*Review* **Curing the Curable: Managing Low-Risk Acute Lymphoblastic Leukemia in Resource Limited Countries**

**Bernice L. Z. Oh 1,2, Shawn H. R. Lee 1,2 and Allen E. J. Yeoh 1,2,\***


**Abstract:** Although childhood acute lymphoblastic leukemia (ALL) is curable, global disparities in treatment outcomes remain. To reduce these global disparities in low-middle income countries (LMIC), a paradigm shift is needed: start with curing low-risk ALL. Low-risk ALL, which accounts for >50% of patients, can be cured with low-toxicity therapies already defined by collaborative studies. We reviewed the components of these low-toxicity regimens in recent clinical trials for low-risk ALL and suggest how they can be adopted in LMIC. In treating childhood ALL, the key is risk stratification, which can be resource stratified. NCI standard-risk criteria (age 1–10 years, WBC < 50,000/uL) is simple yet highly effective. Other favorable features such as *ETV6-RUNX1*, hyperdiploidy, early peripheral blood and bone marrow responses, and simplified flow MRD at the end of induction can be added depending on resources. With limited supportive care in LMIC, more critical than relapse is treatment-related morbidity and mortality. Less intensive induction allows early marrow recovery, reducing the need for intensive supportive care. Other key elements in low-toxicity protocol designs include: induction steroid type; high-dose versus low-dose escalating methotrexate; judicious use of anthracyclines; and steroid pulses during maintenance. In summary, the first effective step in curing ALL in LMIC is to focus on curing low-risk ALL with less intensive therapy and less toxicity.

**Keywords:** childhood acute lymphoblastic leukemia; low-risk ALL; risk-stratified treatment; treatment related toxicity

#### **1. Introduction**

Childhood acute lymphoblastic leukemia (ALL) is curable. Underpinning the cure for ALL is more than half a century of intensive collaborative research [1] that has systematically tested and defined highly effective drug combinations which now form the backbone of contemporary protocols. Although there are minor differences, contemporary ALL protocols are strikingly similar and almost formulaic. However, as the survival of childhood ALL in high-income countries (HIC) surpasses 90% [1], significant disparities in survival have emerged [2]. The high cure rates of ALL achieved in HIC are not seen in low-middle income countries (LMIC) [3]. With 80% of the world childhood ALL burden residing in LMIC [4,5], our success in curing childhood ALL remains limited and geographically restricted [2]. To reduce such glaring disparities, many groups such as the International Pediatric Oncology Society (SIOP), VIVA Foundation for Childhood Cancer, St Jude Global, and the World Health Organization are beginning to tackle the obstacles to widespread adoption of effective treatment. The first steps in improving cures for ALL worldwide, we believe, is for LMIC to focus on curing low-risk ALL as it is highly cost-effective and transformational. In this review, we focus on key components of contemporary trials on curing low-risk ALL, cost-effectively.

**Citation:** Oh, B.L.Z.; Lee, S.H.R.; Yeoh, A.E.J. Curing the Curable: Managing Low-Risk Acute Lymphoblastic Leukemia in Resource Limited Countries. *J. Clin. Med.* **2021**, *10*, 4728. https://doi.org/10.3390/ jcm10204728

Academic Editor: Rupert Handgretinger

Received: 28 August 2021 Accepted: 12 October 2021 Published: 15 October 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

#### **2. Causes of Failures in LMIC**

We reference the World Bank Income Group classification in defining LMIC (https:// datahelpdesk.worldbank.org/knowledgebase/articles/906519-world-bank-country-andlending-groups accessed on 1 October 2021). Specifically, in 2021, LMIC are defined as countries with gross national income per capita of <USD 12,695. Albeit imperfect and simplistic, this definition is used in the Lancet Oncology Commission on Sustainable Care for Children with Cancer [5], in which we also participated.

In HIC, the main cause of failure in treating childhood ALL is relapse [6]. This fear of relapse is so ingrained that overtreatment is rarely questioned. However, for LMIC, the key reason for failure is treatment toxicity [7]. Treatment toxicity disproportionately inflicts suffering and exponentially increases the cost of treatment [5], which then invariably leads to treatment abandonment [8]. In LMIC, treatment toxicity is further amplified by malnutrition [9], suboptimal supportive care [10], and widespread antibiotic resistance [11]. The Malaysia-Singapore ALL study group (Ma-Spore) is based in Malaysia and Singapore, an example of two countries which have emerged from LMIC status within the past three decades [12]. Given the health resource constraints, the Ma-Spore study group focused on testing cost-effective deintensification of therapy in low-risk ALL patients as one of the main strategies.

#### **3. Identifying Low-Risk Groups in Resource Limited Settings**

The key to better cures in ALL is better risk stratification [13]. Specifically, if we can define each patient's risk of relapse early, the optimal intensity of therapy can be given to maximize cure while minimizing side-effects. The two major determinants of relapse are molecular genetics and early response to therapy [6]. Early response to therapy can be measured by (1) widely available peripheral blast counts by light microscopy on day 8; and/or (2) sophisticated minimal residual disease (MRD) quantitation of sub-microscopic disease by flow cytometry or polymerase chain reaction (PCR) [2]. Based on the resources of specific hospitals and the country level of care for ALL and access to tests, we propose various levels for risk assessment of ALL (Table 1).


**Table 1.** Key stratified strategies discussed in this review on management of low-risk childhood ALL.

In LIC with very limited resources, the only diagnostic facility available may be light microscopy to identify ALL lymphoblasts. Flow cytometry may not be available to subtype ALL into B or T-ALL. Given these limitations, a simple chest X-ray revealing a mediastinal mass [28] may help in differentiating T-ALL from B-ALL. In these resource limited settings, early responses to treatment can be assessed morphologically using day 8 peripheral blast count (>1 × 109/L) and day 15 or end-of-induction (EOI) bone marrow morphology [2]. ALL may have to be managed by general pediatricians without access to specialized pediatric oncology nursing care in LIC [10]. Chemotherapy drug availability is also likely to be limited and its supply unreliable [29,30]. Because of these constraints in LIC, it is best to have one simple, common protocol that is minimally myelosuppressive. The SIOP PODC [31] and the Lancet Oncology Asian Consensus Protocol [2] are probably effective stratified regimens.

For MIC, limited panel flow cytometry to diagnose B- and T-ALL may be possible. Working with universities, MIC hospitals can potentially access PCR thermocyclers [12]. With PCR thermocyclers it is possible to run simple oncogene fusion tests to screen for *BCR-ABL1* and *ETV6-RUNX1* fusions [32,33]. In these settings, it is important to put in appropriate positive and negative controls as cross contamination and degraded RNA are common.

The ALL IC-BFM 2002 study group demonstrated the feasibility of a risk stratification approach based on a combination of modified NCI criteria, early morphologic evaluation on days 8, 15, and 33, without PCR-MRD monitoring. The study was successfully implemented in 15 countries across 3 continents in 130 centers. The ALL IC-BFM 2002 [16] study reported an excellent 81% EFS and 90% OS in the standard risk arm. Interestingly, the ALL IC-BFM 2002 standard risk criteria were defined as age 1 to 6 years and lower white blood cell (WBC) count of 20 × 109/L. The ALL IC-BFM 2002 protocol was intensive, and treatment administered in national centers with good supportive care.

#### **4. The Importance of the NCI Standard-Risk (SR) Criteria**

ALL is a genetically heterogeneous disease [13,34]. The National Cancer Institute (NCI) standard-risk (SR) criteria, presenting age of 1 to 10 years and WBC count < 50 × <sup>10</sup>9/L), are simple yet surprisingly effective risk stratification criteria [35]. Favorable genetic drivers, such as hyperdiploidy and ETV6-RUNX1, form the largest proportion of NCI SR patients [34]. In MS2003 [18], age remained prognostically significant for event-free survival (Figure 1). The NCI criteria can be easily applied even in LMIC settings and should remain one of the mainstays of risk stratification [2].

**Figure 1.** (**A**–**C**) Kaplan-Meier 10-year event-free-survival (EFS) curves from the Malaysia-Singapore ALL 2003 study cohort: significant differences in outcomes based on features from the National Cancer Institute (NCI) risk criteria such as total white blood cell (WBC) count at diagnosis and age were found; in this study risk stratification was based primarily on end-of-induction PCR-MRD responses. (**A**) 10-year EFS based on the presence of NCI low versus high-risk criteria, (**B**) 10-year EFS based on total white cell count at diagnosis, (**C**) 10-year EFS based on NCI criteria defined age groups. NCI risk criteria has been shown to clearly define groups of patients with clinically significant differences in long term EFS

regardless of MRD response at the end of induction. This is a finding that is especially relevant to children in LMIC settings who may not have access to MRD monitoring and risk stratification during treatment. Children with lower risk ALL can already be defined early at the point of diagnosis based on NCI risk features.

> Table 2 summarizes the risk stratification criteria used by various clinical trials to characterize patients with ALL who are at the lowest risk of relapse. Overall, the NCI SR criteria remain a cornerstone of ALL risk stratification for 9 of the 13 clinical trials evaluated in this review. However, the upper age limits for inclusion in these low-risk arms still vary from group to group (Table 2).

> Like those of the AIEOP-BFM ALL 2000, the NCI features were not used for risk stratification in MS2003 [18]. However, compared to NCI SR patients, NCI HR patients treated in the MS2003-SR arm did significantly more poorly (Figure 1). Given these findings, in the MS2010 [19], patients aged ≥10 years old were treated in either the intermediate or high-risk arm depending on MRD responses—and not in the lowest risk, SR arm.

> In the AIEOP-BFM ALL 2000-SR arm [20], EOI MRD-negative NCI HR patients had poorer outcomes on the less intensive Protocol III compared to standard Protocol II during delayed intensification (DI) (8-y EFS: 82.9% versus 90.4% *p* = 0.04). In contrast, NCI SR patients did equally well. The AIEOP-BFM ALL 2000 SR study concluded that, despite negative EOI MRD, age >10 years adversely affected outcomes. Taken together, despite EOI MRD negativity, both the MS2003 and ALL AIEOP-BFM ALL 2000 studies suggested that NCI HR patients, specifically teenagers, should *not* be treated with de-intensified treatment.


**Table2.**Riskstratificationcriteria,cumulativechemotherapy dosingandproportionsofpatientstreatedinlowest-risk

 arms.

\* D8 or 15 M1 marrow; \*\* Day 19 MRD < 0.01%; + D29 MRD detectable but <0.01% AND undetectable

used, 2500 units/m2 was calculated to be the equivalent of 40,000 units/m2 given over 1 week; DNA Index: (the ratio of DNA content in leukemic cells to that in normal diploid G0/G1 cells).

 MRD before start of interim maintenance;

 Day 29 responses; ˆ Pirarubicin; ˆˆ Where pegylated L-asp was

#### **5. Favorable ALL Genetics: Hyperdiploidy and ETV6-RUNX1**

Hyperdiploidy (>50 chromosomes) and *ETV6-RUNX1* ALL are associated with excellent outcomes (5-year EFS >90%). Unfortunately, the karyotyping of lymphoblasts to determine ploidy is technically challenging and different from antenatal karyotyping. Hyperdiploidy is characterized by recurrent, non-random gains in specific chromosomes: 4, 10, 17, and 18. To standardize the diagnosis of hyperdiploid ALL, the Children's Oncology Group (COG) focused on double or triple trisomy fluorescent-in-situ-hybridization (FISH) probes for chromosomes 4, 10, and/or 17 to define these favorable hyperdiploid features.

*ETV6-RUNX1* is an oncogene fusion transcript and cannot be defined by conventional karyotyping. UKALL's strategy of low-cost FISH to identify oncogene fusions such as *ETV6-RUNX1*, *BCR-ABL1* fusion probes and a *KMT2A* break-apart probe in a triple probe FISH screening strategy has been tested in low-resource settings [38,39]. In contrast, Ma-Spore and other groups have used reverse-transcription PCR (RT-PCR) to screen for *ETV6- RUNX1*, *BCR-ABL1, TCF3-PBX1,* and *KMT2A-AFF1 (AF4).* RT-PCR can be performed using standard PCR thermocyclers, which are also available in many universities including those in LMIC. Although more expensive, there are also available oncogene fusion screening kits for leukemia, e.g., QuanDx's Leukemia Fusion Gene (Q30) Screening Kit and the HemaVision Screen kit. Given that the instability of mRNA and that PCR reactions may be prone to aerosol contamination, proper positive and negative controls are critical for RT-PCR screening. Referencing and partnering with good laboratories internationally, such as VIVA-NUS CenTRAL and the St Jude Global Alliance, can also be very helpful.

Using a low-intensity protocol, COG AALL0331 [23] demonstrated excellent outcomes in children who achieved EOI MRD negativity with either favorable triple trisomy (38% of population) of chromosomes 4, 10, 17, or ETV6-RUNX1 (62% of population). COG AALL0331 showed that intensification of therapy for these low-risk patients did not improve outcomes. For this low-risk cohort, the successor COG AALL0932 [22] also reported excellent outcomes: 5-year DFS 98.8% and 5-year OS 100% with a P9904-based regimen without alkylating agents or anthracyclines in the LR-M arm. However, only 6.5% of the AALL0932 study population was eligible for this randomization; and only 3.3% of patients were assigned to the P9904-based LR-M arm. Critically, to be able to define this lowest-risk subgroup, there is a requirement for excellent cytogenetics or FISH setup defined hyperdiploidy, oncogene fusion screening for ETV6-RUNX1, and EOI MRD. For LMIC, accurate EOI MRD by flow may not be available.

In the presence of hyperdiploidy and ETV6-RUNX1, poor responses such as high EOI MRD of >1% are fortunately very rare. In MS 2003/2010 and UKALL 2003, only 3% of patients with low-risk genetics had a high EOI MRD of >1% [40]. Hence MRD monitoring for this low-risk genetic group is probably not critical. For MIC, using DNA index >1.16 may be feasible, as demonstrated by the RELLA05 [14] group in low-resource settings in Brazil. FISH for double trisomy 4 and 10 as a surrogate marker for triple trisomy is also feasible.

