 59.6% HiDAC, P = 0.032). The OS and DFS among patients receiving HiDAC was similar before and after 2003.


Excellent EFS was reported (93%) in 14 patients who underwent autologous HCT with no minimal residual disease (MRD) in their peripheral blood stem cells (PBSC). The collected autologous PBSC products were tested with nested PCR for MRD (with a detection threshold of 10−5). EFS was shorter in a small group of four patients without HCT. Importantly, some relevant prognostic mutations were not evaluated [122].

A retrospective analysis of 338 CBF AML patients, who received either an autologous or allogeneic HCT from the Japanese Society of Hematopoeitic cell transplantation database, showed 3 year OS of 50% and 72% for t(8;21) and inv(16), respectively (P = 0.002). Although no survival difference was observed between t(8;21) and inv (16) patients who received allogeneic HCT in first CR, a major survival difference was noted for patients receiving allogeneic HCT in second or third CR (45% for t(8;21) versus 86% inv(16), P = 0.008) [123]. Among patients in first CR, OS at 3 years was similar for allogeneic and autologous transplantation in both the t(8;21) and inv(16) groups [123].

Other therapies

Gemtuzumab ozogamicin(GO)

GO is a humanized anti-CD33 antibody linked to a potent DNA damaging agent calicheamicin derivative (N-acetyl gamma calicheamicin) that has shown improvement in relapse rates and overall survival among AML patients [124-127]. The addition of GO 3mg/m2 on day one of induction therapy was associated with lower incidence of relapse (68% vs. 76%) and an improvement of OS (25% vs. 20%, P = 0.05). The benefits seen with GO were reproducible among the good risk AML including CBF AML. A recent metanalysis of five randomized trials (n = 3325) showed that GO significantly reduced relapse risk and improved OS at 5 years. The absolute survival benefit was more apparent in the favorable cytogenetics including CBF AML (20.7%, P = 0.0006) [124].

Histone deacetylase inhibitors

The current basis of treating CFB AML may evolve as we better understand the molecular basis of this disease. As mentioned previously, the RUNX1/RUNX1T1and CBFB/MYH11 recruit histone deacetylases and DNA methyltransferase 1 [16] leading to transcriptional repression [128, 129]. Histone deacetylases (HDAC) allow tighter binding of histones to DNA by removing acetyl groups and hence prevent transcription [130]. Several HDAC inhibitors (phenylbutyrate, trichostatin A, SAHA, Valproic acid and depsipeptide) are under investigation [131-133] as single agents or in combination with methyltransferase inhibitors. Panobinostat was shown to be effective in an murine AML model bearing t(8;21) [134]. Further studies with HDACs are needed to establish doses, infusion schedules and side effect profiles along with treatment outcome in patients with CBF AML [135].

DNA methyl transferase inhibitors

The CBF mutations in AML recruit, in addition to the histone deacetylases, DNA methyltransferase 1 [16]. Decitabine (5-aza-2′-deoxycitidine) is a pyrimidine analogue which irreversibly inhibits methyl transferases and hence leads to hypomethylation of the promoters of tumor suppressor genes, their activation and cell differentiation [136]. Decitabine has demonstrated efficacy in refractory leukemia [137]. Induction of histone acetylation was also reported in with combination of butyrate and decitabine in t(8;21) AML [138].

Proteasome inhibition

The ubiquitin-proteasome pathway is central in degradation of RUNX1 [19] and hence proteasome inhibitors may have therapeutic potential in CBF AML by stabilizing the RUNX1 and preventing its degradation. Bortezomib is a potent and specific proteasome inhibitor [139] with documented activity in myeloma and CLL, but studies in CBF AML have not been reported.

Tyrosine kinase inhibitors

Constitutively active tyrosine kinases have been identified as attractive therapeutic targets in leukemia therapy including: ABL1, CSF1R, KIT, FLT3, PDGFRA and PDGFRB [140]. The high incidence of KIT mutations in CBF AML patients prompted several trials to investigate the effect of tyrosine kinase inhibitors (TKI) in AML.. In a phase II pilot study by Kindler et al., 5 of 21 patients with refractory AML showed response to imatinib 600 mg daily [141]. Imatinib given alone or in combination with other agents showed variable responses in CBF AML patients with KIT 8 mutation, whereas no response was seen in patients with D816 mutation [142, 143].

