A novel prognostic model based on pretreatment serum albumin and ECOG PS for primary CNS lymphoma: an international, multi-center study

Background Serum albumin has been demonstrated as prognostic parameter in non-Hodgkin lymphoma (NHL). Primary central nervous system lymphoma (PCNSL) is a rare extranodal NHL with highly aggressive behavior. In this study, we aimed at creating a novel prognostic model for PCNSL based on serum albumin levels. Methods We compared several commonly used laboratory nutritional parameters for predicting the survival of PCNSL patients using overall survival (OS) for outcome analysis and receiver operating characteristic curve analysis to determine the optimal cut-off values. Parameters associated with OS were evaluated by univariate and multivariate analyses. Independent prognostic parameters for OS were selected for risk stratification, including albumin ≤ 4.1 g/dL, ECOG PS > 1, and LLR > 166.8, which were associated with shorter OS; albumin > 4.1 g/dL, ECOG PS 0–1 and LLR ≤ 166.8, which were associated with longer OS, and five-fold cross-validation was used for evaluating predictive accuracy of identified prognostic model. Results By univariate analysis, age, ECOG PS, MSKCC score, Lactate dehydrogenase-to-lymphocyte ratio (LLR), total protein, albumin, hemoglobin, and albumin to globulin ratio (AGR) resulted statistically associated with the OS of PCNSL. By multivariate analysis, albumin ≤ 4.1 g/dL, ECOG PS > 1, and LLR > 166.8 were confirmed to be significant predictors of inferior OS. We explored several PCNSL prognostic models based on albumin, ECOG PS and LLR with 1 point assigned to each parameter. Eventually, a novel and effective PCNSL prognostic model based on albumin and ECOG PS successfully classified patients into three risk groups with 5-year survival rates of 47.5%, 36.9%, and 11.9%, respectively. Conclusions The novel two-factor prognostic model based on albumin and ECOG PS we propose represents a simple but significant prognostic tool for assessing newly diagnosed patients with PCNSL.