Dr Anupama Rao, Department of Haematology, Camelia Botnar Laboratories, Great Ormond Street Hospital for Children, London WC1N 3JH, UK. E-mail: firstname.lastname@example.org
Down syndrome (DS) children are at an increased risk of developing myelodysplasia and acute myeloid leukaemia (AML). We retrospectively analysed the population-based data on 81 children with myeloid leukaemia of Down syndrome (ML-DS) from the UK National Registry of Childhood Tumours and experience in the Medical Research Council (MRC) AML 10 and AML 12 trials, which enrolled 46 children with ML-DS from 1988 to 2002. Eight per cent of UK children with AML had DS, but DS children comprised only 5% of children registered in MRC trials. The unique clinical characteristics of ML-DS were confirmed. Overall survival (OS) of ML-DS at 5 years increased from 47% in UK children diagnosed from 1988 to 1995 to 75% in children diagnosed from 1996 to 2002. OS for DS children registered in AML 10 and AML 12 was 74% in 5 years and improved from AML 10 to AML 12 (56% vs. 83%) There was no significant difference in OS between DS and non-DS children (OS: 74% vs. 62%, P = 0·4) in the trials, but this result masked a significant increase in early death amongst DS children, with a significant reduction in mortality later on. Relapse was significantly reduced (3% vs. 39%, P = 0·0003), leading to the improved disease-free survival (83% vs. 56%, P = 0·02). Given the increased number of earl treatment-related deaths, future treatment protocols should aim to reduce chemotherapy dosage or intensity whilst maintaining low rates of resistant and recurrent disease.
The association between Down syndrome (DS) and leukaemia was first described by Krivit and Good (1957). The overall risk of leukaemia is increased 10–20-fold in DS (Robison, 1992). Although both acute lymphoblastic leukaemia (ALL) and acute myeloid leukaemia (AML) have a similar incidence in DS (Lange, 2000), children with DS under the age of 5 years have a fourfold higher incidence of AML to ALL (Ross et al, 2005) accounting for 8–14% of cases (Zeller et al, 2005). Myelodysplasia (MDS) precedes AML in DS children in up to 70% of cases (Zipursky et al, 1992; Creutzig et al, 1996). Both MDS and AML in DS respond well to chemotherapy and it has been proposed that the two disorders are recognised as a single entity – myeloid leukaemia of DS (ML-DS) in the World Health Organization (WHO) classification. (Hasle et al, 2003). Over 80% of DS children have the French–American–British (FAB) M7 subtype of AML (Zipursky et al, 1987). There is a convincing evidence of better survival outcomes in children with ML-DS when compared with AML in other children (Creutzig et al, 1996; Craze et al, 1999). This might be attributable to several unique features of ML-DS, particularly sensitivity to chemotherapy (Ravindranath et al, 1992; Kojima et al, 1993; Ravindranath, 2003), with disease resistance and relapse being uncommon. Conversely, DS children suffer high rates of treatment toxicity, (Lange et al, 1998; Craze et al, 1999) with an increased risk of treatment-related death. Accordingly, some reduction in treatment intensity compared with standard therapy for AML may be plausible, but any reduction in intensity needs to be balanced against the risks of jeopardising the high-cure rate. Refinement of AML therapy to optimise cure rates whilst minimising toxicity is the major focus of modern treatment protocols.
This study compares outcomes for children with and without DS who were treated in the UK Medical Research Council (MRC) AML 10 and AML 12 trials, which were open to recruitment during the period 1988–2002, and differences in outcome for ML-DS between the two trials themselves. The population-based data on trial recruitment and outcome are provided for children with ML-DS in the UK during the same period. This study updates data published by Craze et al (1999) describing the outcomes for the time period of AML10 and presents new information for the time period of AML 12.
Patients and Methods
The MRC AML 10 (1988–1995) and AML 12 (1995–2002) trials were consecutive clinical studies, which together were open to recruitment between 1988 and 2002. The two trials predominantly recruited patients from the UK, but children were also registered from the Republic of Ireland and the Netherlands.
Children aged <15 years with AML, refractory anaemia with excess blasts (RAEB) and RAEB in transformation (RAEBt) were eligible for treatment according to the paediatric section of AML 10 and AML 12. DS infants with transient abnormal myelopoiesis (TAM) were not eligible for trial registration.
The trials required central review of morphology of bone marrow aspirates and cytochemistry by a panel of three haematologists. Bone marrow trephines and the results of immune phenotyping were not available for the review process. Bone marrow cytogenetics at diagnosis was requested for all patients.