#### **6. Democratization of Flow Cytometry**

Interestingly, it is common for hospitals, even in LMIC, to have a good flow cytometer. The widespread availability of good, multi-color flow cytometers makes it possible to do flow cytometry for diagnosis of ALL. However, not many laboratories are trained to properly perform flow cytometry for the diagnosis of childhood leukemias. Access to a supply chain of good quality fluorochrome-labeled antibodies is also potentially a problem.

While several simple low-cost flow cytometry methodologies to measure MRD have been developed [41,42], they have yet to be widely implemented in Asia. Flow MRD needs to be analyzed and interpreted properly. In the presence of a lot of hematogones, simple, low-cost flow MRD using a limited panel of markers can yield misleading results. Flow MRD-lite end-of-induction assessment has been used in limited resource settings such as the RELLA05 study [43]. Although MRD testing is expensive, its key role in risk assignment would offset costs involved in toxicity related hospitalizations. In addition, it

can be cost effective to set up a good flow MRD-lite platform to identify the best responders that can be cured with less therapy.

#### **7. Specific Considerations for T-Lineage ALL**

Treatment de-intensification in T-ALL is much less studied and should be undertaken with caution in the context of a clinical trial. Outcomes in T-ALL have only very recently improved significantly, approaching those of B-ALL. This has been achieved with combinations of (1) the use of dexamethasone (Dexa)-based 4 drug induction, (2) a more intensive Berlin-Frankfurt-Munster (BFM) ALL backbone, (3) Capizzi escalating methotrexate, and (4) optimizing the use of L-asparaginase (L-asp). In the MS2003 study [18], which was Dexa-based, 6-year EFS rates of B and T-ALL patients were 80.7% and 80.5%, respectively.

Use of Dexa throughout all phases of therapy, like in the MS2003 [18] and UKALL 2003 [27] studies, has led to better outcomes in children with T-ALL. This better outcome with Dexa could be due to better CNS penetration, given that CNS relapse is more common in T-ALL. T-ALL patients should receive 4-drug Dexa-based induction but will require prolonged inpatient admission throughout the whole period of induction because of high risk of infections and TRM.

The largest T-ALL study, the COG AALL0434 study [44], surprisingly showed superior outcomes with the Capizzi escalating MTX plus L-asp regimen compared to HDMTX regimen (4-y DFS 92.5% ± 1.8% vs. 86.1% ± 2.4%, *p* = 0.02). However, 90% of T-ALL patients on AALL0434 received cranial radiotherapy. AALL0434 also found that addition of 5 days of nelarabine improved outcomes for IR and HR T-ALL (4-y DFS 88.9% vs. 83.3%). However, the high cost and high neurotoxicity of nelarabine will limit its use in LMIC.

#### **8. Delaying the First Intra-Thecal (IT) Chemotherapy**

Traumatic lumbar puncture (LP) with blasts is a risk factor for CNS relapse. If the first LP is performed at the time of diagnosis [45], traumatic LP occurs in up to 14%. Delaying the first IT until after clearance of circulating blasts at the end of the first week of induction would reduce incidence of traumatic LP with blasts, an adverse risk feature. This was first described in the Tokyo Children's Cancer Study Group study L89-12 [46]. The TPOG-ALL-2002 study [47] also confirmed that the delay of the first triple IT did not adversely affect survival or CNS control despite omission of cranial irradiation.

Delaying of the first IT also reduces the risk of methotrexate-related kidney injury that may be exacerbated by ongoing tumor lysis syndrome during the induction phase. Reduced need for sedation in the first week of therapy may also be advantageous to patients with large mediastinal masses at diagnosis, given the inherent risks of airway obstruction with procedures requiring sedation in such cases.

#### **9. Prednisolone/Dexamethasone-Based and 3/4-Drug-Based Induction**

Pred has historically been used during the BFM ALL induction protocol, while Dexa has been used later during DI (Figure 2). Dexa is more potent than Pred and because of better CNS penetration [48], reduces the rate of CNS relapse. The enduring question of whether Dexa is superior to Pred during induction was tested in the randomized AIEOP-BFM ALL 2000 [36] study in a 4-drug induction including anthracyclines. AIEOP-BFM ALL 2000 showed that patients with a good Pred response who received Dexa during induction had one-third the risk of relapse of those who had received Pred—a remarkable feat. However, these improvements in relapse rates were offset by the higher incidence of life threatening events during induction. Overall, despite the marked reduction in relapse rates in the Dexa arm, there were no differences in OS as relapses in the Dexa arm were less salvageable and more patients died of infections during Dexa-based induction. Subsequently the AIEOP-BFM group reserved 4-drug Dexa-based induction only for a subset of T-ALL patients with good Pred response.

**Figure 2.** Induction regimens in contemporary ALL studies for low−risk patients. Prednisolone (Pred) prophase allows management of tumor lysis. 3 drug induction refers to induction protocols which include the use of Prednisolone or Dexamethasone (Dexa), Vincristine and L-asp, with the exception of patients treated on the CoALL 07-03 [21] who received 3 drug Prednisolone-based induction consisting of Prednisolone, Vinristine, and either Doxorubicin or Daunorubicin, without L-asp. This included a pre-phase comparing the responses after a single dose of either Doxorubicin or Daunorubicin. 4 drug induction protocols include the use of anthracyclines such as Doxorubicin or Daunorubicin. For LMIC, 3-drug Dexa-based induction is safer. For 4-drug induction, Pred-based induction is probably less toxic than one that is Dexa-based.

> The Japanese L95-14 [49] and the Dana-Farber Cancer Institute ALL 91-01P [50] trials also reported a higher rate of infection-related induction deaths in the Dexa arm as compared to those who had received Pred. Infectious deaths also increased during Dexa-based induction in UKALL 97 [51] although there was overall survival benefit.

> To use Dexa during induction, a 3-drug induction without anthracyclines is feasible in LMIC. Of note, many groups such as the COG, UKALL, and the Ma-Spore ALL used Dexa-based, 3-drug induction without anthracyclines. COG and UKALL used it for NCI SR induction while Ma-Spore used it for all B-ALL patients. Without anthracyclines, Dexa-based 3-drug induction can be given safely and mainly as outpatient therapy.

> Dexa-based 4-drug induction is best reserved for T-ALL patients and used in hospitals with good isolation facilities and ability to treat breakthrough secondary infections including fungal infections. These hospitals should also have good microbiological diagnosis platforms for bacterial, fungal, and viral pathogens. Dexa-based 4-drug induction is toxic, even in the context of HIC where supportive care is good. We do not recommend Dexa-based 4-drug induction for LMIC.

#### **10. L-Asp Doses in Induction and Delayed Intensification**

A key drug in treatment of ALL is L-asp. Unlike adult ALL protocols, such as Hyper-CVAD, pediatric-inspired protocols use L-asp as a mainstay drug during induction and DI. L-asp is an enzyme derived from *E. coli* that can cause allergic reactions and silent inactivation. L-asp is moderately myelosuppressive and can cause pancreatitis and thromboembolism, especially in children > 10 years old.

After the first exposure to L-asp during induction therapy for newly diagnosed ALL, the risk of neutralizing antibodies and silent inactivation is low. Because of the lower risk of neutralizing antibodies that have to be overcome with higher doses, Ma-Spore ALL induction starts with a lower dose of L-asp during induction (7500 U/m<sup>2</sup> of Leunase spaced out to twice a week). This lower dose of L-asp during induction also reduces the risk of myelosuppression. During DI, where low levels neutralizing antibodies may already have developed, we use a higher dose of L-asp of 10,000 U/m2 every 3 days. This high dose allows for sufficient asparagine depletion during DI.

The major brands of L-asp available include: Leunase (Kyowa-Hakko), Kidrolase (Kyowa-Hakko), and Spectrila (Medac, Germany). In addition, L-asp is also manufactured by a few companies in India. The various brands of L-asp have different potencies and different risks of allergic reaction. Pegylated (PEG) L-asp has a much longer half-life than regular L-asp. However, PEG-L-asp is expensive and not registered in most LMICs. Because of these limitations, the Ma-Spore ALL study group focused on using L-asp. Erwinase is given to patients with allergic reactions to L-asp and PEG L-asp. However, Erwinase is less potent and has a much shorter half-life requiring dosing of 20,000 U/m<sup>2</sup> every 2 days to ensure complete asparagine depletion.

The St Jude Total XVI [26] study showed prolonged and more intensive asparagine depletion using higher doses of PEG L-asp (3500 U/m<sup>2</sup> versus 2500 U/m2) did not improve outcomes. Instead, it was associated with a higher incidence of toxic deaths than in an earlier study (3.2% vs. 1.4%). This prolonged asparagine depletion is also associated with increased risk of pancreatitis and long-term poor pancreatic function with diabetes mellitus.

In MS2010 [52], single doses of vincristine and L-asp were added during DI to maintain treatment intensity during a rest period at day 15. However, this led to more hospitalizations for fever, increased risk of bacteremia, and critical-care admissions, but fortunately without any increase in treatment-related mortality. The DFCI-ALL 05-01 [53] also previously described the myelosuppressive effects of asparaginase.

Because of its high costs, risks of allergy and silent inactivation, we recommend restricting L-asp use to only the induction and DI phases, especially in LMICs. To reduce the risk of allergy, Ma-Spore delayed L-asp until after at least 2 days of steroid cover had been started. We also caution against any additional doses of L-asp given that it causes increased myelosuppression and a higher risk of infections.

#### **11. Anthracycline-Free Regimens**

Anthracyclines are most used as part of induction and DI in the BFM-ALL treatment backbone. Although effective, anthracyclines cause severe immediate myelosuppression and long-term cardiotoxicity [54,55]. Because of these side effects, Ma-Spore and other groups have attempted to eliminate or reduce anthracyclines in the treatment of low-risk ALL. The COG AALL0932 [22] and MS2010 are both examples of clinical studies where anthracyclines were completely omitted in their low-risk arms. In terms of toxicity, MS2010- SR [52] revealed excellent results comparable to those of other contemporary protocols, yet with reduced toxicity. As mentioned above, in a highly selected subgroup with lowrisk genetics and that were EOI MRD negative, the COG AALL 0932 LR-M arm reported excellent results with 5-year DFS of 98.8% and 5-year OS of 100% with no anthracyclines and alkylating agents.

The CCG-105 [56] study showed that dose dense DI is only critical for *older* children. This is due to residual leukemia cells that persist after induction/consolidation which

are relatively resistant to therapy. Anthracyclines, usually doxorubicin, are used with Dexa, vincristine and L-asp (Protocol II) for intensive DI. However, for younger children who have low-risk ALL, the CCG-105 study showed that dose intensive DI is probably not critical.

#### **12. Is High-Dose Methotrexate Really Necessary?**

In LMIC, it is difficult to administer high-dose methotrexate (HD MTX) safely. This is because IV MTX > 500 mg/m<sup>2</sup> requires folinic rescue dosing and close MTX level monitoring. Although the adjustment of the start time for administration can allow MTX level monitoring during office hours, this infrequently utilized test is generally not available and not cost-effective in most LMIC settings. Although various groups have devised various strategies [57,58] to overcome challenges of giving HDMTX, we review alternatives to HD-MTX in Table 3.


**Table 3.** Consolidation and MTX dosing across clinical trials.

Differences in MTX dosing strategies are summarized and highlighted in this table, from intermediate to high-dose MTX regimens to low-dose MTX regimens including the characteristic COG dose-escalating MTX. Patients on the **CoALL 07-03** [21] trial were treated with intermediate doses of MTX but were given a combination of other drugs such as Teniposide, L-asp, and Cytarabine as well. In the **COG AALL 0932** [22], low-risk patients were randomized to receive either the P9904 regimen A-based (Arm LR-M) which included 6 courses of intermediate dose (1g/m2) MTX without any further alkylating agents or anthracyclines, essentially omitting DI entirely and completing therapy with the maintenance phase; or the CCG 1991 regimen-like outpatient-based regimen (Arm LR-C) with standard COG doseescalating MTX. Patients treated on the **JACLS ALL-02** [25] were randomized to receive either truncated BFM-like consolidation (Arm A) or low-dose Cytarabine containing consolidation (Arm B). Of all the reviewed studies, the **COG AALL 0932** [22] and **COG AALL 0331** [23] studies, together with the **UKALL 2003** [27], included the lowest intensity of MTX treatment. While of interest to LMIC or LIC groups with limited access to serum MTX drug monitoring, the excellent outcomes achieved in these studies were derived from HIC settings with individual protocol-specific caveats such as the more stringent criteria imposed by the COG to be considered as low risk, and later intensification in other parts of the protocol in the UKALL 2003, which have been reported to be toxic even in HIC settings.

> For all NCI-SR patients, instead of HDMTX, the COG study used escalating intravenous MTX that started at 100 mg/m<sup>2</sup> and did not require serum MTX level monitoring.

COG AALL0331 [23,37] and COG AALL0932 [22] reported excellent outcomes in NCI SR patients treated without HD MTX. The UKALL 2003 [27] achieved excellent outcomes without use of HDMTX for all patients. Specifically, for EOI MRD negative patients (Regimen A and B), there was no HDMTX or Capizzi MTX-L-asp; while EOI MRD positive patients had two blocks of Capizzi MTX-L-asp.

Although the COG AALL0232 [59] showed that HDMTX was superior to Capizzi MTX, the reported benefits of HDMTX over Capizzi MTX were in fact in *higher risk* B-ALL but not low-risk ALL. The newer UKALL 2011 randomized EOI MRD negative patients to receive HDMTX compared to interim maintenance.

