The current study validates the predictive performance of the BLAST and BLAST-Mol risk models in a large two-center international cohort with a total of 1101 patients with newly diagnosed CMML. Chronic myelomonocytic leukemia (CMML) is a myeloid neoplasm with overlapping features between myelodysplastic syndrome (MDS) and myeloproliferative neoplasms (MPN) [1]. It has a variable clinical course, complicated by constitutional symptoms, progressive cytopenia or cytosis, and an inherent risk of transformation to acute myeloid leukemia (AML), also known as blast transformation (BT) [2]. Several CMML survival risk models [3-5] have been developed over the years with the objective of guiding treatment decisions such as allogeneic stem cell transplant (ASCT) and facilitating patient selection for participation in clinical trials [6]. The recently published BLAST clinical and BLAST molecular (BLAST-Mol) risk models incorporate clinical and molecular variables, respectively, for prediction of overall survival (OS), while the BLAST-BT risk model was predictive of BT-free survival (BTFS) [6]. Similarly, the contemporary molecular CMML-specific prognostic scoring system (CPSS-mol) utilizes clinical variables including white blood cell count (WBC), bone marrow (BM) blasts, red cell transfusion dependency along with genetic variables like cytogenetics and mutations (ASXL1MUT, NRASMUT, SETBP1MUT, and RUNX1MUT) for stratification of patients into four risk groups with distinct OS and risk of BT [3]. In the current study, the Mayo Clinic (USA) CMML cohort used in the aforementioned study was expanded from 457 to 600 patients and merged with an external cohort of 501 patients from the Humanitas Research Hospital (HRH; Italy) in order to validate the BLAST, BLAST-Mol, and BLAST-BT risk models. The retrospective study was conducted under an institutional review board approved minimal risk protocol. CMML diagnosis and sub-categorization were according to the International Consensus Classification [1], which included CMML-1 based on BM/peripheral blood (PB) blasts < 10%/< 5% versus CMML-2 (BM/PB blasts ≥ 10%/≥ 5%) along with myeloproliferative (CMML-MP; WBC ≥ 13 × 109/L) and myelodysplastic (CMML-MD; WBC < 13 × 109/L) subtypes [1]. The BLAST score is based on clinical risk factors including PB blasts ≥ 2% (1 point), WBC ≥ 13 × 109/L (1 point), and severe anemia (defined as transfusion dependent or hemoglobin < 8 g/dL in women and < 9 g/dL in men, 2 points) or moderate anemia (defined as hemoglobin 8 to < 10 g/dL in women and 9 to < 11 g/dL in men, 1 point) resulting in BLAST low (0 points), intermediate (1 point), and high (2–4 points) risk groups [6]. In addition, BLAST-Mol accounts for mutations and karyotype and identifies PHF6 and TET2 as “favorable” and DNMT3A, U2AF1, BCOR, SETBP1, ASXL1, NRAS, PTPN11, RUNX1, and TP53 as “unfavorable” mutations. Abnormal karyotypes other than sole loss of Y or +8 were also considered unfavorable. Molecular and cytogenetic features result in molecular risk categorization into high: in the presence of unfavorable mutation/adverse karyotype and absence of favorable mutations (2 points), intermediate: neither low or high molecular risk (1 point), or low molecular risk groups: in the presence of favorable mutations alone (0 points) stratifying patients into BLAST-Mol low (0–1 points), intermediate (2–3 points), and high risk (4–6 points) groups [6]. BLAST-BT variables include absence of PHF6MUT (3 points), BM blasts ≥ 10% (2 points), presence of DNMT3AMUT(2 points), PB blasts ≥ 2% (1 point), leukocyte count ≥ 13 × 109/L (1 point), and ASXL1MUT(1 point) resulting in BLAST-BT very low (0–2 points), low (3 points), intermediate (4 points), high (5 points), and very high risk (≥ 6 points) groups [6]. OS analyses were censored for ASCT. Conventional statistical methods were employed, using JMP Pro 18.0.0 software SAS Institute, Cary, NC, USA. Median follow-up was calculated using the reverse Kaplan–Meier method. Cumulative incidence of BT was calculated after accounting for death from another cause as a competing risk. A total of 1101 consecutive patients (median age 73 years; 67% males) with newly diagnosed CMML were considered in the analysis: Mayo cohort (N = 600) and HRH cohort (N = 501). Compared to the HRH cohort, the Mayo cohort was more likely to include CMML-MP (44% vs. 28%; p < 0.01), CMML-1 (87% vs. 81%; p < 0.01), ASXL1MUT (48% vs. 35%; p < 0.01), SRSF2MUT (43% vs. 33%; p < 0.01), NRASMUT (15% vs. 9%; p < 0.01), PHF6MUT (7% vs. 4%; p < 0.01), and less likely to harbor TET2MUT (50% vs. 58%; p < 0.01), or SF3B1MUT (4% vs. 10%; p < 0.01) (Table S1). Median follow-up was longer (57 vs. 51 months; p = 0.02) with higher rates of deaths (66% vs. 58%; p = 0.01) and ASCT (17% vs. 8%; p < 0.01) in the Mayo cohort. At a median follow-up of 55 months (range 0–202) in the whole cohort, 686 (62%) deaths, 218 (20%) BTs, and 142 (13%) ASCTs were documented. The CPSS-mol risk [3] categories at diagnosis included low (N = 190; median OS 67 months), intermediate-1 (N = 271; median 38 months), intermediate-2 (N = 407; median 26 months), and high (N = 231; median 16 months; p < 0.01) (AUC at 5 years 0.75). The corresponding AUCs for CPSS-mol calculated for the Mayo and HRH cohorts, separately, were 0.73 and 0.76, respectively. BLAST clinical risk distribution was low (N = 356; median OS 65 months), intermediate (N = 393; median 30 months), and high (N = 352; median 16 months; p < 0.01) (Figure 1; AUC at 5 years 0.79). The corresponding AUCs for BLAST calculated for the Mayo and HRH cohorts, separately, were 0.81 and 0.76, respectively. BLAST-Mol risk assignments included low (N = 357; median OS 66 months), intermediate (N = 512; median 27 months), and high (N = 232; median 15 months; p < 0.01) (Figure 2; AUC at 5 years 0.79). The corresponding AUCs for BLAST-Mol for the Mayo and HRH cohorts were 0.81 and 0.77, respectively. Compared with BLAST clinical risk model stratification, the BLAST-Mol risk model reassigned 255 patients (23%) into different risk categories: 19% (N = 67) initially classified as BLAST low were reassigned to BLAST-Mol intermediate risk, 17% (N = 68) of those in the BLAST intermediate group were reclassified as BLAST-Mol low risk, and 34% (N = 120) of the BLAST high risk patients were reassigned to BLAST-Mol intermediate risk groups (Figure S1). The 3-year cumulative incidence of BT (accounting for death from another cause) stratified by the BLAST-BT risk model (evaluable = 1093) included 8% for very low risk (N = 54), 11% for low risk (N = 332), 20% for intermediate risk (N = 281), 27% for high risk (N = 250), and 34% for very high risk (N = 176) (p < 0.01; Figure 3). Univariate analysis of the combined Mayo–Humanitas cohort (N = 1101) identified the following variables not included in BLAST or BLAST-Mol risk models with additional prognostic significance; inferior OS was associated with CMML-2 (median 15 vs. 38 months; p < 0.01), age > 65 years (median 29 vs. 40 months; p < 0.01), EZH2MUT (median 18 vs. 33 months; p < 0.01), NPM1MUT (median 8 vs. 32 months; p < 0.01), FLT3MUT (median 12 vs. 32 months; p < 0.01), IDH1MUT (median 16 vs. 32 months; p = 0.04), CEBPAMUT (median 18 vs. 31 months; p < 0.01), and absence of SF3B1MUT (median 30 vs. 48 months; p < 0.01). Multivariable analysis (MVA) confirmed the BLAST-Mol-independent prognostic contribution of age > 65 years (HR 1.6, 95% CI 1.3–2.1; p < 0.01), CMML-2 (HR 2.0, 95% CI 1.6–2.5; p < 0.01), NPM1MUT (HR 2.9, 95% CI 1.6–5.6; p < 0.01), EZH2MUT (HR 1.5, 95% CI 1.1–2.0; p = 0.01), and SF3B1MUT (HR 0.7, 95% CI 0.5–0.95; p = 0.02). Furthermore, an age and sex adjusted all-inclusive MVA highlighted the BLAST-Mol-independent unfavorable prognostic impact of CMML-2 (HR 1.9, 95% CI 1.5–2.4; p < 0.01), NPM1MUT (HR 3.2, 95% CI 1.7–5.9; p < 0.01), EZH2MUT (HR 1.6, 95% CI 1.2–2.1; p < 0.01), U2AF1MUT (HR 1.6, 95% CI 1.2–2.1; p < 0.01), BCORMUT (HR 1.7, 95% CI 1.1–2.7; p = 0.02), TP53MUT (HR 1.7, 95% CI 1.2–2.6; p < 0.01), and a near significant favorable impact of SF3B1MUT (HR 0.7, 95% CI 0.5–1.0; p = 0.05) on OS. The performance of BLAST, BLAST-Mol, and CPSS-mol risk models was further explored in patients with CMML-2 at diagnosis. In this subgroup, all three models performed poorly compared to their performance in CMML-1. The BLAST clinical model stratification showed no significant difference in OS between low and intermediate groups (19 vs. 16 months, p = 0.3) and only a borderline difference between intermediate and high risk (16 vs. 13 months, p = 0.05), while low risk did have an improved survival compared to high risk group (19 vs. 13 months, p < 0.01). Similarly, BLAST-Mol risk stratification showed an improved survival in the low compared to intermediate (21 vs. 15 months, p = 0.02) and high-risk groups (21 vs. 13 months, p < 0.01), while survival was similar in intermediate and high risk (p = 0.1). Finally, there was no difference in OS based on CPSS-mol risk stratification (p > 0.05 in all instances, Figure S2). The clinical management of CMML remains challenging. BT remains a frequent disease complication and