Timmer Antje
Bremer Institut für Präventionsforschung und Sozialmedizin, Universität Bremen.
Z Evid Fortbild Qual Gesundhwes. 2011;105(3):194-200. doi: 10.1016/j.zefq.2011.03.003. Epub 2011 Apr 2.
Publication bias arises from the preferential publication of studies with positive, statistically significant, or particularly strong effects. In meta-analysis on the effectiveness of interventions this may, for example, lead to an overestimation of an effect. In studies on harm, though, wish bias and publication bias frequently take opposing directions. In consequence the overall effect is more difficult to interpret. Generally, non randomized studies might be expected to have a higher potential for publication bias as they have lower publication rates. Specifically, adverse effects are often not reported in a standardized way and tend to be secondary outcomes. Related biases such as positive outcome bias and location bias (problems in locating studies, retrieval bias) are therefore also more likely. As yet, however, there is no empirical proof to support this since research activities on the subject of publication bias is almost exclusively focused on randomized controlled trials.
发表偏倚源于对具有阳性、统计学显著或特别强效应的研究的优先发表。在对干预措施有效性的荟萃分析中，这可能会导致例如对效应的高估。然而，在关于危害的研究中，愿望偏倚和发表偏倚常常方向相反。因此，总体效应更难解释。一般来说，非随机研究可能被认为具有更高的发表偏倚可能性，因为它们的发表率较低。具体而言，不良反应往往没有以标准化方式报告，并且往往是次要结果。因此，相关偏倚如阳性结果偏倚和定位偏倚（查找研究中的问题、检索偏倚）也更有可能出现。然而，迄今为止，尚无实证证据支持这一点，因为关于发表偏倚主题的研究活动几乎完全集中在随机对照试验上。
