Collaborative Efforts Driving Progress in Pediatric Acute Myeloid Leukemia

INTRODUCTION

Outcomes for children with acute myeloid leukemia (AML) have improved significantly over the past 30 years. During this time, multiple international cooperative groups have contributed to an evolving treatment strategy that consists of four to five courses of intensive myelosuppressive chemotherapy; stem-cell transplantation (SCT) is reserved for a subgroup of patients.^1–3 Pediatric AML therapy challenges patients, families, and care providers because of a high incidence of severe and dose-limiting short- and long-term toxicities. Given that AML in children is relatively rare, with an incidence of approximately seven occurrences per 1 million children annually, multicenter clinical trials are required for continued progress to define new therapies and new approaches to ameliorate adverse effects. Achievements in pediatric oncology stem from a willingness to collaborate and coordinate research efforts. Starting in the 1980s and 1990s, several (inter-) national study groups formed with the common goal of improving outcomes among children with cancer through cooperative research.

Collaboration has evolved to encompass different approaches by different cooperative groups. The diversity in approaches allows cooperative groups to ask scientific questions in parallel, which provides multiple opportunities for innovation and validation. A brief history of each major cooperative group is summarized in Table 1, and major recent findings as published in literature are summarized in Table 2.^4–16 As cooperative groups do more to integrate their efforts, it is important to examine and understand how each has grown, adapted, and evolved to find common ground in their interpretation of local and global data sets. This review summarizes important achievements in the field of pediatric AML and the lessons learned through both parallel and integrated international efforts.

PROGNOSTIC FACTORS AND RISK-GROUP STRATIFICATION

There is general agreement across groups about the definition of high-risk (HR) AML. Groups differ in the use of low, standard, and intermediate risks to designate all other types of AML. For the purposes of this review, disease in all patients with non-HR AML will be called standard-risk (SR) AML. Novel data from the genomic era show that only a limited number of gene mutations are required for AML pathogenesis compared with solid tumors,¹⁷ but their spectrum may be broader than the class I (proliferative) and class II (blocking) mutations postulated by Kelly and Gilliland¹⁸ as required in AML pathogenesis. Figure 1 summarizes many of the novel insertion and deletion events recently identified through next-generation sequencing approaches as well as the well-known large translocation events. As a result of efforts in sequencing and cytogenetics, new molecular subsets have been identified through independent and combined data analysis across cooperative groups. Intergroup validation of disease markers is key to building confidence in the true prognostic impact of the marker. This section will not review all genetic events in AML but will focus on those identified across different groups.

The core-binding factor (CBF) mutations—inversion16(p13;1q22), t(16;16)(p13;q22), and t(8;21)(q22;q22)—are recognized by all study groups as markers of SR AML. A recent North American and European collaboration confirmed that mutations in the nucleophosmin1 gene (NPM1)^19,20 and CEBPα gene also predict a favorable prognosis.^21–23 Including CBF, NPM1, and CEBPα mutations, the consensus SR group now comprises roughly 30% to 40% of patients with childhood AML, which is considerably larger than in adult AML.²⁴

Among the common markers of HR disease are the FLT3 mutations, a family that includes internal tandem duplications (FLT3/ITD) and point mutations in the kinase domain (FLT3/KD). FLT3/ITDs occur in approximately 10% to 20% of pediatric AML occurrences, and there is a wealth of data on their significance as a poor prognostic marker.^25,26 Approximately 10% of FLT3 mutations are either gained or lost at relapse, which suggests that FLT3 mutations can be late subclonal events, and they are not always the driver of prognosis and response.^27,28 However, the reports of secondary resistance mutations after treatment with FLT3 inhibitors suggests that these are indeed driver mutations.²⁹ The prognostic significance and biologic role of FLT3/KD mutations is less clear.³⁰

Although FLT3/ITD, monosomy 7, and monosomy 5 are longstanding traditional markers of HR disease, the prognostic impact of KMT2A (previously known as MLL) rearrangements (which occur in 20% to 24% of all patients with childhood AML) has only recently been clearly defined. KMT2A is notoriously promiscuous and has more than 120 translocation partners described.^31,32 A recent large, retrospective effort from 11 international groups included data from greater than 750 pediatric patients with KMT2A-rearranged AML and showed that the translocation partner predicts clinical outcome. AML with t(1;11)(q21;q23) was rare but did exceptionally well, whereas disease with t(6;11)(q27;q23), t(10;11)(p12;q23), or t(10;11)(p11.2;q23) had a dismal outcome.³¹

Repetitive rearrangements that involve NUP98 have been identified in European and North American cohorts. The cryptic translocation NUP98/NSD1 occurs in 15% of patients with cytogenetically normal AML and is associated with a high frequency of FLT3 mutations and a dismal outcome.^33,34 In a collaborative effort, North American and European groups also identified NUP98/JARID1A in acute megakaryoblastic leukemias (AMKL).³⁵ This differs from the well-known t(1;22)(p13;q13) translocation, which results in the RBM15/MLK1 fusion product that is mainly present in infants with AMKL.^36,37 The recently describedCBFA2T3/GLIS2 fusion transcript, found both in non–Down syndrome AMKL and in other AML subtypes, confers a poor outcome in North American and European cohorts.^38,39 As a consequence of these findings, a large international cooperative effort has been established for genome-wide analysis of AMKL.

Infants with AML can also be characterized by a translocation t(7:12)(q36;p13), which is a recurrent translocation involving the MNX1/ETV6 gene in AML. This abnormality is associated with poor outcome.⁴⁰ Recently, other ETV6 abnormalities in both European and North American cohorts were reported at scientific meetings, but these data are not published yet.

The t(6;9)(p22;q34) and the t(8;16)(p11;p13) translocations predict a poor prognosis across cooperative groups. A fusion between the DEK and NUP234 genes results from t(6;9)(p22;q34) translocation. This abnormality occurs in older children and has a poor clinical outcome, which may improve with SCT.^41,42 The t(8;16)(p11;p13) translocation may have an age-dependent impact on prognosis and fuses KAT6A to the CREBBP gene. In very young children (28% of patients are diagnosed in the first month of life), this translocation is associated with spontaneous remission and warrants a watch-and-wait strategy before initiation of chemotherapy. Older children require intensive chemotherapy.⁴³

Despite these findings, there is not always consistency across groups. NPM1 mutations may occur concurrent with FLT3/ITD. The Medical Research Council (MRC) results suggest that the NPM1 mutation with FLT3/ITD predicts a poor prognosis.^44,45 Others have shown that the presence of an NPM1 mutation may abrogate the negative effects of FLT3/ITD.^19,46 Mutations in WT1 have yielded variable outcome reports from Europe (poor prognosis) and North America (no prognostic impact).^47,48 In Europe, rare structural abnormalities in EVI1 failed to demonstrate prognostic significance.^49,50 However, in a Japanese series, EVI1 gene expression was related to outcome, especially in patients with KMT2A/AF9 rearrangements.⁵¹ Although chromosome 7 abnormalities have long been considered a poor prognostic marker, Hasle et al⁵² performed a large retrospective study that demonstrated that monosomy 7 has a worse outcome than del7q.

RAS pathway mutations are present in approximately 20% of pediatric patients with AML and mainly occur in KMT2A-rearranged cases, and AML with inv(16) and NPM1-mutated AML.^53,54 NRAS mutations (8%) and KRAS mutations (13%) are the most common RAS pathway mutations.^53,55 To date, RAS mutations are not associated with a prognostic influence or response to therapy in either North American or European cohorts. However, this should not preclude the study of MEK inhibitors for RAS pathway–mutated AML.^56,57

RESPONSE ASSESSMENT AND RISK ASSIGNMENT

Minimal residual disease (MRD), as defined by multidimensional flow cytometry after induction chemotherapy in AML,^58,59 is prognostic in childhood AML and predicts relapse risk.^60–62 Most groups use a combination of genetic abnormalities and response as detected with multidimensional flow cytometry to allocate patients to HR treatment, which includes the need for SCT. Monitoring of MRD by molecular markers (eg, fusion genes, mutations) is currently being standardized in Europe by several cooperating reference laboratories. These methods may allow enhanced MRD detection and early identification of a molecular relapse.^38,62,63

PROSPECTIVE COLLABORATIVE CLINICAL TRIALS AND STANDARD THERAPEUTIC APPROACH

Several cooperative groups have recently published single-arm or randomized clinical trials in childhood AML, as summarized in Table 2. Collaboration among cooperative groups has led to rapid accrual to randomized trials, and the coordinated efforts of multiple cooperative group studies permits a validation and comparison of results so that the best therapies may be assimilated into future trial designs. As we begin to apply targeted therapies to patients in increasingly smaller subsets of molecularly defined disease, coordination of research questions across cooperative groups becomes even more critical. In this section, the collective work to define a standard therapeutic approach to childhood AML is discussed.

