Introduction

The Philadelphia chromosome, corresponding to the BCR-ABL rearrangement, is the cytogenetic hallmark of chronic myelogenous leukemia (CML) and is frequently found in high-risk acute lymphoblastic leukemia (ALL). In addition, 0.5–3 % of all acute myelogenous leukemia (AML) cases also carry this fusion gene [1–5].

The distinction between this rarely occurring BCR-ABL+ AML and CML in (primary) blast crisis can be difficult, and the diagnostic challenges have led to an ongoing debate as to whether BCR-ABL+ AML does actually exist as a distinctive subgroup of AML. During the past, numerous case reports on BCR-ABL+ AML have been published, diminishing further doubts on its existence.

In this study, all published cases of BCR-ABL+ AML since 1975 as well as cases from our institution were analyzed with respect to their clinical and molecular features. Furthermore, a diagnostic algorithm for this rare but challenging problem in clinical practice will be suggested.

Patients and methods

References were primarily obtained by a PubMed literature search. Cross-references were used to obtain additional cases.

Cases from the literature were collected starting in 1975; cases from our own institution were available in a database that begins in 1998. In total, 164 cases of de novo BCR-ABL+ acute myeloid leukemia and 14 cases of relapsed BCR-ABL+ AML were analyzed. For a better comparability between the cases over this long period of time, published cases were re-classified (if necessary and possible) according to the latest WHO classification [6] or, in the case of acute undifferentiated leukemia (AUL)/mixed phenotype acute leukemia (MPAL), according to EGIL and/or WHO classification. Since classification was based on the EGIL criteria in the majority of cases, a re-classification according to WHO 2016 could not always be made due to a lack of information. Cases re-classified to AUL according to EGIL or to MPAL according to WHO were excluded from the analysis, since MPAL with BCR-ABL re-arrangement is an own entity within MPAL in WHO classification. Eight cases from the literature and two cases from our institution were excluded after being re-classified as AUL/MPAL according to the immunophenotype reported. Fifteen additional cases from the literature were excluded since they were BCR-ABL+ AUL according to the authors’ own statements.

Eight cases with 100 % BCR-ABL-positive metaphases, published before 1990, were excluded from the analysis because no additional features were described helping to distinguish AML from CML. One additional case was excluded because the patient had an antecedent myeloproliferative disorder, and four cases were excluded since no information was given that helped to distinguish de novo BCR-ABL+ AML from CML blast crisis. Cases describing secondary AML with an antecedent MDS are included and analyzed as AML with myelodysplasia-related changes. Finally, data on epidemiology, molecular pathogenesis, and phenotype of BCR-ABL+ AML are based on 126 cases of de novo BCR-ABL + AML, fulfilling the above-mentioned criteria. A detailed summary of all cases is given in the supplementary tables.

Results

Epidemiology

The Philadelphia chromosome is supposed to be a rare event in acute myeloid leukemia. Its reported incidence ranges from 0.5 to 3 % [1–5]. Most cases are described as de novo AML, few authors report on a new acquisition of the Philadelphia chromosome upon disease relapse or transformation of an antecedent myelodysplastic syndrome into AML. In contrast, BCR-ABL is much more common in acute leukemia of ambiguous lineage and thus was classified as an own subgroup in WHO classification [6].

So far, the overall incidence is mainly estimated from larger case series without any standardized definition. The accuracy will soon improve since several large study groups are acquiring these data in a joint effort to improve the molecular characterization of AML. Currently, a precise distinction between AML and CML in blast crisis is still limited by the lack of a clear definition; however, CML in primary blast crisis can be considered an extremely rare event and patients with secondary blast crisis can usually be identified by their medical history.

Molecular pathogenesis and phenotype

By definition, AML is a clonal disease. In comparison to most solid cancers, AML shows a limited spectrum of cytogenetic aberrations and mutations with different roles in pathogenesis [7–9].

Basic features of leukemogenesis according to the model of Gilliland and Griffin [10] are inhibition of differentiation and apoptosis (caused by a class II mutation as the leukemia initiating event) and the subsequent acquisition of a proliferative advantage (due to class I mutations). Typical examples are PML-RARA as a class II event and FLT3 or KIT mutations as class I events [7, 10]. Applying this model, BCR-ABL belongs to the group of class I mutations, conferring a proliferative advantage to the affected cells [11].

