Portal de Periódicos da CAPES

The Management of Neoplastic Disorders of Haematopoeisis in Children with Down's Syndrome

2000; Wiley; Volume: 110; Issue: 3 Linguagem: Inglês

10.1046/j.1365-2141.2000.02027.x

1365-2141

Beverly Lange,

Lymphoma Diagnosis and Treatment

Down's syndrome (DS) is the most common factor predisposing to childhood leukaemia. Children with DS have a 10- to 20-fold excess risk of developing leukaemia ( Table I) ( Krivit & Good, 1957; Stewart et al, 1958 ; Wald et al, 1961 ; reviewed in Fong & Brodeur, 1987; Levitt et al, 1990 ; Robison, 1992; Zipursky et al, 1992 ; Avet-Loiseau et al, 1995 ). Roughly 1% of children with DS develop one or more of the following distinctive types of leukaemia: (i) a spontaneously regressing congenital or neonatal myeloproliferative disorder (TMD) (also known as transient myeloproliferative syndrome, congenital transient leukaemia, congenital leukaemoid reaction, transient leukaemoid proliferation and transient abnormal myelopoiesis); (ii) acute myeloid leukaemia (AML), usually acute megakaryoblastic leukaemia (AMKL) (or erythro/megakaryoblastic leukaemia before the age of 5 years); or (iii) common B-lineage acute lymphoblastic leukaemia (ALL) (for a review, see Zipursky et al, 1987 ; Robison, 1992; Avet-Loiseau et al, 1995 ). The current management strategy for these disorders, with few exceptions and considerable caution, is simple: in TMD, 'do nothing'; in AML, 'do less'; and in ALL, 'do more'. The rationale for this approach is the subject of this review. Table II lists seminal contributions to our understanding of leukaemia in DS. The full anatomical, physiological and functional expression of DS derives from trisomy of chromosome 21 ( Lejeune et al, 1959 ). Predisposition to leukaemia is common not only in the 90% of children with typical DS features and trisomy 21 but also in phenotypically normal or abnormal children with trisomy 21 mosaicism (5%), Robertsonian translocations (1%), partial trisomy of 21 (0·5%) or ring 21 (0·5%) ( Fong & Brodeur, 1987; Iselius et al, 1990 ). There may even be an excess of leukaemia in those with DS mosaicism, perhaps because of a survival advantage in mosaics ( Iselius et al, 1990 ). The trisomy is caused by non-disjunction during the first or, less often, the second meiotic division, or in the case of a DS mosaic during a post-zygotic division (for a review, see Hassold et al, 1995 ). Mapping of centromeric genes to determine parental origin of non-disjunction indicates that in over 90% of cases the non-disjoined chromosome 21 is the maternal chromosome ( Abe et al, 1989 ; Lorber et al, 1992 ; Hassold et al, 1995 ). The non-disjoined maternal chromosome 21 is almost always present in the leukaemia blasts ( Heaton et al, 1981 ; Ferster et al, 1986 ; Lorber et al, 1992 ). The association of DS with advancing maternal age has been appreciated for over a century ( Fraser & Mitchell, 1876). Brewster and Cannon (1930) first called attention to the occurrence of leukaemia in DS in their paper 'Acute Lymphatic Leukaemia: Case Report in 11-Month Mongolian Idiot.' By 1972, the world literature contained 272 reports of acute or chronic leukaemia, congenital leukaemia, myelofibrosis and/or megakaryoblastic marrow infiltration in infants and children with DS and four reports in adults with DS ( Rosner and Lee, 1972). Because of spontaneous regression in 22 cases of DS newborns and because there was often disagreement about whether the leukaemia was lymphoid or myeloid, Rosner and Lee (1972) urged caution in diagnosing leukaemia in DS patients. Numerous descriptions of megakaryoblastic leukaemia and myelofibrosis ( Evans, 1975 ; DenOttolander et al, 1979 ; Bain et al, 1981 ; Lewis, 1981,1984; Sulton et al, 1981 ; Bevan et al, 1982 ; Chan et al, 1983 ; Mirchandani & Palutke, 1983 ; Huang et al, 1984; and others) led to the French–American–British (FAB) criteria for diagnosis of megakaryoblastic leukaemia ( Bennett et al, 1985 ). Recognition of AMKL as a biologically distinct