Acute leukaemias of ambiguous lineage in children: characterization, prognosis and therapy recommendations

Authors


Prof Dr med. Dirk Reinhardt, Hannover Medical School, Department of Pediatric Hematology/Oncology, Carl-Neuberg-Strasse 1, D-30625 Hannover, Germany. E-mail: Reinhardt.Dirk@mh-hannover.de

Summary

Acute leukaemias of ambiguous lineage (ALAL) represent a rare type of leukaemia, expressing both myeloid and lymphoid markers. This study retrospectively analyzed data from 92 children (biphenotypic n = 78, bilineal n = 6, lineage switch n = 8) with ALAL registered in the Berlin-Frankfürt-Münster (BFM) acute myeloid leukaemia (AML) and acute lymphoblastic leukaemia (ALL) studies between 1998 and 2006 (2·4% of all cases with acute leukaemia). Our cohort of ALAL patients was characterized by comparatively high median age (8·9 years), high median white blood cell count (14·9 × 109/l), as well as frequent hyperleucocytosis (18·5%) and central nervous system involvement (24·1%). The most frequent cytogenetic abnormalities were ETV6/RUNX1 fusion (16%) and trisomy 8 (14·6%). Complete remission rate was significantly lower than in ALL-BFM patients (91·8% vs. 99·1%, P < 0·001), but comparable to AML-BFM patients (87·9%). Event-free survival (EFS) and overall survival (OS) of ALAL patients were low, at 62 ± 5%. 5-year probability of EFS was significantly worse than in ALL patients (80 ± 1%, P < 0·001), but better than for AML patients (49 ± 2%, P = 0·027). Our data suggest that an intensive therapy regimen including stem cell transplantation may be favourable for bilineal or lineage switch cases, whereas patients with ETV6/RUNX1 fusion, lymphoid morphology and patients with expression of cyCD22 and cyCD79a should be treated with an ALL-directed therapy.

Most acute leukaemias are classified as lymphoid or myeloid using morphological and cytochemical techniques as well as immunophenotyping demonstrating myeloid and/or lymphoid differentiation antigens. Acute biphenotypic leukaemia, by contrast, is a rare disease entity characterized by leukemic blasts expressing both myeloid and lymphoid markers as determined by immunophenotyping (Bene et al, 1995). The term was defined by the European Group for the Immunological classification of Leukaemias (EGIL) and is based on the number and respective degree of specificity of myeloid/lymphoid antigens expressed by the leukemic blasts (Bene et al, 1995). These criteria greatly reduced the number of biphenotypic leukaemias diagnosed, as cases with aberrant expression of only a few markers from another lineage (i.e. Myeloid + acute lymphoblastic leukaemia (ALL) and Lymphoid + acute myeloid leukaemia (AML) (Sobol et al, 1987) were excluded. The 2001 World Health Organization (WHO) classification established the term ‘acute leukaemia of ambiguous lineage’ including bilineal acute leukaemia, biphenotypic acute leukaemia and undifferentiated acute leukaemia (Brunning et al, 2001). More recently, the 2008 WHO classification introduced the term ‘mixed phenotype acute leukaemia’, which encompassed both bilineal and biphenotypic acute leukaemia (Swerdlow et al, 2008). In addition, new immunophenotypic requirements for assigning more than one lineage to a single blast population were proposed (Swerdlow et al, 2008). Overall, acute leukaemias of ambiguous lineage are rare subtypes of acute leukaemias and account for less than 4% of all cases (Ludwig et al, 2003).