PLANNED AND ONGOING CLINICAL TRIALS

In Table 3, the planned and ongoing clinical trials from the various collaborative groups are summarized. Of interest, several European groups have developed collaborative studies to increase accrual and permit randomized comparisons. The I-BFM-SG is working toward a pan-European backbone for all patients in an effort to standardize approaches and to permit intergroup comparisons and clinical studies in small subgroups. In its last three trials, COG has adopted an approach close to that of the MRC.

SEPARATE TREATMENT PROTOCOLS FOR SPECIFIC SUBGROUPS OF PEDIATRIC AML

TREATMENT FOR RELAPSED AML

Treatment outcome at relapse is dependent on cytogenetics and the interval between diagnosis and relapse.^82,105 These factors are not independent from each other; patients classified as good risk tend to experience relapse later than patients classified as poor risk in their initial treatment period. Several reports showed the poor outcome of relapsed AML in general, despite intensive reinduction and SCT.^80,105–110 The current concept of treatment is intensive reinduction with one or two courses followed by SCT once in CR or in aplasia.⁸⁰ The I-BFM-SG study Relapsed AML 2001/01 was developed in 1999 to study the optimal remission reinduction regimen for relapsed AML. At that time, several groups used high cumulative doses of anthracyclines at initial diagnosis, and it was not clear whether anthracyclines would be beneficial in reinduction at relapse. Therefore, the study randomly assigned patients to FLAG or FLAG plus DNX. The study showed that early treatment response improved when DNX was added to conventional chemotherapy.⁸² OS did not improve for the total group of patients, but the study was not powered to demonstrate that. Of interest, OS improved significantly for the subgroup of patients with CBF AML. Relapsed AML 2001/01 also showed the prognostic significance of early treatment response in pediatric patients with relapsed AML.¹¹¹ Although the I-BFM-SG has chosen to run large, randomized studies of pediatric AML, relapsed AML in COG is considered an indication for studies with experimental agents. It would be beneficial, however, if consensus could be reached on a transatlantic platform for reinduction to which novel agents could be added to expedite recruitment. However, DNX, the favorite anthracycline in relapsed AML in Europe, is not registered in North America, which hampers this approach.

NEW AGENT STUDIES

Drug development is typically achieved through collaborative study groups or consortia, such as Innovative Therapies for Children with Cancer, Therapeutic Advances in Childhood Leukemia, I-BFM-SG, and COG. With the exception of gemtuzumab ozogamicin in Europe, however, no new drugs were licensed for AML in the recent past. A more detailed summary of potential new agents was recently published elsewhere.^112,113 Table 4 summarizes some of the major classes and findings.^{9,29, 108,114–135}

SUPPORTIVE CARE

Without doubt, the advancements in supportive care have contributed significantly to the improvement in outcome for pediatric patients with AML, given the requirement for intensive chemotherapy that is needed to treat AML. Series in the past showed relatively high mortality related to leukemia and treatment complications.^136,137 Patients with AML are at particular risk of viridans group streptococcal and Gram-negative bacteremia and invasive fungal infections. Candida and Aspergillus are prominent causes of fungal infection, although other fungi, such as Scedosporium, Fusarium, and Mucor, are emerging. In terms of infection prevention, approaches evaluated include prophylactic antimicrobial agents, prophylactic G-CSF, and nonpharmacologic approaches.^138–140 Major questions relate to the risk-benefit ratio for antimicrobial prophylaxis to prevent invasive bacterial and fungal infection. In terms of antibacterial prophylaxis, it is unknown whether the risks of prophylaxis, which include Clostridium difficile infection, invasive fungal infection, and promotion of antibacterial resistance, are outweighed by the benefits. However, a Cochrane review that summarizes the available data provides evidence for a reduction in mortality.^140,141 The best agents to use for antifungal prophylaxis are also unknown but include fluconazole; extended-spectrum triazoles, such as voriconazole and posaconazole; echinocandins, such as micafungin and caspofungin; and amphotericin products.^139,142 To address these knowledge deficits, COG is conducting three randomized, controlled trials in patients with AML. One study compares levofloxacin prophylaxis versus no levofloxacin prophylaxis in intensively treated patients with leukemia and evaluates the evolution of resistant pathogens during exposure to levofloxacin. A second randomized trial compares fluconazole prophylaxis versus caspofungin prophylaxis during each cycle of neutropenia in patients with newly diagnosed or relapsed AML. The third study is a double-blind, randomized trial to evaluate if chlorhexidine bathing can reduce central line–associated bloodstream infection. Although prophylactic G-CSF reduces infection rate and infection-related mortality, widespread use has not been promoted, because it does not influence OS.¹⁴³ Moreover, a placebo-controlled trial in adults did not show survival benefit despite quicker neutrophil recovery.¹⁴⁴ G-CSF may even result in a higher relapse incidence in patients with AML that expresses the isoform IV of the G-CSF receptor.¹⁴⁵

