Abstract
BACKGROUND: Although diffuse large B-cell lymphoma (DLBCL) has entered the era of molecular subtyping, gene mutations have not yet been used for risk stratification of patients with DLBCL. Therefore, there is an urgent need for a molecular-based model to comprehensively evaluate the prognosis of DLBCL patients in clinical practice. METHODS: We retrospectively analyzed 761 patients with DLBCL from cBioPortal datasets. Gene mutations and the International Prognostic Index (IPI) were analyzed for their associations with overall survival and were used to determine the set of prognostic variables. A Cox model was utilized to estimate the relative weights of the selected variables. RESULTS: We mapped at least one gene mutation in 94% (718/761) of patients with DLBCL. Survival analysis identified that PIM1, KLHL14, CD79b, and MYC mutations were correlated with adverse outcomes, whereas EZH2, STAT3, and B2M mutations were predictors of favorable outcomes. Using these gene mutations together with IPI, we established a novel prognostic system, IPI-M, which resulted in a unique risk score for individual patients. We further derived five IPI-M risk categories (very low, low, intermediate, high, and very high), demonstrating stronger prognostic separation in DLBCL. Compared with IPI, the discrimination ability of IPI-M showed a better discrimination of survival with concordance index value and higher 1-year, 3-year, and 5-year survival as measured by area under the receiver operating characteristic curve values. Finally, IPI-M model also presented a prognostic separation for overall survival especially for the discrimination of high/very high patients from intermediate/low/very low patients in an external validation cohort. CONCLUSIONS: We developed a novel, feasible integrated prognostic system for newly diagnosed DLBCL that incorporates the IPI and genetic mutations, which facilitates precise prognostic assessment in clinical practice and supports precision therapy. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s12672-026-04817-4.