Clinical Features and Prognostic Impact of Splicing Factor Mutations and Their Co-Mutations in Acute Myeloid Leukemia

The clinical and prognostic significance of splicing factor (SF) mutations in newly diagnosed adult de novo acute myeloid leukemia (AML) remains incompletely defined. We retrospectively analyzed 323 newly diagnosed adult de novo AML patients (non-APL) treated between 2015 and 2023. Targeted next-generation sequencing assessed at least 32 recurrently mutated genes, with SF status defined by mutations in SF3B1, SRSF2, U2AF1, or ZRSR2. Clinical characteristics, co-mutation profiles, treatment regimens, treatment response, and survival outcomes were compared between SF-mutated (SFmut) and SF-wild-type (SFwt) groups. Kaplan-Meier and Cox proportional hazards models were used for survival analyses. SF mutations were identified in 58/323 patients (17.96%) and were largely mutually exclusive. Compared with SFwt, SFmut patients were more often male, older, and had a higher CD4/CD8 ratio; they also more frequently received low-intensity therapy. SFmut cases were enriched for co-mutations in ASXL1, RUNX1, CBL, IDH1, PHF6, SETBP1, BCORL1, and STAG2. The SFmut group had a lower composite complete remission rate (50.9% vs 71.3%, p = 0.003) and significantly shorter overall survival (median 6.73 vs 19.07 months, p < 0.001). In the overall multivariable model, SFmut, age ≥ 65 years, and WBC ≥ 100×109/L were independently associated with inferior OS, whereas intensive therapy was independently protective; intermediate/adverse cytogenetic risk also conferred stepwise worse OS. SFmut showed nonsignificant trends toward shorter OS in both treatment strata and was independently associated with poorer OS in the non-intensive, but not intensive, subgroup; however, the SF × treatment-intensity interaction was nonsignificant. Exploratory, unadjusted co-mutation analyses showed the most consistent adverse signals for SF/ASXL1 and SF/RUNX1, whereas other findings were heterogeneous or based on sparse subgroups; all should be considered hypothesis-generating and require validation.