Acute megakaryoblastic leukaemia (AMKL) in children: a comparison of AMKL with and without Down syndrome

Authors


Seiji Kojima, Department of Paediatrics, Nagoya University Graduate School of Medicine, 65 Tsurumai-cho, Showa-ku, Nagoya 466–8550, Japan.
E-mail: kojimas@med.nagoya-u.ac.jp

Summary

To characterize childhood acute megakaryoblastic leukaemia (AMKL), we reviewed 45 children with AMKL diagnosed between 1986 and 2005 at Nagoya University Hospital and Japanese Red Cross Nagoya First Hospital. Twenty-four patients (53%) had AMKL associated with Down syndrome (DS-AMKL) and 21 (47%) had non-DS-AMKL. The median age of the DS-AMKL patients was 21 months (range, 8–38 months) and that of non-DS-AMKL patients was 15 months (range, 2–185 months). The morphology of blast cells was categorized into three groups according to the stage of megakaryocyte maturation. The blast cells were more immature in DS-AMKL than in non-DS-AMKL in terms of morphology and immunophenotyping. Cytogenetic abnormalities of leukaemic cells were classified into seven categories: normal karyotype including constitutional trisomy 21 in DS-AMKL; numerical abnormalities only; t(1;22)(p13;q13); 3q21q26 abnormalities; t(16;21)(p11;q22); −5/del(5q) and/or −7/del(7q); and other structural changes. The outcome of children with either DS-AMKL or non-DS-AMKL is excellent. The 10-year overall survival estimate was 79% [95% confidence interval (CI): 54–90] for DS-AMKL and 76% (95% CI: 58–91) for non-DS-AMKL (P = 0·81) with a median follow-up of 78 months (range, 20–243 months). Our study shows the diverse heterogeneity of childhood AMKL and the need for subclassification according to cytogenetic and morphological features.

Acute megakaryoblastic leukaemia (AMKL), M7 according to the French-American-British (FAB) classification (Bennett et al, 1985), is a subtype of acute myeloid leukaemia (AML) (Bennett et al, 1985). Until now, only a few series of AMKL have been reported in a consecutive cohort of children (Ribeiro et al, 1993; Athale et al, 2001; Paredes-Aguilera et al, 2003; Reinhardt et al, 2005). Large collaborative studies have shown that AMKL occurs in 4·1% to 15·3% of patients with childhood AML (Ravindranath et al, 1992; Gamis et al, 2003; Creutzig et al, 2005; Zeller et al, 2005; Rao et al, 2006). These previous studies revealed a diverse heterogeneity of the disease including morphology and cytogenetic studies. In children, two major subgroups have been described: AMKL in patients with Down syndrome (DS-AMKL) and AMKL in patients without Down syndrome (non-DS-AMKL). AMKL is the most frequent type of AML in children with DS (Zipursky et al, 1987). Somatic mutations of GATA1 are found in almost all children with DS-AMKL (Wechsler et al, 2002; Hirose et al, 2003; Rainis et al, 2003). It is well known that patients with DS-AMKL have an excellent outcome with less intensive chemotherapy. In general, the remission rate is approximately 90% with event-free survival (EFS) of 70% to 80% (Gamis, 2005). The favourable outcome in patients with DS-AMKL has been explained by the increased sensitivity of leukaemic cells to anticancer drugs (Taub et al, 1999, 2000). On the other hand, patients with non-DS-AMKL appear to be more heterogeneous and several cytogenetic groups have been identified (Lu et al, 1993; Dastugue et al, 2002; Duchayne et al, 2003). Among these groups, the occurrence of t(1;22)(p13;q13) is restricted to infants with AMKL (Carroll et al, 1991; Lion et al, 1992; Bernstein et al, 2000). The outcome of children with non-DS-AMKL is generally poor (Ribeiro et al, 1993; Athale et al, 2001), but a recent study reported long-term survivors after intensive chemotherapy (Reinhardt et al, 2005).

We reviewed 45 children with AMKL (24 DS-AMKL, 21 non-DS-AMKL) diagnosed between 1986 and 2005 at Nagoya University Hospital and Japanese Red Cross Nagoya First Hospital. The main purpose of this study was to compare clinical and biological characteristics of patients with DS-AMKL and non-DS-AMKL.