#### **13. Delayed Intensification—Is More Necessarily Better?**

Although the importance of DI is clear, a balance between dose intensity and treatment toxicity is paramount. The BFM/COG DI Protocol II is intensive with significant toxicity, thus many groups have focused on deintensification of DI. In SR patients who were MRD negative, DCOG-ALL10 [24] successfully removed doxorubicin, cyclophosphamide, cytarabine, and thioguanine, which were replaced with a single, low-intensity Protocol IV, which consists of Dexamethasone, two doses of Vincristine, and single doses of PEG L-asp and intrathecal chemotherapy; with excellent outcomes (93% 5-y EFS and 99% 5-y OS). The randomized CoALL 07-03 study [21] also successfully de-intensified Protocol II, by removing one dose of doxorubicin and one week of Dexa in SR patients. In contrast, the large, randomized AIEOP-BFM-ALL 2000 study [20] showed that the shortened but dosedense Protocol III was paradoxically more toxic and less effective in preventing relapse. The AIEOP-BFM ALL Protocol III is shorter and highly compressed DI, resulting in more toxicity and prolonged post-Protocol III delay.

Recent studies have demonstrated that repeated DI blocks might not improve outcomes. Figure 3 summarizes the various strategies and overviews of major clinical trials in childhood ALL. The randomized ALL IC-BFM 2002 study [16] failed to show any improvement in outcomes with additional DI blocks in both standard and medium-risk patients. Similarly, CCG-1991 [60] showed no added benefit with double DI blocks in patients with standard-risk ALL. Instead, escalating MTX during interim maintenance improved outcomes. UKALL2003 randomized EOI MRD negative patients to single versus two blocks of Protocol II; one block of Protocol II was less toxic without compromising outcomes [27].

**Figure 3.** Overview of contemporary ALL protocols for lower-risk ALL: Major differences in protocol

design of delayed intensification (single versus double blocks interspersed with interim maintenance blocks), types of maintenance phases are highlighted. Consolidation phases described in this figure refer to the period following the completion of induction phase and end prior to the start of delayed intensification phase and therefore include HD MTX phases. The experimental arm of the **ALL IC-BFM 2002 study** [16] comprised two shorter DI blocks (Protocol III), split from the original single Protocol II DI. **MS2003** [18] and **MS2010** [19] studies employed a similar dosing strategy with multiple DI blocks with improvements in toxicity following dosing modifications. The **St Jude Total XVI** [26] protocol embeds dual 3 week blocks of DI in interim maintenance phases which start right after the consolidation phase. Similarly, the **UKALL 2003** [27] protocol also comprises two blocks of DI, albeit longer in duration and higher in dose intensity as compared to the abovementioned studies. **COG AALL 0932** [22] randomized low-risk patients to receive either P9904 regimen A-based (Arm LR-M), which is a very low-intensity protocol without alkylating agents or anthracyclines, or the CCG 1991 regimen-like outpatient-based regimen (Arm LR-C) with reduced vincristine/dexamethasone pulses during maintenance phase (every 12 weeks). For patients assigned to Arm LR-M, the total duration of therapy would be 2 <sup>1</sup> <sup>2</sup> years from diagnosis for both female and male patients. For those assigned to Arm LR-C, the duration of therapy would continue to be gender based: 2 years from the start of interim maintenance for female patients and 3 years from the start of Interim Maintenance I for male patients. In the **COG AALL 0331** [23] study, patients in the lowest defined risk group in the study, the standard risk-low group, were randomized to receive either standard treatment or four additional doses of PEG L-asp at 3 week intervals in an attempt to intensify treatment to improve outcomes in this group of patients. Although intensification failed to improve outcomes, the authors concluded that standard COG therapy without intensification still led to excellent outcomes in this identified low-risk group. **AIEOP-BFM ALL 2000** [36] randomized patients to receive either Prednisolone or Dexamethasone during induction. The **CCLG-ALL 2008** [17] study was based on BFM ALL treatment backbone but modified to reduce toxicity in SR patients by halving the dose intensity of early intensification after induction and before consolidation. DI was modified as per the COG with 25–33% reduction of Dexamethasone and Doxorubicin. Patients were then randomized in the maintenance phase to either receive standard maintenance therapy with vincristine and dexamethasone pulses versus a 1 week rest of mercaptopurine and MTX during the vincristine dexamethasone pulse. The lowest-risk group of patients treated on the **CoALL 07-03** [21] trial was given the reduced intensity LR-R arm with only 1 week of Dexamethasone, two doses of Vincristine and single doses of Doxorubicin with PEG L-asp in a shortened DI protocol. **DCOG ALL10** [24] includes a significantly deintensified DI Protocol IV with only 2 weeks of Dexamethasone, two doses of Vincristine and a single dose of PEG L-asp; this was followed with maintenance therapy consisting only of oral 6-MP and MTX without any pulses. The **JACLS ALL-02** [25] protocol used Prednisolone pulses during maintenance in contrast to most other groups where Dexamethasone was used during pulses with Vincristine during the maintenance phase; Pirarubicin was also used instead of the more commonly used Doxorubicin or Daunorubicin as the anthracycline of choice during induction and DI.

In the MS2003 study [18] SR arm, the DI phase consisted of two blocks of Protocol III like the experimental arms of the ALL IC-BFM 2002 study [16]. Unfortunately, as in the ALL IC-BFM 2002, there was significant toxicity during DI in MS2003 where most of the treatment-related deaths occurred [52]. Although larger randomized clinical trials investigating the effects of DI indicate that a single block of BFM/COG Protocol II is probably sufficient, the Ma-Spore chose to continue with two shorter but further modified DI blocks instead of a single block of Protocol II. In the MS2010, EOI MRD negative patients received two less intensive DI blocks without anthracyclines (Protocol V) with less treatment interruption and toxicity. Toxicity analysis of MS2010 [52] revealed significant reductions in toxicity in terms of infections as well as overall phase delays.

For low-risk ALL, it is not clear whether a strong intensive DI phase is necessary. Taken together, one block of COG Protocol II DI has been shown to be highly effective and is our recommendation for LMIC with good supportive care.

#### **14. The Malaysia Singapore Experience**

The Ma-Spore Study Group is a collaborative group of four pediatric oncology units from Malaysia and Singapore. Ma-Spore started with a MRD risk-stratified, Ma-Spore ALL 2003 (MS2003) treatment protocol [12,18]. Because of moderate resources, MS2003 focused on deintensifying therapy in MRD-negative patients.

MS2003 starts with a less myelosuppressive 3-drug Dexa-based induction to reduce the risk of severe infections during induction. The Ma-Spore treatment mantra is *"Patient first, leukemia second."* The aim was to get the patient to safety first by allowing recovery of marrow function. Depending on MRD response after induction, strength of delayed intensification therapy is tailored later to eliminate residual leukemia. The vast majority of patients with no high-risk genetics and a good day 8 Pred response received 3-drug Dexabased induction without anthracyclines. In Ma-Spore ALL studies, MRD risk stratification is by using a single PCR MRD marker at EOI and at the end of consolidation (EOC) at week 12. MS2003 focused on intensive DI by adopting the experimental ALL IC-BFM 2002 repeated Protocol III blocks. MS2003 [12] achieved a 6-year EFS of 80%, with an overall survival of 88%.

In addition, a strong collaborative network was forged between the two countries, where bone marrow MRD samples were processed and cryopreserved then couriered weekly on dry ice to the centralized laboratory in Singapore. There was also regular exchange of manpower training and knowledge sharing between the different pediatric oncology centers that extended to regional hospitals in rural areas in Malaysia. An important feature was regular telephone calls to regional hospitals to track count recovery on full blood counts, drug doses, and complications. Regional hospitals were educated on complications such as febrile neutropenia, to be able to reach emergency services in a timely manner should the need arise. Healthcare personnel in rural areas were also educated on the management of neutropenic fever and the importance of up triaging and early administration of antibiotics.

The keys to the success of Ma-Spore are the use of centralized academic molecular laboratories in the National University of Singapore and University Malaya, and a protocol design that is cognizant of moderate access to supportive care. Figure 4 summarizes key networks that have also been used in the Malaysia Singapore experience in establishing a program to treat children with ALL in settings with resource limitations.

**Figure 4.** Proposed key networks in establishing an ALL program with limited resources. "Buddy" institutions refer to more established programs that newer growing programs with limited resources can reach out to for help and advice regarding patient care and technical support.

#### **15. Infections**

In ALL, given the growing threat of antimicrobial resistance and, more recently, the COVID-19 pandemic, treatment-related infections are a major concern. The risk of infection during ALL treatment is dependent on the treatment phase and its intensity. The induction phase poses the highest risk of infections due to a combination of prolonged myelosuppression from both disease and induction chemotherapy [61,62]. Because of this exquisite vulnerability to sepsis during induction, the Ma-Spore group focused on 3-drug Dexa-based induction, which is less myelosuppressive, yet sufficiently intense to achieve sufficient complete remission to promote marrow recovery.

Although UKALL 2003 [27] had no HDMTX blocks, which made it feasible in limited resource settings, the risk of sepsis during the 4-drug Dexa-based induction protocol was still significant [62]. Given the septic deaths prevalent even in high income settings, these risks may be exponentially higher in limited resource settings. In areas with hygiene concerns and where access to supportive care may be an issue, it may be prudent to keep patients within a closer proximity during such high-risk periods.

In the St Jude Total XV [61] study, the lack of neutrophil surge after Dexa pulse, as a reflection of decreased marrow reserves, was linked to a high risk of sepsis. Dexa suppresses fever. During Dexa pulse, presence of even low-grade fever of >37.3 ◦C and severe neutropenia (ANC < 0.5 × <sup>10</sup>9/L) confer increased risk of sepsis. In limited resource settings, neutropenic fever during Dexa-based phases should be prioritized for emergency access to supportive care.

Malnutrition aggravates the risk of infection during cancer treatment [9]. In limited resource settings, training shared-care hospitals and educating families to recognize fever and signs of sepsis is critical. This involves providing clear guidelines to parents and shared-care hospitals on how to treat ALL patients with fever, regardless of neutrophil count. Prediction scoring systems could supplement multidisciplinary efforts specifically involving front line emergency department staff to improve early access to antibiotics and supportive care as a whole. The only problem is that most shared-care hospitals in LMIC settings lack staff who can learn and implement such a prediction score system, given that children with ALL probably only form a minority of cases seen.

Protocol-specific analysis of infections during treatment may help inform positive changes in protocol design. In MS2003 [52], after 2 weeks of Protocol III DI, severe neutropenic fevers were observed. These observations led to the one-week mandatory break after Protocol IIIa, which helped reduced infective complications in the successor MS2010 protocol, unlike the AIEOP-BFM ALL 2000 experience.

#### **16. Improving Supportive Care**

Up to now, we have focused on adjusting various aspects of low-risk ALL therapy to the limited supportive care available in LMIC. Good supportive care is the bedrock of our improved cancer outcomes. Without good supportive care, most of what we propose is not possible.

Improving supportive care is cost-effective. Cost-effective measures like setting up appropriate inpatient and outpatient childhood cancer wards is transformational. Cohorting children with cancer who are immunocompromised in a childhood cancer unit reduces cross infections. Common childhood viral infections such as measles and varicella are mild in normal children but can be devastating in immunocompromised ones. With childhood cancer units, both doctors and nurses can be trained to implement life-saving neutropenic fever protocols immediately and give chemotherapy safely. Overcrowding is detrimental. Low-risk ALL can be treated in dedicated outpatient cancer centers where chemotherapy beds can be quickly cleaned and reused after a short IV vincristine or IM L-asp. By recycling outpatient beds, more children can be treated, reducing overcrowding and infections [63].

Setting up a childhood cancer unit must come with improved infectious disease and intensive care unit support. Laboratory tests including FBC and blood cultures, and a safe blood supply, are critical.

#### **17. Overview of Maintenance Therapy**

Maintenance therapy is indispensable in the cure of childhood ALL and is universally part of all chemotherapy protocols for ALL [64,65]. However, the exact reason for its essentiality remains unclear. Compared to the intensive prior phases of ALL therapy, MT is only mildly intensive and simple: it comprises a 2 drug "anti-metabolite" backbone of daily oral mercaptopurine (6-MP), and weekly oral/IV/IM MTX. From initial diagnosis, the duration of ALL therapy should exceed 2 years (104 weeks). Attempts to reduce the duration of ALL therapy to 12 months (TCCSG) to 18 months (BFM) have resulted in poorer outcomes. The only times that MT is omitted is after bone marrow transplantation or CAR-T cell therapy.

#### **18. Duration of Maintenance Phase**

Although MT is only mildly intensive, toxicities remain and even deaths occur [66]. During MT, patients remained mildly immunocompromised, exposing them to bacterial, fungal, and viral infections. The long duration of MT contributes to the cumulative risk of toxicity. Attempts to intensify MT by adding VCR/Dexa pulses and rotating drug pairs such as cyclophosphamide/cytarabine (SJ Total protocols), can add to the risk of infection (see below).

Historically, for ALL outcomes, boys fared worse compared to girls. However, with modern day risk-stratified therapy, this survival gap between boys and girls has narrowed [67]. In MS2003, which is Dexamethasone-based, boys and girls do equally well. Because of this, boys are not treated differently. However, due to inferior outcome in boys, some groups treated boys with an additional year of maintenance chemotherapy [67]. Thus, most contemporary protocols no longer treat them with separate durations, with very few exceptions (such as the TPOG trial group) [68].

Compared to standard duration of 2 years, a longer duration (i.e., 3 years) of MT confers no overall survival. Although longer MT reduces risk of relapse, it increases toxic deaths which erases the survival advantage [66]. Moreover, relapse on therapy is less salvageable. Numerous attempts have been made to shorten the duration of MT down to as little as 6 months, mostly with significantly poorer results [69–71]. However, certain small subsets of patients were cured despite shorter treatment. In the Tokyo Children's Cancer Study Group's L92-13 study, where MT was truncated to 6 months, patients with *TCF3-PBX1* or *ETV6-RUNX1* fusion had favorable survival [69]. Surprisingly, hyperdiploid ALL, which is low genetic risk group, fared the worst with shortened MT. However, these analyses are based on retrospective data of failed attempts to reduce duration of therapy. A shortened duration of MT generally resulted in a higher overall rate of late relapses. Therefore, 2-year ALL therapy including MT remains the current de facto strategy.