form of leukaemia that mimics FAB L2 ALL resolved most issues about the cell of origin of leukaemia in children with DS. Gene dosage or disomy of a leukaemia predisposition gene or haematopoiesis regulatory gene on chromosome 21 has been the presumed mechanism of leukaemogenesis in DS ( Abe et al, 1989 ; Lorber et al, 1992 ; Rogan et al, 1995 ; Huret & Leonard, 1997) . This theory gains credibility from the fact that acquisition of one or more chromosomes 21 is the most common numerical abnormality in acute leukaemia ( Mitelman et al, 1990 ; Berger, 1997 ; Belkov et al, 1999 ; Heerema et al, 1999 ). However, overexpression of the known genes on chromosome 21 has not identified the leukaemia predisposition gene. Moreover, gene mapping and genotypic–phenotypic correlations on chromosome 21 in DS indicate that anomalies as complex as mental retardation in DS span a large part of the chromosome ( Korenberg et al, 1992 , 1995; Delabar et al, 1998 ; Yamakawa et al, 1998 ). Similarly, predisposition to leukaemia, spontaneous regression and response to cytotoxic therapy may be polygenic phenomena. Superoxide dismutase ( Druzhyna et al, 1997 ), tumour invasion and metastasis factor ( Ives et al, 1998 ), multiple genes involved in the immune response such as CD18 and the DS cell adhesion molecule ( Yamakawa et al, 1998 ) and genes predisposing to precocious ageing such as Alzheimer's disease amyloid-associated A4 protein precursor ( Hol et al, 1998 ) may all contribute to the predisposition to leukaemia. Although gene dosage has not yet explained leukaemia, it may in part explain response to therapy. There are on chromosome 21 at least five genes involved in nucleic acid synthesis: phosphoribosyl aminoimidazole synthetase, phosphoribosyl glycinamide synthetase, cystathionine β synthetase, superoxide dismutase and reduced folate carrier ( Taub et al, 1996 , 1999; Belkov et al, 1999 ). Overexpression of cystathionine β synthase appears to confer exquisite sensitivity to cytosine arabinoside in the blasts of children with DS and AMKL, but not necessarily in direct proportion to the number of copies of chromosome 21 ( Taub et al, 1996 , 1999). In contrast, expression of reduced folate carrier confers sensitivity to methotrexate in direct proportion to the number of copies of chromosome 21 ( Belkov et al, 1999 ). Niebuhr et al (1974) localized the Down's syndrome critical region to chromosome 21q22, a region of interest for leukaemia as well. Within the 21q22.1–22·2 region is the CBFA2 (AML-1) gene. CBFA2 is translocated in t(8;21) FAB M2 myeloid leukaemia, in t(3;21) chronic myeloid leukaemia in blast crisis, in treatment-associated myelodysplastic syndrome (MDS) and in the molecular t(12;21) in favourable childhood ALL (reviewed in Nucifora & Rowley, 1995; McLean et al, 1996 ; Downing, 1999). Also CBFA2/AML-1 is implicated in the familial platelet disorder (FPD)/AML syndrome ( Legare et al, 1997 ; Song et al, 1999 ): in five families with this disorder, subtle deletions in CBFA2/AML-1 led to the haploinsufficiency of this gene and loss of function ( Song et al, 1999 ). Because CBFA2 is involved in megakaryocyte differentiation, MDS, AML and ALL, it is an attractive candidate gene for the DS 'leukaemia predisposition' gene. Although trisomy of a putative leukaemia-predisposition gene on chromosome 21 offers an intuitive mechanism for the frequency of leukaemia in DS, there is some evidence that loss of heterozygosity through deletion may be the critical event. Kempski et al (1997) have noted interstitial deletions on part of the long arm of chromosome 21 in the leukaemic cells in four of five DS patients with AMKL. Cavani et al (1998) found significantly increased crossover in pericentric regions of 21q in DS AML cases compared with DS with ALL and DS without leukaemia. Seghezzi et al (1997) described a case of chronic myeloid leukaemia in a DS patient in which maternal allelic loss was present in the leukaemic clone. Despite intense interest in the biology and epidemiology