On 10 January 1998 we published a retraction (1998;316:116) of this paper (1997;314:550-4) requested by the authors on the grounds that examination of the data had revealed inaccuracies and that the conclusions could not be sustained. The University of Leicester, where the work was done, established an inquiry, which has now confirmed that there were significant errors in the data and that the paper's findings were unsound. The errors arose in transcribing data from paper to digital form for statistical analysis, and the individual responsible has been censured. Following this incident the university has introduced a code of practice governing the management and conduct of research.