An intriguing model to reduce the risk of immunosuppression and infectious toxicity is the Brazilian Childhood Cooperative Group for ALL Treatment's (GBTLI) use of an intermittent schedule of 6-MP and MTX [72]. Children were randomized to receive either continuous therapy (i.e., continuous oral 6-MP 50 mg/m2 daily and intramuscular MTX 25 mg/m2/week) or intermittent therapy (i.e., intermittent 6-MP 100 mg/m2 daily for 10 days and 11 days' rest, plus MTX 200 mg/m2 as 6 h IV infusion every 3 weeks, with leucovorin rescue). Here, they found that children with LR ALL treated with the intermittent schedule had improved survival than those receiving the standard continuous schedule. Significantly, there was lower severe toxicity even though the overall cumulative MTX dose was higher in the intermittent group. Notably, boys allocated to the intermittent regimen had significantly better EFS than those receiving the continuous schedule.

#### **19. VCR/Steroid Pulses**

The addition of vincristine plus steroid (VCR/steroid) pulses during maintenance therapy significantly improved EFS by at least 10% in multiple clinical trials in the 1980s [66,73]. This was subsequently adopted by all major study groups. However, the benefits of these VCR/steroid pulses in contemporary more intensive protocols are increasingly questioned. The International BFM (I-BFM) Study Group prospective randomized multi-protocols study found that IR patients who received intensive ALL BFM-backbone protocols did not benefit from six pulses of VCR/Dexa during MT [74]. A recent large randomized trial from China showed that omitting these pulses in LR patients did not impact survival outcomes [75]. However, the EORTC ALL 58951-trial showed better survival [76]. With the contemporary intensive BFM-ALL protocol, LR or IR patients probably do not need VCR/steroid pulses during MT. However, for HR patients, the role of VCR/steroid pulses during MT is still unclear.

For the lower-risk groups, reduction of VCR/steroid pulses has been studied. The COG AALL0932 for NCI-SR B-ALL found that reducing the intensity of VCR/steroid pulses from 4 weekly to 12 weekly maintained the same excellent outcomes (OS 98%), although this was not performed in the context of a non-inferiority trial [77]. In the Ma-Spore trials, VCR/steroid pulses in the SR and IR arms were also given 10 weekly (MS2003) and 12 weekly (MS2010), with excellent outcomes [18,19]. Taken together, reduction or removal of these pulses might be applicable to those with the most favorable risk groups; i.e., favorable molecular subtypes with negative minimal residual disease (MRD) throughout [78]. MS2020 will continue VCR/Dexa pulses during MT every 12 weeks for SR/IR patients and 4 weeks for HR patients.

It is important to remember a successful ALL protocol is tested as one protocol with many phases. The intensity of all the phases contributes to the successful outcome. Currently, most contemporary protocols in HIC utilize sustained and highly intensive induction and reintensification blocks. This intensive ALL-BFM and augmented BFM backbone have been highly effective. In HIC, it is feasible for families to focus on an intensive 1 year of therapy and then a lower-intensity MT. With an intensive first year of therapy, the subsequent use of VCR/steroid pulses in MT is probably less important [73]. Whether a similar finding will result from a less intensive initial backbone, such as earlier trials that showed that VCR/Dexa pulses were useful, remains to be determined. In countries with limited resources for supportive care, a more spaced out and moderate intensity protocol during the first year followed by a slightly more intensive MT (with VCR/steroid pulses) might be more manageable. However, it is to be noted that these monthly VCR/Dexa pulses during maintenance can cause severe infections, especially with prolonged Dexa pulses of 2 weeks. VCR/Dexa pulses during MT may be complicated by varicella, measles, multi-resistant bacteria, and fungal infections. For low-risk ALL in LMIC, starting with less intensive upfront phases, shorter blocks of dexamethasone (6 mg/m2/day for 5–7 days without tailing) and one dose of VCR every 4-6 weeks, is recommended. To further mitigate risk of infections, some groups have even stopped MP/MTX during the weeks of VCR/Dexa pulses with no significant issues.

#### **20. TPMT and NUDT15 Variants on 6-MP Metabolism**

Mercaptopurine (6-MP), the main anti-metabolite medication used in MT, exhibits wide interpatient variability in its efficacy and toxicity. In dosing of 6-MP, the two actionable pharmacogenetic variants are TPMT and NUDT15. TPMT variants are common in Caucasians (10%) while NUDT15 variants occur more frequently in 20% of Asians.

TPMT methylates 6-MP and thioguanine, reducing their efficacy. Low TPMT activity increases the levels of active metabolites of thiopurines (TGNs), causing myelosuppression [79,80]. NUDT15 encodes a nucleoside diphosphatase which degrades thioguanine triphosphates by dephosphorylation. This dephosphorylation of thioGTP reduces its incorporation into DNA and protecting cells from apoptosis [81,82]. TPMT and NUDT15 variants have low enzymatic activities and these act in a co-dominant manner. Specifically,

heterozygosity of TPMT and NUDT15 variants reduces the levels of TPMT and NUDT15 activities, causing mild sensitivity to 6-MP. Yang et al. proposed a thiopurine genetic score incorporating both TPMT and NUDT15 variants. In the MS2010 study [19], score 1 patients who carried either the TPMT or NUDT15 variant tolerated a reduced dose of 6-MP at 40 mg/m2/day. The frequency and type of variants affecting both enzymes vary significantly by ethnicity [83,84].

Pre-emptive testing of both TPMT and NUDT15 for possible dose modification is now standard care, with carefully established guidelines [85]. This is because doses that are customized based on TPMT or NUDT15 status reduce the likelihood of acute and severe toxicities (e.g., myelosuppression), without compromising disease control. Therefore, the risk-benefit ratio of pre-emptive genotyping is favorable and should be implemented in regions likely to have a high allelic frequency of these variants, and where testing resources are available.

#### **21. Conclusions**

Taken together, NCI SR features, low-risk genetics (hyperdiploidy, ETV6-RUNX1) and a rapid early response identify a group of patients who can be cured with lowintensity ALL therapy. Even in LMIC settings, these low-risk patients can be identified and cured cost-effectively with low-intensity protocols. Low-intensity protocols are based on two principles of (1) starting slow, with a 3-drug, anthracycline-free induction, delaying first IT to day 8, and (2) keeping safe, with low-intensity DI and uninterrupted metronomic MP/MTX maintenance. Setting up the appropriate supportive care to support the treatment protocol is as important. Adapting and testing therapy appropriate to resource-constrained supportive care and testing for TPMT/NUDT15 variants in high-frequency areas can be cost-effective. To appropriately adapt the best standards of care, partnering aspirant institutions through St Jude Global and SIOP is key. As a community caring for children with cancer, we have been fortunate. Realizing that childhood cancer is rare and we cannot do it alone make sharing experience and working together the guiding principles. By learning how to better treat low-risk ALL cost-effectively, LMIC could potentially contribute to the global ALL knowledge of how to cure with less. We are hopeful that HIC, in the near future, can learn from LMIC on "curing the curable" with less. As teachers, we learn best from our students.

**Author Contributions:** B.L.Z.O., S.H.R.L., and A.E.J.Y.: data curation, conceptualization, formal analysis, and writing—review and editing of the final manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** Supported by the Singapore National Medical Research Council Clinician Scientist Investigator (NMRC/CSA/0053/2008, NMRC/CSA/0053/2013 and NMRC/CSA/MOH/000227) and Research Training Fellowships (NMRC/RTF/MOH/000616 and Cancer Science Institute, Singapore, and Centre Grant No. NMRC/CG/NCIS/2010, Goh Foundation, Children's Cancer Foundation, Singapore Totalisator Board, and VIVA Foundation for Children with Cancer.

**Institutional Review Board Statement:** The studies involving human participants in the Malaysia -Singapore (MS) ALL 2003 and 2010 clinical trials were reviewed and approved by Domain Specific Review Board (DSRB ref: 2004/00275, 2008/00081 and 2015/00164).

**Acknowledgments:** We would like to thank Edwynn Chiew and Tai Si Ting for their kind assistance with administration and data curation in the writing of this paper.

**Conflicts of Interest:** The authors have no conflict of interest to disclose.

#### **References**


### *Review* **Pediatric Acute Myeloid Leukemia—Past, Present, and Future**

**Dirk Reinhardt \*,†, Evangelia Antoniou † and Katharina Waack †**

Pediatrics III, Department of Pediatric Hematology, Oncology and Stem Cell Therapy, University Children's Hospital, University Duisburg, 45147 Essen, Germany; Evangelia.Antoniou@uk-essen.de (E.A.); katharina.waack@uk-essen.de (K.W.)


**Abstract:** This review reports about the main steps of development in pediatric acute myeloid leukemia (AML) concerning diagnostics, treatment, risk groups, and outcomes. Finally, a short overview of present and future approaches is given.

**Keywords:** acute myeloid leukemia; children; treatment; prognosis

#### **1. Introduction**

The treatment of pediatric acute myeloid leukemia (AML) is a success story in improving prognosis. Whereas, in the 1980s, almost all children suffering from AML died, today, up to 75% of the children survive. However, this is only feasible in a well-structured setting of comprehensive diagnostics, intensive therapy, and effective supportive care. This has been achieved by the cooperative study groups in Europe, North America, and Japan. By contrast, even within Europe, the prognosis of children with AML shows an unacceptable level of inequality of survival rates, ranging from less than 50% to 80% [1].

The incidence of pediatric AML is about seven per million, with only minor differences between continents or countries. The malignant blasts originate from early hematopoietic progenitors as an evolution from (pre-)leukemia stem cells. External/environmental factors could explain only a tiny percentage. In addition, predisposing syndromes or germline mutations are associated with less than 10% of pediatric AML. During childhood and adolescence, infants less than two years old and adolescents have the highest incidence. Whereas MLL-rearranged leukemia dominates during infancy, the frequency of core-binding leukemia (CBL) and AML associated with mutations, such as NPM 1 or FLT3-ITD, increases by age [2].

Except for acute promyeloblastic leukemia (APL), improved survival has been achieved by using long-known conventional drugs, mainly cytarabine and anthracyclines. Scheduling risk group stratification, modifications of allogeneic stem cell transplant (alloHSCT), and management of complications allowed for curing most children (Figure 1) [3,4].

The only new approach within the most recent 20 years was the targeted CD33 antibody gemtuzumab ozogamicin (GO), which showed some advantages in at least one randomized trial, but GO is only approved in North America [5].

The successful therapy of APL with all-trans retinoic acid and arsenic-trioxide is a rare example of curing leukemia targeting specifically leukemia-inducing molecular mechanisms and eradicating the leukemic stem cell [5,6].

Liposomal drug formulation of daunoribiicn allowed treatment intensification without increasing toxicities, but this has disappeared due to economic reasons, such as a limited pediatric market. The actual approach with a liposomal nanoscale co-formulation of cytarabine and daunorubicin seems to be promising but needs confirmatory trials in pediatric AML and marketing approval thereof.

**Citation:** Reinhardt, D.; Antoniou, E.; Waack, K. Pediatric Acute Myeloid Leukemia—Past, Present, and Future. *J. Clin. Med.* **2022**, *11*, 504. https:// doi.org/10.3390/jcm11030504

Academic Editor: Giuseppe Milone

Received: 19 November 2021 Accepted: 13 January 2022 Published: 19 January 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Although stem cell transplantation is still an unspecific treatment option with severe acute and long-term side effects, combined with a better risk group stratification, alloHSCT significantly improved survival in children with high-risk (HR) AML [7,8].

There are a broad number of new compounds explicitly targeting signaling pathways. However, it is not confirmed in children to what extent these approaches will contribute to curing or be able to reduce toxicities by allowing reduction of treatment intensity of conventional drugs.

A fast-growing field are immune and cellular therapies, which show promising results in preclinical and early phase clinical trials (mainly in adults).

**Figure 1.** Improvement of outcome in pediatric AML. Continuous increase of survival in first remission (red), following initial non-response (green), and after relapse (blue). In parallel, the treatmentrelated mortality (black, non-relapse deaths) and non-remission deaths (grey) decreased significantly. This supported the hypothesis that the improved overall survival is based on better treatment and improved supportive care.

#### **2. Past**

Although pediatric AML has been described since the 1900s, a formal classification was established, such as in adults, in 1976 by the French-American-British (FAB)- Classification [9]. There are already six subtypes of AML that have been established and described. The regular introduction of immunophenotyping modified this morphology and cytochemistry-based classification during the 1990s [10–13]. In the WHO classification in 2001, a shift from morphology to a primarily genetically-based classification has been released and continuously extended.

Until 1968, the remission rate was inferior, and the median survival was about 1.5 months [14]. With the implementation of an intensified block therapy, including cyclophosphamide and cytarabine, a survival of 9.5 months was achieved, in 1976 [15].

Since 1975 the first clinical trials for pediatric patients with AML were initiated [16–18]. Cooperative Study groups have been established: AIEOP (Associazione Italiana di Ematologia e Oncologia Pediatrica), AML-BFM (Berlin, Frankfurt, Münster), NOPHO (Nordic Society for Pediatric Hematology and Oncology), MRC (Medical Research Council), and EORTC (European Organization of Research and Treatment of Cancer), CCG (Childhood Cancer Group), POG (Pediatric Oncology Group; merged in 2000 to COG (Children's Oncology Group); and SJCRH (St. Jude Children's Research Hospital). Whereas the MRC conducted combined pediatric and adult trials [19], the AML-BFM 78 study (1978–1983) was a pure pediatric trial to examine the application of two to six courses of daunorubicin, cytarabine, and 6-thioguanine (DAT) after a first induction of the same course [17]. In the AML 10 trial of the MRC group, the comparison of etoposide with thioguanine, as randomization from ADE versus DAT, showed no significant difference [20]. Significant progress in pediatric AML was made by the AML-BFM 83 trial, based on the introduction of block-scheduling. Within this trial, the favorable risk groups of AML with t(8; 21) and, inv(16) have been

identified and confirmed in children. The same favorable cytogenetic criteria were confirmed in the MRC AML 10. As adverse characteristics, -5, -7, del(5q), abn(3q), and complex karyotype were documented [20,21].

