Pure erythroid leukemia (PEL) is an exceptionally uncommon acute leukemia and remains a difficult diagnosis in pediatric hematopathology. In conventional acute myeloid leukemia (AML), the abnormal population is usually recognizable as myeloblasts; in PEL, the malignant compartment consists predominantly of immature erythroid precursors with abrupt maturation arrest. This lineage pattern creates several diagnostic pitfalls. Peripheral blood may contain few obvious blasts, aspirates may be dilute or necrotic, and blastoid erythroid cells can resemble lymphoblasts, minimally differentiated AML, megakaryoblasts, plasma cell neoplasms, or non-hematopoietic small round-cell tumors. Morphology therefore serves as an initial diagnostic entry point rather than the definitive basis for diagnosis. Current classification systems provide the necessary diagnostic framework [1,2], while earlier WHO classifications and hematopathology reviews remain valuable for understanding the evolution of diagnostic thresholds and lineage assignment [3-5]. A defensible diagnosis requires concordant marrow architecture, cytology, flow cytometry, immunohistochemistry, cytogenetics, and molecular testing [6-8]. In children, ribonucleic acid (RNA)-based fusion testing deserves early consideration when the morphology and phenotype suggest erythroid-lineage leukemia [9].

The classification of erythroid-predominant leukemias has narrowed considerably over time. The older French-American-British AML-M6 category combined biologically different disorders, including cases that would now be classified as myelodysplastic neoplasms, AML with myelodysplasia-related genetics, therapy-related AML, or other genomically defined AML subtypes. In contemporary practice, true PEL is not merely characterized by erythroid hyperplasia with dysplasia; rather, it represents a leukemic proliferation dominated by immature erythroid precursors. The 2022 World Health Organization (WHO) classification retains acute erythroid leukemia as a distinct morphological entity, whereas the 2022 International Consensus Classification (ICC) classifies many TP53-mutated adult cases under acute myeloid leukemia (AML) with mutated TP53 [1,2]. This difference matters clinically because terminology affects risk communication, trial eligibility, and how older AML-M6 studies are interpreted.

The pediatric literature requires separate interpretation. Adult PEL is strongly linked to advanced age, antecedent or therapy-related myeloid disease, complex or monosomal karyotype, chromosome 5 and 7 abnormalities, and poor survival [10-12]. Biallelic TP53 inactivation and adverse cytogenetics are central to this adult model [13,14]. Pediatric reports, however, point to broader heterogeneity, including NUP98 rearrangements [9], NFIA::CBFA2T3 and NFIA::RUNX1T1 fusions [15,16], other erythroid leukemia-associated rearrangements [17-19], CNS or soft-tissue erythroblastic sarcoma presentations [20,21], marrow-based adolescent acute erythroid leukemia [22], NFIA::CBFA2T3-associated erythroblastic sarcoma or PEL [23], CIC-rearranged disease [24], multi-institutional erythroblastic sarcoma series [25], and more recent NFIA::CBFA2T3 reports with disseminated or cerebrospinal-fluid-based disease [26,27]. Experimental work on NFIA-ETO2 supports a direct maturation-arrest mechanism, particularly in cooperation with mutant TP53 [28]. Limited existing evidence suggests that some children present with central nervous system or soft-tissue disease before marrow disease is obvious. These observations argue against considering pediatric PEL merely an age-shifted counterpart of adult TP53-mutated AML. They also highlight substantial gaps in the evidence regarding treatment, prognosis, MRD standardization, and genotype-specific risk stratification. This review therefore integrates diagnostic reasoning with treatment principles, prognostic stratification, MRD monitoring, and practical reporting considerations.

This narrative review used a structured search and evidence-weighting process informed by the SANRA framework for narrative reviews [29]. PubMed, including MEDLINE-indexed records, was searched from 1 January 2000 to 13 June 2026, with earlier landmark publications retained when necessary to provide historical context for disease classification and erythroid biology. The review was restricted to English-language full-text publications. Four Boolean search blocks were used: (1) (“pure erythroid leukemia” OR “acute erythroid leukemia” OR erythroleukemia OR “erythroblastic sarcoma”) AND (classification OR morphology OR immunophenotype OR diagnosis OR WHO OR ICC); (2) the same disease block AND (child* OR pediatric* OR infant* OR adolescent*) AND (NUP98 OR NFIA OR CBFA2T3 OR RUNX1T1 OR TP53 OR EPOR OR JAK2 OR CIC OR GATA1); (3) the same disease block AND (treatment OR chemotherapy OR transplantation OR HSCT OR outcome OR survival OR prognosis OR relapse); and (4) the same disease block AND (“measurable residual disease” OR “minimal residual disease” OR MRD OR flow cytometry OR “molecular monitoring” OR cerebrospinal fluid OR CSF). The term “minimal residual disease” was included only as a legacy search synonym; throughout the manuscript, MRD refers to measurable residual disease. Backward and forward citation screening of key articles, together with targeted searches of journal websites, was performed to identify additional classification documents, mechanistic studies, and recent molecularly characterized pediatric reports.