Patients and methods

Patients and diagnostic criteria of AMKL

Forty-five children with newly diagnosed AMKL (24 DS-AMKL, 21 non-DS-AMKL) at Nagoya University Hospital and Japanese Red Cross Nagoya First Hospital between 1986 and 2005 were retrospectively reviewed. Acute leukaemia was diagnosed by the presence of at least 20% blasts in the bone marrow (BM) according to the World Health Organization classification (Harris et al, 1999). In patients with poor quality BM aspiration smears, the presence of more than 20% blasts in the BM core biopsy or 20% or more circulating blasts were used to support the diagnosis of acute leukaemia. The diagnosis of AMKL was established on the basis of FAB classification (Bennett et al, 1985) by studies of cell morphology and cytochemistry and was confirmed by immunophenotyping.

For immunophenotyping of leukaemic blasts, mononuclear cells were isolated by Ficoll-Hypaque density gradient centrifugation from the BM. In patients who did not have an adequate BM aspirate, the immunological studies were performed on peripheral blood (PB) mononuclear cells. The cells were analysed by flow cytometry with a panel of monoclonal antibodies. For the assessment of megakaryocytic lineage, at least 10% of the blast cells needed to be positive for one or more of the platelet-associated antigens (CD36, CD41, CD42 or CD61) (San Miguel et al, 1988). For the other immunological markers, the samples were defined as positive if more than 20% of the cells were stained. In the absence of immunophenotyping, the diagnosis was confirmed by electron microscopic identification of platelet peroxidase (PPO) activity or CD41 expression in the histopathological examination in malignant cells.

Cytogenetic studies were performed on BM or PB samples taken at the time of diagnosis; samples were processed and analysed by standard methods.

Analysis of GATA1 mutation

After obtaining informed consent from the parents of the children for the purpose of sample banking and molecular analysis, BM or PB samples were obtained from all of the patients with AMKL at the time of diagnosis. High-molecular weight DNA was extracted from the samples using standard methods. For screening of GATA1 mutations, we amplified the genomic DNA that corresponded to exon 2 of GATA1 by using polymerase chain reaction that employed one primer pair as previously reported (Hirose et al, 2003). Amplified products were cloned into the pGEM-T Easy vector (Promega, Madison, WI, USA) and sequenced on a DNA sequencer (310; Applied Biosystems, Foster City, CA, USA) using a BigDye terminator cycle sequencing kit (Applied Biosystems).

Treatment

In patients with DS-AMKL, six patients received chemotherapy consisting of cytosine arabinoside (AraC) and dounorubicin (Kojima et al, 1990, 1993) and 16 patients received chemotherapy consisting of AraC, etoposide, and dounorubicin or pirarubicin (Kojima et al, 2000). One patient with mosaic DS received intensive chemotherapy according to non-DS-AML protocol containing high-dose AraC (AML99) (Shimada et al, 2006) and one patient received chemotherapy according to an acute lymphoblastic leukaemia oriented protocol (Kojima et al, 1990). One patient who failed to achieve complete remission (CR) by induction therapy underwent bone marrow transplantation (BMT) from a human leucocyte antigen (HLA)-matched unrelated donor.

In the patients with non-DS-AMKL, 18 patients were treated in one of two national co-operative studies for AML (12 on ANLL91, six on AML99). Two patients received less intensive chemotherapy according to DS-AMKL protocol (Kojima et al, 1993). Twelve patients received BMT (autologous, seven; allogeneic, five) in the first CR. Three patients who failed to achieve CR after induction therapy received allogeneic stem cell transplant (SCT: BM, two; cord blood, one). Two patients who relapsed after allogeneic BMT in the first CR (one from a matched sibling, the other from a matched unrelated donor) received a second allogeneic BMT (one from a matched unrelated donor, the other from a haploidentical related donor). The indication for SCT varied over the study period and was defined by individual protocol. In the ANLL91 study, allogeneic BMT was indicated in the first CR for patients who had a matched sibling donor; patients without a matched sibling donor were eligible for autologous BMT. AML99 study did not indicate SCT for patients with AMKL in the first CR without poor prognostic chromosome abnormalities such as monosomy 7 or t(16;21). Conditioning regimens included busulfan and melphalan; fludarabine was added in the unrelated donor settings. Prophylaxis against graft-versus-host disease consisted of ciclosporin and short-term methotrexate therapy in BMT from a matched related donor and tacrolimus and short-term methotrexate in SCT from an unrelated donor or a mismatched related donor.

