High-dimensional mediation analysis in survival models

Abstract

Mediation analysis with high-dimensional DNA methylation markers is important in identifying epigenetic pathways between environmental exposures and health outcomes. There have been some methodology developments of mediation analysis with high-dimensional mediators. However, high-dimensional mediation analysis methods for time-to-event outcome data are still yet to be developed. To address these challenges, we propose a new high-dimensional mediation analysis procedure for survival models by incorporating sure independent screening and minimax concave penalty techniques for variable selection, with the Sobel and the joint method for significance test of indirect effect. The simulation studies show good performance in identifying correct biomarkers, false discovery rate control, and minimum estimation bias of the proposed procedure. We also apply this approach to study the causal pathway from smoking to overall survival among lung cancer patients potentially mediated by 365,307 DNA methylations in the TCGA lung cancer cohort. Mediation analysis using a Cox proportional hazards model estimates that patients who have serious smoking history increase the risk of lung cancer through methylation markers including cg21926276,

High-Dimensional Mediation Analysis With Confounders in Survival Models

“…We will compare this with the method of adjusting all confounders as covariates and the method of ignoring confounders. The goal of variable selection is to identify S = k : α k β k = 0 , which are the significant mediators between the exposure and the outcome when the number of potential mediators p is much larger than the sample size n, and the traditional statistics methods for Cox regression analysis fail to work (Luo et al, 2020). Besides, there are confounders influence the relationship of exposure, mediators, and outcome.…”

Statistical methods for mediation analysis in the era of high-throughput genomics: Current successes and future challenges

“… Note: we focus primarily on methodological papers and have not listed applied work that employs mediation methods similar to these listed above (e.g., Wu et al (2018) [166] , Luo et al (2020) [96] ). In addition, we only focus on methods that aim to detect active mediators and have not listed mediation methods for effect estimation and decomposition (e.g., VanderWeele and Vansteelandt (2014) [77] , Daniel et al (2015) [121] , Huang and Yang (2017) [92] , Steen et al (2017) [167] , Taguri et al (2018) [168] , Zhou et al (2020) [115] , and Zhao et al (2020) [114] .…”

“…Finally, HIMA evaluates the significance of the mediation effect using P max, s = max( , P as ) and adjusts for multiple comparisons through Bonferroni correction. HIMA has been recently extended to survival outcomes [96] and to yield unbiased mediator on outcome effects with debiased Lasso [159] .

Fig.

…”

“…Detailed statistical methodology for mediation analysis is generally constructed under the counterfactual framework, which is also known as the potential outcome framework or Rubin’s model [64] , [65] , [86] , [87] , [88] , [89] , [90] , [91] , [92] , developed in the field of causal statistical inference. The counterfactual framework facilitates methodological establishment for mediation analysis to accommodate different outcome types that include continuous [64] , [65] , [69] , [70] , binary [76] , [88] , [93] and survival outcomes with censoring [74] , [80] , [92] , [94] , [95] , [96] , as well as to account for possible interactions between exposure and mediator [64] , [88] . Under the counterfactual framework, mediation analysis makes further modeling assumptions to effectively represent the relationship among the exposure, mediator, and outcome through a directed acyclic graph that links the exposure to the outcome through the mediator (see below) [97] , [98] .…”

“…However, existing univariate and multivariate mediation methods do not explicitly model correlation among potential mediators. Altogether, the above new challenges brought by high-throughput genomics have motivated the intense recent development of high-dimensional mediation methods that aim to accommodate a large number of potentially correlated mediators [78] , [79] , [81] , [82] , [96] , [109] , [111] , [112] , [113] , [114] , [115] .…”