Y.Y. performed title and abstract screening, followed by full-text assessment. R.P. verified the final included set and reviewed uncertain eligibility decisions; the authors resolved disagreements through discussion. Because the initial database hit counts were not prospectively archived, the numerical flow was reconstructed during revision from the final deduplicated source library. The reconstructed library contained 61 unique English-language full-text publications. All 61 publications were reassessed against the final eligibility criteria and retained in the narrative synthesis; no additional publications were excluded during this reconstructed audit. Publications were eligible for inclusion if they addressed official classification, diagnostic criteria, erythroid morphology or immunophenotype, pediatric or fusion-defined PEL, erythroblastic sarcoma, molecular pathogenesis, Down syndrome-associated mimics, treatment, transplantation, prognosis, or MRD. For interpretation, evidence was grouped into the same five priority categories shown in Figure 1: Highest, WHO or ICC classification documents and consensus guidance; High, multicenter or cooperative-group cohorts; Moderate, comparative clinicopathologic or molecular studies; Limited, small series and molecularly characterized case reports, including representative single-case reports; and Exploratory, preclinical models, drug-sensitivity studies, or hypothesis-generating targeted-therapy data. Single-case reports are therefore included within the Limited category, whereas the Exploratory category is reserved for preclinical or hypothesis-generating evidence and is not combined with case-report evidence. Adult studies were included only when pediatric evidence was unavailable or when they clarified classification, TP53 biology, treatment resistance, MRD methodology, or investigational targets. Abstract-only reports, duplicate descriptions without additional information, non-English articles, studies limited to reactive erythroid hyperplasia, and reports without sufficient lineage or molecular detail were excluded during the original literature selection and were not represented in the reconstructed final library.

Evidence statements were calibrated according to the strength of the supporting evidence. Diagnostic statements derived from the WHO or ICC are identified as classification-based conclusions. Findings from multicenter or cooperative-group cohorts are presented as the strongest available clinical evidence, whereas comparative clinicopathologic and molecular studies are considered moderate-level evidence. Small series and molecularly characterized case reports, including representative single cases, are treated as limited evidence and are not used as validated prognostic rules. Preclinical models, drug-sensitivity studies, and hypothesis-generating targeted-therapy data are labeled exploratory and do not constitute treatment recommendations. Because this is a narrative review rather than a systematic review or meta-analysis, no pooled effect estimates or formal risk-of-bias score was calculated. Figure 1 summarizes the reconstructed source-library accounting and this five-category evidence hierarchy. The figure is not presented as a PRISMA-compliant systematic-review flow.

2.1. Definition and Classification

PEL is best characterized as an acute leukemia dominated by immature erythroid precursors, typically exhibiting severe maturation arrest at the proerythroblast or pronormoblast stage. The 2022 WHO framework preserves pure erythroid leukemia within acute myeloid leukemia defined by differentiation [1]. Historical WHO revisions and hematopathology reviews describe commonly used thresholds of at least 80% erythroid precursors and at least 30% proerythroblasts, although wording has varied across classifications and reports [3-5]. These thresholds provide a basis for classification; however, sampling-related limitations can compromise the accuracy of cell counting. Aspirate dilution, a dry tap, marrow fibrosis, necrosis, marked dyserythropoiesis, or extramedullary presentation may render the differential cell count unreliable. In such cases, biopsy architecture, erythroid immunohistochemistry, and genomic evidence of clonality carry greater diagnostic weight [6-8].

The central diagnostic challenge is not simply determining whether erythroid cells are increased in number, but establishing whether they are neoplastic and developmentally arrested. Erythroid hyperplasia may be reactive, nutritional, regenerative, drug-induced, or secondary to hemolysis. It can also accompany myelodysplastic neoplasms and AML with myelodysplasia-related features. PEL differs because the abnormal population is not a normally maturing erythroid series; it is a monomorphic or near-monomorphic population of immature erythroid blasts that displaces normal hematopoiesis. A useful diagnostic report should therefore document marrow cellularity, erythroid percentage, proerythroblast proportion, presence or absence of a separate myeloblast population, dysplasia in non-erythroid lineages, fibrosis, p53 staining pattern when performed, flow cytometry findings, and the results of karyotype, fluorescence in situ hybridization (FISH), next-generation sequencing (NGS), and fusion testing [6-8].

In pediatric patients, reporting is best framed by integrating morphological and genomic findings rather than forcing every case into an adult-derived classification category. WHO terminology preserves the erythroid phenotype, while ICC terminology highlights TP53-mutated AML biology when TP53 criteria are fulfilled. A child with a NUP98 or NFIA-related fusion, however, should not be described as though the adult TP53-mutated paradigm fully explains the disease. Practical diagnostic terminology may include acute erythroid leukemia/pure erythroid leukemia, with the specific genetic lesion identified when erythroid-lineage leukemia is supported; AML with mutated TP53 and a pure erythroid phenotype when the diagnostic criteria for TP53-mutated AML are met; or erythroblastic sarcoma, with the specific fusion and bone marrow involvement status reported when the initial presentation is a mass lesion. Such integrated wording communicates lineage, genetics, and risk without overstating certainty [6-8]. Table 1 summarizes key classification issues relevant to pediatric PEL.

