Makowski David, Catarino Rui, Chen Mathilde, Bosco Simona, Montero-Castaño Ana, Pérez-Soba Marta, Schievano Andrea, Tamburini Giovanni
Unit Applied Mathematics and Computer Science (MIA Paris-Saclay), INRAE AgroParisTech Université Paris-Saclay, 91120, Palaiseau, France.
European Commission, Joint Research Centre (JRC), Ispra, Italy.
Environ Evid. 2023 Aug 21;12(1):16. doi: 10.1186/s13750-023-00309-y.
Statistical synthesis of data sets (meta-analysis, MA) has become a popular approach for providing scientific evidence to inform environmental and agricultural policy. As the number of published MAs is increasing exponentially, multiple MAs are now often available on a specific topic, delivering sometimes conflicting conclusions. To synthesise several MAs, a first approach is to extract the primary data of all the MAs and make a new MA of all data. However, this approach is not always compatible with the short period of time available to respond to a specific policy request. An alternative, and faster, approach is to synthesise the results of the MAs directly, without going back to the primary data. However, the reliability of this approach is not well known. In this paper, we evaluate three fast-track methods for synthesising the results of MAs without using the primary data. The performances of these methods are then compared to a global MA of primary data. Results show that two of the methods tested can yield similar conclusions when compared to global MA of primary data, especially when the level of redundancy between MAs is low. We show that the use of biased MAs can reduce the reliability of the conclusions derived from these methods.
数据集的统计合成（荟萃分析，MA）已成为一种流行的方法，用于提供科学证据以指导环境和农业政策。随着已发表的荟萃分析数量呈指数级增长，现在针对特定主题通常有多个荟萃分析，有时会得出相互矛盾的结论。为了综合多个荟萃分析，第一种方法是提取所有荟萃分析的原始数据，并对所有数据进行新的荟萃分析。然而，这种方法并不总是与应对特定政策请求的短时间要求相兼容。另一种更快的方法是直接综合荟萃分析的结果，而无需追溯到原始数据。然而，这种方法的可靠性尚不清楚。在本文中，我们评估了三种不使用原始数据来综合荟萃分析结果的快速方法。然后将这些方法的性能与原始数据的整体荟萃分析进行比较。结果表明，与原始数据的整体荟萃分析相比，所测试的两种方法可以得出相似的结论，特别是当荟萃分析之间的冗余度较低时。我们表明，使用有偏差的荟萃分析会降低从这些方法得出的结论的可靠性。
