Allison D B, Faith M S, Gorman B S
Obesity Research Center, St. Luke's/Roosevelt Hospital, Columbia University College of Physicians Surgeons, NY 10025, USA.
Int J Obes Relat Metab Disord. 1996 Oct;20(10):931-7.
The present investigation examined the extent of publication bias (namely the tendency to publish significant findings and file away non-significant findings) within the obesity treatment literature.
Quantitative literature synthesis of four published meta-analyses from the obesity treatment literature. Interventions in these studies included pharmacological, educational, child, and couples treatments.
To assess publication bias, several regression procedures (for example weighted least-squares, random-effects multi-level modeling, and robust regression methods) were used to regress effect sizes onto their standard errors, or proxies thereof, within each of the four meta-analysis. A significant positive beta weight in these analyses signified publication bias.
There was evidence for publication bias within two of the four published meta-analyses, such that reviews of published studies were likely to overestimate clinical efficacy. The lack of evidence for publication bias within the two other meta-analyses might have been due to insufficient statistical power rather than the absence of selection bias.
As in other disciplines, publication bias appears to exist in the obesity treatment literature. Suggestions are offered for managing publication bias once identified or reducing its likelihood in the first place.
本研究调查了肥胖症治疗文献中发表偏倚的程度（即发表显著结果而将非显著结果存档的倾向）。
对肥胖症治疗文献中四项已发表的荟萃分析进行定量文献综合分析。这些研究中的干预措施包括药物治疗、教育治疗、儿童治疗和夫妻治疗。
为评估发表偏倚，在四项荟萃分析中的每一项中，使用了几种回归程序（例如加权最小二乘法、随机效应多层次建模和稳健回归方法），将效应量回归到其标准误或其替代指标上。这些分析中显著的正β权重表示存在发表偏倚。
在四项已发表的荟萃分析中有两项存在发表偏倚的证据，因此对已发表研究的综述可能高估了临床疗效。另外两项荟萃分析中缺乏发表偏倚的证据，可能是由于统计效力不足而非不存在选择偏倚。
与其他学科一样，肥胖症治疗文献中似乎也存在发表偏倚。针对一旦发现发表偏倚如何进行管理或首先降低其可能性提出了建议。
