Hepatocellular Carcinoma Prognosis Prediction based on Model Averaging Method
10.11783/j.issn.1002-3674.2025.03.010
- VernacularTitle:基于模型平均法的肝癌预后模型研究
- Author:
Yibo LUO
1
;
Nana HE
;
Jieyu HE
Author Information
1. 东南大学公共卫生学院流行病与卫生统计学系(210009)
- Publication Type:Journal Article
- Keywords:
Hepatocellular carcinoma;
Prognostic prediction model;
Frequentist model averaging;
Bayesian model av-eraging
- From:
Chinese Journal of Health Statistics
2025;42(3):369-377,381
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the integrating modeling strategy of prognosis prediction models,and provide support for the methodological selection of establishing effective clinical prediction models.Methods Based on the SEER liver cancer clinical follow-up data,the predictive performance and estimation accuracy of the classical Cox proportional hazard regression model,frequentist model averaging,and Bayesian model averaging methods were compared,and the applicability of model averaging methods was explored.Simulated studies were used to investigate the predictive performance of the models.Results Results of simulation analysis:the C index obtained by Bayesian model averaging was slightly higher than the Cox regression and frequentist model averaging.The 95%confidence interval of the C index tended to become narrower as the sample size increased.For variables with larger effect sizes,the Bayesian model averaging method obtained the smallest deviation of regression coefficients and the largest 95%interval coverage.Results of the case study:the validation set C index obtained by Cox regression,Bayesian model averaging and frequentist model averaging were 0.7845(95%CI:0.7613～0.8076),0.7851(95%CI:0.7619～0.8083)and 0.7845(95%CI:0.7613～0.8076),respectively.Conclusion Bayesian model averaging method can improve the ability of predicting prognosis when the sample size is small and the predictor variables are correlated.