Statistical analysis

Non-parametric Mann–Whitney test was used to analyse statistical differences in the distribution of continuous variables. χ2 test or Fisher’s exact test was used for differences in frequencies. Survival distributions were estimated by using the method of Kaplan and Meier (1958) and were compared using the log-rank test. All estimates of outcomes are reported with 95% confidence intervals (CI). The duration of overall survival (OS) was defined as the period between the date of diagnosis and the date of death from any cause or the date of the last follow-up examination. The duration of EFS was defined as the period between the date of diagnosis and the date of an adverse event (relapse, death from any cause) or the most recent follow-up examination. Early death or remission induction failure was recorded as an event at zero time, with an EFS value of zero. SAS release 6.12 software (SAS Institute, Cary, NC, USA) was used to perform the statistical analysis. P values <0·05 were considered significant.

Results

Patients

During the period of study, 194 children with AML were diagnosed in two hospitals; 45 (23·2%) of these had AMKL. Twenty-four patients (53·3%) had DS-AMKL and 21 (46·7%) had non-DS-AMKL. Among them, 11 patients with DS-AMKL had been previously reported (Kojima et al, 1990, 1993, 2000). Twenty-seven children with DS-AML/myelodysplastic syndrome (MDS) were diagnosed during the same period of study. Among them, one patient was diagnosed as AML (M0) and two were diagnosed as MDS. One of the patients with DS-AMKL was trisomy 21 mosaicism. In the patients with DS-AMKL, eight patients (33·3%) had a history of transient myeloproliferative disorder and 12 (50·0%) had prior MDS. Congenital heart anomalies were present in 11 patients (45·8%) with DS-AMKL. None of the patients with AMKL had secondary leukaemia or mediastinal germ cell tumour. The median age of DS-AMKL patient was 21 months (range, 8–38 months) and that of non-DS-AMKL was 15 months (range, 2–185 months). Patients younger than 4 years at diagnosis accounted for 95·6% (43 of 45) in both groups.

Clinical and laboratory features

Table I shows the clinical and laboratory findings at the time of diagnosis. There were no statistically significant differences in these findings between DS-AMKL and non-DS-AMKL patients. The initial leucocyte count, BM blast cell count, and lactic dehydrogenase activity tended to be higher in patients with non-DS-AMKL than in those with DS-AMKL, but differences were not significant. Nine patients (DS-AMKL, eight; non-DS-AMKL, one) had less than 20% blasts in the BM; the BM of all nine patients was difficult to aspirate, contributing to the low estimates of percentage of blasts. Six patients (DS-AMKL, five; non-DS-AMKL, one) underwent BM biopsy, which confirmed the presence of more than 20% blast cells and the diagnosis of acute leukaemia. Three DS-AMKL patients had more than 20% blasts in the PB; which supported the diagnosis of acute leukaemia.

Table I.   Clinical and laboratory findings at the time of diagnosis.
FeatureDS-AMKL (n = 24)non-DS-AMKL (n = 21)P
  1. DS-AMKL, Down syndrome-associated acute megakaryoblastic leukaemia; LDH, lactic dehydrogenase.

Median age, months (range)21 (8–38)15 (2–185)0·52
Sex
 M/F ratio12:128:130·42
Clinical findings
 Fever, n (%) 12 (50)14 (67)0·33
 Lymphadenopathy, n (%)  2 (8·3) 6 (29)0·12
 Hepatomegaly, n (%) 16 (67)14 (67)1
 Splenomegaly, n (%) 11 (46) 9 (43)1
Laboratory findings
 Haemoglobin, g/l (median, range) 83 (52–130)82 (36–110)0·38
 Platelet count, ×109/l (median, range) 26·0 (2–143)24·0 (3–492)0·72
 Leucocyte count, ×109/l (median, range)  6·8 (2·4–107·0)17·9 (0·1–134·8)0·07
 Circulating blast cells, % (median, range)15·0 (0–88)13·0 (0–82)0·95
 Bone marrow blast cells, % (median, range)25·0 (2–90)50·0 (10–81)0·09
 Serum LDH, IU/L (median, range)662 (260–6450)1639 (338–8660)0·11