Previous difficulties in classification were associated with problems in defining an optimal therapy. Hitherto, the treatment of childhood ALAL cases partially followed an ALL-directed therapy including steroids, asparaginase and methotrexate and partially included AML-directed cytotoxic drugs, such as anthracyclines and cytarabine (Mirro et al, 1985; Pui et al, 1991). While a concordantly unfavourable prognosis for adult ALAL patients has been described in the majority of studies (Matutes et al, 1997; Legrand et al, 1998; Killick et al, 1999), conflicting results were reported for childhood cases with expression of antigens of a different lineage: some studies identified an adverse prognosis for these patients (Schmitt-Graff et al, 1988; Wiersma et al, 1991), whereas others found outcomes similar to the other ALL and AML patients (Mirro et al, 1985; Pui et al, 1990, 1991; Killick et al, 1999). Killick et al (1999) reported a higher survival rate for childhood than for adult patients. So far, only one study examined specifically childhood ALAL patients according to the 2001 WHO criteria (Brunning et al, 2001) and proposed an initial AML-based therapy for all patients (Rubnitz et al, 2009). As ALAL appears to be a heterogeneous group, it seems reasonable to establish treatment stratification criteria for specific subgroups rather than proposing a general treatment recommendation (Altman, 1990). However, clear prognostic factors and treatment stratifications have remained elusive.

To address these issues, we retrospectively analyzed clinical, biological and therapeutic data from a total of 92 children with de novo ALAL enrolled in the Berlin-Frankfürt-Münster (BFM) ALL and AML studies between January 1998 and March 2006. Our study defined new pathological predictors of outcome and gives treatment recommendations, contributing to the stratified treatment of children with ALAL.

Patients, materials and methods

Patient population

Between January 1998 and March 2006, 92 children enrolled in the AML-BFM and ALL-BFM studies were diagnosed with de novo ALAL according to the 2001 WHO classification (Brunning et al, 2001). One patient with de novo ALAL suffered from ataxia telangiectasia and was excluded from the study, as chemotherapy could not be performed according to standard protocols due to his disease. After confirmation of diagnosis, parental or patient informed consent for data registration and follow-up was obtained in all patients according to local laws and regulations. All investigations had been approved by the Institutional Ethics Committees.

Immunophenotyping

Diagnosis of ALAL was made by immunophenotyping using a wide panel of monoclonal antibodies against myeloid- and lymphoid-associated antigens, as proposed by EGIL (Bene et al, 1995). Such extensive immunophenotyping was done on almost all patients registered in the BFM studies during our study period. Flow cytometry was performed according to standard protocols, as described earlier (Langebrake et al, 2006; Rhein et al, 2007).In this study, biphenotypic leukaemia was defined according to the EGIL definition, when blasts scored over 2 points for the myeloid and one of the lymphoid lineages (Bene et al, 1995; EGIL 1998). The diagnosis of bilineal acute leukaemia was applied to leukaemias containing separate populations of blasts of more than one lineage based on immunophenotyping. Patients were diagnosed with lineage switch leukaemia when blasts showed morphological and immunophenotypic features of a different lineage during the course of therapy as compared to initial diagnosis. While the majority of ALAL cases were identified at initial diagnosis, we scanned our immunophenotypic datatbase for ALAL criteria and included a few additional formerly mis-diagnosed patients in our retrospective study.

Analysis of genetic aberrations

Cytogenetic and molecular genetic analyses were mainly carried out in the BFM reference laboratory at the University Hospital Giessen and partly in local laboratories following standard protocols. Karyotypes were classified according to the International System of Chromosome Nomenclature (ISCN 2005).

Treatment

Nineteen patients were treated according to AML protocols (AML-BFM 98 and 2004)(Creutzig et al, 2005), 46 according to ALL protocols (ALL-BFM 95 and 2000)(Moricke et al, 2008) and 27 patients received a combination of AML and ALL protocols. Thirty-three patients (35·9%) received an allogeneic haematopoetic stem cell transplantation (HSCT).

