PROC PHREG: add survival-based c-statistic for competing risk

[greveam]

Hi, I would love to see c-statistics added to proc phreg for Fine and Gray competing risk outcomes.

Thanks,

Anders

[ChrisHemedinger]

@greveam Thanks for this and for following through after your Twitter question.  It's not my area of expertise, but in looking around I want to make sure that this existing option (CONCORDANCE) doesn't already cover what you need?

[greveam]

Hi Chris, the current version of concordance statistics for proc phreg only works for events with no "rival" way of exiting the riskset, e.g. 1 for death and 0 for censored(alive). When there are 2 or more competing events and they are not mutually exclusive (alive, sick, dead), then the current version of concordance statistics is not valid: https://ifsv.sund.ku.dk/biostat/annualreport/images/f/f4/Research_Report_16-04.pdf

[ChrisHemedinger]

@greveam Thanks for the additional detail.  That will help to shape the suggestion in our system for tracking these. 

[cbguo]

@greveam Many thanks for your question and suggestion. I am the developer at SAS who implemented the CONCORDANCE option in PHREG. Looking forward, we are definitely considering adding new features for evaluating competing risks models such as the Fine and Gray.

[BeverlyBrown]

Hi @greveam, R&D is taking a close look at this, so I'm changing its status to reflect that. Thank you for creating this ballot.