Efficiency of the cross-over design: an empirical estimation

The aim of the present study is to estimate the relative efficiency of cross-over clinical trials as compared to the corresponding parallel designs. A MEDLINE 'full text' search covering years 2000-2003 over the word 'cross-over' identified 40 true cross-over studies with the parametric approach published in leading medical journals. A parallel design is expected to need, in order to achieve the same power, between 4 and 10 times more subjects than the corresponding cross-over design, meaning that cost of a parallel design can be between 2 and 5 times higher. There is a high proportion (18/40 = 45.0 per cent) of cross-over studies recently published in leading medical journals that, disregarding the CONSORT recommendations, does not provide estimates for the effect size and its standard error, hence, not allowing for the posterior use of the data in a meta-analysis. Copyright (C) 2004 John Wiley Sons, Ltd.