Time to or duration of an event

Time-to-event analyses are frequently used in cancer trials, usually of course time to death, but this can equally be used to assess the time to worst, or best, QL. This is probably a valid comparison when the two therapies have similar survival, but if there are differences, censoring at death can give a biased result. This is because, with the Kaplan-Meier analyses, censored patients are assumed to follow the same pattern of response or side-effects as those with full information. This is perfectly reasonable if a patient is censored because they have not been in the trial very long, or have been lost to follow-up, but is often not so logical if they are censored because of death.

0 0

Post a comment