In the AML 10 trial, all children were scheduled to receive four courses of intensive chemotherapy (Hann et al, 1997). The two induction courses were based on the daunorubicin and cytarabine, and children were eligible for randomisation for the third drug, either thioguanine or etoposide (DAT versus ADE). The two consolidation courses were amsacrine, cytarabine and etoposide (MACE) and mitoxantrone and cytarabine (MidAC). Drug doses were reduced by 25% for children aged <1 year or under 10 kg body weight. No dose reductions were specifically made for children with DS. Children with a matched sibling donor were eligible for bone marrow transplantation (BMT) in first remission after the completion of chemotherapy, whereas the remainders were eligible for randomisation between an autograft and no further therapy. Following the analyses of outcomes in AML 10, cytogenetic abnormalities were classified as favourable, intermediate and poor risk (Grimwade et al, 1998).
In the AML 12 trial, children were allocated between three risk groups based on the results of cytogenetics (see above) and response to the first induction course (Wheatley et al, 1999). At induction, patients were randomised between daunorubicin and mitoxantrone, and the selected drug was given with cytarabine and etoposide (ADE versus MAE). All children received consolidation therapy with MACE and MidAC. Children were eligible for randomisation between four and five courses of treatment in total. Children with DS or who were in the good risk group were ineligible for an allogeneic BMT in the first remission. Children randomised to a fifth course and who were not scheduled for BMT received high-dose cytarabine and asparaginase (CLASP). Autografting in first remission was discontinued.
Demographic and clinical features and clinical outcomes were compared between children with and without DS registered in the trials. The population-based data for children with AML, RAEBt and ML-DS diagnosed in the UK from April 1988 to May 2002, the era of entry to the AML 10 and AML 12 trials, were obtained from the National Registry of Childhood Tumours (NRCT). These data were less detailed, and analyses were limited to the population-based assessments of demographics, proportions receiving antileukaemic treatment and recruited to trials, and overall survival. Significance tests for differences in survival rates were not carried out because data on the FAB subtype and presenting white blood cell (WBC) count were incomplete and data on cytogenetics were unavailable.
Actuarial estimates of OS (time from diagnosis to death from any cause), event-free survival (EFS, time from diagnosis to any event including induction failure, relapse or death) and disease-free survival (DFS, time from remission to relapse or death) were calculated.
Statistical comparisons were carried out using the standard techniques, with differences in baseline demographics being compared using either chi-squared or Mantel–Haenszel tests. Unadjusted comparisons of complete remission (CR), induction death and resistant disease were performed using the Mantel–Haenszel test, with analyses adjusted for imbalances in baseline variables performed using the logistic regression. Time-to-event outcomes were analysed using either the log-rank test (unadjusted) or Cox regression (adjusted analyses). By convention, odds ratios (ORs) or hazard ratios (HRs) greater than one indicated an adverse outcome in DS children compared with non-DS (NDS). The proportional hazards hypothesis was tested by introducing a time-dependent covariate to the model. The significance of this additional variable was tested to determine if there was an evidence of a trend towards the different HRs over time.
To allow for differences between DS and NDS children in important prognostic baseline variables: age, WBC count and cytogenetics, comparisons of children with and without DS were stratified by these variables, as otherwise the imbalances would lead to a misleading comparison. Both the unadjusted headline comparison and the comparison adjusted for age, WBC count and cytogenetics are given, although estimates of effect size are given for the adjusted analyses only. P-values <0·05 were considered significant.
A total of 868 children (477 boys, 391 girls) were registered at diagnosis on AML 10 and AML 12, of whom 46 (5·3%; 24 boys, 22 girls) had DS (15 ML-DS enrolled on AML 10 and 17 ML-DS on AML 12). There were six children with DS from the Netherlands and eight from the Republic of Ireland registered on these trials. Details for all children are shown in Table I. Children with ML-DS were significantly younger than the NDS patients, with a median age at diagnosis of 1·5 compared with 7 years (P < 0·0001 for trend). Almost 90% of ML-DS children were diagnosed between 1 and 4 years of age, with only four children aged over 4 years and just one child in the first year of life.
Table I. Clinical features: data for Down syndrome AML (ML-DS) and non-Down syndrome AML (NDS-AML) from the AML 10 and AML 12 trials.
ML-DS (n = 46)
NDS-AML (n = 822)
Tests are either chi-squared test for heterogeneity or Mantel–Haenszel test for trend (indicated by *).