of leukaemia in DS, there was for many years a universal reluctance to treat leukaemia, particularly AML, and other lethal diseases in children with mental deficiency ( Churchill, 1989). In 1982, Baby Doe, a neonate with DS and multiple potentially lethal anomalies, was allowed to die in a nursery in Indiana ( Pless, 1983). Baby Doe's death engendered legislation mandating full access to medical care for all children in the USA. Thereafter, more or less systematic inclusion of children with DS in clinical trials became the standard. Only then were the unique features of leukaemia in DS revealed. Transient myeloproliferative disorder is a form of self-limited leukaemia that occurs almost exclusively in neonates with DS. The mean maternal age of children with TMD is 29 years and that of mothers with DS children with AML 33 years ( Iselius et al, 1990 ) . This difference probably reflects the older age of DS children with leukaemia rather than a predisposition of younger mothers to give birth to infants with TMD. Zipursky et al (1992, 1997, 1999) estimated that at least 10% of DS newborns have TMD. TMD may be present at birth or in stillborns. It usually is diagnosed by 3 weeks of age. TMD may be an incidental finding on a complete blood count (CBC). Typically, TMD regresses spontaneously by the age of 2 or 3 months. The peripheral white blood cell (WBC) count is relatively low in most cases of TMD and profound cytopenias are rare ( Zipursky et al, 1992 ). Often the percentage of blasts in the peripheral blood is higher than in the marrow ( Nagao et al, 1970; Shinbo et al, 1977 ). Although TMD is by definition transient, it has many clinical and laboratory features of malignancy. Light microscopy, histochemistry, immunophenotyping and electron microscopy indicate that the blasts in TMD are dysplastic megakaryoblasts ( Lazarus et al, 1981 ; Kojima et al, 1990 ). Characteristically, the cells are pleomorphic, ranging in size from 7 µm to 20 µm ( Bennett et al, 1985 ). The blasts may manifest cytoplasmic blebbing and megakaryocytic cytoplasmic fragments may circulate in the peripheral blood. Cells are negative for peroxidase and Sudan black, but are sometimes focally positive for periodic acid–Schiff (PAS) and non-specific esterase (NSE), which is fluoride inhibited. They do not react with butyrate acetate esterase ( Bennet et al, 1985 ). Sometimes they contain basophilic granules. There is also one well-documented case of basophilic differentiation of TMD megakaryoblasts in vitro ( Suda et al, 1985 ). The leukaemic cell in AMKL and TMD is derived from a common erythroid and megakaryoblastic progenitor ( Zipursky et al, 1992 ; Tchernia et al, 1996 ). The cells express the transcription factor GATA -1 and erythroid-associated gamma-globin and delta-amino laevulinate synthase mRNA ( Ito et al, 1996 ). Immunological evaluation may be limited by difficulties in aspirating cells. Reactivity with platelet-related glycoproteins IIb/IIIa (CD41, CD42 or CD61) or, on biopsy sections, with factor VIII antigen supports the diagnosis of FAB M7, but their absence does not rule it out. As with other forms of AML, up to half the cells may express B-lineage- or T-lineage-related antigens ( Kuerbitz et al, 1992 ; Smith et al, 1992 ; Slordahl et al, 1993; Creutzig et al, 1996 ). They usually express the myeloid-related antigens CD13 or CD33 ( Creutzig et al, 1996 ). Slordahl et al (1993) described leukaemic megakaryoblasts displaying markers of four lineages in an infant with DS. TMD is clonal. Studies of female infants for X-linked phosphoglycerate kinase (PGK), hypoxanthine guanosine ribosyl transferase (HGPRT) or immunoglobulin heavy chain and T-cell antigen receptor (TCR) indicate that TMD is derived from a single cell ( Kurahashi et al, 1991; Miyashita et al, 1991 ). Occasionally, blasts have clonal cytogenetic abnormalities ( Lazarus et al, 1981 ; Barnett et al, 1990 ; Iselius et al, 1990; Bhatt et al, 1995 ). An acquired chromosome 21 (i.e. tetrasomy 21) t(21;21) and i21 are