Background: IDH2 gene mutations (mIDH2) occur in up to ̃20% of patients (pts) with acute myeloid leukemia (AML), most commonly as R140Q (in ̃75% of cases) or R172K (̃25%) point mutations. The functional effects and prognostic relevance of mIDH2-R140 and mIDH2-R172 can vary (Papaemmanuil 2016). In the randomized, phase 3 IDHENTIFY trial, enasidenib (ENA), an oral mIDH2 inhibitor, did not significantly improve overall survival (OS) vs conventional care regimens (CCR) as salvage treatment (Tx) for older pts with mIDH2 relapsed/refractory (R/R) AML in ITT analysis, but a trend for improved OS with ENA was detected in pts with IDH2-R172 mutations. We further investigated OS and correlative biomarkers in IDHENTIFY pt subgroups defined by mIDH2 variant (R140/R172). Methods: This open-label trial (NCT02577406) enrolled pts ≥ 60 years of age who had received 2 or 3 prior AML Tx. Pts were preselected to a CCR (azacitidine, intermediate- or low-dose Ara-C, or supportive care), and were then randomized 1:1 to ENA 100 mg/d or CCR in 28d cycles. Co-occurring gene mutations were identified by targeted next-generation sequencing (37-gene panel) of bone marrow mononuclear cell (BMMC) DNA. Total 2-HG levels were determined by LC/MS. Results: Of 319 pts enrolled, 88 pts (28%; 43 ENA, 45 CCR) had mIDH2-R172 and 229 (72%; 115 ENA, 114 CCR) had mIDH2-R140. Median baseline (BL) 2-HG levels were similar between Tx arms and mIDH2 variant subgroups, as were IDH2 variant allele frequencies. Pts with mIDH2-R172 had fewer median BL co-mutations (4 [range 2–8]) than did pts with mIDH2-R140 (5 [1–11]) (P< 0.0001). The most frequently co-occurring mutations were SRSF2 and RUNX1 in the R140 cohort (59% each) and DNMT3A in the R172 cohort (57%). Compared with the R172 cohort, the R140 group was enriched with SRSF2, FLT3 (-ITD/-TKD), NPM1, RUNX1, and JAK2 mutations, whereas DNMT3A and TP53 mutations were more common in the R172 group. In Cox multivariate analysis including mIDH2 variant (R140/R172), DNMT3A mutation status, and number of gene mutations at BL, mIDH2-R172 was significantly (P = 0.04) correlated with improved OS (vs. R140) in the ENA arm, whereas the number of BL gene mutations was significantly (P< 0.01) associated with OS in the CCR arm. Median OS in the R172 subgroup was 14.6 mo with ENA vs 7.8 mo with CCR (HR, 0.59 [95%CI 0.35-0.98]; P = 0.039) and 1-yr survival rates were 62% and 30%, respectively. In mIDH2-R140 pts, median OS was 5.7 mo in both Tx arms (0.93 [0.70-1.24]; P = 0.61), and 1-year survival rates were 29% and 25% with ENA and CCR, respectively. Conclusions: Mutational burden and co-mutational profiles differed between pts with mIDH2-R140 and mIDH2-R172 R/R AML. ENA improved survival outcomes for pts with IDH2-R172 mutations, with median OS and 1-year survival rate approximately double those in the CCR arm. Clinical trial information: NCT02577406.