Article type
Abstract
Background: In 2002, researchers chose two methods of statistical analysis for meta-epidemiological study, namely, fixed-effect of logistic regression and meta-analytic approach. To date, several additional statistical analysis methods are used in the meta-epidemiological studies.
Objectives: We aim to investigate the methods of statistical analysis used in the meta-epidemiological studies.
Methods: A literature search was performed in PubMed, EMBASE, Web of Science and Cochrane Library. The studies that had terminologies related to meta-epidemiology reported in the title or where the methods of statistical analysis were those of meta-epidemiology were eligible for inclusion. Two independent reviewers identified the eligible studies and extracted the data by using standardised forms. Disagreements were resolved by discussion or consulting the third reviewer.
Results: 3528 records were obtained from the search and 75 articles were included. 73 (97.3%) synthesised the data. Of which, 45 (60.0%) reported the methods of data synthesis. There were a total of nine methods of statistical analysis including meta-regression (22/50.0%), logistic regression (14/31.8%), meta-meta-analysis (3/6.8%), meta-epidemiologic regression (1/2.3%), Egger’s linear regression method (1/2.3%), Multivariate regression analysis (1/2.3%), the formula: Exp{ln(OR[FDA])—ln(OR[Pub])} in Microsoft Excel (1/2.3%), weighted linear regression (1/2.3%).
Conclusions: Only about half of the meta-epidemiological studies reported the statistical analysis methods of data synthesis, but almost all the methods are related to regression and meta-meta-analysis.
Objectives: We aim to investigate the methods of statistical analysis used in the meta-epidemiological studies.
Methods: A literature search was performed in PubMed, EMBASE, Web of Science and Cochrane Library. The studies that had terminologies related to meta-epidemiology reported in the title or where the methods of statistical analysis were those of meta-epidemiology were eligible for inclusion. Two independent reviewers identified the eligible studies and extracted the data by using standardised forms. Disagreements were resolved by discussion or consulting the third reviewer.
Results: 3528 records were obtained from the search and 75 articles were included. 73 (97.3%) synthesised the data. Of which, 45 (60.0%) reported the methods of data synthesis. There were a total of nine methods of statistical analysis including meta-regression (22/50.0%), logistic regression (14/31.8%), meta-meta-analysis (3/6.8%), meta-epidemiologic regression (1/2.3%), Egger’s linear regression method (1/2.3%), Multivariate regression analysis (1/2.3%), the formula: Exp{ln(OR[FDA])—ln(OR[Pub])} in Microsoft Excel (1/2.3%), weighted linear regression (1/2.3%).
Conclusions: Only about half of the meta-epidemiological studies reported the statistical analysis methods of data synthesis, but almost all the methods are related to regression and meta-meta-analysis.