Abstract 431


Primary myelofibrosis (PMF) has the worst prognosis among myeloproliferative neoplasms with median overall survival (OS) of 4.6y in the International Prognostic Scoring System (IPSS) series (Cervantes F, Blood 2009;113:2895) and 6.5y in patients (pts) seen more recently (Cervantes F, JCO, 2012 in press). OS is predicted by the four risk categories of IPSS, dynamic-IPSS and IPSS-plus system, and these scores are used for therapeutic choices particularly allogeneic stem cell transplantation. Nevertheless, pts heterogeneity still remains within these categories, necessitating improved risk stratification. A number of molecular abnormalities have been reported in PMF pts, but their prognostic relevance is incompletely understood, particularly with regard to transformation to leukemia (AL). The aim of this work was to analyze the prognostic impact of known mutations detected close to diagnosis in an international series of 429 pts.

Patients and methods.

PMF diagnosis had to satisfy the 2008 WHO criteria. Mutations in JAK2V617F, MPLW515, EZH2, ASXL1, TET2, IDH1/2, DNMT3A, CBL, SRSF2 were genotyped in whole blood or granulocytes using allele specific RTQ-PCR, HRM and direct sequencing; all mutations were confirmed at least twice. Missense, nonsense and frameshift mutations were considered; in case of novel mutations, SNPs were excluded by database searching and when feasible by germline DNA genotyping. The prognostic value of the molecular variables with regard to overall survival (OS) was analyzed by Cox regression and adjusted for the IPSS category. The association of molecular features with the risk of progression to AL was investigated in the framework of competing risks by the Fine & Gray regression method. Replicability of the prognostic models for both OS and progression to AL was assessed by replication in 1000 bootstrap samples randomly taken from the original series.


Patient median age was 60y. Median follow-up was 3.7y (95% CI, 0.02–27.9), death occurred in 157 pts (32%). Frequency of pts with constitutional symptoms was 28%, splenomegaly 74%, anemia 27%, leukocytosis 8%, >1% blasts 16%, thrombocytopenia 12%. Abnormal karyotype was found in 24% (of 229 evaluated). IPSS risk category: low-risk 35%, Int-1 30%, Int-2 21%, High-risk 14%. Frequency of mutations was: JAK2V617F 59.8% with 49% of pts having <25% allele burden; MPLW515 14%; EZH2 5%; ASXL1 21.3%; TET2 9.5%; IDH1-2 2.4%; DNMT3A 5.6%; CBL 4.3%; SRSF2 8.4%.

Survival Model.

Median survival was 9.7y (CI, 7.9–12.2), 22y in low-risk, 10y Int-1, 6.2y Int-2, 2.5y high-risk (P<0.0001). We found a strong association between IPSS risk categories and ASXL1 and SRSF2 mutated cases that clustered in the high-risk category (41.5% and 25.4%, respectively, P<0.001). ASXL1 mutations were associated with leukocytosis, blasts and constitutional symptoms; mutations in SRSF2 with older age, leukocytosis and symptoms. No other relevant associations between IPSS and molecular parameters was found. In the final prognostic model, only mutations in ASXL1 (Hazard ratio, HR:2.02; P<0.001) were found to add to IPSS (HR: 2.40; P<0.001) with regards to OS independently of the association of ASXL1 mutations and high-risk category. Within the Int-2/high-risk category, ASXL1 mutations were associated with significantly shorter survival (median survival 2.6 years months for mutated versus 5.8 years for unmutated; P=0.0004). According to bootstrap analysis, ASXL1 mutations were selected as significant predictor for OS in 74.6% of the samples.

Leukemia Model. AL occurred in 75 pts (15.2%) after of a median of 3.8y (95%, 0.04–26.5) from diagnosis. Mutations in ASXL1 and IDH1/2 were the only molecular variables associated with higher risk of AL with a SHR of 2.33 (P<0.001) and 3.63 (P=0.008), respectively. Bootstrap validation approach resulted in ASXL1 and IDH1/2 mutations being significant predictors for AL in 60% and 54% of the samples, respectively.


In this comprehensive series of mutations profiled in PMF pts, mutations in ASXL1 emerged as a powerful prognostic variable for survival refining prognosis in the Int-2/high risk IPSS categories. ASXL1 and IDH1/2 mutations predicted for death due to AL. Therefore, genotyping for ASXL1 and IDH1-2 mutations at diagnosis may help to tailor therapy for pts with IPSS Int-2/high risk PMF.


No relevant conflicts of interest to declare.

Author notes


Asterisk with author names denotes non-ASH members.

This icon denotes a clinically relevant abstract