Morphological features

Forty-two of 45 BM smears were studied, as three smears of patients with non-DS-AMKL were unevaluable. The leukaemic cells of all 42 patients were negative for myeloperoxidase, chloroacetate esterase, and alpha naphthyl butyrate esterase activity. The morphology of blast cells was extremely varied and was categorized it into three groups according to the stage of megakaryocyte maturation: type 1, completely undifferentiated blasts with nucleolus or vacuoles in the cytoplasm (Fig 1A and B); type 2, intermediately differentiated blasts with cytoplasmic blebs, sometimes a large cytoplasm and azurophilic granules (Fig 1C and D); and type 3, blasts with dysmegakaryocytopoiesis (Fig 1E) including the presence of micromegakaryocytes (Fig 1F). The blast cells with deep blue cytoplasm (type 1b, 2b) (Fig 1B and D) were distinguishable from type 1a or 2a blasts (Fig 1A and C). The morphology of blast cells in DS-AMKL and non-DS-AMKL were distributed as follows: type 1 (63%, 39%), type 2 (25%, 39%), and type 3 (12%, 22%) (Tables II and III). Type 1b and 2b blasts were detected in eight of 24 patients (33%) with DS-AMKL. The blast cells tended to be less mature in DS-AMKL than in non-DS-AMKL in terms of morphology. Seven patients presented with type 3 morphology: four patients (DS-AMKL, one; non-DS-AMKL, three) had increased numbers of micromegakaryocytes with type 1 blasts, two patients with DS-AMKL had increased numbers of dysmegakaryocytopoiesis, and one non-DS-AMKL patient with t(16;21) showed type 3 blasts with emperipolesis and cytophagocytosis, which are characteristic in patients with t(16;21) (Imashuku et al, 2000). Dysplasia in trilineage blood cells was seen in two patients (DS-AMKL, one; non-DS-AMKL, one). Prominent dyserythropoiesis without involvement of the myeloid cell lineage was found in five patients (DS-AMKL, three; non-DS-AMKL, two). Emperipolesis was observed in three patients (DS-AMKL, two; non-DS-AMKL, one).

Figure 1.

 Morphological categories of the blasts. May-Giemsa staining of the bone marrow smears. (A) Type 1a: completely undifferentiated blasts. (B) Type 1b: completely undifferentiated blasts with deep blue cytoplasm. (C) Type 2a: intermediately differentiated blasts with cytoplasmic blebs. (D) Type 2b: intermediately differentiated blasts with cytoplasmic blebs and deep blue cytoplasm. (E) Type 3: blasts with dysmegakaryocytopoiesis. (F) Type 3: blasts with micromegakaryocytes. Original magnification ×1000 for all panels.

Table II.   Morphological classification, immunophenotype and karyotype of the 24 patients with Down syndrome-associated acute megakaryoblastic leukaemia.
Patient no.Age (months)SexMorphological classificationImmunophenotype (%)Karyotype
CD7CD13CD36CD41CD42GlyA
  1. GlyA, glycophorin A; Tri, trilineage dysplasia; NA, not available; Ep, emperipolesis; dE, dyserythropoieis; m, micromegakaryocytes.

A: Constitutional trisomy 21 only
  132F3+Tri5719421014NA47,XX,+21c
  221F1a68364212NA1347,XX,+21c
  321M2b72175556NA2047,XX,+21c
  421F3+Ep1845NA3144NA47,XX,+21c
  518M2b53788143472447,XX,+21c
B: Numerical abnormalities only
  621M1a8320NA4142251,XY,+8,+19,+21c,+21,+22
  78M1a5610NA1623NA48,XY,+8,+21c
C: −5/del(5q) and/or −7/del(7q)
  820F1b7830352434NA47,XX,del(6)(q23q25),−7,+ring(7?),+21c
  923F1a3629837675NA46,XX,−1,+der(1)t(1;?),−5,+der(5)t(5;7),−7,+21c
 1024F1a+dE9030352434NA47,XX,del(7)(q32),+21c
 1117F3+m4075NA14121247,XX,add(5)(p15),−7,+8,+21c
 1220F2b1479NA4259990,idem×2,−3,−7,−9,del(11)(q?),−18
 1338M2b+dE8183NA1520447,XY,−7,+8,+21c,+rl
D: Other structural changes
 1420F1a8327NA5987NA48,X,der(X)t(X;1)(q28;q25),+11,+21c
 1518M1a701234223847,XY,der(3)t(3;3)(p25;p10),i(7q), der(17)t(1;17) (q25;q25),+21c
 1617F1a514NA1192848,idem,add(5)(q1?),+11
 1712M1a8936NA74191547,XY,der(7)t(1;7)(q23;q36),del(20)(q11q13.1),+21c
 1829M2b60436914346247,XY,del(11)(p?),+21c
 1925F1a+dE7123NA41NA8447,XX,t(5;12)(p15;q21),add(7)(p11),+21c,add(22)(q13)
 2018M1b5237NA10263248,idem,+add21
 2122F1b+Ep99788859669347,XX,add(7)(p11),add(19)(p13),+21c
 228M1a72536472754NA
 2321M2a87123320234NA
 2419F1a8726NA40NA11NA
Table III.   Morphological classification, immunophenotype and karyotype of the 21 patients with non-Down syndrome-associated acute megakaryoblastic leukaemia.
Patient no.Age (months)SexMorphological classificationImmunophenotype (%)Karyotype
CD7CD13CD36CD41CD42GlyA
  1. GlyA, glycophorin A; Tri, trilineage dysplasia; NA, not available; Ep, emperipolesis; dE, dyserythropoiesis; m, micromegakaryocytes.