2.2. Pediatric Epidemiology and Clinical Presentation

PEL is rare across all age groups and is particularly uncommon in children. The true incidence in the pediatric population remains uncertain because earlier reports used to overlap diagnostic terminology, including AML-M6, erythroleukemia, pure erythroid leukemia, erythroblastic sarcoma, acute undifferentiated leukemia, and AML not otherwise specified. Reclassification under current WHO and ICC systems further limits comparison across cohorts. Adult AML studies indicate that PEL represents only a small fraction of cases [10]. Pediatric evidence is much thinner and comes mainly from cooperative group data [9], NFIA-associated case reports [15,16], recent erythroblastic sarcoma reports [20,21], and cytogenetic or mechanistic descriptions of rarer erythroid fusion events [17-19, 28]. This limited evidence base has practical consequences: clinicians may not consider PEL early, and no pediatric PEL-specific therapeutic standard exists.

Clinical presentation usually reflects bone marrow failure rather than a distinctive leukemia syndrome. Children may present with pallor, fatigue, fever, recurrent infection, bruising, mucosal bleeding, petechiae, bone pain, or a generally poor clinical condition. Hepatomegaly and splenomegaly can occur but are non-specific. Laboratory findings typically include anemia and thrombocytopenia, with variable leukocyte counts. Circulating blasts may be absent, low, or morphologically ambiguous. Lactate dehydrogenase and uric acid may rise with high tumor burden. Because erythroid blasts may not be readily recognized as leukemic blasts on automated or manual peripheral blood examination, persistent or unexplained cytopenias accompanied by abnormal nucleated erythroid forms should prompt bone marrow evaluation. Extramedullary disease deserves particular attention in children. Erythroblastic sarcoma may involve the central nervous system, skull, spine, soft tissue, lymph nodes, or other sites and can be mistaken for lymphoma, myeloid sarcoma, Ewing sarcoma, rhabdomyosarcoma, poorly differentiated carcinoma, or inflammatory lesions. Immunohistochemistry is often decisive because the tumor may be CD45-negative or weak and may lack conventional myeloid markers. A panel including CD71, E-cadherin, CD36, glycophorin A, hemoglobin, GLUT1, CD117, CD34, myeloperoxidase, CD41, CD61, cytokeratin, desmin, and lymphoid markers helps prevent misclassification. CNS or infant erythroblastic sarcoma presentations have been associated with NFIA-related fusions [15,16], and recent pediatric reports continue to broaden this spectrum [20,21]. Sufficient tissue should be reserved for cytogenetic and RNA-based testing because rare or cryptic fusions may provide the decisive diagnostic clue [17-19, 28].

Record age and constitutional context carefully. In neonates and infants, transient abnormal myelopoiesis (TAM) and myeloid leukemia of Down syndrome (ML-DS) must remain in the differential diagnosis, including in phenotypically subtle or mosaic trisomy 21. These disorders may exhibit megakaryocytic and erythroid features, as well as GATA1 mutations, and may be mistaken for erythroid leukemia when the lineage work-up is incomplete [30,31]. Their progression patterns and clinical spectrum differ from classic PEL [32-34]. Broader Down syndrome leukemia risk and the molecular evolution from TAM to ML-DS have been characterized separately [35,36], and additional genomic studies have clarified cooperating lesions during progression [37-39]. In suspected pediatric PEL, clinical history, physical examination, karyotype, assessment for constitutional or mosaic trisomy 21 when indicated, and GATA1 mutation testing are therefore not peripheral details; they are part of the diagnostic framework. Table 2 summarizes the pediatric reports most relevant to acute erythroid leukemia/pure erythroid leukemia and erythroblastic sarcoma. These cases should not be interpreted as constituting a mature epidemiologic dataset. Rather, they are better viewed as evidence that childhood erythroid-lineage leukemia and sarcoma may represent a heterogeneous, fusion-enriched spectrum that overlaps with, but is not identical to, adult TP53-mutated PEL [9, 15-16]. Early CNS and soft-tissue report highlight NFIA-associated presentations [20,21]. Later reports broaden the marrow-based, CIC-rearranged, and multi-institutional spectrum [22-25], while recent NFIA: CBFA2T3 reports strengthen the molecular subgroup and support cerebrospinal-fluid-based molecular diagnosis [26,27]. Table 2 summarizes the pediatric reports most relevant to AEL/PEL and erythroblastic sarcoma.

