[Introduction of landmarking approach and its application in dynamic prediction].

Conventional prediction model, as a static prediction model, can be only used to predict the probability of the occurrence of an event during the observation period using the information available at baseline survey. However, based on current clinical demands, dynamic prediction, which obtains prediction probabilities for both baseline survey and later time points given the history of the events and covariates up to that time, is gaining a growing attention. As a dynamic prediction model, the landmarking approach is simple, easy to use, computationally efficient and has a comparable performance of joint modeling, which makes it to be widely used in recent researches. Because of its limited application in China, this paper makes a brief introduction of its ideas and basic application to further promote its applications in clinical dynamic prediction.传统的生存预测属于“静态预测”，即其只能利用基线暴露信息预测结局的发生情况。而目前出于临床研究实际需求，越来越多的研究者开始关注“动态预测”，该方法能利用截至预测起点（界标时点）的暴露信息对预测模型进行更新，从而实现对患者结局的动态预测。界标模型即一种动态预测模型，其结构简单、容易实现、计算效率高，且其性能与相较之下更为复杂的联合模型相近，目前已得到广泛应用。但其在国内的应用仍较少，本文从界标模型的基本思想出发，通过案例简要介绍其应用方法，以期促进其在临床生存分析动态预测场景中的应用。.