There is increasing evidence that the presence of age-dependent clonal hematopoiesis is a risk factor for the development of hematological malignancies [12, 13]. It is well known that BCR-ABL may also temporarily occur at very low levels in the blood of healthy individuals [14–17]. The co-incidence of spontaneously occurring BCR-ABL within clonal hematopoiesis could possibly explain the development of a BCR-ABL+ AML and the underlying “mutational background”; the maturation stage of the affected cell might decide whether the additional acquisition of BCR-ABL leads to AML, CML, or just a very small and only temporarily detectable BCR-ABL+ subclone.

BCR-ABL1 has been described in AML together with different class II aberrations such as CBFB-MYH11, RUNX1-RUNX1T1, PML-RARA, and NPM1 [1, 11, 18–35].

Based on the aforementioned model of AML pathogenesis, possible interplays of BCR-ABL with co-occurring genetic events are conceivable but the proof-of-principle experiment in an appropriate pre-clinical model is missing due to the lack of BCR-ABL+ AML models. The generation of a BCR-ABL+ AML cell line or the overexpression of various known genes together with BCR-ABL in hematopoietic precursor cells could be the first step to improve our understanding of molecular interactions during pathogenesis.

Recently, two major clonal evolution patterns during AML relapse were defined based on a genome-wide sequencing approach of several AML cases [36]. The first pattern implicates that the founding clone of the disease gains mutations and evolves into the relapse clone(s). The second suggests that a subclone of the initial clone survives therapy, gains additional mutations, and expands at relapse. In both models, a founding mutation is followed by a cooperating mutation able to convey a proliferative advantage [36]. Interestingly, mutations in genes involved in epigenetic regulation appear to occur early in the evolution of AML [9]. In view of this model, the acquisition of BCR-ABL is an interesting option to explain leukemia relapse and will be discussed later.

Interestingly, Nacheva EP et al. [37] recently showed that BCR-ABL+ AML displays characteristics of lymphoid disease as it shares deletions of IKZF and CDKN2A/B as well as the concomitant loss within the immunoglobulin and T cell receptor gene complexes that are all also found in BCR-ABL+ ALL and CML. Remarkably, all these aberrations have been found to be absent in myeloid blast crisis of CML [38], possibly adding a helpful tool for differential diagnosis in difficult cases.

However, larger data sets are needed before these biomarkers can be integrated into routine differential diagnosis. Confirmation of this intriguing finding would provide further evidence for a different and possibly unique biology of BCR-ABL+ AML, namely the appearance of lymphoid genomics in the guise of a myeloid phenotype.

A possible role of BCR-ABL within AML pathogenesis is depicted in Fig. 1.

Fig. 1
figure 1

BCR-ABL in AML pathogenesis. In this simplified model of leukemogenesis, two driver mutations are required: a class-II mutation, conferring inhibition of differentiation and apoptosis, and a class-I mutation conferring a proliferative advantage. BCR-ABL is likely to represent a class-I mutation

Full size image

Characteristic features of BCR-ABL+ AML

Regarding the WHO 2016 classification, most cases of BCR-ABL+ AML belong to the subgroup AML-NOS (38 %, corresponding to 48/126 of de novo BCR-ABL+ AML cases), followed by AML with myelodysplasia-related changes (32.5 %, 41/126 cases) and AML with recurrent genetic aberrations (in particular CBF leukemia with 16.7 %, corresponding to 21/126 cases). It seems very likely that the accumulation of cases in the subgroup AML-NOS may be caused by incomplete cytogenetic and/or molecular analyses, particularly in older case reports. Considering coinciding cytogenetic/molecular events, CBF leukemia with inv(16) or t(8;21) and NPM1 mutations are reported by several authors [1, 11, 18–35]:

  1. 1)

    BCR-ABL in CBF-AML

Interestingly, 30 cases (23.8 %) of the 126 collected de novo BCR-ABL+ AML cases were AML with recurrent genetic alterations, mainly CBF leukemias in 21/126 cases. The most common recurrent genetic alteration was inv(16) in 17 cases; only one case was reported as promyelocytic leukemia carrying the t(15;17) translocation and three cases with the translocation t(8;21). This accumulation of “good risk” phenotypes (irrespective of BCR-ABL) is remarkable, since the vast majority of cases with BCR-ABL+ AML belongs to the intermediate or high-risk group according to ELN. On the one hand, this may explain why BCR-ABL+ AML is perceived as a “high risk” AML; on the other hand, it clearly shows that BCR-ABL may also arise in a “good risk” molecular background (see Fig. 2c).

