AUTHORS
Yudi Zhang, MD, Wei Wang, PhD2, Xinping Zhou3, Liya Ma4, Yingwan Luo, MD, PhD5, Lingxu Jiang, MD6, Hongyan Tong, MD, PhD7
AFFILIATIONS
1 the First Affiliated Hospital of Zhejiang University, Hangzhou, China
2 the First Affiliated Hospital of Zhejiang University, Hangzhou, China
3 the First Affiliated Hospital of Zhejiang University, Hangzhou, China
4 the First Affiliated Hospital of Zhejiang University, Hangzhou, China
5 the First Affiliated Hospital of Zhejiang University, Hangzhou, China
6 the First Affiliated Hospital of Zhejiang University, Hangzhou, China
7 the First Affiliated Hospital of Zhejiang University, Hangzhou, China
ABSTRACT
Introduction
IPSS-M-Evo was recently developed by adding age and evolution-informed molecular features to the IPSS-M framework. Its incremental value over IPSS-M in Asian patients with myelodysplastic neoplasms (MDS), and the relative contributions of its added components, remain uncertain. We compared the prognostic performance of IPSS-R, IPSS-M, and IPSS-M-Evo for overall survival (OS) and leukemia-free survival (LFS), evaluated reclassification from IPSS-M to IPSS-M-Evo, and assessed individual IPSS-M-Evo components.
Methods
We retrospectively included 666 adults with newly diagnosed MDS treated at the First Affiliated Hospital of Zhejiang University between November 2018 and December 2025. Targeted sequencing used a 96-gene panel. GNB1, an IPSS-M residual gene, was not covered and was treated as missing in IPSS-M calculations; all other required clinical, cytogenetic, and molecular variables were available. OS and LFS were estimated using Kaplan–Meier methods and analyzed using Cox regression. Discrimination was compared using Harrell C-indices with 95% confidence intervals and paired bootstrap tests, with Holm adjustment for five prespecified comparisons per endpoint. Reclassification from IPSS-M to IPSS-M-Evo was evaluated using a 6×6 transition matrix. Component-wise analyses assessed the incremental contributions of age and the combined five-variable evolution-informed molecular component.
Results
The median age was 63 years (interquartile range, 55–70), and 64% of patients were male. After a median follow-up of 30.6 months, 272 (41%) patients died and 97 (14.6%) progressed to acute myeloid leukemia. The most frequent mutations were TP53 (25.7%), ASXL1 (24.6%), TET2 (17.9%), U2AF1 (14.3%), RUNX1 (13.2%), DNMT3A (12.5%), and SF3B1 (10.5%). At least one evolution-informed molecular variable was detected in 21 (3.2%) patients, including ASXL1→KRAS in 4 (0.6%), SRSF2→NRAS in 2 (0.3%), NRAS/RUNX1 co-occurrence in 7 (1.1%), ATRX in 1 (0.15%), and JAK2 in 8 (1.2%). IPSS-R, IPSS-M, and IPSS-M-Evo each significantly stratified OS and LFS (all log-rank P < 0.001). Their Harrell C-indices were 0.726, 0.734, and 0.755 for OS and 0.709, 0.721, and 0.736 for LFS, respectively. Compared with IPSS-R, IPSS-M-Evo increased the C-index by 0.029 for OS (95% CI, 0.006–0.052; P = 0.039) and 0.026 for LFS (95% CI, 0.005–0.050; P = 0.048); corresponding differences versus IPSS-M were 0.021 (95% CI, 0.012–0.030; P = 0.005) and 0.015 (95% CI, 0.006–0.025; P = 0.005), respectively (all P values Holm-adjusted).
IPSS-M-Evo reclassified 58.6% of patients relative to IPSS-M: 1.8% were upgraded, 56.8% downgraded, and 7.2% shifted by at least two categories. Reclassification was most frequent in the moderate-high, high, and very-high groups (74.7%, 61.4%, and 58.5%, respectively). Among reclassified patients, IPSS-M-Evo provided additional prognostic separation, most clearly within the original high and very-high groups. Outcomes after upgrading could not be reliably assessed because only 12 patients were upgraded and two events occurred; downgraded patients had superior OS and LFS within the original high and very-high groups compared with those remaining in their original categories.
Age yielded the largest C-index improvement for OS and LFS (ΔC-index, 0.017 and 0.011), particularly for OS. After Holm adjustment, the age increment remained significant for OS (P = 0.004) but not LFS (P = 0.054); the combined molecular component was significant for LFS (ΔC-index, 0.004; P = 0.015) but not OS (ΔC-index, 0.003; P = 0.054).
Conclusion
In this independent Chinese MDS cohort, IPSS-M-Evo showed superior prognostic discrimination compared with IPSS-M and IPSS-R and reclassified 58.6% of patients, particularly in the moderate-high to very-high groups. Age provided the largest incremental improvement, particularly for OS, while the evolution-informed molecular component improved LFS discrimination. These findings support the external applicability of IPSS-M-Evo and the value of age adjustment in refining prognostic assessment for MDS.
DISCLOSURES
Yudi Zhang, MD: No relevant conflicts of interest to declare.