Premium
Interim analysis of failure time data — A Bayesian approach
Author(s) -
Bartolucci Alfred A.,
Katholi Charles R.,
Birch Robert
Publication year - 1992
Publication title -
environmetrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.68
H-Index - 58
eISSN - 1099-095X
pISSN - 1180-4009
DOI - 10.1002/env.3170030407
Subject(s) - interim , bayesian probability , statistics , sample size determination , interim analysis , computer science , sample (material) , mathematics , econometrics , clinical trial , medicine , chemistry , archaeology , pathology , chromatography , history
Since the final results of a comparative intervention (environmental or clinical) trial are usually in doubt at an interim analysis, it is necessary to quantify the uncertainty associated with data yet to be observed. A natural way of quantifying such uncertainty is the Bayesian predictive distribution. The use of such a distribution requires an investigator to specify his opinions in the form of a prior distribution. We reference how this distribution may be obtained by eliciting opinions from an investigator. Further, the methodology is applied to determine an appropriate stopping rule in an experiment to compare survival times of two treatments. This is done by deriving the probability that the ratio of mean or total survival times given the current sample size exceeds a given fixed value. If this probability is below a certain specified value, a dynamic estimate of the required additional sample size may be obtained. We show examples from cancer clinical trials.