Editorial

Abstract

Medical decision making in critically ill patients is often challenging. Although sound decision making should be grounded in the best available scientific evidence, busy critical care practitioners who seek to apply best evidence at the bedside may encounter obstacles along the way. Many of us have neither the time nor the skills to locate, evaluate, and synthesize the results of the relevant clinical trials. Because of this, many clinicians rely on summary information provided by review articles to keep abreast of new developments and to guide their practice decisions. In recent years, attention has focused on the strengths and weaknesses of traditional narrative review articles (1,2). Narrative reviews are usually broad in scope, summarizing the epidemiology, pathophysiology, diagnosis, and treatment of a given condition. Although their broad perspective provides a reader with an excellent introduction to a topic, information about specific clinical questions is often incomplete. In addition, narrative reviews often mix evidence with the expert opinion of the authors, which may be biased (3). New methods have been developed to overcome these limitations. Many readers are likely familiar with meta-analysis, which is one method for summarizing evidence, but may not be aware that meta-analysis belongs to a larger family of systematic review methods. Systematic reviews differ from traditional narrative reviews in several important ways (2). They usually address a specific clinical question, in contrast to the broad scope of narrative reviews. Unlike narrative reviews, systematic reviews follow planned methods that are described in sufficient detail so as to be reproducible. Authors of systematic reviews attempt to identify all potentially relevant articles through a comprehensive literature search. Articles are selected according to prespecified criteria, and are critically appraised for methodological quality. In systematic reviews, data may be synthesized either qualitatively or quantitatively, as in a meta-analysis (4). Because systematic reviews use methods to minimize random error and bias, they are more likely to produce valid, evidence-based conclusions than narrative reviews. That discrepancies exist between the results of systematic meta-analyses and the recommendations found in traditional review articles points to limitations in the validity of narrative reviews (1,5). Systematic reviews are also potentially useful for highlighting and exploring sources of variation in study results, prompting the observation that meta-analyses and other systematic reviews are “method[s] for studying studies” (6). In this issue of the Journal, Saint and Matthay (7) review the epidemiology and prevention of three common, potentially serious, and often preventable nosocomial complications of critical illness: venous thromboembolism, upper gastrointestinal bleeding, and catheter-related vascular infection. Their article illustrates and clarifies some of the differences between narrative and systematic reviews. By embracing some of the methods of systematic reviews, the authors have taken an important step in the direction of evidence-based critical care medicine. Saint and Matthay selected a broad perspective for their article, and thus, it is fundamentally a traditional narrative review. Focusing on two aspects (epidemiology and prevention) of three complications, the authors do not cover any single complication in great detail. For many of their recommendations, clinicians will still need to consult and critically review the primary literature before they can be certain that the cited evidence is valid and applicable to a particular clinical situation. As in a systematic review, Saint and Matthay describe their methods for identifying potentially relevant studies. However, other aspects of their article are more characteristic of a narrative review. In a formal systematic review, we would expect a complete description of methods for selecting or excluding studies, and a detailed appraisal of each study’s methodological quality. These omissions do not necessarily invalidate their recommendations, but do make it more difficult for the reader to render judgment. In the scale employed by Saint and Matthay, evidence from at least one randomized controlled trial in patients admitted to the intensive care unit (ICU) is designated grade A evidence. We certainly agree that evidence from a well designed and conducted randomized trial is better than evidence from an observational study or case report. But methodological quality varies among randomized Am J Med. 1998;105:551–553. From the Division of Pulmonary and Critical Care Medicine (MKG, CTL, TAR), Stanford University School of Medicine, Stanford, California, and Section of Pulmonary and Critical Care Medicine (ABW), Louisiana State University Medical Center, New Orleans, Louisiana. Requests for reprints should be addressed to Michael K. Gould, MD, MSc, Acting Assistant Professor of Medicine, Division of Pulmonary and Critical Care Medicine, Stanford University School of Medicine, H-3147, Stanford, California 94305-5236.