Comprehensive Clinical-Molecular Transplant Risk Model for Myelofibrosis Undergoing Allogeneic Stem Cell Transplantation

Background

The dynamic International Prognostic Scoring System (DIPSS) is commonly applied to predict survival among patients with primary myelofibrosis (PMF) but has been shown to perform less precisely in secondary myelofibrosis (SMF) and after transplantation. Furthermore, the prognostic relevance of mutation profile resulted in the mutation-enhanced IPSS (MIPSS) in PMF, as well as in a model specific to SMF (MYSEC-PM) after essential thrombocythemia (ET) or polycythemia vera (PV). The aim of the current study was to develop a comprehensive prognostic system including clinical and molecular information, specifically in myelofibrosis undergoing transplantation.

Methods

Previously published methods were used to sequence myelofibrosis-associated genes (i.a. CALR1/2, JAK2, MPL,ASXL1, SRSF2, EZH2, IDH1/2, DNMT3A, TET2, TP53). Outcome was calculated from date of transplant (95% confidence interval). Variables associated with overall survival (OS) constructed a Cox regression with a stepwise selection procedure. Hazard ratios (HR) were used as weights for model development. Validation was done using repeated random subsampling. Performance of the model was verified via Harrel's concordance index C and was also tested in predefined cohorts: disease (PMF, SMF), conditioning, and ruxolitinib pretreatment.

Results

Population. The total cohort consisted of 361 patients from four different centers in Germany and France (260 PMF, 101 SMF). Median age at transplant was 57 years (range, 22-75), 58% were male and 42% had a Karnofsky performance score (KPS) <90. The median follow-up was 62 months and was similar between PMF and SMF (p=0.50). Overall 5-year OS was 60% (54-67) being similar in PMF (63%) and SMF after ET (59%) and slightly lower after PV (45%). Most frequent mutations were: JAK2 V617F (57%), CALR (20%; types 1/2/other 66%/23%/11%), MPL (5%), ASXL1 (31%), TET2 (19%), SRSF2 (9%), DNMT3A (6%), TP53 (6%). Two or more mutations were present in 60%. Most transplants were received from matched unrelated (46%), mismatched unrelated donors (MMUD, 27%), identical siblings (27%), and mismatched siblings (1%). Reduced intensity was given more frequently (64%) than myeloablative conditioning (36%). Frequencies at transplant were 9% (low), 29% (intermediate-1), 48% (intermediate-2), 14% (high) according to DIPSS and 3% (low), 40% (intermediate), and 57% (high) for MIPSS.

Factors on outcome.In univariate analysis, mutations in CALR and MPL showed better OS (79% and 76%) vs. JAK2 (53%) and triple negative (50%; p=0.001). Outcome was similar according to CALR type (p=0.99). ASXL1 and DNMT3A mutations also entered the multivariate model. The following eight clinical, molecular and transplant-related variables were identified (corresponding HR): leukocytes >25x10⁹/l (1.71), platelets <150x10⁹/l (1.53), KPS <90 (1.63), age >57 years (1.69), recipient/donor CMV serostatus (+/- vs. other, 1.68), ASXL1 (1.74), JAK2/triple negative (2.10), and MMUD (2.11).

Myelofibrosis Transplant Scoring System (MTSS). A weighted score of 1 was assigned to leukocytosis, thrombocytopenia, KPS <90, age >57, recipient/donor CMV serostatus (+/-), and ASXL1 mutation, whereas 2 points were assigned to JAK2/triple negative and MMUD. Four risk groups constructed the MTSS: low (score 0-2), intermediate (score 3-4), high (score 5-6), and very high (score 7-9). The 5-year OS according to risk groups was 88%, 71%, 50%, and 20% (Figure 1). The hazard for death (with low-risk as reference) was 2.36 for intermediate-risk, 4.65 for high-risk, and 9.72 for very high-risk. The score was predictive of OS overall as well as for PMF and SMF (p<0.001, respectively). The MTSS showed overall C statistics of 0.718 (0.707-0.730) after cross-validation yielding a median of 0.727 in PMF and 0.708 in SMF indicating improved performance and replicability vs. DIPSS (0.572), MIPSS (0.577), and MYSEC-PM (0.601). The system was also predictive of OS in different conditioning settings (reduced intensity and myeloablative) and in patients with ruxolitinib pretreatment (p<0.001, respectively).

Conclusions

The new MTSS includes modern disease- and transplant-associated risk variables pertinent to both PMF and SMF. This proposed system consistently predicts outcome facilitating posttransplant decision-making and can be applied to different conditioning settings and to patients receiving ruxolitinib pretreatment.

Figure 1.

View largeDownload slide

Figure 1.

View largeDownload slide

Close modal