Another major supportive care issue faced by patients with AML is hyperleukocytosis, and approximately 20% of children and adolescents with AML have an initial WBC count greater than 100 × 10⁹/L at diagnosis. Pulmonary leukostasis, CNS ischemia and hemorrhage, and early deaths are common in these patients. Patients with hyperleukocytosis should receive aggressive hydration; administration of rasburicase; optimization of coagulation, which includes the use of fresh frozen plasma and platelet transfusion; and urgent administration of chemotherapy. The role of leukopheresis is controversial in improving outcomes and does not seem to improve survival.¹⁴⁶ Creation and dissemination of clinical practice guidelines to manage supportive care for patients with AML should be a priority. Recent international collaborations have been focused on developing guidelines for the management of pediatric fever and neutropenia¹⁴⁷and for mucositis prevention.¹⁴⁸

THE FUTURE OF INTERNATIONAL COOPERATIVE GROUP COLLABORATION

In the 1980s and 1990s, institutions across the globe organized into cooperative groups to better serve the clinical and scientific challenges in childhood cancer. Though initially these groups functioned independently, the interest and imperative for larger international networks of investigators has grown in response to the requirement for increasingly large and complex trial designs. Prospective clinical trials, with randomized questions when possible, remain the foundation for progress in improved outcomes for children with AML. As molecular aberrations define smaller and smaller subsets of childhood AML that require a targeted or a risk-adapted approach, the need for intergroup trials has been amplified. More and more countries and institutions are joining cooperative groups in Europe and North America. Inclusion of countries across the globe is growing. Challenges remain in defining standards for risk stratification, indications for bone marrow transplantation, patient-appropriate dose reductions, and optimal therapy for targetable molecular lesions as well as for optimal supportive care guidelines. The fate of children with AML who live in countries with fewer resources also must not be forgotten. Through cooperative and integrated efforts, the international research community can address these challenges.

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Disclosures provided by the authors are available with this article at www.jco.org.

AUTHOR CONTRIBUTIONS

Conception and design: C. Michel Zwaan, Edward A. Kolb, Dirk Reinhardt, Henrik Hasle, Franco Locatelli, Marry M. van den Heuvel, Gertjan J.L. Kaspers

Provision of study materials or patients: All authors

Collection and assembly of data: Edward A. Kolb, Jonas Abrahamsson, Souichi Adachi, Eveline S.J.M. De Bont, Barbara De Moerloose, Michael Dworzak, Brenda E.S. Gibson, Henrik Hasle, Guy Leverger, Christine Ragu, Raul C. Ribeiro, Carmelo Rizzari, Jeffrey E. Rubnitz, Owen P. Smith, Lillian Sung, Daisuke Tomizawa, Ursula Creutzig, Gertjan J.L. Kaspers

Data analysis and interpretation: C. Michel Zwaan, Jonas Abrahamsson, Souichi Adachi, Richard Aplenc, Eveline S.J.M. De Bont, Brenda E.S. Gibson, Guy Leverger, Christine Ragu, Raul C. Ribeiro, Jeffrey E. Rubnitz, Owen P. Smith, Lillian Sung, Ursula Creutzig, Gertjan J.L. Kaspers

Manuscript writing: All authors

Final approval of manuscript: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST