Current Hematologic Malignancy Reports

, Volume 5, Issue 4, pp 200–206

Prognostic Factors in Pediatric Acute Myeloid Leukemia


    • Children’s Mercy Hospital
  • Soheil Meshinchi
    • Fred Hutchinson Cancer Research Center
  • Alan Gamis
    • Children’s Mercy Hospital

DOI: 10.1007/s11899-010-0060-z

Cite this article as:
Radhi, M., Meshinchi, S. & Gamis, A. Curr Hematol Malig Rep (2010) 5: 200. doi:10.1007/s11899-010-0060-z


Acute myeloid leukemia (AML), a heterogeneous group of diseases with variable responses to the same therapy, comprises nearly a quarter of childhood acute leukemias. Although historically very few prognostic markers have been incorporated into therapeutic decision making in AML, recent advances in technology have enabled identification of numerous factors associated with disease outcome. This review provides a detailed analysis of most clinically relevant factors associated with disease outcome in childhood AML.




Childhood acute myeloid leukemia (AML) represents a relatively heterogeneous group of leukemias despite their common myeloid origin. Outcome for AML has improved over the past several decades as therapy has significantly intensified.

Historically, the prognosis for AML patients was determined by a few limited disease and host factors such as white blood cell count and age, but now we recognize a larger and more specific group of factors that can influence the outcome of AML, whereas many of the original prognostic factors have lost their significance with modern therapy. Over the past 10 to 15 years, major advances in identifying molecular and cytogenetic risk factors in AML therapy have provided insight into the molecular heterogeneity of the disease that may underlie the variable response. These discoveries, coupled with improved outcome with more intensive therapy and the associated treatment-related toxicity, have all combined to drive efforts to better define the prognostic risk factors in AML. With the delineation of several independent risk factors, efforts have been under way to stratify therapy based upon these factors, in order to reduce toxicity for those who can be cured with less intensity, to better target therapy for specific types of AML for which targeted agents are available, and to further refine therapy for those who remain at high risk of relapse despite intense therapy. Two subtypes of AML, t(15;17)-associated acute promyelocytic leukemia and Down syndrome–associated AML, have been treated differently for the past decade. Now, most protocols in the childhood cooperative groups are further stratifying therapy based upon cytogenetic, molecular, and response-based factors. This article details these risk factors and potential future factors, to aid the clinician and the researcher as they approach this group of myeloid leukemias, with its increasingly recognized heterogeneity.

Traditionally, the major prognostic factors used to predict response in pediatric AML are morphology/phenotype (eg, FAB M3), cytogenetics (eg, monosomy 7, del5q, t[15;17]), and response to therapy [1]. These risk factors have helped to stratify AML into high-risk and low-risk types. For example, acute promyelocytic leukemia with the t(15;17) translocation has been successfully treated with all-trans retinoic acid (ATRA), with overall survival (OS) up to 90% and event-free survival (EFS) of 71% without the use of myeloablative transplant [2]. Similarly, in the Children’s Oncology Group (COG) Study 2891, Gamis et al. [3] showed superior survival in Down syndrome (DS) patients compared with non-DS patients, despite using a less-intense standard-timing therapy than was used in the non-DS AML cohorts. This improved outcome was a result of significantly greater remission rates, equivalent or lower toxicity, and less subsequent relapse. Importantly, in the unique DS cohort with a unique AML (GATA1 mutation), an important covariable affected outcome [4]: increasing age (specifically, age >4 years) was associated with a significantly worse outcome [3]. Yet to be determined is whether it was age that affected outcome or whether it was the presence or absence of GATA1 mutations in the AML clone. Newer data are now showing that a multitude of host factors—not just factors related to the therapy given or the biology of the leukemia—can have an impact on disease.

