Cox proportional hazards regression model (Cox model) is the most commonly used multivariate approach in time-to-event data analysis. A vital issue in fitting Cox model is choosing the appropriate time scale related to the occurrence of the outcome events. However, few domestic studies have focused on selecting and applying time scales for Cox model in the analysis of cohort study data. This study briefly introduced and compared several time scales in the reports from literature; and used data from the Shanghai Women's Health Study to illustrate the impact of different time scales on data analysis results, using the association between central obesity and the risk of liver cancer as an example. On this basis, several suggestions on selecting time scales in Cox model are proposed to provide a reference for the analysis of cohort study data.
Cox比例风险回归模型（Cox模型）是时间-事件数据分析中常用的多因素分析方法，拟合Cox模型时一个关键问题是如何选择合适的与结局事件发生相关的时间尺度。目前国内开展的队列研究在资料分析中较少关注Cox模型的时间尺度选择问题。本研究对文献报道中常见的几种时间尺度选择策略进行简要介绍和比较；并利用上海女性健康队列资料，以中心性肥胖与肝癌发病风险的关联为例，说明选择不同时间尺度的Cox模型对数据分析结果的影响；在此基础上提出几点Cox模型时间尺度选择上的建议，以期为队列研究资料的分析提供参考。.