I have the following setup:
- Population who all experience a qualifying event (age and year of event vary per person)
- Population is followed until outcome of interest (cancer) or competing risk of death, censored at last follow-up
- There is a desired stratification variable Treatment, coded as 0/1. This value is time-dependent. A person who never receives treatment is coded as 0 throughout follow-up. A person who receives treatment at baseline is coded as 1 throughout follow-up. A person who receives treatment during the follow-up period is coded as 0 until treatment and 1 thereafter.
- The time variable of interest for cumulative incidence is latency since initial exposure.
I am requested to provide cumulative incidence figures for cancer with competing risk of death, stratified by treatment status (treated as time-dependent). I could use PROC LIFETEST to code this except that I need to start follow-up after the initial exposure for patients who received treatment after baseline. There does not seem to be a way to handle this (what I believe is 'left truncation') in SAS. I do have macro code from https://www.nature.com/articles/s41409-021-01435-2 that appears to allow for late entry, but it does not handle competing risks.
Does anyone have a solution that would work for this situation, or advice that this is not a valid approach? Essentially the request is to provide separate cumulative incidence curves for each stratification (treatment yes vs. no) but if this is not reasonable then I would like to know.