Host Factors

Several host factors, including age, gender, race, body mass index (BMI), and host polymorphism, have been implicated in disease outcome [57]. Racial differences have been demonstrated to be associated with disease outcome, as African American patients had a significantly worse outcome than their white counterparts [8]. These studies, confirmed in other studies [9], demonstrated that poor survival was primarily due to higher rate of relapse. Underlying mechanisms for disease resistance have not been specifically elucidated but are thought to be due to host polymorphisms. Several studies have demonstrated that polymorphisms in genes involved in drug metabolism, DNA repair, or regulation of hematopoietic development may contribute to variable response to therapy [10]. A null phenotype of the glutathione S-transferase enzyme has been shown to affect survival adversely through excessive toxicity [5]. More recently, the presence of a polymorphism in the Wilms’ tumor 1 (WT1) gene was shown to be associated with favorable outcome [11].

Extremes in patients’ BMI was clearly shown to affect outcome in AML. In a retrospective analysis of the data from CCG-2961 (n = 768), Lange et al. [12] found that both extremely overweight and underweight children had survival inferior to that of children with BMI in the range of the 11th to 94th percentiles. Possible causes that may have contributed to the inferior outcome included early treatment-related mortality, mostly from infection.

Despite early suggestions that AML patients less than 1 year of age had poor outcome, more recent studies from large cooperative groups have shown that young patients with AML (in contrast to ALL) have better survival rates than older patients because of a lower relapse rate [13].

Response to Therapy

Response to induction therapy is intuitively a major prognostic factor in AML. This strong factor was included in both the original and revised recommendations of the International Working Group on Acute Myeloid Leukemia [14]. Response to therapy has traditionally been based upon morphologic exam of the marrow, but these methods have a limit of detection that misses minimal residual leukemia, which may be more prognostic. More recent methods, including methods based on multiparameter flow cytometry and polymerase chain reaction (PCR), are more sensitive [15, 16].

The impact of therapy response upon survival and relapse rates was clearly illustrated in the Medical Council Research (MRC) AML 10 trial (n = 1711; ages 0–55 years). This study ascertained that patients with 5% to 15% blasts (partial remission [PR]) after the first course fared better than patients with 15% to 20% blasts, who fared as poorly as patients with more than 20% blasts (OS 42%, 23%, and 22%, respectively). Importantly, the PR patients had a high remission rate (89%) after induction course II (INDII), resulting in an OS for PR patients that was only slightly inferior to the rate of patients with complete remission (CR) and was identical if they entered remission after INDII, whereas patients with resistant disease (>15% blasts) had an extremely poor prognosis even if they entered remission after INDII [17]. Of note, this response-based adverse impact upon OS was completely abrogated in the presence of favorable-risk cytogenetics: t(8;21), inv(16), or t(15;17). Rapid response to induction chemotherapy was also shown to be an independent predictor of superior OS in AML in other studies of both pediatric and adult patients [18, 19].

Minimal residual disease (MRD) has also been investigated in pediatric patients with AML [20, 21]. Both studies reported the significance for overall survival of negative MRD using multicolor flow cytometry at the end of first induction chemotherapy. Patients who were MRD positive at the end of first induction fared significantly worse than patients who were MRD negative. Langebrake et al. [21] also showed that in 95 patients, MRD positivity at three time points fared worse than those with positivity at no more than two time points. The COG also used four-color flow cytometry data from two consecutive clinical trials and specifically showed that high MRD (defined as >0.5%) at the end of induction and before intensification increased the risk of relapse severalfold [15, 22•]. The impact of MRD status on relapse was also seen in patients who received stem cell transplantation (SCT) [23]. Patients who were MRD negative after SCT had a much lower relapse rate than patients who were MRD positive, regardless of the type of transplant or their postinduction MRD status.


Cytogenetics in AML is widely accepted as one of the major prognostic factors in all age groups and was the initial factor considered in the reclassification of AML by the WHO [5, 6, 24].

