Scholten R J, Assendelft W J, Kostense P J, Bouter L M
Vrije Universiteit, Faculteit der Geneeskunde, Instituut voor Extramuraal Geneeskundig Onderzoek, Amsterdam.
Ned Tijdschr Geneeskd. 1999 Apr 17;143(16):843-8.
Meta-analysis comprises the analysis of the data of similar studies gathered in a systematic review. If the results of the studies differ (strongly), this is called heterogeneity. Possible causes of heterogeneity are: a wrong choice of type of measure of treatment effect, differences of methodological quality between studies, or real differences between studies. Heterogeneity between studies can be assessed by thorough examination of the differences between study characteristics in combination with a visual inspection of the degree of overlap of the confidence intervals of the estimates of effect of the different studies. By applying the random effects model for pooling, (non-systematic) heterogeneity between studies can be addressed. In case of heterogeneity between studies, sources of heterogeneity should be explored by means of subgroup analysis according to strict, present criteria. Subgroup analysis should address subgroups of patients within studies.
荟萃分析包括对系统评价中收集的类似研究的数据进行分析。如果各研究结果（差异很大）不同，则称之为异质性。异质性的可能原因有：治疗效果测量类型选择错误、研究之间方法学质量的差异或研究之间的实际差异。可以通过全面检查研究特征之间的差异并结合直观检查不同研究效应估计的置信区间的重叠程度来评估研究之间的异质性。通过应用随机效应模型进行合并，可以处理研究之间的（非系统性）异质性。如果研究之间存在异质性，应根据严格的现有标准通过亚组分析来探索异质性来源。亚组分析应针对研究中的患者亚组。