Statistical methods

Event-free survival (EFS) was calculated from date of diagnosis to last follow-up or to the first event [early death (ED) within 6 weeks of diagnosis, resistant leukaemia (non-response, NR), relapse, secondary malignancy, or death of any cause]. Failure to achieve remission (ED + NR) was considered as an event on day 0. Overall survival (OS) was defined as the time from diagnosis to death from any cause. Survival times for patients without event were evaluated considering the date of last follow-up. Rates were calculated according to Kaplan-Meier and compared by log-rank test. Standard errors of the estimates were obtained using the Greenwoods formula. Cumulative incidence functions of relapse in continuous CR (CCR) were constructed by the method of Kalbfleish and Prentice and the functions were compared with the Gray test. Tests were two-sided with 0·05 significance level. Analyses were carried out using sas (Statistical Analysis System Version 9.1; SAS Institute, Cary, NC, USA).

Results

Clinical data

From January 1998 to March 2006, a total of 3,849 patients were enrolled and treated in the AML-BFM and ALL-BFM studies. Ninety-two patients (=2·4%) were diagnosed with ALAL according to the 2001 WHO classification and EGIL definition (see Table SI for patients’ characteristics). Sex distribution for these patients was similar in the AML and ALL studies (male n = 50, 54·3% vs. 54·7%, female n = 42, 45·7% vs. 45·3%). Clinical data of ALAL patients at presentation compared to patients in the ALL-BFM 95 and AML-BFM 98 study are shown in Table I. The median age of ALAL patients at diagnosis as well as percentage of patients with initial hyperleucocytosis (≥100 × 109/l) was significantly higher compared to ALL patients. Initial CNS involvement was found in 24·1% of ALAL patients, which was significantly higher than in both AML and ALL patients.

Table I.   Initial clinical presentation of ALAL patients compared to AML-BFM 98 and ALL-BFM 95 patients.
 ALALAML-BFM 98ALL-BFM 95
  1. P values in comparison to ALAL patients, respectively. CNS, central nervous system.

Median age (years)
range
 8·9
 0·08–17·7
 9·2, P = 0·835·0, P < 0·001
Median leucocyte count (×109/l)
range
14·9
 0·63–964·3
17·5, P = 0·43
0·4–711
102, P = 0·13
10·0–1000
Hyperleucocytosis (%)18·520·1, P = 0·7810·9, P = 0·04
Hepatosplenomegaly (%)65·056·2, P = 0·3275·0, P = 0·04
CNS involvement (%)24·111·5, P = 0·02 3·2, P < 0·001

Morphology

Morphological assessment at initial diagnosis showed mainly lymphoid features in 50 and mainly myeloid features in 10 patients. In those cases with myeloid features, the most common FAB subtypes were M0 (n = 3) and M1 (n = 3), respectively. Fifteen cases showed an undifferentiated or unclassifiable morphology, and the remaining 17 patients presented with dimorphic blast populations, one resembling lymphoblasts and the other myelo- or monoblasts. Four out of six patients with immunophenotypically bilineal acute leukaemia also showed two morphologically different blast populations. However, 10 immunophenotypically biphenotypic cases and 3 lineage switch cases also presented with two different populations that differed by morphology.

Immunophenotyping

Seventy-eight of the 92 cases with ALAL (85%) were consistent with the EGIL criteria for biphenotypic acute leukaemia (2·0% of all AML and ALL cases). Of these, 45 showed a B-lymphoid and myeloid phenotype (57·7%), 27 a T-lymphoid and myeloid (34·6%) and six cases a trilineage phenotype (7·7%), expressing myeloid-, B-, and T-cell-associated antigens. Six cases (6·5%) were diagnosed as bilineal leukaemia (B-lymphoid and myeloid: n = 4, T-lymphoid and myeloid: n = 2), representing 0·2% of all ALL and AML patients. Eight cases (8·7%) showed a lineage switch during treatment, seven cases switched their immunophenotype from B-lymphoid to myeloid, one case from T-lymphoid to myeloid. For six of eight cases, the lineage switch occurred during the first 15 d of treatment. None of these cases were acute undifferentiated leukaemia based on immunophenotypic features.

Cytogenetic and molecular genetic data

Cytogenetic data were available in 52 patients, whereas for 40 patients cytogenetic analysis had not been performed or failed. In the latter patients, molecular genetic screening was accomplished by polymerase chain reaction (PCR) and/or fluorescence in situ hybridization (FISH). Thirty-nine were analyzed for the fusion genes BCR/ABL1 and ETV6/RUNX1, 38 for MLL/AFF1, and in a minority of cases the appearance of other rearrangements (MLL/MLLT3, MLL/MLLT1, RUNX1/RUNX1T1, CBFB/MYH11) was tested. While most of these analyses revealed normal results, 11 additional patients showed a ETV6-RUNX1 fusion, and in two others a MLL rearrangement was detected. Cytogenetic and moleculargenetic results of ALAL patients are listed in Table II. No typical abnormality differentiating ALAL from AML or ALL cases was identified. The most frequent chromosomal abnormality detected in the ALAL cohort was a ETV6/RUNX1 fusion (16·0%), followed by trisomy 8 (13·5%) and 11q23 rearrangements (10·9%). For four out five patients with MLL/11q23 rearrangement, reverse transcription (RT)-PCR analysis was performed and showed MLLT1 (n = 2), MLLT3 (n = 1) and AFF1 (n = 1), respectively, as translocation partners of MLL.

Table II.   Cytogenetic and molecular results of ALAL patients compared to AML patients.
 ALALAML-BFM 98
  1. FISH, fluorescence in situ hybridization; RT-PCR, reverse transcription polymerase chain reaction.

Normal karyotype19·2% (10/52)22%, P = 0·73
ETV6/RUNX1 fusion (detected by RT-PCR)16% (13/81)N.A.
Trisomy 813·5% (7/52)6·6%, P = 0·09
11q23 rearrangements (detected by cytogenetics and FISH analysis)10·9% (5/46)21·1%, P = 0·12
t(9;22)(q34;q11) (detected by cytogenetics and RT-PCR)4·4% (4/90)0·2%, P = 0·003
Complex aberrant karyotype11·5% (6/52)7·9%, P = 0·42

Compared to non-ALAL patients enrolled in the AML-BFM 98 study, translocation t(9;22)(q34;q11) was significantly more frequent in patients with ALAL (0·2% vs. 4·4%, P = 0·003), while trisomy of chromosome 8 tended to be more frequent in ALAL patients (6·6% vs. 13·5%, P = 0·09). For most of the patients in the ALL-BFM-95 study cytogenetic analysis was not done, but a molecular genetic screening for the BCR/ABL1, MLL/AFF1 and ETV6/RUNX1 rearrangements was performed. The percentage of the BCR/ABL1 rearrangement tended to be higher in ALAL patients than ALL patients (4·4% vs. 2·1%, P = 0·14), while ETV6/RUNX1 fusion occurred at a similar frequency (16·5% vs. 21·4%, P = 0·32).

Treatment and outcome

Results from our investigations were compared to AML and ALL patients treated in the AML-BFM 98 (Creutzig et al, 2006) and ALL-BFM 95 (Moricke et al, 2008) studies. Of 92 patients with ALAL, 91·8% achieved complete remission (CR), which is comparable to children with AML (87·9%, P = 0·36), but significantly lower than in children with ALL (99·1%, P < 0·001). The estimated probability of EFS at 5 years (5-year-pEFS) and the estimated probability of OS at 5 years (5-year-pOS) for ALAL patients were both 62 ± 5%. EFS in the ALAL cohort was significantly higher than that of AML patients (5-year-pEFS 49±2%, P = 0·027), while OS appeared to be similar (AML: 63 ± 2%, P = 0·86). Both EFS and OS were significantly higher in the ALL study (5-year-pEFS 80 ± 1%, 5-year-pOS 88 ± 1%, P < 0·001) (Fig 1A, B). Detailed analysis of ALL and AML patients in different risk groups revealed different results: EFS in ALAL patients was comparable to that of standard risk (SR)-AML patients (5-year-pEFS 63 ± 4%, P = 0·89) and high risk (HR)-ALL patients (5-year-pEFS 51 ± 3%, P = 0·15) but significantly better than in patients with HR-AML (5-year-pEFS 40 ± 3%, P = 0·0005) and significantly worse than in patients with medium risk (MR)-ALL (5-year-pEFS 80 ± 1%, P < 0·001) and SR-ALL cases (5-year-pEFS 90 ± 1%, P < 0·001) (Fig 1C, D).

