Molecular predictors of survival in patients with myeloproliferative neoplasm-blast Phase (MPN-BP) treated with venetoclax and decitabine

Abstract Background. MPN-BP after chronic phase MPN is characterized by dismal outcome, with OS <6 mo. Owing to disappointing results of intensive chemotherapy in MPN-BP, there is increasing interest for non-intensive regimens, such as venetoclax (VEN) plus hypomethylating (HMA) agents, successfully used in unfit, untreated, as well as relapsed/refractory, pts with de-novo AML. Genetic risk score specific for de-novo AML receiving VEN/HMA were developed: ELN24/mPRS (includes TP53, FLT3-ITD, NRAS and KRAS mut; Dohner H, Blood 2024); ELN24-refined (DDX41, NPM1, IDH1, IDH22, N/KRAS, PTPN11, FLT3-ITD, TP53 mut; Lachowiez CA, Blood 2024); Mayo Genetic Risk Model (adverse karyotype, KMT2Ar, TP53mut, KRASmut, absence of IDH2mut; Gangat N, AJH 2025). However, the cytogenetic and molecular landscape of MPN-BP is highly heterogeneous and usually lacks abnormalities typical of de-novo AML. Aim Using data from ENABLE, a phase 2 trial of VEN and decitabine in MPN-BP (submitted to ASH2025, AM Vannucchi et al), we aimed to define a prognostic risk profile associated with OS. Patients and Methods. The trial enrolled 101 pts. DNA was obtained from purified blasts at diagnosis. Mutations in driver and a 40-myeloid gene panel were determined by PCR assays and NGS sequencing, respectively. Karyotype abnormalities were classified according to Tefferi A (JCO 2018;1169) and ELN22 (Dohner H, Blood 2022). Results Karyotype and molecular information were available in 80 patients, of which 72 (90.0%) had a driver mutation (JAK2 73.8%, CALR 12.5%, MPL 3.7%) and 93.8% had >1 (median, 4) myeloid gene mutations. An abnormal karyotype was found in 59% of the pts, classified as adverse in 40 (50%) according to Tefferi and 22 (27.5%) according to ELN22. We first evaluated the performance of available predictive scores. We found that either ELN17 (p=0.82) and ELN22 (p=0.10) score failed to accurately predict outcome, as it could be anticipated since they were originally developed in the settings of intensive regimens for de-novo AML pts. On the other hand, ELN24/mPRS (p=0.05) better separated the favorable category (54% of the pts; median OS, 20.7mo) from the intermediate (17%, OS 8.9mo) and the adverse one (29%; OS 5.1mo); the latter 2 categories, however, were almost superimposable (p=0.35). In ELN24-ref system (p=0.12), OS was not reached for pts in favorable category (9% of the pts), compared to 14mo and 10mo in intermediate (57%) and adverse (24%), respectively, that again did not differ each to the other (p=0.10). Conversely, the Mayo Genetic Risk Model score failed to discriminate risk categories (p=0.88), possibly because of the lack of karyotype abnormalities typical of de-novo AML in our MPN-BP patients, except for chr17 involvement. In univariable analysis, OS was predicted by achievement of CR/PR within 2 cycles (55% of the pts; p=0.02), SRSF2mut (25%; p=0.048), U2AF1mut (7.5%; p<0.001), EZH2mut (16.3; p=0.047) and TP53mut (28.8%; p=0.01) irrespective of being multihit (17.5%). High molecular risk mutations (Guglielmelli P, JCO 2018; 36:301) (p=0.24) and adverse karyotype, according to either Tefferi (p=0.32) and ELN22 (p=0.72), were not significant. In multivariable analysis, the factors that maintained significance were SRSF2mut (HR 3.0; 95%CI 1.4-6.3; p=0.004), U2AF1mut (HR 4.2, 1.6-10.7; p=0.003) and TP53mut (HR2.7, 1.3-5.3; p=0.005). We attributed 2-points to SRSF2mut and U2AF1mut and 1-point to TP53mut, and obtained 3 survival curves in Kaplan-Meier analysis (p=0.006). In the favorable category (point 0; n=15 pts, 18.8%) OS was 28.1 mo (95%CI, 16.1-28.5) compared to 11.5mo (6.3-16.7) in the intermediate (point 1; n= 35 pts, 43.8%) and 4.4mo (0.1-11.5) in the adverse (n=30 pts, 37.4%) risk category. By using the favorable as reference category, the HR was 2.4 (95%CI, 1.0-7.1; p=0.049) and 2.7 (95%CI, 5.4-15.9; p=0.004) for intermediate and adverse category, respectively. With C-statistical analysis, the area under the curve was 0.71 for the 3-gene model, versus 0.60 for ELN22, 0.65 for ELN24 and 0.64 for ELN-ref. Corresponding concordance index (CI) value were 0.68 compared to 0.56, 0.59 and 0.57. Conclusions A 3-gene signature, including SRSF2, U2AF1 and TP53, was able to identify prognostically different risk categories among MPN-BP pts treated with VEN/DEC, resulting more accurate than genetic risk models specifically developed for de-novo AML pts. These findings warrant confirmation in independent series. [On behalf of ENABLE investigators].