most common ( Barnett et al, 1990 ). It is not clear why a disease that has all the cellular properties of leukaemia regresses. Presumably, there are host factors that enable regression. Infants with TMD have thrombopoietin (TPO) levels lower than age-matched controls and their blasts have receptors from c-MPL, the ligand for TPO ( Bonno et al, 1998 ). TPO levels rise as the TMD wanes. TPO levels correlate inversely with blast number but not platelet number. No other host factors have been investigated in this context. A rising conjugated bilirubin is a sign of serious liver disease in infants with TMD. TMD is sometimes complicated by hepatic fibrosis which is life-threatening and often fatal ( Becroft & Zwi, 1990; for a review, see Ruchelli et al, 1991 ; Schwab et al, 1998 ). The frequency of hepatic fibrosis is not known. It may be the cause of stillbirth in DS. In one series of eight patients with the complication, two were long-term survivors ( Miyauchi et al, 1992 ). Infants may die of fulminant liver failure while the TMD is regressing. Autopsy shows profound sinusoidal lobular and intralobular fibrosis and mild to moderate haemosiderosis, often with evidence of extramedullary haematopoiesis with excess numbers of maturing megakaryoblasts. Pancreatic fibrosis was noted in 8 out of 18 cases in another series ( Ruchelli et al, 1991 ) . Extensive myelofibrosis and generalized visceral fibrosis are occasionally present ( Becroft & Zwi, 1990; Zipursky et al, 1992 ). Miyauchi et al (1992) offered an interesting theory: the signs and symptoms of hepatic fibrosis may dominate TMD because the leukaemic blasts arise from fetal hepatic blood-forming cells. Normally, the transition from hepatic to medullary myelopoiesis accompanies the regression of TMD. Failure to make a timely transition to medullary haematopoiesis and overproduction of cytokines, such as platelet-derived growth factor or transforming growth factor β, by megakaryoblasts may lead to overwhelming fibrosis ( Miyauchi et al, 1992 ; Schwab et al, 1998 ). The management of TMD is conservative, most often involving watchful waiting or supportive care. When the WBC count exceeds 200 × 109/l, exchange transfusion or leucapheresis may avoid the complications of hyperleucocytosis ( Nakagawa et al, 1988 ) . In the rare cases of TMD where cytopenias or hyperleucocytosis become life-threatening, cytotoxic therapy may be considered. Low dose cytosine arabinoside (Ara-C) may be an appropriately conservative measure as it is sometimes curative in AMKL in DS ( Tchernia et al, 1996 ). There are as yet no data in this situation. In the face of imminent death from liver fibrosis, therapy may also be indicated. Options include low dose Ara-C for its efficacy in AMKL ( Tchernia et al, 1996 ), or standard AML therapy because it can reverse myelofibrosis in AMKL ( Cairney et al, 1986 ). Another strategy would be to inhibit fibroblastic proliferation with a biological response modifier such as interferon gamma-1b, which has been shown to reduce fibrosis in patients with idiopathic pulmonary fibrosis ( Ziesche et al, 1999 ). Again, there are no data for these suggestions. Because some neonates with trisomy 21 mosaicism are phenotypically normal, it is prudent to rule out DS by cytogenetics or fluorescence in situ hybridization (FISH) before initiating cytotoxic therapy for leukaemia in any neonate with leukaemia. It is estimated that about 10–20% of DS infants who develop AMKL or MDS had an antecedent TMD ( Lin et al, 1980 ; Morgan et al, 1985 ; Zipursky et al, 1992 ). Although the estimate of both the 10% incidence of TMD and the subsequent AMKL were derived prospectively from the Canadian DS registry ( Zipursky, 1996), both figures seem high as they imply that one in 50 infants with DS has both these disorders. Among recent series of DS patients with AML, 10–20% had a history of TMD ( Ravindranath et al, 1992 ; Zipursky et al, 1992 ; Creutzig et al, 1996 ). If one calculates that 1/300 DS children develops AML, 10–20% of