2.3. Morphology and Immunophenotyping

Morphologic assessment begins by determining whether the marrow or tissue is dominated by immature erythroid cells. Proerythroblasts are usually large, with round to irregular nuclei, fine chromatin, prominent nucleoli, deeply basophilic cytoplasm, and frequent cytoplasmic vacuolization. Multinuclearity, nuclear budding, megaloblastoid change, cytoplasmic blebbing, and dyserythropoiesis may be present. The bone marrow is often hypercellular and may show extensive replacement by immature erythroid elements, accompanied by suppression of granulopoiesis and megakaryopoiesis. Auer rods are typically absent, and myeloperoxidase staining is negative in the neoplastic erythroid population. These features are suggestive, but none is sufficiently specific to establish lineage in isolation. Representative bone marrow morphology is shown in Figure 2. Use flow cytometry and immunohistochemistry together. Erythroid precursors are fragile, may be underrepresented following red-cell lysis, and often exhibitweak or absent CD45. CD71 is highly sensitive and typically shows strong expression; however, it is not entirely specific, as transferrin receptor expression may also reflect cellular proliferation [40,41]. E-cadherin is useful for highlighting immature erythroid precursors in marrow biopsy and tissue sections. CD36 supports erythroid differentiation but may also be expressed by monocytes and megakaryocytic cells. Glycophorin A and hemoglobin are more specific markers of erythroid maturation; however, their expression may be weak or absent in highly immature proerythroblastic disease [40]. GLUT1 is a useful adjunctive erythroid marker in tissue sections [42]. Given the overlap between erythroid and megakaryoblastic leukemias, interpretation should remain panel-based rather than dependent on a single marker [43].

CD34 and CD117 require cautious interpretation in suspected pure erythroid leukemia. CD117 may be expressed in a subset of PEL cases, whereas CD34 may be negative, dim, or limited to only a minor population; therefore, absence of CD34 does not exclude acute leukemia, and CD117 positivity alone should not be interpreted as definitive evidence of myeloid lineage [5, 8, 10, 43]. Myeloperoxidase is usually negative in neoplastic erythroid blasts, while megakaryocytic markers, including CD41, CD61, and CD42b, help exclude acute megakaryoblastic leukemia, although aberrant or partial expression must be interpreted within the full morphologic and immunophenotypic context [8,43]. Lymphoid markers should also be included to exclude B- or T-lymphoblastic leukemia, particularly when the abnormal cells exhibit a blastoid morphology and dim CD45 expression. In tissue-based or extramedullary presentations, epithelial, neural, muscle, germ-cell, and other lineage-specific tumor markers may be required to exclude nonhematopoietic small round cell tumors [43]. The diagnosis is strongest when morphology, immunophenotype, and genetic findings are concordant. A marrow replaced by proerythroblasts with strong CD71 and E-cadherin expression, supportive CD36 or GLUT1 staining, negative myeloperoxidase, absence of lineage-defining lymphoid markers, and a clonal cytogenetic or fusion abnormality strongly supports PEL [10, 40-43]. By contrast, isolated CD71 positivity is insufficient because CD71 also reflects erythroid proliferation and may be seen in non-neoplastic erythroid expansion [40,41]. Likewise, negativity for glycophorin A does not exclude PEL when the cells are highly immature and other erythroid markers provide convincing evidence of erythroid differentiation [40]. In children, p53 immunohistochemistry should be regarded as supportive rather than obligatory. Abnormal overexpression or a null pattern supports TP53-altered disease, whereas a wild-type pattern should prompt careful evaluation for alternative pediatric drivers, including fusion-defined disease, rather than dismissal of the erythroid diagnosis [13]. Table 3 summarizes the main immunophenotypic markers used in suspected pediatric PEL.

2.4. Cytogenetic and Molecular Pathophysiology in Pediatric Pure Erythroid Leukemia

Genetics and pathophysiology are difficult to disentangle in PEL because its morphology reflects disrupted erythroid development. Pediatric PEL can be conceptualized as a neoplastic state in which a clone retains erythroid identity, acquires self-renewal capacity, and fails to undergo terminal maturation. Normal erythroid differentiation depends on coordinated transcriptional and chromatin-regulatory networks that progressively establish erythroid identity and maturation [44-46]. Distinct lesions may converge on this phenotype through different roles. NUP98 and NFIA fusions act as transcriptional drivers that impose self-renewal or maturation arrest [9, 28, 47], while TP53 loss can confer genomic tolerance and a survival advantage [13,14]. EPOR/JAK2 activation often appears to function as a proliferative cooperating pathway, particularly in TP53-altered disease [48]. GATA1s drives trisomy 21-associated myeloid disease, but its main relevance to PEL is as a developmental comparator and diagnostic mimic. Figure 3 summarizes these relationships.

NUP98 and NFIA fusions are considered primary transcriptional or lineage-driving lesions; TP53 alterations may confer genomic instability and adverse-risk biology; EPOR/JAK2/STAT5 activation may function as a cooperating proliferative pathway; and GATA1s in the setting of trisomy 21 indicates transient abnormal myelopoiesis (TAM) or myeloid leukemia associated with Down syndrome (ML-DS), rather than classic P.

