Predicting relapse risk in childhood acute lymphoblastic leukaemia

Transcription

1 review Predicting relapse risk in childhood acute lymphoblastic leukaemia David T. Teachey 1 and Stephen P. Hunger 2 1 Pediatric Hematology and Oncology, Children s Hospital of Philadelphia, University of Pennsylvania School of Medicine, Philadelphia, PA, and 2 Pediatric Hematology/Oncology/BMT, University of Colorado School of Medicine and Children s Hospital Colorado, Aurora, CO, USA Summary Intensive multi-agent chemotherapy regimens and the introduction of risk-stratified therapy have substantially improved cure rates for children with acute lymphoblastic leukaemia (ALL). Current risk allocation schemas are imperfect, as some children are classified as lower-risk and treated with less intensive therapy relapse, while others deemed higherrisk are probably over-treated. Most cooperative groups previously used morphological clearance of blasts in blood and marrow during the initial phases of chemotherapy as a primary factor for risk group allocation; however, this has largely been replaced by the detection of minimal residual disease (MRD). Other than age and white blood cell count (WBC) at presentation, many clinical variables previously used for risk group allocation are no longer prognostic, as MRD and the presence of sentinel genetic lesions are more reliable at predicting outcome. Currently, a number of sentinel genetic lesions are used by most cooperative groups for risk stratification; however, in the near future patients will probably be risk-stratified using genomic signatures and clustering algorithms, rather than individual genetic alterations. This review will describe the clinical, biological, and response-based features known to predict relapse risk in childhood ALL, including those currently used and those likely to be used in the near future to risk-stratify therapy. Keywords: acute leukaemia, acute lymphoblastic leukaemia, childhood haematological malignancies, cytogenetics of leukaemia, minimal residual disease. One of the greatest accomplishments in cancer medicine is the dramatically improved survival of children with acute lymphoblastic leukaemia (ALL), the most common paediatric Correspondence: Stephen P. Hunger, Center for Cancer and Blood Disorders, Children s Hospital Colorado, East 16th Ave. Box B115, Aurora, CO 80045, USA. Stephen.Hunger@childrenscolorado.org malignancy. In the 1960s, <10% of children survived; now over 85% of children are cured using intensified multi-agent chemotherapy (Hunger et al, 2012). A significant reason for this improved survival was the introduction of risk-stratified therapy, using clinical and biological features and early treatment response to tailor chemotherapy intensity. Some cooperative groups divide patients into eight or more risk groups that receive different therapy (Schultz et al, 2007). While this allows therapy to be tailored more precisely, it creates challenges in trial design. As survival has improved over time with therapy intensification, some factors used in risk allocation have lost prognostic significance, mandating that the value of prognostic factors be reassessed periodically. Risk allocation schemas remain imperfect. Many low-risk patients relapse, and many high-risk patients are treated more intensively than necessary for cure. Minimal residual disease (MRD) is currently the single most powerful prognostic factor in childhood ALL (Campana, 2012). Nevertheless, half of all relapses on Children s Oncology Group (COG) ALL trials occur in patients with an excellent early MRD response (Borowitz et al, 2008). More work is needed to predict which children are at highest risk for relapse, as the prognosis for relapsed ALL is poor, and efforts to prevent relapse by improving de novo therapy are paramount (Nguyen et al, 2008). This review will describe the clinical, biological, and response-based features that are known to predict relapse in childhood ALL, including features currently used to risk-stratify therapy and those likely to be used in the near future (Table I). Some risk factors can be used to allocate high-risk patients to targeted therapies, thereby improving survival (Schultz et al, 2009; Arico et al, 2010; Biondi et al, 2012). Clinical features Historically, many clinical variables known at initial diagnosis, including age, sex, race, white blood cell count (WBC), haemoglobin, platelet count, degree of organomegaly and lymphadenopathy, and presence of a mediastinal mass or extramedullary disease have been used in risk allocation schemas (Smith et al, 1996; Schultz et al, 2007). Most of these First published online 29 June 2013 doi: /bjh ª 2013 John Wiley & Sons Ltd

2 Table I. Prognostic features for childhood ALL. Favourable prognostic features Unfavourable prognostic features Clinical Clinical Age >1<10 years Age <1 year or Age 10 years* WBC < /l WBC /l NCI SR NCI HR Caucasian, Asian, Black, Native American, Pacific Islander Alaskan Native, Hispanic CNS1 CNS3 Response to chemotherapy Response to chemotherapy PGR PPR Day 7 BM M1/M2 Day 7 BM M3 by morphology by morphology Day 14 BM M1 Day 14 BM M2/M3 by morphology by morphology End induction BM End induction BM M2/M3 M1 by morphology by morphology Day 8 PB MRD <001% Day 8 PB MRD 001% End induction BM End induction BM MRD 001% MRD <001% End consolidation End consolidation BM BM MRD <01% MRD 01% (T-ALL only) (T-ALL only) Blast biology Blast biology B-cell immunophenotype T-cell immunophenotype ETV6-RUNX1 BCR-ABL1 High hyperdiploid MLL-r Trisomy +4, +10 TCF3-HLF iamp21 Hypodiploid <44 chromosomes BCR-ABL1-Like (Ph + -like) expression profile CRLF2r IKZF1 mutation or deletion TP53 mutation M1: <5% lymphoblasts by morphology. M2: 5 25% lymphoblasts by morphology. M3 > 25% lymphoblasts by morphology. Prednisone good response (PGR): < leukaemic blasts/l in peripheral blood after a 1 week prednisone prophase. Prednisone poor response (PPR): leukaemic blasts/l in peripheral blood after a 1-week prednisone prophase. Other definitions and abbreviations are provided in main text. Some of these features lose prognostic value in multivariate analysis. MRD is a continuous variable, and some cooperative groups use different cut-offs to define risk. The thresholds provided in the Table are ones confirmed and used by multiple independent groups. *Age 16 years worse than ETP phenotype particularly poor outcome. Improved survival with addition of TKIs. Only unfavourable prognosis for CRLF2r if NCI HR. have been replaced by more reliable predictors of outcome: immunophenotype, treatment response, and presence/absence of sentinel genetic lesions (Schultz et al, 2007). Age at diagnosis is a strong predictor of outcome. Infants <1 year old and children 10 years have worse prognoses (Smith et al, 1996). During infancy, there is an inverse linear relationship between age and prognosis, with survival of infants under 3 months half that of infants 9 12 months old (Pieters et al, 2007). Unfortunately, survival for infants with ALL has not improved in the past 20 years despite therapy intensification (Hunger et al, 2012). After 10 years of age there is a near-linear relationship between older age and decreased survival, with adolescents 16 years having inferior survival to children years old. There is also a linear relationship between initial WBC and prognosis in childhood ALL, with higher WBC portending a worse survival. Even with the identification of novel biomarkers and the introduction of response-based risk allocation, initial WBC remains a strong independent predictor of risk in B-lineage ALL (B-ALL) (Schultz et al, 2007). Different cooperative groups have historically used different cut-offs for age and WBC for risk stratification. While it is challenging to dichotomize a continuous variable, an international consensus conference divided B-ALL patients into two risk groups: (i) standard risk (SR) with WBC < /l and age 100 to 999 years and (ii) high risk (HR) with WBC /l and/or age 1000 years (Smith et al, 1996). These National Cancer Institute (NCI)/ Rome risk criteria remain in use today. Patients with NCI HR B-ALL treated on Children s Cancer Group (CCG) protocols from 1996 to 2002 had a 5-year event-free survival (EFS) rate of 701%, vs. 819% for NCI SR B-ALL patients (Gaynon et al, 2010). Similar differences in EFS are reported in patients treated on other cooperative group trials from this era (Conter et al, 2010a; Kamps et al, 2010; Mitchell et al, 2010; Moricke et al, 2010; Pui et al, 2010; Silverman et al, 2010). Age and WBC are less accurate predictors of outcome for T-ALL. Sex and race both have prognostic importance but are not used in risk stratification. Males still have a slightly worse prognosis than females (Hunger et al, 2012). In the past, this was largely because of the risk of testicular relapse; however, with modern chemotherapeutic approaches the worse prognosis for males is probably caused by a number of factors, including a higher incidence of T-ALL, a less favourable DNA index, and likely undefined intrinsic biological, endocrine, and metabolic differences. Historically, outcome for blacks has been worse than whites, partly because they have higher incidence of T-ALL and other high-risk features (Pollock et al, 2000). Nevertheless, the absolute difference in survival comparing blacks with whites decreased significantly over the past 20 years (Hunger et al, 2012). Compared with whites, some studies suggest Native American, Alaskan Native, and Hispanic ethnicity have worse outcome; whereas, Asian and Pacific Islanders have similar outcomes (Kadan- Lottick et al, 2003). Socioeconomic and cultural differences may cause a portion of the outcome disparity; however intrinsic biological differences in the blasts, including a higher incidence of high-risk genetic alterations, and in the hosts, such as germline polymorphisms that may affect drug ª 2013 John Wiley & Sons Ltd 607

3 metabolism, may help to delineate racial and ethnic survival differences (Yang et al, 2011). Multiple studies have demonstrated that ALL patients with central nervous system (CNS) leukaemia have worse prognosis (Pui & Thiel, 2009). CNS status is classified as: (i) no identifiable leukaemic cells in the cerebrospinal fluid (CSF) (CNS1); (ii) <5 WBC/ll in CSF but blasts on cytospin (CNS2); and, (iii) 5 WBC/ll in CSF with blasts or cranial nerve palsy (CNS3). CNS3 patients have worse outcomes than CNS1. The prognostic significance of CNS2 is less clear. Studies that included early CNS-directed therapy found that CNS2 patients have similar outcomes as CNS1; however, studies without early CNS-directed therapy found similar outcomes for CNS2 and CNS3 (Pui & Thiel, 2009). Finally, comorbid immunodeficiencies, bone marrow failure syndromes, or chromosomal breakage syndromes, as well as malnutrition, pre-existing organ dysfunction, and trisomy 21 can affect outcomes in children with ALL (Viana et al, 1994; Maloney, 2011). These conditions often lead to treatment changes but are not commonly used for risk allocation. Response to therapy Multiple studies have demonstrated the prognostic importance of the rapid morphological clearance of blasts in peripheral blood (PB) and bone marrow (BM) during the initial phases of chemotherapy (Miller et al, 1983, 1989; Campana, 2012). For decades, morphological response assessment at different time-points early in therapy has been used for risk group allocation in ALL trials (Schultz et al, 2007; Lauten et al, 2012). Recently, MRD assessment has largely replaced morphological assessment in risk allocation. Patients who fail to achieve morphological remission after induction therapy (induction failure) have a poor prognosis; however, not all patients have dismal outcomes. Comparing outcomes between studies can be difficult, as duration of induction-remission therapy can vary (4 6 weeks), and some cooperative groups classify patients who have an M3 marrow (>25% blasts by morphology) as induction failures, while others classify patients who either have M2 or higher marrow morphology as induction failures (>5% blasts). Schrappe et al (2012) reported the outcomes of 1041 patients with induction failure treated by 14 study groups between 1985 and Induction failure was defined as persistence of leukaemic blasts in blood, marrow or any extramedullary site after 4 6 weeks of remission-induction therapy. Induction failure was rare, occurring in 24% of patients, with a 10-year overall survival (OS) of 32%. These patients often presented with high-risk features, including T cell immunophenotype, MLL-rearrranged (MLL-r), BCR- ABL1 translocation, older age, and higher WBC. Those with T-ALL, MLL-r, age 10 years, or M3 marrow had particularly poor outcome. In contrast, children less than 6 years old with B-ALL and no high-risk genetic features who had induction failures had a 72% 10-year OS when treated with chemotherapy alone and showed no benefit from haematopoietic stem cell transplant (HSCT). Many studies have evaluated the prognostic import of morphological assessment of PB and BM during the first 4 weeks of therapy, before remission is assessed. Patients with a rapid response to a 1-week prednisone prophase (prednisone good responders (PGR); < /l leukaemic blasts/l in PB) have improved survival over patients with a poor prednisone response (PPR; /l leukaemic blasts/l) (Schrappe et al, 2000; Lauten et al, 2012). Residual leukaemia in BM on