Figure 1.

 Prognosis of patients with ALAL compared to AML and ALL patients. 5-year-pEFS (A) and 5-year-pOS (B) compared to children treated in the AML-BFM 98 and ALL-BFM 95 studies. 5-year-pEFS in comparison to AML-BFM 98 (C) and ALL- BFM 95 (D) subgroups. SE, standard error.

The 5-year EFS in patients treated with AML-directed therapy alone (n = 19, 5-year-pEFS 41 ± 12%) was significantly worse than for patients who received an ALL-based therapy alone (n = 46, 5-year-pEFS 81 ± 6%, P = 0·0009).

Cumulative incidence functions of relapse revealed a higher probability for AML patients to relapse compared to ALAL patients (33 ± 2% vs. 21 ± 4%, P = 0·02), while ALL patients tended to relapse less frequently than ALAL patients (16 ± 1%, P = 0·09).

Within the group of patients with biphenotypic or bilineal acute leukaemias, five of seven patients who failed to respond to initial ALL therapy achieved CR after having changed to AML therapy. Similarly, four patients who did not respond to initial AML therapy entered CR with ALL therapy.

Immunological subgroups.  Patients diagnosed with bipheno-typic leukaemia (n = 78) had a 5-year-pEFS of 67 ± 6%. Outcome tended to be worse for patients with lineage switch (n = 8, 5-year-pEFS 38 ± 17%, P = 0·08) and bilineal acute leukaemia (n = 6, 5-year-pEFS 33 ± 19%, P = 0·07) (Fig 2A).

Figure 2.

 Event free survival of children with ALAL. Depending on immunophenotypic (A and B), morphological (C) and cytogenetic (D) subgroups. SE, standard error.

For detailed information on the clinical course of patients with bilineal and lineage switch leukaemia see Figure S1A, B. Of six patients with bilineal leukaemia, only two survived, both after HSCT. Outcome for the subgroups of biphenotypic patients showed that patients with a B-lymphoid/myeloid phenotype (n = 45) tended to have a higher 5-year-pEFS than patients with a T-lymphoid/myeloid phenotype (n = 27, 74 ± 7% vs. 53 ± 10%, P = 0·06) while outcome for patients with trilineage phenotype was similar to the two other groups (n = 6, 67 ± 19%, P = 0·55 and 0·73, respectively) (Fig 2B).

Morphological subgroups.  Patients presenting with a blast population resembling lymphoblasts at initial diagnosis had a significantly better prognosis than patients with myeloid, undifferentiated or dimorphic blast populations (5-year-pEFS 79 ± 6% vs. 46 ± 17%, P = 0·015 vs. 39 ± 9%, P < 0·0001, Fig 2C). A subgroup of patients with blasts resembling lymphoblasts who were treated with ALL-directed therapy only (n = 42) had a significantly better 5-year-pEFS than the remainder patients (84 ± 6% vs. 42 ± 7%, P < 0·0001).

Cytogenetic subgroups.  Regarding the cytogenetic aberrations, trisomy 8 cases tended to have a favourable prognosis with a 5-year-pEFS of 86 ± 13% compared to patients without trisomy 8 (5-year-pEFS 47 ± 8%, P = 0·10). Patients with a ETV6/RUNX1 fusion showed a significantly better outcome than patients who were negative for this aberration (5-year-pEFS 100% vs. 57 ± 6%, P = 0·0084, Fig 2D), all of them achieved long-term complete remission with an ALL-directed therapy only. ALAL patients with t(9;22) or 11q23 rearrangement had an unfavourable prognosis (see Figure S1C, D for clinical details): Of four patients with t(9;22), two are alive: one after ALL-based therapy followed by HSCT, one after combination of AML and ALL therapy combined with imatinib followed by HSCT. Of five patients with 11q23 rearrangement, only one patient is still alive after good response to ALL-based therapy.