Although most patients were defined just by morphology, due to the high correlation between FAB M2 with Auer rods and t(8;21), the presence of atypical eosinophils in FAB M4 Eo with inv(16) or, later, FAB M2 and NPM1-mutated AML allowed event-free survival (EFS) rates of 70% and more [22,23]. Interestingly, despite further intensification, this has not changed until today. In the LAME90/91 study, patients were classified into two groups dependent on whole blood count (WBC) and cytogenetics [24]. The standard risk group included t(8;21), inv(16), t(15;17) (defined as FAB M2, M4, and M3), and also patients with <100 000/μL leukocytes, initially. The AML BFM Study Group described hyperleukocytosis as a poor prognostic marker but was not used for risk stratification [25].

Within the 20th century, the relevance of anthracycline analogs has been discussed. Randomized trials, such as AML-BFM 98 and MRC 12/15, tested idarubicin versus daunorubicin or mitoxantrone. Significant achievements were the identification of idarubicin as the most effective anthracycline if applied in a 1 to 5 conversion rate compared to daunorubicin [26,27].

In the NOPHO-93, all the patients initially underwent the same treatment (ATE-Dox), but, dependent on the response after the first induction, an extra AM (cytarabine, Mitoxantrone) induction was recommended [28]. Those patients with excess blasts after the AM course received an HA2E course.

The tested antileukemic drug in the NOPHO 2004 trial was GO and did not show any significant effect on the recurrence rate of leukemia or overall survival (OS) [29]. An extra criterium of this study was the presence of mixed lineage leukemia (MLL) rearrangements other than t(9;11), suppoeritng for the first time the independent involvement of cytogenetics in the risk stratification. During this study, especially in 2009, the criteria for the high-risk (HR) patients were restricted to the poor response [7].

The St. Jude AML02 Study stratified the patients into two subgroups according to morphologic and genetic characteristics. Patients were randomized to receive daunorubicin (50 mg/m<sup>2</sup> on days 2, 4, and 6) and etoposide (100 mg/m2 on days 2–6), and high-dose cytarabine (3 g/m<sup>2</sup> every 12 hours) [30].

The Japanese AML99 study (JPLSG) implemented a risk stratification of three groups. The initial stratification was made for the low-risk patients, and including the HR group criteria of WBC (>100.000 μL) and the age of the patients (<2 years). Additionally, the response after the first induction and the karyotype led to the allocation of the patients in the final stratification group. An alloHSCT was indicated only for the intermediate and HR groups, especially for the last group; a "not familiar" donor was suggested [31].

In the AML05 study (JPLSG), the reduced cumulative anthracycline dose (<300 mg/m2) was tested in the low and intermediate-risk patients. At the same time, 50% of the etoposide dose was used in the AML99 protocol. A higher incidence of relapse was noticed, but the OS was not influenced [32].

In the early 2000s, international cooperative projects, analyzing a larger cohort of patients, finally defined further prognostic factors [4]. This also led to a harmonization of risk group definition worldwide. All major study groups agreed on AML with t(8;21) and inv(16) as a favorable prognostic group that could be cured with chemotherapy only [23,31]. There are still controversies about the definition and post-remission therapy of the patients belonging to the intermediate-risk group. Some groups recommend alloHSCT, while others stick to chemotherapy. In addition, the HR group was not finally defined, hence remaining heterogenous. In addition, in that period, large treatment groups failed to demonstrate the advantage of alloHSCT in the HR group [33–35].

#### *2.1. Hematopoietic Stem Cell Transplantation*

AlloHSCT was introduced in the 1980s as post-consolidation therapy in pediatric AML. Although effective in some cases, it never achieved the status as a general standard in contrast to adults. This is explained by the relevant side and long-term effects and the effective chemotherapy in children, which already allowed in the 1990s a long-term survival of about 60% [36,37]. Finally, the improved risk group stratification allowed the identification of those children who benefit. Nevertheless, it is an ongoing process to identify the pediatric AML subgroups who finally benefit from alloHSCT. This includes all associated issues, such as donor selection, prevention and treatment of graft versus host disease, management of virus reactivation, and immune reconstitution.

Between the mid-1980s and the 1990s, progress in pediatric AML was limited. Autologous HSCT or alloHSCT from a matched sibling donor have been introduced to the therapy [38,39]. Although there was a reduction of relapses, this could not be translated to improved OS. Transplant-related mortality counterbalanced the potentially increased antileukemic effect [40]. Considering the significantly higher risk of post-transplant late sequelae, long-lasting controversies about the relevance and importance of alloHSCT in first complete remission occurred [33,40–44].

In the AML-BFM 87 study and the MRC AML 12 trial, the response on day 15 after the first induction was added as a criterium for stratification [27,45]. Whereas alloHSCT was not generally recommended in the AML-BFM trial, the British study group limited alloHSCT to the intermediate and poor-risk group [27].

#### *2.2. CNS Prophylaxis and Treatment*

A monotherapy with intrathecal cytarabine was used as prophylactic therapy in the CCG-2891 and as a treatment twice a week in a total of six doses for the central nervous system (CNS) positive patients [46]. In Italy, the VAPA protocol, the first conducted study in children, also included monotherapy with cytarabine with 12 doses for all included patients [47]. Triple intrathecal therapy with methotrexate, cytarabine, and hydrocortisone was administered in the MCR AML10 and NOPHO 2004 twice a week by CNS positive patients until cerebrospinal fluid (CSF) clearance. The difference was the clearance duration after the extra intrathecal inductions, one and two weeks, respectively, for the MRC and NOPHO treatment groups [29].

In the AML-BFM-87 study, prophylactic irradiation was randomized. However, an increased rate of relapses in children without cranial irradiation led to a premature stop [45]. Since then, the AML-BFM studies have included cranial irradiation as a mandatory treatment for all patients, except those who received alloHSCT [48]. In contrast, the CCG and MRC group applied cranial irradiation only if the CSF was not cleared after the intrathecal therapy [35,49]. The standard therapy in the AML02 study of the JPLSG included triple intrathecal therapy in each course but no prophylactic cranial irradiation [31]. Based on the more intensive and CNS-effective chemotherapy, since 2012 the AML-BFM group no longer applied prophylactic cranial irradiation. Only in patients with CNS involvement is irradiation still recommended [50].

To summarize, different intrathecal therapies, such as monotherapy or triple treatment, were administered from the studies mentioned above. Cranial irradiation was excluded from the standard treatment and is now only recommended for children with CNS involvement.

#### *2.3. Development of Minimal Residual Disease (MRD) Diagnostics*

In parallel to the intensified therapy, the relevance of genetic risk groups and treatment response became obvious. Improved techniques, such as multicolor immunophenotyping and quantitative PCR, allowed a more precise response detection.

Detection of MRD by multicolor flow cytometric immunophenotyping started during the 1990s. The leukemic blasts were selected via different antigens (CD45 and CD34, CD117, CD13, CD15, CD33, etc.).

Langebrake et al. defined different response measurements within AML subtypes, including the prognostics relevance [51]. In contrast, in the SJCRH AML02 study, a higher incidence of relapse was noticed in patients with MRD 1% after the first induction and

>0.1% after the second induction. MRD was characterized as a poor prognostic factor EFS and OS [52].

In the NOPHO 2004 study, a difference in the EFS, but not in OS, comparing the MRD positive with the morphologic positive patients was noticed [53].

Within the Dutch Childhood Oncology Group (DCOG) ANLL, 97, and the MRC 12 trials, MRD levels measured by flow were prognostically favorable for the patients achieving MRD negativity after the first/second induction [54].

Polymerase chain reaction (PCR) is also used to detect MRD for different fusion transcripts. The correlation of expression with the clinical progress of the patients is supported in many studies. The data presented are promising for using this method as a sensitive diagnostic tool. The studies include a small number of patients for the specific subpopulations in the pediatric population, such as t(8;21), in(16) [55].

#### **3. Present**

The cooperative trial groups achieved significant improvements in overall survival. Table A1 (Appendix A) summarizes recent results, showing similar despite different chemotherapy schemes.

The analysis of the AML-BFM trials between 1993 and 2010 revealed a continuous improvement of OS but limited progress of EFS [50]. This suggests that 2nd line treatment plays a relevant role in explaining the increasing gap between EFS and OS [56].

Treatment regimens include initial double induction and 2- or 3-consolidation blocks. Even if the definitions vary, it is evident that about 60% of children with AML can be cured by chemotherapy only. On the other hand, a significant achievement within the ongoing trials was the establishment of a risk group dependent indication for alloHSCT in the 1st CR [4].

Along with the improvements in the transplant procedure and the option to rescue refractory AML, EFS improved [8]. A condition is supportive care to prevent and manage expected complications and more precise diagnostics to allow a genetic and response-based stratification [4].

#### *3.1. alloHSCT*

The AML-BFM Study group and the NOPHO proved that alloHSCT in the 1st CR of high-risk pediatric AML improves EFS, which is not significantly lower than intermediaterisk (IR) or standard risk (SR) [7,8]. However, due to the intensified therapy, including alloHSCT, the salvage treatment is ineffective, compared to IR/SR, resulting in a still inferior OS.

In case of relapse, all patients indicate alloHSCT in the 2nd CR. Although the prognostic characteristics of relapsed AML are impaired, the survival rate has been maintained or improved [56,57].

This allows treating the "right" patient group with the "right" intensity. In particular, the precise definition of the high-risk group by genetics and response associated with the indication of alloHSCT in the 1st CR eliminated significant differences in EFS.

To achieve this, the results of the AML-BFM 2004 trial have been re-analyzed. Based on the genetic characteristics and augmented by the treatment response to the 1st and 2nd induction, three risk groups could be defined [58].

In the AML-BFM 2012 Registry, this risk group stratification has been implemented. The NOPHO Group and others have published similar reports [7].

Figure 2 shows the improvement of the HR Group in the AML-BFM 2012 Registry, including the suspension of significant differences between the risk groups.

In addition to the improved risk group stratification, the selection of conditioning regiments, preparation of the transplanted stem cells, the donor identification and availability, and the graft-versus-host prophylaxis significantly contributed to a better outcome in the 1st/2nd CR but also in children with a refractory AML [59]. Within the de-novo AML patients who were transplanted in the 1st CR, the OS increased continuously between 1981 and 2019, documenting the improvement of the treatment approach over time (Figure 3).

**Figure 2.** EFS of risk groups in Study AML-BFM 2004 [58] and AML-BFM registry 2012 [8].

**Figure 3.** Increase of overall survival of children with AML and HSCT in CR1 (any donor/conditioning) since 1981; 10 year-periods (AML-BFM trials).

Most groups have accepted the standard for myeloablative conditioning with busulfan, melphalan, and cyclophosphamide. Earlier studies with less intensive conditioning (busulfan/cyclophosphamide) resulted in unacceptably high relapses [37]. However, concerns about severe acute toxicities, especially in adolescents, supported the application of alternative regimens, such as treosulfan, fludarabine, and thiotepa [60,61]. Other regiments include clofarabine, busulfan, and fludarabine [59].

Regarding stem cell selection, the CD3/CD19-depleted graft transplantation of bone marrow or apheresis cells is the most widely used approach. The donor selection included matched sibling donors (MSD) or matched unrelated donors (MUD), defined as a 9/10 or 10/10 allele match for the HLA loci A, B, C, DR, and DQ, as determined by molecular 4-digit high-resolution typing. For the HR patients without a matched donor, a haploidentical donor is accepted [59].

An unexpected, good outcome has been achieved in children with refractory AML, who got fludarabine/amsacrine (FLAMSA) and reduced conditioning (TBI/DLI). The reported 4-years EFS of 41% seems to be promising because, in the past, almost all patients of this cohort died [59].

#### *3.2. Diagnostics*

The diagnostic of pediatric AML requires morphology, immunophenotyping, and comprehensive cyto- and molecular genetics of the leukemic blasts. All available methods, such as multicolor flow with at least eight colors, panel-next-generation sequencing (NGS), and RNA seq, must be integrated. In general, the risk groups definition can be based mainly on genetics augmented by response measurement by flow and morphology (AIEOP/AML- BFM/FRANCE/UK/COG/Japan), or visa-versa, preferentially MRD-driven augmented by genetics (NOPHO) [62].

Table 1 shows the definition of risk groups according to genetics aberrations and response. The rarity and, in several cases, cryptic translocations require high qualification of the reference laboratories to provide reliable results within a short time frame.

**Table 1.** Risk Group definition by genetics and response, an example from the AIEOP-BFM AML 2020 Study.


Today, measurement of residual disease by immunophenotyping is the most appropriate method to define initial treatment response. The ongoing treatment protocols use residual disease detection either by immunophenotyping only or in combination with morphology for treatment stratification. The different response kinetics of fusion genes (KMT2A; AML1/Eto, CBL/M) and mutations (NPM1, FLT3-ITD, WT1), measured by quantitative PCR, makes this approach suitable and prognostically relevant only in some subgroups, such as PML/RARA. However, continuous monitoring after remission allows the early detection of molecular relapse. Although it is not entirely proven yet, the treatment of molecular relapse might be feasible with less intensive chemotherapy as bridge to transplant option. The international AmoRe 2017 trial (conducted by GPOH as sponsor) should allow alloHSCT without toxic re-induction in children with a molecular relapse by applying the epigenetically-effective low dose azacytidine. A reduction of MRD -levels to less than 10−<sup>3</sup> should allow direct alloHSCT.