Fig. 2
figure 2

BCR-ABL + AML: clinical and molecular features. a Distribution of patients according to FAB classification. b Distribution according to WHO classification. c Risk groups according to the European LeukemiaNet (ELN) criteria. n.a. not available

Full size image

However, the occurrence of inv(16) is not restricted to AML. It can also be found in blast crisis of CML [reviewed in 18, 39]. Therefore, a clear distinction of de novo AML and CML blast crisis remains challenging even in these inv(16)+ cases. The distribution of BCR-ABL p190 and p210 in inv(16)-CBF-AML was notably different from CML blast crisis: 53 % (9 cases) carried the p190 transcript in AML, whereas p190 was only found in one of 19 published cases of CML-AP/BC with inv(16) [18, 19, 39–48]. Therefore, the detection of the minor BCR-ABL transcript in AML with inv(16) might be helpful in identifying de novo BCR-ABL+ AML. Furthermore, inv(16) is a characteristic class II event and therefore present in every leukemic clone in AML, whereas it is mostly not in blast crisis of CML.

In AML with t(8;21), two thirds of cases were also p190 positive underlining the aforementioned hypothesis [11, 31].

Whether the overall prognosis of this favorable AML subgroup is worsened by BCR-ABL cannot be determined due to the small number of patients, but several long-term survivors are reported in the cohort of BCR-ABL+ CBF leukemia.

  1. 2)

    BCR-ABL+ AML with concomitant mutations

Considering the 126 cases of de novo BCR-ABL+ AML, 6 had an additional NPM1 mutation at primary diagnosis [1, 11, 32–34]. In total, information on NPM1 mutational status was given in 15 cases. Notably, 3 of these 6 NPM1+ AML patients were long-term survivors, and for the remaining 3 cases, no survival data were available. This may again show that the overall prognosis of AML is not determined by BCR-ABL alone. However, the small number of cases does not allow any final conclusion.

The FLT3 mutational status was reported in 13 cases with two of them being positive (one FLT3-ITD, one FLT3-TKD) [1, 49]. This is remarkable, since it shows that two class I mutations, each conferring a proliferative advantage, can be acquired subsequently.

  1. 3)

    Acute myeloid leukemia with myelodysplasia-related changes

Eighteen cases of MDS that became BCR-ABL positive upon transformation into secondary AML were reported [3, 11, 50–60]. In nine cases, information regarding the type of the BCR-ABL transcript was available. In 17 % of all AML cases (3/18), expression of p190 was reported; another 22 % (4/18) carried the p210 transcript. In 11 % (2/18) both transcripts were detectable. Fifty percent (9/18) of these patients had a complex-aberrant karyotype at the time of BCR-ABL acquisition. In most cases, BCR-ABL was documented for the first time upon transformation into AML (the same finding as reported by [61]).

The WHO classification of 2008 and its 2016 revision include cytogenetic criteria for the diagnosis of “acute myeloid leukemia with myelodysplasia-related changes” [6]. According to these criteria, 23 additional cases can be classified as AML with myelodysplasia-related changes although a history of documented MDS is not reported [1, 4, 30, 34, 61–66]. Therefore, AML with myelodysplasia-related changes represents the second largest subgroup with 41 cases in total after AML-NOS.

  1. 4)

    Therapy-related BCR-ABL+ AML

Five cases of therapy-related BCR-ABL+ AML were described, two more cases can possibly be considered as “therapy-related”, but a lack of information prevents a definitive classification [31, 35, 60, 67–70]. No correlation with radiation or a specific therapy-regimen can be established due to the low number of cases.

  1. 5)

    BCR-ABL at AML relapse

In 14 published cases, BCR-ABL was acquired upon AML relapse [11, 55, 71–77]. In 7 of these 14 patients, the morphologic classification was given. Four cases were classified as AML with maturation (FAB-M2), one case as AML with minimal differentiation (AML-FAB M0), and another case as myelomonocytic AML (AML-FAB M4). In these cases of BCR-ABL+ AML, the p190 transcript was reported in seven cases (one case expressed p210; no information was given in six cases). These data underline p190 as the predominant BCR-ABL transcript in AML.

Ten of these 14 patients succumbed to their disease or infectious complications; no outcome was reported for the remaining cases. At relapse, most patients acquired further, mostly high-risk cytogenetic aberrations such as deletions of chromosome seven. Therefore, the unfavorable outcome is not necessarily due to BCR-ABL alone; however, BCR-ABL may contribute to an increased genetic instability.

