Proc Causalmed for Mediation Analysis with Time-to-event data

[Lyansu]

I have a project related to mediation analysis with time-to-event data (Cox model is preferred), and tried to rerun the SAS code for "Example 38.5 Causal Mediation Analysis of a Time-to-Event Outcome".

https://documentation.sas.com/doc/en/pgmsascdc/v_008/statug/statug_causalmed_examples05.htm

However, it did NOT in the SAS on my computer. I got the follow error message from the log:

4546  proc causalmed data=RedDye ciratio=log poutcomemod pmedmod;
                                 -------
                                 22
                                 202
WARNING: Ignoring second data set reference.
NOTE: PROCEDURE CAUSALMED used (Total process time):
      real time           0.00 seconds
      cpu time            0.00 seconds

NOTE: The SAS System stopped processing this step because of errors.
ERROR 22-322: Syntax error, expecting one of the following: ;, (, ALL, ALPHA, CASECONTROL, DATA,
              DECOMP, DESC, DESCEND, DESCENDING, EPSILON, NAMELEN, NAMELENGTH, NLOPTIONS,
              NOPRINT, ORDER, PALL, PMEDMOD, POUTCOMEMOD, PSHORT, PSUMMARY, RORDER, SINGULAR,
              THREADS.
ERROR 202-322: The option or parameter is not recognized and will be ignored.
4547     class RedDye Female Tumor;
4548     model Weeks*Censor(1) = RedDye | Tumor / aft;
4549     mediator Tumor(event='1') = RedDye(treat='3');
4550     covar Female;
4551     freq Freq;
4552  run;

Could anyone please let me know why this happened? How can I fix it? Or you can suggest other procedures. Thank you very much.

[ChrisHemedinger]

Some of these options were added in SAS Viya 2020.1 and might not work in SAS v9.4-based systems.

See What's New  and search for CAUSALMED.

Excerpt:

CAUSALMED Procedure

The following enhancements have been made to the CAUSALMED procedure:

• The procedure now supports the causal mediation analysis of time-to-event outcomes. You can specify the following two types of models for the time-to-event outcome in the MODEL statement:
  • Cox proportional hazard models (with the COXPH option)
  • accelerated failure time models (with the AFT option)
• The new CIRATIO= option enables you to compute confidence intervals for ratio-type causal mediation effects on the log scale, original (nontransformed) scale, or both. The z-value and the p-value for testing null hypothesis of no effect are also computed separately for each of the chosen scales.
• You can now specify the treated or control level of a binary treatment variable by using the TREAT= or CONTROL= option, respectively, in the MEDIATOR statement. You can specify the event or reference level of a binary outcome variable by using the EVENT= or REF= option, respectively, in the MODEL statement.

[soosas]

So is there a way to run PROC CAUSALMED with time-to-event data in SAS v9.4-based systems?

[soosas]

I just answered this for myself. The answer is no, referencing this thread:

https://communities.sas.com/t5/SAS-Procedures/error-73-322-when-using-proc-causalmed-for-survival-an...