Department of Clinical Research, Basel Institute for Clinical Epidemiology and Biostatistics, University Hospital Basel, University of Basel, Basel, Switzerland; Swiss Tropical and Public Health Institute, Basel, Switzerland; University Medical Library, University of Basel, Basel, Switzerland.
Cochrane Austria, Department of Evidence-based Medicine and Evaluation, Danube University Krems, Krems, Austria.
J Clin Epidemiol. 2020 Dec;128:1-12. doi: 10.1016/j.jclinepi.2020.08.002. Epub 2020 Aug 8.
The objective of this study was to assess the agreement of treatment effect estimates from meta-analyses based on abbreviated or comprehensive literature searches.
This was a meta-epidemiological study. We abbreviated 47 comprehensive Cochrane review searches and searched MEDLINE/Embase/CENTRAL alone, in combination, with/without checking references (658 new searches). We compared one meta-analysis from each review with recalculated ones based on abbreviated searches.
The 47 original meta-analyses included 444 trials (median 6 per review [interquartile range (IQR) 3-11]) with 360045 participants (median 1,371 per review [IQR 685-8,041]). Depending on the search approach, abbreviated searches led to identical effect estimates in 34-79% of meta-analyses, to different effect estimates with the same direction and level of statistical significance in 15-51%, and to opposite effects (or effects could not be estimated anymore) in 6-13%. The deviation of effect sizes was zero in 50% of the meta-analyses and in 75% not larger than 1.07-fold. Effect estimates of abbreviated searches were not consistently smaller or larger (median ratio of odds ratio 1 [IQR 1-1.01]) but more imprecise (1.02-1.06-fold larger standard errors).
Abbreviated literature searches often led to identical or very similar effect estimates as comprehensive searches with slightly increased confidence intervals. Relevant deviations may occur.
本研究旨在评估基于简化或全面文献检索的荟萃分析中治疗效果估计的一致性。
这是一项荟萃流行病学研究。我们简化了 47 项全面的 Cochrane 综述检索，并单独或联合检索了 MEDLINE/Embase/CENTRAL，同时检查了参考文献（共进行了 658 次新检索）。我们比较了每个综述中的一个荟萃分析与基于简化检索重新计算的荟萃分析。
47 项原始荟萃分析包括 444 项试验（每项综述的中位数为 6 项[四分位距（IQR）3-11]），涉及 360045 名参与者（每项综述的中位数为 1371 名[IQR 685-8041]）。根据检索方法的不同，简化检索在 34%-79%的荟萃分析中导致了相同的效果估计，在 15%-51%的荟萃分析中导致了相同方向和统计学意义水平的不同效果估计，在 6%-13%的荟萃分析中导致了相反的效果（或无法再估计效果）。50%的荟萃分析中效应大小的偏差为零，75%的荟萃分析中偏差不超过 1.07 倍。简化检索的效应估计值并不总是更小或更大（中位数比值比为 1[IQR 1-1.01]），但更不精确（标准误差增加 1.02-1.06 倍）。
简化文献检索通常会导致与全面检索相同或非常相似的效果估计，置信区间略有增加。可能会出现相关偏差。
