Somatic mutations at diagnosis in patients with chronic Phase CML receiving frontline imatinib are associated with a higher rate of treatment failure: First analysis from the international CML foundation (iCMLf) genomics alliance on the harmony platform

Abstract

Abstract Background Evidence has accumulated that somatic mutations in myeloid genes other than BCR::ABL1 at CML diagnosis are associated with treatment failure. Genomic profiling in chronic phase diagnosis samples of unselected cohorts revealed mutations in 17-24% of patients. At diagnosis, ASXL1 is the most frequently mutated gene at 7.3-9%. Genomic profiling using next-generation sequencing is now standard practice for many hematologic cancers where pathogenic variants define molecular classes that stratify risk and assign therapy. However, the genomic profile at diagnosis is not integrated into risk stratification for CML and data is lacking to guide management decisions. Combining many datasets derived from different centers through the iCMLf Genomics Alliance and the HARMONY Alliance Foundation Platform could facilitate the incorporation of genomics-based algorithms into the clinical management of CML based on strong clinical evidence. Objective To pool global genomic data to assess the clinical relevance of pathogenic variants at diagnosis of chronic phase CML. Methods The HARMONY Platform is a central repository of anonymous data from multiple contributors. CML patient whole exome or targeted gene panel sequencing data was submitted to the HARMONY Platform from 12 centers: Europe 6; Middle East 2; Australia 1; Taiwan 1; Singapore 1; and Canada 1. This first analysis was performed for the largest cohort on the platform of 468 frontline imatinib-treated patients at chronic phase diagnosis. The data comprised a mix of consecutively-treated patients at diagnosis or patients retrospectively selected for sequencing based on treatment failure. Kaplan–Meier survival curves were generated to estimate time-to-event outcomes, and the log-rank test was used to assess differences in survival distributions between groups. Five-year survival probabilities are reported to provide time-specific context. Progression was defined as blast phase, accelerated phase or death. Failure was defined as per the ELN 2020 criteria. Survival outcomes were considered from the start date of imatinib to the date of the first event under investigation or death, whichever came first. Probabilities of acquiring BCR::ABL1 kinase domain mutations were calculated using Gray's test with death as the competing risk. Level of significance was 0.05. Results At diagnosis, somatic variants in myeloid genes that met strict criteria for pathogenicity were detected in 96/468 imatinib-treated patients, 20.5%. ASXL1 was the most frequently mutated gene: 53/468 patients, 11.3%. Of 20 mutated genes, only 3 others were mutated at a frequency of >1.5%: TET2, DNMT3A and RUNX1, 2.6%, 2.4% and 1.7%, respectively. The median follow-up time was 4.2 years, range 0.04-20.8. Comparing outcomes for patients with and without pathogenic variants at diagnosis, there was no difference in overall survival. Five-year progression-free survival (PFS) for patients with pathogenic variants at diagnosis compared to patients with no pathogenic variants was 76.3% vs 83.1% (overall PFS comparison P=0.041). Five-year failure-free survival (FFS) for patients with pathogenic variants compared to patients with no pathogenic variants was 43.8% vs 65.2% (overall comparison P <0.0001). BCR::ABL1 kinase domain mutation analysis was performed after diagnosis for 432/468 patients. The 5-year probability of acquiring BCR::ABL1 mutations for patients with pathogenic variants at diagnosis compared to patients with no pathogenic variants was 20.0% vs 11.4% (overall comparison P=0.067, Gray test). Since the frequency of mutated ASXL1 at diagnosis was higher than other genes, the association with outcome was assessed. There was no difference in overall survival or PFS. Patients with mutated ASXL1 had the lowest 5-year FFS compared to patients with other pathogenic variants and those with no variants: 41.0% vs 49.3% vs 65.2% (overall comparison P value=0.0005). The overall FFS comparison P values for mutated ASXL1 vs no variants was 0.0012 and other pathogenic variants vs no variants 0.007. Conclusion This is the largest genomic analysis of frontline imatinib-treated patients that has been undertaken. Pathogenic variants at diagnosis were associated with lower PFS and FFS. This first report from the global alliance awaits validation but provides evidence-based insights that could enhance patient management by refining risk stratification through the integration of large-scale clinical and genomic data.