Ten Basic Guidelines for Conducting and Publishing a Meta-analysis

Meta-analysis is important in evidence-based medicine because it combines information from numerous sources and brings new insights. Meta-analysis research has grown in popularity as the amount of published scientific literature has increased. They are widely mentioned in biomedical sciences. Meta-analysis methodologies and criteria have developed throughout time. While there are several guidelines for certain types of meta-analysis, researchers interested in generating meta-analysis want a condensed piece containing general and up-to-date advice. This blog provides 10 straightforward guidelines for doing and publishing meta-analyses.

• To know more about Meta-Analysis Services, check our study guide. How to do a meta-analysis for a manuscript?

How do you carry out a meta-analysis?

Meta-analysis is a statistical technique used to synthesize and analyze data from multiple independent studies, aiming to derive more robust and precise conclusions than individual studies alone. To carry out a meta-analysis, meta-reviewer researchers follow specific steps. First, they define a clear research question and set inclusion/exclusion criteria for selecting relevant studies. Next, a comprehensive literature search is conducted to identify eligible studies. Data is extracted from each study, and its quality and bias are assessed. Statistical methods, such as effect size calculation and pooling, are then used to combine the data. Finally, results are interpreted, addressing heterogeneity and drawing meaningful conclusions to contribute to the existing knowledge base.

Conducting and publishing meta-analysis involves systematically synthesizing data from multiple studies on a specific research question. Here is an overview of the process:

Rule 1: Be specific about the meta-analysis’s topic and approach.

A systematic review is crucial for meta-analysis, using the PICO model to formulate research questions. Verify that no published meta-analysis on the specific topic to avoid duplication. Meta-analysis can be conducted for multiple types of studies, including epidemiological variables, genetic association studies, gene expression studies, genome-wide association studies, and animal experiments—Preregister systematic review protocols at PROSPERO to avoid duplication of efforts.

Rule 2: Adhere to the available instructions for the various forms of meta-analysis.

General guidelines for meta-analysis include the Quality of Reports of Meta-analyses of Randomized Controlled Trials (QUORUM) and the Meta-analysis of Observational Studies in Epidemiology (MOOSE) statements. The Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) are widely used. Specific guidelines for clinical, genetic association, genome-wide expression, GWASs, and animal studies have been developed.

Rule 3: Identify the essential variables and establish the inclusion criterion

Establish inclusion and exclusion criteria for studies, avoiding strict language or sample size criteria. Clearly define variables extracted from each article. Broad inclusion criteria increase heterogeneity, while narrow criteria can make it difficult to find studies. A compromise should be found, and prospective meta-analysis, typically conducted by international consortia, offers the advantage of including individual-level data.

Rule 4: Conduct a systematic search across several databases and extract important data.

To systematically search published studies, use various bibliographic databases like PubMed, Embase, The Cochrane Central Register of Controlled Trials, Scopus, Web of Science, and Google Scholar. Specific databases like BIOSIS, CINAHL, PsycINFO, Sociological Abstracts, and EconLit can help identify additional articles and data. The Web of Science database can identify publications with key articles. Adequate extraction and recording of key data from primary articles is crucial for conducting a meta-analysis. Quality assessment of included studies, such as the Jadad scale for randomized clinical trials, the Newcastle-Ottawa scale for nonrandomized studies, and QUADAS-2 for diagnostic accuracy studies, is essential for determining inclusion criteria, sensitivity analysis, and differential weighting. However, quality assessment has been criticized for reducing studies to a single “quality” score.

Rule 5: Request missing data from primary article authors.

Key data in primary articles is often unavailable, requiring authors to contact them for missing information. Standards like MIAME and STREGA promote primary data availability, but strict criteria and specialized permissions may be needed in certain areas, like genetics, where aggregated statistics can identify individuals.

Rule 6: Choose the best statistical models for your problem.

Comprehensive meta-analysis are useful for analyzing high-quality primary data for quantitative summaries. However, analysts must be prepared for various transformations, such as extracting and transforming continuous variables, 2 × 2 tables, or survival data. These analyses often use fixed-effects or random-effects statistical models, combining ranks or p-values, or multivariate methods. Additional statistical examinations include sensitivity analyses, metaregressions, subgroup analyses, and calculation of heterogeneity metrics. It is crucial to assess and explain possible sources of heterogeneity, identify sources of between-studies variation, and quantify their impact on effect size. Publication bias, such as the “Proteus phenomenon” or “winner’s curse,” is important in certain scientific fields, and cumulative meta-analysis is suggested to identify them.