A: Normal karyotypes
  1147F3+m6695425319146,XX
  218M2a+Tri25544042446,XY
B: Numerical abnormalities only
  33FNA188NA7754848,XX,+21,+22
  415F2a429NA6810NA58,XX,+X,+2,+2,+6,+7,+8,+10,+13,+15,+19,+19,+22
  538F1a3830NA104NA49,XX,+12,+18,+22
  612FNANA30NA95NANA57,XX,+2,+4,+6,+7,+8,+10,+14,+15,+19,+19,+22
  745M1a953NA863148,idem,+8
  82M2a1932NA4618551,XY,+6,+7,+8,+19,+21
C: t(1;22)(p13;q13)
  92F2a+dENANANANANANA46,XX,t(1;22)(p13;q13)
 1012M1a513NA6770NA51,XY,+der(1)t(1;22)(p13;q13),t(1;22)(p13;q13), +6,+7,+10,+19
D: 3q21q26 abnormalities
 1112M1a143644847NA46,XY,−11,der(11)t(3;11)(q21;p15)
E: t(16;21)(p11;q22)
 1238M3+Ep369NA3333NA46,XY,t(16;21)(p11;q22)
F: −7
 138F1a307NA8280NA46,XX,−7,−7,del(11)(p11),+2mar,inc
 1441M3+m6643NA2832847,XY,−7,+21,+ring(1)
G: Other structural changes
 157M1a1265NA57NANA46,XY,t(2;7)(p12;p22)
 1628F1aNANANA3017151,XX,+X,6p+,6p+,−13,+21,+21,+2mar
 1735F3+m3182781NA47,XX,add(16)(p13),+21
 1812F2a+dE305626755546,XX,t(2;11;19)(q31;q13;q13.4),del(3)(q23),13q−
 196FNANA10NA5253NA51,XX,4p+,11q+,14q+,+17,+19,+21,+22,+mar
 2033F2a521251146,XX,del(2)(q11),del(2)(q31),der(5)t(2;5)(q11;q22), der(5)t(5;13)(q35;q14),−13,add(16)(p13),+mar
 21185F2a5990NA164NA49,XX,+5,+8,i(17)(q10),+21

Immunophenotyping

All but one patient had immunophenotyping studies performed (Tables II and III); the BM sample of one patient with non-DS-AMKL who showed CD41 expression in blast cells by histopathological examination and translocation t(1;22) was insufficient and could not evaluated. The leukaemic cells of 44 patients expressed at least one platelet-associated antigen (CD36, CD41, CD42, or CD61). The blast cells with low expression of platelet-associated antigens in two patients (DS-AMKL, one; non-DS-AMKL, one) were positive for PPO. Among the myeloid antigens, CD13/CD33 was expressed by leukaemic blast cells in 78%/53% and 60%/78% of patients with DS-AMKL and non-DS-AMKL, respectively. Atypical expression of lymphoid-associated antigens CD7 was detected in 88% and 53% of patients with DS-AMKL and non-DS-AMKL, respectively (P = 0·003). Glycophorin A was detected only on the leukaemic cells of 47% of patients with DS-AMKL (P = 0·009). Interestingly, 86% of the type 1b and 2b blasts were positive for glycophorin A.

Cytogenetic findings

Cytogenetic studies were performed for 21 patients with DS-AMKL and 21 patients with non-DS-AMKL (Tables II and III). Three patients with DS-AMKL had insufficient BM samples. Cytogenetic abnormalities of leukaemic cells were classified into seven categories: normal karyotype including constitutional trisomy 21 in DS-AMKL; numerical abnormalities only; t(1;22)(p13;q13); 3q21q26 abnormalities; t(16;21)(p11;q22); −5/del(5q) and/or −7/del(7q); and other structural changes. Normal karyotype including constitutional trisomy 21 was found in five patients with DS-AMKL and in two patients with non-DS-AMKL. Numerical chromosomal abnormalities were common in non-DS-AMKL. Patients with non-DS-AMKL had trisomy 8 (six patients), trisomy 19 (five patients), trisomy 21 (seven patients), and monosomy 7 (two patients). Six patients with DS-AMKL had −7/del(7q) and one of them had both monosomy 5 and 7. The translocation t(1:22) was found in two patients with non-DS-AMKL and 3q21q26 abnormalities, which are common in adult AMKL (Lu et al, 1993; Tallman et al, 2000; Dastugue et al, 2002; Duchayne et al, 2003), was found in one patient with non-DS-AMKL. The translocation t(16;21) was found in one patient with non-DS-AMKL. These recurrent structural changes were not observed in patients with DS-AMKL. The 11q23 abnormalities and the Philadelphia chromosomes were not detected in either group. Other structural changes, such as t(5;12)(p15;q21) was found in DS-AMKL and t(2;7)(p12;p22), t(2;11;19)(q31;q13;q13) were found in non-DS-AMKL.

GATA1 mutations

GATA1 mutation analysis was performed in 17 of 24 patients with DS-AMKL and 11 of 21 patients with non-DS-AMKL. GATA1 mutations were observed in all patients with DS-AMKL and one patient with non-DS-AMKL. Ten of 17 patients with GATA1 mutations in DS-AMKL were previously reported (Hirose et al, 2003).

Outcome

Twenty-three of 24 (96%) DS-AMKL patients achieved CR. Three patients relapsed and died, and two other patients with congenital heart anomalies died of congestive heart failure. Of the 24 patients with DS-AMKL, 19 (79%) are currently alive. Of the 21 patients with non-DS-AMKL, 16 (76%) achieved CR. One patient with t(1;22) did not receive induction therapy because of multiple organ failure on the day of admission. Three of four non-responders to induction therapy underwent successful allogeneic SCT and one patient died of pneumonia. Twelve patients received BMT in the first CR, four of whom relapsed and three of whom died. Five patients received chemotherapy only, four of whom have remained in CR. Overall, 16 of 21 patients (76%) are currently alive.

The estimate of 10-year OS was 79% (95% CI: 54–90) for patients with DS-AMKL and 76% (95% CI: 58–91) for patients with non-DS-AMKL with a median follow-up of 78 months (range, 20–243 months) (P = 0·81, Fig 2). The estimate of 10-year EFS was 79% (95% CI: 58–91) for patients with DS-AMKL and 57% (95% CI: 36–77) for patients with non-DS-AMKL (P = 0·09, Fig 3). The outcome of DS-AMKL and non-DS-AMKL was comparable. In non-DS-AMKL, the estimated OS of the 15 children who received SCT (79%, 95% CI: 51–93) did not differ from five children treated with chemotherapy alone (80%, 95% CI: 30–97) (P = 0·95).

Figure 2.

 Estimated probability of overall survival for 24 patients with AMKL associated with Down syndrome (DS-AMKL) and 21 patients with non-DS-AMKL. The estimate of 10-year overall survival was 79% [95% confidence interval (CI): 54–90] for patients with DS-AMKL and 76% (95% CI: 58–91) for patients with non-DS-AMKL (P = 0·81).

Figure 3.

 Estimated probability of event-free survival for 24 patients with AMKL associated with Down syndrome (DS-AMKL) and 21 patients with non-DS-AMKL. The estimate of 10-year event-free survival was 79% (95% CI: 58–91) for patients with DS-AMKL and 57% (95% CI: 36–77) for patients with non-DS-AMKL (P = 0·09).

