Abstract
Background: Current prognostication in primary myelofibrosis (PMF) utilizes international prognostic scoring systems that rely on clinical parameters that are sensitive to day-to-day variations and subjective interpretation. Recent studies in PMF have disclosed important prognostic information attached to additional cytogenetic details (Blood. 2011;118:4595) and somatic mutations, including CALR and ASXL1 (NEJM. 2013;369:2379; Leukemia. 2013;27:1861).
Methods: PMF diagnosis and definition of blast transformation (BT) were according to World Health Organization criteria (Blood. 2009;114:937). Cytogenetic analysis and reporting was done according to the International System for Human Cytogenetic Nomenclature (Cytogenetic and genome research. 2013. Prepublished on 2013/07/03 as DOI 10.1159/000353118). Previously published methods were used for analyses of CALR, JAK2, MPL and other prognostically-relevant mutations, including ASXL1, SRSF2, EZH2 and IDH(Leukemia. 2014;28:1472).
Results: The training set included 964 Mayo Clinic patients (median age 65 years; 62% males) in whom informative karyotype or mutation information was available; cytogenetic information was available in 903 (94%) cases, JAK2/CALR/MPL mutational status in 532 (55%), ASXL1 in 425 (44%), SRSF2 in 434 (45%), IDH1/2 in 376 (39%) and EZH2 in 268 (28%). DIPSS-plus (JCO. 2011;29:392) risk distribution was high in 37% of patients, intermediate-2 in 37%, intermediate-1 in 15% and low in 11%.
We used a revised risk stratification for cytogenetics (see accompanying ASH 2014 abstract) to distinguish four distinct cytogenetic risk categories: very high (monosomal karyotype, inv(3), i(17q), -7/7q-, 11q or 12p abnormalities; n=67), high (complex non-monosomal, two abnormalities not included in very high risk category, 5q-, +8, other autosomal trisomies except +9, and other sole abnormalities not included in other risk categories; n=164), intermediate (sole abnormalities of 20q-, 1q+ or any other sole translocation, and -Y or other sex chromosome abnormality; n=133) and low (normal or sole abnormalities of 13q- or +9; n=539).
Mutational frequencies were 58% for JAK2, 25% CALR, 7% MPL, 36% ASXL1, 11% SRSF2, 5% IDH1/2 and 6% EZH2. The 131 cases with CALR mutations were further subclassified into two prognostically different groups: type 1/type 1-like (n=110) and type 2/type 2-like (n=21) (see accompanying ASH 2014 abstract).
At a median follow-up time of 4.2 years for patients who are alive, 664 (69%) deaths and 70 BT (7%) were recorded. Age-adjusted multivariable analysis that included cytogenetic and mutational risk groups disclosed the following as independent predictors of shortened survival: very high risk karyotype (HR 4.2; 3 points), high risk karyotype (HR 1.9; 1 point), triple-negative (HR 2.8; 2 points), JAK2 (HR 3.1; 2 points), MPL (HR 3.1; 2 points), type 2/type 2-like CALR (HR 3.6; 2 points), ASXL1 (HR 1.9; 1 points) and SRSF2 (HR 1.9; 1 point); EZH2 (p=0.24) and IDH1/2 (p=0.68) and intermediate risk karyotype (p=0.87) were not significant.
The above-mentioned significant variables and age demarcated at 60 years (2 points), were subsequently used to develop an HR-derived, genetics-based prognostic scoring system (GPSS) for 369 patients who were fully informative for both karyotype and all significant mutations: low risk (0 points; n=31), intermediate-1 (1 or 2 points; n=90), intermediate-2 (3 or 4 points; n=133) and high (5 or more points; n=115); the corresponding median survivals were >17, 9 (HR 4.7, 95% CI 1.7-13.0), 5 (HR 10.7, 95% CI 3.9-29.3) and 2.2 (HR 29.2, 95% CI 10.6-80.0) years (Figure 1). High risk GPSS was also associated with higher BT rate (HR 7.4, 95% CI 2.1-26.3).
The prognostic distinction between high/intermediate-2 and low/intermediate-1 risk GPSS, in terms of both overall (median 5 vs 26.4 years; HR 7.1, 95% CI 3.3-14.9) and leukemia-free survival (median 11.6 years vs not reached; HR 9.4, 95% CI 2.2-41.0) was validated in an independent cohort of 183 patients from the University of Florence (Figure 2).
Conclusions: The current study demonstrates the feasibility of genetics-based prognostic models in PMF that rely on objective parameters that are amenable to further refinement as new genetic information becomes available.
No relevant conflicts of interest to declare.
