The need for sample size calculations is briefly reviewed: many of the
arguments against small trials are already well known, and we only cu
rsorily repeat them in passing. Problems that arise in the estimation
of sample size are then discussed, with particular reference to surviv
al studies. However, most of the issues which we discuss are equally a
pplicable to other types of study. Finally, prognostic factor analysis
designs are discussed, since this is another area in which experience
shows that far too many studies are of an inadequate size and yield m
isleading results.