Induction day 7 and/or 14 correlates with worse outcome, and augmentation of therapy in slow responding patients improves outcome (Nachman et al, 1998). Accordingly, many groups use early marrow response to risk stratify patients. MRD is currently the single most powerful prognostic factor in childhood ALL (Borowitz et al, 2008). Several large studies have investigated the sensitivity and specificity of MRD at different time points on a variety of treatment protocols, using either flow- or polymerase chain reaction (PCR)-based technologies (Coustan-Smith et al, 1998; Conter et al, 2010b; Schrappe et al, 2011). MRD is helpful in determining prognosis in de novo ALL patients, after reinduction chemotherapy for first relapse, and prior to and after HSCT (Campana, 2012). Currently, MRD can be measured in the clinic by three validated methods (Table II): (i) PCR amplification of immunoglobulin (Ig) and T cell receptor (TCR) genes; (ii) PCR amplification of oncogenic fusion transcripts; and, (iii) flow cytometric detection of aberrant immunophenotypes (Bruggemann et al, 2010). Several studies have shown that flow cytometry and PCR amplification of antigen receptor genes yield remarkably close measurements if MRD is present at 001% (Neale et al, 2004; Kerst et al, 2005). PCR amplification of fusion transcripts is less commonly used because many patients do not have identifiable fusion transcripts, and many of the most common gene fusions have not been quantitatively validated. PCR amplification and quantification of BCR-ABL1 transcripts in PB and BM are powerful for determining prognosis (Campana, 2012). PCR amplification of Ig/TCR rearrangements is commonly used for MRD detection in many cooperative group trials, including most European trials. Because it uses DNA, specimens remain stable with prolonged shipping times. It is highly sensitive, accurately detecting MRD to 10 4 with the capacity to detect to PCR-based MRD is more timeconsuming, expensive, and technically complex than standard flow-based assays, and can miss blasts that undergo clonal evolution, leading to false negative results (Bruggemann et al, 2010). The COG uses flow cytometric detection of aberrant immunophenotypes (Borowitz et al, 2008). This rapid test can have results returned in 1 d. It is less sensitive than PCR, accurately detecting MRD to 10 4 using 3- to 4-colour flow cytometry, and requires a high level of expertise to interpret (Campana, 2012). It can also miss blasts that undergo clonal 608 ª 2013 John Wiley & Sons Ltd

4 Table II. MRD techniques (Bruggemann et al, 2012; Wu et al, 2012). Flow cytometry PCR amplification Deep sequencing 10 4 to 10 5 Sensitivity 3- to 4-colour: 10 3 to to to 9-colour: 10 4 to 10 5 Specificity 90 95% 90 95% 90 95% Specimen stability Need fresh viable cells Prolonged (DNA test) Prolonged (DNA test) Cost Relatively less expensive Relatively more expensive Relatively most expensive Time Rapid turnaround Slower turnaround Slower turnaround Labour Intensity Low High: patient specific primers Low Availability Standard in Europe Standard in US Experimental only Clonal evolution Can miss immunophenotypic shifts Can miss new TCR and Ig rearrangements Can identify new clones and subclones MRD, minimal residual disease; PCR, polymerase chain reaction. evolution, and it has the disadvantage of requiring fresh specimens with viable cells. Six- to 9-colour flow cytometry techniques have been developed, which increase the sensitivity to that of PCR; however, this increases cost and complexity. Recently, high-throughput deep-sequencing approaches of Ig/TCR rearrangements have been developed to detect MRD (Faham et al, 2012). This technique has many potential advantages, including improved sensitivity, reliably identifying one blast in greater than one million cells, and the ability to detect clonal evolution. It may replace other MRD technologies in the near future. MRD is a continuous variable, and prognosis improves as MRD decreases. Nevertheless, most studies dichotomize MRD for risk stratification, and the cut-offs used by many cooperative groups have changed over time. Early studies from St. Jude Children s Research Hospital (SJCRH), found that MRD 001% on Day 19 (mid-induction) and Day 46 (end-induction) were strongly predictive of relapse, and MRD remained an independent predictor of prognosis regardless of other clinical and genetic features (Coustan- Smith et al, 1998). On COG P9904/5/6 trials, multivariate analysis found Day 29 MRD measured by flow cytometry using a cut-off of 001% was the most significant predictor of outcome in patients with B-ALL (Borowitz et al, 2008). While MRD was an independent predictor of outcome, the effect of MRD was quantitatively different among genetic subgroups and NCI risk groups. These studies also found that Day 8 PB MRD level was an independent predictor of worse outcome in multivariate analysis. Both SJCRH and COG use a combination of genetic features, NCI risk classification, MRD obtained during and at the end of induction to risk stratify B-ALL patients. The recently completed Italian Association of Pediatric Haematology-Oncology (AIEOP)-Berlin-Frankfurt-Muenster (BFM) 2000 study used PCR to measure MRD on Days 33 (end induction) and Day 78 (end of consolidation, EOC) in 3184 B-ALL patients, finding MRD predicted outcome more precisely than genetic factors, prednisone response, or NCI risk category (Conter et al, 2010b). B-ALL patients with MRD <00 l% at both time-points had a 5-year EFS of 92%, compared with 501% if MRD >01% on day 78, and 776% in all other patients. While most studies have focused on the prognostic significance of end induction MRD in B-ALL, these data suggest that EOC MRD may also be an important predictor of adverse outcome in B-ALL. The successor AIEOP-BFM ALL 2009 study uses a combination of Day 33 and 78 MRD, genetic features and prednisone response to risk-allocate B-ALL patients. The Medical Research Council (MRC) adopted an MRDbased risk allocation in UK ALL 2003 and found that MRD was the single most important predictor of relapse (Vora et al, 2013). Patients with unfavourable MRD responses received intensified therapy, and low-risk patients (MRD <001% at Day 29 and undetectable by week 11) were randomized to receive one vs. two delayed intensification (DI) courses. Patients with day 29 BM MRD 001% (HR group) had a threefold higher relapse rate (5-year EFS 798%) compared with low-risk patients (5-year EFS 944% with single DI and 955% with double DI). While the majority of data suggest end induction MRD assessment is the best predictor of outcome in B-ALL, data from AIEOP BFM 2000 show that EOC MRD is a better predictor of adverse outcome in T-ALL (Schrappe et al, 2011). Patients with Day 33 MRD <001% had the best outcome (7-year EFS 911%). Patients with MRD <01% on Day 78 had a favourable outcome, regardless of other factors including end of induction MRD. Patients who had Day 33 MRD 001% but had Day 78 MRD <01% had 7-year EFS of 806%. In contrast, patients with EOC MRD >01% had a much worse outcome (7-year EFS 498%), regardless of other factors. While there is a relative paucity of data assessing the prognostic value of Day 8 PB MRD in T-ALL, PB and BM are more concordant in T-ALL than in B-ALL, suggesting Day 8 PB MRD could be helpful for risk allocation in T-ALL (van der Velden et al, 2002). Finally, MRD can predict outcome in infants with ALL. On Interfant-99, all infants with HR MRD (MRD 001% at EOC) relapsed (Van der Velden et al, 2009). In contrast, low-risk patients (MRD <001% at end induction and EOC) ª 2013 John Wiley & Sons Ltd 609