TP53-altered PEL remains the best-characterized adult model and should therefore also be evaluated in pediatric populations. Under normal conditions, TP53 regulates the DNA damage response, cell-cycle arrest, cellular senescence, and apoptosis. Adult series link the pure erythroid phenotype to adverse cytogenetics, treatment resistance, and poor outcome [10-12]. Biallelic TP53 loss allows erythroid progenitors with chromosomal damage to evade elimination and survive, creating conditions that promote complex karyotypes, chromosome 5q or 7q losses, and therapy resistance [13,14]. Abnormal p53 immunohistochemistry can help distinguish TP53-altered PEL from reactive erythroid hyperplasia [49]. Mechanistically, TP53 loss may permit damaged erythroid-committed progenitors to proliferate despite failed maturation. Experimental AML data indicate that the antileukemic effect of venetoclax combined with cytarabine and the anthracycline idarubicin is TP53-dependent. These findings support a potential role for TP53 status in determining chemotherapy response but do not establish a PEL-specific or pediatric PEL-specific resistance mechanism [50].

NUP98 fusions offer a pediatric-relevant mechanism of fusion-driven erythroid leukemogenesis. Rearranged NUP98 proteins can alter transcriptional regulation, activate stem-cell and HOX/MEIS-associated programs, maintain chromatin accessibility at self-renewal loci, and impair maturation. In erythroid progenitor, these effects may preserve erythroid identity while preventing normal terminal differentiation. The practical implication is important: NUP98-rearranged pediatric AEL/PEL may be missed if testing relies only on morphology, conventional karyotype, or DNA-only panels. Early RNA-based fusion testing should therefore be strongly considered in suspicious pediatric cases [9, 47, 51].

NFIA-related fusions provide one of the clearest mechanisms underlying an erythroid maturation-arrest phenotype. NFIA participates in the transcriptional regulation of erythroid development, whereas CBFA2T3/ETO2 and RUNX1T1/ETO function as transcriptional corepressor partners involved in hematopoietic differentiation. Clinical reports have described: NFIA: CBFA2T3 and NFIA::RUNX1T1 in infant or CNS erythroblastic sarcoma/PEL-like disease [15,16]. Cytogenetic and sequencing studies have also linked CBFA2T3 rearrangements and other rare fusions to acute erythroid leukemia [17-19]. Recent pediatric reports further expand the spectrum of NFIA-associated erythroid tumors [20,21]. Experimental data show that NFIA-ETO2 blocks erythroid maturation and, with mutant TP53, induces PEL in model systems. These findings support a two-component model in which the fusion establishes an abnormal erythroid transcriptional state and maturation block, while cooperating lesions enhance survival or clonal fitness. Clinically, this helps explain why NFIA-rearranged tumors can present as erythroblastic sarcoma or PEL-like disease with strong erythroid markers, few conventional myeloblasts, and occasional CNS involvement [28].

GATA1 alterations in the context of Down syndrome should be discussed as a distinct developmental pathway rather than as a classic PEL subtype. In constitutional trisomy 21, acquired exon 2 GATA1 mutations generate the short GATA1s isoform and initiate transient abnormal myelopoiesis, while additional cohesin, epigenetic, or signaling lesions can support progression to myeloid leukemia of Down syndrome [30-39]. These disorders typically exhibit megakaryoblastic or mixed megakaryocytic–erythroid differentiation rather than a monomorphic proliferation of proerythroblasts. Their importance in a PEL review is therefore diagnostic: neonates or infants with circulating blasts, nucleated erythroid cells, dyserythropoiesis, CD36 or CD117 expression, or marrow erythroid expansion require assessment for constitutional or mosaic trisomy 21 and GATA1 mutation [30-31, 52].

A GATA1s-positive trisomy 21-associated myeloid proliferation should be classified and treated within the TAM or ML-DS framework rather than relabeled as PEL. This distinction differentiates a biologically and therapeutically defined pediatric entity from fusion-driven or TP53-associated erythroid leukemia.

EPOR/JAK2/STAT5 activation and epigenetic dysregulation may function as cooperating mechanisms. Amplification of EPOR and JAK2 can enhance erythropoietin receptor pathway signaling, thereby conferring survival and proliferative advantages on erythroid progenitors [48]. If the same clone harbors a transcriptional lesion that blocks maturation, enhanced signaling may promote expansion of an already arrested erythroid compartment. Similarly, abnormalities in chromatin regulators may stabilize a leukemic erythroid transcriptional program and reduce responsiveness to normal differentiation cues. Integrated genomic studies have identified molecular subgroups of acute erythroid leukemia [47,51], while transcriptomic and model-based work has shown how transcriptional and epigenetic abnormalities can drive disease rather than serve as bystanders [53,54].

Clinical correlations remain provisional because pediatric datasets are small and depend heavily on one cooperative-group reclassification study, small series, and individual reports [9, 15-16, 20-27]. Limited existing evidence suggests that marrow replacement by immature erythroid cells produces anemia, thrombocytopenia, neutropenia, infection risk, bleeding, and transfusion dependence. In contrast, some fusion-defined pediatric tumors are predominantly extramedullary or CNS-based and have little circulating or marrow disease at presentation. Diagnostic reporting should therefore document karyotype complexity, TP53 mutation and allelic status, copy-number abnormalities, NUP98 or NFIA-related fusions, EPOR/JAK2 pathway alterations, GATA1 mutation when trisomy 21-associated disease is possible, and the method used to detect cryptic rearrangements [1-2, 9-10, 13, 30-31, 48, 52]. Table 4 summarizes the genetic lesions and mechanisms that may produce a pediatric pure erythroid leukemia phenotype.