#### *3.3. Myeloid Leukemia of Down Syndrome (ML-DS)*

Until almost the end of the 20th century, patients with AML and Down syndrome (DS) were treated identically with the whole group of pediatric AML [63]. In the NOPHO AML-93, after the same treatment, a better 5-year survival was obsereved [64]. In CCG Studies 2861 and 2981, a significantly better 4-year-EFS and no benefit of the BMT was achieved in the patients with DS [65]. A significantly lower relapse rate was noticed in the MRC AML10 study and the BFM-83 and 98 studies [66,67]. Consequently, the therapeutic schema was modified to minimize the toxicities for this favorable group. In the AML02 Trial (JCCSG) patients with ML-DS were treated separately, and a stratification depended on the response after the inductions were implemented [68,69]. In the MRC AML 12, they were allocated for only four courses of chemotherapy and were not eligible for alloHSCT [70]. In the AML-BFM-93 study, the treatment of the DS patients included reduced doses of anthracycline and no high-dose cytarabine/mitoxantrone or cranial irradiation [71]. Since then, these patients have been treated with lower doses of chemotherapy. No maintenance therapy is recommended. Contrary to the excellent response and survival in the case newly diagnosed ML-DS, the overall survival after a relapse, which affects less than 10% of those patients, remains disappointing [65,72].

#### *3.4. Supportive Care*

All the improvements of the recent decades would be impossible without the progress in supportive care. The introduction of prophylactic antimycotic and, effective antibiotic regimens as well as improved intensive care, including sufficient, sensitive, and specific microbiologic diagnostics, enables intensive treatment with a limited rate of treatmentrelated death and toxicity [73–75]. In addition, recent data confirmed that strict separation of children under immunosuppressive therapy might not be required. The best strategy could be to react immediately with a very high level of awareness and structures in pediatric oncology sites [76]. Unfortunately, these structures are only given in some developed countries.

#### *3.5. Long-Term Toxicities*

Pediatric AML and intensive treatment are associated with relevant long-term sequelae. All organ systems could be involved. Although much attention has been spent on cardiotoxicities, especially anthracycline-induced cardiomyopathy, severe damages of the liver, renal function, and endocrinology must be considered. Recent data showed the increased risk of early-onset cardiovascular diseases, neurology, and mental diseases [68]. In addition, treatment-induced malignancies occur in 2 to 5% within 10 to 20 years posttreatment. Unfortunately, to date, there is no plateau of the cumulative incidence [77].

#### **4. Future**

The therapy of pediatric AML is still based on intensive chemotherapy and, if necessary, alloHSCT. Despite significantly improved survival chances, this therapy has severe acute and long-term side effects [68,78,79]. Therapy-related toxicity is also relatively high at 2 to 4% [75] The aim of new, innovative therapies must be a more targeted treatment, presumably with fewer side effects, without jeopardizing achieved the results. Another aspect is that cure has a the highest priority. While, in adult disease with an age peak above 70 years, it may be beneficial to gain control of the disease for several years, cure must remain the primary aim in children.

Early deaths from AML in children and adolescents continue to be a significant problem [80–82]. While, in some cases, the course is fateful due to the disease dynamics, on the other hand, a higher awareness of pediatricians, general practitioners, and pediatric hospitals could rescue some children. This is especially true for APL and monoblastic leukemia. These must be considered acute emergencies and treated immediately in cooperation with an experienced pediatric oncology center. Effective, quality-assuring structures (central consultation, reference laboratories), as established in some European countries, improve the chances of survival [83].

Several achievements will allow more reliable and precise diagnostics, mainly based on NGS, genome mapping, RNA seq, acetylation/methylation assays, and molecular single-cell characterization [84–86]. The challenge will be to integrate the complex data into meaningful results, allowing clinical decisions, better stratification, and more precise treatments. The reliable measurement of MRD by NGS-based approaches will cover all patients and give better insights into the fate of leukemic stem cells, clonal hierarchies, and evolution [86–89].

Regarding more precise treatment options, the differentiating therapies with all-trans retinoic acid (ATRA) and arsenic trioxide (ATO) in APL are already in use. For the first time, this allows the cure without chemotherapy and with significantly reduced side effects [90].

The use of antigen-mediated therapies was successful, especially with the CD33 specific and ozogamicin-coupled antibody GO (Mylotarg) [91]. However, even the positive randomized trials in adults and children have not yet led to a general marketing authorization in pediatric AML [5]. Other specific-acting agents, such as FLT3 or IDH1/2 inhibitors, have also shown efficacy in children and adolescents. Still, it has not been conclusively investigated whether this improves the chances of a cure or only means an effective but transient blast reduction [61]. The same is probably true for epigenetic approaches, which have been used very successfully in older adults. Combining BCL-2 inhibitors (venetoclax) with low-dose, mainly epigenetically active chemotherapy (e.g.azacytidine) modifies the clinical course of myelodysplastic syndromes/AML, especially with low proliferation activity, very positively with significantly improved survival [92]. Other approaches combined venetoclax with MDM2 or FLT3-ITD inhibitors; however, experience in children is lacking. The addition of venetoclax to high-dose cytarabine with or without idarubicin revealed a promising overall response of 69% [93]. Nevertheless, the contribution of venetoclax to this response rate needs to be evaluated. In summary, the relevance of venetoclax in pediatric AML needs to be confirmed in further trials (such as planned within the LLS-PedAL initiative).

Another complex area of AML therapy includes various immunotherapies. One approach will be post-HSCT immunomodulation. The effectiveness has already been shown in post-HSCT treatment with donor-lymphocyte infusions (DLI) but definitively provides more options. Cytokine-induced killer (CIK) cells have already been introduced to clinical trials [94]. Other approaches use activated NK-cells [95]. In particular, the combination with immunomodulatory agents that optimize cellular treatments shows promising efficacy in preclinical and initial clinical studies. Directly related to these approaches are the current research findings on the importance of the microenvironment [96], its interaction with leukemic blasts, and the effects it induces on the selective proliferation of malignant cells, support of escape mechanism of leukemic stem cells, and inhibition of immunocompetence of effector cells, such as T/NK cells [97].

The successful cellular therapy approaches in B-cell lymphocytic leukemia/lymphomas raise high hopes for myeloid neoplasms [98]. However, the challenge in pediatric AML is more complex, while it is relatively easy to compensate for B lymphocyte eradication with antibody substitution, the reconstitution of myelopoiesis is only feasible with alloHSCT. Accordingly, to date, these therapeutic options must be viewed primarily as "bridge-totransplant" regime. Nevertheless, cellular treatment options, such as gene-engineered T-CAR or NK-CAR cells, are promising approaches to enable more precise and hopefully less side-effective therapy in the future [99–101] Several targets have been addressed so far (CD33, CD123, CLL1, and others) [101–103].

Overall, it is unlikely that there will be "the one" effective therapy for a heterogeneous disease like AML. Only the optimized combination of all available options will allow a further, significant improvement of cure rates so that likely different treatments adapted to the AML subtype are needed. This underlines the need for further comprehensive research into the mechanisms of leukemogenesis, specific therapies, and, above all, systematic clinical research to develop scientifically-validated treatments, despite the small number of cases. This will only be possible if the international collaboration between the study groups will be further improved on a global level. It is important to learn more about small subgroups and more precise treatment, as well as reducing inequalities between countries and continents. The recently established Leukemia & Lymphoma Society (LLS), Pediatric Acute Leukemia (PedAL) initiative and the European Pediatric Acute Leukemia (EuPAL) foundation are on the way to launching such a platform in North America, Europe, Australia, and, hopefully, in Japan [3,104].

In this context, the therapies of AML in children, even if they have low economic impact, should not be considered exclusively as a "waste product" of adult medicine but should have a right to their own, child-specific therapy development. This applies to both, research on pediatric therapies and the timely establishment of therapies that have been successfully used in adult AML.

**Author Contributions:** Conceptualization: D.R. and K.W.; resources: D.R., E.A. and K.W.; writing and original draft preparation: D.R., E.A. and K.W.; writing, review and editing: D.R., E.A. and K.W.; visualization: D.R.; supervision: D.R.; project administration: D.R.; funding acquisition: D.R. and K.W. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Appendix A**

**Table A1.** The outcome of pediatric AML. Clinical trials by the Cooperative Study Groups.


AIEOP (Associazione Italiana di Ematologia e Oncologia Pediatrica), AML-BFM (Berlin, Frankfurt, Münster), COG (Childhood Oncology Group), JACLS (Japanese Association of Childhood Leukemia Study), JPLSG (Japanese Pediatric Leukemia Study Group), NOPHO (Nordic Society for Pediatric Hematology and Oncology), MRC (Medical Research Council), PPLLSG (Polish Pediatric Leukemia/Lymphoma Study Group), SJCRH (St. Jude Children's Research Hospital).

#### **References**


### *Review* **Current Treatment of Juvenile Myelomonocytic Leukemia**

**Christina Mayerhofer 1, Charlotte M. Niemeyer 1,2 and Christian Flotho 1,2,\***


**Abstract:** Juvenile myelomonocytic leukemia (JMML) is a rare pediatric leukemia characterized by mutations in five canonical RAS pathway genes. The diagnosis is made by typical clinical and hematological findings associated with a compatible mutation. Although this is sufficient for clinical decision-making in most JMML cases, more in-depth analysis can include DNA methylation class and panel sequencing analysis for secondary mutations. *NRAS*-initiated JMML is heterogeneous and adequate management ranges from watchful waiting to allogeneic hematopoietic stem cell transplantation (HSCT). Upfront azacitidine in *KRAS* patients can achieve long-term remissions without HSCT; if HSCT is required, a less toxic preparative regimen is recommended. Germline *CBL* patients often experience spontaneous resolution of the leukemia or exhibit stable mixed chimerism after HSCT. JMML driven by *PTPN11* or *NF1* is often rapidly progressive, requires swift HSCT and may benefit from pretransplant therapy with azacitidine. Because graft-versus-leukemia alloimmunity is central to cure high risk patients, the immunosuppressive regimen should be discontinued early after HSCT.

**Keywords:** juvenile myelomonocytic leukemia; RAS signaling; hematopoietic stem cell transplantation; 5-azacitidine; myelodysplastic/myeloproliferative disorders; targeted therapy

#### **1. Introduction**

JMML is a pediatric leukemia with shared features of myelodysplastic and myeloproliferative neoplasms, usually manifesting during early childhood with leukocytosis, thrombocytopenia, pronounced monocytosis, splenomegaly, immature precursors on peripheral blood (PB) smear, and bone marrow (BM) blast count below 20% [1–3]. Its clinical and hematological picture, as well as natural history and outcome, are remarkably diverse [4]. The common molecular denominator of JMML is the deregulation of the intracellular Ras signal transduction pathway, caused in >90% of cases by mutations in one (or, rarely, more than one) of five primordial genes (*PTPN11*, *NRAS*, *KRAS*, *NF1*, or *CBL*) [5]. For most patients, allogeneic hematopoietic stem cell transplantation (HSCT) is the only curative treatment option, in contrast to a smaller percentage of children who survive long-term without HSCT and eventually experience spontaneous clinical remissions [6,7]. Clinical and molecular risk factors were established to help predict the disease course and guide therapeutic decisions, including age at diagnosis, percentage of fetal hemoglobin (HbF), platelet count, and aberrant DNA methylation patterns [8,9]. In this article, we review the current knowledge of genetic and epigenetic properties of JMML and provide detailed recommendations for the clinical management of children diagnosed with this challenging disorder.

#### **2. The Origin of JMML: The Ras Pathway**

The Ras pathway is a sequence of kinases in the cell that serves as a chain of communication between extracellular mitogens and the cell nucleus [10]. External cytokine signals, relayed through receptor tyrosine kinases and intracellular adapter proteins, lead

**Citation:** Mayerhofer, C.; Niemeyer, C.M.; Flotho, C. Current Treatment of Juvenile Myelomonocytic Leukemia. *J. Clin. Med.* **2021**, *10*, 3084. https:// doi.org/10.3390/jcm10143084

Academic Editor: Rupert Handgretinger

Received: 22 June 2021 Accepted: 10 July 2021 Published: 13 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

to guanosine exchange factor-mediated transformation of Ras proteins into their active guanosine triphosphate (GTP)-bound state (reviewed in more detail in [11,12]). The Ras signal is terminated by intrinsic Ras phosphatase activity, which converts Ras back to an inactive guanosine diphosphate (GDP)-bound configuration. An additional layer of regulation is provided by GTPase activating proteins (GAPs). Effects of Ras activation include the subsequent phosphorylation of Raf, Mek, and Erk kinases [13–17], activation of the mammalian target of rapamycin (mTOR) axis via phosphoinositide 3-kinase (PI3K) [18], and others [19]. Among nuclear targets are the transcription factors Jun and Fos [20].

Genetic mutations in specific Ras pathway components (*PTPN11*, *NRAS*, *KRAS*, *NF1*, or *CBL*), resulting in net hyperactivation of the Ras-GTP-GDP loop, are present in hematopoietic cells of >90% of children diagnosed with JMML [4,21–26]. These can be traced back to early myeloid stem/progenitor cell compartments [27–29], and they are found in patient cord blood samples [24], substantiating their role as initiating events and suggesting the inception of the leukemogenic sequence before birth [30].

Somatic mutations in exons 3 or 13 of the *PTPN11* gene are present in ~35% of JMML cases [22,31], resulting in a gain-of-function of the nonreceptor tyrosine phosphatase Shp2 [32]. Somatic mutations in *NRAS* or *KRAS* codons 12, 13, or 61, accounting for ~25% of JMML cases [4,25,33], freeze Ras in its active GTP-bound form by inhibition of GTPase activity or resistance to GAPs [4]. Somatic *PTPN11*, *NRAS*, and *KRAS* mutations occur in heterozygous form in JMML, indicating strong cell-transforming capacity already in monoallelic fashion.

