Goodin Douglas S
Department of Neurology, Room M794, University of California San Francisco, 505 Parnassus Avenue, San Francisco, California 94143-0114, USA.
J Neurol. 2004 Sep;251 Suppl 5:v3-v11. doi: 10.1007/s00415-004-1503-z.
Several types of errors are commonly made during the conduct, analysis, and interpretation of clinical trials in multiple sclerosis (MS). These include statistical errors of overestimating the significance of trial results, particularly when multiple endpoints are evaluated. They also include errors arising from the use of inappropriate covariate analyses, meta-analyses, and post hoc subgroup analyses. Interpretation of trial results can also be confounded by regression to the mean, by post hoc data re-analysis, and by the use of a non-concurrent control population. As these kinds of errors continue to plague the medical literature, it has become important for physicians to be able to assess critically the reports of clinical trial results. In turn, this has made it necessary for physicians to become familiar with the rudiments of the fields of statistics, epidemiology, and trial design. It is the purpose of this manuscript, therefore, to provide an overview of these principles through a detailed analysis of these kinds of clinical trial errors, together with examples that have actually occurred in the recently published MS literature.
