KM curve: how to account for patients with short follow-up period?

#1
Dear all,

I need to generate a KM curve for a cohort of patients that underwent a certain vascular surgery. The event is defined as one particular postop complication. We would like to calculate the survival probability until at least one year after the surgery. The problem I'm facing right now is that some patients underwent the surgery earlier this year, meaning their follow-up is only about 6 months. For these patients, I'm wondering whether it's possible to include these patients in our study and if so, what assumptions I need to make for these patients?

P.S. I'm not a stats major but I did do a bit of research before posting. I came up with two solutions but not sure if it's accepted in the field:
1. Treat them as subjected died from other causes: then I'll just censoring them out after 6 month
2. Treat them as lost to follow up: so assume they are complication-free until the end of one year (not sure about this option).

Thank you very much.
 

hlsmith

Not a robit
#2
Do you have patients that die of other causes? If so, that is a completing event and you need to control for it if there are more than a few of them.

Yes, you will likely censor them at last contact - which isn't technically the same as loss to follow-up. So for patients that have short follow-ups, they will be right censored (if I have my terminology right), so you will lose some precision in 1-year estimates when running them before all people have been followed that long and potential power loss (possible type I error maybe, if you are under-powered).

Overall your scenario isn't uncommon in such analyses, you just have to be very transparent in writing up your results.

Why can't you just wait until the whole sample has a year of follow-up???
 
Last edited: