Clinical and Molecular Insights into the Classification Dilemma: Are MDS/AML Cases Distinct from Secondary AML?

MDS/ AML has been introduced as a new entity by the ICC classification among patients (pts) with 10-19% blasts in the absence of AML- defining molecular alterations. Clonal evolution suggests that pts with secondary AML (sAML) would exhibit a molecular profile similar to that of MDS/AML cases with additional mutations (MT) acquired during disease progression. To date, distinct clinical and molecular differences have not been identified. Moreover, risk stratification for this particular entity remains unclear, as pts can be classified as high-risk MDS (per WHO 2022) or MDS/AML overlap (per ICC 2022).

Here, we aim to analyze pts with MDS/AML overlap to highlight differences in molecular features and clinical outcomes using MDS-based IPSS-M and AML-based ELN risk stratifications. Also, we identified distinct molecular signatures of MDS/AML cases based on our previously reported machine learning unsupervised molecular clustering approach (Kewan & Durmaz, et al. Nat Comm. 2023).

We analyzed 3,598 pts with myeloid neoplasms from a multicenter cohort. Pts with bone marrow (BM) blasts <10% were excluded. Cases were classified into MDS/AML (blasts: 10-19%), oligoblastic sAML (OB-sAML, blasts: 20-29%) and sAML (blasts ≥30%). Only pts with available molecular data were included and IPSS-M was calculated for pts with BM blasts ≤30%. Overall survival (OS) was assessed from the time of diagnosis. The performance of IPSS-M and ELN risk groups on OS prediction was assessed using the Harrell's C-index. Pts were assigned to one of the 14 identified molecularly distinct clusters (C).

Overall, 898 pts were included, of whom 367 (41%) had MDS/AML, 213 (24%) had OB-sAML, and 318 (35%) had sAML. Only 80 (9%) pts received allogenic BM transplantation. The median age was 71 (IQR: 65-77) years. Normal karyotype was observed in 424 (47%) of the cases. Compared to pts with OB-sAML, pts with MDS/AML had significantly higher median age (73 vs. 71 years, p=0.01), higher hemoglobin level (9.9 vs. 9.4 g/dL, p=0.001), and higher platelet count (91 vs. 50 x10⁹/L, p=<0.001). MDS/AML cases had a significantly higher frequency of normal karyotype vs OB-sAML/sAML (55% vs. 42%, p=0.006). Trisomy 8 (15% vs. 7%) was more often encountered in sAML (BM blasts ≥30%) compared to MDS/AML (p= 0.002). Complex karyotype was not different between the two groups (20% vs. 24%, P=0.297).

For all cases, TET2 (24%), ASXL1 (24%), RUNX1 (23%), and SRSF2 (22%) were the most common identified MT. Pts with MDS/AML had more ASXL1^MT (31% vs. 19%, p=0.002), SRSF2^MT (31% vs. 17%, p=0.003), TET2^MT (38% vs. 13%, p<0.001) and U2AF1 ^MT (10% vs. 5%, p=0.03). DNMT3A^MT (11% vs. 6%) was significantly higher in sAML cases with BM blasts ≥30% compared to MDS/AML (p=0.02).

According to our molecular clustering model, more MDS/AML cases were classified to C6 (normal karyotype, SRSF2^MT, and RAS^MT, 17% vs. 3%) and C9 (normal karyotype, SRSF2^MT, ASXL1^MTand RUNX1^MT, 11% vs. 6%). While more OB-sAML pts were classified to C2 (normal karyotype, DNMT3A^MT, and RAS^MT 27% vs. 15%) and C14 (Del-Y and SF3B1^MT, 11% vs 3%). In reverse analysis, MDS/AML cases constituted most of C3 (Del-Y, TET2^MT, ZRSR2^MT, and ASXL1^MT, 88%), C6 (91%), C7 (other MTs, 79%), C9 (76%), C10 (normal karyotype, SF3B1^MT, DNMT3A^MT, and TET2^MT, 14%), and C12 (normal karyotype, TET2^MT, ASXL1^MT, SRSF2^MT, and RUNX1^MT, 81%). OB-sAML cases composed most of C5 (Del-20q, U2AF1^MT, and ASXL1^MT, 55%) and C14 (72%).

Most of the MDS/AML cases were assigned to adverse ELN risk (82% vs. 62%) and 34% had high risk IPSS-M. IPSS-M very-high risk was more frequent among OB-sAML cases (68% vs. 37%). The median OS for MDS/AML pts (25.5, 95% CI: 20.1-34.2 mo.) was significantly higher than OB-sAML (14, 95%CI: 11-22 mo.) and sAML with BM blasts ≥30% (10, 95%CI: 8-13 mo.) pts, p<0.001. Compared to ELN risk groups, IPSS-M had better C-index (95% CI) for OS in MDS/AML (0.724 vs. 0.546) pts and OB-sAML pts (0.608 vs. 0.526).

In summary based on our machine learning model, pts with MDS/AML had unique clinical and molecular features compared to pts with OB-sAML and sAML. ASXL1, SRSF2, TET2 and U2AF1 genes were more frequently mutated in MDS/AML pts. OB-sAML and sAML generally had worse OS, which reflects the advanced stage of the disease. Despite being validated in cases with blasts up to 19%, the IPSS-M provided better OS predictions for both MDS/AML and OB-sAML cases compared to the ELN risk groups, highlighting the urgent need for improved prognostic models for sAML cases.

Carraway:Novartis: Membership on an entity's Board of Directors or advisory committees; Abbvie: Membership on an entity's Board of Directors or advisory committees; Celgene: Research Funding; BMS: Membership on an entity's Board of Directors or advisory committees; Stemline: Membership on an entity's Board of Directors or advisory committees; Daiichi: Membership on an entity's Board of Directors or advisory committees; Jazz: Membership on an entity's Board of Directors or advisory committees, Speakers Bureau; Servier: Membership on an entity's Board of Directors or advisory committees. Balasubramanian:Alexion AstraZeneca: Speakers Bureau; Kura Oncology: Research Funding. Bat:Recordati Rare Diseases: Other: Advisory Board; Novartis: Other: Advisory Board; Alexion: Other: Advisory Board; Sanofi: Other: Advisory Board. Madanat:OncLive, MD Education, Sierra Oncology, Stemline, MorphoSys: Consultancy; Blueprint Medicines, MD Education, and Morphosys: Other: travel; Taiho Oncology, Rigel Pharmaceuticals, Novartis: Consultancy; Sierra Oncology, Stemline Therapeutics, Blueprint Medicines, Morphosys, Taiho Oncology, SOBI, Rigel Pharmaceuticals, Geron, Cogent Biosciences and Novartis: Other: Advisory Board; BMS, Kura Oncology, BluePrint Medicines, Geron: Consultancy. Maciejewski:Alexion: Consultancy, Honoraria; Novartis: Consultancy, Honoraria, Speakers Bureau.