While myeloperoxidase (MPO) is the only myeloid marker that is assigned 2 points according to the EGIL system, expression of markers cytCD79a, cytCD22 and cytCD3 is considered most specific for the B or T lymphoid lineage, respectively (Brunning et al, 2001).

Seventy-eight patients expressed one of the specific B-lymphatic markers (cyCD22 or cyCD79a) at presentation, 35 of these patients were treated with an ALL-directed therapy only and showed a significantly better outcome compared to the those treated with AML-directed or combined therapy (5-year-pEFS 91 ± 5% vs. 39 ± 10%, P < 0·001). Thirty-three patients expressed cyCD3 as a specific T-lymphatic marker. For this group, outcome was comparable for patients receiving ALL-directed therapy (n = 14) and the remaining patients (n = 19) (5-year-pEFS 56 ± 13% vs. 42 ± 12%, P = 0·86). EFS for patients who were MPO-positive (n = 63, based on cytochemistry and/or immunophenotype) was better than for those who were MPO-negative (n = 29) (5-year-pEFS 69 ± 6% vs. 45 ± 10%, P = 0·029).

A good response to prednisone, defined as fewer than 1 × 109 blasts/l in the peripheral blood on day 8 of ALL therapy (Moricke et al, 2008), was a favourable prognostic factor [5-year-pEFS 91 ± 5% (n = 37) vs. 28 ± 16% (n = 11), P < 0·001].

Occurrence of hyperleucocytosis, hepatosplenomegaly or CNS involvement had no statistically significant effect on OS or EFS. Patients >2 years of age (n = 82) tended to have a better outcome compared to patients <2 years (n = 10) (65 ± 5% vs. 40 ± 15%, P = 0·069).

Discussion

ALAL, i.e. biphenotypic, bilineal and acute undifferentiated leukaemia according to the 2001 WHO classification, is a rare subgroup of acute leukaemias. The overall incidence of biphenotypic leukaemia was estimated to be 2–5% in studies that used the EGIL scoring system (Carbonell et al, 1996; Killick et al, 1999; Thalhammer-Scherrer et al, 2002; Ludwig et al, 2003; Owaidah et al, 2006). All these studies were based on combined groups of adults and children or adults only. More recently, Rubnitz et al (2009) published a study of 35 childhood cases with acute mixed lineage leukaemia and found an incidence of 1·8% in their patients. Our study reports on 92 children suffering from ALAL, thus representing the largest cohort of children to date. Interestingly, regarding clinical features, CNS involvement was significantly higher for ALAL than for both AML and ALL patients. At least for ALL patients, this difference can be partly explained by a significantly higher occurrence of hyperleucocytosis.

The diagnosis of ALAL is based upon immunophenotypic analysis. In our study, co-expression of B-lymphoid and myeloid antigens on leukemic blasts was found in 45 of 78 biphenotypic patients (57·7%) and of T-lymphoid and myeloid antigens in 34·6% of cases (n = 27). In line with our results, most previous studies reported the B-lymphoid/myeloid subtype to be more frequent (60–70%) than the T-lymphoid/myeloid subtype (26–30%) (Legrand et al, 1998; Killick et al, 1999; Tiribelli et al, 2004; Owaidah et al, 2006). In contrast, Rubnitz et al (2009) observed a higher frequency of biphenotypic acute leukaemias with expression of T and myeloid markers (62·5%) compared to cases with B and myeloid marker expression (31·3%) in children with ALAL. While Rubnitz et al (2009) reported that the biphenotypic subtype does not appear to have an influence on outcome, we found a significantly higher EFS for patients with B-lymphoid/myeloid expression compared to patients of the T-lymphoid/myeloid subtype.