whom have a history of MDS, the frequency of both disorders could be as low as 1/1500 to 1/3000. If leukaemia develops following TMD, it is almost always AMKL and it usually occurs between the ages of 1 and 3 years ( Lin et al, 1980 ; Nito et al, 1983 ; Ruiz-Argielles et al, 1986 ; Barnett et al, 1990 ; Zipursky et al, 1992 ). That both TMD and AML are usually phenotypically AMKL suggests that the second disease is derived from the first. In some cases, the karyotype of the AML is more complex but contains features of the original TMD, implying clonal evolution ( Morgan et al, 1985 ; Wong et al, 1988 ; Barnett et al, 1990 ). However, there are rare cases in which the AML has a translocation or deletion unrelated to the clone in the TMD. The morphology is not M7, and the age is usually over 5 years ( Morgan et al, 1985 ; Wang et al, 1987 ; Sato et al, 1997 ; Yamaguchi et al, 1997; Kounami et al, 1998 ). These clonally unrelated cases implicate host factors that predispose to the development of myeloid leukaemia rather than clonal evolution of residual leukaemic cells. Infants with a history of TMD warrant surveillance with complete blood counts for several years and parental education regarding the presentation of AMKL. In their review, Barnett et al (1990) could not discern any clinical, haematological or cytogenetic differences between those infants with TMD who went on to develop AML and those who did not. Table III lists the presenting features of AML, most often AMKL, in DS patients compared with non-DS patients. Estimates of the proportion of AML that is AMKL in children with DS vary from 40% to 100% ( Table III; Zipursky, 1996) . The 100% figure is probably an overestimate as 1/10 or 20 children with DS and leukaemia will have developed their leukaemia independently of their predisposing DS ( Table I). Many of the non-M7 cases are well-documented. In the population-based Nordic NOPHO-84 and NOPHO-88 studies, 13% of children with AML had DS ( Lie et al, 1996 ) . In sequential Children's Cancer Group (CCG) studies, the proportion of children with AML who have Down's syndrome has risen from 1–2% to nearly 10% ( Table IV) ( Robison, 1992 ; Lange et al, 1998 ). Progressively higher numbers of DS children with AML probably derive from several sources: (i) routine entry of children with DS and AML on large clinical trials after Baby Doe legislation ( Lange et al, 1998 ); (ii) reduced death from congenital heart disease and other anomalies, thus increasing the pool of DS children to develop leukaemia ( Levitt et al, 1990 ); (iii) reduced parental refusal of therapy ( Lie et al, 1996 ); and (iv) recognition of AMKL as a distinct form of leukaemia that in the past may have been misdiagnosed as ALL ( Bennett et al, 1985 ; Zipursky et al, 1987 ). Only the last factor accounts for the observed increase in DS patients with AML ( Table III) and the relatively stable proportion of DS ALL patients. Current estimates of the ratio of ALL to AML have altered from 4:1, i.e. the same as in non-DS children, to 1:1 ( Table I) ( Levitt et al, 1990 ; Robison, 1992; Lange et al, 1998 ), and in the first 3 years of life AML is 100 times more common than ALL ( Creutzig et al, 1996 ). The unique features of AML in DS patients are the young age, an antecedent MDS, FAB M7 or M6/M7 morphology and the response to therapy. Almost all AML in children with DS occurs between the ages of 1 and 5 years, with a median of 2 years ( Table III) ( Zipursky et al, 1987 ; Kojima et al, 1990 ; Levitt et al, 1990 ; Ravindranath et al, 1992 ; Creutzig et al, 1996 ; Lie et al, 1996 ; Lange et al, 1998 ). The blasts consistently express the myeloid surface antigens CD33 and/or CD13 or CD11b, but only rarely have the glycophorin A, glycoprotein IIb/IIIa or factor VIII reactivity ( Table III) ( Ravindranath et al, 1992 ; Fisher et al, 1994 ) . As with other forms of AML and TMD, lymphoid antigens are commonly expressed ( Kuerbitz et al, 1992 ; Smith et al, 1992 ; Creutzig et al, 1996 ). It has been