Evidence-based practice requires clinicians to understand the current research relevant to clinical decisions. This is a tall order, given both the number of clinical decisions made each day and the large and ever-increasing number of research articles published each year. The Cochrane Collaboration ( www.cochrane.org ) provides systematic reviews that are meant to guide both health care and health policy. Although Cochrane has more than 5000 such reviews, it cannot possibly explore every clinical question. In addition, a systematic review does not necessarily focus on specific, real-world clinical situations. A critically appraised topic (CAT) can fill these gaps, providing clinicians with the information necessary for their specific clinical decisions and enabling effective communication with patients. Such communication is the foundation of patient-centered care and is the first step in shared decision-making. A well-written CAT may even be directly useful to motivated patients who are seeking information on their own.
Sadigh, Parker, Kelly, and Cronin are to be congratulated for their comprehensive description of the process of writing a high-quality CAT, which appears in this issue of Academic Radiology . As they demonstrate, writing a high-quality CAT is a complex task that requires a broad knowledge base. For a clinician planning to write a CAT, the article by Sadigh and colleagues outline key steps in gathering and describing the evidence on a chosen clinical topic and includes many relevant and important references. In this editorial, we address a key step neglected by their tutorial: building a strong, multidisciplinary CAT writing team. Only a few, rare individuals possess both the clinical and statistical expertise needed to write a CAT without assistance. For this reason, writing a CAT provides an excellent opportunity for collaboration between clinicians and statisticians.
The first step identified by Sadigh and colleagues is to clearly articulate the clinical question to be addressed in the CAT. Clinicians are certainly able to identify important and relevant clinical questions, yet statisticians can provide important support at this stage. Specifically, statisticians can assist in refining and focusing the question of interest to ensure that it can be addressed given the available evidence. This is an important component of statistical training and is analogous to consulting with a statistician in the early phases of a research project.
The next step in writing a CAT is identification of articles that are potentially related to the clinical question at hand. In most cases this search will yield many published articles. A single high-quality study is rarely comprehensive enough to completely answer a clinical question: conditions need to be varied, subgroups examined, and findings replicated. A CAT article provides an essential link in the evidence chain, yet a synthesis of available evidence is only as informative as the quality of the literature review at its base. As Sadigh et al point out, a good literature search requires a broad familiarity with various types of studies and the strength of evidence each provides. A multidisciplinary CAT writing team that includes statistical expertise helps ensure a comprehensive examination of the literature.
After potential articles have been identified, each one must be critically appraised—that is, evaluated for study quality and relevance. The quality of a study is determined by a collection of factors including study design, how the data are analyzed, and how the results are reported and interpreted. Although many clinicians are statistically savvy, the skills required to evaluate study quality are particularly aligned with the skills and training of a statistician. Statisticians are trained in study design, selection of appropriate analytic methods, and appropriate interpretation of results, making them well-equipped to identify potential problems and biases. Just as a clinician develops clinical insights with experience in clinical practice, a statistician continues to gain insights, beyond his or her classroom training, in study validity, generalizability, and bias with experience collaborating with research teams studying different types of diseases and populations. Statisticians are better able to identify potential problems in specific clinical areas after learning about the disease process through discussions with clinician-collaborators. Thus, the task of assessing study quality builds partnerships between clinicians and statisticians, with both contributing unique expertise required for assessing the importance and rigor of a scientific study.
Although study quality is critical to weighing the available evidence in a CAT, threats to study quality can be subtle. For example, although randomized controlled trials are considered the gold standard for estimating treatment efficacy, analytic issues can still arise in randomized trials, such as the proper estimation of standard errors or bias from missing data or low compliance. Statisticians regularly wield statistical tools to minimize bias in these settings, so they know what to look for in the statistical analysis section of a high-quality study. Observational studies are useful for providing evidence beyond randomized trials and can answer scientific questions that cannot be addressed in a randomized setting (eg, for ethical reasons). However, as Sadigh and colleagues note, observational studies have much greater potential for bias than randomized studies. Although statistical approaches can be used to minimize bias in observational studies, these methods can be highly technical and often involve assumptions that must be validated in the context of a particular study. Statisticians are trained to be cognizant of the assumptions necessary for various statistical methods, the appropriate ways to validate these assumptions, and the red flags that can indicate problems with an analysis.