2.5. Genotype-Phenotype Heterogeneity and Prognostic Stratification

Available pediatric evidence supports several recurring genotype–phenotype patterns; however, these patterns should not be interpreted as validated therapeutic subgroups. TP53-altered, complex-karyotype PEL represents the dominant adult pattern but appears to be uncommon in children. NUP98-rearranged disease is enriched in pediatric AEL and PEL, while NFIA-related fusions are repeatedly associated with infants or young children, CNS or soft-tissue erythroblastic sarcoma, and occasionally minimal marrow involvement [9, 15-16, 20-28]. CIC-rearranged and other fusion-defined cases further expand this spectrum, although the available evidence remains limited to isolated cases and small case series [19,24].

Prognostic interpretation should integrate genotype, morphology, disease distribution, and treatment response. The five PEL cases in the Children’s Oncology Group series had 5-year overall and event-free survival of approximately 20%, but the sample was too small for reliable partner-specific analysis [9]. NUP98:: KDM5A was associated with poor outcome in the available pediatric AEL cohort, while NFIA-fusion reports show heterogeneous outcomes ranging from early death to durable remission after AML-directed treatment and, in selected cases, transplantation [9, 15, 20, 26-27]. These data support explicit risk communication but do not justify a definitive pediatric PEL risk score. Table 5 summarizes these genotype-phenotype patterns, the reported treatment responses, and their provisional clinical implications.

2.6. Differential Diagnosis

The differential diagnosis should begin with non-neoplastic erythroid expansion. Severe vitamin B12 or folate deficiency, copper deficiency, hemolysis, recent bleeding, erythropoietin exposure, marrow recovery after chemotherapy, congenital dyserythropoietic anemia, and viral or inflammatory marrow stress may produce erythroid hyperplasia and dyserythropoiesis. These conditions usually retain a spectrum of erythroid maturation, lack leukemia-defining clonal genetics, and improve when the underlying cause is corrected. Neoplastic mimics include myelodysplastic neoplasm with erythroid predominance, AML with myelodysplasia-related genetics, acute megakaryoblastic leukemia, minimally differentiated AML, acute undifferentiated leukemia, lymphoblastic leukemia or lymphoma, plasma-cell neoplasms, and metastatic pediatric small round-cell tumors. Acute megakaryoblastic leukemia is especially difficult because erythroid and megakaryocytic programs share developmental ancestry, and Down syndrome-associated disease occupies this lineage boundary. A broad panel and genomic testing reduce diagnostic error. PEL should be diagnosed only when an immature erythroid differentiation is convincingly demonstrated and relevant alternative diagnoses have been systematically excluded [5, 8, 43]. Table 6 summarizes the principal differential diagnoses.

2.7. Practical Diagnostic Approach

The practical work-up should proceed as an integrated sequence rather than as a descriptive summary. First, establish the clinical trigger and obtain adequate marrow, biopsy, peripheral blood, tissue, and CSF specimens when indicated. Second, determine whether the dominant abnormal population shows immature erythroid morphology with maturation arrest. Third, prove erythroid lineage with a panel that includes sensitive and maturation-associated erythroid markers while actively excluding myeloid, lymphoid, monocytic, megakaryocytic, and non-hematopoietic mimics. Table 3 and Table 6 summarize the morphology and immunophenotype criteria.

Genetic testing should proceed in parallel with phenotyping characterization. Conventional karyotype and copy-number assessment identify complex, monosomal, 5q, 7q, 17p, EPOR, or JAK2 abnormalities. DNA-based NGS should include TP53 and relevant myeloid genes, but a negative DNA panel does not exclude fusion-driven PEL. Early RNA-based fusion testing is strongly recommended in suspicious pediatric cases because NUP98, NFIA, CIC, and other rearrangements may be cryptic [9, 15-19, 24, 28]. GATA1 mutation and constitutional or mosaic trisomy 21 assessment should be ordered early when neonatal presentation, Down syndrome, or megakaryocytic-erythroid ambiguity is present [30-31, 52]. Table 4 and Table 5 summarize the distinction between likely drivers and cooperating lesions, along with provisional genotype-phenotype patterns.

The final report should state the WHO and ICC classification basis; the quantitative and architectural evidence for erythroid predominance; the immunophenotypic evidence for erythroid lineage; the genetic lesion and TP53 allelic status; the marrow and CNS or extramedullary distribution; the major excluded mimics; and the baseline MRD target. Suggested wording is: “acute erythroid leukemia or pure erythroid leukemia, with [genetic abnormality], [TP53 allelic status], and [marrow, CNS, or extramedullary status].” The report should also indicate the strength of evidence supporting unusual fusion-defined interpretations and link the diagnosis to the treatment and MRD frameworks summarized in Table 7 and Table 8. Figure 4 integrates the diagnostic, genotypic, treatment, and MRD steps into one workflow.

