how to set up dataset for survival analysis with time-dependent covariate

[Merdock]

I have a dataset with patients who had liver transplant. Some of them were in group 1 (if on drug A at baseline), others were in group 2 (if on drug B at baseline), depending on the immunosuppressive med they were taking. I am interested in seeing how the time of first switch from med A to med B impacts graft survival. The issue is that I am relatively new to survival analysis and have definitely no experience with time-dependent covariates.

Given this, I would like to know if anybody can give me a step-by-step on how I could model time of first switch as a time-dependent covariate in a Cox model with time to graft loss as the outcome? I have read a few theoretical, and some more practical tutorials for how to do this in R, so I know I need the data in long format or so-called counting process format with tstart and tstop and all that, but I am still having trouble getting situated and figuring out an action plan for how to analyze this, and in SAS too.

Below is a test dataset similar to the one I have, where GROUP=1, if on drug A; 2 if on drug B, and 3 if they switched from A to B;

CENSOR=1, if event (graft failure) happened, 0 otherwise;

FUTIME = total follow-up time (months) (=CENSDT – transplant date), where CENSDT (not shown here, is either date of event, or date of last follow-up/death);

EVTIME= time of event (months); EVTIME=FUTIME when CENSOR=1, else EVTIME= . ;

SWITCHDATE (not shown here) =date of first switch from drug A to drug B;

SWITCHTIME=time of first switch from group 1 to group 2 (=SWITCHDATE – transplant date); this will be 0 for those patients who start off on drug B at baseline, and it is FUTIME for those who start off on drug A and never switch.

data have;
input ID$ CENSOR$ FUTIME EVTIME SWITCHTIME GROUP$;
datalines;
001	0	1229	.		333		3
002	0	1659	.		343		3
003	0	733	 	.		509		3
004	0	6998	.		1630	3
005	1	1005	1005	558		3
006	1	4726	4726	147		3
007	0	3790	.		2856	3
008	1	672		672		504		3
009	0	5224	.		2648	3
010	0	4143	.		149		3
011	0	4973	.		500		3
012	1	3626	3626	3624	3
013	0	4296	.		3998	3
014	0	977		.		0		2
015	0	2898	.		331		3
016	0	1382	.		1187	3
017	1	1164	1164	1164	1
018	0	1232	.		336		3
019	1	1599	1599	143		3
020	0	1795	.		111		3
021	0	3171	.		3171	1
022	1	483		483		333		3
023	1	824		824		351		3
024	0	662		.		662		3
025	1	597		597		0		2
026	0	1269	.		0		2
027	0	4120	.		4120	1
028	1	621		621		0		2
029	0	3452	.		3452	1
030	0	1842	.		0		2
;
run;
proc print data=have; run; 

[pink_poodle]

Please check out this paper:
https://support.sas.com/resources/papers/proceedings12/168-2012.pdf

[Merdock]

Thank you for the reference. I was actually reading that exact paper right before I posted but I didn't find it very helpful since it uses hypertension as an example of time-dependent binary variable but I'd like to know how to do this for my continuous time-dependent covariate, which is time of first switch. How would I create my (start, tstop] intervals and account for both the event (graft failure/loss) and time to first switch as the time-dependent predictor? I guess I don't really see how time to first switch can even be seen as a time-dependent covariate..

[pink_poodle · Accepted solution]

On this page, ctrl-F “ Notice the creation of start and stop variables, which denote the beginning and end intervals defined by hospitalization and death”:
https://stats.oarc.ucla.edu/sas/seminars/sas-survival/. Anything useful?

[Merdock]

@pink_poodle, thanks! At first glance, I think there might actually be something in this article that could indeed be of help. I will need to read it more carefully though and see if I can adapt some of this for my data.