Although the most recent WHO5 and ICC classifications [4, 5] of MDS have begun to incorporate gene mutations, the MDS classification framework remains more grounded in historical morphologic classification systems than do classifications of AML. A recent large MDS-IWG study attempted to develop a molecular classification of myelodysplastic syndrome, reporting numerous clinicopathologic associations [15]. We sought to apply the MDS-IWG classification to a cohort of patients diagnosed at our institution to assess its utility and to test its applicability in an independent cohort of patients.

The application of the MDS-IWG hierarchical classification to our cases was straightforward, though it should be emphasized that the MDS-IWG categories do not entirely map onto those currently in clinical use. As an example, since the SF3B1 category is late in the hierarchy, many SF3B1-mutated MDS cases will have already been sorted into other genetic groups: indeed, only 76% of WHO5/ICC-defined MDS with SF3B1 (and only 61% of all cases with SF3B1 mutation) were in the SF3B1 MDS-IWG genetic class. Similarly, the bi-TET2 group included cases with SRSF2 and TET2 co-mutations, in addition to those with biallelic inactivation of TET2. In addition, the TP53-complex group is more expansive than are the currently accepted TP53-mutated MDS categories.

Overall, our findings closely align with those reported by Bernard et al. [15], including association of monocytosis and older patient age with the bi-TET2 group and lower blast counts in the SF3B1 group. Bone marrow blast percentages were lower in certain groups, such as SF3B1 and bi-TET2, while increased blasts were enriched in the TP53-complex and IDH-STAG2 groups. Some associations reported by Bernard et al. were not seen in our smaller cohort, such as the reported association between the BCOR/L1 group and thrombocytopenia [15].

Recently, the relevance of blast quantification in the classification of MDS has been questioned [12, 14]. While it is evident that across all MDS cases taken as a whole, blast counts provide prognostic information [18], the significance of blast counts within specific genetic groups may vary [14, 19]. The MDS-IWG study suggested that blast counts were prognostically relevant primarily in lower-risk molecular groups (del (5q), bi-TET2, SF3B1, CCUS-like, mNOS) but were less relevant in the higher-risk molecular groups (−7/SETBP1, AML-like, DDX41, EZH2-ASXL1) [15]. Our study demonstrated overall similar findings, with clear importance of increased blasts in the SF3B1 and bi-TET2 groups, but without significance in the poor-prognosis EZH2-ASXL1 group. We also saw significance of blast counts in the more intermediate-risk IDH-STAG2 group, and in contrast to Bernard et al., in our cohort, blast counts appeared to be a significant prognostic factor in the −7/SETBP1 group. As our cohort contained fewer cases than did the MDS-IWG cohort, the significance of this finding is uncertain.

The importance of TP53 mutations in MDS is well recognized, and TP53 status was incorporated into current classification systems [4, 5] due in part to prior work by the MDS-IWG that demonstrated the significance of multihitTP53 aberrations in MDS [20]. The MDS-IWG genetic TP53-complex group, however, includes both cases with multihit TP53 alterations and cases that lack TP53 mutations but have complex karyotypes. In our cohort, blast percentage was not significant within this very high-risk group, though in the MDS-IWG cohort, it was with the caveat that all blast tiers had very poor survival [15]. Although blast percentage was not significant in TP53-complex cases in our cohort, the presence of a TP53 mutation was with TP53-mutated cases, exhibiting much worse outcomes than cases that only exhibited genetic complexity. Similar findings appear to have been present in the MDS-IWG cohort [15], suggesting that grouping cytogenetically complex cases without TP53 mutations together with multihit TP53 cases in a biologic classification may not be appropriate.

Although the MDS-IWG-defined genetic subgroups exhibit differences in survival, we confirmed that the IPSS-M risk stratification retains better predictive performance regarding patient outcomes. This finding was also reported in the MDS-IWG cohort and in other proposed genetic classification systems [13, 15]. This emphasizes the difference of purpose between the proposed genetic taxonomy, in which blast percentage and adverse hematologic findings may reflect an advanced stage of disease within a biologically defined group, and prognostic models that fully incorporate these features to predict for patient outcome.

Our study has several limitations, including the smaller size of our patient cohort compared with that of the MDS-IWG. Our study is retrospective, and patients were heterogeneously treated, which could affect assessment of survival data. A major limitation of our study was the lack of information regarding DDX41 mutation for all our patients. DDX41 is the first category in the MDS-IWG mutational hierarchy, but DDX41 was not included in the sequencing panels used for this study. However, DDX41 mutations are relatively uncommon (3% of the MDS-IWG cohort), and they tend to have relatively few co-occurring mutations [15]. As we identified our patient cohort partly by the presence of somatic mutations, it is likely that very few DDX41-mutated patients are present and misclassified in the cohort. Similarly, assessment for KMT2A partial tandem duplication (part of the AML-like signature) was not available for 67% of our patients, and STAG2 mutations were not assessed in 43% of the cohort. As these mutations are fairly uncommon and require co-mutations to be relevant to the molecular classification, the number of misclassified patients due to their absence is likely to be low and should not significantly affect the overall conclusions of our study. Due to our cohort identification methods, we also would not have identified relatively infrequent cases of MDS with chromosomal abnormalities but lacking somatic mutations.

In this study, we validated numerous findings described in the MDS-IWG paper on the molecular taxonomy of MDS in an independent cohort, and our findings support that the MDS-IWG taxonomy could be a basis for a future revision to MDS classification. We demonstrate genetic-subtype-specific differences in the prognostic importance of blast percentages, and we confirm the association of different molecular groups with overall survival and certain clinicopathologic features. We also highlight the continued relevance of the existing IPSS-M risk stratification, as it predicted for patient outcome better than did the biologic classification. Our findings support retention of the current multihit TP53 MDS category included in the ICC and WHO5 and suggest that MDS with complex karyotype but without TP53 mutations should be classified separately.