AUTHORS
Lucia Masarova, MD, Andrew Dunbar, MD2, Prithviraj Bose, MD3, Naveen Pemmaraju, MD4, Hussein Ali Abbas, MD, PhD5, Danielle Hammond, MD6, Mahesh Swaminathan, MD7, Dzifa Duose, PhD8, Taghi Manshouri9, Hagop Kantarjian, MD10, Guillermo Garcia-Manero, MD11, Keyur P. Patel, MD12
AFFILIATIONS
1 The University of Texas MD Anderson Cancer Center, Houston, Texas, United States
2 The University of Texas MD Anderson Cancer Center, Houston, Texas, United States
3 The University of Texas MD Anderson Cancer Center, Houston, Texas, United States
4 The University of Texas MD Anderson Cancer Center, Houston, Texas, United States
5 The University of Texas MD Anderson Cancer Center, Houston, Texas, United States
6 The University of Texas MD Anderson Cancer Center, Houston, Texas, United States
7 The University of Texas MD Anderson Cancer Center, Houston, Texas, United States
8 The University of Texas MD Anderson Cancer Center, Houston, Texas, United States
9 The University of Texas MD Anderson Cancer Center, Houston, Texas, United States
10 The University of Texas MD Anderson Cancer Center, Houston, Texas, United States
11 The University of Texas MD Anderson Cancer Center, Houston, Texas, United States
12 The University of Texas MD Anderson Cancer Center, Houston, Texas, United States
ABSTRACT
Introduction: Myelofibrosis (MF) is characterized by broad clinical and molecular heterogeneity. Although currently approved JAK inhibitors have tremendous impact on patients’ lives, the disease still progresses, often driven by increasing genomic complexity. Limited understanding exists about the association of clinical disease behavior, use of JAK inhibitors and genomic patterns over time.
Objective: We aimed to investigate how initial and subsequent genomic signatures relate to clinically defined MF phenotypes on long-term JAK inhibitor therapy and their dynamic correlations.
Methods: We performed targeted next generation DNA sequencing (blood or bone marrow; sensitivity 1.5%) of 110+ genes using Illumina NextSeq to identify single nucleotide variants (SNVs) and insertions-deletions (indels). Patients must have ≥1 subsequent SEQ over ≥3 years. Genomic results were integrated with clinical phenotype, serial blood counts and JAK inhibitor use. Analysis included descriptive statistics, Kaplan-Meier method with log rank test for overall survival (OS) and cox regression models for variables predicting outcomes as appropriate (significant p values < 0.05).
Results: 34 MF patients (65% primary MF) underwent longitudinal SEQ over a median of 7.5 years (range 3-11). Median SEQ times per patient were 4 (range, 2-19). All patients had SEQ in the chronic phase (CP; < 10% blasts); 11 patients additionally in accelerated/blastic phase (AP; ≥10%/20% blasts). Ten patients (29%) had SEQ prior to JAK inhibitor (9 ruxolitinib, RUX; 1 momelotinib, MMB); 33 (97%) had SEQ on JAK inhibitor (32 RUX, 1 MMB, 2 fedratinib). Median (range) overall follow-up from SEQ1 was 9 years (3.5-16). At SEQ1, baseline characteristics included median age of 64.5 years (range 31-80), 56% males, 50% higher risk by DIPSS, 68% splenomegaly ≥5 cm. 12%, 35% and 15% had WBC > 25 x 109/L, hemoglobin < 10 g/dL and platelets < 100 x109/L, respectively.
Across all SEQ timepoints, we identified 488 somatic coding mutations (all recurrent: 347 SNVs and 141 indels). Overall median MUT burden per patient was 12 (range, 3-47). JAK2, MPL, CALR occurred in 27 (79%), 4 (12%) and 3 (8.8%) patients with median variant allele frequency (VAF) of 69%, 60% and 5.8%. Additional MUT (AddMUT) were detected in 27 (79%) patients: 21 (62%) at SEQ1 and 24 (71%) at subsequent SEQ (clonal evolution; CE; median time to detection 5 years; range, 2-10). ASXL1, TET2, RAS, TP53, SH2B3, SF3B1, EZH2 were the most common AddMUT.
Ten (29%) and 14 (41%) patients with CE developed AP (median time CE-AP of 3 years; range, 0.3-7) or remained in CP (median observation CE-last visit of 3.1 years (range, 0.1-12). One patient (SEQ1: MPL+IDH2+SRSF2) developed AML without CE. CP without CE (CP-CE; n 9):CP with CE (CP+CE; n 14):AP (n 11) at SEQ 1 differed significantly in the following: AP were the oldest (median 63:63:68 years); CP-CE had the highest platelets (median [x 109/L] 357:182:162), lowest LDH (median 849:1513:1460); lowest AddMUT (% -22:64:91); and CP+CE had the most frequent non-diploid karyotype (% - 11:50:35). CP-CE had the longest duration of RUX therapy (median 9:7:4 years). As CE, ASXL1, RAS, TET2, TP53 were found in CP and AP, ETNK1 and MPL only in CP and EZH2, RUNX1, CBL, NF1 only in AP. AP (vs CP+CE) had more of ≥3 CE (50% vs 36%) and more of consequent CE (60% vs 29%). Median JAK2/MPL/CALR VAF were similar between SEQ1-SEQ/last, SEQ1-SEQ/CE and SEQ1-SEQ/AP in all groups.
AddMUT/SEQ1 was significantly predictive of CE (p 0.02, 95% CI 1.3-36), AP (p 0.034, 95% CI 2.5-45) and OS (p 0.002, CI 2-27). OS [95% CI] for the entire cohort was 105 months [87-123]. OS of CP-CE (unreached) was superior to CP+CE and AP (both p < 0.01), it was comparable between CP+CE and AP (median 112 vs 96 months; p 0.33). AddMUT+/SEQ1, rather than CE, showed a distinct effect on OS, despite stratification per age >/<65 years at SEQ1: patients without AddMUT (n 13, all ages) vs <65 years +AddMUT (n 8) vs >65 years +AddMUT (n 13) had OS of 181 [102-NE], 104 [61-147] and 86 [67-105] months, respectively (each p < 0.02).
Conclusion: In our longitudinal genomic and clinical dataset of MF patients treated with JAK inhibitors over a median of 9 years, we noted 3 distinct trajectories distinguishing stable MF from progressive MF and MF transformed to AML. Possible roles of JAK inhibitors in late onset of CE and observed unexpectedly long OS continue to be investigated.
DISCLOSURES
Lucia Masarova, MD: Karyopharm - Honoraria; PharmaEssentia - Honoraria; GSK - Honoraria; Cogent - Honoraria; PharmaEssentia - Consultancy (Includes expert testimony); GSK - Consultancy (Includes expert testimony); MorphoSys - Membership on a Board/Advisory Committee