Comparative evaluation of ipss-m and cpss-mol in CMML: Impact of molecular markers on prognosis

Abstract Introduction Chronic myelomonocytic leukemia (CMML) is a heterogeneous hematologic malignancy with overlapping myelodysplastic (MD) and myeloproliferative (MP) features, making accurate risk stratification essential for guiding clinical management. The CMML-specific CPSS-Mol score (2016) integrates clinical parameters (e.g., bone marrow (BM) blasts, white blood cell count (WBC), transfusion dependency, cytogenetic risk groups) with selected genetic alterations (ASXL1, NRAS, RUNX1, SETBP1) to classify patients into 4 prognostic groups. In contrast, the more recently established IPSS-M (2022), primarily developed for myelodysplastic neoplasms (MDS), incorporates 31 recurrently mutated genes alongside key clinical parameters such as blood counts and cytogenetics, refining risk stratification into 6 categories. Although IPSS-M has shown potential applicability in CMML, it remains uncertain which clinical and molecular factors are most critical for risk stratification, particularly across MD-CMML and MP-CMML subtypes. Methods We retrospectively analyzed 480 patients with CMML comprising the international IWG cohort (n=389) and the Japanese J-MDS cohort (n=91). Both CPSS-Mol and IPSS-M scores were calculated based on the respective specifications. Overall survival (OS) was analyzed using Kaplan-Meier estimates, and differences between risk groups were tested with the log-rank method. Model discrimination was quantified by Harrell's concordance index (c-index). To statistically compare the discriminative performance of the two models, we used the R package “compareC” to test for significant differences between concordance indices. We first performed univariate Cox regression to identify significant and clinically relevant parameters, which were then included in a multivariate Cox model to determine independent prognostic factors. Results Of 480 patients, 279 (58%) had MD-CMML and 186 (39%) MP-CMML. IPSS-M outperformed CPSS-Mol in the overall cohort (c-index: 0.760 vs. 0.695, p<0.001) and across subtypes (MD-CMML: 0.780 vs. 0.732, p<0.001; MP-CMML: 0.724 vs. 0.614, p<0.001), with the most pronounced difference seen in MP-CMML. Especially when comparing the extreme risk categories, a significant difference was observed between the high-risk groups, with a median OS of 9.6 months for IPSS-M (very high-risk) versus 13.6 months for CPSS-Mol (high-risk, p=0.031). A similar trend was observed in the low-risk groups, where IPSS-M (very low-risk) reached a median OS of 110.7 months compared to 66.0 months for CPSS-Mol (low-risk, p=0.036). Following an initial univariate analysis, significant and clinically relevant variables were included in a multivariate Cox model. In this model, several classical CPSS-Mol parameters, including BM blasts, WBC, platelet counts, NRAS and SETPB1 did not retain independent prognostic value. In contrast, hemoglobin (HR 0.87, p<0.001), age (HR 1.04, p<0.001), and a low cytogenetic risk profile (HR 0.50, p=0.021) were independent prognostic factors. Moreover, multiple genetic alterations were also associated with worse OS: RUNX1 (HR 2.02, p=0.001), BCORL1 (HR 2.75, p=0.044), EZH2 (HR 2.95, p<0.001), STAG2 (HR 3.06, p=0.022), GATA2 (HR 3.74, p=0.007), and del5q (HR 3.90, p=0.027). The overall model achieved a c-index of 0.783 (se=0.019), indicating strong prognostic performance. Conclusions Our analysis confirms the applicability of IPSS-M in a large CMML cohort and demonstrates its superior discriminative capacity compared to CPSS-Mol, particularly in MP-CMML and across the highest and lowest risk categories. Our findings indicate that classical CPSS-Mol parameters have limited prognostic impact once comprehensive molecular data are included, emphasizing the relevance of the extended genetic profiling as used in IPSS-M. These findings validate the clinical utility of IPSS-M and suggest that further refinement of CMML-specific models should focus on incorporating additional molecular determinants, particularly for proliferative disease variants.