leading cause of death [6, 7]. Current drug therapy, typically with hydroxyurea or hypomethylating agents (HMAs), has not been shown to prevent BT or prolong OS [2, 8]. ASCT is the only potential curative therapy; however, given its substantial treatment-associated morbidity and mortality, it is not a suitable option for all patients [2]. Accurate risk prognostication of patients can guide treatment and ASCT decisions. The current study validates the predictive performance of the BLAST, BLAST-Mol, and BLAST-BT risk models in a large two-center international cohort of patients with CMML. The BLAST and BLAST-Mol model performance was superior to that of the widely utilized CPSS-mol risk stratification [3]. Blast count based subclassification (CMML-1 vs. CMML-2) was first proposed in the 3rd edition of the WHO classification [9] and remains prognostically relevant [1, 10], with CMML-2 being associated with an adverse prognosis independent of BLAST-Mol, suggesting subtype-adjusted risk models for CMML-1 may improve clinical utility. Our observations also confirm the unfavorable prognostic impact of rare AML-like mutations (NPM1/FLT3/CEPBA) [11, 12] in CMML, particularly NPM1MUT, which remained independently significant and identifies a subgroup of patients with dismal prognosis warranting early ASCT consideration. In the current study, only 14 cases were informative cases for NPMMUT, but we observed clustering with DNMT3AMUT (71% vs. 29%, p < 0.01), and it was noted to be mutually exclusive of ASXL1MUT (0% vs. 100%, p < 0.01) and RUNX1MUT (0% vs. 100%, p = 0.02). These findings warrant investigation in a larger cohort. Interestingly, we found BLAST-Mol-independent adverse impact of U2AF1MUT, BCORMUT, and TP53MUT, which are already included in the BLAST-Mol molecular risk stratification [6], suggesting their negative impact may be underestimated in the current model. Previously, BCORMUT has been reported to be associated with CMML-2, but prognostic impact remained uncertain [13]. Similarly, TP53MUT has been observed to be associated with CMML-2, therapy-related CMML, and HMA resistance [14, 15]. TP53MUT, particularly multihit configuration, is associated with dismal OS and leukemia-free survival [14]. Finally, our findings are consistent with a previous report from Xicoy et al. [16], who have previously shown the favorable prognosis of the SF3B1MUT in 491 patients with CMML. OS was significantly longer in patients with ring sideroblasts (RS)/SF3B1MUT (6.75 vs. 3.17 years for CMML without RS/ SF3B1MUT, p < 0.001). Similarly, cumulative incidence of BT was lower with CMML with RS/ SF3B1MUT (12% vs. 24%, p < 0.001) [16]. Considering the complementary nature of the BLAST clinical and BLAST-Mol risk models, a stepwise approach to their use appears reasonable. Prognostication can begin with the clinically based BLAST model, which is readily applicable in most settings, and be refined by incorporating molecular risk through BLAST-Mol when available. Practically, molecular risk stratification can help identify BLAST low-risk patients who are re-stratified as BLAST-Mol intermediate or high risk and may warrant consideration of earlier ASCT. Conversely, BLAST intermediate-risk patients who are reclassified as low-risk by BLAST-Mol might benefit from the deferral of ASCT. These observations collectively validate the BLAST, BLAST-Mol, and BLAST-BT models and suggest the possibility of further enhancement of the BLAST-Mol model by the incorporation of independent prognostic relevance from additional mutations in the genetic risk stratification. S.F., M.Y., P.F., A.A., C.C., M.N., L.L., A.C., G.M., M.G.D.P., K.K.R., R.H., and A.T. were involved in study design, gathering, and analysis of data. N.G., A.A.M., A.P., M.M.P., and A.T. participated in patient care. S.F. and A.T. wrote the paper. The authors have nothing to report. A.A.M.: Research funding from Nssovartis, BMS, Solu Therapeutics, and Sanofi. N.G.: Advisory board to DISC Medicine and Agios. M.M.P.: Research funding from Kura Oncology, Stemline Therapeutics, Epigenetix, Solu Therapeutics, Polaris, and has served on the advisory board for AstraZeneca and SOBI Pharmaceuticals. The other authors declare no conflicts of interest. By email request to the corresponding author. Figure S1: Supporting Information. Figure S2: Supporting Information. Table S1: Supporting Information. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.
No takes yet. Share an insight, caveat, or question.
Fathima et al. (2025) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: