Prediction Model and Its Value of IrAEs Based on Peripheral Blood Markers
10.12259/j.issn.2095-610X.S20250408
- VernacularTitle:基于外周血标志物初步探讨irAEs预测模型及价值
- Author:
Jun DENG
1
;
Jun WANG
;
Xi WANG
;
Change GAO
;
Xiao CHEN
;
Mingxia SHI
Author Information
1. 昆明医科大学第一附属医院肿瘤内科,云南 昆明 650032
- Keywords:
ICIs;
irAEs;
Predictors;
Predictive model
- From:
Journal of Kunming Medical University
2025;46(4):57-66
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the predictive model and its value of irAEs based on peripheral blood markers.Methods The baseline clinical data,laboratory tests,and irAEs follow-up results of 825 malignant tumor patients treated with PD-1/PD-L1 antibodies in the First Affiliated Hospital of Kunming Medical University were retrospectively collected from December 2020 to December 2023.The patients were divided into irAEs group and non-irAEs group according to the presence or absence of irAEs.The differences between and within groups were analyzed by t-test,rank-sum test,chi-square test and Fisher exact probability method.LASSO,Ridge and Elastic-net logistic regressions were used to screen the predictors and establish the risk prediction models for irAEs.Results 136 patients experienced 178 irAEs,of which endocrine toxicity accounted for 42.64％,hepatitis 35.29％,pneumonia 20.58％,grade≥G3 accounted for 19.07％,involving more than two organs accounted for 24.26％of the total number of irAEs.Univariate analysis showed that baseline CD4+T cell count,IL-6,IL-17,TSH,GLB and ALB were associated with irAEs.GLB,ALB,IL-17 and TSH were selected as the important risk factors by Ridge,LASSO and Elastic-Net logistic regression.The results showed that the AUC of the three algorithms were over 0.800.The AUC of internal validation set by LASSO-Logistic was 0.800(95％CI 0.739～0.862).The AUC of external validation set was 0.800(95％CI 0.739～0.861)and the DCA curve results indicated the highest net return for this predictive model.Conclusion GLB,ALB,IL-17 and TSH are independent predictors of irAEs,and the predictive model of irAEs based on them is effective.
