AUTHORS
Xinzhu Tang, N/A, Siyuan Chen2, Yuhui Chen3, Yutong Wang4, Lang Min5, Xinai Gan6, Laduona Wang7, Jiang Yu8, Zhenyuan Zhuang9, Shuxia Yang10, Wei Li Zhao III, Certified Doctor11, He LI12, Ailin Zhao, PhD13, Ting MD, PhD Niu, MD, PhD14
AFFILIATIONS
1 West China Hospital, Sichuan University, Chengdu, Sichuan, China
2 Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China
3 West China Hospital, Sichuan University, Chengdu, Sichuan, China
4 West China Hospital, Sichuan University, Chengdu, Sichuan, China
5 West China Hospital, Sichuan University, Chengdu, Sichuan, China
6 West China Hospital, Sichuan University, Chengdu, Sichuan, China
7 West China Hospital, Sichuan University, Chengdu, Sichuan, China
8 West China Hospital, Sichuan University, Chengdu, Sichuan, China
9 West China Hospital, Sichuan University, Chengdu, Sichuan, China
10 West China Hospital, Sichuan University, Chengdu, Sichuan, China
11 Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China
12 West China Hospital, Sichuan University, Chengdu, Sichuan, China
13 West China Hospital, Sichuan University, Chengdu, Sichuan, China
14 West China Hospital, Sichuan University, Chengdu, Sichuan, China
ABSTRACT
Objective:
EBV-positive DLBCL (EBV+ DLBCL) is an aggressive subtype with inferior outcomes, and few dedicated externally validated prognostic models exist. Plasma EBV-DNA may hold prognostic value but has not been integrated into a practical clinical tool. Given these considerations, we conducted a retrospective, multi-center study to evaluate the prognostic value of plasma EBV-DNA level in patients with EBV+ DLBCL. We aimed to determine an optimal cut-off value for plasma EBV-DNA, to develop a robust prognostic model incorporating plasma EBV-DNA, and to derive a simplified clinical scoring system for routine risk stratification.
Methods:
In this study, 107 patients with EBV+ DLBCL from West China Hospital (WCH) (training set, n=50) and Ruijin Hospital, Shanghai Jiao Tong University School of Medicine (RJH) (external validation set, n=57) were enrolled. ROC analysis was performed to determine the optimal cut-off for baseline plasma EBV-DNA. Independent prognostic factors were identified by multivariable Cox regression, from which the EBV-DNA/ECOG Cox model and the simplified EES (EBV-DNA and ECOG-PS score) was derived. Model performance was evaluated using C-index, time-dependent ROC curves, calibration plots, and Kaplan-Meier curves, and externally validated in the RJH cohort. The EES and the EBV-DNA/ECOG Cox model were further compared with existing prognostic tools, including IPI, NCCN-IPI, Oyama model and GELL score. A two-sided P < 0.05 was considered statistically significant. All statistical analyses were performed using Python.
Results:
After a median follow-up of 62.6 months (95% CI: 44.1-81.1) in the WCH cohort, the 1-year, 3-year, and 5-year OS rates were 71.7%, 65.0%, and 57.6%, respectively, and the corresponding PFS rates were 68.9%, 58.5%, and 51.4%. In the RJH cohort, after a median follow-up of 79.3 months (95% CI: 54.6-104.0), the 1-year, 3-year, and 5-year OS rates were 92.3%, 80.4%, and 74.4%, respectively, and the corresponding PFS rates were 77.4%, 61.2%, and 53.5%.
The optimal EBV-DNA cut-off was 9,285 copies/mL. Multivariable Cox regression identified ECOG-PS≥2 and log10(EBV-DNA) as independent prognostic factors for OS.
The EBV-DNA/ECOG Cox model demonstrated good discrimination in the WCH cohort (C-index: 0.835; 1-, 3-, and 5-year AUCs: 0.880, 0.807, and 0.857), with excellent calibration (E/O ratios: 1.039, 1.037, and 1.037) and significant risk stratification (P < 0.001). External validation in the RJH cohort confirmed acceptable discrimination (C-index: 0.702; AUCs: 0.810, 0.692, and 0.687), with E/O ratios of 1.887, 0.903, and 0.787 at 1, 3, and 5 years, though risk stratification didn’t reach statistical significance (P = 0.526).
The simplified EES assigned 1 point each for ECOG-PS≥2 and high EBV-DNA (dichotomized by the previously established cut-off), stratifying patients into low (0), intermediate (1), and high risk (2) groups. The EES demonstrated good discrimination in the WCH cohort (C-index: 0.761; 1-, 3-, and 5-year AUCs: 0.813, 0.815, and 0.806), with excellent calibration (E/O ratios: 1.030, 1.031, and 1.031) and significant risk stratification (P < 0.001). External validation in the RJH cohort confirmed its prognostic value (C-index: 0.694;1-, 3-, and 5-year AUCs: 0.813, 0.667, and 0.675), with E/O ratios of 1.851, 0.880, and 0.767, and significant discrimination across the three risk groups (P = 0.029).
Both the EBV-DNA/ECOG Cox model and the EES achieved the highest predictive accuracy during the first year, surpassing all other prognostic tools.
Conclusion:
The EES is a practical, externally validated tool for early risk stratification in EBV+ DLBCL, offering favorable early predictive performance with two readily available parameters.
DISCLOSURES
Xinzhu Tang, N/A: No relevant conflicts of interest to declare.