Shinohara Kiyomi, Suganuma Aya M, Imai Hissei, Takeshima Nozomi, Hayasaka Yu, Furukawa Toshi A
Department of Health Promotion and Human Behavior, Kyoto University Graduate School of Medicine / School of Public Health, Konoe-cho, Sakyo-ku, Kyoto, Japan.
PLoS One. 2017 Sep 13;12(9):e0184786. doi: 10.1371/journal.pone.0184786. eCollection 2017.
Abstracts of scientific reports are sometimes criticized for exaggerating significant results when compared to the corresponding full texts. Such abstracts can mislead the readers. We aimed to conduct a systematic review of overstatements in abstract conclusions in psychiatry trials.
We searched for randomized controlled trials published in 2014 that explicitly claimed effectiveness of any intervention for mental disorders in their abstract conclusion, using the Cochrane Register of Controlled Trials. Claims of effectiveness in abstract conclusion were categorized into three types: superiority (stating superiority of intervention to control), limited superiority (intervention has limited superiority), and equal efficactiveness (claiming equal effectiveness of intervention with standard treatment control), and full text results into three types: significant (all primary outcomes were statistically significant in favor of the intervention), mixed (primary outcomes included both significant and non-significant results), or all results non-significant. By comparing these classifications, we assessed whether each abstract was overstated. Our primary outcome was the proportion of overstated abstract conclusions.
We identified and included 60 relevant trials. 20 out of 60 studies (33.3%) showed overstatements. Nine reports reported only significant results although none of their primary outcomes were significant. Large sample size (>300) and publication in high impact factor (IF>10) journals were associated with low occurrence of overstatements.
We found that one in three psychiatry studies claiming effectiveness in their abstract conclusion, either superior to control or equal to standard treatment, for any mental disorders were overstated in comparison with the full text results. Readers of the psychiatry literature are advised to scrutinize the full text results regardless of the claims in the abstract.
University hospital Medical Information Network (UMIN) Clinical Trials Registry (UMIN000018668).
科学报告的摘要有时会因与相应全文相比夸大显著结果而受到批评。此类摘要可能会误导读者。我们旨在对精神病学试验摘要结论中的夸大陈述进行系统评价。
我们使用Cochrane对照试验注册库，检索了2014年发表的随机对照试验，这些试验在摘要结论中明确声称任何干预措施对精神障碍有效。摘要结论中的有效性声明分为三种类型：优越性（表明干预措施优于对照）、有限优越性（干预措施有有限优越性）和等效有效性（声称干预措施与标准治疗对照等效），全文结果分为三种类型：显著（所有主要结局在统计学上均显著支持干预措施）、混合（主要结局包括显著和非显著结果）或所有结果均不显著。通过比较这些分类，我们评估了每篇摘要是否存在夸大陈述。我们的主要结局是夸大的摘要结论的比例。
我们识别并纳入了60项相关试验。60项研究中有20项（33.3%）存在夸大陈述。9篇报告仅报告了显著结果，但其主要结局均无显著性。大样本量（>300）和发表于高影响因子（IF>10）期刊与夸大陈述的低发生率相关。
我们发现，在声称对任何精神障碍在摘要结论中具有有效性（优于对照或等同于标准治疗）的精神病学研究中，三分之一与全文结果相比存在夸大陈述。建议精神病学文献的读者仔细审查全文结果，而不管摘要中的声明如何。
大学医院医学信息网络（UMIN）临床试验注册库（UMIN000018668）。