Two congenital developmental disorders predispose to JMML: NF-1 and CBL syndrome [26,34–36]. Here, the germline of the patient carries a monoallelic loss-of-function mutation of the *NF1* or *CBL* gene, which may have been inherited or arisen de novo. JMML develops after somatic biallelic inactivation of the respective gene in hematopoietic progenitor cells, predominantly by mitotic gene recombination resulting in uniparental isodisomy [21,37]. *NF1* functions as a Ras-GAP and thus negatively regulates the Ras pathway [38,39]. Indicative features in children with JMML/NF-1 are the presence of ≥6 cutaneous café au lait spots and/or the family history; other characteristics of NF-1, such as neurofibromas, optic pathway gliomas, bone lesions and neurological abnormalities, usually manifest only later. Overall, 10–15% of JMML cases are driven by *NF1* [33,40]. CBL is a E3 ubiquitin ligase mediating the decay of receptor tyrosine kinases in the Ras pathway. Mutations targeting exons 8 or 9account for ~15% of JMML cases [26,33]. CBL syndrome, a Noonan-like rasopathy, has a wide phenotypic spectrum. Features include impaired growth, facial anomalies, developmental delay, cryptorchidism, autoimmune phenomena, and notably, neurovasculitis [26,37]. However, it is not rare for patients with JMML and *CBL* germline mutation to display no abnormalities at all [26,41,42].

Noonan syndrome (NS), the most common rasopathy with an incidence of 1 in 1000–2500 children [43], bears clinical similarities with Turner syndrome. Patients with NS exhibit a short statue, facial dysmorphism, congenital heart defects, skeletal defects, a webbed neck, mental retardation, and cryptorchidism. The genetic basis is a germline mutation in *PTPN11* (around 50% of NS cases), *SOS1*, *RAF1*, *KRAS*, *BRAF*, *NRAS* or other members of the RAS pathway [5,22,44,45]. Children with NS may experience a polyclonal myeloproliferative disorder (MPD) at a very young age, sometimes shortly after birth [4,46]. Although the condition is indistinguishable from JMML by clinical and hematological features, it has a self-limiting course in the vast majority of cases. Only a small fraction of children with NS/MPD progress to JMML, presumably after the acquisition of additional genetic changes [5,47]. Although the landscape of *PTPN11* mutations is not identical in JMML and NS/MPD [31], there is considerable overlap, and it is not well understood how the same mutation elicits a transient disorder when present in the germline and a fatal disorder when acquired somatically. Obviously, the occurrence of germline and somatic Ras pathway mutations in the same clinical context requires analysis of non-hematopoietic tissue (e.g., hair follicles or skin fibroblasts) to differentiate these conditions [9].

Systematic exome sequencing studies revealed that JMML is generally characterized by a paucity of somatic mutations in the neoplastic clone when compared to most other types of cancer [48]. However, subclonal secondary gene mutations can be found in up to half of the cases [23,24,48]. These mutations primarily target the *SETBP1*, *JAK3*, *SH2B3*, or *ASXL1* genes. Not infrequently, the secondary mutations affect the Ras pathway itself ("Ras double mutants"). In addition, a role for subclonal mutations in the Polycomb Repressive Complex 2 network was highlighted [23]. Several studies have linked the presence of secondary mutations with an aggressive clinical course or disease progression [23,49]. Furthermore, an association with an increased risk of recurrence after allogeneic HSCT was demonstrated [23,49].

Less than 10% of JMML cases are negative for the five canonical driver mutations. Rarely, these children harbor germline or somatic activating *RRAS* mutations [23,50]. Recently, a *CCDC88C-FLT3* fusion responsive to sorafenib was described in a pediatric patient with clinical features of JMML and monosomy 7 [51]. Other fusions detected in children with myeloproliferative disease include *ALK* [52,53], *ROS1* [52,53], *FIP1L1-RARA* [54], *HCMOGT-1-PDGFRB* [55], *NDEL1-PDGFRB* [56], and *NUP98-HOXA11* [57]. Although kinase fusion-positive cases without Ras pathway mutation may fulfill the clinical and diagnostic criteria of JMML, they likely represent a genetically distinct myeloproliferative neoplasm in childhood. When identified, these tyrosine kinase fusions offer an attractive target for personalized therapies [51,53].

#### **3. Clinical and Hematological Features of JMML**

JMML occurs in 1.2 children per million per year, accounting for 2% of pediatric hematopoietic malignancies [58]. One half of the children with JMML are diagnosed below the age of two years and two-thirds are male [59]. Clinical signs at diagnosis include nonspecific symptoms such as infections, fatigue, or failure to thrive. Splenomegaly is noted in nearly all cases, often accompanied by hepatomegaly and lymphadenopathy. Pulmonary infiltration by leukemic cells manifests with dry cough, tachypnea and, radiologically, interstitial infiltrates [59,60]. Abdominal symptoms may arise in patients with intestinal infiltration [59]. Variable cutaneous features may be present, ranging from eczematous lesions to erythematous papules or nodules and/or petechiae [60]. In contrast to other pediatric leukemias, JMML does not usually invade the central nervous system. As a substantial proportion of JMML cases arise on the basis of an underlying predisposition syndrome, the clinician needs to examine the patient carefully, paying attention to growth, facial dysmorphism, congenital heart defects, skeletal anomalies, developmental status, and skin lesions such as café-au-lait macules or juvenile xanthogranulomas [45,61–63].

The PB smear typically shows mild to pronounced leukocytosis with monocytosis without a significantly increased number of blasts (median 2% myeloblasts) [60,64]. Immature precursor cells of the granulocytic lineage (myelocytes, metamyelocytes), immature monocytes and nucleated erythropoietic cells are found, giving the blood film examination a pivotal diagnostic role [59]. Platelet and erythrocyte counts are usually decreased, whereas the absolute monocyte count is increased to >1 × 109 G/L in all but exceptional cases [1]. Bone marrow examination is necessary to exclude acute leukemia, but is per se insufficient to confirm a suspected diagnosis of JMML. BM findings in JMML include hypercellularity from myelomonocytic proliferation, reduction of megakaryocytes and moderate increase of blasts (<20% myeloblasts) [59].

The combination of young age, splenomegaly, skin lesions, appearance of myeloid and erythroid precursors in the PB, and/or elevated levels of HbF should prompt the pediatric oncologist to suspect JMML and initiate specific tests. First of all, this involves the molecular analysis of driver mutations in the *PTPN11*, *KRAS*, *NRAS*, and *CBL* genes, and a meticulous search for features of NF-1 including family history. Genetic analysis of *NF1* can be added, but it is laborious, and the interpretation of findings is not always straightforward. On cytogenetics, two-thirds of cases exhibit a normal karyotype. Monosomy 7 is the most frequent aberration [33,53,65], occurring in combination with *PTPN11* and *KRAS*

mutations, but rarely with *NF1*, *NRAS*, or *CBL*. A traditional hallmark of clonogenic JMML cells is their hypersensitivity to GM-CSF in vitro [66,67]. However, laboratory tests of this feature are poorly standardized and not widely available. In the era of efficient mutational analysis, GM-CSF hypersensitivity has become largely dispensable, but may potentially be helpful in occasional cases without a canonical driver mutation. The direct antiglobulin test may come back positive due to autoantibodies, but this is usually not accompanied by clinical or laboratory hemolysis [59]. Similarly, increased levels of IgG, IgM and IgA can be observed [59]. It was suggested that flow cytometric analysis of STAT5 hyperphosphorylation after stimulation with GM-CSF may aid in distinguishing JMML from other conditions [68].

Bacterial [69] and viral (e.g., Epstein–Barr virus [70], cytomegalovirus [71], and herpesvirus 6 [72]) infections can mimic the clinical and laboratory findings of JMML in infants, including fever, splenomegaly, leukocytosis with monocytosis, hypersensitivity to GM-CSF, and STAT5 hyperphosphorylation. Genetic or non-leukemic hematological disorders, such as infantile malignant osteopetrosis, leukocyte adhesion deficiency, Wiskott–Aldrich syndrome, or Ras-related autoimmune lymphoproliferative disease (RALD), must also be differentiated from JMML [69,73,74]. The latter is a non-malignant, chronic condition induced by an apoptosis defect in lymphocytes [74,75]. RALD is characterized by monocytosis, lymphoproliferation and autoimmune phenomena. Blood leukocytes exhibit similar somatic *NRAS* and *KRAS* mutations as in JMML [74], but these patients do not require aggressive treatment. Two cases of JMML evolving from RALD were described in the literature [75,76], highlighting the need for close observation. Differentiation of both entities can be difficult in the absence of monosomy 7. Functional apoptosis assays might be helpful to diagnose RALD [75].

#### **4. The Emerging Role of Epigenetics**

The genetic subtypes discussed above account for the phenotypic diversity of JMML only incompletely. For example, long-identified prognostic parameters, such as age of the patient, sex, platelet count, or elevated levels of fetal hemoglobin [6,40,59], do not correspond to a specific Ras genotype. Further molecular factors related to the course of the disease were observed in JMML, including micro and long non-coding RNA expression [77,78], AML-like expression profile [79], secondary mutations [48,49] and alterations of the fetal hematopoietic regulator gene *LIN28B* [80]. In addition, the previous observation of epigenetic dysregulation during Ras-mediated transformation suggested aberrant DNA methylation as a potential disease modifier [81,82].

The first study examining the role DNA methylation changes in a large European series of 127 children with JMML revealed CpG island hypermethylation of a candidate gene set in up to half of the cases [81]. Importantly, CpG hypermethylation at diagnosis was an independent risk factor for poor overall survival (OS) and risk of relapse after HSCT [81]. The conclusions were largely confirmed in a Japanese study investigating a similar candidate gene set [83]. Both studies observed that hypermethylation in JMML affected a narrow subset of gene promoters, as opposed to broad random distribution across all genetic regions examined, suggesting that high-risk JMML is characterized by a CpG island hypermethylation phenotype, as discovered previously in other specific cancer types [82–84]. Several follow-up studies corroborated this concept at the candidate gene level [85–87].

Extending these findings to a genome-wide scope using array-based methods, study groups in Europe [33], Japan [53] and United States [65] analyzed independent JMML cohorts with the aim to establish a methylation based risk-stratification. Comparing the methylome patterns of 167 children with JMML, the European Working Group of Myelodysplastic Syndromes in Childhood (EWOG-MDS) discriminated 3 distinct methylation groups, again highlighting epigenetic dysregulation as a strong prognostic risk factor [33]. Factors associated with hypermethylation were repressed chromatin, Ras pathway double mutants and upregulation of methyltransferases DNMT1 and DNMT3B.

This supported the emergence of DNA hypermethylation as a consequence of hyperactive Ras signaling [81,83]. Several associations between genetic driver mutation and DNA methylation pattern were noted. The group of patients with highest hypermethylation was dominated by somatic *PTPN11* mutation and older children, both known factors for inferior clinical course, whereas the low-methylation group was enriched for patients with NS/MPD, CBL cases, and young children with somatic *NRAS* mutations. The group with intermediate hypermethylation was characterized by somatic *KRAS* mutations and occurrence of monosomy 7 [33]. The Japanese cohort, consisting of 106 JMML cases, was split into two methylation groups [53]. In addition to known clinical risk factors, the highmethylation group involved cases with *NF1* or *PTPN11* mutations, secondary mutations, *LIN28B* overexpression and AML-like expression profile [53]. The North American study defined three similar methylation classes in 39 patients [65]. Interestingly, some JMML patients with good transplantation-free outcome and all patients with NS/MPD exhibited a DNA methylation signature closer to healthy controls than to other JMML cases [65,88], again underlining the significance of disrupted epigenetic control for the biology of JMML. The fact that all three methylome studies had used a comparable technical platform provided the unique opportunity for a comprehensive overarching meta-analysis. These collaborative efforts succeeded in developing and validating an international standard classifier of three different methylation categories matching those above and correlating with disease biology and outcome [88]. The prospective use of methylation analysis as a biomarker in JMML will aid in adapting treatment strategies, e.g., use of pretransplant therapy or low-intensity graft-versus-host disease (GVHD) prophylaxis, and support the generation of internationally comparable JMML study data.

#### **5. Current Recommendations for the Management of JMML**

With extensive molecular diagnostic work-up of JMML established in major international study groups and large centers around the world, it has become evident that there can no longer be a uniform one-size-fits-all approach for this disorder (Table 1). The authors recommend that therapeutic decisions in a newly diagnosed case of JMML be based on the following diagnostic information (Figure 1):


#### *5.1. Somatic NRAS Mutation*

The disease course in this group is remarkably heterogeneous. In older children with severe thrombocytopenia, increased HbF and high methylation class, a rapidly progressive course with a considerable risk of recurrence after HSCT is to be expected, likening the disorder to *PTPN11*- or *NF1*-driven JMML. On the other hand, a group of patients exist who are clinically well and have low HbF. Here, spontaneous clinical regression of the disease can occur in the long run without therapy. The search for an unrelated stem cell donor may be deferred in these cases. In between these two ends of the spectrum, the prospective identification of patients who benefit from watchful waiting is the real challenge. Factors suggesting surveillance without therapy include infant age, clinical status, age-appropriate levels of HbF, and low methylation class [41,89,90]. However, this must be balanced with the concern that delaying HSCT may compromise the outcome in some patients. EWOG-MDS data shows that the survival curve of JMML patients without HSCT keeps dropping during the first ten years and then plateaus at 25%.

**Table 1.** Management of JMML according to driver mutation. HSCT, hematopoietic stem cell transplantation; GVHD, graft-versus-host disease; DLI, donor lymphocyte infusions; NF-1, neurofibromatosis 1.


It is very rare for a germline *NRAS* or *KRAS* mutation to be detected in a suspected JMML case. Most of these children have additional syndromic features of the rasopathy spectrum [91,92] or correspond in phenotype to Noonan syndrome [93]. Anecdotal observations contradict the paradigm that the canonical tumor-associated Ras mutations in codons 12, 13 or 61 are not tolerated in the germline; such cases are sometimes based on mosaicism [94,95]. Because of their rarity, no general recommendation can be given for the treatment of these highly individual cases.

