I am trying to run mediation analysis to see if one continuous biological variable mediates the relationship between another continuous biological variable and a time-to-event survival outcome. I am using the mediate function in the mediation R package, the output of which gives "control" and "treated" ACME/ADE/Proportion Mediated as well as "average" ACME/ADE/Proportion Mediated.
I have looked through vignettes and documentation for this package, which say that the user can choose values for "treated" and "control" groups in the case of continuous predictors. However, this is not relevant to my use case: both the predictor and mediator are levels of different circulating proteins, both 'high' and 'low' levels of which may be indicative of a diseased state or some kind of dysregulated process. There is no real way to say that one value is "control" and one value is "treated," especially since "typical" values of these proteins may vary greatly with demographic factors (e.g. age, sex, etc.).
This being the case, can I safely ignore the "control" and "treated" output values from this package and just take the "average" ACME/ADE/Proportion Mediated as the overall effect? Or is this inappropriate? If it is inappropriate, is there a better package to use that can work with survival data? Some collaborators have suggested lavaan, but it is my understanding that mediation analysis in the lavaan package cannot take time-to-event data.