  1. 6)

    BCR-ABL+ mixed phenotype acute leukemia

The occurrence of a Philadelphia chromosome is a frequent event in acute leukemia of ambiguous lineage as defined by the WHO classification [6].

In a larger case series, 16 % of the evaluated MPAL patients had a BCR-ABL rearrangement [78]. In a detailed analysis by the German ALL group, 30 % of their biphenotypic acute leukemia (BAL) patients were BCR-ABL positive [79]. In another case series, Afty et al. [80] found a shorter survival in the BCR-ABL+ BAL as compared to BCR-ABL+ AML. Furthermore, patients with BAL were older than patients with AML. None of the patients of this case series was treated with a tyrosine kinase inhibitor.

Clinical features and diagnosis

Little is known about characteristic clinical features of BCR-ABL+ AML. In general, its nature as a high-risk leukemia is assumed although there are no systematic data on outcome. Thus, the collected cases were analyzed regarding their clinical features.

As mentioned above, the distinction between BCR-ABL+ AML and CML in primary myeloid blast crisis can be difficult. Although both diseases are rare events, differential diagnosis is of major importance in order to choose the most appropriate therapy (intensive induction chemotherapy vs. tyrosine kinase inhibitor followed by an early allogeneic stem cell transplant).

The first step of differential diagnosis is a thorough history and physical examination: Has the patient ever presented before with unexplained leukocytosis? Does the patient have an antecedent blood disorder or a hepatosplenomegaly?

Splenomegaly is one of the clinical hallmarks of CML. Among our 126 cases of de novo BCR-ABL+ AML, we only found 16.7 % of cases with splenomegaly (corresponding to 21/126 cases); the absence of splenomegaly was explicitly stated in 28.6 % of all cases (36/126), whereas no definitive statement was given in 54.7 % (69/126) suggesting that in most of these cases, splenomegaly was at least not a prominent clinical feature. Thus, the absence of relevant splenomegaly seems to support the diagnosis “de novo BCR-ABL+ AML” rather than “CML blast crisis.” This is underlined by a recent publication by Soupir CP et al. [4].

Another hallmark of CML is the concomitant basophilia of ≥2 % WBC. This feature was only present in 5.6 % (7/126 cases) of BCR-ABL+ AML cases. 57.2 % (4/7) of these cases with basophilia were accompanied by splenomegaly and 85.7 % (6/7) by 100 % BCR-ABL-positive metaphases on cytogenetic analysis. Thus, cases with basophilia may occur in both, de novo AML and CML-BC, but are definitely rare. Of note, some authors had already excluded questionable cases with basophilia (e.g. [1]) so that our data might underrepresent cases with basophilia. Nevertheless, it seems appropriate to conclude that the absence of basophilia supports the diagnosis of de novo BCR-ABL+ AML.

In 32.5 % (41/126) of all cases, the immunophenotype was reported. Remarkably, co-expression of lymphoid markers was reported in 34.2 % of these cases (14/41), but was not sufficient for a classification as MPAL or BAL according to WHO or EGIL respectively. This is in line with a recent finding by Atfy et al. reporting an aberrant lymphoid co-expression in seven out of nine BCR-ABL+ AML cases [80].

For the collected 126 patients, the FAB classification was reported in 79 cases. The most common subtypes were M4eo, M0, and M1, followed by M7, M4, and M2; M3 was only seen once. For further information regarding FAB and WHO classification, see Fig. 2a, b.

Interestingly, megakaryoblastic AML (FAB M7) was reported in approximately 6.3 % of AML-NOS cases in contrast to less than 1 % in a population of AML-NOS [81].

Bacher and colleagues [11] as well as Berger [5] also suggested that the presence of the Philadelphia chromosome in less than 100 % of metaphases during conventional karyotyping is the major criterion for the diagnosis of BCR-ABL+ AML, since CML-BC/AP is characterized by the presence of the Philadelphia chromosome in virtually all metaphases. This is due to the fact that CML is an early progenitor disease carrying the key cytogenetic abnormality in all cells.

Since 70 % of all cases with convincing de novo BCR-ABL+ AML in our collection express the Philadelphia-chromosome in 100 % of all metaphases, we suggest a modified use of this criterion: In the absence of basophilia, typical AML-like co-aberrations such as a complex-aberrant karyotype or a deletion of chromosome 7 may nevertheless lead to the diagnosis of AML.