#### *5.2. Somatic KRAS Mutation*

Children diagnosed with *KRAS*-JMML are typically very young, often infants. Concurrent monosomy 7 in the neoplastic clone is often observed (see below). Autoimmune phenomena (hyperimmunoglobulinemia, autoantibodies) should be searched for, and a diagnostic differentiation from RALD [74,96,97] should be kept in mind. The clinical presentation of *KRAS*-driven JMML tends to be aggressive, requiring rapid intervention. In the past, long-term survival without HSCT has not been reported for this group, but the picture is now changing with the introduction of azacitidine. *KRAS*-mutated JMML responds particularly well to low-dose azacitidine with long-lasting clinical and molecular remissions [98,99]. Regimens use 100 mg/m2/day on five consecutive days or 75 mg/m2/day on seven consecutive days, repeated every 28 days; due to instability, immediate intravenous or subcutaneous application of the cold reconstituted solution must be observed.

Azacitidine has a favorable toxicity profile in children with JMML, mainly including lowergrade cytopenias, gastrointestinal discomfort, and infections [64,98]. It is variable how many cycles of azacitidine are necessary to achieve a response; between 6 and 9 cycles are usually administered. Frequently, the earliest sign of response to azacitidine is the improvement of thrombocytopenia. The spleen size diminishes after three to six cycles. Possibly, long-term cure with azacitidine alone will be achievable in *KRAS*-mutated JMML with a low risk profile.

**Figure 1.** Diagnostic approach for children with JMML. \* Helpful for clinical decision making, but not mandatory.

#### *5.3. Somatic PTPN11 Mutation*

This is the numerically largest group that also carries the highest risk of rapid progression and early death. As no curative chemotherapy regimen is known, expeditious planning of allogeneic HSCT is mandatory. Within this group, risk factors for an unfavorable course are: age at diagnosis ≥2 years, significantly increased level of HbF, presence of secondary mutations, and/or high methylation class. In these cases, the aim should be to perform HSCT within a period of 3 months after diagnosis (recommendations for implementation see below). Low-dose azacitidine is the preferable option for upfront therapy until HSCT in non-high risk cases, with the goal to achieve a more favorable biological status of the leukemia. In many affected children, this treatment leads to a clinical partial remission or at least sufficient disease control until HSCT [64,98–100] However, high-risk cases carry the potential of immediate progression under azacitidine and may therefore be better off with more instant cytoreduction using 6-mercaptopurine (50 mg/m2/day, to be adjusted according to clinical course) and/or cytarabine (40 mg/m2/day × 5 days) [7]. If a patient presents in a critically ill condition that requires rapid reduction of tumor burden, the use of intensive chemotherapy with cytarabine (2 g/m2/day × 5 days) and fludarabine (30 mg/m2/day × 5 days) may also be considered [7]. However, this involves a substantial risk of organ toxicity and life-threatening infections. Removal of the spleen, which is often

grossly enlarged, can be justified in individual cases to control respiratory impairment. A systematic beneficial effect of splenectomy on the further course of the disease has not been proven [40,101]. It is not yet clear whether treatment with azacitidine before HSCT also improves the long-term prognosis in this patient group. A recent non-randomized prospective study using a matched historical cohort as control group supports this assumption [98]. Until more precise evidence is available, the authors emphasize that treatment with azacitidine must not delay urgent HSCT in patients with *PTPN11* mutation.

#### *5.4. Germline Mutation in PTPN11*

Germline mutations in *PTPN11* cause Noonan syndrome, a condition of the rasopathy spectrum predisposing to a myeloproliferative disorder imitating JMML. For this reason, genetic diagnostics in suspected cases of JMML must always be carried out in both hematopoietic and non-hematopoietic material. Cultivated fibroblasts from a skin biopsy are ideal; hair follicles are less invasive, but more challenging to analyze. An oral mucous membrane swab has a high probability of contamination with hematopoietic cells, even if it is macroscopically not bloody, and should therefore only be scored if the result is negative.

In cases of suspected JMML in very young patients or with clinical evidence of Noonan syndrome, the diagnostic test for *PTPN11* should not just encompass the hotspot exons 3 and 13 because Noonan syndrome mutations may also be found in exons 4 or 8. The spectrum of somatic *PTPN11* mutations in nonsyndromic JMML and germline *PTPN11* mutations in Noonan-associated MPD overlaps to a great extent but not completely [31].

The distinction between non-syndromic JMML and Noonan-associated MPD is important, as the latter is usually self-limiting. However, these patients must be monitored carefully, as there can be relevant clinical compromise from cell infiltrates, making it necessary to begin cytoreductive therapy. In individual cases, a clonal development towards a bona fide neoplastic disease is possible [47,102,103].

#### *5.5. Neurofibromatosis Type 1*

If not already recognized in the patient, the syndrome can usually be diagnosed clinically and/or through family history at the time of onset of JMML. In younger children, only café-au-lait spots but not the other typical signs of neurofibromatosis may be present. Six or more café-au-lait spots as stipulated in the NIH criteria are noted in the majority of children, but in exceptional cases there may be none or fewer. Genetic analysis of *NF1* usually confirms the clinical diagnosis in children with JMML/NF-1 [104]. The typical finding is an *NF1*-inactivating heterozygous variation in the germline of the patient that arose de novo (~50% of cases) or was inherited. These lesions are often, but not always, deletions or truncating missense mutations reported as recurrent aberrations in the NF-1 literature. In addition, the neoplastic clone exhibits somatic loss of heterozygosity at the *NF1* locus or an independent second *NF1* mutation, leading to biallelic *NF1* inactivation [21,104,105]. The judgment is more difficult if no clinical signs of neurofibromatosis are present and the genetic findings correspond to the above paradigm only incompletely (for example, in case of monoallelic lesions, variants of unclear significance, no germline findings or low allelic frequency). In such cases, a myeloid disorder with a secondary *NF1* lesion, but driven by an unrelated event, may be present. With careful work-up, however, such dilemmas are rare.

JMML on the basis of NF-1 manifests more frequently at an older age than the other groups and typically does not involve a drastically reduced platelet count. Some children initially show little clinical impairment. However, long-term survival without HSCT has not yet been observed in this group, so that proceeding to transplant and inception of therapy with azacitidine is advisable as in patients with somatic *PTPN11* mutation. Consistent with this recommendation, JMML/NF-1 cases almost always have an intermediate or high methylation profile [88].

#### *5.6. CBL Mutation*

The typical configuration of *CBL* mutations in JMML is a heterozygous missense point mutation in *CBL* exons 8 or 9 in the germline, accompanied by uniparental isodisomy of the 11q chromosome arm as a somatic event in hematopoietic cells, leading to loss of heterozygosity [26,36,37]. Many, but not all, children show syndromic rasopathy features, such as facial dysmorphia and growth retardation. A particular phenomenon in this patient group is the frequent occurrence of autoimmunity and vasculitis. Some children with *CBL*-JMML have massive organ enlargement and may require splenectomy for symptom relief. Most patients do not require swift HSCT but can be managed with watchful waiting; many of these experience spontaneous resolution of the myeloproliferation. The homozygous *CBL*-mutant status in hematopoietic cells may persist until adulthood even in the absence of hematologic abnormalities [106]. Patients who undergo transplant often revert to stable mixed chimerism with sufficient disease control [6,26,36]. It is still unclear if the allograft also prevents the later development of symptoms related to autoimmune vasculitis. A recent report highlighted the role of somatic-only *CBL* inactivation in five patients with a clinical course that required HSCT [42], again illustrating the need for proper germline analysis in the diagnostic evaluation of JMML.

#### *5.7. None of the Above*

Suspected cases of JMML with negative panel sequencing for all five primordial genes and no clinical evidence of NF-1 are called "quintuple-negative" or "all-negative". In a third of these cases, in-depth multimodal genetic analysis uncovered a driving role of the *NF1* gene in the absence of clinical NF-1 features [104]. In other cases, RNA sequencing identified activating fusions involving *ALK*, *ROS1*, or *FLT3* [51,53]. Some authors argue in favor of diagnosing such myeloproliferative disorders as JMML due to the indistinguishable clinical and hematologic presentation. Excluding non-neoplastic causes of myelomonocytic proliferation, a maximum of 5–10% of suspected JMML cases remain genetically unexplained. It is advisable to refer these patients to an extended rasopathy work-up, especially if additional syndromic stigmata are present.

#### *5.8. Monosomy 7*

The significance of monosomy 7 for the biology of JMML is unclear, and there seems to be no association with clinical features or relevance for outcome [40]. Interestingly, monosomy 7 is observed more frequently in European compared to Japanese patients with JMML [33]. In a large international series, all patients with monosomy 7 and intermediate methylation class carried *KRAS* mutations, in contrast to an association between monosomy 7 and *PTPN11* or *NF1* in the high methylation group, and the absence of monosomy 7 in patients with low methylation pattern [88]. The mechanistic connection between this particular chromosomal lesion and aberrant DNA methylation patterns is not understood. Overall, it is likely that the presence of monosomy 7 plays a supportive role in JMML rather than being an independent pathogenetic factor [107]. This concept is also supported by the observation of secondary monosomy 7 in a watch-and-wait patient with Noonan syndrome and neonatal myeloproliferative disorder [102].

#### *5.9. Allogeneic HSCT*

Busulfan-based myeloablative conditioning regimens are commonly chosen and achieve 55–73% OS with a moderate 10–15% rate of transplant-related mortality but significant probability of leukemia relapse in the order of 25–35% [7,40,101,108–112]. The EWOG-MDS currently recommends a three-alkylator regimen consisting of busulfan (0.8–1.2 mg/kg/dose given 4 doses per day, day 7 to day 4), cyclophosphamide (60 mg/kg/d, day 3 to day 2), and melphalan (125–140 mg/m2/d on day 1) [40]. In an attempt to reduce toxicity, a recent prospective randomized trial compared busulfan, cyclophosphamide, and melphalan with busulfan and fludarabine alone but terminated early due to excessive disease recurrence in the latter arm [113]. Matched sibling donors

(MSD) or matched/1-antigen-disparate unrelated donors (MUD) are considered the most suitable stem cell sources [40]. Matched cord blood units are a viable alternative, especially for smaller patients [114–117]. Although haploidentical relatives are readily available for urgent transplant and highly motivated, this should still be viewed as an approach with limited experience [101,112,118]. A recent study from China in 47 JMML patients suggested a lower relapse incidence in mismatched/haploidentical donor transplants compared to matched donors with similar rates of acute/chronic graft-versus-host disease and non-relapse mortality [101].

The North American group noted better post-HSCT outcome of patients with JMML who experienced molecular response to pretransplant chemotherapy [119], similar to other pediatric leukemias [120,121]. However, a limitation was that only a minority of patients responded to chemotherapy, conceivably those with favorable disease biology. Biomarkers predicting response to chemotherapy are lacking. Therefore, it cannot be generalized that pretransplant chemotherapy benefits survival in JMML, and the risk of unwarranted organ damage remains a concern [122].

Beside the leukemia biology factors discussed above, the way the transplant procedure is handled significantly influences the risk of relapse. It is likely that it is not so much the conditioning regimen but rather the establishment of a graft-versus-leukemia effect that is decisive for the success of allogeneic HSCT in JMML [101,109–111]. For this reason, EWOG-MDS recommends keeping immunosuppressive therapy with cyclosporine A at low levels (trough levels around 80 μg/L) and tapering early (from day +40 in the absence of grade II-IV GVHD). It is advisable to determine the recipient-donor chimerism at very close intervals (up to weekly in high-risk patients), as the reappearance of even small autologous cell populations mandates immediate withdrawal of the immunosuppressive therapy [123–125].

Age at diagnosis ≥2 years, *NF1* or somatic *PTPN11* mutation, and high DNA methylation define a patient group whose risk of JMML recurrence after HSCT is even higher than 50%, bringing up the question of post-transplant prophylaxis. On the basis of favorable data for other myeloid neoplasms [126–131] and in the absence of better alternatives, the authors consider it appropriate to recommend azacitidine (started as soon as safe and tolerable after engraftment; 32 mg/m2/day for five consecutive days, every 28 days) plus donor lymphocyte infusions (started after 3 cycles of azacitidine and 4 weeks after discontinuation of immunosuppressive prophylaxis, CD3<sup>+</sup> cell dose 1–5 × <sup>10</sup>6/kg, repeated every 8 weeks with increasing cell dose up to 1–5 × <sup>10</sup>7/kg). However, we emphasize that there are no systematic data for this approach in JMML.

#### *5.10. Experimental Agents and Targeted Therapy*

Despite the prominent role of the Ras/MAPK network, attempts to target this complex signal cascade have shown limited therapeutic benefit in JMML [10,132]. The Children's Oncology Group is currently recruiting patients for a phase II trial to examine the safety and efficacy of oral trametinib, a MEK1/2 inhibitor, in refractory or relapsed JMML (NCT03190915). In vitro data from induced pluripotent stem cell lines suggests mutation-specific sensitivity to kinase inhibition, with a preferential sensitivity of *PTPN11* driven JMML to trametinib [133]. BCL2 inhibition gave impressive results when combined with azacitidine in elderly AML patients [134,135] and early results argue for a benefit in pediatric patients with advanced MDS/AML [136,137]. With an upregulation of the macrophage immune checkpoint CD47 in myeloid malignancies, ongoing preclinical and clinical trials test CD47-directed agents in MDS/AML, with encouraging efficacy results in combination with azacitidine [138–140].

**Author Contributions:** Conceptualization, C.M., C.M.N. and C.F.; writing—original draft preparation, C.M.; writing—review and editing, C.M.N. and C.F.; funding acquisition, C.M.N. and C.F. All authors have read and agreed to the published version of the manuscript.

**Funding:** This manuscript was supported by Bundesministerium für Bildung und Forschung MyPred 01GM1911A (to C.M.N. and C.F.) and Deutsche Forschungsgemeinschaft CRC992-C05 (to C.F.).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Acknowledgments:** The authors wish to thank all physicians and scientists collaborating in the European Working Group of Myelodysplastic Syndromes in Childhood for their valuable contributions.

**Conflicts of Interest:** C.M.N. has a consultancy with Bristol-Myers Squibb. The other authors declare no conflict of interest. The funders had no role in the writing of the manuscript.

#### **References**

