AML classification in the year 2023: How to avoid a Babylonian confusion of languages

This study addresses how recent classification updates—WHO 2022 and the International Consensus Classification (ICC)—affect acute myeloid leukemia (AML) and myelodysplastic syndromes (MDS) diagnoses and risk stratification. WHO 2017 predominantly used morphology with some genetics, whereas WHO 2022 emphasizes genetics, dividing AML into AML with defining genetic abnormalities (DGA) and AML defined by differentiation, expanding entities (e.g., broader KMT2A and MECOM rearrangements, inclusion of NUP98) and removing AML with RUNX1 mutation as an entity. AML with myelodysplasia-related changes (AML-MRC) was redefined as AML myelodysplasia-related (AML-MR) using specific gene mutations and removing dysplasia criteria. WHO 2022 removes blast cutoffs for most genetically defined AML except for BCR::ABL1, CEBPA, and AML-MR; AML defined by differentiation retains the 20% blast cutoff. ICC differs by setting a 10% blast cutoff for AML-DGA, introducing an MDS/AML overlap category, defining AML with mutated TP53 as a new entity, narrowing CEBPA to in-frame bZIP variants, specifying partner genes in KMT2A/MECOM rearrangements, and separately considering AML with MR gene mutations and MR cytogenetic abnormalities (with mutations having higher hierarchy). ELN 2022 risk recommendations align with ICC genetic criteria. The study evaluates the diagnostic reclassification and prognostic impact of these changes in a large sequenced cohort.

The paper situates its analysis within evolving AML/MDS taxonomy literature: WHO 2017 as prior standard; WHO 2022 and ICC updates integrating genomic data; prior studies identifying MR-defining mutations (ASXL1, BCOR, EZH2, SF3B1, SRSF2, STAG2, U2AF1, ZRSR2) and their ontogeny implications; evidence informing CEBPA criteria (favorable impact of bZIP mutations; in-frame specification); mixed data on prognostic differences by KMT2A and MECOM fusion partners; poor outcomes associated with TP53 alterations and the complexity of TP53 biology (allelic status, co-mutations). ELN 2022 risk stratification changes, including removal of FLT3-ITD allelic ratio and inclusion of MR mutations and non-GATA2 MECOM rearrangements, reflect these updates.

Design: Retrospective cohort study of 1,451 non-therapy-related cases (717 MDS; 734 AML) diagnosed per WHO 2017. Samples were referred to MLL Munich Leukemia Laboratory between September 2005 and January 2020. All patients provided written informed consent; study approved by the institutional review board. Clinical data were obtained from treating physicians and updated at least annually. Genomics: Whole-genome sequencing (WGS; median coverage 100x) and whole-transcriptome sequencing (WTS; median yield 50 million reads) were performed (details in Supplementary Methods). Cytogenetics, cytomorphology, and immunophenotyping were part of the diagnostic workup. Reclassification: Cases were re-assigned to WHO 2022 and ICC categories, including assessment of AML-DGA, AML-MR (mutations and cytogenetic criteria), AML defined by differentiation, and the ICC MDS/AML category. Specific genetic entities included KMT2A- and MECOM-rearrangements (with partner gene resolution), NUP98 rearrangements, NPM1, CEBPA, and TP53. Statistics: Analyses were performed in SPSS v19. Overall survival (OS) estimated by Kaplan–Meier; comparisons via two-sided log-rank tests. OS defined from diagnosis to death or last follow-up. Harrell’s c-index assessed concordance between ELN risk categorization and outcomes. Significance threshold p<0.05.