As evidenced repeatedly across a variety of AML regimens, the use of cytogenetics alone can divide AML patients into three broad, major risk groups impacting relapse risk (RR), disease-free survival (DFS), and OS: favorable, intermediate, and unfavorable. Favorable AML cytogenetics include the core binding factor (CBF) leukemias t(8;21) (RUNX1-ETO) and inv(16) (MYH11-CBFB), as well as t(15;17) (PML-RARA). Unfavorable cytogenetics include complex cytogenetics (three or more distinct cytogenetic abnormalities in a leukemic clone), monosomy 7, monosomy 5, del(5q), and abnormal chromosome 3. The intermediate-risk cytogenetics include all the remaining chromosomal abnormalities commonly seen in AML, such as +8, + 21, and 11q23 (MLL)–associated abnormalities, as well as AML patients with a normal karyotype [5, 25, 26]. The MRC AML-10 trial accrued a total of 1966 patients up to the age of 55, including 364 children. Overall, the OS was 65% in the favorable-risk group, 41% in the intermediate-risk group, and 14% in the unfavorable-risk group. The RR was 35%, 51%, and 76% respectively. In the pediatric patients (age <15 years), the OS values were 78%, 55%, and 42%, and the RR values were 32%, 40%, and 61%, respectively. Patients with favorable cytogenetics were found to have a better response to induction, lower RR, and improved OS. The 5-year OS and RR were 63% and 37% for t(15;17), 69% and 29% for t(8;21), and 61% and 42% for inv(16). The comparable rates for patients with unfavorable cytogenetics were 21% and 68% for those with complex cytogenetics, 10% and 80% for monosomy 7, 12% and 85% for del(5q), and 4% and 90% for monosomy 5. Rates for those in the intermediate-risk group were between those for the other two groups, with a variable OS between 29% and 59% depending upon the specific cytogenetic subgroup.

Patients with poor-risk cytogenetics are not a uniform group; their survival rates vary. In a large, international retrospective analysis of AML cytogenetics in 258 children, Hasle et al. [27] reported on survival of children with AML and the chromosome 7 abnormalities monosomy 7 or del(7q) with or without other cytogenetic changes. Outcome was particularly poor in patients with monosomy 7 and inv(3)(q21q26), monosomy 5/del(5q), or +21, with OS of 5%, compared with 34% for those who had monosomy 7 without additional factors. The OS for those with del(7q) with or without additional factors was even higher; in current regimens, this group is not included among the high-risk karyotypes. A recent manuscript from Löwenberg’s group [28] describes the grouping of monosomal karyotypes into one prognostic group, which overall conveys a very poor prognosis [29]. This idea has not yet been validated in a pediatric cohort.

It is important to understand the impact of the therapeutic regimen upon the survival influence of cytogenetic abnormalities. For pediatrics, this is most clearly illustrated by 11q23 abnormalities, which once had a poor prognosis [30]; this impact upon survival has been lost with increasingly intensive therapeutic regimens [31•]. Also, as cytogenetic techniques improve and molecular detection has been incorporated, subsets of cytogenetic abnormalities can be better segregated by their impact upon prognosis. In the 11q23-related leukemias, t(10;11) and t(6;11) had worse EFS (HR, 2.5 and 2.2, respectively), whereas t(1;11) conferred a very favorable EFS (HR, 0.1); in the historically poor-risk leukemias containing t(6;9), only those with FLT3-ITD mutations had a poor prognosis [32]. Specific molecular abnormalities are being explored for prognostic impact and, perhaps more importantly, for therapeutic targeting.

Molecular Risk Factors

There are several limitations on the use of karyotype as a risk-stratification tool. These include failed cytogenetic analyses, the presence of cryptic chromosomal rearrangements, and, notably, the fact that about 20% of children with AML do not have identifiable karyotypic alterations. Overall, cytogenetics identifies only a small population of high-risk patients. More recently, genomic alterations of a number of genes involved in hematopoietic development have been identified and have proven to have prognostic significance. Adding these factors to the cytogenetic data increases the proportion of patients with specific risk factors predicting outcome. Currently, mutations in three genes have been proven to be associated with outcome in childhood AML: the Fms-like tyrosine kinase 3 (FLT3) gene, the nucleophosmin (NPM) gene, and the CCAAT/enhancer-binding protein-alpha (CEBPA) gene. As patients with AML who are in morphologic remission may harbor residual disease when tested for molecular targets, molecular remission (CRm) has been suggested by the International Group on AML for identification of disease response in those with a specific molecular marker [14].

