Clinical outcomes of WHO2022 and ICC MDS entities with risk stratification by IPSS-R and IPSS-M

The classification and prognostic stratification of myelodysplastic neoplasms (MDS) are undergoing substantial refinement, driven by the integration of molecular genetic data with established morphologic assessment. In 2022, two major diagnostic frameworks were introduced: the 5^th edition of the World Health Organization classification (WHO2022) and the International Consensus Classification (ICC).^1,2 Both are designed to provide globally applicable, evidence-based definitions of MDS entities, facilitating a shared diagnostic language for clinical care and research. Apart from these two disease classifications, there has been a significant shift in the risk stratification model predicting survival and leukemia progression in MDS, which also emerged in 2022. The Molecular International Prognostic Scoring System (IPSS-M) was introduced,³ with its enhanced predictive capabilities over the previous revised IPSS (IPSS-R) primarily attributed to the incorporation of molecular genetic alterations.⁴ To explore how these frameworks function in practice, we retrospectively reclassified a large single-center cohort of Korean patients with MDS - originally diagnosed according to WHO2016 - under both WHO2022 and ICC criteria. We then examined how outcomes varied within each classification and assessed the prognostic performance of IPSS-R and IPSS-M when applied within those diagnostic categories.

The study included 639 adult MDS patients per WHO2016 criteria aged 18 years or older who were treated at Seoul St. Mary’s Hospital, a national tertiary referral and transplant center, between 2007 and 2022. Patients were enrolled if they had available next-generation sequencing (NGS) data from untreated samples and complete clinical, pathologic, and cytogenetic information. The Institutional Review Board of the Catholic Medical Center approved this study (KC18TESE0700). The analyses were conducted in accordance with the guidelines of the Institutional Review Board and followed the ethical principles outlined in the Declaration of Helsinki. The diagnosis was independently re-reviewed by five experienced hematopathologists and reclassified into new classifications considering additional genetic information. The baseline characteristics of the patients are summarized in Table 1. The median age was 58 years (range, 18-88), and 380 (59.5%) were male. Patients with therapy-related MDS were included; this subgroup comprised 5.3% of the cohort. Disease-modifying therapy - including lenalidomide, hypomethylating agents, or allogeneic hematopoietic stem cell transplantation (HSCT) -was given to 388 patients (60.7%). Cytogenetic analysis was performed using conventional G-banding techniques, with abnormalities reported according to the 2024 International System for Human Cytogenetic Nomenclature.⁵ Fluorescence in situ hybridization assays targeted common recurrent alterations, including del(5q), del(20q), chromosome 7 abnormalities, MECOM rearrangement, and TP53 deletion. NGS used the St. Mary’s customized 87-gene myeloid panel, and variants were classified in accordance with the Association for Molecular Pathology guidelines and somatic oncogenicity criteria.⁶ Variants with at least 20 supporting reads and variant allele frequency (VAF) ≥5% were considered pathogenic; known driver hotspot mutations were included even if below this VAF threshold. All detected DDX41 mutations underwent germline confirmation via buccal DNA testing.⁷ Multiplex ligation-dependent probe amplification was used to identify KMT2A partial tandem duplications.

Table 1.Baseline characteristics.

Among the 639 enrolled patients, 25 subjects (3.9%) were classified as AML by WHO2022 due to KMT2A, MECOM, and NUP98 rearrangements or an NPM1 mutation. Six (0.9%) were classified as clonal cytopenia of undetermined significance (CCUS) and 17 (2.7%) as chronic myelomonocytic leukemia (CMML). The remaining 592 (92.6%) had MDS, including 80 (13.5%) as ‘MDS with defining genetic abnormalities’ and 512 (86.5%) as ‘morphologically-defined MDS’ (Figure 1). By ICC, five patients (0.8%) were classified as AML, seven (1.1%) as CCUS, and 17 (2.7%) as CMML. Of the 610 (95.5%) defined as MDS, 81 (13.3%) were assigned to the distinct subgroup MDS/AML. In direct comparison, WHO2022 classified 19 additional cases as AML due to low blast counts. Except for cases with TP53 mutation, most patients in the ICC-defined MDS/AML category were classified as MDS with increased blasts-2 (MDS-IB2) under WHO2022 (N=68, 95.9%). Another key difference lies in the definition of MDS with BM blast <5%: WHO2022 applies BM cellularity, whereas ICC relies on the number of dysplastic lineages. There was no significant difference in the proportion of hypoplastic MDS (hMDS) between patients classified as MDS, not otherwise specified, with single-lineage dysplasia (MDS-NOS-SLD) and those with multilineage dysplasia (MDS-NOS-MLD) (18/77, 23.4% vs. 55/251, 21.2%). In our cohort, the distribution of MDS subtypes was largely consistent with previous large unselected series including hypoplastic MDS (hMDS).^8-10