NB: The total numbers of trial patients include data from the UK, the Republic of Ireland and the Netherlands
Presenting WBC count (×109/l)
There were significant differences in FAB type between the DS and NDS children (P < 0·0001 for heterogeneity); RAEBt and FAB type M7 were more common in ML-DS, FAB types M2 and M5 were less common. Children with ML-DS also had lower presenting WBC counts than NDS children (P = 0·0006 for trend), with 69% of children with ML-DS having a WBC count <10 × 109/l at diagnosis. None of the children with ML-DS in this sample had favourable cytogenetics and there was a significant trend for DS children to have less favourable cytogenetics than NDS children (P = 0·0004).
Although rates of CR were similar [DS 89%, NDS 92%, unadjusted P = 0·5, adjusted OR 1·63 (95% confidence interval (CI) 0·52–5·07), adjusted P = 0·4], children with DS were more likely to die during induction [11% vs. 4%, unadjusted P = 0·02, adjusted OR 3·20 (95%CI 1·00–10·38), P = 0·04; Table II]. Induction deaths in DS children were because of infection (one respiratory syncitial virus pneumonia, one parainfluenza pneumonia, one pseudomonas septicaemia) and one respiratory failure with no stipulated cause and one death because of multiple causes. Induction death rates were not significantly lower in AML12 than AML10 (AML 10 19%, AML 12 7%, P = 0·2) but the numbers were too small for reliable analysis, and such historical comparisons may pick up both improvements in care, and also possible selection biases. There was no significant difference in the rate of resistant disease between DS and NDS children (0% vs. 4%, unadjusted P = 0·17, adjusted P = 0·3, OR not estimable using logistic regression). There were three ML-DS children who received allografts in first CR during the era of AML 10, all survived.
Table II. Data comparing the morbidity and survival of Down syndrome children with AML (ML-DS) and children without DS (NDS-AML).
*Tests used are logistic regression (CR, resistant disease, induction death) or Cox regression.
Complete remission (CR)
OS at 5 years
DFS at 5 years
EFS at 5 years
Relapse at 5 years
Death in CR at 5 years
Relapse rates were significantly lower amongst DS children than amongst NDS (adjusted P = 0·0003), with only one relapse in the DS children. By contrast, when adjusted for age, WBC count and cytogenetics, there was evidence of significantly worse mortality in CR amongst DS children (adjusted HR 2·98, 95% CI 1·18–7·48, P = 0·02): amongst the 1–4-year-old group, 5-year death in CR was 16% in the DS children, compared with 7% in the NDS children. Amongst the DS children, there were six deaths in CR, five of them because of gram-negative infections. One child died of cardiac failure because of cardiomyopathy. Overall, however, DFS was significantly improved in the DS group and there was a suggestion of better EFS (Fig 1) (DFS-adjusted HR 0·39, 95% CI 0·17–0·89, P = 0·02; EFS adjusted HR 0·56, 95% CI 0·30–1·06, P = 0·07). The proportional hazards hypothesis showed no significant evidence of any departure from the proportional hazards for DFS, relapse or death in CR.
Overall, there were no significant differences in OS between DS and NDS children (adjusted HR 0·77, 95% CI 0·41–1·47, P = 0·4), with 5-year survival rates of 74% vs. 62%. However, model checking of the proportional hazard assumption showed evidence (P = 0·01) of a difference in HR over time: early on, DS patients had an increased risk of death, whereas later, these positions were reversed. This feature can be seen graphically in the 1–4 year age group, which contained nearly 90% of the DS children, where the survival curves show the increased induction death rate amongst DS children compared with NDS children (Fig 2); there was no evidence of significant difference in survival in any particular age group (Fig 3), although numbers were small in some groups. Splitting the time period in to deaths before 6 months, and deaths after 6 months, there was significant interaction, with an adverse hazard in the 6 months following the entry, and a favourable hazard thereafter (Fig 3). Thus, DS was associated with an increased risk of early death, but it was associated with the improved survival amongst those who survive the initial treatment period.
Amongst DS children, there was some evidence that outcomes have improved over time: comparing AML10 with AML12 showed a significant improvement in 5-year survival from 56% to 83%, with concomitant improvements in DFS and EFS and a reduction in induction death and deaths in CR (31% vs. 7%) However, such historical comparisons are notoriously unreliable and need to be interpreted cautiously.
