Etiévant Lola, Gail Mitchell H
Division of Cancer Epidemiology and Genetics, Biostatistics Branch, National Cancer Institute, Rockville, MD, USA.
Int J Epidemiol. 2025 Feb 16;54(2). doi: 10.1093/ije/dyaf016.
The case-cohort design only requires covariate measurements for individuals experiencing the outcome of interest (cases) and individuals in a subcohort randomly selected from the cohort. Stratified subcohort sampling and calibration of the design weights increase efficiency of relative hazard and pure risk estimates, but require specifically adapted variance estimators. Yet, the 'robust' variance formula is often inappropriately used with stratified case-cohort data. Also, weight calibration and pure risk estimation are underused, possibly because of the lack of convenient software.
An influence-based method for inference of case-cohort Cox model relative hazards and pure risks is implemented in the CaseCohortCoxSurvival R package.
CaseCohortCoxSurvival allows estimation of parameter and variance of Cox model relative hazards and pure risks from case-cohort data. It can handle stratified subcohort sampling and calibrate the design weights. Both features are properly accounted for in the variance estimation.
CaseCohortCoxSurvival is available on the Comprehensive R Archive Network at [https://cran.r-project.org/package=CaseCohortCoxSurvival].
病例队列设计仅需要对经历感兴趣结局的个体（病例）以及从队列中随机选取的一个亚队列中的个体进行协变量测量。分层亚队列抽样和设计权重的校准可提高相对风险和纯风险估计的效率，但需要专门适配的方差估计量。然而，“稳健”方差公式在分层病例队列数据中常常使用不当。此外，权重校准和纯风险估计未得到充分利用，可能是因为缺乏便捷的软件。
基于影响的病例队列Cox模型相对风险和纯风险推断方法在CaseCohortCoxSurvival R包中得以实现。
CaseCohortCoxSurvival可根据病例队列数据估计Cox模型相对风险和纯风险的参数及方差。它能够处理分层亚队列抽样并校准设计权重。这两个特性在方差估计中均得到了恰当考虑。
CaseCohortCoxSurvival可在综合R存档网络上获取，网址为[https://cran.r-project.org/package=CaseCohortCoxSurvival]。