Figure 1.Redistribution of myelodysplastic syndromes patients according to WHO2016, WHO2022 and ICC classifications. MDS: myelodysplastic neoplasm; WHO: World Health Organization classification; ICC: International Consensus Classification; MDS-5q: MDS with low blasts and 5q deletion; MDS-SLD: MDS with single-lineage dysplasia; MDS-RS: MDS with ring sideroblasts; MDS-MLD: MDS with multilineage dysplasia; MDS-EB1: MDS with excess blasts-1; MDS-EB2: MDS with excess blasts-2; MDS-U: MDS unclassifiable; MDS-SF3B1: MDS with low blasts and SF3B1 mutation; MDS-biTP53: MDS with biallelic TP53 inactivation; MDS-LB: MDS with low blasts; hMDS: hypoplastic MDS; MDS-IB1: MDS with increased blasts-1; MDS-IB2: MDS with increased blasts-2; MDS-F: MDS with increased blasts and fibrosis; AML: acute myeloid leukemia; CCUS: clonal cytopenia of undetermined significance; CMML: chronic myelomonocytic leukemia; NOS: not otherwise specified: MDS-TP53: MDS with mutated TP53; MDS/AML-gene: MDS/AML with myelodysplasia-related gene mutations; MDS/AML-cyto: MDS/AML with myelodysplasia-related cytogenetic abnormalities.

Of all 639 enrolled patients, pathogenic mutations were detected in 446 (69.8). These included 153 (23.9%) with one mutation, 129 (20.2%) with two mutations, and 164 subjects (25.7%) with three or more mutations. Nine genes were mutated in more than 5% of patients: ASXL1 (N=120), U2AF1 (N=114), DDX41 (N=58), TP53 (N=56), RUNX1 (N=55), SF3B1 (N=47), DNMT3A (N=45), TET2 (N=43), and STAG2 (N=31). Among the 193 patients without a detectable mutation, 105 had karyotypic abnormalities, resulting in 551 (86.2%) with at least one genetic abnormality. Among the 592 patients with MDS classified by WHO2022, 410 (69.3%) had at least one mutation. Of the 182 without a detectable mutation, 96 had karyotypic abnormalities, giving a total of 506 (85.5%) with genetic abnormalities. Of the 610 MDS cases classified by ICC, mutations were detected in 425 (69.7%), and 524 (85.9%) had either mutations or karyotypic abnormalities. This frequency is somewhat lower than the 80-90% reported in Western cohorts¹¹ but is comparable to other Asian series, where overall mutation rates are 60-70% and SF3B1 mutations are notably less frequent.¹² Conversely, ASXL1 and U2AF1 were among the most common drivers, consistent with prior Asian reports.¹³

After excluding patients classified as CCUS, CMML, or AML within each system, survival analysis was conducted in 592 patients by WHO2022 and 610 by ICC. Both classifications significantly discriminated overall survival (OS) and leukemia-free survival (LFS) (Online Supplementary Figure S1A-D). The worst outcomes were seen in MDS-biTP53 (median OS 1.0 year) and MDS with increased blasts and fibrosis (MDS-F) (1.0 year) by WHO2022, and MDS/AML-TP53 (0.9 year) and MDS-TP53 (1.0 year) by ICC. The most favorable outcomes were observed in MDS with low blasts (MDS-LB) (median not reached [NR]) and hMDS (14.6 years) in WHO2022, and MDSNOS-SLD (median NR) in ICC. Consistently, MDS-SF3B1 (9.8 years) and MDS-5q (9.3 years) showed prolonged survival in both systems. LFS showed similar patterns to OS. Also, when we assessed IPSS-R and IPSS-M in 592 WHO2022-defined MDS patients, both systems showed effective discrimination against survival (Online Supplementary Figure S1E-H). However, IPSS-M demonstrated superior predictive ability, with a statistically higher Harrell’s C-index (0.751 vs. 0.713; P<0.001),¹⁴ and further stratified prognosis within the same IPSS-R groups (Online Supplementary Figure S1I-M).

To further clarify prognostic capabilities of diagnostic classifications, we integrated WHO2022 with IPSS-R and IPSS-M. When analyzing the distribution of IPSS-M-based risk groups within each WHO2022 category, the ‘very low’ to ‘moderate low’ risk groups constituted the majority of the MDS-5q, MDS-SF3B1, hMDS, and MDS-LB subsets (Figure 2A-D). Conversely, MDS-biTP53 and MDS-F were classified as ‘very high’ risk by the IPSS-M (Figure 2E-H). The prognostic utility of IPSS-M was more pronounced in MDS-LB, MDS-IB1, and MDS-SF3B1 categories. For other MDS categories, the discriminatory power of IPSS-M seemed to largely be limited by an inadequate number of patients. Online Supplementary Figure S2 displays the IPSS-R/IPSS-M risk categories by WHO2022 classification.

