DNMT3A Mutations Do Not Affect Treatment Response or Outcomes for Patients with Polycythemia Vera Treated with Interferon Alpha

Two recent papers indicate that DNMT3A mutations should be considered when recommending interferon-alpha (IFN) for patients (pts) with polycythemia vera (PV). Usart et al. (Blood 2024) showed that deletion of DNMT3A promotes JAK2V617F (JAK2) hematopoietic stem cell (HSC) self-renewal and resistance to IFN in mice. Similarly, Knudsen et al. (Blood Adv 2022) identified treatment-emergent DNMT3A mutations (DNMT3A+) in patients with PV treated with IFN (IFN-PV) who did not attain a complete hematologic response (CR). The mutation-enhanced international prognostic system did not associate DNMT3A+ with adverse risk in PV, suggesting an IFN-specific effect (Tefferi et al. Br J Haematol 2020). Given these concerning findings and IFN's role as a first-line treatment (Rx) in PV associated with myelofibrosis-free, event-free, and overall survival (MFS/EFS/OS), it is critical to determine if DNMT3A+ affects IFN-PV outcomes. To address this potentially practice-changing issue, we conducted a retrospective study of long-term outcomes in 246 pts with PV who had next generation sequencing (NGS) data available in the largest single-center database of IFN-PV. Data were collected from 437 NGS reports and electronic health records. Fisher's test was used to compare frequencies, t-test for means, Kaplan-Meier and log-rank test for MFS/EFS/OS. Median follow-up was 12 years (yr), with 912 pts-yr of IFN Rx (143 IFN-PV pts with 4 years median Rx time). DNMT3A+ was not more common in IFN-PV than in non-IFN pts (18/143 (13%) vs 14/103 (14%), p=0.8). All DNMT3A+ mutations were known/likely pathogenic, and the median initial VAF was 14% (range 1.1-50%). Between DNMT3A+ (n=32) and DNMT3A- (n=214) pts, there were no significant differences in age at Dx (median 55 yr vs 56 yr, p>0.9), sex (females 59% vs 46%, p=0.2), race (90% vs 89% identified as White, p>0.9), thrombosis history (19% vs 16%, p=0.7), JAK2 variant allele frequency (VAF) (median 32% vs 36%, p=0.9), time on interferon (median 3.4 vs 3.8 yr, p=0.9), or median interferon dose. The same was true for the subset of IFN-DNMT3A+ (n=18) compared to IFN-DNMT3A- (n=125). The percentage of patients achieving CR-defined as 3 or more months of phlebotomy-free hematocrit<45%, platelets <450 x 103/mL, and white blood count <11.5 x 103/mL -was similar for IFN-DNMT3A+ and IFN-DNMT3A- pts (94% and 93%, p>0.9) with median time to CR of 0.8 yr and 0.6 yr, respectively. Molecular response (MR), defined by a relative decrease in JAK2 >20% from baseline, was attained in 5 of 13 (38%) IFN-DNMT3A+ pts and 34 of 79 (43%) of IFN-DNMT3A- pts with serial VAFs (Fisher's p>0.9). Serial NGS testing for DNMT3A >1 year apart was available for 9 IFN-DNMT3A+ pts and none had an increasing DNMT3A VAF during Rx (6 stable and 3 with DNMT3A MR > 20% from baseline). In a Cox proportional hazards multivariable analysis considering age and IFN, DNMT3A status showed no significant association with MFS/EFS/OS (HR = 0.45/0.69/0.89, p=0.1/0.2/0.8 respectively). The 20-yr EFS (composite of thrombosis, myelofibrosis, leukemia and death) was favorable in IFN-DNMT3A+ at 73%, and not significantly different from IFN-DNMT3A- (44%, p=0.093). In conclusion, despite preclinical evidence that DNMT3A loss can promote JAK2-HSC self-renewal and IFN resistance, we found no influence of DNMT3A+ on clinical response or MFS/EFS/OS outcomes in IFN-treated pts with PV. Therefore, DNMT3A+ should not influence the decision to use IFN in pts with PV.