5 had a relapse-rate of only 13%. All others had a 31% relapse rate. Blast and host biology Immunophenotype Immunophenotype is one of the most important prognostic factors in childhood ALL. Historically, T-ALL has been shown to have a worse prognosis than B-ALL (Pui et al, 2012). This difference has decreased considerably over time and is probably explained primarily by the high frequency of unique genetically defined low-risk subsets in B-ALL (e.g. high hyperdiploidy (HeH) and ETV6-RUNX1 fusion) that do not exist in T-ALL. With modern chemotherapeutic approaches the outcome for patients with T-ALL and B-ALL are more similar when these low risk groups are excluded. Leukaemia arises from initial transforming events in early progenitors followed by multi-step acquisition of multiple genetic alterations that affect cell proliferation, differentiation, and survival. Normal B cells derive from progenitors in the marrow and mature through a series of discrete developmental stages identifiable by expression of cell surface markers and specific transcription factors and by immunoglobulin gene rearrangements (Campos-Sanchez et al, 2011). The vast majority of paediatric B-ALLs are developmentally arrested at the pre-pro-b and pre-b stages. With the exception of mature B cell (Burkitt) ALL, which has a worse prognosis and requires very different therapy, the developmental stage does not independently predict risk with modern treatment regimens and risk allocation schemes. Leukaemic or preleukaemic blasts can acquire mutations at multiple stages of differentiation, at times leading to a predominant clone with multiple subclones. These subclones can be at different stages of B cell development compared to the main clonal population and may emerge as a dominant clone at relapse (Mullighan et al, 2008a; Campos-Sanchez et al, 2011). Unlike acute myeloid leukaemia (AML), leukaemia propagating cells are present at high frequency in diverse B ALL populations, and there does not appear to be a stem cell hierarchy (Rehe et al, 2013). In T-ALL, maturational stage of the dominant clone is more predictive of outcome. The European Group for the Immunological Classification of Leukaemias (EGIL) classified T-ALL into multiple immunophenotypic subtypes (Bene et al, 1995). Cortical T-ALL has the best prognosis, and earlier stage T-ALL tends to do worse (Schrappe et al, 2011). With the possible exception of early T cell precursor ALL (ETP ALL), patients are not typically risk-stratified by T cell developmental stage, because stage is not an independent risk factor once treatment response is factored into risk algorithms. The unique ETP ALL subtype is defined by blasts expressing T-lineage markers and myeloid/early progenitor markers; however, ETP ALL is not an ambiguous-lineage leukaemia (Coustan-Smith et al, 2009). Despite their T cell origin, ETP ALL blasts have a gene expression profile more similar to myeloid/early progenitor leukaemia and have mutations more commonly associated with myeloid leukaemia (Zhang et al, 2012). The initial report from SJCRH and older AIEOP trials showed that ETP ALL patients had a very poor outcome (Coustan-Smith et al, 2009). Results from the more recent AIEOP BFM 2000 trial suggest that many, but not all, ETP ALL patients do poorly and outcome can be predicted by early response characteristics (Schrappe et al, 2011). Both of these studies had a very small sample size. While some institutions and cooperative groups consider ETP ALL an independent risk factor for very poor outcome and recommend HSCT for all ETP ALL patients in first complete remission (CR1), most are waiting for more mature data and do not alter therapy based on ETP status. Sentinel translocations and gene fusions The number of prognostic genetic alterations in childhood ALL is increasing exponentially with next generation sequencing (NGS) (Fig. 1). Many of these alterations are currently used to stratify risk and some to allocate therapy, while others still need to be validated. It is critical to assess whether mutations are independent predictors of outcome in the context of clinical factors and MRD. BCR-ABL1 The BCR-ABL1 oncogene created by the t(9;22)(q34;q11) (Philadelphia chromosome; Ph + ) is identified in nearly all cases of chronic myeloid leukaemia (CML), 25% of adult ALL cases, and 3-5% paediatric ALL cases, and produces either a 210 (CML) or 190 kda (90% of paediatric Ph + ALL cases) oncoprotein with aberrant tyrosine kinase activity (Suryanarayan et al, 1991; Hunger, 2011). Prior to the development of BCR-ABL1 tyrosine kinase inhibitors (TKIs), Ph + ALL had a dismal prognosis. Two reports from major paediatric ALL international cooperative groups demonstrated very poor survival, with 7-year EFS and OS rates of 31% and 44% in the cohort (Arico et al, 2000, 2010). While best available donor HSCT in CR1 improved outcome over chemotherapy alone, survival was still poor. Based on promising results from early phase trials, the COG AALL0031 Ph + ALL trial added the TKI imatinib to an intensive multi-agent chemotherapy backbone with or without HSCT, finding 3-year EFS rates over 75% in children treated with continuous imatinib plus chemotherapy (Schultz et al, 2009). These results are stable, with 7-year EFS of 71% for patients treated with chemotherapy plus imatinib and without HSCT (Stephen P. Hunger, unpublished observations). The European EsPhALL group has also reported very promising results with imatinib in Ph + ALL (Biondi et al, 2012). While data are limited, Ph + ALL patients treated with chemotherapy plus TKI appear to be less likely to develop 610 ª 2013 John Wiley & Sons Ltd

