Quantification of the prentice criteria for surrogate endpoints

In a recent discussion oil the Prentice criteria for surrogate endpoints (1989) and the role of meta. analyses in this context, Alonso et al. (2004, Biometrics 60, 724-728) expand on a number of points. Among these is a suggestion to base a, measure of the degree at which all endpoint meets the Prentice criteria on a likelihood reduction factor. Although new in this context, the suggestion has been studied elsewhere and the measure of Alonso et al., as presented, requires some modification if it is to be useful in general situations. As it stands it depends upon the censoring mechanism, even when independent of the failure mechanism. This is, however, easily Fixed. The measure suggested by Alonso et al. can be seen to fit in with a well-established theory so that, once corrected for the presence of censoring, it can be employed with confidence. Of particular interest is the wide availability of a standard SAS program producing a coefficient which call be immediately transformed to provide all estimate of the population quantity behind the Alonso et al. proposal.