• Yafang Huang, Jinling Tang, Wilson Wai-San Tam, Chen Mao, Jinqiu Yuan, Mengyang Di, and Zuyao Yang.
• From the School of General Practice and Continuing Education, Capital Medical University (YH), Beijing, China; Alice Lee Centre for Nursing Studies, Yong Loo Lin School of Medicine, National University of Singapore (WW-ST), Singapore; and Division of Epidemiology, The Jockey Club School of Public Health and Primary Care, The Chinese University of Hong Kong (JT, CM, JY, MD, ZY, YH), Hong Kong SAR, China.
• Medicine (Baltimore). 2016 Apr 1; 95 (14): e3312e3312.

AbstractThe aim of the study was to examine how well aggregate data meta-analyses (ADMAs) and individual patient data meta-analyses (IPDMAs) agree in their overall results and how frequently interactions are detected in IPDMAs and ADMAs.ADMA articles immediately published before the IPDMA and matching the research topic were identified. Agreement in the overall result was achieved if the estimate was in the same direction. The number of subgroup analyses, in particular that of significant interactions, was compared between the 2 types of meta-analyses.A total of 829 IPDMA articles were identified; 129 (15.6%) were found to have a matched ADMA article and 204 paired meta-analyses were identified. Agreement in the overall effect was observed in 187 (91.7%) of the 204 paired meta-analyses. Fifty-three (26.0%) ADMAs and 121 (59.3%) IPDMAs conducted subgroup analyses and presented 150 and 634 subgroup analyses, respectively. The IPDMAs conducted 7 times more subgroup analyses on interaction (544 in IPDMAs vs 68 in ADMAs) and identified 14 times more potential interactions (44 in IPDMAs vs 3 in ADMAs).ADMAs will almost always agree with their corresponding IPDMAs in the overall result if greater efforts are made to improve the methodology in conducting ADMAs. The IPDMA is required mostly if interactions are suspected.

1 keywords...

hide…