In our case series, the p190 transcript was found in 27.8 % (35/126) of all cases, p210 in 30.9 % (39/126), both transcripts in 2.4 % (3/126), e6a2 in 1.6 % (2/126), and no information was given in 37.3 % (47/126) of all cases (see Fig. 3). Since p190 is very rare in CML, the presentation with a p190 transcript is in favor of the diagnosis “AML” rather than “CML.” Thus, the type of transcript can sometimes help to distinguish AML from CML as suggested [5, 11, 82].

Fig. 3
figure 3

Distribution of different BCR-ABL transcripts in de novo AML as a whole group (a) and among CBF leukemia patients (b)

Full size image

The overall prognosis of BCR-ABL+ AML is generally assumed to be unfavorable. Thus, the NCCN Clinical Practice Guidelines in Oncology for AML [83] include this entity into the poor-risk group, comparable with complex-aberrant karyotype AML. The ELN risk classification [84] does not include BCR-ABL+ AML so far. Hence, we determined the prognostic subgroup of all collected cases according to ELN criteria [84] independently of BCR-ABL: 21.4 % (27/126) were classified as favorable, 39.7 % (50/126) belonged to the intermediate-I and intermediate-II risk groups, 30.2 % of the patients (38/126) had an adverse prognosis, and for 8.7 % (11/126) of the patients, no information was available (see Fig. 2c). Therefore, the majority of cases—irrespective of BCR-ABL—already belonged to the less favorable ELN risk groups. Thus, the influence of BCR-ABL itself can hardly be determined. The same accounts for the favorable risk group, but it is at least remarkable that a considerable part of these patients showed a good response to therapy with long-term survival without allogeneic stem-cell transplantation being reported (e.g., in CBF leukemia/inv(16): 5/17 patients survived without stem cell transplantation, 4/17 patients who did not receive a SCT died, 3/17 patients who received SCT died, 3/17 patients survived after SCT, and for 2 patients, no survival data were available). In conclusion, BCR-ABL alone is not a valuable prognostic parameter so far. It seems more likely that the prognosis of BCR-ABL+ AML rather depends on the genetic background (concurrent aberrations) than on BCR-ABL itself [25].

Treatment of BCR-ABL+ AML

Currently, no standardized treatment for BCR-ABL+ AML exists. Considering the small number of cases, a randomized trial testing different treatment regimens can hardly ever be expected. Furthermore, an evaluation of the treatment response to tyrosine kinase inhibitors (TKIs) is complicated by the fact that the patients concomitantly received chemotherapy and allogeneic bone marrow transplants. In addition, many cases were published before TKIs were integrated into clinical practice. For the future, a world-wide registry would be helpful to address the efficacy of imatinib and other second- or third-generation TKIs in this rare disease.

In our case collection, 24.6 % of de novo BCR-ABL+ AML patients received a therapy with a TKI (imatinib: 29/31 TKI-treated cases; dasatinib: 4 cases; nilotinib: 1 case; some of the patients received a sequential therapy with more than one TKI). Only in two patients, a severe adverse event was reported leading to a withdrawal of the TKI (hepatotoxicity and diarrhea CTC grade 3 upon imatinib treatment [85]).

In relapsed BCR-ABL+ AML, seven patients received a tyrosine kinase inhibitor (imatinib: five cases, dasatinib: one case and nilotinib: two cases; some of the patients received a sequential therapy with two different TKIs). One patient achieved a CCyR, but died of a brain abscess, two patients had a transient response [73, 76], and two patients had no response [66, 75]. Interestingly, in three patients, the TKI treatment led to an eradication of the Ph-positive clone despite an overall refractory disease [65, 74, 77]. This clearly shows the clonal diversity of BCR-ABL+ AML in contrast to chronic phase CML, which is strictly addicted to BCR-ABL. In summary, therapy with a TKI alone does not induce sustained responses in BCR-ABL+ AML, although such a response was occasionally reported [e.g., 86]. This may be due to a very rapid clonal evolution, leading to resistance in a much higher proportion of patients and in a substantially shorter time than in CML. Furthermore, in contrast to CML, BCR-ABL is not supposed to be the key driver of the disease, but nevertheless might temporarily provide a proliferative advantage to a particular BCR-ABL+ subclone. However, as most reports show the eradication of the BCR-ABL+ subclone in AML, TKI therapy alone does not seem to control the disease, possibly owing to the fact that BCR-ABL is not a driving mutation in the founder clone. These aspects emphasize the importance of a differentiation between BCR-ABL+ AML and CML blast crisis: Whereas TKI therapy without chemotherapy is at least an option for remission induction in CML blast crisis before allogeneic stem cell transplantation (SCT), intensive induction chemotherapy remains the cornerstone of curative AML therapy and is usually followed by consolidation with chemotherapy or allogeneic stem cell transplantation according to ELN risk stratification, age, comorbidities, performance status, and donor availability [9, 87, 88]. Furthermore, allogeneic SCT is the standard of care for CML blast crisis after remission induction in eligible patients, whereas in the light of our current study, the need for SCT can at least be discussed in all cases of BCR-ABL+ AML belonging to the favorable ELN risk group.

