Enhancement of the International prognostic index with β2-microglobulin, platelet count and red blood cell distribution width: a new prognostic model for diffuse large B-cell lymphoma in the rituximab era
Haizhu Chen, Qiaofeng Zhong, Yu Zhou, Yan Qin, Jianliang Yang, Peng Liu, Xiaohui He, Shengyu Zhou, Changgong Zhang, Lin Gui, Sheng Yang, Liqiang Zhou, Yuankai Shi
Abstract
BACKGROUND: This study aimed to propose a new user-friendly, cost effective and robust risk model to facilitate risk stratification for diffuse large B-cell lymphoma (DLBCL) treated with frontline R-CHOP regimens. METHODS: Data on 998 patients with de novo DLBCL diagnosed between Jan 1st, 2005 and Dec 31st, 2018 at our center, who received frontline R-CHOP or R-CHOP-like regimens, were retrospectively collected. Patients were randomly divided into the training cohort (n = 701) and the validation cohort (n = 297). A new prognostic model for overall survival (OS) was built based on the training cohort. The performance of the new model was compared with International prognostic index (IPI), revised IPI (R-IPI) and National Comprehensive Cancer Network (NCCN)-IPI (NCCN-IPI). The new model was validated in the validation cohort. RESULTS: The multivariate analysis of the training cohort showed that the IPI, β2-microglobulin, platelet count and red blood cell distribution width were independent factors for OS, which were incorporated into the new prognostic model. Patients were stratified into low risk, low-intermediate risk, high-intermediate risk, high risk and very high risk groups, with distinct survival outcomes. The new model achieved good C-indexes for 5-year OS prediction of 0.750 (95%CI 0.719-0.781) and 0.733 (95%CI 0.682-0.784) in the training and validation cohorts, respectively, and displayed well-fitted calibration curves. The C-index and the time-dependent ROC analysis demonstrated better performance of the new model than the IPI, R-IPI and NCCN-IPI in both training and validation cohorts. The integrated Brier score for predicting 5-year OS of the new model was lower than that of the IPI, R-IPI and NCCN-IPI in both cohorts, and decision curve analysis also showed a higher net benefit, indicating the superiority of the new model over the conventional models. CONCLUSION: The new prognostic model might be a useful predictive tool for DLBCL treated with R-CHOP regimens. Further external validation is warranted.