established that 20–69% of AML cases in DS present with MDS ( Kojima et al, 1990 ; Zipursky et al, 1992 , 1997; Creutzig et al, 1996 ; Lange et al, 1998 ) ( Table III). The MDS is characterized by months of worsening thrombocytopenia followed by anaemia. Neutropenia and infection are rare ( Zipursky, 1996). Frequently, both erythroid and megakaryoblastic precursors are dysplastic and increased numbers of megakaryocytes in the presence of thrombocytopenia suggest ineffective megakaryopoiesis (for a review, see Zipursky et al, 1992 ). In the Berlin–Frankfurt–Munster (BFM) series, 17 out of 27 patients with AMKL had MDS. The time of progression to AMKL, i.e. to > 30% blasts, ranged from days to 30 months ( Creutzig et al, 1996 ). In the absence of therapy, progression to AMKL appears inevitable. However, one patient diagnosed with MDS at 22 months of age experienced two spontaneous regressions at 9 and 13 months before diagnosis of AMKL ( Creutzig et al, 1996 ). Although there is no evidence of an advantage to pre-emptive therapy during MDS, it is probably wise to initiate therapy when repeated platelet or erythrocyte transfusions are required to control bleeding or to treat anaemia. Response of MDS in DS patients appears to be as favourable as the response of their AML ( Ravindranath et al, 1992 ). The profile of cytogenetic abnormalities is different in DS and non-DS patients with AML ( Table III) ( Kaneko et al, 1981 ; Hecht et al, 1986 ; Groupe Francaise de Cytogenetique Haematologique, 1988; Ravindranath et al, 1992 ; Creutzig et al, 1996 ; Lange et al, 1998 ). About 25% of DS patients show only constitutional trisomy 21. The favourable t(8:21), t(15;17), inv(16) and the t(9;ll) translocations of AML are present in 40% of the non-DS patients with AML, but are exceptional among children with DS ( Table III). Reduced recombination is a characteristic of the non-disjoined chromosome 21, which may account for the relative absence of translocations ( Warren et al, 1987 ). Numerical changes, most commonly +8 and +21, occur in about one-third of the DS cases, i.e. twice as often as in the non-DS cases ( Creutzig et al, 1996 ; Lange et al, 1998 ). AKML-associated translocations, t(1;22) and t(1;3), do not occur in DS AMKL, but Creutzig et al (1996) noted five translocations involving chromosome 1q in their series of 40 DS patients. Other less frequent AML translocations such as t(5;7) and t(8;16) also occur ( Ma et al, 1997 ). Karyotype is not a prognostic factor in DS AML; even patients with monosomy 7 have responsive disease ( Bunin et al, 1991 ; Lange et al, 1998 ). An acquired 21, i.e. tetrasomy 21, in DS AML is as favourable as other cytogenetic variants, whereas acquired +21 in non-DS is a neutral or unfavourable cytogenetic variant ( Lange et al, 1998 ). The most remarkable feature of AML in DS patients is its extraordinary responsiveness to AML therapy ( Ravindranath et al, 1992 ; Kojima et al, 1993 ; Lie et al, 1996 ; Lange et al, 1998 ). Response to therapy for DS infants with AMKL is the same as that of DS patients with other FAB subtypes ( Creutzig et al, 1996 ; Lange et al, 1998 ). In contrast, AMKL is the only consistently unfavourable FAB subtype in non-DS AML ( Schaison et al, 1990 ; Barnard et al, 1997 ; Lange et al, 1998 ). Table IV compares the remission induction rate and event-free survival (EFS) among those with and without DS in recent series. In general, the DS patients treated according to AML protocols have significantly better outcomes than those who had no treatment or minimal treatment ( Levitt et al, 1990 ; Ravindranath et al, 1992 ; Creutzig et al, 1996; Lie et al, 1996 ). There are two exceptions. Tchneria et al (1996) reported that three out of seven DS patients with erythroblastic or megakaryoblastic leukaemia treated with low-dose Ara-C (10 mg/m2/d b.d.) were cured of their disease. Others ( Zipursky et al, 1992 ; Zipursky, 1996) reported sustained remissions in seven out of nine children with DS and MDS treated with low-dose