The task of a statistical referee.

The results are presented of an audit of the statistical standard of papers published in The British Journal of Surgery. A number of deficiencies are highlighted, many of which stem from an over-emphasis on statistical significance at the expense of any assessment of the clinical relevance of research findings. The flaws in the design of published studies, and in particular the many instances of inadequate sample sizes, emphasize that statistical input should be sought at the beginning of a research project rather than at the end.