Schmucker C, Meerpohl J J, Blümle A
Institut für Evidenz in der Medizin (für Cochrane Deutschland Stiftung), Medizinische Fakultät, vertreten durch das Universitätsklinikum Freiburg, Albert-Ludwigs-Universität, Breisacher Str. 153, 79110, Freiburg, Deutschland.
HNO. 2020 Apr;68(4):291-300. doi: 10.1007/s00106-020-00835-y.
Results from clinical studies are often subject to the risk of bias (deviation from the truth, systematic error). Therefore, a critical appraisal of studies provides a useful strategy in evidence-based healthcare to safeguard against wrong decisions and resulting in overtreatment or undertreatment. This article explains the frequently encountered types of bias, differentiates between them and provides strategies for avoidance of systematic errors. In addition, the two established Cochrane tools with which the risk of bias can be assessed in randomized and non-randomized studies are presented. To highlight the most important components of these tools for bias assessment, examples of randomization, confounding, blinding, completeness of data and selective reporting are provided. Finally, it is shown that bias should not be confused with other study limitations, such as external validity and imprecision.
临床研究结果常常存在偏倚风险（与事实不符、系统误差）。因此，对研究进行严格评价是循证医疗中的一项有用策略，可防止做出错误决策以及由此导致的过度治疗或治疗不足。本文解释了常见的偏倚类型，对它们进行了区分，并提供了避免系统误差的策略。此外，还介绍了两种既定的Cochrane工具，可用于评估随机和非随机研究中的偏倚风险。为突出这些偏倚评估工具的最重要组成部分，文中提供了随机化、混杂、盲法、数据完整性和选择性报告的示例。最后表明，不应将偏倚与其他研究局限性（如外部效度和不精确性）相混淆。