6 TAL1 7% TLX3 2% LYL1 1% ETP 2% TLX1 <1% MYC 2% Other T-ALL 2% ETV6-RUNX1 20% ERG 3% HNF1 -HLF 1% Other B-ALL 8% CRLF2-r 4% High Hyperdiploid 25% BCR-ABL1 Like 9% TCF3-PBX1 4% BCR-ABL1 3% MLL-r 6% Hypodiploid 1% iamp21 2% Fig 1. Frequency of cytogenetic subtypes of childhood ALL. Shows the relative frequencies of B-ALL (blue) and T-ALL (red) genetic subtypes. This pie chart does not include submicroscopic genetic alterations. Data from Pui et al (2012). BCR-ABL1 mutants that confer TKI resistance than CML patients treated with TKI alone (Chang et al, 2012). The second generation ABL1-class TKI dasatinib is more potent than imatinib in vitro and has better CNS penetration, and preclinical evidence suggests abnormal signalling through Src family kinases is critical for development of Ph + ALL but not CML; therefore, there is interest in testing dasatinib in paediatric Ph + ALL (Hu et al, 2004). Deletions or mutations in IKZF1, which encodes the lymphoid regulator IKAROS, are found in 80% of childhood Ph + ALL cases (Mullighan et al, 2008b). IKZF1 mutations may confer resistance to chemotherapy, and prospective identification of IKZF1 mutations in Ph + ALL may be helpful in future studies for risk allocation and treatment (Virely et al, 2010). ETV6-RUNX1 (TEL-AML1) The most common chromosome translocation in paediatric ALL is t(12;21)(p13;q22), which occurs in 20 25% of B-ALL and creates a chimaeric ETV6-RUNX1 transcription factor (Harrison et al, 2010). The translocation is cryptic and generally not visible on standard karyotyping, but ETV6-RUNX1 fusion can be detected readily via fluorescence in situ hybridization (FISH) or reverse transcription (RT)-PCR. Like other ALL fusion proteins, ETV6-RUNX1 is not sufficient for leukemogenesis and requires additional cooperating genetic events. A number of studies have shown that children with ETV6-RUNX1 + ALL have an outstanding prognosis as compared to children lacking this gene fusion, leading many cooperative groups to classify patients with ETV6-RUNX1 as lower risk and treat with less intensive chemotherapy (Harrison et al, 2010). While the vast majority of these studies have demonstrated ETV6-RUNX1 is an independent predictor of improved outcome, Loh et al (2006) found ETV6-RUNX1 lost prognostic significance in multivariate analysis in the Dana Farber Cancer Institute (DFCI) trial Not all children with ETV6-RUNX1 ALL are cured. Like other ALL subsets, MRD response is highly predictive of outcome in ETV6-RUNX1 ALL, and patients with poor responses have an increased risk of treatment failure (Borowitz et al, 2008). ETV6-RUNX1 ALL has been shown to have recurrent submicroscopic genetic alterations in EBF1, PAX5, BTLA, TOX, NR3C1, BMF, TBL1XR1, and BTG1 (Mullighan et al, 2007; Parker et al, 2008). Further studies are needed to determine if these alterations are helpful for risk stratification, as some are associated with drug resistance and are more common in ETV6-RUNX1 ALL at relapse (van Galen et al, 2010; Kuster et al, 2011). MLL translocations and fusion proteins Translocations involving the chromosome 11q23 gene MLL are present in over 70% of infant leukaemias, ~2 5% of childhood ALL, and 5 10% of childhood AML, as well as in chemotherapy-associated leukaemias arising after topoisomerase II inhibitor therapy. Over 100 different fusion partners ª 2013 John Wiley & Sons Ltd 611

7 have been described, but three of them, t(4;11)(q21;q23) (MLL-AFF1 [AF4]), t(9;11)(p22;q23) (MLL-MLLT3 [AF9]), and t(11;19)(q23;p13.3) (MLL-MLLT1), are identified in over 70% cases of MLL-r ALL (Krivtsov & Armstrong, 2007). MLL is a methyltransferase that is ubiquitously expressed in haematopoietic cells and is an important developmental regulator of haematopoietic stem cells. Many MLL translocations are strongly leukaemogenic and, unlike other subtypes of ALL, additional genetic alterations are uncommon (Mullighan et al, 2007). All types of MLL-r ALL confer a worse prognosis in infancy, with the 4-year EFS for MLL-r ALL being nearly half that of MLL-germline (<40% vs. >70%) (Pieters et al, 2007). In children 1 year of age, the prognostic significance of MLL-r is less clear. Recent data suggest the most common MLL rearrangement, t(4;11), may be associated with a higher risk of relapse in children younger than 4 years but not in those four and older (Moorman et al, 2010). Some MLL translocations lose prognostic significance in multivariate models, and in T-ALL certain MLL translocations, including t(11;19), confer a favourable prognosis (Moorman, 2012). The location of the breakpoint within MLL may predict prognosis, with breaks in MLL intron 11 carrying a worse prognosis than others (Emerenciano et al, 2013). Modern genomic techniques may be helpful in further risk-stratifying patients. Kang et al (2012) performed gene expression profiling on 97 infant ALLs treated on COG P9407 and identified a 7-gene classifier that split MLL-AFF1 + infants into two groups with markedly different outcomes (20 vs. 65% 5-year EFS). Some cooperative groups, including the COG and MRC, consider all forms of MLL-r B-ALL higher risk and recommend intensified therapy, regardless of other disease markers or treatment response. SJCRH does not treat non-infant MLL-r ALL differently, while other groups, including AIEOP and BFM, only intensify therapy for patients with t(4;11) (Conter et al, 2010b). Ploidy/DNA index Whole chromosome losses or gains are found in blasts from the majority of children with ALL. With the increased use of single nucleotide polymorphism (SNP) arrays and other high throughput genomic technologies, subtle chromosomal gains and losses are identified at even higher frequency. For over 25 years, many groups have used ploidy in childhood ALL risk stratification. Ploidy is determined by counting chromosome number in a metaphase karyotype or measuring DNA content to determine DNA index, the ratio of fluorescence in BM blasts compared to a normal diploid cell, using flow cytometry (diploid = 10; hypodiploid <10; hyperdiploid; > 10 hyperdiploid). For the purposes of risk stratification, cases are classified as hypodiploid (<46 chromosomes, 6% of cases), normal/diploid (46 chromosomes, 30% of cases), pseudodiploid (diploid with chromosomal translocations, 28%), low hyperdiploid (47 to 50 chromosomes, 11%), and HeH (>50 chromosomes, 25%) (Sutcliffe et al, 2005). Hypodiploid ALL with <44 chromosomes is present in 15% of paediatric ALL and is an independent risk factor for poor prognosis, with worsening outcome with decreasing chromosome number. Most hypodiploid ALLs (80%) have 45 chromosomes, and outcome for these patients is similar to that of diploid ALL (Nachman et al, 2007). Consistent across studies is that survival is markedly worse for low hypodiploid (~30 39 chromosomes), and very poor for near haploid (~<30 chromosomes) (Harrison et al, 2004; Moorman, 2012). The MRC reported on 141 children