2.8. Treatment and Prognosis

No prospective pediatric PEL trial and no accepted PEL-specific regimen have been reported. Children are generally treated within a contemporary pediatric AML protocol or clinical trial, using protocol-defined cytarabine- and anthracycline-based induction followed by risk-adapted consolidation [9, 15, 20-27, 56]. This approach reflects the lack of an alternative evidence-based standard, not proof that all molecular subtypes respond equally. Reported CNS or other extramedullary cases have generally received systemic AML therapy plus protocol-directed intrathecal treatment. Local radiotherapy or surgery has been used selectively for diagnostic tissue acquisition, neurologic compromise, or local control.

The role of allogeneic hematopoietic stem-cell transplantation (HSCT) after first complete remission remains unresolved for PEL. Current pediatric AML practice uses adverse genetics, induction response, MRD, primary refractory disease, and relapse status to determine transplant eligibility [56,57]. The Children’s Oncology Group AEL series was too small to establish a PEL-specific benefit from transplantation, and the reported transplanted patients belonged to the erythroid or myeloid subgroup rather than the five-case PEL subset [9]. Limited case evidence describes durable remission after HSCT in selected NFIA::CBFA2T3-associated disease, but selection and publication bias prevent routine recommendation for every NFIA-fusion case [15,26].

Genotype-directed therapy remains investigational. NUP98-rearranged AML is associated with poor induction response and high MRD, and co-occurring FLT3-ITD should be treated according to pediatric AML protocol or clinical-trial criteria [55,56]. Menin inhibition suppresses NUP98-fusion leukemia in preclinical and patient-derived models, but PEL-specific clinical efficacy remains unestablished [58,59]. A small retrospective pediatric NUP98-rearranged AML study reported responses with combinations containing BCL-2, FLT3, or CDK6 inhibitors, but only 10 NUP98-rearranged patients were included, and the findings do not define a standard regimen [55]. EPOR/JAK2-amplified AEL shows ruxolitinib sensitivity in vitro and in xenograft models, but no validated pediatric clinical strategy exists [48]. Adult TP53-altered PEL generally responds poorly to intensive chemotherapy, hypomethylating agents, and hypomethylating agent–venetoclax combinations; therefore, adult data should not be extrapolated to assume therapeutic benefit in children [10].

Prognosis is generally adverse but heterogeneous. In the largest pediatric cooperative-group analysis, the five patients with PEL had approximately 20% 5-year overall survival and event-free survival [9]. Pediatric erythroblastic sarcoma series also show poor aggregate survival, but mixed adult and pediatric cohorts, variable definitions, and frequent publication of severe extramedullary cases limit direct comparison [25]. Potential adverse features include PEL morphology, complex or monosomal karyotype, biallelic TP53 alteration, NUP98::KDM5A, induction failure, persistent MRD, relapse, and disseminated CNS or soft-tissue disease. These features should guide risk communication and trial referral, but they do not yet constitute a validated pediatric PEL prognostic model.

2.9. Measurable Residual Disease Monitoring

MRD is a central prognostic tool in pediatric AML; however, no PEL-specific threshold or validated MRD assay has been established [60,61]. Morphology alone is inadequate because its conventional detection limit is approximately 5%, early erythroid regeneration can mimic residual disease, and PEL may persist in extramedullary sites despite a morphologically negative marrow. Morphology should therefore describe response and marrow recovery but should not serve as the sole MRD method.

Multiparameter flow cytometry should use the diagnostic sample to define a leukemia-associated immunophenotype and should also apply a different-from-normal strategy at follow-up [60,61]. A PEL-oriented panel can include CD71, CD36, CD117, CD34, CD45, glycophorin A or another erythroid marker suitable for flow, and exclusion markers for myeloid, lymphoid, monocytic, and megakaryocytic differentiation. Practical limitations include erythroblast fragility, cell loss during red-cell lysis, hemodilution, variable CD34 or CD117 expression, and phenotypic shifts after treatment. Reserve the first marrow pull for MRD when feasible, and interpret flow results alongside biopsy findings and regeneration status.

Molecular MRD is potentially more specific in fusion-defined pediatric disease. When the genomic breakpoint is known, patient-specific RT-qPCR, digital PCR, or validated RNA- or DNA-based sequencing can track NUP98, NFIA, or another stable fusion. A negative broad DNA panel at diagnosis does not provide a molecular MRD target, and a mutation such as TP53 should be interpreted with allelic context and paired with flow because persistence may not always represent the same active leukemic population. The laboratory report should state assay sensitivity, specimen type, target stability, and the quantitative reporting threshold [60].