There is no known single chromosomal aberration uniquely associated with ALAL (Carbonell et al, 1996; Rubnitz et al, 2009). However, chromosomal abnormalities appear to be common, with only 19·2% of patients exhibiting a normal karyotype, which is lower than for adult ALAL patients (32–35%) (Legrand et al, 1998; Owaidah et al, 2006). Interestingly, the most frequent aberrations in our study were ETV6/RUNX1 fusion (16·0%) and trisomy 8 (13·5%), which have not been reported before to be common in ALAL. Patients with ETV6/RUNX1 fusion had a significantly better outcome than the remainder while patients with trisomy 8 tended to have a better prognosis than patients without this aberration. Similar to results from earlier studies, our study also found a high incidence of Philadelphia chromosome and MLL rearrangements (4·4 and 10·9%, respectively) (Ludwig et al, 1989; Carbonell et al, 1996; Legrand et al, 1998; Killick et al, 1999; Owaidah et al, 2006). The occurrence of t(9;22) or t(4;11) is considered an adverse prognostic factor in ALL patients (Moricke et al, 2008). In patients with biphenotypic acute leukaemia, the presence of Philadelphia chromosome has also been associated with a poor prognosis (Killick et al, 1999). These findings were confirmed in our study. Only one of five patients with 11q23 rearrangement and two of four patients with Philadelphia translocation, both having received an allogeneic HSCT (one after treatment with imatinib in addition to ALL and AML based therapy), have survived. Taking into account the observed adverse outcome of patients with Philadelphia chromosome or MLL rearrangement, ALAL patients with either of these translocations should receive an intense chemotherapy. Furthermore, an allogeneic HSCT in first CR should be considered.

The pathogenetic events underlying the development of bilineal leukaemia have remained controversial. It is still unclear whether it is a distinct disease entity or if bilineal cases represent biphenotypic leukaemia with a greater degree of maturation (Catovsky et al, 1991). Our observation that only four of six bilineal cases, and also 13 cases with biphenotypic and lineage switch leukaemia, presented with two morphological distinct population indicates a close relationship between the subtypes. In the 2008 WHO classification, the term mixed phenotype acute leukaemia encompasses both bilineal and biphenotypic cases (Swerdlow et al, 2008). Weir et al (2007) recently reported 19 cases with bilineal leukaemia, including eight children, representing 1·3% of their leukaemia cohort. Incidence of bilineal leukaemia in our paediatric cohort was lower, at 0·2% of all ALL and AML patients. Previous reports associated bilineal leukaemia with a poor outcome (Ludwig et al, 1988; Schmitt-Graff et al, 1988). Consistent with these data, the EFS of our patients was poor (33 ± 19%). Only two of the patients are still alive, both of them underwent an allogeneic HSCT, similar to the findings reported by Weir et al (2007). Although treatment experience with bilineal leukaemia is limited due to small case numbers, a HSCT in CR should thus be considered in all cases of bilineal leukaemia.

Similar to bilineal cases, a consistently adverse prognosis for patients with lineage switch leukaemia has been described (Gagnon et al, 1989; Cuneo et al, 1994). A low EFS rate (38 ± 17%) also applied to the patients in our study. Treatment recommendations for lineage switch leukaemia are scarcely found in the literature. Stass et al (1984) reported an achievement of remission with treatment adjusted to the phenotype evident at lineage switch after initial lineage-specific chemotherapy. Only three of our eight lineage switch patients, all carrying a trisomy 8, are still alive (Figure S1B). All were switched to AML therapy after initial ALL-based therapy, and two remain in CR following HSCT. Based on these findings, we suggest initial treatment according to phenotype, i.e. ALL-directed therapy for lymphoid, AML-directed therapy for myeloid phenotype. Following the lineage switch patients should be treated with the other regimen or a combination of both ALL and AML therapy. Taking into account the poor outcome of lineage switch cases, a HSCT in first remission should be considered.