When analysing data from the NRCT, we followed Zipursky et al (1992) in assuming that infants recorded with ML-DS before 6 months of age had TAM. After excluding 10 such infants, all diagnosed before the age of 4 months, the NRCT included 1053 children with a diagnosis of AML, RAEBt or ML-DS during the 14-year study period. Of these, 81 (7·7%) had DS. The great majority of the DS children (77, 95%) were diagnosed at 1–4 years of age, and they represented 21% of all children with AML or RAEBt in this age group; only two (2·5%) were diagnosed during the first year of life, the youngest being 9 months of age, and two (2·5%) at the age 5–14 years. Numbers of DS and NDS children and the proportions that received antileukaemic treatment and that were entered in trials are shown in Table III, together with overall 5-year survival for all children and for the treated children. In the eras of both AML 10 and AML 12, over 90% of DS and NDS children received antileukaemic treatment and there was no significant difference between the proportions of DS and NDS children who were treated. It is notable, however, that all DS children diagnosed in the era of AML 12 were treated. The proportions of treated DS and NDS children who were recruited to a trial increased between the two periods, but the difference was not significant for DS children. The proportion of treated children who were entered in a trial was significantly lower for DS than for NDS children during both the era of AML 10 (χ2 = 16·4, 1 degree of freedom (d.f.), P < 0·0001) and that of AML 12 (χ2 = 14·9, 1 d.f., P < 0·0001). The overall 5-year survival of DS children increased from 47% to 75% between the two periods, and survival of treated DS children increased from 51% to 75%. All children who received palliative treatment died within 13 months of diagnosis. So far, two deaths of DS children have been recorded beyond 2 years from diagnosis, both amongst those diagnosed in the earlier period. One child died 13 years after diagnosis from cardiomyopathy and one died 15 years after diagnosis from AML.
Table III. The population-based data on children with AML, RAEBt and ML-DS diagnosed in the UK during the eras of entry to the trials AML 10 (April 1988 to March 1995) and AML 12 (April 1995 to May 2002). Source: NRCT.
April 1988–March 1995
April 1995–May 2002
April 1988–March 1995
April 1995–May 2002
Number (%) of children treated
Number (%) of treated children entered in trial
OS at 5 years (95% CI) for all children
47% (32%, 60%)
75% (57%, 86%)
50% (45%, 54%)
62% (58%, 66%)
OS at 5 years (95% CI) for treated children
51% (35%, 65%)
75% (57%, 86%)
53% (48%, 58%)
65% (60%, 69%)
Children with DS comprised 5·3% of patients registered in AML 10 and AML 12. This percentage was lower than that recorded for other collaborative group studies (Lange et al, 1998; Creutzig et al, 2005; Zeller et al, 2005), and was lower than in the UK population. In the period 1971–1986, fewer than half of all children with ML-DS in Britain received intensive chemotherapy (Levitt et al, 1990). The proportion treated increased substantially to 91% in 1988–1995, and by 1995–2002 all children with ML-DS were given specific antileukaemic therapy. Nevertheless, nearly two-thirds of children treated in the era of AML 10 and nearly one-half of those treated in the era of AML 12 were not formally entered in a trial, despite the excellent results achieved for DS children within the trials. The shortfall in registration in AML 10 and AML 12 was because of clinicians electing to treat DS children outside the trials for reasons that included concerns over the associated congenital anomalies, general state of health, treatment intensity and, especially, cumulative exposure to anthracyclines. Previous reports have identified a reluctance of clinicians to treat ML-DS intensively (Levitt et al, 1990), but trial registration in several countries improved once the high response rate of ML-DS to chemotherapy was demonstrated (Lange et al, 1998). Modifications in drug doses, avoidance of stem cell transplantation (SCT), and recommendations regarding supportive care have all been emphasised as important in achieving the optimum outcomes.
The great majority of children with ML-DS are under 5 years of age (Lange et al, 1998; Webb et al, 2001; Ross et al, 2005; Zeller et al, 2005). In our study, 90% of the DS children were aged 1–4 years at diagnosis. Age is a recognised prognostic factor, with poorer outcome in children over 4 years (Creutzig et al, 2005). The Childrens Cancer Group identified poorer outcomes from the age of 2 years (Gamis et al, 2003) but this has not been verified by other groups or in our study. It has been postulated that DS children who present over 4 years represent standard AML occurring in a DS child and not ML-DS, but this has not been confirmed. In AML 10 and AML 12, there were four children with DS-AML above the age of 4 years (three with FAB type M4 and 1 with M7). Two died of toxicity and two remained alive and free of disease, but the treatment intensity was the same as that employed in NDS children.
Children with ML-DS have a high prevalence of FAB type M7, and AML is preceded by MDS in up to 70% of cases (Zipursky et al, 1992; Lange et al, 1998; Craze et al, 1999; Creutzig et al, 2005). These findings were confirmed in AML 10 and AML 12, where M7 and RAEBt were the commonest FAB types on morphology review. FAB types M2, M3 and M5 were especially uncommon. The number of cases diagnosed with FAB type M7 was still relatively low compared with other reports, and may have been reduced artificially as bone marrow trephines were not available for review, and the diagnosis may be missed on blood films and bone marrow aspirates.
