European LeukemiaNet (ELN) acute myeloid leukemia (AML) genetic risk classification systems were based on response to intensive chemotherapy; their ability to discriminate outcomes in older patients treated with venetoclax-azacitidine may be suboptimal. Here, pooled analysis of patients in the phase 3 VIALE-A trial (NCT02993523) and phase 1b study (NCT02203773) examined prognostic stratification according to 2017 and 2022 ELN risk classifications. A bioinformatic algorithm derived new molecular signatures differentiating venetoclax-azacitidine-treated patients based on median overall survival (OS). 279 patients treated with venetoclax-azacitidine and 113 patients treated with placebo-azacitidine were analyzed. When classified by ELN 2017 or 2022 prognostic criteria, most patients had adverse-risk AML (60.2% and 72.8% for venetoclax-azacitidine and 65.5% and 75.2% for placebo-azacitidine, respectively). While outcomes with venetoclax-azacitidine were improved across all ELN risk groups compared with placebo-azacitidine, ELN classification systems poorly discriminated venetoclax-azacitidine outcomes. By applying a bioinformatic algorithm, new molecular signatures were derived differentiating OS outcomes with venetoclax-azacitidine; the mutational status of TP53, FLT3-ITD, NRAS, and KRAS categorized patients into higher-, intermediate-, and lower-benefit groups (52%, 25%, and 23% of patients, respectively), each associated with a distinct median OS (26.5 months [95% CI, 20.2 to 32.7], 12.1 months [95% CI, 7.3 to 15.2], and 5.5 months [95% CI, 2.8 to 7.6], respectively). ELN prognostic classifiers do not provide clinically meaningful risk stratification of OS outcomes for patients with AML treated with venetoclax-azacitidine. TP53, FLT3-ITD, NRAS, and KRAS mutation status allows classification of these patients into three risk groups with distinct differences in median OS.
