BookmarkSubscribeRSS Feed

Dear SAS-users,

I work with registry data from national registries and often, when doing survival analysis, I have to account for left-truncation. I am a big fan of using the 'assess ph' statement to evaluate proportionality and linearity of covariates, since it is superior to manual, individual testing of each covariate and has been proven/evidenced to be reliable. Unfortunately, when using Proc Phreg in tandem with delayed entry, using either syntax :

"model (start,stop) * event(0) = " OR "model (stop) * event(0) = / entry = start"

The 'assess ph' statement is not included/functional. The zph statement is a poor substitute, as the supremum test P-values are not valid (no scientific backing) and relies on graphical interpretation. 

This forces me to use R and the packages (prodlim) and (mets) instead. However, I would really like to stay within the SAS working environment, not having to skip back and forth between the two. 

Would it be possible to add this feature?


In advance thanks for reading my suggestion 🙂 

1 Comment
Obsidian | Level 7

Minor addition: This also applies to a Phreg with time-dependent/variying covariates.