GATA 1 is a transcription factor, which is essential for the maturation of erythroid cells and megakaryocytes. The recent discovery of GATA -1 mutations that occur in nearly every DS child with TAM and acute megakaryoblastic leukaemia (Gurbuxani et al, 2004) provides the biological basis of the ML-DS phenotype, but not all children with TAM will develop AML as additional, as yet unidentified, genetic events are necessary for leukaemogenesis.
Children with ML-DS generally have a low WBC count at diagnosis (Zeller et al, 2005). This was confirmed in AML 10 and AML 12. No child with ML-DS had MRC good risk cytogenetics. After adjusting for cytogenetics, age and WBC count, there was no significant difference in survival between DS and NDS children. However, it appears that this overall non-significant result is because of two distinct processes: an excess of treatment-related deaths early on, and then a lower risk of disease-related death, indicated by the presence of only one relapsing child (who subsequently died because of unknown cause). However, larger datasets are needed to give reliable evidence and, in particular, validate whether the known AML prognostic factors of age, WBC count and cytogenetics retain their effect in children with DS.
Accordingly, ML-DS has a high sensitivity to chemotherapy. This confirms findings from previous studies by the Pediatric Oncology Group, Childrens Cancer Group, Berlin Frankfurt Munster group and Nordic Organisation for Paediatric Haematology and Oncology (NOPHO) (Ravindranath et al, 1992; Zipursky et al, 1994; Lie et al, 1996; Lange et al, 1998; Craze et al, 1999; Zeller et al, 2005). This sensitivity to chemotherapy is at least partly because of overexpression of cystathionine B synthase, which confers increased sensitivity to cytarabine in the leukaemic blasts of ML-DS (Taub et al, 1996, 1999; Ge et al, 2005). A recent study (Zwaan et al, 2002) demonstrated a 12-fold increase in sensitivity to cytarabine in ML-DS compared with NDS AML, as well as increased sensitivity to anthracyclines (two- to sevenfold), mitoxantrone (ninefold) and etoposide (20-fold).
Our study confirms previous experience that DS children are at high risk of treatment-related deaths (Lange et al, 1998; Zubizarreta et al, 1998; Craze et al, 1999), primarily because of infection. There were fewer treatment-related deaths in DS children on AML 12 compared with AML 10, and the same effect has been reported for NDS children treated on the trials (Riley et al, 1999). It is not possible to draw firm conclusions regarding these differences, but improvements in supportive care and familiarity with the management of complications were most likely to be responsible.
Despite dosage reductions for children with ML-DS in the BFM 98 trial (Creutzig et al, 2005), treatment-related deaths still equalled those because of disease and this suggests that further reductions in chemotherapy doses and in dose intensity (Gamis et al, 2003) may be justified. With the efficacy of chemotherapy and the increased toxicity of SCT in DS children (Lange, 2000), there is no role for either autologous or allogeneic SCT in first line therapy (Rubin et al, 1996; Lange et al, 1998). Anthracyclines carry particular toxicity in DS children, especially with the severity of mucositis and risk of infection that largely account for the excess of treatment-related deaths. Several groups have now explored dose reduction of anthracyclines by around 25% compared with their standard AML protocols, and confirmed efficacy (Creutzig et al, 2005; Zeller et al, 2005). AML 10 and AML 12 had high cumulative doses of anthracycline (300 mg/m2 daunorubicin plus 50 mg/m2 mitoxantrone) compared with recent BFM and NOPHO studies (Creutzig et al, 2005; Zeller et al, 2005) and these data indicate that cumulative doses of anthracycline in MRC protocols have been higher than necessary for the treatment of ML-DS. Hence, the major question in therapy of ML-DS is how much can the treatment further reduce whilst maintaining a low rate of resistant and recurrent disease. To this end, the International BFM group has developed an observational study based on the treatment for ML-DS in BFM 98 with low cumulative dose of daunorubicin 240 mg/m2 and using a total of 29 g/m2 of cytarabine. This study emphasises the need for specific protocols in children with ML-DS.
The authors are indebted to the members of the UK Childhood Leukaemia Working Party, United Kingdom Childrens Cancer Study Group and Dutch Childhood Leukaemia Study Group who treated the children described in this report. The Childhood Cancer Research Group is supported by the Department of Health and the Scottish Executive. The views expressed here are those of the authors and not necessarily of the Department of Health or the Scottish Executive.