Much effort has been made during the last couple of decades (not least by the work of the Cochrane Collaboration and by the development of the CONSORT Guidelines) to improve the the design, analysis and reporting of randomized clinical trials.
To investigate the quality of publications from subfertility trials, Vail and Gardener (1) reviewed 39 articles published during 2001 in Human Reproduction and Fertility and Sterility.
There was analysis unit errors in 32 (82%) of the articles, i.e. the analysis unit was pregnancy or birth instead of woman.
Six of the articles (15%) presented trials that were fatally flawed due to inappropriate use of cross-over design or to systematic allocation described as random.
Only 6 (15%) complied with the fundamental intention-to-treat principle (that the primary analysis includes all randomized subjects in the groups to which they were allocated), and the principle had been misunderstood in 4 of these 6 articles.
Vail and Gardener state that their assumption when they planned the review was that the articles would represent trials of the highest methodology, but their conclusion from the review is that many subfertility trials have insufficient design, analysis or results reporting, and that the results of these cannot be reliably interpreted.
1. Vail A, Gardener E. Commons statistical errors in the design and analysis of subfertility trials. Hum Rep 2003;18:1000-1004.