FLT3 Mutations

FLT3 is a receptor tyrosine kinase expressed on hematopoietic progenitors. The first report that indicated the presence of genetic aberrations in the FLT3 gene was in 1996 [33]. FLT3 mutations are the most common somatic mutations observed in AML, and their presence (with 30% to 35% incidence in adult patients) may be a prognostic factor for outcome [34]. Several studies have documented the prevalence and prognostic significance of FLT3 mutations in pediatric AML [34, 35]. The overall prevalence of FLT3 mutations was about 20%, with FLT3/ITD comprising about 12% and FLT3/ALM, about 8% [34]; an age-related increase in FLT3/ITD prevalence is demonstrated by a prevalence less than 2% in younger patients (<2 years) and nearly 20% in teenagers.

FLT3 mutations have been associated with a poor prognosis in AML patients because of a higher risk of relapse in both children and adults [33, 34]. With time, this effect has been further delineated, and now this poor risk is restricted primarily to those with high FLT3-ITD allelic ratios [34]. In children, a high allelic ratio is defined as greater than 0.4. Progression-free survival was 16% in those with high FLT3-ITD allelic ratios, compared with 51% for those with FLT3 point mutations, 55% for those without FLT3 mutations (wild-type), and 72% with low allelic ratios. Further, in AML patients with a high FLT3-ITD allelic ratio, the poor prognostic impact upon progression-free survival could be completely abrogated by allogeneic transplantation, which is now the standard of care for consolidation of this particularly high-risk group.

NPM1 Mutations

NPM1 encodes a nuclear protein and is shown to be involved in biosynthesis of ribosomes, prevention of aggregation of proteins in the nucleus, and regulation of centrosome duplication. Several mutations of the NPM1 gene have been associated with pathogenesis of malignancies. NPM1 mutation is present in both adult and pediatric AML with variable prevalence. In general, AML patients with NPM1 mutation have improved survival and patients are typically classified as low-risk. Hollink et al. [36] found an EFS and OS of 80% and 85% in a cohort of pediatric AML patients with NPM1 mutations, compared with 39% and 60%, respectively, in those without these mutations.

CEBPA Mutations

CEBPA is a gene encoding a protein member of the family of basic region leucine zipper (bZIP) transcription factors that plays an essential role in granulopoiesis. Loss of activity by either mutation or epigenetic silencing can result in a block in normal hematopoietic differentiation [33, 37]. CEBPA mutations confer a favorable prognosis in cytogenetically normal AML patients, regardless of the age of the patient [3840]. In one of the largest studies of pediatric AML, Ho et al. [40] reported the outcome of 847 patients treated over three consecutive cooperative clinical trials. CEBPA mutation was an independent risk factor for outcome, with a 5-year EFS of 70% versus 38% and a cumulative incidence of relapse from CR of 13% versus 44% for those with and without CEBPA mutations.

Thus, at this time, for AML patients with traditionally intermediate-risk cytogenetics, including those with normal karyotype, three gene mutations permit further risk stratification: high FLT3-ITD allelic ratio indicates poor risk, and NPM1 and CEBPA suggest good risk. Two other mutations—WT1 and c-KIT—have been analyzed in numerous studies but have not been shown to be prognostic in pediatric trials.

WT1 Mutations

The WT1 gene, located in the 11p13 locus, encodes a zinc-finger protein. The gene is constitutively expressed in the renal tissue, and in some patients the mutated form (associated with Wilms tumor) is overexpressed in acute and chronic leukemia. Early studies suggested that a high level of expression of WT1 was significantly associated with relapse and worse outcome [24]. However, more recent studies in pediatric AML have failed to show any significance of the WT1 expression level [41]. Genomic alterations have also been identified in the WT1 gene and its presence has been implicated in clinical outcome in AML. Paschka et al. [42] identified mutations in the WT1 gene in 11% of patients and demonstrated that it was an independent prognostic factor: patients with WT1 mutations had a 3-year DFS of only 13%, versus 50% in those without the WT1 mutation. More recent adult studies failed to confirm the association of WT1 mutations with worse outcome [43]. A similar large-scale study in pediatric AML, which screened 842 patients for the presence of WT1 mutations, demonstrated a prevalence of 8%, with nearly 30% overlap with FLT3/ITD [44]. Although the presence of WT1 was associated with worse survival (35% vs 52% OS, and 27% vs 41% EFS), this prognostic impact was lost with correction for the presence of FLT3-ITD mutations. Among those with the WT1 mutation but without the FLT3-ITD mutation, no effect on outcome was seen (OS, 51% vs 54% without these mutations; P = 0.5). Similarly, the OS was not different in patients with normal karyotype and negative for FLT3/ITD (OS, 40% vs 55%; P = 0.23). Thus, it was concluded that the prognostic significance of the WT1 mutation is absent when corrected for other mutations.

