Cause specific hazard ratios - data structure

kc — Sun, 04 Sep 2022 03:56:53 GMT

I have a dataset (one row per subject) with the possibility that a subject can experience 3 events, one of them being death. I am using the following code to get cause specific hazard ratios for treatment group A vs. B at different levels of variable 'Category' for each of these 3 events. So, the 'Status' variable has 4 values, 0-no events, 1-event1, 2-event2, 3-death. And there are 5 effects including an interaction term in the model.

I have the following questions -

1. Should the time variable be the number of days until the first event (among all 3) or no event occurs

2. Is the code below correct. If so, the only changes when running the code for event types 2 and 3 are the censoring values

3. When asked for an interaction p-value for the 'Category*Trt' term, does reporting the p-value from the joint tests of the SAS output the right thing to do

Code below for Event Type 1 (Status=1)

proc phreg data=ds;

class Trt(ref='B') Category(ref='1') Sex(ref='Female');

model T*Status(0, 2, 3) = Category Trt Category*Trt Age Sex;

hazardratio "A vs. B" Trt / at(Category=all) diff=pairwise;

run;

Thanks!

topic Cause specific hazard ratios - data structure in Statistical Procedures

Cause specific hazard ratios - data structure