with hypodiploid ALL, finding a 29% 3-year EFS for patients with chromosomes compared to 66% for patients with chromosomes (Harrison et al, 2004). In addition, clonal evolution is common in near-haploid ALL, and sub-clones often have reduplicated chromosomal copies of the primary clone, appearing diploid or hyperdiploid (masked hypodiploidy) and retaining a poor prognosis. New discoveries on the genomic complexity of hypodiploid ALL have provided novel insights into this high risk ALL subset, including the finding that about half of paediatric low-hypodiploid ALL cases have germline TP53 mutations, meaning that this disease can be a manifestation of Li-Fraumeni syndrome (Holmfeldt et al, 2013). Further genomic analyses may refine risk stratification of hypodiploid ALL. In contrast to hypodiploidy, HeH (51 65 chromosomes, DNA index 116) is associated with a favourable prognosis in B-ALL. Chromosome gain in hyperdiploid ALL is nonrandom, with eight chromosomes (X, 4, 6, 10, 14, 17, 18, and 21) accounting for most gains (Paulsson et al, 2010). Some cooperative groups use HeH and/or specific trisomies in risk stratification, while others do not. The COG previously used both trisomy and in riskstratification, but now use to help define low-risk patients (Sutcliffe et al, 2005; Schultz et al, 2007). The MRC found that trisomy 18 was a better predictor of survival compared to other trisomies in their population but do not use it or any form of hyperdiploidy for risk stratification (Moorman et al, 2010). Paulsson et al (2010) used SNP array analysis to investigate 74 HeH ALLs, finding additional abnormalities missed on routine karyotype in 80% of cases. Recurrent abnormalities included whole-chromosome uniparental isodisomies of 9 and 11 and microdeletions of CKDN2A, PAX5, and PAN3; however, the study was not powered to determine if these alterations predicted risk. Further investigation may help identify patients in this favourable risk group who have a higher relapse risk. TCF3-PBX1 and TCF3-HLF The t(1;19)(q23;p13) occurs in about 5% of childhood B-ALL, with most cases having TCF3-PBX1 fusion (formerly termed E2A-PBX1) (Hunger et al, 1998; Moorman et al, 612 ª 2013 John Wiley & Sons Ltd

8 2010; Moorman, 2012). A small percentage (<05%) of paediatric B-ALLs have a t(17;19)(q22;p13) translocation and TCF3-HLF fusion (Hunger et al, 1998; Mullighan, 2012). Some groups previously considered the t(1;19) to be an independent predictor of poor outcome; however, with modern therapy and MRD-based risk allocations schemas, t(1;19) is no longer an independent risk factor for prognosis in childhood ALL (Hunger, 1996; Hunger et al, 1998; Jeha et al, 2009). While the number of cases is small, TCF3-HLF is associated with a very poor outcome (Hunger, 1996). Because of its rarity, only the UKALL group currently uses this abnormality in risk stratification. Intrachromosomal amplification of chromosome 21 (iamp21) Approximately 2% of paediatric B-ALLs have iamp21, defined as three or more extra copies of RUNX1 on a single abnormal chromosome 21 (Moorman et al, 2007). Either FISH to identify AML1-RUNX1 fusion or routine karyotyping can identify iamp21. This abnormality was originally associated with a very poor outcome, leading to suggestions that HSCT in CR1 might be the best therapy, but those studies were small and many used therapy that is not as intensive as now used (Moorman et al, 2007). More recent data from the BFM, UKALL 2003 and COG show better outcomes with more intensive chemotherapy without SCT, with MRD predicting those with very poor outcome (Stephen P. Hunger, unpublished observations, Attarbaschi et al, 2008). The COG found that the low intensity therapy used for children with SR ALL was not effective for patients with iamp21 and now treats all patient with iamp21 with high-risk regimens regardless of other features (Stephen P. Hunger, unpublished observations). Recently, Rand reported recurrent abnormalities in RB1 (37%), IKZF1 (22%), ETV6 (19%), CDKN2A/B (17%), and PAX5 (8%) in iamp21 patients, but none of these helped to refine risk prediction (Rand et al, 2011). Philadelphia chromosome (Ph-like or BCR-ABL1-like) ALL Recently, a subset (about 15%) of B-ALL has been identified that has a gene expression profile similar to that of Ph + ALL, a high frequency of IKZF1 deletions/mutations, and a very poor outcome (Den Boer et al, 2009; Mullighan et al, 2009a; Harvey et al, 2010a; Loh et al, 2013). The similarity of the expression profile to Ph + ALL suggested that aberrant kinase activity may drive this subset of ALL. About half of Ph-like ALLs have the cryptic rearrangements P2RY8-CRLF2 or IGH- CRLF2 affecting CRLF2 (cytokine receptor-like factor 2) leading to overexpression of this component of the heterodimeric cytokine receptor for thymic stromal lymphopoietin (TSLP), with approximately half of CRLF2-rearranged cases having point mutations in Janus kinase (JAK) family members JAK1 and JAK2 (Mullighan et al, 2009b,c; Russell et al, 2009; Harvey et al, 2010b; Yoda et al, 2010; Zhang et al, 2011). Interestingly, CRLF2-r, particularly P2RY8-CRLF2, is present in over half of Down syndrome-associated ALL (with JAK mutations in half of these cases) (Mullighan et al, 2009b; Hertzberg et al, 2010). About half of the Ph-like ALL cases that lack CRLF2-r have novel rearrangements and mutations involving a variety of tyrosine kinase and cytokine signalling genes that may be amenable to targeted therapy with ABL1-class TKIs (Roberts et al, 2012). Some Ph-like ALLs have activating point mutations in IL7R and FLT3 or deletion of the JAK2 negative regulator SH2B3 (LNK), but other TK gene point mutations have not been observed and are probably rare (Roberts et al, 2012; Loh et al, 2013). Because the Ph-like gene signature and IKZF1 alterations (see below) are independent risk factors for poor prognosis, they will probably be used for risk stratification and allocation to targeted therapies in the near future. Robust clinical diagnostic assays to identify these abnormalities will soon be available. Submicroscopic/secondary genomic abnormalities in B-ALL Approximately 25% of B-ALL patients lack an identifiable chromosomal translocation. New genomic techniques, including SNP arrays, array-comparative genomic hybridization (acgh), transcriptome profiling, epigenetic profiling, and whole exome and genome sequencing, are identifying abnormalities in paediatric ALL at a very rapid pace. These novel techniques have identified dozens of recurring genetic alternations. Certain ALL subtypes, including MLL-r have few additional alterations, whereas others, including BCR- ABL1 and ETV6-RUNX1 ALL have multiple additional alterations (Mullighan et al, 2007, 