Comparison of prognostic models for patients with diffuse large B-cell lymphoma in the rituximab era

Several revisions of International Prognostic Index (IPI) have been proposed for patients with diffuse large B-cell lymphoma (DLBCL) after the introduction of rituximab. Expanding evidence suggests that baseline absolute lymphocyte count (ALC) is also an independent factor for outcome prediction. We investigated the optimal prognostic model for these patients in the rituximab era. The study enrolled 274 consecutive patients with DLBCL receiving first-line cyclophosphamide, doxorubicin, vincristine, and prednisone based chemotherapy with rituximab between 2003 and 2009. Five factors within IPI and ALC were entered for Cox regression analysis. Overall survival (OS) and progression-free survival were calculated for different risk groups of models. Efficacy of models was compared by the value of Akaike information criterion (AIC). Revised IPI (R-IPI) and ALC/R-IPI, but not IPI, were informative to discriminate between different risk groups. In multivariate analysis for individual factors of the prognostic models, performance status >1 [odds ratio (OR) 3.59], Ann Arbor stage III or IV (OR 2.24), and ALC <1 × 109/L (OR, 2.75) remained significant. Another modified score based on the three factors divided patients into four risk groups and the 3-year OS rate was 93, 77, 39, and 13 %, respectively. By comparing AIC values in the Cox proportional hazards model, the modified three-factor model was the superior prognostic model followed by established ALC/R-IPI, R-IPI, and standard IPI. In conclusion, the addition of the novel factor, ALC, interacts with other established factors in outcome prediction for DLBCL. Development of a new score is needed for a better risk stratification in the rituximab era and would be helpful in the design of future clinical trials. The proposed three-factor model should be validated in large-scale studies.