Ara-C, vincristine and retinyl palmitate. In more recent studies, excess toxic deaths rather than relapse composed the most common event among those with DS and AML ( Lange et al, 1998 ). Argentinean investigators found the BFM protocols too toxic in their DS patients, as all 7 events in 11 patients were treatment-related deaths ( Zubizarreta et al, 1998 ) . CCG-2891, the only randomized comparison of intensive therapy to standard AML therapy in DS, indicates that DS patients benefit from therapy that is less intensive than that currently used for other children with AML ( Lange et al, 1998 ). In CCG-2891, in contrast to the non-DS patients, those with DS randomly assigned to the less intensive standard induction experienced significantly higher induction rate and event-free survival than those given an intensively timed induction ( Table IV) ( Woods et al, 1996 ; Lange et al, 1998 ). Neither autologous nor allogeneic marrow transplant conferred a better outcome than high-dose Ara-C consolidation therapy among the DS patients. The high mortality rate with intensive AML therapy and the responsiveness of AMKL/MDS to low-dose Ara-C in some patients with AMKL justify abandoning the generic AML therapy for patients with DS and investigating Ara-C-based strategies of diminished intensity. Low-dose Ara-C alone may be insufficient: the approach of Tchernia et al (1996) cured less than half the patients and the study of Zipursky (1996) had two out of seven patients with short follow-up. These figures seem lower than those in the large studies using standard AML therapy ( Ravindranath et al, 1992 ; Lange et al, 1998 ). Taub et al (1996 , 1999) have hypothesized that altered Ara-C metabolism accounts for the relatively good outcome of DS patients with AML. They demonstrate that, in the tetrazolium bromide reduction assay (MTT), DS myeloblasts are 10-fold more sensitive to Ara-C than non-DS myeloblasts. Mean intracellular Ara-C triphosphate is fivefold higher in DS myeloblasts than in non-DS myeloblasts. Epstein-Barr virus (EBV)-transformed cell lines from DS generated fivefold more Ara-C TP than normal diploid cell lines ( Taub et al, 1996 ). These metabolic differences result from reduced intracellular folate and deoxynucleotide pools. At best, they correlated roughly with overexpression of the cystathionine-β-synthase gene in DS cells ( Taub et al, 1999 ). Treatment of refractory or recurrent AML in DS has not been the subject of systematic study. Arensen and Ford (1989) reported that inclusion of DS patients in marrow transplant trials occurred at 25% of the predicted rate. Although the CCG-2891 study could show no benefit in transplantion in first remission, stem cell transplantation may be a consideration for consolidation therapy for those with recurrent disease. Rubin et al (1996) found that among 27 DS patients with AML (n = 11), ALL (n = 14) or aplastic anaemia (n = 2) 3-year relapse-free survival after transplant was 44% and death due to non-leukaemic causes was 39%. Compared with non-DS patients, those with DS appeared to be at risk of increased fatal pulmonary complications and airway problems requiring intubation. Although MDS, AKML and AKML/erythroleukaemia in infants and young children with DS are distinct entities, there is no archetypal form of ALL in DS. Sex, white blood cell count and haemoglobin are similar in DS and non-DS patients ( Table III). Median age is slightly higher in DS patients but distribution of age range is similar; mean platelet count is modestly lower and the proportion of B-lineage ALL is higher ( Robison et al, 1984 ; Kalwinsky et al, 1990 ; Levitt et al, 1990 ; Ragab et al, 1991 ; Robison, 1992; Pui et al, 1993 ; Dordelman et al, 1998 ; Heerema et al, 1999 ). T-cell ALL and FAB L3 ALL are uncommon among those with DS. There have been no cases of central nervous system (CNS) disease at diagnosis among 389 patients with DS and ALL ( Table V). It appears that about half the cases of ALL in DS have a normal karyotype ( Berger, 1997); +21 