2008b). The most prevalent submicroscopic genomic alterations in paediatric ALL occur in CDKN2A/B (30 40%), IKZF1 (15%), PAX5 (20%), and ETV6 (10 15%) (Hunger et al, 2011; Moorman, 2012; Mullighan, 2012). Many of these submicroscopic genomic abnormalities are found across ALL subtypes, while others are more subtype-specific. A large number of additional genetic alterations have been identified in paediatric ALL, including deletions, amplifications, or mutations in genes involved in B cell development (EBF1, ERG, IKZF2, LEF1, RAG1/2), histone acetylation (CREBBP), apoptosis (BTG1), and drug response (NR3C1, TBL1XR1), as well as known oncogenes and tumor suppressor genes including TP53 and RB1 (Hunger et al, 2011; Moorman, 2012; Mullighan, 2012). Most of these lesions have not been shown to impact prognosis; however, many are recently identified, and studies were not powered to assess prognostic significance. Based on the rarity of paediatric ALL, international collaborations are needed to determine the prognostic value of many of these genetic abnormalities accurately. In addition, immunophenotypic subclones within a patient can be genetically heterogeneous, making it more complicated to interpret the prognostic ª 2013 John Wiley & Sons Ltd 613

9 significance of abnormalities (Greaves & Maley, 2012). Only mutations or deletions of IKZF1, located at 7p12, have been shown to be independent predictors of outcome in childhood ALL. IKZF1 is mutated in 15 20% of ALL cases, including 30% of high-risk cases and 80% of Ph + and Ph-like cases, and is associated with poor prognosis in all of these (Mullighan et al, 2009b; Harvey et al, 2010a). Genetics and prognostic factors in T-ALL Children with T-ALL have a worse prognosis than children with B-ALL, largely due to the limited number of good risk T-ALL patients. High-risk T- and B-ALL patients have similar outcomes. While a reasonable percentage of patients with relapsed B-ALL can be cured, relapsed T-ALL has a dismal prognosis with 3-year EFS rates of <15% (Einsiedel et al, 2005; Raetz et al, 2008). Predicting which T-ALL patients are more likely to relapse has significant clinical implications, as these patients could be treated more intensively (with or without HSCT) or with novel agents at diagnosis. Unfortunately, while knowledge of the biology of T-ALL has improved greatly over the past few decades, the vast majority of identifiable genetic lesions do not independently predict prognosis. Response to treatment, including assessment of MRD is the most important predictor of outcome in T-ALL (Schrappe et al, 2011). As with B-ALL, some patients who have excellent treatment response and become MRD-negative eventually relapse. New insights into the molecular biology of T-ALL should improve risk stratification, allowing early identification of these patients. Constitutive activation of NOTCH1 signalling is the most common abnormality in childhood T-ALL. Notch signalling can be dysregulated in T-ALL by activating mutations in NOTCH1 (>50% of cases of T-ALL), FBXW7 mutations (15% of cases), and t(7;9)(q34;q34.3) (<1% of cases), with most studies suggesting none of these have a strong independent correlation with outcome (Van Vlierberghe & Ferrando, 2012). Other lesions commonly found in T-ALL can be clustered based on function, including genes that control cell cycle-regulation, signal transduction, cell growth, and chromatin remodelling, as well as homeobox, LMO, and bhlh family members (Van Vlierberghe & Ferrando, 2012). None of the more common alterations have been demonstrated to predict outcome independently and consistently across trials once MRD is included. The genetics underlying T-ALL are complex, and modern genomic techniques may be more able to risk-allocate patients by identifying gene signatures that correlate independently with outcome. Cleaver et al (2010) studied 84 T-ALL patients using qrt-pcr and microarray gene expression profiling and identified a five-gene signature that was 82% accurate in predicting patients who would relapse at diagnosis. Gutierrez evaluated the prognostic significance of absence of biallelic TRG deletion (ABD) in childhood T-ALL by performing acgh on 47 children with T-ALL (Gutierrez et al, 2010). ABD was strongly correlated with induction failure, suggesting it could be used as a marker to allocate these patients to alternative therapy early. Because TRG rearrangements occur early in T-lineage development, the authors hypothesized there would be overlap between ABD and ETP phenotype. They did find ABD and ETP have similar gene expression profiles; however, only one of the 41 ABD subjects had the classic ETP immunophenotype. Genetics and prognostic factors in infant ALL Infants (<1 year of age) with ALL have poor outcome (Pieters et al, 2007). In addition, very young infants (<3 months), WBC > /l, CD10 negativity, and MLL-r predict even worse outcome in this high-risk group (Pieters et al, 2007). Not surprisingly, infant ALL is biologically distinct from its older aged counterpart. MLL-r (reviewed above) is found in the blasts of >70% of infants with ALL. Gene expression profiling of infant ALL demonstrates markedly different transcriptomes compared with childhood ALL, including frequent activation of JAK/STAT signalling (Qazi & Uckun, 2010). Infant ALL blasts frequently overexpress FLT3 and genes involved in chemotherapy resistance, including MCL1 (Stam et al, 2006). The prognostic relevance of these alterations are not known, however, future studies may determine if any of these may be helpful for risk allocation or be potential targets for therapy. Pharmacogenomics and host polymorphisms Major new insights have come from investigating how host germline variability influences leukemogenesis, treatment response, drug metabolism, the immune system, and the microenvironment. Some SNPs predict outcome, and future studies may include these in risk allocation. Others may be used to tailor therapy, such as polymorphisms in thiopurine methyltransferase (TPMT), an enzyme involved in the conversion of thiopurines to inactive metabolites. Patients with homozygous TPMT deficiency are at higher risk for toxicities, because of the inability to metabolize thiopurines (Relling et al, 2011). The AIEOP-BFM 2000 trial genotyped 614 children for deletions of the glutathione S-transferase genes, GSTM1 and GSTT1 (Franca et al, 2012). GSTs encode for a class of enzymes involved in metabolism of various chemotherapeutics. In Caucasians, homozygous deletions of GSTM1 and GSTT1 occur in 40 60% and 13 25% of the population, respectively. Neither had an overall effect on outcome; however, GSTM1 deletions were associated with better outcome in PPR patients, and GSTT1 deletions were associated with worse outcomes in BFM standard risk patients and in PGR patients. A genome wide association study (GWAS) of children enrolled in COG and SJCRH ALL trials identified SNPs associated with early MRD response, relapse, and/or drug 614 ª 2013 John Wiley & Sons Ltd

10 disposition, generally linking MRD eradication with greater drug bioavailability (Yang et al, 2009). Another GWAS study showed components of genomic variation that co-segregated with Native American ancestry were independently associated with an increased risk of relapse that could be overcome with therapy intensification (Yang et al, 2011). Cloppenborg et al (2005) performed PCR-based genotyping on B-ALL patients and healthy controls to determine if host interferon-gamma (IFNG) expression levels were prognostic. They found that different IFNG alleles correlated with risk group, age at presentation, and response to corticosteroids, suggesting differential effects of IFNG on early response to treatment and immunosurveillance. Conclusion The survival of children with ALL has greatly improved due to more accurate risk stratification, therapy intensification, and improved understanding of ALL biology. Incorporation of MRD in risk allocation has markedly improved the ability to determine prognosis early in therapy; however, many patients who relapse are MRD-negative. High-throughput deep-sequencing MRD technologies may improve sensitivity and prognostic accuracy, and additional work is needed to determine the best time to evaluate MRD in different ALL subsets. A number of recent studies have provided biological insights into why some patients do not respond to therapy and others relapse, including the identification of novel genes that can lead to chemotherapy resistance and the discovery of significant copy number alterations and epigenetic changes at in blasts at relapse (Mullighan et al, 2008a, 2011; Hogan et al, 2011; Meyer et al, 2013; Tzoneva et al, 2013). Improved understanding of ALL blast biology may lead to superior therapies, better risk stratification, and less treatment failures. Future patients will probably be risk-stratified using genomic signatures and clustering algorithms, rather than individual genetic alterations. The myriad genetic and epigenetic changes found in blasts may allow clinicians to predict risk more accurately and provide targeted therapies. Identification of Ph-like ALL is a recent example of this approach: a high-risk subset with significant genetic heterogeneity identified by a similar genetic signature. Clinical use of these newer technologies is currently limited by cost, reproducibility, and the need for expert interpretation (Schrijver et al, 2012; Xuan et al, 2012; Ginsburg, 2013). Nevertheless, as technologies improve, these hurdles are disappearing, and in 3 5 years, newly diagnosed ALL patients may have SNP arrays to identify losses, gains, and polymorphisms not only in the leukaemic blasts, but also in the germline DNA to identify clinically relevant host genetic variation; whole genome sequencing to identify relevant mutations; and epigenetic and transcriptome profiling to help select the best therapies. References Arico, M., Valsecchi, M.G., Camitta, B., Schrappe, M., Chessells, J., Baruchel, A., Gaynon, P., Silverman, L., Janka-Schaub, G., Kamps, W., Pui, C.H. & Masera, G. (2000) Outcome of treatment in children with Philadelphia chromosome-positive acute lymphoblastic leukemia. The New England Journal of Medicine, 342, Arico, M., Schrappe, M., Hunger, S.P., Carroll, W.L., Conter, V., Galimberti, S., Manabe, A., Saha, V., Baruchel, A., Vettenranta, K., Horibe, K., Benoit, Y., Pieters, R., Escherich, G., Silverman, L.B., Pui, C.H. & Valsecchi, M.G. (2010) Clinical outcome of children with newly diagnosed Philadelphia chromosome-positive acute lymphoblastic leukemia treated between 1995 and Journal of Clinical Oncology: Official Journal of the American Society of Clinical Oncology, 28, Attarbaschi, A., Mann, G., Panzer-Grumayer, R., Rottgers, S., Steiner, M., Konig, M., Csinady, E., Dworzak, M.N., Seidel, M., Janousek, D., Moricke, A., Reichelt, C., Harbott, J., Schrappe, M., Gadner, H. & Haas, O.A. (2008) Minimal residual disease values discriminate between low and high relapse risk in children with B-cell precursor acute lymphoblastic leukemia and an intrachromosomal amplification of chromosome 21: the Austrian and German acute lymphoblastic leukemia Berlin-Frankfurt-Munster (ALL-BFM) trials. Journal of Clinical Oncology: Official Journal of the American Society of Clinical Oncology, 26, Bene, M.C., Castoldi, G., Knapp, W., Ludwig, W.D., Matutes, E., Orfao, A. & van t Veer, M.B. (1995) Proposals for the immunological classification of acute leukemias. European Group for the Immunological Characterization of Leukemias (EGIL). Leukemia, 9, Biondi, A., Schrappe, M., De Lorenzo, P., Castor, A., Lucchini, G., Gandemer, V., Pieters, R., Stary, J., Escherich, G., Campbell, M., Li, C.K., Vora, A., Arico, M., Rottgers, S., Saha, V. & Valsecchi, M.G. (2012) Imatinib after induction for treatment of children and adolescents with Philadelphia-chromosome-positive acute lymphoblastic leukaemia (EsPhALL): a randomised, open-label, intergroup study. The Lancet Oncology, 13, Borowitz, M.J., Devidas, M., Hunger, S.P., Bowman, W.P., Carroll, A.J., Carroll, W.L., Linda, S., Martin, P.L., Pullen, D.J., Viswanatha, D., Willman, C.L., Winick, N. & Camitta, B.M. (2008) Clinical significance of minimal residual disease in childhood acute lymphoblastic leukemia and its relationship to other prognostic factors: a Children s Oncology Group study. Blood, 111, Bruggemann, M., Schrauder, A., Raff, T., Pfeifer, H., Dworzak, M., Ottmann, O.G., Asnafi, V., Baruchel, A., Bassan, R., Benoit, Y., Biondi, A., Cave, H., Dombret, H., Fielding, A.K., Foa, R., Gokbuget, N., Goldstone, A.H., Goulden, N., Henze, G., Hoelzer, D., Janka-Schaub, G.E., Macintyre, E.A., Pieters, R., Rambaldi, A., Ribera, J.M., Schmiegelow, K., Spinelli, O., Stary, J., von Stackelberg, A., Kneba, M., Schrappe, M. & van Dongen, J.J. (2010) Standardized MRD quantification in European ALL trials: proceedings of the Second International Symposium on MRD assessment in Kiel, Germany, September Leukemia, 24, Bruggemann, M., Gokbuget, N. & Kneba, M. (2012) Acute lymphoblastic leukemia: monitoring minimal residual disease as a therapeutic principle. Seminars in oncology, 39, Campana, D. (2012) Minimal residual disease monitoring in childhood acute lymphoblastic leukemia. Current opinion in hematology, 19, Campos-Sanchez, E., Toboso-Navasa, A., Romero- Camarero, I., Barajas-Diego, M., Sanchez-Garcia, I. & Cobaleda, C. (2011) Acute lymphoblastic ª 2013 John Wiley & Sons Ltd 615