Dasatinib was evaluated in leukemic cell lines, with and without KIT mutations. Dasatinib could suppress src kinase and kit activity, inhibit the phosphorylation of the downstream target AKT pathway, block proliferation and induce apoptosis in CBF cell lines [144]. A recent study (CALGB 10801) evaluated the effect of adding dasatinib to standard induction and consolidation therapy for patients with newly diagnosed CBF AML [145]. The 1 year DFS and OS were 90% and 87%, respectively.

Monitoring of genetic products for minimal residual disease (MRD)

Patients with t(8;21) in morphologic CR may be positive for RUNX1/RUNX1T1 using highly sensitive molecular testing to identify MRD [146]. The significance of detecting MRD in CBF AML is unclear with some, but not all studies showing prediction of relapse. Several studies showed that patients with high levels of RUNX1/RUNX1T1 at diagnosis or poor response to treatment (defined as ≤2 logs after induction or high levels (>10−5) after consolidation) might be at high risk for relapse [101, 147, 148]. The median reduction of initial RUNX1/RUNX1T1 expression level was 4 logs (range 0–5) after both induction and consolidation therapies [149]. Among 45 patients with RUNX1/RUNX1T1 AML, those who did not achieve the median level in RUNX1/RUNX1T1 expression after induction therapy had a worse EFS (36 months vs. 13 months; P = 0.004) and OS (not reached vs. 22 months) compared to those who did.

Several studies followed CBF AML patients with quantitative real-time PCR and failed to identify a level that predicted relapse after either induction or consolidation. Most of these studies found that increases in copy numbers of the RUNX1/RUNX1T1 transcript after treatment was followed by relapse [150, 151]. LFS was shorter in patients with >1 log increase in transcript levels (median LFS not reached vs. 14 months, HR 8.6, P = 0.008) [150]. Others have suggested that extramedullary relapse should be suspected if RUNX1/RUNX1T1 levels were persistently higher in peripheral blood compared to bone marrow [152].

Lately, other studies focused on the prognostic importance of alternate splicing of RUNX1/RUNX1T1 [101, 147, 148]. Alternate splicing of exon 9a in RUNX1/RUNX1T1 transcripts can contribute to leukemogenesis and its persistence may indicate relapse [153]. The same group reported that leukemia clones with each molecular marker have different doubling times, therefore sampling intervals for detecting relapse should be different [154]. With relapse detection at 90% at a median time to relapse of 60 days, they suggest blood sampling every 6 months should be performed for CBFB/MYH11 leukemias and bone marrow sampling every 4 months. Similarly, Jiao reported that in 118 patients with higher levels of RUNX1/RUNX1T1 exon 9a had more KIT mutations and worse prognosis (2 year EFS of 15.4% and OS 27.3%) vs. those with low levels (EFS 44.6%; OS 58.6%).

In a recent study, 116 patients with t(8;21) were monitored for RUNX1/RUNX1T1 after their second consolidation. Patients were then assigned as low versus high risk based on MRD. Low risk patients received chemotherapy-autologous HCT while the high risk group received allogeneic HCT. MRD status was an independent prognostic indicator in a multivariate analysis [155]. Elevated expression of BAALC and WT1 by PCR were also predictive of inferior OS and higher relapse post HCT [156].

The UK MRC AML-15 trial included prospective sample collection for MRD assessment by PCR in 278 patients [163 with t(8;21) and 115 with inv(16)] [157]. Pretreatment levels of both fusion genes were not associated with clinical presentation or outcomes. While >3 log reduction in RUNX1-RUNX1T1 BM transcripts was significant in patients with t(8;21), the absolute copy number (>10 CBFBMYH11) in peripheral blood during CR1 was the strongest predictive factor for relapse risk. However, OS in modest (1–3 log reduction) and deeper responses (>3 log reduction) were similar. There were no significant cut-off levels identified during consolidation. All patients with positive MRD (t(8;21) BM > 500 copies or PB > 100 copies; inv(16), BM > 50 copies or PB > 10 copies) relapsed compared to MRD- negative patients (7% for each group). These higher relapse rates in MRD+ patients led to poorer OS (57% and 59% vs. 94% and 95% for t(8;21) and inv(16) patients, respectively). In addition, rising fusion gene levels were associated with clinical relapse within 3 to 4 months. In other studies using PCR monitoring, reduction in fusion gene levels after induction from levels at diagnosis (<1% [158] and >3 log reduction [147]) were predictive of impending relapse. Similarly time between molecular MRD detection and hematologic relapse was 3 months [158].

Peripheral blood versus bone marrow for MRD monitoring

The site for sampling MRD as a significant predictor of relapse for AML differed among studies [157] and some discrepant results between BM and PB samples were noted. In general, finding MRD+ in PB and MRD− in BM was rare. Discrepancies between BM and PB (MRD+ in BM, but MRD− in PB) ware more common and decreased after each line of therapy (20%, 28%, and 16% after induction, consolidation and observation, respectively). Corbacioglu et al found the MRD sensitivity gap between BM and PB was 31%, 43%, and 11% at similar time points. In both studies, the percentage of MRD− patients in both PB and BM increased after each line of therapy. Boeckx et al. found no correlation between MRD levels in PB and corresponding BM samples during follow-up in CBFB-MYH11+ AML [159]. In contrast, comparison of BM and PB samples showed similar sensitivity for detecting RUNX1/RUNX1T1transcripts [147].

These studies show promising opportunity to use the fusion gene products as a measure to guide decisions on further therapy before relapse occurs. The standardization of testing methods, preferred cut-off level for MRD, timing of sampling, and best sample site has not been established. Moreover, first relapsed CBF patients still do well following allogeneic HCT. Therefore, it may be premature to treat patients based solely on molecular MRD testing as it may cause overtreatment in some patients. The treatment algorithm (algorithm 1) takes into consideration the latest data on MRD, targeted therapy, transplantation and high risk CBF AML categories.

Figure Algorithm 1.

Treatment Algorithm for Core Binding Factor Leukemia. 1We recommend c-kit testing on all newly diagnosed CBF AML patients. 2Allogeneic transplantation is an option in this situation if no available clinical trials

Outcome of relapsed CBF-AML

Outcomes of CBF-AML patients not only very good after upfront therapy but also after first relapse. Recent French study showed that 88% of 145 relapsed CBF-AML patients achieved CR2. The estimated 5-year LFS and OS were 50% and 51%, respectively. Using GO was asscoaited with higher 5-year DFS whereas older age and shorter CR1 duration was associated with a shorter OS. After achieving a second CR, the 78 patients who received allo HCT has superior DFS and OS compared to those who did not receive HCT,DFS(57% vs. 42%, HR 0.58) and OS (59% vs. 45%, HR 0.55) for Allo HCT receipietns and non transplant receipients respectively [160]. In this study there was no difference between patients with translocation t(8 ;21) and inversion inv(16)/t(16;16) regarding CR, DFS, and OS; however some studies can argue that outcomes are inferior in patients with t(8;21) [83, 84].


In summary, CBF AML constitutes 15% of all AMLs and with appropriate therapy leads to favorable OS and RFS. The finding that Inv (16) and t(8;21) have significant differences in presentation and outcome suggests that future studies to decipher the molecular basis of these differences may guide the development of new treatment approaches. Importantly, incorporating TKIs and GO or other targeted therapies into the management of CBF AML and active MRD monitoring may further improve the outcomes of this already favorable subgroup of AML and increase the long term DFS and cure. The benefit of alloHCT in a subgroup of MRD+ or other high risk patients in CR1 remain to be clarified in the current era wheresupportive care and outcomes of allogeneic HCT have improved.