Contemporary Outcome Measures in Acute Stroke Research

See related articles, p 1163 and 1171.The 2 articles in this issue1,2 are based on the recent European Stroke Organization Outcomes Workshop and provide an excellent basis for discussion of outcomes used in stroke trials. Bath, Lees, and colleagues raise important points to consider when choosing an outcome measure or an analysis plan. The authors speculate that trials to identify therapeutic benefits of effective drugs may fail to find benefits because of problems inherent in choice of outcome measures or methods of analysis. Early in their article, Bath and colleagues give examples of multiple ways to define an outcome and show that trial results vary based on the definitions used. The emphasis throughout the article is on increasing power to detect differences so effective treatments will not be missed. There are many ways to increase power, but not all lead to the identification of clinically meaningful differences. The authors give little mention to the possibility of a Type 1 error, that is, declaring an ineffective therapy effective. Type I errors can result from searches for outcomes and analytic approaches to increase the power to detect treatment differences after a trial is completed. It is as important to patients that ineffective treatments are kept out of clinical practice as it is to assure that effective treatments are not missed.The authors recognize that 1 outcome and 1 analytic approach may not fit all trials. In choosing an outcome or approach, the most important consideration is the research question of interest. It is the trial hypothesis that should drive the choices. As an example, at a workshop to identify outcomes and an analytic approach for the National Institute of Neurological Disorders and Stroke Tissue Plasminogen Activator (NINDS tPA) Stroke Trial,3 Thomas Brott stated that he would …