Dear all,
I have 2 queries about cumulative incidence analysis (competing risks).
Question 1
Say we have 2 causes of failure, event A and event B. For arguments
sake, say we want to calculate the cumulative incidence of type A event
(in the presence of event B).
Usually texts talk about the case where failure is "death". Causes of
death can be heart attack, cancer etc. and of course when a person dies
he/she cannot die again at a later date...but what happens when an
individual's causes of failure can each happen several times in their
lifetime? Say one of our individuals 'failed' due to event B at time 8
months, then they were treated for this failure and recovered but then
they failed again due to event A at 23 months.
My question is, for analysis, for this individual, would it be the time
to the *first* failure that I recorded? i.e. for this individual :
time = 8 months Event = B
would be entered into the data set for analysis.
So, we are saying that for each individual in a competing risk analysis,
only one cause of failure can be assigned from a given set of p
causes....this one cause of failure is the *first* cause that occurs.
Question 2
One example I found on cumulative incidence procedures dealt with bone
marrow transplantation. The example talked about 'failure of therapy'
and involved two events, 'relapse' and 'non relapse mortality'. Does
this imply that the two events have to be non-related? If we had two
events, A and B, where A and B were *related illnesses* and somone died
of event A at 8 months and someone else got event B at 11 months, could
event A and B be entered as two separate competing risks of failure or
not?
Many thanks for your views on this,
Best Wishes,
Kim
|