Although the addition of a TKI to standard chemotherapy for BCR-ABL+ AML is reasonable and probably safe, we believe that there are no sufficient clinical data in support of a routine use in addition to chemotherapy during first-line therapy. The use of a TKI alone or in combination with chemotherapy in relapsed AML patients remains an individual decision, but it should not defer standard salvage therapy including allogeneic stem cell transplantation in eligible patients. Treatment of BCR-ABL+ AML is also reviewed elsewhere [4, 34, 85].

Conclusions and discussion

During the past years, an increasing number of data has accumulated which argues in favor of the existence of BCR-ABL+ AML. Therefore, BCR-ABL+ AML is now included as a new provisional entity in the 2016 revised WHO classification of hematopoietic malignancies [6]. In AML, BCR-ABL seems to cooperate with several AML-specific aberrations such as inv(16) and myelodysplasia-related cytogenetic aberrations; however, the molecular mechanism of both disease initiation and cooperation with other aberrations remains largely unknown. A better molecular characterization of BCR-ABL+ AML will always be hampered by its very low incidence. Furthermore, routine cytogenetic testing might lead to an underestimation of its real incidence since less dominant small BCR-ABL+ subclones could be missed and PCR for BCR-ABL is not routinely performed in AML. Thus, a BCR-ABL rearrangement might not be found upon routine analysis.

In clinical routine diagnostics, the distinction between de novo BCR-ABL+ AML and CML in blast crisis remains challenging in every single case. Therefore, we have generated the following algorithm for primary differential diagnosis which relies on a thorough analysis of all published cases (see Fig. 4).

Fig. 4
figure 4

Clinical path for the differential diagnosis of BCR-ABL + AML and CML blast crisis

Full size image

Briefly, after exclusion of acute leukemia of ambiguous lineage by flow cytometry (which constitutes a separate entity according to WHO), the patient’s medical history with regard to an antecedent hematologic disease is of major importance. In particular, unexplained leukocytosis and/or splenomegaly rather point towards the diagnosis of CML blast crisis than AML. In contrast, prior signs of MDS in peripheral blood or bone marrow may support the diagnosis of (secondary) AML. Furthermore, basophilia or significant splenomegaly at diagnosis strongly diminishes the probability of de novo AML. Detection of p190-transcript and occurrence of any BCR-ABL transcript in less than 100 % of metaphases supports the diagnosis of AML rather than CML. Finally, persistent CCyR after conventional chemotherapy is extremely unusual for CML in blast crisis and supports the diagnosis of AML.

In summary, the following basic requirements for diagnosis of BCR-ABL+ AML are recommended:

  1. 1)

    No antecedent hematological anomaly

  2. 2)

    No basophilia upon disease onset

Our proposed diagnostic algorithm is depicted in Fig. 4.

With regard to therapy of BCR-ABL+ AML, we feel that TKI therapy cannot be routinely recommended as a part of first-line therapy. However, TKI as a part of salvage therapy is a reasonable approach, although any prognostic benefit remains unproven. Furthermore, it is not clear whether to use imatinib, dasatinib, or nilotinib. For bosutinib and ponatinib, no data are available. If indicated, we would choose the most appropriate TKI by taking into account the patient’s comorbidities in analogy to CML treatment. Whether the TKI should be used concomitantly with salvage chemotherapy, after chemotherapy or as a maintenance therapy or a “bridging” to transplant after a remission has been achieved, cannot be answered so far. In our clinical practice, we would use a TKI primarily in relapsed BCR-ABL+ AML, starting after a standard salvage regimen has been administered. However, possible benefits have always to be weighed against the risk of clonal selection in favor of a BCR-ABL-negative subclone. In case of BCR-ABL+ MPAL, we would prefer a therapy protocol in analogy to BCR-ABL+ ALL since these protocols are well established [89] and there is no clear evidence supporting the use of AML protocols in this particular situation.

In prospect of the recently revised WHO classification, we are looking forward to emerging new aspects on this interesting new entity, both in basic research and clinical management.