Methods/Statistics
MetaAnalysisOnline.com: a simplified solution for conducting meta-analysis Janos Tibor Fekete* Janos Tibor Fekete Balazs Gyorffy
Background. Meta-analysis is a robust quantitative research method widely employed in epidemiology and clinical medicine to systematically evaluate and synthesize existing studies, enabling comprehensive conclusions on specific topics. This study introduces a new bioinformatics tool designed to enhance visualization and analysis options for conducting meta-analyses.
Methods. A meta-analysis can utilize binary or continuous data as its basis for analysis. The primary statistical models, such as the random effect model and the fixed effect model, are delineated along with their corresponding statistical methodologies. Moreover, we incorporated graphical representations including forest plots, funnel plots, and Z-scope plots. While the forest plot effectively illustrates heterogeneity and pooled results, a funnel plot can uncover potential publication bias, and a Z-score plot demonstrates the robustness of the sample size used. In addition to traditional pairwise meta-analysis network meta-analysis (NMA) is also adapted.
Results. The outlined models and visualization options have been integrated into a new online web portal accessible without requiring registration. The web tool works on an Ubuntu server running Apache and enables researchers to conduct meta-analyses using results typically used in epidemiology and clinical trials including binary (total and event numbers), continuous (mean and standard deviation data), and time-to-event data (hazard rate and CI data). Results from commonly used spreadsheet applications such as Excel can be directly inputted into the system. The online tool makes use of the meta, metafor and netmeta packages in the R programming environment (R version 4.2.2) for statistical calculations and for drawing the plots. The Shiny interface is powered by shinyjs, shinydashboard, and rhandsontable R libraries. The portal generates a forest plot to summarize meta-analysis results, a funnel plot to visually detect potential bias, and a Z-score plot indicating the sufficiency of the cumulative sample number. The new bioinformatic tool is accessible at www.metaanalysisonline.com and does not necessitate any programming knowledge or the use of command lines.
Conclusion. MetaAnalysisOnline.com is a user-friendly and reproducible platform tailored for epidemiological research and clinical trials. It enables the swift integration and visualization of findings from multiple studies, bringing advanced meta-analytic techniques within reach for researchers without requiring programming skills.