Survival analyses with two outcomes

[Al Bothwell]

Hello everyone,

I have clinical trial data with a follow-up period of 5 years. I have two outcomes, let's call them outcome A (recorded as occured or not) and outcome B (recorded as occured or not). Using descriptive statistics, I found that 85% of patients who suffer outcome A also suffer B.

My hypothesis is that outcome A increases the chances of outcome B occuring . Is there a type of survival analyses that will allow me to model the hypothesized temporal precedence? I looked at competing risks but it looks like it is for situations where outcome A prevents the occurence of outcome B.

Suggestions will be most appreciated.

Al Bothwell

[Weiwen Ng]

Al, I am not an expert in survival analysis. However, seeing as you have no other responses yet, my two cents are that this sounds like it can be handled using regular survival analysis with time-dependent covariates. You’d want to create a new record for each person when they get outcome A. Then, you could just use outcome A as a covariate. If the hazard ratio is more than 1, that is consistent with your hypothesis.

You’re correct that this isn’t a competing risk problem unless you ended observation when a subject got either A or B.