Discussion

In the current series, 23·2% of patients with AML were identified as having AMKL; this frequency was higher than those found in several collaborative group studies of childhood AML (Ravindranath et al, 1992; Gamis et al, 2003; Creutzig et al, 2005; Zeller et al, 2005; Rao et al, 2006). The relative frequency of AMKL has varied markedly, ranging from 4·1% to 15·3%. There are possible explanations for the high proportion of AMKL in our series. One is the difference in prevalence of DS-AMKL. The proportion of DS-AMKL in this study was 53·3%, which was much higher than those of other institutions. For example, the ratio of DS-AMKL to non-DS-AMKL was 6:35 in the report from St. Jude Children’s Research Hospital (Athale et al, 2001). In the 1980s and early 1990s, the outcome of patients with DS-AMKL was generally poor (Levitt et al, 1990). Most DS-AMKL patients have not been enrolled in clinical studies. In the German (Berlin-Frankfurt-Münster; BFM) co-operative group studies, the percentage of patients with DS has gradually increased since study 78 (1·9%), study 83 (5·6%), study 87 (8·1%), study 93 (9·7%) and study 98 (12·9%) (Creutzig et al, 2005). On the other hand, among patients registered in the population-based Nordic study between 1984 and 2001, 72 of 515 (14·0%) children with AML had DS (Zeller et al, 2005), which is similar to the percentage in our study, in which 25 of 194 patients (12·9%) with AML had DS. Our clinical trial for DS-AMKL used a less intensive regimen that had been specifically designed for DS-AMKL in the mid-1980s (Kojima et al, 1990, 1993, 2000). Therefore, patients with DS-AMKL were not excluded from the data file, which may account for the higher proportion of DS-AMKL in our series. In a report from Mexico, 29 of 152 (19·1%) children with AML were diagnosed as AMKL among whom only one patient had DS (Paredes-Aguilera et al, 2003).

The incidence of non-DS-AMKL was 12·4% in our non-DS-AML series, which might also be higher than those of other reports. Immunophenotyping with platelet-associated antigens and electron microscopic identification of PPO were introduced over 15 years ago in our hospitals (Kojima et al, 1990). The incidence of AMKL might have been underestimated because of its diverse clinical presentation and requirement of specific laboratory methods in early reports. As previously described, 19·1% of children with AML had AMKL in a report from Mexico (Paredes-Aguilera et al, 2003). The incidence of AMKL might be different between western countries and non-western countries. We speculate that AMKL represents approximately 10% of all cases of non-DS-AML in children.

In the current study, the morphology of blast cells was categorized into three groups (type 1, type 2, or type 3) according to the stage of megakaryocyte maturation, modelled according to the FAB classification of myeloid leukaemia (Fig 1). The blast cells with deep blue cytoplasm (type 1b, 2b) were only detected in patients with DS-AMKL and were positive for glycophorin A in all but one patient. Erythroid-specific mRNAs encoding γ-globin and erythroid δ-amino levulinate synthase were expressed in blasts from all patients with DS-AMKL (Ito et al, 1995). These findings suggest that type 1b and 2b blasts arise from bipotent megakaryocyte/erythrocyte progenitors. A high incidence of the co-expression of the T cell-associated marker CD7 in patients with DS-AMKL compared with patients with non-DS-AMKL was also observed. In addition to mature T cells, the CD7 antigen is expressed on immature haematopoietic cells (Kita et al, 1993; Creutzig et al, 1995). In terms of morphology and immunophenotype, the blast cells were more immature in patients with DS-AMKL than in those with non-DS-AMKL.

The cytogenetic profile of AMKL is complex, which reflects the heterogeneity of the disease. In the current study, seven cytogenetic groups were identified. The translocation t(1;22)(p13;q13), which produces the RBM15-MKL1 fusion gene (Ma et al, 2001; Mercher et al, 2001, 2002), was detected in two patients (4·3%) with non-DS-AMKL. This frequency was lower than in another study (Dastugue et al, 2002; Duchayne et al, 2003). Dastugue et al (2002) reported that RBM15-MKL1 transcript was detected in one patient with a normal karyotype, suggesting that the molecular determination method may increase the detection of t(1;22) translocation. The 3q21q26 abnormalities were observed in one patient with non-DS-AMKL. The 3q21q26 abnormalities are rare in childhood AMKL (6·7% in Dastugue et al, 2002), while they are seen in 17% to 20% of adult AMKL patients (Lu et al, 1993; Tallman et al, 2000; Dastugue et al, 2002; Duchayne et al, 2003). The translocation t(16;21)(p11;q22) generating a FUS-ERG transcript (Kong et al, 1997) was found in one patient with non-DS-AMKL. This translocation has been found in all subtypes of AML except M3, including several cases with AMKL. The patient with t(16;21) in our series morphologically showed type 3 blasts with emperipolesis and cytophagocytosis. These morphological findings of BM are characteristic of patients with t(16;21), regardless of FAB classification (Imashuku et al, 2000). These recurrent structural changes were not observed in patients with DS-AMKL. On the other hand, monosomy 7/del(7q) was more frequent in patients with DS-AMKL (29%) than non-DS-AMKL (9·5%). Monosomy 5/del(5q) was found in only one patient with DS-AMKL who had also monosomy 7. In non-DS-AMKL, trisomies (+8, +19, +21) were more frequent than DS-AMKL, as previously reported (Lu et al, 1993; Dastugue et al, 2002; Duchayne et al, 2003; Reinhardt et al, 2005). Acquired trisomy 21 is a common chromosome gain of childhood AMKL, reported in 23% to 43% of patients with non-DS-AMKL (Ribeiro et al, 1993; Athale et al, 2001). In our study, acquired trisomy 21 was found in 33% of patients with non-DS-AMKL. Coupled with other reports, acquired trisomy 21 seems to have a higher incidence in non-DS-AMKL in children than in other childhood non-DS-AML. The 11q23 abnormalities are often detected in childhood AMKL (Athale et al, 2001; Reinhardt et al, 2005), although not in the current study. The Philadelphia chromosomes or i(12)(p10) with mediastinal germ cell tumour, which were mainly found in adult AMKL (Dastugue et al, 2002; Duchayne et al, 2003), were also not found in the current study.