C-KIT Mutations

Activating mutations in the c-KIT receptor gene have been reported in AML, myelodysplastic syndrome, and mastocytosis, with a prevalence of 2% to 15% [5]. Recent work by several groups has shown that the c-KIT receptor mutation is prevalent in the core binding factor (CBF) leukemias t(8;21), and inv(16) in both pediatric and adult patients [4547]. Results from these studies suggested that this mutation has an adverse impact on survival when present in these leukemias. The Cancer and Leukemia Group B (CALGB) found a c-KIT mutation prevalence of 22% to 29% in CBF leukemia patients, with an adverse impact upon RR despite similar remission rates [47]. More recent and more comprehensive studies from the same author have brought the significance of KIT mutations into question in t(8;21) [48].

A definitive study in pediatric patients with CBF AML demonstrated that although KIT mutations are prevalent in childhood CBF AML, their presence does not indicate worse outcome [49•]. Investigators performed KIT mutational analysis (exon 8 and exon 17) on diagnostic specimens from 203 pediatric patients with CBF AML enrolled in four pediatric AML protocols. KIT mutations were detected in 38 of 203 patient samples (19%; 95% CI, 14%–25%), of which 20 (52.5%) of 38 involved exon 8, 17 (45%) of 38 involved exon 17, and 1 involved both locations. Patients with KIT mutations had a 5-year EFS of 55%, compared with 59% for patients with wild-type KIT (P = 0.86). The conclusion of the study was that c-KIT mutation lacked any prognostic significance in pediatric AML.

Combining Risk Factors to Establish Risk Groups

As already mentioned, when two risk factors coexist in the same patient, their impact may be lost. Wheatley et al. [17] in the MRC created a robust model combining cytogenetics and response; it is the basis for many AML risk classifications today. In this model, three risk groups (favorable, intermediate, poor) were defined, first by cytogenetics and then by morphologic marrow response. In general (with recent refinements), favorable risk factors include t(8;21), inv(16), CEBPA, or NPM mutations. Factors suggesting high risk generally include (in pediatrics) monosomy 7, monosomy 5, del(5q), high FLT3-ITD allelic ratio, and poor morphologic response (>15% remaining blasts after the first induction course). In the upcoming COG de novo phase 3 trial, MRD will also be incorporated, to further refine the risk classification for these patients and influence subsequent therapeutic choices.

Impact on Therapy Stratification

As noted, in the past risk factors have been used solely for prognostication at the beginning of most AML protocols. Over the past decade, however, these risk factors have begun to influence the therapy chosen. For example, the t(15;17) PML-RARA + acute promyelocytic leukemias and the DS-associated leukemias in children less than 4 years of age both have completely separate standard-of-care regimens and research protocols. More recently, the cytogenetic and response-based risk groups identified above have influenced the use of transplantation; the favorable-risk groups receive no transplant in first remission, the intermediate-risk groups use matched, related donor transplants if available, and the high-risk groups will often use transplants from related or unrelated donors. As new targeted agents emerge (eg, FLT3 inhibitors), further variations in therapy will be developed based upon these and future risk factors.


AML is a heterogeneous disease arising from a variety of genetically acquired mutations, many of which have not yet been detected. As diagnostic techniques have evolved and therapy has intensified, the traditional prognostic methods of morphology and host factors at presentation are no longer adequate to classify patients with AML. Over the past two decades, a growing list of prognostic factors has helped greatly in understanding this disease better and treating it more precisely.


No potential conflicts of interest relevant to this article were reported.

Copyright information

© Springer Science+Business Media, LLC 2010