AUTHOR=Wu Zhentian , Wang Chenyi , Lyu Yao , Lin Zheshen , Lu Ming , Wang Shixiong , Wang Bingxuan , Yang Na , Li Yeye , Wang Jianhong , Duan Xiaohui , Zhang Na , Gao Jing , Zhang Yuan , Hao Miaowang , Wang Zhe , Gao Guangxun , Liang Rong TITLE=A novel inflammation-related prognostic model for predicting the overall survival of primary central nervous system lymphoma: A real-world data analysis JOURNAL=Frontiers in Oncology VOLUME=13 YEAR=2023 URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2023.1104425 DOI=10.3389/fonc.2023.1104425 ISSN=2234-943X ABSTRACT=Background

Primary central nervous system lymphoma (PCNSL) is a type of extranodal non-Hodgkin lymphoma. Although there are widely used prognostic scores, their accuracy and practicality are insufficient. Thus, a novel prognostic prediction model was developed for risk stratification of PCNSL patients in our research.

Methods

We retrospectively collected 122 patients with PCNSL from two medical centers in China from January 2010 to June 2022. Among them, 72 patients were used as the development cohort to construct a new model, and 50 patients were used for the validation. Then, by using univariate and multivariate Cox regression analsis and Lasso analysis, the Xijing model was developed and composed of four variables, including lesion number, β2-microglobulin (β2-MG), systemic inflammation response index (SIRI) and Karnofsky performance status (KPS). Finally, we evaluated the Xijing model through internal and external validation.

Results

Compared with the original prognostic scores, the Xijing model has an overall improvement in predicting the prognosis of PCNSL according to the time-dependent area under the curve (AUC), Harrell’s concordance index (C-index), decision curve analysis (DCA), integrated discrimination improvement (IDI) and continuous net reclassification index (NRI). For overall survival (OS) and progression-free survival (PFS), the Xijing model can divide PCNSL patients into three groups, and shows more accurate stratification ability. In addition, the Xijing model can still stratify and predict prognosis similarly better in the elderly with PCNSL and subgroups received high-dose methotrexate (HD-MTX) or Bruton’s tyrosine kinase inhibitors (BTKi). Finally, external validation confirmed the above results.

Conclusions

Integrating four prognostic factors, including imaging findings, tumor burden, systemic inflammation response index, and comprehensive physical condition, we provided a novel prognostic model for PCNSL based on real-world data and evaluated its predictive capacity.