Treatment recommendations based on our data were difficult as our subgroups were small and thus statistically significant results insufficient to yield. Moreover, as our study was retrospective and patients were not stratified to the different treatment options, subgroups treated with different regimens were not homogeneous. For example, EFS for patients treated with ALL-directed therapy alone was better than for patients treated with AML-directed therapy. Yet, one cannot deduce from this observation that AML therapy is not a valid therapy for ALAL patients as the two groups were not homogenous.

AML-based treatment is a very intense therapy with a high risk of infectious complications. Rubnitz et al (2009) proposed an initial AML-directed therapy for all children with ALAL and a switch to ALL therapy for non-responding patients. The results of our study lead us to a different recommendation. Fifty-five of our patients (excluding lineage switch cases) initially received an ALL-based therapy. Of these, 47 patients achieved CR and 37 are still alive after treatment with ALL directed therapy only, two patients are alive after ALL therapy and HSCT. Moreover, our statistical data show that a subgroup of patients who expressed a specific B-lymphoid differentiation marker (i.e. cytCD79a and/or cytCD22) and also patients who initially showed a lymphoblastic morphology had a significantly better outcome with ALL-directed therapy compared to the remaining patients. Also, all 13 patients with ETV6/RUNX1 fusion achieved long-term complete remission with an ALL-directed therapy only (5-year-pEFS 100%). A good response to prednisone therapy on day 8 is a favourable marker regarding outcome (EFS 91 ± 5%, n = 37). Thus, in contrast to Rubnitz et al (2009), we recommend initial therapy according to the ALL protocol for these patients to prevent over-treatment.

As all of the patients in this study were treated in the BFM studies before 2008, we made the diagnosis of ALAL according to the 2001 WHO classification. One major difference is that the 2008 WHO classification accepts only MPO positivity as evidence of myeloid differentiation. While our data partly support this new definition (all 13 patients with ETV6/RUNX1 translocation did not express MPO, but showed coexpression of CD13, 33, 117 or 15 and survived with ALL-based therapy only), the observation that the total group of MPO-positive cases still had a significantly better outcome than MPO-negative patients (5-year-pEFS 69 ± 6% vs. 45 ± 10%, P = 0·029) indicates a strong influence of expression of myeloid markers other than MPO. We suggest that, in order to confirm the reasonability of the 2008 WHO classification, larger studies are needed to examine whether MPO as a sole marker is really sensitive enough to diagnose biphenotypic leukaemia.

In summary, our treatment recommendations for paediatric patients with ALAL are as follows:

  • 1 To prevent over-treatment, patients with expression of cytCD79a or cytCD22, patients with leukemic cells resembling lymphoblasts and patients with ETV6/RUNX1 fusion should be initially treated with ALL-directed therapy.
  • 2 Bilineal and lineage switch patients, as well as ALAL patients with Philadelphia chromosome or 11q23 rearrangement should receive an intensive chemotherapy, and may benefit from allogeneic HSCT.
  • 3 Lineage switch patients should be initially treated according to phenotype and the therapy changed to the respectively other treatment following the lineage switch.

Consequently, our study defines new pathological predictors of outcome and gives treatment recommendations, contributing to the stratified treatment of children with ALAL. In order to further optimize therapy for childhood ALAL patients and clarify open questions, prospective studies are necessary. Due to the rarity of this disease, international multi-centre studies are urgently required.

Acknowledgements

We thank J.-H. Klusmann and U. Bernsmann for critical reading, C. Augsburg, T. Reinke and M. Wackerhahn for technical assistance, as well as J.-E. Müller and N. Mühlegger for data management. This study was supported in part (JB) by the ‘Parents’ Foundation for Children with Cancer, Giessen’.

Ancillary