The use of state of the art statistical methodology in biomedical research: the example of methodology for interim analyses in clinical trials.

BACKGROUND: Clinical research stresses the need for evidence as a basis for setting research agendas. The extent to which current clinical research involves recent advances in biostatistics is not well known. AIM: This manuscript focuses on the use of interim analyses accumulating data in clinical trials. MATERIAL AND METHODS: A PubMed search was conducted for the period 1990-2000. Keywords used included 'interim analysis', 'alpha-spending function', and 'early termination', after subsetting to 'randomised controlled trials (RCT)' that were 'multicenter' since these particular studies are more likely to have external monitoring boards conducting interim analyses. Studies were grouped by clinical discipline and also by journals of high or low impact index. RESULTS: Thirteen thousand two hundred eighty two articles with both 'RCT' and 'multicenter' were found. Of these, the 'interim analysis' keyword yielded 879 articles, of which only 163 (19%) studies reported their methodology. The percentages varied by clinical discipline and by journal type. A 25% random sample of the articles were further reviewed to investigate the appropriateness of the interim analysis methods used. DISCUSSION: Recent articles in the medical literature attempt to make the interim analysis methods more accessible to clinicians. However, further training of statisticians involved in clinical trials is needed as well.