Evans M, Pollock A V
Br J Surg. 1985 Apr;72(4):256-60. doi: 10.1002/bjs.1800720403.
We devised 33 rules for the evaluation of random control trials (RCT) and used them to assess 56 RCTs listed in Index Medicus during 1980-82 on antibiotic prophylaxis of surgical wound infection. We asked 15 questions about design and conduct, 10 about analysis and 8 about presentation. Out of a maximum score of 100 only 16 papers scored over 70, the highest being 89 and the lowest 34. Defects in presentation were not common, and 17 papers scored over 90 per cent. Defects in analysis included the incorrect use of statistical tests and ignoring the Type II error in 'negative' trials; only 13 papers scored over 70 per cent. Defects in design and conduct included transgressions of ethical principles, inappropriate regimens, ill-defined end points and biased randomization or assessment; only 20 papers scored over 70 per cent. We conclude that there is room for improvement in the performance of RCT and that the application of the 33 rules would be helpful not only to researchers but also to editors and referees of scientific journals, and to their readers.
我们制定了33条评估随机对照试验（RCT）的规则，并运用这些规则对1980年至1982年《医学索引》中列出的56项关于外科伤口感染抗生素预防的随机对照试验进行了评估。我们提出了15个关于设计与实施的问题、10个关于分析的问题以及8个关于呈现的问题。在满分100分的情况下，只有16篇论文得分超过70分，最高分是89分，最低分是34分。呈现方面的缺陷并不常见，有17篇论文得分超过90%。分析方面的缺陷包括统计检验的错误使用以及在“阴性”试验中忽略II型错误；只有13篇论文得分超过70%。设计与实施方面的缺陷包括违反伦理原则、方案不合适、终点定义不明确以及随机化或评估存在偏差；只有20篇论文得分超过70%。我们得出结论，随机对照试验的表现仍有改进空间，应用这33条规则不仅对研究人员有帮助，对科学期刊的编辑、审稿人及其读者也会有所帮助。
