Retrospective clinical studies in surgery: potentials and pitfalls.

As many surgical studies are performed retrospectively, it is desirable to improve the conduct, analysis and reporting of such research designs. This article reviews some of the most common forms of bias encountered in clinical research and outlines various types of study design and analysis that can help to minimize the amount of bias introduced into the results. We describe the advantages and disadvantages of 'raw' data analysis, stratified analysis, matched pair analysis and multivariate analysis, and apply them to an example of a hypothetical hand surgery study. Copyright 2002 The British Society for Surgery of the Hand.