Causal mediation analysis on survival data

[Giuseppe Criaco]

Dear members,

I would like to perform causal mediation analysis on survival data in Stata. Initially, I would run that on a discrete-time survival models (xtcloglog specification). However, I would also try to run it on a Cox model.

I have tried to see whether any command has been developed in Stata but I could not find any. The closest I could find are:

1) medeff (which apparently does not take into account the longitudinal nature of the data)

2) binarymediation (which apparently has the same limitation as 1)

I would thus be interested to know whether I could follow some existing command/procedure for Stata which deals specifically with causal mediation analysis on survival data .

Any help would be much appreciated.

Thanks,

Giuseppe

[Dick Campbell]

I would love to stand corrected, but I don't think there is a Stata package that does what you want. There is a paper in Epidemiology that propose a method for doing mediation in a survival framework by Lange and Hansen. "Direct and indirect effects in a survival context" Epidemiology. 2011 Jul;22(4):575-81. doi: 10.1097/ but to the best of my knowledge that method has not been programmed in Stata. In the same issue of the journal there is a nice summary piece by Van der Weele that you might check.

[Giuseppe Criaco]

Hi Dick,

thanks a lot for your quick answer. Those are exactly the papers I was referring to. I know someone has implemted a command/procedure for R (http://sfi.nr.no/sfi/index.php/How_t..._survival_data) and I was checking whether a similar command/procedure existed for Stata.

Let's hope so!

[Quint Koevoets]

Hi Giuseppe,

I would like to know if your analysis has worked out well for you? I am interested in how you have managed to perform a caussal mediation analysis in Stata with longitudinal data. Since for my analysis I would like to use the medeff command on a panel data set. I am looking forward to your reply.

Kind regards,

Quint Koevoets

[Giuseppe Criaco]

Hi Quint,

unfortunately I haven't found any solution to this.

Cheers,

Giuseppe

[Steven Barger]

Greetings

I just ran across a Stata implementation of mediation for survival models. The article is here:

Discacciati A, Bellavia A, Lee JJ, Mazumdar M, Valeri L. Med4way: a Stata command to investigate mediating and interactive mechanisms using the four-way effect decomposition. Int J Epidemiol. 2019;48(1):15-20. https://doi.org/10.1093/ije/dyy236

The program is called med4way. It does not come up when I enter net search mediation from the command window in Stata 15.1. To directly install it from GitHub enter

net install med4way, from("https://raw.githubusercontent.com/anddis/med4way/master/") replace

I hope I can adapt it for complex survey data...

Best

Steven