GATA1 mutation analysis was performed in 17 of 24 patients with DS-AMKL and 11 of 21 patients with non-DS-AMKL. GATA1 mutations were observed in all patients with DS-AMKL as previously reported (Wechsler et al, 2002; Hirose et al, 2003). In contrast to DS-AMKL, GATA1 mutations were rarely found in patients with non-DS-AMKL. To date, only four children with GATA1 mutations in non-DS-AMKL have been reported (Rainis et al, 2003; Bourquin et al, 2006). Interestingly, all of them had acquired trisomy 21 in their leukaemic cells. Our non-DS-AMKL patient with GATA1 mutation did not have acquired trisomy 21 in his leukaemic cells.

Before the 1990s, most patients with DS-AML were treated outside of clinical studies and received suboptimal therapies, resulting in poor outcomes (Levitt et al, 1990). Following the recognition of the favourable outcome when treated with protocols of the collaborative study group for AML (Ravindranath et al, 1992), there has been an increase in recruitment into protocol studies. However, it has become apparent that resistant disease is rare but treatment-related deaths are frequent in most series (Creutzig et al, 1996; Lange et al, 1998), and several collaborative groups adapted their AML protocols for DS-AML by reducing the dose of chemotherapeutic agents (Creutzig et al, 2005; Zeller et al, 2005; Rao et al, 2006). In recent reports, 5-year survival rate have been in excess of 80%, largely because of reductions in treatment-related deaths with a decrease from 30% to 40% in the early 1990s to around 10% in recent studies (Creutzig et al, 2005; Zeller et al, 2005; Rao et al, 2006). Since the mid-1980s, we have used a less intensive regimen specifically designed for DS-AML (Kojima et al, 1990, 1993, 2000). The excellent outcome of DS-AMKL may originate from early use of a regimen specific for DS-AML.

The 10-year OS in our series was 76% (95% CI: 58–91) for patients with non-DS-AMKL, which was superior to those of other reports (Ribeiro et al, 1993; Athale et al, 2001; Reinhardt et al, 2005). The prognosis for children with non-DS-AMKL was poor in previous reports. According to the report from St Jude Children’s Research Hospital, the 2-year OS was only 14%, which was significantly higher after allogeneic SCT (30%) than after chemotherapy alone (0%) (Athale et al, 2001). The result of a recently published report on AMKL from the European Group for Blood and Marrow Transplantation was excellent (Garderet et al, 2005). Three-year OS was 82% in 19 children after allogeneic SCT and 61% in 38 children after autologous SCT. The authors recommended allogeneic SCT when an HLA-matched sibling is available; otherwise, autologous SCT should be used for children with AMKL in first CR. However, this report included 11 children with DS and analysed the outcome of children with DS and without DS together, which confused the interpretation of the results. In the current study, 5-year OS was 79% in patients who received SCT, which did not differ from patients achieving CR and being treated with chemotherapy alone (80%, P = 0·98). In a recent report from the BFM collaborative group, the 5-year OS was 43% in the SCT group and 54% in the chemotherapy group (P = 0·37) (Reinhardt et al, 2005). The recent use of intensified chemotherapy may abrogate the indication of allogeneic SCT for children with non-DS-AMKL.

In conclusion, our study shows the diverse heterogeneity of childhood AMKL and the differences in the clinical and biological presentation between DS-AMKL and non-DS-AMKL. Subclassification according to megakaryocyte maturation and cytogenetic abnormalities in childhood AMKL is warranted.

Ancillary