In addition, we assessed the survival impact of newly defined subcategories within WHO2022 and ICC by Cox models adjusted for sex, IPSS-M, and HSCT (Online Supplementary Table S1). In WHO2022, newly defined category MDS-F showed poorer median OS (1.0 year) than MDS-IB1 (6.1 years) and MDS-IB2 (3.2 years) (Online Supplementary Figure S1A), but fibrosis was not an independent prognostic factor in multivariable model (adjusted hazard ratio [HR] vs. MDS-IB1: MDS-IB2, =1.12, 95% confidence interval [CI]: 0.69-1.85; MDS-F adjusted HR=1.44, 95% CI: 0.64-3.23). In patients with lower blast counts, no significant difference was observed in hMDS, a recently identified entity, compared to MDS-LB (adjusted HR=1.24, 95% CI: 0.74-2.08). In ICC, survival did not differ between MDS-NOS-SLD and MDS-NOS-MLD (adjusted HR=1.17, 95% CI: 0.67-2.04). Similarly, no differences were observed among MDS/AML subcategories (MDS/AML with MDS-related cytogenetic abnormalities and MDS/AML-NOS vs. MDS/AML with MDS-related gene mutations: adjusted HR=2.08; 95% CI: 0.77-5.65 and HR=1.38, 95% CI: 0.56-3.37, respectively). Finally, within the TP53-mutated subsets, survival was comparable between MDS-TP53 and MDS/AML-TP53 adjusted HR of MDS/AML-TP53 vs. MDS-TP53 adjusted HR=0.81, 95% CI: 0.29-2.29).

This study clarified the role of contemporary MDS classifications in clinical practice and highlights the complementary but distinct functions of diagnostic classification and prognostic scoring. Both WHO2022 and ICC provided meaningful population-level outcome discrimination, while IPSS-M demonstrated superior prognostic accuracy across morphologically defined subgroups.⁹ Importantly, TP53-mutated categories consistently showed the poorest outcomes, reaffirming their distinct significance. After adjustment for IPSS-M (and HSCT where applicable), morphologic features such as hypocellularity, fibrosis, and the number of dysplastic lineages did not retain independent prognostic value. Although morphology-defined entities such as hMDS and MDS-F retain value beyond prognosis clarifying the differential diagnosis with related mimics (e.g., aplastic anemia and primary myelofibrosis) and, in selected patients, informing entity-aligned management (e.g., immunosuppression in hMDS), our findings underscore the need to refine the interface between diagnostic classification and prognostic scoring models.¹⁵

This retrospective, single-center analysis spans a long accrual period and has a relatively short follow-up. Our cohort skews older patients with a high proportion of HSCT recipients and small numbers in several subgroups. Nevertheless, our results offer actionable guidance: WHO2022/ICC remain indispensable for disease definition and biologic categorization, whereas IPSS-M provides the most reliable patient-level survival prediction - including within diagnostic entities. Moreover, morphology-defined features contribute little independent prognostic information once IPSS-M is considered. Taken together, further refinements should integrate molecularly defined categories, more closely with prognostic modeling to improve risk stratification and ultimately guide therapeutic strategies.

Figure 2.Overall survival according to IPSS-M within each WHO2022 category. (A-H) IPSS-M: Molecular International Prognostic Scoring System; MDS: myelodysplastic neoplasm; WHO: World Health Organization classification; MDS-5q: MDS with low blasts and 5q deletion; MDS-SF3B1: MDS with low blasts and SF3B1 mutation; MDS-biTP53: MDS with biallelic TP53 inactivation; MDS-LB: MDS with low blasts; hMDS: hypoplastic MDS; MDS-IB1: MDS with increased blasts-1; MDS-IB2: MDS with increased blasts-2; MDS-F: MDS with increased blasts and fibrosis.

Footnotes

• Received April 23, 2025
• Accepted October 20, 2025

Correspondence

*DK and BB contributed equally as first authors

M. KIM - microkim@catholic.ac.kr

YJ. KIM - yoojink@catholic.ac.kr

Disclosures

No conflicts of interest to disclose.

Contributions

Funding

This research was funded by National Research Foundation of Korea (NRF) grant (NRF-2022R1A2C2006746) (to YJK) and the Korea Health Industry Development Institute (KHIDI) (grant number: RS-2025-22842969/ RS-2025-22852971) (to MK).

Acknowledgments

This study utilized the cancer clinical-genomic database developed as part of the project “Establishment of Real-World Evidence and Technology Development for Precision Medicine Based on targeted NGS Data”.