Methods and Applications of Statistics in Clinical Trials, Volume 2: Planning, Analysis, and Inferential Methods

Chapter 35

Publication Bias

Hannah Rothstein

35.1 Publication Bias and the Validity of Research Reviews

Publication bias is the general term used to describe the problem that results whenever the published research literature on a topic is systematically unrepresentative of the entire body of completed studies on that topic. The major consequence of this problem is that, when the research that is readily available differs in its results from the results of all the research that has been done on the topic, both readers and reviewers of that research are in danger of drawing the wrong conclusion about what that body of research shows. In some cases, the result can be danger to the public, as, for example, when an unsafe or ineffective treatment is falsely viewed as safe and effective. Two events that received much media attention in 2004 and 2005 serve as cautionary examples. Controversy surrounded Merck’s recall of Vioxx (rofecoxib), a popular arthritis drug; Merck maintained that it recalled Vioxx as soon as the data indicated the high prevalence of cardiovascular events among those who took Vioxx for more than 18 months, but media reports claimed that Merck had hidden adverse event data for years. In another controversy, the New York State attorney general filed a lawsuit against GlaxoSmithKline charging that the company had concealed data about the lack of efficacy and increased suicide risk associated with the use of Paxil (paroxetine), a selective serotonin reuptake inhibitor, ...