n medical and pharmaceutical research, interest in using biomarkers as surrogate endpoints
for target clinical endpoint has stemmed from various reasons. Because of importance of
statistical evaluation of surrogate marker, very diff
erent methods are suggested.
Alonso et al. proposed the ?likelihood reduction factor? (LRF) as a unifi
ed approach when
neither the biomarker nor the true endpoint is normally distributed. Th
is measure of individual-
level association can be used under any genera
lized linear model for a single trial or meta-
Flowing of these criteria for surrogate evaluation, in this study, we have explored the
Bayesian approach to the evaluation of the validity of a surrogate at the individual level, based
on the Bayes factor for choosing the best model in the context of generalized linear modeling.
It is suggested that the Bayesian LRF denoted by LRFB which benefi
ts from the prior
knowledge on the situation under study would perform yet better in comparison to other
By a Th
eorem we proof, for large sample size, Lim LRF
e relation between the
Bayesian likelihood reduction factor (LRF
) and its frequentist counterpart (LRF) have been
shown by a small scale simulation also.
We have simulated diff
erent trials with diff
erent priors in the logistic regression models by R
e results show that LRF can be viewed as a special case of LRFB relative to a certain
prior. Hence, the importance of prior knowledge and Bayesian analysis for surrogate?s validation
Shohreh Jalaie has completed her Ph.D in Biostatistics from Tarbiat Modares University in 2008. She passed a
scholarship in Melbourne University. She is faculty member of Tehran University of Medical Science. She has
published more than 20 papers in reputed journals.
Peer Reviewed Journals
Make the best use of Scientific Research and information from our 700 + peer reviewed, Open Access Journals