• Cohort and baseline: 734 AML and 717 MDS; median age AML 68.4 years, MDS 73.2. Aberrant karyotype: AML 63.4%, MDS 43.6%. NPM1 mutations: AML 22.8%, MDS 0.8%. TP53 mutations: AML 8.4%, MDS 11.2%.
• Shift toward genetic entities and reduction of morphology-only AML:
  • Genetically defined AML proportion: WHO 2017 65% (477/734) vs WHO 2022 67% (502/746) vs ICC 69% (515/742) (excluding AML-MR for comparability).
  • Morphology-only AML decreased from 13% (99/734) AML-NOS (WHO 2017) to 5% (36/746) AML defined by differentiation (WHO 2022); AML-NOS reduced to 34 cases (5%) in ICC.
• Entity-specific changes:
  • Expanded KMT2A- and MECOM-rearranged AML in WHO 2022: KMT2A-r n=45 (42% with partner ≠ MLLT3); MECOM-r n=69 (48% with partner ≠ GATA2). Newly recognized: 5 NUP98-r (partners KDM5A n=2; NSD1 n=2; STIM1 n=1) and 1 KAT6A::CREBBP.
  • AML with RUNX1 mutation (WHO 2017) largely reclassified as AML-MR: 77% (37/48) in WHO 2022; 96% (46/48) in ICC (remaining 2 had other MECOM-r).
  • AML-MR increased from 22% (158/734) AML-MRC to 28% (208/746) AML-MR in WHO 2022; in ICC, AML-MR (mutations n=174; MR cytogenetics n=19) totaled 26% (193/742). The largest driver was presence of MR-defining mutations alone (44% of AML-MR in WHO 2022).
  • 23% (36/158) of WHO 2017 AML-MRC were not AML-MR in WHO 2022 (33 had new defining genetic abnormalities—55% MECOM-r; 3 were defined by differentiation).
  • ICC introduced AML with mutated TP53 (n=52), 92% (48/52) formerly AML-MRC; frequent complex karyotype and biallelic TP53; limited overlap with MR mutations.
• Upstaging from MDS to AML based on genetics/blasts:
  • WHO 2022: 12 former MDS (8 MDS-EB-2) upstaged to AML due to DGA (MECOM-r n=5; KMT2A-r n=1; NPM1 n=6).
  • ICC: 8 former MDS-EB-2 upstaged due to NPM1 or in-frame bZIP CEBPA.
  • Consistency: Only 4/16 cases were concordantly upstaged by both WHO 2022 and ICC.
  • ICC created MDS/AML category: 137/1,451 (9%) overall; subgroups—TP53 mutated 14% (19/137), MR gene mutations 72% (99/137), MR cytogenetic abnormalities 4% (6/137), NOS 10% (13/137).
• Survival (OS) analyses (overall p<0.001 for all classification systems):
  • CEBPA-mutated AML median OS: WHO 2017 5.0 years; WHO 2022 4.1 years; ICC not reached (reflecting narrower in-frame bZIP criteria).
  • Myelodysplasia-related AML: AML-MRC (WHO 2017) 0.4 years; AML-MR (WHO 2022) 0.5 years; AML-MR (ICC) 1.0 years (improved by excluding TP53-mutated cases with OS 0.1 years). Trend toward worse OS in biallelic vs monoallelic TP53.
  • No significant OS difference between KMT2A::MLLT3 vs other KMT2A-r; similarly, GATA2::MECOM vs other MECOM-r. MR cytogenetic vs MR gene mutation (ICC) OS similar: 0.71 vs 0.98 years (p=0.958).
• ELN risk reclassification impact (excluding APL): Favorable decreased 256→229; intermediate 155→125; adverse increased 275→332. Median OS: favorable 4.23→6.56 years; intermediate ~0.81→0.80; adverse 0.66→0.71. C-indices similar: ELN 2017 0.5967; ELN 2022 0.5961.
• Concordance between WHO 2022 and ICC AML subgroups was high (86%; 643/750), but 14% differed, mainly due to ICC’s TP53 entity, CEBPA criteria, partner gene splits, and MDS/AML category.

The study demonstrates that both WHO 2022 and ICC substantially shift AML diagnosis toward genetically defined entities, reducing morphology-only categories and increasing AML-MR when molecular criteria are applied. This realigns diagnoses with underlying biology and reveals prognostic distinctions, particularly in CEBPA-mutated and MR categories. ICC’s narrower CEBPA definition (in-frame bZIP) enriches for patients with superior outcomes, while its segregation of TP53-mutated AML captures a biologically and clinically adverse group, improving the prognostic profile of AML-MR. However, splitting KMT2A and MECOM by partner genes did not yield distinct survival differences in this cohort, questioning clinical utility of such granularity for outcome prediction. The introduction of MDS/AML by ICC affects 9% of patients and may facilitate trial inclusion but creates discrepancies with WHO 2022 regarding upstaging, underscoring the need for harmonization. ELN 2022 risk shifts more patients into the adverse group and improves median OS within the favorable group but does not enhance overall prognostic concordance versus ELN 2017. Overall, while there is high agreement between WHO 2022 and ICC, remaining differences—especially TP53 categorization, CEBPA criteria, and MDS/AML—have tangible diagnostic and therapeutic implications and warrant consensus-building informed by additional data.

In a large WGS/WTS-characterized cohort, WHO 2022 and ICC produced broadly concordant, genetics-focused AML classifications, markedly reducing morphology-only diagnoses and refining AML-MR definitions. Key differences—ICC’s TP53-mutated AML entity, stricter CEBPA criteria, specific KMT2A/MECOM partner delineations, and the MDS/AML category—alter diagnostic assignment and observed outcomes. ELN 2022 adjustments shift risk group distributions without improving overall prognostic discrimination versus ELN 2017. Approximately 86% of AML cases mapped to corresponding subgroups between WHO 2022 and ICC, but the residual non-comparability underscores the need for a unified classification to avoid diagnostic ambiguity. Future research should clarify the biological/clinical boundaries of TP53-altered AML, validate the utility of the MDS/AML category, assess whether KMT2A/MECOM partner splits confer actionable differences, and further refine risk stratification frameworks.

Single-laboratory cohort with multiple referring centers introduces heterogeneity in treatment approaches and follow-up. Inclusion of cases dating back to 2005 provides long follow-up but fewer patients treated with contemporary therapies. Referral patterns and WGS/WTS case selection may enrich for de novo AML and rare genetic subtypes; therapy-related myeloid neoplasms per WHO 2017 were not included. Some subgroup sample sizes (e.g., specific KMT2A/MECOM partners, TP53 allelic subsets) were limited, and outcomes were not stratified by treatment modalities, potentially impacting generalizability.