基于半参数转移模型的复发事件与终止事件内积混合终点预测研究

Prediction Study on Inner Product Composite Endpoint of Recurrent and Terminal Events Based on Semiparametric Transition Models

摘要: 针对生物医学领域中复发事件与终止事件共存的删失数据建模难题，本文提出两类半参数模型：一是复发事件与终止事件内积混合终点的半参数混合模型，无需指定事件间相依关系，可根据事件严重程度赋予权重；二是带时变回归系数的半参数转移模型，刻画协变量对复合事件的动态影响。通过数值模拟验证模型估计的有效性，利用模拟的慢性心力衰竭和膀胱癌患者数据进行实证分析，并开展敏感性分析探究模型稳健性。结果表明，所提模型在估计精度和灵活性上优于传统Cox比例风险模型，可为医学治疗方案评估、患者生存状况预测提供科学依据。

Abstract: To address the challenge of modeling censored data with coexisting recurrent and terminal events in the biomedical field, this paper proposes two semiparametric models: First, a semiparametric mixture model with an inner product composite endpoint for recurrent and terminal events, which dispenses with the need to specify the dependence between events and allows weights to be assigned based on event severity. Second, a semiparametric transition model with time-varying regression coefficients, which characterizes the dynamic impact of covariates on composite events. The effectiveness of model estimation is verified via numerical simulations. Empirical analysis is conducted using simulated data from patients with chronic heart failure and bladder cancer, and sensitivity analysis is performed to explore the robustness of the models. The results indicate that the proposed models outperform the traditional Cox proportional hazards model in terms of estimation accuracy and flexibility, providing a scientific basis for the evaluation of medical treatment regimens and the prediction of patients’ survival status.