The Dynamic International Prognostic Scoring System-plus (DIPSS-plus) uses eight adverse factors to predict survival in primary myelofibrosis (PMF): age >65 years, hemoglobin <10 g/dL, leukocytes >25 x 109/L, circulating blasts 31%, constitutional symptoms, red cell transfusion need, platelets <100 x 109/L and unfavorable karyotype (J Clin Oncol. 2011;29:392). The latter two also predicted leukemic transformation. Most recently, an international study of 879 PMF patients identified ASXL1, SRSF2, EZH2 and IDH1/2 mutations as inter-independent predictors of shortened overall or leukemia-free survival (Leukemia 2013Apr 26 doi: 101038/leu2013119. 2013). Molecular prognostication is now well established for acute myeloid leukemia with normal karyotype (NK). There is currently little information on clinical or molecular prognostication in NK-PMF.
Diagnosis of PMF was according to WHO criteria. The study population included all referrals to the Mayo Clinic, Rochester, MN, USA or the University of Florence, Florence, Italy. Information on clinical and laboratory parameters and karyotype was available in all study patients, at time of referral, which coincided with time of sample collection for mutation screening. The current study considered only those patients with NK-PMF. ASXL1, EZH2, SRSF2, IDH1/2 and JAK2 mutations were performed on variable number of patients depending on bone marrow or granulocyte DNA availability. Clinical parameters examined for prognostic relevance included those listed for DIPSS-plus, less unfavorable karyotype.
A total of 690 patients with NK-PMF patients were studied. Median age was 65 years (14-89). The percentage of patients with age >65 years was 48%, red cell transfusion dependent 35%, hemoglobin <10 g/dL 49%, platelets <100 x 10(9)/L 19%, leukocyte count >25 x 10(9)/L 15%, circulating blasts ≥1% 49% and constitutional symptoms 33%. At a median follow-up of 29 months, 351 (51%) deaths and 39 (6%) leukemic transformations were recorded.
The respective frequencies (mutated/number of patients studied) of JAK2V617F, ASXL1, EZH2, SRSF2 and IDH1/2 mutations were 60% (284/473), 34% (72/214), 8% (14/179), 15% (37/249) and 5% (14/262). There was no significant association between JAK2V617F and any one of the aforementioned mutations. Inter-mutation association was evident only between SRSF2 and IDH1/2 (p=0.0005). Each one of the DIPSS-plus risk parameters was examined for correlation with a specific mutation: ASXL1 was associated with leukocyte count >25 x 10(9)/L (p<0.0001) and circulating blasts ≥1% (p=0.0005); EZH2 with leukocyte count >25 x 10(9)/L (p=0.008); and SRSF2 with age >65 years (p=0.0007), transfusion need (p=0.04) and hemoglobin <10 g/dL (p=0.03). JAK2V617F was associated with age>65 years (p=0.003) and leukocyte count >25 x 10(9)/L (p=0.02).
In univariate analysis, all 7 DIPSS-plus parameters, as well as ASXL1, EZH2, SRSF2 and IDH1/2 mutations showed significant association with shortened survival: When each of these analyses was adjusted for age, all except constitutional symptoms and SRSF2 mutations retained their significance. Multivariable analysis of the six age-independent DIPSS-plus variables identified all but transfusion-need as independent predictors of inferior survival. A similar analysis restricted to mutations identified ASXL1 and EZH2 as being independently adverse. When both mutations and DIPSS-plus variables were included in the multivariable model, only four parameters remained significant: age >65 years (HR 4.2; p<0.0001), platelets <100 x 10(9)/L (HR 3.4; p<0.0001), ASXL1 mutations (HR 2.2; p=0.0001) and EZH2 mutations (HR 2.7; p=0.001). A similar analysis identified SRSF2 mutations (HR 5.9; p=0.0003) and platelets <100 x 10(9)/L (HR 4.3; p=0.005) as the only predictors of leukemia-free survival.
In NK-PMF, molecular markers might be prognostically more useful than conventional models that rely on clinical features. In the current study, thrombocytopenia was the only clinical variable, other than age, with additional value to molecular prediction of survival and leukemic transformation.
No relevant conflicts of interest to declare.
Asterisk with author names denotes non-ASH members.