For documented CNS disease, CSF cytology should be combined with flow cytometry and, when technically validated, fusion-specific molecular testing. Reported use of CSF flow and NFIA::CBFA2T3 RNA detection shows diagnostic and monitoring feasibility, but current evidence is limited to individual cases, and no standardized CSF MRD threshold or schedule exists [27]. Reasonable AML-based time points include diagnosis, the end of induction courses, pre-HSCT assessment, post-HSCT surveillance when a stable molecular target exists, and any clinical suspicion of relapse. Treatment adjustment should follow the applicable pediatric AML protocol or multidisciplinary trial criteria rather than an unvalidated PEL-specific MRD cutoff.

2.10. Critical Appraisal and Evidence Gaps

The evidence base has major structural limitations. The largest pediatric study included only five cases of PEL, while most observations involving NFIA-, CIC-, and other fusion-defined subgroups derive from individual case reports or small case series [9, 15-16, 20-27]. Classification changes from FAB AML-M6 through successive WHO and ICC systems create substantial case-mixing and prevent direct comparison of older survival data. Publication bias likely favors unusual extramedullary presentations, molecularly solved cases, and severe outcomes. Conversely, children who respond well without extensive molecular testing may be underrepresented. The search was restricted to PubMed and English-language full-text publications. Studies indexed only in other databases, non-English reports, and grey literature may therefore have been missed. Initial database retrieval and exclusion counts were not prospectively archived. Figure 1 therefore represents a reconstructed audit of the final deduplicated source library rather than a prospectively documented PRISMA flow diagram.

Treatment and MRD conclusions require similar caution. No randomized or prospective PEL-specific therapeutic trial has been reported, HSCT selection is confounded by response and genetic risk, and targeted-therapy evidence is largely preclinical or derived from broader NUP98-rearranged pediatric AML [48, 55-59]. Flow-cytometric and molecular MRD methods have not been specifically validated for detecting fragile erythroid blasts or CNS erythroblastic sarcoma. Future progress requires international registry-based case capture, harmonized WHO and ICC annotation, central morphology and immunophenotype review, paired DNA and RNA sequencing, standardized marrow and CSF MRD collection, and prospective reporting of treatment exposure and outcome.

Pediatric PEL is a rare, high-risk erythroid-lineage leukemia that requires an integrated diagnostic approach and careful exclusion of reactive erythroid hyperplasia, myelodysplastic neoplasms, AML with myelodysplasia-related genetics, acute megakaryoblastic leukemia, and Down syndrome-associated myeloid proliferations. The adult model of biallelic TP53 alteration and complex karyotype remains important, but childhood disease includes NUP98-, NFIA-, CIC-, and other fusion-defined presentations with distinct marrow and extramedullary patterns. No PEL-specific pediatric treatment standard or validated prognostic score exists. Current management should follow a high-quality pediatric AML protocol or clinical trial, with HSCT decisions based on genetics, treatment response, MRD, relapse status, and disease distribution. MRD should combine flow cytometry with fusion-specific molecular monitoring when available, while CSF-based monitoring remains case-supported and unstandardized. Immediate priorities include transparent evidence grading, early RNA fusion testing, genotype-aware reporting, harmonized MRD methods, and international prospective registration.

Conceptualization, methodology, investigation, data curation, visualization, project administration, and writing—original draft: YY.; Validation, formal analysis, and supervision: R.P.; Writing—review and editing: Y.Y. and R.P. All authors have read and agreed to the published version of the manuscript.

All literature-derived data discussed in this narrative review are available in the cited publications. Figure 2 comprises original, previously unpublished, anonymized patient-derived bone marrow aspirate images supplied by the authors; these images are not available in the cited literature. The original source image files are retained for editorial verification and are not publicly deposited because they derive from clinical material and remain subject to applicable ethical and privacy restrictions. The source files may be provided confidentially to the editorial office by the corresponding author upon request.

This narrative review did not involve prospective human-subject research or collection of new research data. Figure 2 comprises anonymized, previously unpublished bone marrow aspirate images obtained during routine clinical care and supplied by the authors. No additional research procedure was performed for the purposes of this review. The ethical basis for publication of these clinical images is the written informed consent for publication obtained from the patient’s parent/legal guardian, as stated below.

Figure 2 contains anonymized, previously unpublished patient-derived bone marrow aspirate images obtained during routine clinical care and supplied by the authors. The images contain no directly identifying patient information. The patient's parent/legal guardian provided written informed consent for publication of the anonymized clinical images.

The authors declare no conflicts of interest.

The study did not receive any external funding and was conducted using only institutional resources.

Not applicable. No individuals or organizations other than the listed authors contributed to the manuscript through technical assistance, administrative support, language editing, proofreading, or other non-author contributions.

The use of artificial intelligence (AI) tools was limited to editorial and organizational assistance. ChatGPT (OpenAI) assisted with language polishing, response organization, and preliminary layout planning. Google NotebookLM assisted with organization of author-selected source materials and preliminary visual planning. No AI tool generated original research data, performed independent literature selection, interpreted patient findings, or produced unsupported scientific or clinical recommendations. The authors manually reviewed and verified every statement, reference, table, and figure and take full responsibility for the integrity, accuracy, and final content of the work.