and +8 are common changes, whereas an isolated +X is an acquired abnormality almost unique to ALL in DS ( Pui et al, 1993 ). The prognostically unfavourable t(1:19), t(8;14), t(4;11) and t(11q23) are rare ( Alimena et al, 1985 ; Hecht et al, 1986 ; Groupe Francaise de Cytogenetique Haematologique, 1988; Kalwinsky et al, 1990 ; Pui et al, 1993 ; Watson et al, 1993; Dordelmann et al, 1998 ; Heerema et al, 1999 ). Hyperdiploidy occurs but is usually modest, ranging from 47 to 51 chromosomes. The favourable 51–58 hyperdiploidy is uncommon ( Pui et al, 1993 ; Heerema et al, 1999 ). Among 11 cases of ALL in DS, the prognostically favourable Tel/AML-1 fusion transcripts could not be detected by reverse transcriptase polymerase chain reaction ( Lanza et al, 1997 ). By contrast, trisomy 21 is the most common acquired numerical aberration in paediatric ALL with t(Tel/AML-1) ( Loncarevic et al, 1999 ). Table V shows the outcome of recent controlled therapeutic trials of ALL in children with and without DS. Most studies conducted in the 1970s and early 1980s showed significantly worse outcome for the DS patients ( Pui et al, 1981 ; Robison et al, 1984; Kalwinsky et al, 1990 ; Levitt et al, 1990 ; Dordelmann et al, 1998 ). Both induction failure and induction death rate in DS are high. Patients die of pneumonitis or sepsis and have increased incidence of hepatitis and pancreatitis ( Pui et al, 1981; Kalwinsky et al, 1990 ). Even without leukaemia, children with DS experience early death, both from congenital anomalies and impaired immunity ( Baird & Sadovnick, 1987 ; Levin, 1987). DS patients tend to be slow early responders, suggesting either that their blasts are inherently less chemosensitive or that there are host metabolic factors that determine drug usage. Belkov et al (1999) have established that the gene copy number of the reduced folate carrier, which maps to chromosome 21, determines the amount of methotrexate polyglutamates in leukaemic lymphoblasts. They hypothesized that the same mechanism may account for the increased toxicity of methotrexate in DS children with ALL ( Belkov et al, 1999 ). Table V shows that DS patients treated with current standard ALL therapy have outcomes similar to non-DS patients. Of note, Dordelmann et al (1998) compared those who could tolerate with those who could not tolerate full BFM therapies, whereas others divided them according to intention to treat ( Robison et al, 1984 ; Levitt et al, 1990 ; Ragab et al, 1991 ; Pui et al, 1993 ; Lange et al, 1998 ). Outcomes are similar. During maintenance therapy, DS patients show delayed clearance of methotrexate, poor haematological tolerance, excessive mucosal toxicity and skin rashes ( Blatt et al, 1986 ; Lejeune et al, 1986 ; Kalwinsky et al, 1990 ; Levitt et al, 1990 ; Ragab et al, 1991; Dordelman et al, 1998 ). This intolerance is believed to result from endogenously depleted folate stores caused by excessive purine synthesis ( Lejeune et al, 1986 ; Garre et al, 1987 ; Ueland et al, 1990 ). Delays in therapy are frequent and long. Supportive care requires dogged persistence and is heroic and expensive. Treating ALL in DS children is especially challenging. We have achieved a satisfactory working classification of the leukaemia syndromes in children with DS and have developed treatment strategies tailored to their disease and their tolerance for therapy. Population-based studies that will document the incidence of TMD and its evolution to AML are in progress. Differential display by microarray technologies may reveal what genes are turned off and on during regression of TMD and subsequent development of AML. So far, molecular studies have not yielded a unifying explanation for the predisposition for leukaemia in DS children. Nonetheless, the continued investigation of DS and leukaemia will undoubtedly provide answers to our questions about aetiology and pathogenesis of leukaemia in all children. 1 wish to thank Christine Curran for typing and Anna Meadows for reviewing this manuscript.