IntroductionAcute myeloid leukemia (AML) is a genetically heterogeneous disease. A recent study (NEJM, 2016) classified 1540 patients into 14 subgroups using mutation information from targeted next generation sequencing data as well as cytogenetic information [1]. The classification criteria of 7 of these subgroups rely solely on mutation information. NK-AML is characterized by its lack of cytogenetic abnormalities. In this study, we attempted to replicate the prognostic stratification in an independent set of NK-AML patients using the NEJM study's genomic classification criteria.Patients and MethodsThis study included a total of 393 patients who met the following eligibility criteria: 1) age ≥ 15 years; 2) a diagnosis of NK-AML confirmed by conventional cytogenetic analysis; 3) treatment with induction chemotherapy using a standard protocol (a 3-day course of anthracycline with a 7-day course of cytosine arabinoside). The median follow-up duration was 55.1 months (range, 0.7-182.9). Analysis of genetic mutations were performed using targeted sequencing by Illumina Hiseq 2000 (Agilent custom probe set targeting entire exon regions of a myeloid panel consisting of 94 genes).ResultsWe identified driver mutations across 28 genes or genomic regions, with 2 or more driver mutations identified in 15/393 patients (3.8%). Based on the genomic classification criteria, the patients were classified as follows: 136 patients (34.6%) with NPM1 mutations, 42 patients (10.7%) with mutated chromatin modifiers and/or RNA-splicing genes, 6 patients (1.5%) with TP53 mutations, 40 patients (10.2%) with biallelic CEBPA mutations, 8 patients (2.0%) with IDH2-R172 mutations and no other class-defining lesions, 108 patients (27.5%) with driver mutations but no detected class-defining lesions, 38 patients (9.7%) with no detected driver mutations, and 15 patients (3.8%) who met the criteria of more than one genomic subgroup.Of the 393 patients, 325 patients (82.7%) achieved complete remission (CR). CR rates vary depending on the genomic subgroup (75.9%-97.4%). The CR rate for each subgroup was as follows: •86.8% (118/136) of patients with NPM1 mutations•61.9% (26/42) of patients with mutated chromatin and/or RNA-splicing genes•83.3% (5/6) of patients with TP53 mutations•97.5% (38/40) of patients with biallelic CEBPA mutations•87.5% (7/8) of patients with IDH2-R172 mutations and no other class-defining lesions•75.9% (82/108) of patients with driver mutations but no detected class-defining lesions•97.3% (37/38) of patients with no detected driver mutations•80.0% (12/15) of patients meeting criteria of more than one subgroup5-year OS and 5-year relapse incidence (RI) for each subgroup was as follows: •49.3% (95% CI, 40.1-58.5) and 39.8% (95% CI, 30.1-49.2) of patients with NPM1 mutations•11.6% (95% CI, 1.4-21.8) and 71.4% (95% CI, 45.7-86.5) of patients with mutated chromatin and/or RNA-splicing genes•50.0% (95% CI, 10.0-90.0) and 20.0% (95% CI, 0.4-61.2) of patients with TP53 mutations•68.3% (95% CI, 53.4-83.2) and 19.7% (95% CI, 8.5-34.4) of patients with biallelic CEBPA mutations•56.3% (95% CI, 17.3-95.3) and 21.4% (95% CI, 0.3-67.3) of patients with IDH2-R172 mutations and no other class-defining lesions•26.6% (95% CI, 17.4-35.8) and 53.2% (95% CI, 40.7-64.3) of patients with driver mutations but no detected class-defining lesions•29.1% (95% CI, 14.2-44.0) and 43.8% (95% CI, 27.1-59.3) of patients with no detected driver mutations•40.0% (95% CI, 15.3-64.7) and 33.3% (95% CI, 9.2-60.3) of patients that meet the criteria of more than one subgroup.The CR rates of the subgroup with mutated chromatin and/or RNA-splicing genes was significantly lower than the rest of the cohort (61.9% vs. 85.2%, p=0.00016). The 5-year OS and 5-year RI of the subgroup were also poorer than the others [61.9% vs. 85.2% in OS (p=0.00016), 71.4% vs. 40.1% in RI (p < 0.0001)].ConclusionOur NK-AML cohort showed similar survival patterns to the cohort in Papaemmanuil et al (NEJM 2016). The subgroup in AML with mutated chromatin and/or RNA-Splicing genes had the poorest prognosis with respect to CR rate and overall survival. This analysis replicates the result of recently published genomic classification and supports its use for categorizing NK-AML patients.Reference[1] Genomic Classification and Prognosis in Acute Myeloid Leukemia. Papaemmanuil E et al. N Engl J Med, 2016 vol. 374 (23) pp. 2209-2221. [Display omitted] DisclosuresJang:Kyowa Hakko Kirin Co., Ltd.: Research Funding.